Articles

Atezolizumab plus nab-paclitaxel as first-line treatment for unresectable, locally advanced or metastatic triple-negative breast cancer (IMpassion130): updated efficacy results from a randomised, double-blind, placebo-controlled, phase 3 trial

Peter Schmid*, Hope S Rugo*, Sylvia Adams, Andreas Schneeweiss, Carlos H Barrios, Hiroji Iwata, Véronique Diéras, Volkmar Henschel, Luciana Molinero, Stephen Y Chui, Vidya Maiya, Amreen Husain, Eric P Winer, Sherene Loi, Leisha A Emens, for the IMpassion130 Investigators†

Summary Lancet Oncol 2020; 21: 44–59 Background Immunotherapy in combination with chemotherapy has shown promising efficacy across many different Published Online tumour types. We report the prespecified second interim overall survival analysis of the phase 3 IMpassion130 study November 27, 2019 assessing the efficacy and safety of atezolizumab plus nab-paclitaxel in patients with unresectable, locally advanced or https://doi.org/10.1016/ metastatic triple-negative breast cancer. S1470-2045(19)30689-8 See Comment page 3 Methods In this randomised, placebo-controlled, double-blind, phase 3 trial, done in 246 academic centres and *Contributed equally community oncology practices in 41 countries, patients aged 18 years or older, with previously untreated, histologically †Investigators are listed in the documented, locally advanced or metastatic triple-negative breast cancer, and Eastern Cooperative Oncology Group appendix performance status of 0 or 1 were eligible. Patients were randomly assigned (1:1) using a permuted block method Barts Cancer Institute, Queen Mary University of London, (block size of four) and an interactive voice–web response system. Randomisation was stratified by previous taxane London, UK (Prof P Schmid MD); use, liver metastases, and PD-L1 expression on tumour-infiltrating immune cells. Patients received atezolizumab University of California 840 mg or matching placebo intravenously on day 1 and day 15 of every 28-day cycle and nab-paclitaxel 100 mg/m² of San Francisco Comprehensive body surface area intravenously on days 1, 8, and 15 until progression or unacceptable toxicity. Investigators, patients, Cancer Center, University of California, San Francisco, CA, and the funder were masked to treatment assignment. Coprimary endpoints were investigator-assessed progression- USA (Prof H S Rugo MD); free survival per Response Evaluation Criteria in Solid Tumors version 1.1 and overall survival, assessed in the Perlmutter Cancer Center, intention-to-treat population and in patients with PD-L1 immune cell-positive tumours (tumours with ≥1% PD-L1 New York University Langone expression). The final progression-free survival results were previously reported at the first interim overall survival Medical Center, New York, NY, USA (Prof S Adams MD); analysis. The prespecified statistical testing hierarchy meant that overall survival in the subgroup of PD-L1 immune National Center for Tumor cell-positive patients could only be formally tested if overall survival was significantly different between the treatment Diseases, Heidelberg University groups in the intention-to-treat population. This study is registered with ClinicalTrials.gov, NCT02425891. Hospital and German Cancer Research Center, Heidelberg, Germany Findings Between June 23, 2015, and May 24, 2017, 902 patients were enrolled, of whom 451 were randomly assigned (Prof A Schneeweiss MD); to receive atezolizumab plus nab-paclitaxel and 451 were assigned to receive placebo plus nab-paclitaxel (the intention- Centro de Pesquisa Clínica, to-treat population). Six patients from each group did not receive treatment. At the second interim analysis (data Hospital São Lucas, Pontifícia cutoff Jan 2, 2019), median follow-up was 18·5 months (IQR 9·6–22·8) in the atezolizumab group and 17·5 months Universidade Católica do Rio Grande do Sul, Porto (8·4–22·4) in the placebo group. Median overall survival in the intention-to-treat patients was 21·0 months (95% CI Alegre, Brazil (C H Barrios MD); 19·0–22·6) with atezolizumab and 18·7 months (16·9–20·3) with placebo (stratified hazard ratio [HR] 0·86, 95% CI Latin American Cooperative 0·72–1·02, p=0·078). In the exploratory overall survival analysis in patients with PD-L1 immune cell-positive tumours, Oncology Group, Porto Alegre, Brazil (C H Barrios); Grupo median overall survival was 25·0 months (95% CI 19·6–30·7) with atezolizumab versus 18·0 months (13·6–20·1) Oncoclínicas, Porto Alegre, with placebo (stratified HR 0·71, 0·54–0·94]). As of Sept 3, 2018 (the date up to which updated safety data were Brazil (C H Barrios); Aichi Cancer available), the most common grade 3–4 adverse events were neutropenia (38 [8%] of 453 patients in the atezolizumab Center Hospital, Nagoya, Japan group vs 36 [8%] of 437 patients in the placebo group), peripheral neuropathy (25 [6%] vs 12 [3%]), decreased neutrophil (Prof H Iwata MD); Department of Medical Oncology, Institut count (22 [5%] vs 16 [4%]), and fatigue (17 [4%] vs 15 [3%]). Treatment-related deaths occurred in two (<1%) patients in Curie, Paris, France the atezolizumab group (autoimmune hepatitis related to atezolizumab [n=1] and septic shock related to nab-paclitaxel (V Diéras MD); Department of [n=1]) and one (<1%) patient in the placebo group (hepatic failure). No new treatment-related deaths have been Medical Oncology, Centre reported since the primary clinical data cutoff date (April 17, 2018). Eugène Marquis, Rennes, France (V Diéras); F Hoffmann-La Roche, Basel, Interpretation Consistent with the first interim analysis, this second interim overall survival analysis of IMpassion130 Switzerland (V Henschel PhD, indicates no significant difference in overall survival between the treatment groups in the intention-to-treat population A Husain MD); Genentech, but suggests a clinically meaningful overall survival benefit with atezolizumab plus nab-paclitaxel in patients with South San Francisco, CA, USA (L Molinero PhD, S Y Chui MD, PD-L1 immune cell-positive disease. However, this positive result could not be formally tested due to the prespecified V Maiya MD); Dana-Farber statistical testing hierarchy. For patients with PD-L1 immune cell-positive metastatic triple-negative breast cancer, Cancer Institute, Boston, MA, atezolizumab plus nab-paclitaxel is an important therapeutic option in a disease with high unmet need. USA (Prof E P Winer MD); Peter MacCallum Cancer Funding F Hoffmann-La Roche and Genentech.

44 www.thelancet.com/oncology Vol 21 January 2020 Articles

Copyright © 2019 Elsevier Ltd. All rights reserved. Centre, University of Melbourne, Melbourne, VIC, Introduction survival estimates of approximately 18 months or less Australia (Prof S Loi MD); and University of Pittsburgh Triple-negative breast cancer constitutes 15–20% of cases with available treatments. Because of the promising Medical Center Hillman Cancer of breast cancer and is defined by the absence of efficacy of checkpoint inhibition in other tumour types,9–13 Center, Pittsburgh, PA, USA oestrogen receptors, progesterone receptors, and over­ inhibitors of PD-L1 and PD-1 as monotherapy or in (Prof L A Emens MD) expression or gene amplification of HER2 on the surface combination with chemotherapy have been investigated Correspondence to: of cancer cells.1 Until the first report of the IMpassion130 in triple-negative breast cancer.14–16 The higher prevalence Prof Peter Schmid, Barts Cancer Institute, 2 trial, chemo­therapy was the standard of care for first-line of PD-L1 expression in triple-negative breast cancer Queen Mary University of systemic treatment for patients with triple-negative compared with hormone receptor-positive breast cancer London, London EC1M 6BQ, UK breast cancer that has advanced or metastasised.3–5 In subtypes further supports this therapeutic approach.17–21 [email protected] several countries, such as Japan, and in Europe, other The phase 3 study, IMpassion130,2 evaluated atezoli­ See Online for appendix approved treatment options are bevacizumab in zumab, a monoclonal antibody targeting PD-L1, plus nab- combination with chemotherapy and poly (ADP-ribose) paclitaxel compared with placebo plus nab-paclitaxel as a polymerase inhibitors (eg, olaparib and talazoparib) for first-line treatment for patients with unresectable locally patients with BRCA-mutant, HER2-negative tumours.6–8 advanced or metastatic triple-negative breast cancer. At Prognoses remain poor, with reported median overall the time of primary analysis of IMpassion130, the final

Research in context Evidence before this study improvement in median overall survival in patients with PD-L1 We searched PubMed for clinical trials published between immune cell-positive disease. This preplanned second interim June 30, 2009, and July 30, 2019, in any language using the overall survival analysis, which was done after 534 (59%) of search terms “PD-L1” or “PD-1” or “immunotherapy” or 902 deaths had occurred in the intention-to-treat population “immune checkpoint” in combination with “TNBC” or “triple- (80% information fraction) also showed no significant negative breast cancer”. We identified a feasibility study of a difference in overall survival between the groups in the personalised peptide vaccine and a phase 1–2 study planning to intention-to-treat population. Although the overall survival enrol patients with triple-negative breast cancer to evaluate an results provide evidence for the clinical benefit in patients with RNA vaccine. Several studies reported increased expression of PD-L1 immune cell-positive disease, consistent with the primary PD-L1 and PD-1 in triple-negative breast cancer, suggesting analysis, this result could not be formally tested due to the that inhibition of the PD-L1–PD-1 immune checkpoint could prespecified statistical testing hierarchy and thus these findings result in clinical benefit in this disease. Promising responses and can only be regarded as exploratory. The reported safety update duration of response in phase 1 studies evaluating single-drug supports an unchanged safety profile with atezolizumab plus atezolizumab, pembrolizumab, or avelumab had been reported. nab-paclitaxel relative to the first interim analysis. A tolerable safety profile and increased proportions of patients Implications of all the available evidence achieving a response compared with single-drug approaches The randomised controlled phase 3 IMpassion130 study did not had been reported for the combination of atezolizumab plus suggest a significant overall survival benefit from the addition nab-paclitaxel in a phase 1b study. of atezolizumab to nab-paclitaxel in the intention-to-treat Added value of this study population of patients with unresectable, locally advanced or To our knowledge, IMpassion130 is the first reported phase 3 metastatic triple-negative breast cancer. However, providing study that evaluated an antibody targeting the PD-L1–PD-1 evidence to support promising clinical activity observed in immune checkpoint in patients with unresectable locally phase 1 studies evaluating drugs targeting PD-L1–PD-1 advanced or metastatic triple-negative breast cancer. The study signalling, the study results did indicate clinically meaningful met its coprimary progression-free survival endpoint in the improvement in median progression-free and overall survival primary analysis, showing a significant progression-free survival of atezolizumab plus nab-paclitaxel in patients with PD-L1 benefit with atezolizumab plus nab-paclitaxel compared with immune cell-positive unresectable locally advanced or placebo plus nab-paclitaxel as a first-line treatment for patients metastatic triple-negative breast cancer. The findings of this with advanced or metastatic triple-negative breast cancer. second interim overall survival analysis, reported here, provide Because no treatment effect was seen in the PD-L1 immune cell- evidence that the clinical benefit is consistent and maintained negative subgroup, the clinical benefit in the intention-to-treat with extended follow-up. Approvals by US and European Union population was driven by the positive results in the PD-L1 regulatory agencies, and the inclusion of this regimen in US immune cell-positive subgroup of patients. The first interim National Comprehensive Cancer Network treatment guidelines, overall survival analysis (concurrent with the primary highlight that atezolizumab plus nab-paclitaxel could be a new progression-free survival analysis) showed no significant standard of care for previously untreated patients with PD-L1 difference in overall survival between the treatment groups in immune cell-positive, unresectable, locally advanced or the intention-to-treat population but a clinically meaningful metastatic triple-negative breast cancer.

www.thelancet.com/oncology Vol 21 January 2020 45 Articles

progression-free survival analysis and the first interim centres and community oncology practices in 41 countries overall survival analysis were done (clinical data cutoff in Europe, North America, Asia, and Latin America April 17, 2018). The study met its co­primary progression- (appendix pp 2–6). Full details of the study design and free survival endpoints, showing clinical benefit with methods have previously been published.21 Patients were atezolizumab plus nab-paclitaxel when compared with aged 18 years or older and had histologically documented, placebo plus nab-paclitaxel in both the intention-to-treat unresectable, locally advanced or metastatic triple- population and the subgroup of patients with tumours negative breast cancer and a life expectancy of at least that express PD-L1 on immune cells that cover 1% or more 12 weeks. HER2-negative, oestrogen receptor-negative, of the tumour area (PD-L1 immune cell-positive patients).2 and progesterone receptor-negative status was evaluated The treatment effect in the PD-L1 immune cell-positive by local institutions before enrolment per guidelines of patient subgroup was a clinically meaningful benefit: the the American Society of Clinical Oncology and the median progression-free survival was 7·5 months (95% College of American Pathologists.24,25 Other key inclusion CI 6·7–9·2) with atezolizumab plus nab-paclitaxel versus criteria were that patients had to provide representative 5·0 months (3·8–5·6) with placebo plus nab-paclitaxel tumour specimens that were evaluable for prospective (hazard ratio [HR] 0·62, 95% CI 0·49–0·78], p<0·0001).2 testing of PD-L1 expression by immunohistochemistry Based on these final progression-free survival results, the (SP142 PD-L1 immunohisto­chemical assay [Ventana US Food and Drug Administration granted accelerated Medical Systems, Oro Valley, AZ, USA); they were approval on March 8, 2019, to atezolizumab plus nab- eligible for taxane monotherapy; they had no previous paclitaxel as a first-line treatment for patients with chemotherapy or targeted therapy for metastatic triple- unresectable locally advanced or metastatic triple-negative negative breast cancer; they had measurable disease per breast cancer with tumours that express PD-L1 on Response Evaluation Criteria in Solid Tumors (RECIST) immune cells that cover 1% or more of the tumour version 1.1; they had an Eastern Cooperative Oncology area (PD-L1 immune-cell positive).22 The European Group (ECOG) performance status score of 1 or less; Commission has also approved atezolizumab plus nab- and they had adequate organ and haematological paclitaxel for the treatment of adult patients with function. Radiotherapy and previous curative chemo­ unresectable locally advanced or metastatic triple-negative therapy completed 12 months or more before random­ breast cancer with tumours that are PD-L1 positive (≥1% isation was allowed. Patients with treated asymptomatic PD-L1 expression) and who have not received previous CNS metastases were also allowed. Key exclusion chemotherapy for metastatic disease.23 criteria included untreated CNS disease; previous history As previously reported,2 the first interim overall survival of auto­immune disease; recent treatment (ie, within analysis and concurrent final progression-free survival 4 weeks or five half-lives of the drug [whichever analysis were done after 43% of deaths in the intention- was shorter] before randomisation) with a systemic to-treat population had occurred, with a median follow- immunostimulatory drug; use of glucocorticoids or up of 12·9 months (IQR 8·8–16·5) and 59% information immunosuppressive drugs; and previous immune fraction, which is the percentage of the total information checkpoint-targeting therapies. expected at the scheduled end of the trial. With an HR for The trial was done in accordance with Good Clinical overall survival of 0·84 (95% CI 0·69–1·02, p=0·08) in Practice guidelines and the Declaration of Helsinki. the intention-to-treat population, the first interim overall Protocol approval was obtained from independent review survival analysis did not cross the prespecified boards or ethics committees for each site. An independent significance boundary (α=0·0065). A Kaplan-Meier data monitoring committee reviewed unmasked safety analysis of overall survival in the PD-L1 immune cell- and study conduct data every 6 months. All patients positive subgroup that was not formally tested because of provided written informed consent. For the protocol see the prespecified statistical testing hierarchy (since Our study protocol has been published online and is https://www.nejm.org/doi/ significance in the intention-to-treat population was also available in the appendix. suppl/10.1056/ required to enable subsequent testing in the PD-L1 NEJMoa1809615/suppl_file/ nejmoa1809615_protocol.pdf immune cell-positive subgroup) showed a median overall Randomisation and masking survival of 25·0 months (95% CI 22·6–not estimable) The principal investigators identified and enrolled with atezolizumab plus nab-paclitaxel versus 15·5 months patients, as described previously.2 Using a permuted (13·1–19·4) with placebo plus nab-paclitaxel (HR 0·62, block method (with a block size of four) and an interactive 95% CI 0·45–0·86).2 voice–web response system, we randomly assigned In this paper, we report the prespecified second interim patients in a 1:1 ratio to receive atezolizumab plus nab- overall survival analysis from the IMpassion130 study. paclitaxel or placebo plus nab-paclitaxel. We stratified randomisation by the presence of liver metastases (yes vs Methods no), previous neoadjuvant or adjuvant taxane therapy Study design and participants (yes vs no), and PD-L1 expression on tumour-infiltrating The IMpassion130 trial is a randomised, double-blind, immune cell as a percentage of tumour area (<1% [PD-L1 placebo-controlled trial that was done in 246 academic immune cell negative] vs ≥1% [PD-L1 immune cell

46 www.thelancet.com/oncology Vol 21 January 2020 Articles

positive]). The investigators, study site personnel, and any cause, whichever occurred first), and time to patients were masked to treatment assignment. Masking deterioration in global health status (in items 29 and 30 of was achieved by ensuring that the placebo was not the European Organisation for Research and Treatment distinguishable from atezolizumab. of Cancer Quality of life Questionnaire Core 30). Quality- of-life outcomes will be reported in a separate publication. Procedures Exploratory endpoints included predictive, prognostic, Patients received either atezolizumab 840 mg or and pharmacodynamic bio­markers and their association matching placebo on days 1 and 15 of every 28-day with disease status and response to study treatment. treatment cycle and nab-paclitaxel 100 mg/m² of body Exploratory analyses of the distribution of PD-L1-positive surface area on days 1, 8, and 15 of every cycle. All drugs immune cells and its interaction as a continuous variable were administered intravenously, and patients received and the activity of atezolizumab plus nab-paclitaxel in study treatments until disease progression, as assessed patients with PD-L1 expression in tumour cells are part of by the investigators per RECIST 1.1, or occurrence of a separate biomarker analysis, which will be reported in a unacceptable toxicity (according to the medical assess­ separate publication. Remaining exploratory endpoints ment of the treating investigator who was directly caring were related to additional patient-reported outcomes, for the patient). If no unacceptable toxicity occurred, nab- including health economic data, which will be reported paclitaxel was administered for at least six cycles of elsewhere. Definitions for prespecified adverse events of treatment. Dose reduction of atezolizumab or placebo special interest, which might be suggestive of an immune- was not permitted, but treatment could be temporarily mediated disorder, are sum­marised in the protocol. suspended for as long as deemed necessary by the treating investgator, if required to manage an adverse Statistical analysis event (of any severity); decisions regarding suspension of The trial was originally designed to enrol 350 patients, treatment were made on an individual basis by the with a primary progression-free survival endpoint. On treating investigator and medical monitor by assessing Nov 20, 2015, the study protocol was amended to increase the benefits and risks to that patient. The treating the sample size to approximately 900 patients (intention- physician could use discretion in modifying or to-treat population), supporting the promotion of overall accelerating the dose modification guidelines depending survival from a secondary endpoint to a coprimary on the severity of toxicity and an assessment of the risk endpoint (protocol version 4). Progression-free survival versus benefit for the patient, with the goal of maximizing and overall survival were analysed by intention-to-treat patient compliance and access to supportive care. The and in the PD-L1 immune cell-positive population. Two dose of nab-paclitaxel could be reduced to 75 mg/m², interim analyses and a final overall survival analysis were 50 mg/m², and then discontinued, if required because planned. The first interim analysis for overall survival of haematological toxicity (of any severity). The treating was done with an information fraction of 59% (ie, when investigator assessed patients for adverse events 59% of the total information expected at the scheduled (according to the Common Terminology Criteria for end of the trial was available, after approximately Adverse Events [version 4.0] of the National Cancer 352 deaths had occurred); the definitive progression-free Institute) and laboratory values (serum chemistry) on survival analysis was done concur­rently.21 The second days 1, 8, and 15 of every 28-day cycle. Atezolizumab or interim analysis for overall survival was planned with an nab-paclitaxel could be discontinued independently in information fraction of 80% available (available alpha the absence of disease progression based on RECIST 1.1. was 0·0212 obtained from an alpha spending function). The investigators imaged tumours by CT or MRI at Overall survival and progression-free survival were baseline, every 8 weeks for 12 months, then every compared between treatment groups using a stratified 12 weeks until disease progression. We followed up log-rank test, and HRs for death and disease progression patients for survival every 3 months after treatment were estimated using a stratified Cox proportional discontinuation until death. hazards model. The analysis was stratified by the same stratification factors that were used for the randomisation. Outcomes The type I error for the study (α=0·05) was controlled for The coprimary efficacy endpoints were progression-free treatment comparisons of the coprimary endpoints of survival, assessed by investigators per RECIST 1.1, and progression-free survival and overall survival and for the overall survival. These endpoints were evaluated in the secondary efficacy endpoint of objective responses in the intention-to-treat population and the PD-L1 immune cell- intention-to-treat and PD-L1 immune cell-positive positive subgroup of patients. Prespecified secondary populations, using a stepwise testing procedure. Post- efficacy endpoints were the proportion of patients hoc exploratory results in the PD-L1 immune cell- achieving an objective response (ie, a complete or partial negative population are displayed for comparison. The response according to RECIST 1.1) and duration of interim analysis significance boundary was based on the response (measured from the first objective response to Lan-DeMets implementation of the O’Brien-Fleming use disease progression per RECIST version 1.1 or death from function. The null hypothesis of no difference in overall www.thelancet.com/oncology Vol 21 January 2020 47 Articles

1235 patients assessed for eligibility

333 ineligible* 54 had known CNS disease 51 declined to participate 27 did not have documented TNBC 21 had inadequate haematological and end-organ function 19 had no measurable disease per RECIST 1.1 19 had previous chemotherapy or targeted systemic therapy 16 had no representative FFPE tumour specimens 13 had Eastern Cooperative Oncology Group performance status ≥2

902 randomly assigned

451 assigned to atezolizumab plus nab-paclitaxel 451 assigned to placebo plus nab-paclitaxel

6 did not receive treatment 6 did not receive treatment 8 received one dose of atezolizumab and were evaluated in atezolizumab plus nab-paclitaxel safety-evaluable population

453 safety evaluable 437 safety evaluable 445 received randomly assigned treatment 437 received randomly assigned treatment 8 in placebo plus nab-paclitaxel group received one atezolizumab dose

406 discontinued atezolizumab 426 discontinued nab-paclitaxel 445 discontinued placebo 432 discontinued nab-paclitaxel 330 disease progression 283 disease progression 351 disease progression 318 disease progression 30 adverse event 76 adverse event 41 other reasons 38 adverse event 17 withdrawal by patient 20 withdrawal by patient 25 symptomatic 31 physician decision 14 symptomatic 24 physician decision deterioration 23 symptomatic deterioration 13 symptomatic 12 withdrawal by patient deterioration 6 death deterioration 7 physician decision 14 withdrawal by patient 5 physician decision 6 death 5 adverse event 5 other reasons 3 other reasons 4 other reasons 3 death 3 death 1 non-compliance 1 protocol violation

255 death 279 death 24 lost to follow-up 24 lost to follow-up

172 patients in intention-to-treat population remain on study 148 patients in intention-to-treat population remain on study

133 in survival follow-up 39 remain on treatment 13 remain on treatment 135 in survival follow-up 39 on atezolizumab 0 on placebo 19 on nab-paclitaxel 13 on nab-paclitaxel

451 included in intention-to-treat 451 included in intention-to-treat analysis analysis

Figure 1: Trial profile FFPE=formalin fixed and paraffin embedded. RECIST 1.1=Response Evaluation Criteria in Solid Tumors version 1.1. TNBC=triple-negative breast cancer. *Most common reasons for ineligibility.

48 www.thelancet.com/oncology Vol 21 January 2020 Articles

Intention-to-treat population (n=902) PD-L1 immune cell-positive population (n=369) Atezolizumab plus Placebo plus Atezolizumab plus Placebo plus nab-paclitaxel group nab-paclitaxel group nab-paclitaxel group nab-paclitaxel group (n=451) (n=451) (n=185) (n=184) Age, years 55 (46–64) 56 (47–65) 53 (44–63) 53 (44–63) Age group, years 18–40 63 (14%) 51 (11%) 31 (17%) 24 (13%) 41–64 284 (63%) 285 (63%) 111 (60%) 117 (64%) ≥65 104 (23%) 115 (25%) 43 (23%) 43 (23%) Sex Female 448 (99%) 450 (>99%) 184 (99%) 184 (100%) Male 3 (1%) 1 (<1%) 1 (1%) 0 Race White 308 (68%) 301 (67%) 125 (68%) 129 (70%) Asian 85 (19%) 76 (17%) 38 (21%) 28 (15%) Black or African-American 26 (6%) 33 (7%) 9 (5%) 14 (8%) American Indian or Alaska Native 17 (4%) 23 (5%) 8 (4%) 9 (5%) Native Hawaiian or other Pacific islander 1 (<1%) 0 0 0 Multiple 2 (<1%) 3 (1%) 0 0 Unknown 12 (3%) 15 (3%) 5 (3%) 4 (2%) Eastern Cooperative Oncology Group performance status 0 256/450 (57%) 270/450 (60%) 107/185 (58%) 112/184 (61%) 1 193/450 (43%) 179/450 (40%) 77/185 (42%) 72/184 (39%) 2* 1/450 (<1%) 1/450 (<1%) 1/185 (1%) 0 Metastatic disease 404/450 (90%) 408/450 (91%) 162/185 (88%) 159/183 (87%) Number of metastatic sites 0–3 332/450 (74%) 341/449 (76%) 149/185 (81%) 140/183 (77%) ≥4 118/450 (26%) 108/449 (24%) 36/185 (19%) 43/183 (23%) Site of metastatic disease Liver† 126 (28%) 118 (26%) 44 (24%) 39 (21%) Bone 145 (32%) 141 (31%) 54 (29%) 49 (27%) Brain 30 (7%) 31 (7%) 15 (8%) 11 (6%) Lung 226 (51%) 242 (54%) 86 (46%) 98 (53%) Lymph node only 33/450 (7%) 23/449 (5%) 18/185 (10%) 13/183 (7%) Previous neoadjuvant or adjuvant treatment 284 (63%) 286 (63%) 125 (68%) 117 (64%) Previous taxane use† 231 (51%) 230 (51%) 96 (52%) 94 (51%) Previous anthracycline use 243 (54%) 242 (54%) 109 (59%) 101 (55%) Previous radiotherapy 268 (59%) 280 (62%) 119 (64%) 112 (61%) Radiotherapy of the brain 25 (6%) 27 (6%) 14 (8%) 11 (6%) Time from last surgery until diagnosis with 24·5 (15·9–38·9) 24·8 (17·0–44·6) 21·5 (15·0–36·2) 22·1 (14·1–39·3) unresectable locally advanced or metastatic disease, months

Adapted from Schmid and colleagues,21 by permission of Massachusetts Medical Society. Data are median (IQR), n (%), or n/N (%). Summary statistics are based on the full population indicated in the header, unless otherwise indicated. If the baseline characteristic was not available for all patients, the total number of patients evaluable for this characteristic is shown. *Two patients (one in each group) had a performance status of 2 before the start of treatment; they had a lower performance status at randomisation, hence why they were included in the study. †As recorded in the case report form.

Table 1: Baseline characteristics survival between treatment groups was tested with an PD-L1 immune cell-positive subgroup be tested by re- allocated type I error at an α level of at least 0·04 for using the type I error used for overall survival testing overall survival and hierarchical testing for overall (≥0·04). The Kaplan-Meier method was used to analyse survival, first in the intention-to-treat population and progression-free survival and overall survival, with the then in the PD-L1 immune cell-positive subgroup Brookmeyer-Crowley method used to estimate the (appendix p 7). Only if the difference in overall survival 95% CI for each median duration. The proportional between the treatment groups in the intention-to-treat hazards assumption for overall survival was assessed in population was significant could overall survival in the an exploratory manner. The same methods described for www.thelancet.com/oncology Vol 21 January 2020 49 Articles

A 100 Atezolizumab plus nab-paclitaxel Placebo plus nab-paclitaxel 90 Stratified HR 0·86 (95% CI 0·72–1·02); log-rank p=0·078 80

70

60

50

40 Overall survival (%) 30

20

10

0 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 Number at risk (number censored) Atezolizumab plus nab-paclitaxel 426 (5) 389 (13) 342 (17) 312 (17) 270 (20) 235 (25) 162 (75) 88 (128) 56 (155) 35 (169) 19 (178) 8 (188) 3 (193) ·· Placebo plus nab-paclitaxel 420 (5) 376 (14) 329 (16) 291 (20) 252 (21) 216 (28) 145 (68) 87 (111) 51 (134) 33 (146) 17 (160) 4 (168) 1 (171) ··

B 100 Atezolizumab plus nab-paclitaxel (PD-L1 immune-cell positive) Placebo plus nab-paclitaxel (PD-L1 immune-cell positive) 90 Atezolizumab plus nab-paclitaxel (PD-L1 immune-cell negative) Placebo plus nab-paclitaxel (PD-L1 immune-cell negative) 80

70

60

50

40 Overall survival (%) 30

20

10

0 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 Time since randomisation (months) Number at risk (number censored) Atezolizumab plus nab-paclitaxel 177 (1) 160 (5) 145 (6) 135 (6) 121 (7) 106 (9) 69 (37) 43 (58) 28 (72) 21 (76) 10 (81) 6 (85) 3 (88) ·· (PD-L1 immune-cell positive) Placebo plus nab-paclitaxel 170 (2) 147 (8) 129 (10) 111 (11) 93 (12) 81 (15) 47 (36) 26 (54) 20 (59) 15 (63) 10 (68) 1 (73) ·· ·· (PD-L1 immune-cell positive) Atezolizumab plus nab-paclitaxel 249 (4) 229 (8) 197 (11) 177 (11) 149 (13) 129 (16) 93 (38) 45 (70) 28 (83) 14 (93) 9 (97) 2 (103) ·· ·· (PD-L1 immune-cell negative) Placebo plus nab-paclitaxel 250 (3) 229 (6) 200 (6) 180 (9) 159 (9) 135 (13) 98 (32) 61 (57) 31 (75) 18 (83) 7(92) 3 (95) 1 (97) ·· (PD-L1 immune-cell negative)

Figure 2: Second interim overall survival analysis Kaplan-Meier analysis of overall survival in the (A) intention-to-treat and (B) PD-L1 immune cell-positive and PD-L1 immune cell-negative populations. HR is from stratified log-rank tests. HR=hazard ratio.

the analysis of progression-free survival were used for (ie, patients who received any amount of any study drug). the analysis of duration of response except that the SAS version 9.4 was used for all statistical analyses. analysis was not stratified. Subgroup analyses of overall This study is registered with ClinicalTrials.gov, survival, prespecified in the statistical analysis plan, are NCT02425891. provided. Interaction between PD-L1 status and overall survival benefit with atezolizumab plus nab-paclitaxel Role of the funding source was tested using a stratified Cox-model including the The funders provided the study drug (atezolizumab) and covariate treatment group, PD-L1 status, and their the matched placebo, and collaborated with academic interaction. Safety was evaluated in the safety population authors on the study design, data collection, data analysis,

50 www.thelancet.com/oncology Vol 21 January 2020 Articles

data interpretation, and writing of the report. Celgene provided final approval to publish, and agreed to be provided nab-paclitaxel, but the company had no role in accountable for all aspects of the manuscript. VH, LM, study design, data collection, data analysis, or data SYC, VM, and AH had full access to the raw data and PS interpretation, although they did review the manuscript. and HSR had final responsibility for the decision to All authors contributed to drafting the manuscript, submit for publication.

A Atezolizumab plus Placebo plus Hazard ratio nab-paclitaxel nab-paclitaxel (95% CI) Events (n)/ Median OS, Events (n)/ Median OS, patients (N) months (95% CI) patients (N) months (95% CI) PD-L1 status Positive 94/185 25·0 (19·5–30·7) 110/184 18·0 (13·6–20·1) 0·71 (0·54–0·94) Negative 161/266 19·6 (16·3–21·6) 169/267 19·6 (16·9–22·2) 0·99 (0·80–1·23) Age (years) 18–40 44/63 16·8 (13·2–22·1) 37/51 13·1 (10·7–17·1) 0·81 (0·52–1·25) 41–64 158/284 20·8 (18·5–26·0) 170/285 19·9 (16·9–22·6) 0·88 (0·71–1·10) ≥65 53/104 21·9 (19·0–31·1) 72/115 19·4 (16·0–22·7) 0·78 (0·55–1·12) Race White 180/308 21·0 (17·2–22·8) 198/301 17·6 (14·9–19·6) 0·80 (0·65–0·98) Asian 39/85 29·4 (20·5–30·9) 34/76 30·3 (21·5–NE) 1·17 (0·74–1·87) Black or African-American 14/26 18·5 (10·6–NE) 21/32 15·7 (10·0–22·6) 0·75 (0·38–1·49) ECOG performance status* 0 127/256 23·2 (21·1–30·7) 145/270 22·2 (19·7–24·6) 0·85 (0·67–1·08) 1 127/193 16·3 (14·0–19·5) 132/179 13·2 (10·7–16·2) 0·85 (0·66–1·08) Baseline disease status* Locally advanced 21/46 26·0 (21·0–NE) 13/42 34·9 (24·6–34·9) 1·53 (0·76–3·06) Metastatic 234/404 20·4 (17·5–22·1) 266/408 17·9 (15·9–19·6) 0·82 (0·69–0·98) Number of metastatic sites* 0–3 172/332 23·2 (21·4–28·8) 194/341 20·8 (19·4–24·0) 0·83 (0·68–1·02) >3 83/118 12·7 (11·8–14·7) 83/108 12·1 (9·3–16·0) 0·90 (0·66–1·22) Brain metastases Yes 22/30 14·3 (8·3–21·1) 19/31 16·2 (12·2–NE) 1·34 (0·72–2·48) No 233/421 21·3 (19·3–23·2) 260/420 19·4 (17·3–20·8) 0·83 (0·70–1·00) Bone metastases Yes 92/145 16·8 (13·7–19·7) 103/141 14·9 (11·6–18·2) 0·80 (0·61–1·07) No 163/306 22·1 (20·8–27·3) 176/310 20·1 (18·2–22·7) 0·88 (0·71–1·09) Liver metastases Yes 88/126 14·0 (11·2–17·1) 95/118 12·1 (9·8–13·6) 0·77 (0·58–1·03) No 167/325 22·6 (20·8–28·0) 184/333 21·1 (19·6–24·6) 0·88 (0·72–1·09) Lung metastases Yes 138/227 17·8 (14·7–21·1) 153/242 17·6 (15·2–20·0) 0·94 (0·74–1·18) No 117/224 22·8 (20·4–29·4) 126/209 20·0 (17·9–24·0) 0·80 (0·62–1·02) Lymph node-only disease* Yes 12/33 NE (22·1–NE) 11/23 34·7 (21·1–35·1) 0·74 (0·32–1·67) No 243/417 20·0 (17·6–21·9) 266/426 18·2 (16·3–20·0) 0·88 (0·74–1·05) Previous neoadjuvant or adjuvant chemotherapy Yes 160/284 21·3 (18·2–25·0) 166/286 19·9 (18·1–23·8) 0·92 (0·74–1·15) No 95/167 20·6 (17·2–23·4) 113/165 16·3 (14·1–18·7) 0·75 (0·57–0·99) Previous taxane treatment Yes 138/231 20·5 (17·3–23·2) 136/230 19·7 (17·3–23·8) 0·95 (0·75–1·20) No 117/220 21·1 (18·1–23·7) 143/221 17·9 (15·2–20·0) 0·76 (0·59–0·97) Previous anthracycline treatment Yes 143/243 19·7 (17·1–22·7) 144/242 19·6 (17·1–23·6) 1·00 (0·79–1·26) No 112/208 21·9 (19·4–27·3) 135/209 18·0 (15·3–20·3) 0·71 (0·55–0·92) Overall 255/451 21·0 (19·0–22·6) 279/451 18·7 (16·9–20·3) 0·86 (0·73–1·02)

0·25 1·0 2·5

Favours atezolizumab Favours placebo plus plus nab-paclitaxel nab-paclitaxel

(Figure 3 continues on next page) www.thelancet.com/oncology Vol 21 January 2020 51 Articles

B Atezolizumab plus Placebo plus Hazard ratio nab-paclitaxel nab-paclitaxel (95% CI) Events (n)/ Median OS, Events (n)/ Median OS, patients (N) months (95% CI) patients (N) months (95% CI) Age (years) 18–40 19/31 15·3 (8·8–30·7) 18/24 13·2 (9·0–17·7) 0·74 (0·39–1·42) 41–64 57/111 23·4 (19·4–30·3) 66/117 19·4 (14·4–28·6) 0·77 (0·54–1·10) ≥65 18/43 30·7 (19·0–NE) 26/43 18·7 (11·1–23·1) 0·54 (0·29–0·98) Race White 68/125 23·4 (16·9–30·7) 82/129 16·5 (12·8–19·8) 0·71 (0·52–0·98) Asian 17/38 30·3 (19·1–NE) 12/28 NE (17·7–NE) 1·04 (0·49–2·17) Black or African-American 1/9 NE (NE–NE) 10/13 10·0 (8·4–20·1) 0·07 (<0·01–0·55) ECOG performance status* 0 52/107 28·8 (21·1–33·0) 56/112 23·1 (18·0–34·9) 0·83 (0·57–1·21) 1 41/77 19·5 (16·2–32·2) 54/72 11·5 (8·4–16·5) 0·55 (0·37–0·83) Baseline disease status* Locally advanced 8/23 33·0 (25·0–NE) 9/24 34·9 (21·1–34·9) 0·73 (0·28–1·91) Metastatic 86/162 21·6 (17·6–30·3) 101/159 16·5 (12·8–19·4) 0·69 (0·52–0·92) Number of metastatic sites* 0–3 69/149 30·3 (22·6–32·2) 77/140 20·0 (16·5–28·6) 0·71 (0·51–0·98) >3 25/36 13·9 (10·0–18·5) 32/43 12·1 (8·0–17·6) 0·82 (0·49–1·39) Brain metastases Yes 10/15 14·7 (5·8–NE) 5/11 NE (14·9–NE) 2·00 (0·68–5·88) No 84/170 27·3 (20·8–30·7) 105/173 17·7 (13·3–19·9) 0·66 (0·49–0·88) Bone metastases Yes 28/54 19·1 (12·2–NE) 37/49 11·5 (8·0–14·9) 0·49 (0·29–0·81) No 66/131 27·3 (20·8–31·1) 73/135 20·1 (17·7–34·7) 0·81 (0·58–1·13) Liver metastases Yes 28/44 13·9 (9·6–22·6) 30/39 10·6 (7·9–14·9) 0·68 (0·41–1·15) No 66/141 30·0 (21·6–31·1) 80/145 19·9 (17·1–28·6) 0·70 (0·51–0·98) Lung metastases Yes 47/86 20·0 (14·7–30·7) 59/98 17·6 (12·4–20·3) 0·81 (0·55–1·19) No 47/99 27·3 (21·6–33·0) 51/86 18·7 (14·8–34·7) 0·64 (0·43–0·95) Lymph node-only disease* Yes 5/18 NE (25·0–NE) 10/13 24·6 (13·1–34·9) 0·28 (0·09–0·83) No 89/167 22·6 (18·5–30·3) 99/170 17·7 (13·3–20·0) 0·78 (0·59–1·04) Previous neoadjuvant or adjuvant chemotherapy Yes 65/125 25·0 (17·5–31·1) 66/117 19·6 (13·6–24·6) 0·82 (0·58–1·16) No 29/60 27·3 (19·5–NE) 44/67 14·9 (13·1–18·7) 0·53 (0·33–0·85) Previous taxane treatment Yes 55/96 22·6 (16·2–30·7) 52/94 19·6 (16·5–28·6) 0·96 (0·65–1·40) No 39/89 27·3 (20·0–NE) 58/90 14·9 (12·8–19·9) 0·50 (0·33–0·75) Previous anthracycline treatment Yes 59/109 21·6 (16·8–30·7) 57/101 19·6 (13·6–28·6) 0·92 (0·64–1·33) No 35/76 27·3 (21·1–NE) 53/83 14·9 (13·1–18·7) 0·49 (0·32–0·75) All PD-L1 positive status 94/185 25·0 (19·5–30·7) 110/184 18·0 (13·6–20·1) 0·71 (0·54–0·94)

0·25 1·0 2·5

Favours atezolizumab Favours placebo plus plus nab-paclitaxel nab-paclitaxel

Figure 3: Overall survival in patient subgroups Overall survival in subgroup populations in the (A) intention-to-treat and (B) PD-L1 immune cell-positive populations. ECOG=Eastern Cooperative Oncology Group. NE=not estimable. OS=overall survival. *Excludes patients with unknown or other values for indicated categories. The two enrolled patients with ECOG performance status 2 are also not plotted.

Results followed up for a median of 18·5 months (IQR 9·6–22·8) Between June 23, 2015, and May 24, 2017, we enrolled in the atezolizumab group and 17·5 months (8·4–22·4) 902 patients. The clinical cutoff for the primary in the placebo group. 451 patients were assigned to progression-free survival and first interim overall survival atezolizumab plus nab-paclitaxel and 451 to placebo plus analysis was on April 17, 2018. At the time of clinical data nab-paclitaxel and were included in the primary cutoff reported here (Jan 2, 2019), participants had been intention-to-treat efficacy analysis. Six (1%) patients in

52 www.thelancet.com/oncology Vol 21 January 2020 Articles

A 100 Atezolizumab plus nab-paclitaxel Placebo plus nab-paclitaxel 90

80

70

60

50

40

30 Progression-free survival (%) 20

10

0 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 Number at risk (number censored) Atezolizumab plus nab-paclitaxel 361 (3) 227 (13) 165 (22) 100 (25) 73 (26) 58 (30) 37 (42) 17 (56) 11 (61) 8 (64) 3 (69) 1 (71) 1 (71) ·· Placebo plus nab-paclitaxel 329 (7) 186 (13) 134 (17) 78 (18) 62 (20) 46 (23) 24 (33) 13 (39) 5 (43) 2 (46) 1 (47) ·· ·· ··

B 100 Atezolizumab plus nab-paclitaxel (PD-L1 immune-cell positive) Placebo plus nab-paclitaxel (PD-L1 immune-cell positive) 90 Atezolizumab plus nab-paclitaxel (PD-L1 immune-cell negative) Placebo plus nab-paclitaxel (PD-L1 immune-cell negative) 80

70

60

50

40

30 Progression-free survival (%) 20

10

0 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 Time since randomisation (months) Number at risk (number censored) Atezolizumab plus nab-paclitaxel 146 (1) 104 (5) 75 (9) 51 (10) 40 (11) 32 (14) 19 (23) 8 (29) 4 (32) 3 (33) 1 (35) 1 (35) 1 (35) ·· (PD-L1 immune-cell positive) Placebo plus nab-paclitaxel 128 (3) 63 (7) 45 (7) 30 (7) 23 (9) 16 (9) 9 (16) 4 (19) 3 (19) 2 (20) 1 (21) ·· ·· ·· (PD-L1 immune-cell positive) Atezolizumab plus nab-paclitaxel 215 (2) 123 (8) 90 (13) 49 (15) 33 (15) 26 (16) 18 (19) 9 (27) 7 (29) 5 (31) 2 (34) ·· ·· ·· (PD-L1 immune-cell negative) Placebo plus nab-paclitaxel 201 (4) 123 (6) 89 (10) 48 (11) 39 (11) 30 (14) 15 (17) 9 (20) 2 (24) ·· ·· ·· ·· ·· (PD-L1 immune-cell negative)

Figure 4: Progression-free survival Kaplan-Meier analysis of progression-free survival in the (A) intention-to-treat and (B) PD-L1 immune cell-positive and PD-L1 immune cell-negative populations.

the atezolizumab group and six (1%) patients in the The patient baseline characteristics, as previously placebo group did not receive any treatment. described,2 were generally well balanced between Eight (2%) patients in the placebo group received treatment groups (table 1), and they were consistent one dose of atezolizumab and were therefore included in between the intention-to-treat and PD-L1 immune cell- the safety-evaluable population for atezolizumab. Thus, positive populations. Of the intention-to-treat population, the safety-evaluable population comprised 453 patients 369 (41%) patients were PD-L1 immune cell positive in the atezolizumab group and 437 in the placebo group (185 in the atezolizumab group and 184 in the placebo (figure 1). group). www.thelancet.com/oncology Vol 21 January 2020 53 Articles

analyses in this population, even though overall survival Atezolizumab plus Placebo plus nab-paclitaxel nab-paclitaxel in the intention-to-treat population was not significantly (n=453) (n=437) different between the treatment groups. In patients with Total number of reported adverse events of any cause 7741 6061 PD-L1-positive tumours, median overall survival was All-cause adverse events 25·0 months (95% CI 19·6–30·7) with atezolizumab and Any grade 450 (99%) 428 (98%) 18·0 months (13·6–20·1) with placebo (figure 2B). The Grade 3 or 4 224 (49%) 187 (43%) stratified HR for overall survival was 0·71 (95% CI Grade 5 6 (1%)* 3 (1%)† 0·54–0·94). 24-month overall survival was 50·7% (95% CI 42·9–58·5) in the atezolizumab group versus Serious adverse events‡ 105 (23%) 81 (19%) 36·9% (29·0–44·9) in the placebo group. Median overall Adverse events leading to discontinuation of any study drug 74 (16%) 35 (8%) survival in the PD-L1 immune cell-negative population Adverse events leading to discontinuation of atezolizumab or 30 (7%) 5 (1%) placebo only was 19·7 months (95% CI 16·3–21·6) with atezolizumab Adverse events leading to discontinuation of nab-paclitaxel 74 (16%) 35 (8%) and 19·6 months (16·9–22·2) with placebo (stratified only HR 0·97; 95% CI 0·78–1·20). Subgroup analyses of Adverse events leading to dose reduction or interruption of 196 (43%) 173 (40%) overall survival in the intention-to-treat and PD-L1 nab-paclitaxel§ immune cell-positive populations are shown in figure 3. Treatment-related adverse events The final progression-free survival was already done Any-grade adverse events 437 (97%) 410 (94%) at the first interim analysis of overall survival on Grade 3 or 4 181 (40%) 133 (30%) April 17, 2018. However, because more progression-free Grade 5 2 (<1%)¶ 1 (<1%)|| survival event information had been collected at the Serious adverse events 57 (13%) 32 (7%) time of the second interim overall survival analysis, we Adverse events of special interest** present a post-hoc update to better characterise the Any-grade adverse events of special interest 262 (58%) 183 (42%) long-term progression-free survival results, especially Grade 3 or 4 36 (8%) 19 (4%) in the PD-L1 immune cell-positive patient population. Grade 5 1 (<1%)†† 1 (<1%) ‡‡ At the time of clinical data cutoff for this second Adverse events leading to atezolizumab or placebo 9 (2%) 2 (<1%) analysis (Jan 2, 2019), 379 (84%) of 451 patients in the discontinuation atezolizumab group and 404 (90%) of 451 patients in Treated with systemic corticosteroids within 30 days of onset 62 (14%) 28 (6%) the placebo group had disease progression or had of the adverse event of special interest died. In the intention-to-treat population, median Data are n (%). *Aspiration (n=1), pneumonia (n=1), autoimmune hepatitis (n=1), (unexplained) death (n=1), pulmonary progression-free survival was 7·2 months (95% CI embolism (n=1), and septic shock (n=1). †Hepatic failure (n=1), acute myocardial infarction (n=1), and (unexplained) death 5·6–7·4) with atezolizumab and 5·5 months (5·3–5·6) (n=1). ‡For serious adverse events with prevalence of 1% or more see appendix (p 9). §Dose reductions of atezolizumab or placebo were not permitted as per protocol. ¶Atezolizumab-related autoimmune hepatitis (n=1) and septic shock related with placebo (stratified HR 0·80 [95% CI 0·69–0·92], to nab-paclitaxel (n=1). ||Hepatic failure related to either drug (n=1). **As predefined by the study funder. ††Autoimmune p=0·0021; figure 4A). 24-month progression-free hepatitis (n=1). ‡‡Hepatic failure (n=1). survival was 10% (95% CI 7–13) in the nab-paclitaxel Table 2: Safety summary group and 6% (4–9) in the placebo group. Sites of new lesions per investigator assessment are shown in the appendix p 8. At the time of clinical data cutoff for our analysis In the PD-L1 immune cell-positive population, the (Jan 2, 2019), 255 (57%) of 451 patients in the atezolizumab updated median progression-free survival was 7·5 months group and 279 (62%) of 451 patients in the placebo group (95% CI 6·7–9·2) with atezolizumab and 5·3 months had died. In the intention-to-treat population, median (3·8–5·6) with placebo (stratified HR 0·63 [95% CI overall survival was 21·0 months (95% CI 19·0–22·6) 0·50–0·80], p<0·0001; figure 4B). 12-month progression- with atezolizumab and 18·7 months (16·9–20·3) with free survival was 30·3% (95% CI 23·5–37·2) in the placebo (figure 2A). The stratified HR for overall survival atezolizumab group and 17·3% (11·7–22·9) in the placebo was 0·86 (95% CI 0·72–1·02, p=0·078), with a p value of group, and 24-month progression-free survival was 12·4% 0·070 for the interaction between PD-L1 status and (95% CI 6·5–18·3) versus 7·4% (2·8–12·0), respectively. overall survival benefit with atezolizumab. In a post-hoc In the PD-L1 immune cell-negative patients, median analysis (for which timepoints were planned before progression-free survival was 5·6 months (95% CI unblinding of the study), 24-month overall survival was 5·5–7·3) in the atezolizumab group versus 5·6 months 42·4% (95% CI 37·3–47·4) in the atezolizumab group (5·4–7·3) in the placebo group (HR 0·93 [95% CI and 38·7% (33·7–43·6) in the placebo group. Visual 0·77–1·11). Final analyses of the secondary efficacy inspection of the Kaplan-Meier curves did not indicate endpoints objective responses and duration of response any non-proportionality in the hazard ratio. have previously been reported.2 Because PD-L1 immune cell-positive status predicts The safety update with an additional 4·5 months of clinical benefit with atezolizumab plus nab-paclitaxel follow-up since the primary analysis (data cutoff (as reflected in health authority approvals for patients Sept 3, 2018) are summarised in tables 2 and 3. The with PD-L1-positive tumours), we did exploratory 890 safety-evaluable patients were followed up for safety

54 www.thelancet.com/oncology Vol 21 January 2020 Articles

Atezolizumab plus nab-paclitaxel group (n=453) Placebo plus nab-paclitaxel group (n=437) Grade 1–2 Grade 3 Grade 4 Grade 5 Grade 1–2 Grade 3 Grade 4 Grade 5 Any adverse event 223 (49%) 197 (44%) 24 (5%) 6 (1%) 240 (55%) 157 (36%) 28 (6%) 3 (1%) Alopecia 256 (57%) 2 (<1%) 0 0 249 (57%) 1 (<1%) 0 0 Nausea 204 (45%) 5 (1%) 0 0 160 (37%) 8 (2%) 0 0 Fatigue 195 (43%) 17 (4%) 0 0 180 (41%) 15 (3%) 0 0 Diarrhoea 139 (31%) 8 (2%) 0 0 143 (33%) 9 (2%) 0 0 Cough 114 (25%) 0 0 0 85 (20%) 0 0 0 Constipation 113 (25%) 3 (1%) 0 0 106 (24%) 1 (<1%) 0 0 Anaemia 112 (25%) 14 (3%) 0 0 105 (24%) 13 (3%) 0 0 Headache 106 (23%) 2 (<1%) 0 0 91 (21%) 4 (1%) 0 0 Decreased appetite 87 (19%) 3 (1%) 0 0 77 (18%) 3 (1%) 0 0 Vomiting 85 (19%) 4 (1%) 0 0 70 (16%) 5 (1%) 0 0 Pyrexia 85 (19%) 3 (1%) 0 0 48 (11%) 0 0 0 Arthralgia 81 (18%) 1 (<1%) 0 0 70 (16%) 1 (<1%) 0 0 Rash 76 (17%) 2 (<1%) 0 0 71 (<16%) 2 (1%) 0 0 Neuropathy peripheral 74 (16%) 25 (6%) 0 0 86 (20%) 12 (3%) 0 0 Dyspnoea 69 (15%) 3 (1%) 0 0 62 (14%) 3 (1%) 0 0 Peripheral oedema 68 (15%) 1 (<1%) 0 0 62 (14%) 6 (1%) 0 0 Back pain 66 (15%) 6 (1%) 0 0 57 (13%) 2 (1%) 0 0 Myalgia 65 (14%) 2 (<1%) 0 0 64 (15%) 3 (1%) 0 0 Dysgeusia 65 (14%) 0 0 0 59 (14%) 0 0 0 Dizziness 64 (14%) 0 0 0 46 (11%) 0 0 0 Peripheral sensory neuropathy 63 (14%) 9 (2%) 0 0 44 (10%) 8 (2%) 0 0 Pruritus 63 (14%) 0 0 0 45 (10%) 0 0 0 Hypothyroidism 63 (14%) 0 0 0 16 (4%) 0 0 0 Neutropenia 59 (13%) 28 (6%) 10 (2%) 0 32 (7%) 23 (5%) 13 (3%) 0 Asthenia 54 (12%) 3 (1%) 0 0 46 (11%) 4 (1%) 0 0 Urinary tract infection 52 (12%) 4 (1%) 0 0 43 (10%) 3 (1%) 0 0 Insomnia 52 (12%) 0 0 0 50 (11%) 3 (1%) 0 0 Pain in extremity 50 (11%) 2 (0·4) 0 0 42 (10%) 1 (<1%) 0 0 Nasopharyngitis 48 (11%) 0 0 0 37 (9%) 0 0 0 Abdominal pain 45 (10%) 2 (<1%) 0 0 52 (12%) 1 (<1%) 0 0 Aspartate aminotransferase increased 36 (8%) 8 (2%) 1 (<1%) 0 34 (8%) 8 (2%) 1 (<1%) 0 Neutrophil count decreased 35 (8%) 17 (4%) 5 (1%) 0 33 (8%) 13 (3%) 3 (1%) 0 Hypertension 18 (4%) 4 (1%) 0 0 15 (3%) 10 (2%) 1 (<1%) 0 Hypokalaemia 17 (4%) 8 (2%) 3 (1%) 0 6 (1%) 4 (1%) 0 0 Pneumonia 12 (3%) 10 (2%) 0 1 (<1%) 4 (1%) 3 (1%) 0 0

Data are n (%). All grade 1–2 adverse events that occurred in at least 10% of patients in either group, and grade 3–5 events that occurred in at least 2% of patients in either group, are shown. A full list of all adverse events in the safety population is in appendix pp 10–14.

Table 3: Adverse events in the safety population for a median duration of 15·6 months (IQR 8·8–20·0). adverse events were balanced across treatment groups, Treatment duration and exposure at the time of the safety with the exception of peripheral neuropathy (grade 3 update are shown in the appendix (p 9). only; 25 [6%] of 453 patients in the atezolizumab group vs In the atezolizumab, 224 (49%) of 453 patients had 12 [3%] of 437 patients in the placebo group). The grade 3 grade 3 or 4 adverse events compared with 187 (43%) of peripheral neuropathy was deemed to be taxane-related, 437 patients in the placebo plus nab-paclitaxel group. The which is known to be cumulative. No cases of Guillain- most common grade 3–4 adverse events were neutro­ Barré syndrome were reported. 74 (16%) of 453 patients penia (38 [8%] of 453 patients in the atezolizumab group in the atezolizumab group had an adverse event leading vs 36 [8%] of 437 patients in the placebo group), peripheral to discontinuation of any study drug (vs 35 [8%] of neuropathy (25 [6%] vs 12 [3%]), decreased neutrophil 437 patients in the placebo group; appendix p 9). Adverse count (22 [5%] vs 16 [4%]), and fatigue (17 [4%] vs 15 [3%]). events leading to discontinuation of atezolizumab or As was the case in the primary analysis, all grade 3–4 placebo occurred in 30 (7%) patients in the atezolizumab www.thelancet.com/oncology Vol 21 January 2020 55 Articles

group and in five (1%) patients in the placebo group. Consistent with the primary analysis, the leading cause for treatment discontinuation of atezolizumab plus nab- paclitaxel due to toxicity was peripheral neuropathy (in 1 (<1%) 0 0 0 0 0 0 Any grade Adverse events events Adverse of special interest leading to withdrawal of placebo 20 [4%] patients). Serious adverse events were reported in 105 (23%) of 453 patients in the atezolizumab group compared with 81 (19%) of 437 patients in the placebo group 1·0 0·8 0·5 0·8 NA (0·4–7·2) (0·7–1·0) (0·2–1·4) NE (NE–NE) NE (1·9–NE) (0·5–1·5) Median duration, months (appendix p 9). 262 (58%) of 453 patients in the atezolizumab group time and 183 (42%) of 437 in the placebo group had an adverse event of special interest (table 2, table 4). One death due (2·3–7·8) (1·4–6·4) NA (0·4–2·8) (1·9–3·6) (1·0–1·0) (1·8–6·5) to onset, months First adverse event of special event First adverse interest (any grade) to an adverse event of special interest occurred in each treatment group (hepatitis in both cases). Consistent with the primary analysis, the adverse events of special interest that differed substantially between the atezolizumab 0 0/1 (0%) 1·0 8/20 (40%) 2·0 6/6 (100%) 3·2 group and the placebo group were any-grade rash, hypo­ Resolved adverse adverse Resolved of events special interest thyroidism, hyperthyroidism, pneumonitis, and adrenal insufficiency (table 4). Adverse events leading to death occurred in six (1%) patients in the atezolizumab group and three (1%) patients in the placebo group. Treatment- related deaths occurred in two (<1%) patients in the atezolizumab group (one due to autoimmune hepatitis 1 (<1%) 0 7 (2%) 1 (<1%) 6/7 (86%) 5·4 3 (1%) 1 (<1%) 3/3 (100%) 2·1 0 0 6 (1%) 0 related to atezolizumab and one due to septic shock 20 (5%) 0 Placebo plus nab-paclitaxel group (n=437) Placebo plus nab-paclitaxel Adverse events of special events Adverse interest related to nab-paclitaxel only) and one (<1%) patient in the placebo group (hepatic failure). No new treatment- related deaths have been reported since the primary clinical data cutoff date (April 17, 2018). 274 (61%) of 451 patients from the atezolizumab group 1 (<1%) 1 (<1%) 1 (<1%) 115 (26%) 2 (<1%) 89/115 (77%) 1·0 1 (<1%) 2 (<1%) 0 0 Any grade Any gradeor 4 Grade 3 Any grade Median Adverse events events Adverse of special interest leading to withdrawal of atezolizumab and 295 (65%) of 451 patients from the placebo group received at least one subsequent anticancer therapy (appendix pp 15–17). Anti­metabolites (eg, capecitabine, gemcitabine, fluoro­uracil, and methotrexate) were 1·0 1·9 3·2 2·0 5·5 0·8 (0·3–6·3) 13·0 (1·0–NE) (0·8–2·3) (2·7–NE) (1·1–8·1) (1·5–5·8) (2·2–NE) Median duration, months the most common subsequent therapy, received by 191 (42%) of 451 patients from the atezolizumab group and 204 (45%) of 451 patients from the placebo group. Furthermore, 122 (27%) of 451 patients in each treatment group received at least one subsequent therapy with a (2·9–8·7) (4·9–8·1) (4·7–8·9) (0·4–4·0) (1·9–5·5) (1·9–6·8) (2·2–6·0) Median time to onset, months First adverse event of special event First adverse interest (any grade) platinum compound. 17 (4%) of 451 patients from the atezolizumab group and 26 (6%) of 451 patients from the placebo group received at least one subsequent therapy with a mono­clonal antibody targeting PD-L1, PD-1, or CTLA-4. 38/80 (48%) 3·8 Resolved adverse adverse Resolved of events special interest Discussion To our knowledge, IMpassion130 was the first phase 3 study reported that evaluated an antibody targeting the PD-L1–PD-1 signalling pathway in patients with unresect­able, advanced or metastatic triple-negative breast cancer. In the primary analysis, IMpassion130 5 (1%) 1 (<1%) 4/5 (80%) 6·7 4 (<1%) 1 (<1%) 3/4 (75%) 4·9

11 (2%) 7 (2%) 6/11 (55%) 4·6 met its coprimary progression-free survival endpoint, 155 (34%) 4 (<1%) 126/155 (81%) 1·3 Adverse events of special events Adverse interest Any gradeor 4 Grade 3 Any grade Atezolizumab plus nab-paclitaxel group (n=451) plus nab-paclitaxel Atezolizumab showing a significant progression-free survival benefit with atezolizumab plus nab-paclitaxel compared with placebo plus nab-paclitaxel as a first-line treatment for patients with advanced or metastatic triple-negative 2 Hepatitis (diagnosis)* Colitis Adrenal insufficiency* Rash Hyperthyroidism 21 (5%)Pneumonitis 1 (<1%) 16 (4%) 16/21 (76%) 2 (<1%) 3·8 12/16 (75%) 4·7 Hypothyroidism 80 (18%) 0 breast cancer. Since no treatment effect was seen in the Data are n (%), n/N (%), or median (IQR). The table lists adverse events of special interest as medical concepts (grouped by Medical Dictionary for Regulatory Activities-preferred terms) occurring in more than ten patients in the atezolizumab plus nab- ten patients in than occurring in more terms) Activities-preferred Medical Dictionary for Regulatory of special interest as medical concepts (grouped by events table lists adverse The or median (IQR). Data are n (%), n/N of a hepatitis suggestive that represent events terms of group NE=not estimable. NA=not applicable. *Funder-defined hepatitis (laboratory abnormalities), and additionally including colitis adrenal insufficiency. excluding group, paclitaxel only). test abnormalities function with liver associated to events opposed diagnosis (as of special interest events Clinically relevant adverse 4: Table PD-L1 immune cell-negative subgroup, the clinical

56 www.thelancet.com/oncology Vol 21 January 2020 Articles

benefit of atezolizumab in the intention-to-treat survival benefit with atezoli­zumab, with an improvement population was driven by the improvement in in median progression-free survival of 2·2 months with progression-free survival in the PD-L1 immune cell- atezolizumab plus nab-paclitaxel compared with placebo positive subgroup.21 The first prespecified interim overall plus nab-paclitaxel in this population. survival analysis was aligned with the primary The consistent results across several analyses of progression-free survival analysis and was immature IMpassion130 have established the benefit of PD-L1 with an information fraction of 59%. In this Article, we blockade with atezolizumab in combination with nab- report the second prespecified interim overall survival paclitaxel as a first-line therapy in PD-L1 immune cell- analysis, a concurrent progression-free survival update, positive unresectable, locally advanced or metastatic and a safety update for IMpassion130. These new data triple-negative breast cancer. Any overall survival benefit provide valuable information to health-care providers in seemed to be limited to PD-L1 immune cell-positive the USA, EU, and other territories where atezolizumab patients, suggesting that the combination of atezo­ plus nab-paclitaxel is an approved treatment regimen. lizumab plus nab-paclitaxel can effectively and specifically Consistent with the first interim overall survival harness pre-existing anti-tumour immunity in patients analysis,2 the results from this second interim overall with unresectable, locally advanced or metastatic PD-L1 survival analysis showed that although median overall immune cell-positive triple-negative breast cancer. survival in the intention-to-treat population was longer Consistent with the benefit pattern observed in other with the combination of atezolizumab plus nab-paclitaxel cancer immunotherapy studies,26,27 the absolute benefit than with placebo plus nab-paclitaxel, the significance in overall survival was greater than the absolute benefit boundary was not crossed. Therefore, since median for progression-free survival in IMpassion130. overall survival did not differ significantly between the No new safety signals were observed, and toxicity data treatment groups in this analysis, the statistical testing remained consistent with those from the previously hierarchy did not allow for formal testing of overall published progression-free survival analysis.2 No survival in the PD-L1 immune cell-positive patient cumulative toxicities or new or late-onset safety signals population. However, the magnitude of overall survival were seen with longer follow-up. Potentially immune- benefit with atezolizumab in PD-L1 immune-positive mediated adverse events of special interest, a toxicity patients remained clinically meaningful, with an increase concern with checkpoint inhibitors as a class,28 were of 7 months in median overall survival with atezolizumab manageable, but represent a new class of toxicity that is plus nab-paclitaxel compared with placebo plus nab- important to identify and manage appropriately. paclitaxel (HR 0·71 [95% CI 0·54–0·94]). With a median The median times to onset of adverse events of special follow-up of 18·0 months and an information fraction of interest were consistent with those observed in 80%, the reported overall survival results can now be atezolizumab monotherapy trials.23 considered mature. The event-driven final overall survival The limitations of this study include the ineligibility of analysis will be reported in a subsequent publication. patients who had received previous radiotherapy or A similar proportion of patients in both treatment chemotherapy (including taxanes) in the curative setting groups (61% of those in the atezolizumab group and less than 12 months before randomisation for this study. 65% of those in the placebo group) received post-protocol Findings from this study should therefore not be treat­ments. The remaining patients were either still extrapolated to the excluded patient population with receiving treatment or were not eligible for further rapid disease progression. Furthermore, the statistical treatment because of progression. In some cases, the testing hierarchy did not allow us to do formal statistical second-line therapy was not recorded. The decision for or testing of overall survival results in the PD-L1 immune against post-protocol treatment was not recorded in our cell-positive population because the requirement for a study. The most frequent post-protocol therapies, which significant overall survival improvement in the intention- might have affected overall survival in the study population, to-treat population was not met. were antimetabolites. Very few patients received immune In conclusion, our analysis indicates that there might be checkpoint inhibitors, and the propor­tion who received a clinically meaningful benefit of using atezolizumab in these drugs was similar between the two treatment groups. combination with nab-paclitaxel as a first-line treatment­ The final primary progression-free survival analysis for patients with PD-L1 immune cell-positive unresectable, results of this trial were previously reported in 2018, with locally advanced or metastatic triple-negative breast 358 (79%) of the 451 patients in the atezolizumab group cancer. The atezolizumab plus nab-paclitaxel combination and 378 (84%) of 451 in the placebo group having shown is, to the best of our knowledge, the first therapy to cross disease progression or died,2 and the progression-free the 2-year landmark overall survival benefit in PD-L1 survival results were updated at this second interim immune cell-positive metastatic triple-negative breast overall survival analysis timepoint to further support the cancer. Routine testing for PD-L1 immune-cell expression results of the primary analysis. The updated progression- in unresectable, locally advanced or metastatic triple- free survival results in PD-L1 immune cell-positive negative breast cancer with the approved companion patients support the previously reported progression-free diagnostic (VENTANA PD-L1 SP142 assay) should be used www.thelancet.com/oncology Vol 21 January 2020 57 Articles

to identify patients who might benefit from treatment Bristol Myers-Squibb, Corvus, the US Department of Defense, with atezolizumab plus nab-paclitaxel. EMD Serono, Genentech, Maxcyte, Merck, the National Cancer Institute, the NSABP Foundation, Roche, the Translational Breast Cancer Research Contributors Consortium, and HeritX; and reports royalties from Aduro. PS, HSR, SA, AS, HI, VD, LM, SYC, AH, SL, and LAE designed the study. PS, AS, VH, LM, SYC, AH, and SL developed the study methods. Data sharing PS, HSR, SA, AS, CHB, HI, VD, VH, SYC, AH, SL, and LAE collected Qualified researchers can request access to individual patient-level data For the data request platform data. VH, LM, SYC, and AH had access to the raw data. PS, HSR, SA, through the clinical study data request platform. Details on Roche’s criteria see https://www. AS, CHB, HI, VD, VH, LM, SYC, VM, AH, EPW, SL, and LAE analysed for eligible studies are available online. Those interested in accessing study clinicalstudydatarequest.com and interpreted the data. All authors contributed to manuscript writing, data should view Roche’s Global Policy on the Sharing of Clinical For Roche’s criteria see approved the final version, and are accountable for all aspects of the Information and how to request access to related clinical study documents. https://clinicalstudydatarequest. report. All authors contributed to drafting the manuscript, provided final Acknowledgments com/Study-Sponsors/Study- approval to publish, and agree to be accountable for all aspects of the This study was funded by F Hoffmann-La Roche/Genentech, a member Sponsors-Roche.aspx manuscript. of the Roche Group. We thank the patients participating in this trial, For the Roche policy on data Declaration of interests patient families, nurses, research coordinators, data managers, and sharing see http://www.roche. All authors received support from F Hoffmann-La Roche during the clinical study site investigators (appendix p 2). Medical writing assistance com/research_and_ conduct of the study. PS reports grants and non-financial support from for this manuscript was provided by Steffen Biechele (Health Interactions, development/who_we_are_ F Hoffmann-La Roche, during the conduct of the study; grants and San Francisco, CA, USA) and was funded by F Hoffmann-La Roche. how_we_work/clinical_trials/ research support to their institution from AstraZeneca, Genentech, F Hoffmann-La Roche/Genentech funded the study, provided the study our_commitment_to_data_ F Hoffmann-La Roche, Oncogenex, Novartis, and Astellas; personal fees drug (atezolizumab) and the matched placebo, and collaborated with sharing.htm from Pfizer, AstraZeneca, Novartis, F Hoffmann-La Roche, Merck, academic authors on the study design, data collection, data analysis, and Boehringer lngelheim, Bayer, Eisai, Celgene, and Puma, outside the interpretation. Celgene provided nab-paclitaxel, but the company had no submitted work; is an uncompensated steering committee member for the role in the study design, data collection, or analysis, although they did IMpassion130 trial; and his spouse is an employee of F Hoffman-La Roche. review the manuscript. HSR reports research support for the IMpassion130 study, and editorial References support from F Hoffman-La Roche; institutional support from Pfizer, 1 Bonotto M, Gerratana L, Poletto E, et al. Measures of outcome in Novartis, Eli Lilly, Merck, OBI, Eisai, and Plexxikon; institutional and travel metastatic breast cancer: insights from a real-world scenario. support from Genentech/Roche and MacroGenics; travel support from Oncologist 2014; 19: 608–15. Puma, Mylan, Novartis, and Pfizer; research support from 2 Schmid P, Adams S, Rugo HS, et al. Atezolizumab and Immunomedics; research and travel support from Daiichi Sankyo; nab-paclitaxel in advanced triple-negative breast cancer. and honoraria from Celltrion. SA reports institutional support from N Engl J Med 2018; 379: 2108–21. Genentech, Merck, Amgen, Bristol-Meyers Squibb, Novartis, and Celgene. 3 National Comprehensive Cancer Network. NCCN clinical practice AS reports research grants from Celgene, F Hoffman-La Roche, AbbVie, guidelines in oncology. Breast cancer. V1.2019. https://www.nccn. and Molecular Partners; consulting fees and travel expenses from org/professionals/physician_gls/pdf/breast.pdf (accessed F Hoffman-La Roche and AstraZeneca; and honoraria from F Hoffman-La March 7, 2019). Roche, Celgene, AstraZeneca, Novartis, Merck Sharp and Dohme, Tesaro, 4 Cardoso F, Senkus E, Costa A, et al. 4th ESO-ESMO International and Eli Lilly; and honoraria and travel expenses from Pfizer. CHB reports consensus guidelines for advanced breast cancer (ABC 4). Ann Oncol 2018; 29: 1634–57. research grants from Pfizer, Novartis, Amgen, AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, Roche/Genentech, Eli Lilly, Sanofi, Taiho 5 Senkus E, Kyriakides S, Ohno S, et al. Primary breast cancer: ESMO clinical practice guidelines for diagnosis, treatment and follow-up. Pharmaceutical, Mylan, Merrimack, Merck, AbbVie, Astellas, BioMarin, Ann Oncol 2015; 26 (suppl 5): v8–30. Bristol Myers-Squibb, Daiichi Sankyo, Abraxis BioScience, AB Science, 6 Roche. Avastin (bevacizumab): summary of product characteristics. Asana BioSciences, Medivation, Exelixis, ImClone Systems, LEO Pharma, 2017. https://www.ema.europa.eu/en/documents/product- and Millennium; and consulting fees from Roche/Genentech, Boehringer information/avastin-epar-product-information_en.pdf (accessed Ingelheim, GlaxoSmithKline, Novartis, Pfizer, Eisai, Bayer, Merck Sharp Nov 5, 2019). and Dohme, and AstraZeneca. HI reports honoraria, consulting fees, and 7 AstraZeneca. Lynparza (olaparib): summary of product research support for the IMpassion130 study; editorial support from characteristics. 2014. https://www.ema.europa.eu/en/documents/ F Hoffman-La Roche and Chugai; serves as an uncompensated member of product-information/lynparza-epar-product-information_en.pdf the steering committee for the IMpassion130 trial; and reports honoraria (accessed Nov 5, 2019). and consulting fees from Novartis, AstraZeneca, Pfizer, Eli Lilly, and 8 Pfizer. Talzenna (talazoparib): package insert. 2018. http://labeling. Daiichi Sankyo. VD reports honoraria for serving on advisory boards pfizer.com/ShowLabeling.aspx?id=11046 (accessed Nov 5, 2019). for F Hoffman-La Roche, Genentech, Pfizer, Eli Lilly, Novartis, 9 Tutt A, Tovey H, Cheang MCU, et al. Carboplatin in Daiichi Sankyo, AstraZeneca, AbbVie, and Odonate. VH, SYC, VM, BRCA1/2-mutated and triple-negative breast cancer BRCAness and AH report employment and received stock from Genentech/Roche subgroups: the TNT Trial. Nat Med 2018; 24: 628–37. during the conduct of the study. LM reports stock ownership in 10 Bajaj P, Latremouille-Viau D, Guerin A, et al. What are the Genentech/Roche and has a use patent pending during the conduct of the treatment patterns and overall survival (OS) in patients with study. EPW reports honoraria from Eli Lilly, Leap, Genentech, Infinite MD, metastatic triple-negative breast cancer (mTNBC) in US clinical Carrick Therapeutics, GlaxoSmithKline, Jounce, Genomic Health, Merck, practice? Ann Oncol 2017; 28: mdx365·031. and Seattle Genetics; and is a scientific advisory board member for Leap. 11 den Brok WD, Speers CH, Gondara L, Baxter E, Tyldesley SK, SL reports research funding to her institution from Novartis, Lohrisch CA. Survival with metastatic breast cancer based on initial Bristol-Myers Squibb, Merck, Roche, Genentech, Puma, Pfizer, and presentation, de novo versus relapsed. Breast Cancer Res Treat 2017; Eli Lilly; has acted as an unpaid consultant to Seattle Genetics, Pfizer, 161: 549–56. Novartis, Bristol Myers-Squibb, Merck, AstraZeneca, and 12 Gobbini E, Ezzalfani M, Dieras V, et al. Time trends of overall Roche/Genentech; and acted as consultant to Aduro Biotech (fees paid to survival among metastatic breast cancer patients in the real-life ESME cohort. Eur J Cancer 2018; 96: 17–24. her institution). LAE is co-chair of the steering committee for the IMpassion130 study and chair of the KATE2 study steering committee; 13 Yardley DA, Coleman R, Conte P, et al. nab-Paclitaxel plus carboplatin or gemcitabine versus gemcitabine plus carboplatin as reports honoraria from AbbVie, Amgen, Celgene, Gritstone, MedImmune, first-line treatment of patients with triple-negative metastatic breast Peregrine, and Syndax; honoraria and travel support from AstraZeneca, cancer: results from the tnAcity trial. Ann Oncol 2018; 29: 1763–70. Bayer, MacroGenics, Replimune, Vaccinex; travel support from Bristol 14 Adams S, Loi S, Toppmeyer DL, et al. Pembrolizumab monotherapy Myers-Squibb, Genentech/Roche, and Novartis; has potential future stock for previously untreated, PD-L1-positive, metastatic triple-negative from Molecuvax; reports institutional support from Aduro Biotech, breast cancer: cohort B of the phase II KEYNOTE-086 study. AstraZeneca, the Breast Cancer Research Foundation, Ann Oncol 2019; 30: 405–11.

58 www.thelancet.com/oncology Vol 21 January 2020 Articles

15 Adams S, Diamond JR, Hamilton E, et al. Atezolizumab plus 22 Genentech. Tecentriq (atezolizumab): package insert. 2019. nab-paclitaxel in the treatment of metastatic triple-negative breast https://www.gene.com/download/pdf/tecentriq_prescribing.pdf cancer with 2-year survival follow-up: a phase 1b clinical trial. (accessed Nov 5, 2019). JAMA Oncol 2019; 5: 334–42. 23 Roche. Tecentriq (atezolizumab): summary of product 16 Emens LA, Cruz C, Eder JP, et al. Long-term clinical outcomes and characteristics. 2019. https://www.ema.europa.eu/en/documents/ biomarker analyses of atezolizumab therapy for patients with product-information/tecentriq-epar-product-information_en.pdf metastatic triple-negative breast cancer: a phase 1 study. (accessed Nov 5, 2019). JAMA Oncol 2019; 5: 74–82. 24 Wolff AC, Hammond ME, icksH DG, et al. Recommendations for 17 Sabatier R, Finetti P, Mamessier E, et al. Prognostic and predictive human epidermal growth factor receptor 2 testing in breast cancer: value of PDL1 expression in breast cancer. Oncotarget 2015; American Society of Clinical Oncology/College of American 6: 5449–64. Pathologists clinical practice guideline update. J Clin Oncol 2013; 18 Ghebeh H, Mohammed S, Al-Omair A, et al. The B7-H1 (PD-L1) 31: 3997–4013. T lymphocyte-inhibitory molecule is expressed in breast cancer 25 Hammond ME, Hayes DF, Dowsett M, et al. American Society of patients with infiltrating ductal carcinoma: correlation with Clinical Oncology/College Of American Pathologists guideline important high-risk prognostic factors. Neoplasia 2006; 8: 190–98. recommendations for immunohistochemical testing of estrogen 19 Cimino-Mathews A, Thompson E, Taube JM, et al. PD-L1 (B7-H1) and progesterone receptors in breast cancer. J Clin Oncol 2010; expression and the immune tumor microenvironment in primary 28: 2784–95. and metastatic breast carcinomas. Hum Pathol 2016; 47: 52–63. 26 Rittmeyer A, Barlesi F, Waterkamp D, et al. Atezolizumab versus 20 Mittendorf EA, Philips AV, Meric-Bernstam F, et al. PD-L1 docetaxel in patients with previously treated non-small-cell lung expression in triple-negative breast cancer. Cancer Immunol Res cancer (OAK): a phase 3, open-label, multicentre randomised 2014; 2: 361–70. controlled trial. Lancet 2017; 389: 255–65. 21 Emens LA, Loi S, Rugo HS, et al. IMpassion130: efficacy in 27 Socinski MA, Jotte RM, Cappuzzo F, et al. Atezolizumab for first- immune biomarker subgroups from the global, randomized, line treatment of metastatic nonsquamous NSCLC. N Engl J Med double-blind, placebo-controlled, phase III study of atezolizumab 2018; 378: 2288–301. + nab-paclitaxel in patients with treatment-naïve, locally advanced 28 Brahmer JR, Lacchetti C, Schneider BJ, et al. Management of or metastatic triple-negative breast cancer. San Antonio Breast immune-related adverse events in patients treated with immune Cancer Symposium; San Antonio, TX, USA; Dec 4–8, 2018 checkpoint inhibitor therapy: American Society of Clinical Oncology (abstr GS1-04). clinical practice guideline. J Clin Oncol 2018; 36: 1714–68.

www.thelancet.com/oncology Vol 21 January 2020 59 Letter https://doi.org/10.1038/s41586-019-1600-x

The Drug Rediscovery protocol facilitates the expanded use of existing anticancer drugs ­D. L. van der Velden1,2,21, L. R. Hoes1,2,3,21, H. van der Wijngaart2,3,4,21, J. M. van Berge Henegouwen2,3,5,21, E. van Werkhoven6, P. Roepman7, R. L. Schilsky8, W. W. J. de Leng9, A. D. R. Huitema10,11, B. Nuijen11, P. M. Nederlof12, C. M. L. van Herpen13, D. J. A. de Groot14, L. A. Devriese15, A. Hoeben16, M. J. A. de Jonge17, M. Chalabi1,18, E. F. Smit2,19, A. J. de Langen19, N. Mehra13, M. Labots4, E. Kapiteijn5, S. Sleijfer2,17, E. Cuppen3,7,20, H. M. W. Verheul4,13, H. Gelderblom5 & E. E. Voest1,2,3*

The large-scale genetic profiling of tumours can identify potentially is taken into consideration. However, with regards to drug sensitivity, actionable molecular variants for which approved anticancer the importance of a given genetic or molecular variant is usually tested drugs are available1–3. However, when patients with such variants in the subtype of cancer that most frequently contains this variant. are treated with drugs outside of their approved label, successes The importance of the same variant in other cancers often remains and failures of targeted therapy are not systematically collected unknown. Third, as drug development is challenging for rare subtypes or shared. We therefore initiated the Drug Rediscovery protocol, of cancer, this can create inequality in care12. Finally, with growing an adaptive, precision-oncology trial that aims to identify signals pressure from society to increase the success rate of drug-development of activity in cohorts of patients, with defined tumour types and trials13, there is hesitation amongst payers to reimburse large-scale molecular variants, who are being treated with anticancer drugs sequencing efforts before they have proof that these efforts will make outside of their approved label. To be eligible for the trial, patients healthcare more sustainable. As a result, we are not using the full have to have exhausted or declined standard therapies, and have potential of rapidly expanding technological advances, knowledge of malignancies with potentially actionable variants for which no biomarkers and the spectrum of approved anticancer drugs for our approved anticancer drugs are available. Here we show an overall patients. rate of clinical benefit—defined as complete or partial response, or The Center for Personalized Cancer Treatment was founded in as stable disease beyond 16 weeks—of 34% in 215 treated patients, 201014 to address these issues. In this network (which now connects comprising 136 patients who received targeted therapies and 45 hospitals in the Netherlands), patients with all types of metastatic 79 patients who received immunotherapy. The overall median cancer are offered the opportunity to undergo a fresh tumour biopsy duration of clinical benefit was 9 months (95% confidence interval of for whole-genome sequencing (WGS) before starting systemic 8–11 months), including 26 patients who were experiencing ongoing anticancer treatment. The WGS results are combined with treatment clinical benefit at data cut-off. The potential of the Drug Rediscovery outcomes in a national, centralized database for research purposes, and protocol is illustrated by the identification of a successful cohort returned to the physician who is treating the patient for future planning of patients with microsatellite instable tumours who received of treatment. This initiative has contributed to the identification of nivolumab (clinical benefit rate of 63%), and a cohort of patients potentially actionable variants in cancers that are not routinely tested with colorectal cancer with relatively low mutational load who for these variants. To provide treatment opportunities for patients in experienced only limited clinical benefit from immunotherapy. The whom such variants were identified (while simultaneously collecting Drug Rediscovery protocol facilitates the defined use of approved clinical outcomes), we began the Drug Rediscovery protocol (DRUP), drugs beyond their labels in rare subgroups of cancer, identifies in which we seek to expand the use of targeted therapies that have been early signals of activity in these subgroups, accelerates the clinical approved by the European Medicines Agency (EMA) and/or US Food translation of new insights into the use of anticancer drugs outside and Drug Administration (FDA) beyond the approved indications of of their approved label, and creates a publicly available repository these therapies. of knowledge for future decision-making. The DRUP is an ongoing, prospective multi-drug and pan-cancer The precision treatment of cancer holds great promise for patients in trial. Patients who are eligible are those who have progression of an terms of life extension and quality of life1,2,4–7. However, early studies advanced or metastatic solid tumour, multiple myeloma or B-cell and experiences with genetically and molecularly informed decisions non-Hodgkin lymphoma, and no suitable standard-treatment options. regarding treatment have also identified considerable hurdles, which A potentially actionable genetic or molecular variant, which can be may jeopardize the way in which we capitalize on precision medi- matched to one of the drugs available in the study (Extended Data cine8–11. First, populations of patients who are eligible for specific Table 1), must have been identified via regular diagnostics or by the treatments or trials become smaller and trials accrue slower, owing Center for Personalized Cancer Treatment. to pre-selection by targeted sequencing of candidate variants and to In recognition of the importance of tissue context, the trial design slow implementation of pre-selection tests. Second, these candidate allows for an unlimited number of parallel cohorts (each defined by variants can, in general, be appreciated only when their tissue context tumour type, molecular variant and study treatment) (Fig. 1). For

1Department of Molecular Oncology & Immunology, Netherlands Cancer Institute, Amsterdam, The Netherlands. 2Center for Personalized Cancer Treatment, Rotterdam, The Netherlands. 3Oncode Institute, Utrecht, The Netherlands. 4Department of Medical Oncology, Amsterdam UMC, Vrije Universiteit Amsterdam, Cancer Center Amsterdam, Amsterdam, The Netherlands. 5Department of Medical Oncology, Leiden University Medical Center, Leiden, The Netherlands. 6Biometrics Department, Netherlands Cancer Institute, Amsterdam, The Netherlands. 7Hartwig Medical Foundation, Amsterdam, The Netherlands. 8American Society of Clinical Oncology, Alexandria, VA, USA. 9Department of Pathology, University Medical Center Utrecht, Utrecht, The Netherlands. 10Department of Clinical Pharmacy, University Medical Center Utrecht, Utrecht University, Utrecht, The Netherlands. 11Department of Clinical Pharmacology, Netherlands Cancer Institute, Amsterdam, The Netherlands. 12Department of Pathology, Molecular Diagnostics Laboratory, Netherlands Cancer Institute, Amsterdam, The Netherlands. 13Department of Medical Oncology, Radboud University Medical Center, Nijmegen, The Netherlands. 14Department of Medical Oncology, University Medical Center Groningen, Groningen, The Netherlands. 15Department of Medical Oncology, University Medical Center Utrecht, Utrecht, The Netherlands. 16Department of Medical Oncology, Maastricht University Medical Center, Maastricht, The Netherlands. 17Department of Medical Oncology, Erasmus MC Cancer Institute, Rotterdam, The Netherlands. 18Department of Gastrointestinal Oncology, Netherlands Cancer Institute, Amsterdam, The Netherlands. 19Department of Thoracic Oncology, Netherlands Cancer Institute, Amsterdam, The Netherlands. 20Center for Molecular Medicine, University Medical Center Utrecht, Utrecht, The Netherlands. 21These authors contributed equally: D. L. van der Velden, L. R. Hoes, H. van der Wijngaart, J. M. van Berge Henegouwen. *e-mail: [email protected]

3 OCTOBER 2019 | VOL 574 | NATURE | 127 RESEARCH Letter

The Drug Rediscovery protocol

Drug A Drug B Etc Every new combination of drug/tumour type/tumour pro le forms a new cohort

Cohort 1 Cohort 1 • Tumour type X • Tumour type X • Pro le I •Pro le I Cohort 2 Cohort 2 • Tumour type Y • Tumour type Y • Pro le II •Pro le II Cohort 3 Cohort 3 • Tumour type Z • Tumour type Z • Pro le III •Pro le III Cohort 4 • Tumour type .. • Pro le ..

Each cohort follows a two-stage design for accrual Stage I Stage II

≥5 × clinical Cohort bene t successful ≥1 × clinical + 16 patients bene t <5 × clinical Close cohort 8 patients bene t <1 × clinical Close cohort bene t?

Fig. 1 | Study design. Schematic overview of the study and cohort design. tumour profile and study treatment. In each cohort, patients are enrolled For each study drug, a theoretically unlimited number of cohorts can be in a two-stage design. Clinical benefit is defined as either complete or opened in parallel, depending on the tumour types and tumour profiles partial response, or absence of disease progression for ≥16 weeks, and of submitted patients and the amount of the drug being studied that is must be measured 2 or more times and ≥28 days apart. available. A new cohort is opened for each combination of tumour type, selected variant categories (such as mutational load, microsatellite Table 1 | Baseline characteristics of the first 215 patients who instability and DNA-repair deficiency), the protocol allows for cohorts started study treatment in which tumour types are combined. A Simon-like two-stage design Baseline characteristics 15,16 is used per cohort , in which 8 patients are enrolled in stage I and Median Range up to 24 patients are enrolled in stage II—provided that clinical ben- Approximate age at consent 62 23–87 efit (which we define as complete or partial response, or stable dis- ease beyond 16 weeks, measured 2-or-more times, ≥28 days apart) is Gender n % observed at least once in stage I. A drug warrants further investigation Male 114 53 in a particular cohort if ≥5 out of 24 patients experience a clinical Female 101 47 benefit. If fewer responses are observed, the cohort is closed; results WHO performance status n % will be made public whether or not the cohort is successful. This design WHO 0 60 28 has 85% power to reject a rate of clinical benefit of 10%, if the true WHO 1 116 54 percentage is 30% (α error rate of 7.8%). The analysis of closed cohorts with some activity allows for the opening of new cohorts with refined WHO 2 14 7 criteria for inclusion. Not available 25 12 Between September 2016 and September 2018, over 600 cases were Primary tumour types n % submitted for central review and 294 patients started study treat- Colorectal cancer 49 23 ment. Extended Data Fig. 1 provides details of the review process, and Non-small cell lung cancer 37 17 Extended Data Fig. 2 provides an overview of case submissions. To Prostate cancer 19 9 allow for sufficient follow-up (≥5 months for patients on study treat- ment), here we present the results of the first 215 patients who started Breast cancer 16 7 study treatment. The enrolment of these 215 patients resulted in the Gastro-intestinal stroma cell tumour 9 4 initiation of 76 cohorts (Extended Data Table 2); the baseline charac- Cervical cancer 8 4 teristics of these patients are provided in Table 1. Salivary gland carcinoma 8 4 Overall, clinical benefit was observed in 74 patients (34%) (Extended Urothelial cell carcinoma 8 4 Data Table 3) with a median duration of 9 months (95% confidence Sarcoma 7 3 interval, 8–11 months). Clinical benefit was observed across all types of treatment, comprising immunotherapy (n = 79 patients, clinical Ovarian cancer 7 3 benefit rate of 38%), treatment with small-molecule inhibitors (includ- Other 47 22 ing PARP inhibitors) (n = 81 patients, clinical benefit rate of 36%) and Median Range with monoclonal antibodies (n = 55 patients, clinical benefit rate of Number of prior systemic therapies 3 0–12* 27%). The median progression-free survival and overall survival were WHO, World Health Organization. 3 months (95% confidence interval 2–4 months) and 10 months (95% *All patients were required to have exhausted standard therapies, but some patients refused standard chemotherapy owing to fear of toxicity. In addition, on occasion the treating physician confidence interval 7–13 months), respectively (Fig. 2). To put this in had well-argued reasons to refrain from a given standard therapy (such as the low response rate perspective, a large database of 854 patients who were participating to standard therapies in specifc subgroups of patients).

128 | NATURE | VOL 574 | 3 OCTOBER 2019 Letter RESEARCH

a b 1.0 c 1.0 400 Clinical benet No 0.8 Yes 0.8 300

0.6 0.6 200 0.4 0.4

100 0.2 0.2 Overall survival probability 0 Change from baseline (%) 0 0 Progression−free survival probability 024681012 024681012 Time since start treatment (months) Time since start treatment (months) −100 No. at risk No. at risk 1255075 100 125 150 167 215 116 79 60 43 27 13 215 163 118 85 61 43 26 Patients in order of percentage change Fig. 2 | Response and survival plots. a, Waterfall plot of the best (on the basis of non-target lesions or non-RECIST measurements only) percentage change in the sum of target lesions compared to baseline and patients who went off-study before their response could be evaluated tumour measurements according to ‘Response Evaluation Criteria in are not included in this graph (n = 49 patients). b, Kaplan–Meier curve for Solid Tumours’ (RECIST) 1.1, for all patients with ≥1 response evaluation estimated progression-free survival. c, Kaplan–Meier curve for estimated and with a known change in the sum of target lesions (n = 166 patients). overall survival, with 95% confidence interval (dashed lines). Patients with unequivocal disease progression at the first evaluation in phase I studies and were treated with molecularly targeted agents rare tumour profiles, but also allows further confirmation of clinical indicated a median progression-free survival and overall survival of benefit in a larger cohort of patients20. 2 and 8 months, respectively17. Another immunotherapy cohort—pembrolizumab treatment for One hundred and forty-one patients (66%) did not experience a patients with microsatellite-stable colorectal cancer, with a tumour clinical benefit, either because of progressive disease (n = 117 patients) mutational load of between 140 and 290 (which corresponds to or because they went off-study before they could be classified as having 11–22 mutations per megabase)—showed limited clinical benefit in experienced a clinical benefit or not (n = 24 patients). Reasons for stage I, and was therefore closed (Extended Data Table 6). Together, early withdrawal from the study without obtaining radiologic or clinical these two cohorts illustrate the potential for the DRUP to identify sub- diagnosis of progressive disease included death (n = 9 patients), groups of patients who may benefit from a broader use of approved adverse events (n = 5 patients), patient preference (n = 3 patients) or drugs, and to prevent unnecessary treatment in other subgroups. were unknown (n = 7 patients). Adverse events were consistent with Upon enrolment, the DRUP mandated a fresh baseline tumour those observed in standard of care (Extended Data Table 4). Overall, biopsy for WGS. Baseline WGS results were used for the confirmation ten patients discontinued treatment owing to toxicity. Two suspected of previously identified variants, and for exploratory biomarker analyses. unexpected severe adverse reactions were reported: bacterial peritonitis In the first 215 patients, baseline WGS was successfully performed in a patient with ovarian carcinoma and sinus thrombosis in a patient in 131 patients (61%); the main reason for failure was insufficient with breast cancer. tumour cells in the baseline biopsy (Extended Data Fig. 3). The To date, two cohorts have completed accrual: the first is a tumour- variant on the basis of which patients were included was confirmed type-agnostic cohort of patients with microsatellite-instable (MSI) in 121 patients (92%) with successful baseline WGS. Notably, in tumours treated with nivolumab. In total, 30 patients with 8 types of 112 patients (85%) with baseline WGS, potentially relevant additional tumour were enrolled in this cohort. As of 3 May 2019, one patient (3%) information was revealed (Supplementary Table 1). This information had a complete response. Eleven patients (37%) had a partial response, and seven patients (23%) had stable disease at ≥16 weeks. Four patients Colon (13%) had progressive disease as a best overall response, and seven Prostate Colon patients (23%) went off study before evaluability was reached (that is, Bladder Cervix after fewer than two cycles of nivolumab treatment and/or with insuffi- Colon Colon cient response evaluations to determine clinical benefit). In this cohort, Breast Colon the rate of clinical benefit was 63%. The median progression-free Colon Endometrium survival was not reached after a median follow-up of 16.5 months. A Colon Small intestine summary of the clinical benefits to individual patients is presented in Colon Colon Fig. 3. The results are consistent with previous reports for immunother- Colon 18,19 Colon apy in MSI tumours . Overall, nivolumab was tolerated well, and Colon Breast End of treatment adverse events were largely consistent with those that have previously Colon Treatment ongoing 18,19 Brain Treatment been reported (Extended Data Table 5). One patient developed Prostate Colon No treatment a grade-5 abdominal infection upon intestinal perforation, owing to Colon Colon shrinkage of a peritoneal tumour deposit. One patient experienced Colon Colon grade-5 dyspnoea, possibly attributable to disease progression. Baseline Bladder Endometrium WGS for this cohort was successfully performed in 20 patients (67%) Endometrium (Table 2). Assessment of MSI on the basis of WGS was highly repre- 0816 24 32 40 48 56 64 72 80 88 96 104112 120128 sentative for MSI identification on the basis of immunohistochemis- Time since start treatment (weeks) try and PCR. On average, MSI tumours had 866 mutations (range of 614–1,111 mutations) in the genome. Fig. 3 | Treatment efficacy of nivolumab in completed MSI cohort. Swimmer plot of the time on treatment (in weeks) for each patient The EMA has not yet approved checkpoint inhibitors for the MSI (n = 30 patients). Patients marked with an arrow were still on treatment indication. However, on the basis of these DRUP data, the Dutch at the point of data cut-off (3 May 2019). The white bars represent the Health Care Institute and insurance agencies have now embraced a time period for which nivolumab treatment was interrupted (which was pay-for-performance model for this and future successful cohorts from optional per protocol after 12 months of treatment) for patients, who still the DRUP. This not only creates access to these drugs for patients with experienced clinical benefit.

3 OCTOBER 2019 | VOL 574 | NATURE | 129 RESEARCH Letter

Table 2 | Tumour and biomarker details per patient of the MSI cohort Pre-enrolment Baseline WGS

Somatic or JAK1 and/or JAK2 No. Tumour type Lynch MSI profle MSIseqa ML mutations CD274b COSMIC signaturec CB PFS 1 CRC Somatic MLH1/PMS2 loss − NE NA 2 CRC Somatic MLH1/PMS2 loss,MLH1 methylation 76.9 1,589 3× 6+, 9, 15, 12, 17 PR >105 3 UCC Lynch MSH2 mut, MSH2 loss, MSI 35.2 973 3× 1+, 6+ PR >92 4 CRC Somatic MLH1/PMS2 loss − PD 8 5 Cervix Lynch MSH2 mut, MSH2/MSH6 loss 23.1 776 2× 6+, 1+, 14, 12 PR >88 6 CRC Somatic MLH1/PMS2 loss,MLH1 methylation 66.7 1,301 2× 12+, 6+, 9, 20 PR >83 7 CRC Somatic MLH1/PMS2 loss,MLH1 methylation 17.1 638 2× PR >90 8 CRC Somatic MLH1/PMS2 loss − NE NA 9 CRC Somatic MLH1/PMS2 loss, MSI,MLH1 methyl- − SD 23 ation 10 Breast Somatic MLH1/PMS2 loss,MLH1 not methylated 20.9 287 JAK2 p.Tyr20Asn 8× 6+, 12, 9, 1 PR 23 11 CRC Somatic MLH1/PMS2 loss,MLH1 methylation 65.9 1,036 JAK1 p.Ala639Val 2× 12+, 6+, 9 SD 24 12 CRC Somatic MLH1/PMS2 loss 10.6 346 2× 6+, 12 SD >68 13 CRC Somatic MLH1/PMS2 loss,MLH1 methylation 58.6 912 JAK1 p.Lys860fs 2× 6+, 12, 21, 9 PR >69 14 Breast Somatic MLH1/PMS2 loss,MLH1 not methylated − PR >72 15 CRC Somatic MLH1/PMS2 loss, MSI, BRAFV600E, 0.5 73 2× 1+, 8, 17, 18, 11 NE NA MLH1 methylation 16 CRC Lynch MSH2 mut − SD 36 17 EC Somatic MLH1/PMS2 loss,MLH1 methylation 37.7 798 JAK2 p.Lys1055Glu, 4× 12+, 26+, 6 NE NA JAK2 c.3291 + 8delT 18 EC Somatic MLH1/PMS2 loss,MLH1 methylation − SD >58 19 CRC Somatic MSI, MLH1 not methylated − PR >49 20 CRC Lynch PMS2 mut, MLH1/PMS2 loss, MSI, 57.7 1,048 2× 6+, 15, 1, 9 SD 48 MLH1 not methylated 21 CRC Somatic MLH1/PMS2 loss,MLH1 methylation 85.8 1,558 JAK1 p.Lys860fs 2× 6+, 12+, 15, 9 SD 39 22 GBM Lynch PMS2 mut, PMS2 loss,MSS 42.4 6,137 JAK1 p.Gln437His, 2× 14+, 1+, 15 PD 13 JAK2 p.Ala586Thr 23 Duode- Somatic MSH6 loss, MSI 35.6 1,425 JAK2 p.Arg158Gln 2× 6+, 1, 12, 9 PR >63 num 24 CRC Somatic MLH1/PMS2 loss,BRAFV600E 46.0 1,048 JAK2 c.3291 + 8delT 2× 6, 14, 12 PR >120 25 UCC Lynch MSH6 mut, MSH6 loss 7.2 637 2× 1+, 6+ PD 5 26 Prostate Lynch MSH6 mut − SD >98 27 Prostate Somatic MSI 12.0 543 3× 1+, 6+, 12 PD 15 28 CRC Somatic MLH1/PMS2 loss 60.4 820 2× PD 6 29 CRC Somatic MSH2/MSH6 loss − SD >51 30 EC Somatic MLH1/PMS2 loss 39.3 387 JAK1 2× 6+, 12+ NE NA p.Pro430fs JAK1 c.2115 + 1G>A Histologic tumour type, biomarker characteristics and classifcation of clinical beneft per patient in the MSI cohort. CB, clinical beneft; PFS, progression-free survival (given in weeks, at data cut-of; > denotes beneft ongoing at cut-of); CRC, colorectal cancer; EC, endometrial cancer; GBM, glioblastoma multiforme; Lynch, Lynch syndrome; ML, mutational load; MSIseq, microsatellite instability score; MSS, microsatellite stable; mut, mutation; NA, not applicable; NE, not evaluable; no., case number; PD, progressive disease; PR, partial response; SD, stable disease; somatic, diferent from germline DNA; UCC, urothelial cell carcinoma. aThe use of − denotes that no baseline WGS data are available. bCopy number gain of CD274. cUse of + denotes >20% prevalence of ‘Catalogue of Somatic Mutations in Cancer’ (COSMIC) mutational signature. Signatures are only mentioned in case of >10% prevalence (https://cancer.sanger. ac.uk/cosmic/signatures). included high mutational load, variants associated with therapy much larger cohort21: we combined three large repositories of knowl- response or resistance, and variants that were potentially actionable edge—‘Clinical Interpretation of Variants in Cancer’ (CIViC)22, with experimental or off-label agents (other than the current treatment ‘Precision Oncology Knowledge Base’ (OncoKB)23 and ‘Cancer that the patient was receiving in the DRUP). The latter may lead to Genome Interpreter’ (CGI)24. We also followed the ‘European Society re-enrolment upon failure of the first treatment administered to the for Medical Oncology (ESMO) Scale for Clinical Actionability of patient in the DRUP. molecular Targets’ (ESCAT)25 wherever possible. Some limitations of the DRUP should be taken into account. One The efficacy endpoints bear additional, inherent limitations26: the important caveat is the absence of comparator groups, owing to the objective response rate can detect tumour growth (or reductions in non-randomized trial design. With the increasing availability of large, size) but cannot detect reductions in the rate of growth. By contrast, clinically and molecularly annotated databases, this may be addressed survival statistics cannot differentiate between a true effect of the treat- by methodologies such as trials within cohorts. Another concern is ment and a naturally slow growth rate. The progression-free-survival that, in a heterogeneous study population such as that of the DRUP, ratio (in which each patient serves as their own control) might be able the correct interpretation of molecular aberrations is challenging. to overcome some of these challenges, but has its own limitations (as Fortunately, we were able to draw upon previous experiences with a pre-study progression-free-survival data are collected retrospectively).

130 | NATURE | VOL 574 | 3 OCTOBER 2019 Letter RESEARCH

Taken together, the DRUP shows the feasibility of performing pre- 5. Massard, C. et al. High-throughput genomics and clinical outcome in hard-to-treat advanced cancers: results of the MOSCATO 01 trial. Cancer Discov. cision medicine in multiple, parallel cohorts driven by tumour type 7, 586–595 (2017). and tumour profile. It provides a framework through which patients 6. Meric-Bernstam, F. et al. Feasibility of large-scale genomic testing to facilitate with all types of tumours are able to acquire access to existing targeted enrollment onto genomically matched clinical trials. J. Clin. Oncol. 33, 2753–2762 (2015). therapies and immunotherapies, and in which treatment outcomes are 7. Stockley, T. L. et al. Molecular profling of advanced solid tumors and patient monitored and publicly reported. This improves on current practice, outcomes with genotype-matched clinical trials: the Princess Margaret in which individual physicians obtain anticancer drugs ‘off label’ for IMPACT/COMPACT trial. Genome Med. 8, 109 (2016). 8. Hyman, D. M., Taylor, B. S. & Baselga, J. Implementing genome-driven oncology. their patients without subsequent public reporting of clinical outcomes. Cell 168, 584–599 (2017). The public availability of these data is especially relevant given recent 9. Le Tourneau, C. et al. Molecularly targeted therapy based on tumour molecular concerns that the increasingly widespread use of genetic profiling could profling versus conventional therapy for advanced cancer (SHIVA): a 27 multicentre, open-label, proof-of-concept, randomised, controlled phase 2 trial. escalate the demand for off-label treatment . Furthermore, the impor- Lancet Oncol. 16, 1324–1334 (2015). tance of publicly reporting negative results cannot be underestimated, 10. Prasad, V. Perspective: the precision-oncology illusion. Nature 537, S63 (2016). as it prevents patient exposure to ineffective agents with all their accom- 11. Tannock, I. F. & Hickman, J. A. Limits to personalized cancer medicine. N. Engl. J. Med. 375, 1289–1294 (2016). panying toxicities and financial costs. Another important advantage of 12. Sleijfer, S., Bogaerts, J. & Siu, L. L. Designing transformative clinical trials in the the DRUP is that it enables the rapid incorporation of new drugs and cancer genome era. J. Clin. Oncol. 31, 1834–1841 (2013). scientific insights into clinical practice: matching rules can be adapted 13. Ellis, L. M. et al. American Society of Clinical Oncology perspective: raising the quickly, and cohorts based on new biomarkers may be opened almost bar for clinical trials by defning clinically meaningful outcomes. J. Clin. Oncol. 32, 1277–1280 (2014). instantaneously. In addition, our use of WGS identified many poten- 14. Bins, S. et al. Implementation of a multicenter biobanking collaboration for tially actionable variants that were not identified by smaller gene panels, next-generation sequencing-based biomarker discovery based on fresh frozen immunohistochemistry and/or in-situ hybridization. Eventually, WGS pretreatment tumor tissue biopsies. Oncologist 22, 33–40 (2016). 15. Simon, R. Optimal two-stage designs for phase II clinical trials. Control. Clin. may thus identify more, or more-appropriate, treatment options for Trials 10, 1–10 (1989). each patient. An integral part of our approach is a tiered review process 16. Jung, S. H., Lee, T., Kim, K. & George, S. L. Admissible two-stage designs for that includes reviews of the literature and by multidisciplinary boards phase II cancer clinical trials. Stat. Med. 23, 561–569 (2004). 17. Moreno Garcia, V. et al. Dose–response relationship in phase I clinical trials: of experts, before patients are enrolled. This prevents the prescription a European Drug Development Network (EDDN) collaboration study. Clin. of anticancer drugs when negative clinical data are available, or when Cancer Res. 20, 5663–5671 (2014). the actionability of the variant is unknown or unlikely. Finally, our 18. Le, D. T. et al. PD-1 blockade in tumors with mismatch-repair defciency. N. Engl. J. Med. 372, 2509–2520 (2015). 10 study design and informed consent both enable the sharing of data 19. Overman, M. J. et al. Nivolumab in patients with metastatic DNA mismatch internationally. By combining cohorts from similar international repair-defcient or microsatellite instability-high colorectal cancer (CheckMate studies, we will improve our knowledge of rare subsets of cancer, and 142): an open-label, multicentre, phase 2 study. Lancet Oncol. 18, 1182–1191 (2017). the outcomes of their treatment. Most importantly, our approach shows 20. van Waalwijk van Doorn-Khosrovani, S. B. et al. Personalised reimbursement: that existing anticancer drugs may have value beyond their approved a risk-sharing model for biomarker-driven treatment of rare subgroups of indications, which potentially expands the range of patients who may cancer patients. Ann. Oncol. 30, 663–665 (2019). benefit from their use. 21. Priestley, P. et al. Pan-cancer whole genome analyses of metastatic solid tumors. Preprint at https://www.biorxiv.org/content/10.1101/415133v4 (2018). Online content 22. Grifth, M. et al. CIViC is a community knowledgebase for expert crowdsourcing Any methods, additional references, Nature Research reporting summaries, the clinical interpretation of variants in cancer. Nat. Genet. 49, 170–174 (2017). source data, extended data, supplementary information, acknowledgements, peer 23. Chakravarty, D. et al. OncoKB: a precision oncology knowledge base. JCO Precis. Oncol. https://doi.org/10.1200/PO.17.00011 (2017). review information; details of author contributions and competing interests; and 24. Tamborero, D. et al. Cancer Genome Interpreter annotates the biological and statements of data and code availability are available at https://doi.org/10.1038/ clinical relevance of tumor alterations. Genome Med. 10, 25 (2018). s41586-019-1600-x. 25. Mateo, J. et al. A framework to rank genomic alterations as targets for cancer precision medicine: the ESMO scale for clinical actionability of molecular Received: 21 July 2018; Accepted: 14 August 2019; targets (ESCAT). Ann. Oncol. 29, 1895–1902 (2018). Published online 30 September 2019. 26. Sleijfer, S. & Wagner, A. J. The challenge of choosing appropriate end points in single-arm phase II studies of rare diseases. J. Clin. Oncol. 30, 896–898 (2012). 27. Ray, T. CMS-proposed coverage of NGS cancer tests could lead to of-label scripts, 1. Hainsworth, J. D. et al. Targeted therapy for advanced solid tumors on the basis oncologists worry, https://www.genomeweb.com/molecular-diagnostics/ of molecular profles: results from MyPathway, an open-label, phase IIa multiple cms-proposed-coverage-ngs-cancer-tests-could-lead-label-scripts- basket study. J. Clin. Oncol. 36, 536–542 (2018). oncologists#.WsJTUOkUk5k (2018). 2. Hyman, D. M. et al. HER kinase inhibition in patients with HER2- and HER3-mutant cancers. Nature 554, 189–194 (2018). Publisher’s note Springer Nature remains neutral with regard to jurisdictional 3. Hyman, D. M. et al. Vemurafenib in multiple nonmelanoma cancers with BRAF claims in published maps and institutional affiliations. V600 mutations. N. Engl. J. Med. 373, 726–736 (2015). 4. Colwell, J. NCI-MATCH trial draws strong interest. Cancer Discov. 6, 334 (2016). © The Author(s), under exclusive licence to Springer Nature Limited 2019

3 OCTOBER 2019 | VOL 574 | NATURE | 131 RESEARCH Letter

Methods local principal investigator and could be used to re-assess eligibility if a patient The DRUP is a national, prospective, non-randomized multi-drug and multi- progressed on initial study treatment. As the baseline biopsy was obtained after tumour study, designed and conducted on behalf of the Center for Personalized enrolment, the baseline WGS results did not affect the initiation of the study treat- Cancer Treatment (CPCT) (clinicaltrials.gov: NCT02925234). The trial was ment. For each patient, a unique patient identification number was generated by approved by the Medical Ethical Committee of the Netherlands Cancer Institute the electronic case-report file system. This number was used by the study team in Amsterdam, and was conducted in accordance with good clinical practice and external researchers for data and sample collection and analysis, and could be guidelines and the Declaration of Helsinki’s ethical principles for medical research. tracked back to the individual patient only by the local sub-investigator. Written informed consent was obtained from all study subjects. Patients were In addition to a summary of somatic variants across cancer-related genes, the accrued at multiple hospitals throughout the Netherlands, and followed for sequencing report contained information regarding complex molecular features 30 days after end of study treatment, or death, respectively, for toxicity and survival of the tumour, including the mutational load and microsatellite instability. The analyses. Figure 1 provides a schematic overview of the study design. tumour mutational load represents the total number of somatic missense variants Patient population. Patients who were eligible for the study had an advanced or across the protein-coding region of the tumour genome. The microsatellite (in) metastatic solid tumour, multiple myeloma or B cell non-Hodgkin lymphoma, and stability score represents the number of somatic insertions and deletions in short had exhausted standard-treatment options. A tumour genetic or protein-expression repeat sections across the tumour genome per megabase. This metric can be con- test (CPCT or regular diagnostics) had to have revealed a potentially actionable sidered as a good marker for instability in microsatellite repeat regions34, and has variant, for which FDA- and/or EMA-approved targeted therapy was available— extensively been validated against the standard MSI–PCR assay used in routine but not for the tumour type in question. In addition, patients were required to be practice (data not shown). ≥18 years of age, with acceptable organ function and performance status (Eastern Cohort design. The study comprised multiple parallel cohorts, each defined by one Cooperative Oncology Group (ECOG) score ≤ 2), and to have objectively evalu- histologic tumour type, one molecular tumour variant and one study treatment. able disease of which a fresh baseline tumour biopsy could safely be obtained. For For the purposes of cohort definition, the variant category was defined at the level every study drug, further drug-specific selection criteria applied. of the gene or receptor that contains the mutation, translocation, amplification, Matching rules. Upon case submission, the study team attempted to match each overexpression or homozygous deletion; for example, the EGFR mutant that was patient to the appropriate study treatment (Extended Data Fig. 1), according to defined as the variant for purposes of cohort definition included all detected EGFR pre-defined matching rules (Extended Data Table 1). For matching purposes, a mutations. potentially actionable molecular variant was defined following a previous publi- The rate of clinical benefit for each treatment was analysed per cohort. Clinical cation28, as either one of the following options: (1) the variant is the target of an benefit was defined as objective response, or absence of disease progression for approved drug for any cancer indication, or is known to predict sensitivity to an ≥16 weeks (counted from treatment initiation until end of treatment or measure- approved drug for any cancer indication; (2) the variant is in the same molecular ment of progressive disease, whichever came first), measured 2 or more times and pathway, but located upstream of the target of an approved drug for any cancer ≥28 days apart (defined as a confirmed response). Per cohort, a rate of clinical indication, and has been reported as an oncogenic or pathogenic mutation; (3) benefit of <10% was considered to be of no clinical interest. A rate of clinical mutations that result in unique susceptibility to a specific molecular intervention benefit of ≥30% was considered relevant and of sufficient interest to warrant (such as BRCA1 and BRCA2 mutations and PARP inhibitors, or MSI and PD-1 further investigation. A Simon-like two-stage ‘admissible’ design15 was used for inhibitors); and (4) other variants that have appropriate justification for selection each cohort: in stage I, eight patients were enrolled. If no clinical benefit was on the basis of published scientific evidence regarding their susceptibility to specific observed in these first eight patients, the cohort was closed. Otherwise, 16 addi- targeted therapies. tional patients were enrolled. Cohorts with clinical benefit in ≤4 out of 24 patients If multiple variant–drug matches could be made for one patient, the drug with were considered ineffective, whereas cohorts with clinical benefit in ≥5 patients the highest level of evidence was selected unless there was a rationale (such as were considered effective. This monitoring rule has 85% power and an α error rate drug intolerance) that justified selecting an agent with a lower level of evidence. of 7.8%. These operating characteristics were selected to represent a reasonable Levels of evidence were adapted from a previous publication28 and were defined compromise between high power, low false-positive rates and a desire for small as: the drug met a clinical endpoint (objective response, PFS or overall survival) in sample sizes, especially in stage I. a prospective trial, in patients with the same variant and tumour type, and has not Study endpoints. The main study endpoints included (i) the percentage of yet received regulatory approval for use in the tumour type of the patient (level 1); submitted patients that started study treatment, and the main reasons for non- clinical studies have demonstrated an association between presence of the vari- enrolment; (ii) the efficacy, including best overall response, response duration ant and drug activity against the tumour type of the patient (level 2); the drug is and rate of clinical benefit; and (iii) toxicity, including all treatment-related commercially available in the US and/or European Union (EU) for use in another adverse events of grade 3 or higher. The sequencing-success rate of pre-treatment tumour type that contains the same variant (level 3); and preclinical evidence of biopsies, and comparison of ‘historic’ and baseline tumour profiles formed an anti-tumour activity and target inhibition in model systems of the tumour type exploratory endpoint (endpoint iv). All endpoints were prospectively decided. of the patient (level 4). For endpoint (i), all cases that were submitted for review were considered Study treatment and assessments. If a slot for a matching study treatment was evaluable, and the reasons for non-enrolment were classified by two reviewers available (to which the patient consented) the patient could be enrolled, if all independently. For endpoint (ii), the best overall response was considered drug-specific selection criteria were met. Once a fresh baseline tumour biopsy for evaluable in patients who received at least one cycle of oral study medication biomarker analyses was obtained, the study treatment could be initiated. Treatment or two cycles of intravenous study medication, and for whom response was and follow-up were conducted according to the approved indication. All treat- radiologically or clinically evaluable (at the discretion of the treating physician). ment-related adverse events (following the Common Terminology Criteria for Clinical-benefit calculations included all enrolled patients, regardless of evalua- Adverse Events (CTCAE) version 4.03) of grade 3 or higher were documented. The bility of the best overall response. All patients without clinical benefit had been response to the treatment was evaluated every 2 months (up to every 3 months for followed for at least 16 weeks at the time of the analysis, so no censoring was patients who remained in the study for ≥6 months), and classified by local inves- necessary. All patients who received study treatment were considered evaluable tigators according to internationally accepted criteria for each tumour type29–33. for endpoint (iii), and all patients who were formally enrolled were considered The study treatment could continue until progressive disease (patients who were evaluable for endpoint (iv). receiving immune-system-stimulating agents were permitted to continue treatment Statistics. All statistical analyses were performed using R version 3.5.0 (http:// in case of pseudo-progression), unacceptable treatment-related toxicity, death, www.R-project.org/). This trial was not randomized and investigators were not pregnancy, consent withdrawal or withdrawal from the study at the discretion of blinded to treatment allocation or outcome assessments. Patient characteristics, the investigator. adverse events and tumour responses were summarized using descriptive statistics. Baseline tumour biopsies. A fresh, frozen tumour biopsy specimen was manda- In addition, a waterfall plot was used to illustrate maximum tumour shrinkage tory before treatment initiation (baseline biopsy had to be obtained ≤2 months compared to baseline sizes. Percentages were calculated with 95% confidence inter- before enrolment, and without any anticancer therapy within those ≤2 months), vals using the Clopper–Pearson method. Kaplan–Meier methods were used to and was optional during and after study treatment. All biopsies were sent to the estimate overall survival (calculated from the first day of treatment administration central sequencing institute of the CPCT (Hartwig Medical Foundation (HMF), to the date of death from any cause, censoring patients who were alive at the final Amsterdam, The Netherlands), together with a 10-ml blood sample to determine follow-up), PFS (from the start of treatment to progression or death from any cause, the background variation of the germline DNA of the patient. If the tumour-cell whichever came first, and censoring patients who were alive without progression percentage was ≥ 30% and the DNA yield was ≥ 300 ng, WGS and biomarker at final follow-up and time on treatment (censoring patients who had not finished analyses were performed. treatment at the time of analysis). The WGS data and treatment details were stored in a national centralized Reporting summary. Further information on research design is available in database (at the HMF). In addition, a sequencing report was returned to the the Nature Research Reporting Summary linked to this paper. Letter RESEARCH

Data availability Dohme, Novartis, Pfizer and Roche for their in-kind and financial support; the HMF for their in-kind support by performing sequencing and biomarker All data described in this study are freely available for academic use, and can be analyses on baseline biopsies; the Center for Personalized Cancer Treatment obtained through the Netherlands Cancer Institute and the HMF through stand- Multidisciplinary Expert Board for supporting the central case-review process; ardized procedures and request forms. These can be found at https://www.hartwig- the Independent Data Monitoring Committee for their advice on cohort medicalfoundation.nl/en. In brief, a data request can be initiated by filling out the decisions and the monitoring of preliminary safety data; the Netherlands standard form, which requires the applicant to provide the intended use of Cancer Institute’s Biobank Facility, Scientific Department and Pharmacy for their facilitating services; A. P. Hamberg, L. V. Beerepoot and J. M. Meerum- the requested data. Next, an independent data-access board will evaluate whether Terwogt for their contributions to trial recruitment; and all participating the intended use of the data is compatible with the consent given by the patients, hospitals for supporting and facilitating the conduct of the DRUP trial. and whether there would be any applicable legal or ethical constraints. Upon formal approval by the data-access board, a standard license agreement (which does Author contributions All authors contributed extensively to the work presented not have any restrictions regarding intellectual property resulting from the data in this paper. H.M.W.V., H.G. and E.E.V. initiated and led the trial as principal analysis) needs to be signed by an official representation of the organization before investigators. R.L.S. and S.S. contributed to its design and initiation. D.L.v.d.V., L.R.H., H.v.d.W. and J.M.v.B.H. coordinated the trial. D.L.v.d.V. analysed the data access to the data is granted. All WGS data, including raw data files, will be made and wrote the manuscript. L.R.H., H.v.d.W. and J.M.v.B.H. assisted with the available through a dedicated download portal with two-factor authentication, via data analysis, and substantively revised the manuscript. C.M.L.v.H., D.J.A.d.G., the HMF. Clinical, outcome and safety data can be obtained at a per-patient level L.A.D., A.H., M.J.A.d.J., M.C., E.F.S., A.J.d.L., N.M., M.L. and E.K. contributed to for all patients reported in the study, by emailing the Institutional Review Board of patient enrolment and clinical data collection. P.R., W.W.J.d.L. and P.M.N. helped the Netherlands Cancer Institute ([email protected]). Both the WGS and clinical datasets with the interpretation of tumour profiles by performing variant calling and pathogenicity assessments. A.D.R.H. and B.N. coordinated the distribution of have unique identifiers for individual patients, which allows a combined analysis. study drugs. P.R. and E.C. performed sequencing of tumour biopsies, generated For any other data, correspondence should be addressed to [email protected] and sequencing reports and conducted biomarker analyses. E.v.W. contributed to requests for materials to [email protected]. data extractions, and statistical design and analyses. All authors discussed the results and implications, and commented on the manuscript. 28. Meric-Bernstam, F. et al. A decision support framework for genomically informed investigational cancer therapy. J. Natl Cancer Inst. 107, djv098 (2015). Competing interests E.E.V. is legally responsible for all contracts with 29. Cheson, B. D. et al. Recommendations for initial evaluation, staging, and pharmaceutical companies at the Netherlands Cancer Institute. H.M.W.V. response assessment of Hodgkin and non-Hodgkin lymphoma: the Lugano and H.G. have, through the DRUP and other studies, received support from classifcation. J. Clin. Oncol. 32, 3059–3067 (2014). pharmaceutical companies that are participating in the DRUP. R.L.S serves 30. Rajkumar, S. V. et al. International Myeloma Working Group updated criteria for as principal investigator for the TAPUR trial, in support of which the American the diagnosis of multiple myeloma. Lancet Oncol. 15, e538–e548 (2014). Society of Clinical Oncology receives financial support from Astra-Zeneca, Bayer, 31. Rustin, G. J. et al. Defnitions for response and progression in ovarian cancer Bristol-Myers Squibb, Genentech, Lilly, Merck and Pfizer. C.M.L.v.H. has received clinical trials incorporating RECIST 1.1 and CA 125 agreed by the Gynecological funding from AstraZeneca, BMS, Ipsen, Merck, MSD and Novartis. A.J.d.L. has Cancer Intergroup (GCIG). Int. J. Gynecol. Cancer 21, 419–423 (2011). served as an advisor for AstraZeneca, BMS, Boehringer, Pfizer, Lilly and MSD, 32. Therasse, P. et al. New guidelines to evaluate the response to treatment in solid and has received research grants from AstraZeneca, BMS and MSD. E.K. has tumors. European Organization for Research and Treatment of Cancer, National served as advisor for Amgen, BMS, Eisai, Genzyme-Sanofi, MSD, Novartis and Cancer Institute of the United States, National Cancer Institute of Canada. Roche. The other authors declare no competing interests. J. Natl. Cancer Inst. 92, 205–216 (2000). 33. Wen, P. Y. et al. Updated response assessment criteria for high-grade gliomas: Additional information response assessment in neuro-oncology working group. J. Clin. Oncol. 28, Supplementary information is available for this paper at https://doi.org/ 1963–1972 (2010). 10.1038/s41586-019-1600-x. 34. Huang, M. N. et al. MSIseq: software for assessing microsatellite instability from Correspondence and requests for materials should be addressed to E.E.V. catalogs of somatic mutations. Sci. Rep. 5, 13321 (2015). Peer review information Nature thanks David Hyman, Christophe Le Tourneau and the other, anonymous, reviewer(s) for their contribution to the peer review Acknowledgements We thank the Barcode for Life Foundation and the of this work. Dutch Cancer Society for their financial support; Amgen, Astra Zeneca, Reprints and permissions information is available at http://www.nature.com/ Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Eisai, Merck Sharp and reprints. RESEARCH Letter

Extended Data Fig. 1 | Study flowchart. Patients may be identified via trials or access programmes is preferred, if available. Input for stages (iii) regular diagnostics or by WGS performed within the context of a CPCT and (iv) of the review is derived from PubMed (https://www.ncbi.nlm. sequencing study. Adult patients with advanced cancers and without nih.gov/pubmed/), ClinicalTrials.gov and weekly automatic updates on standard-treatment options (but with a known potentially actionable publications that mention any drug in the study in their titles and/or variant in their molecular tumour profile) can be submitted for review. abstracts. If the general selection criteria are met and the appropriate study The central review is done by two or more reviewers independently, treatment is available, the patient can be informed, screened and enrolled supported by the CPCT Molecular Tumour Board, and includes review (if all drug-specific selection criteria are also met). Once a fresh baseline of (i) the medical history of the patient, (ii) tumour-profiling test results, tumour biopsy is obtained, study treatment can be initiated. Patients are (iii) available literature and (iv) potential drug-access alternatives. Patients treated and followed according to the labelled indication for each drug. who are eligible for standard treatments are referred back to their treating Response is evaluated once every two months. Patients can continue study physician. Genomic variants of unknown significance (VUS) that are not treatment as long as clinical benefit is observed. Patients who discontinue likely to be actionable are not considered acceptable drug targets. Negative study treatment can be resubmitted if their molecular tumour profile (as trials are not repeated, nor are positive or ongoing phase II or III trials, revealed by the baseline biopsy) contains additional actionable variants. unless drug access is not (or is not yet) facilitated. Drug access via other CR, complete response, PR, partial response. Letter RESEARCH

Extended Data Fig. 2 | Case submissions and reasons for non-accrual. these 642 case submissions, and as an absolute number per category. Cases Overview of the first 642 case submissions (submitted between 1 that were erroneously submitted (owing to incomplete understanding September 2016 and 1 September 2018), as well as the reasons for not of the study protocol and/or retraction of the submission by the treating being enrolled in the study. Values are displayed as a percentage relative to physician) are not included in this overview (n = 58 cases). RESEARCH Letter

Extended Data Fig. 3 | Baseline biopsies for biomarker analyses. relative to the 131 successfully sequenced biopsies. CPCT-02, the national Overview and success rate of WGS on pre-treatment tumour biopsies. WGS programme of the CPCT; HML, high mutational load (defined as The bottom panel displays the number of patients for whom WGS ≥140 somatic missense variants across the tumour genome overall); WGS- succeeded, and indicates whether the initial variant (on the basis of which MSI, microsatellite instability suspected on the basis of WGS results; ampl, the patient started the study treatment) was also present in the fresh amplification; mut, mutation; wt, wild type. baseline biopsy. Values are displayed as absolute numbers and percentages, Letter RESEARCH

Extended Data Table 1 | Available drugs and matching rules

A list of the participating pharmaceutical companies, and the drugs that were available for this study. To be eligible for a given treatment, a patient needed to meet the description for the treatment that is marked with a + in this table (exclusion criteria are marked with −). RESEARCH Letter

Extended Data Table 2 | Opened cohorts

Overview of cohorts that have been opened for the frst 215 patients who started study treatment. For each cohort cell in the table, the tumour type is indicated in top line, the tumour profle is indicat- ed in the middle line and the number of evaluable or enrolled patients is indicated in bottom line. All patients were required to be refractory or intolerant to standard therapies. ACUP, adenocarcinoma of unknown primary; amp, amplifcation; cholangio, cholangiocarcinoma; esoph, oesophageal cancer; esthesioneurobl, aesthesioneuroblastoma; fus, fusion; GIST, gastrointestinal stromal tumour; HML, high mutational load (defned here as the sum of all somatic missense variants across the protein-coding region of the tumour genome); HNSCC, head and neck squamous cell carcinoma; HRR, homologous recombination repair; hydradeno., hidradenocarcinoma; IMT, infammatory myofbroblastic tumour; NEC, neuro-endocrine carcinoma; NSCLC, non-small-cell lung cancer; pre-specifed indicates breast, gastric, ovarian, pancreatic, prostate and small-cell lung cancer pre-specifed for olaparib. ∗Three cohorts have been closed for enrolment. The dabrafenib + trametinib cohort for NSCLC with BRAF mutation has been closed because this treatment is now registered and reimbursed for this indication. The nivolumab–MSI cohort has completed stage I as well as stage II and is thus closed for further inclusion: the overall clinical beneft rate was 67%. The pembrolizumab cohort of patients with colorectal cancer with a high mutational load (140 to 290) has completed stage I. As no patient experienced clinical beneft, the cohort has been closed and will not be graduated to stage II. All other cohorts with >8 evaluable patients have been graduated to stage II, as clinical beneft was observed once or more in stage I. Letter RESEARCH

Extended Data Table 3 | Rates of response and clinical benefit

Rates of clinical beneft and response in the frst 215 patients who started study treatment. Clinical beneft is defned as a complete or partial response or absence of disease progression at ≥16 weeks; it must be measured 2 or more times ≥28 days apart (defned as a confrmed response). Given that 29 patients had ongoing clinical beneft at the time of analyses, the actual mean duration is expected to exceed the current mean duration. The reasons for early withdrawal from study (other than progressive disease) included death (n = 9 patients), adverse events (n = 5 patients), patient preference (n = 3 patients) or were unknown (n = 7 patients). RESEARCH Letter

Extended Data Table 4 | Adverse events in the first 215 patients

All reported adverse events of grade 3 or higher that were (or could possibly be) attributed to the study treatment. For each adverse event, the number of patients is displayed in whom it was reported at grade 3, 4 or 5 as the highest grade (according to CTCAE version 4.03). Letter RESEARCH

Extended Data Table 5 | Adverse events in the MSI cohort

All reported adverse events of grade 3 or higher that were (or could possibly be) attributed to treatment with nivolumab. Grades are given according to CTCAE version 4.03. a, For each adverse event, the number of patients is displayed in whom it was reported at grade 3, 4 or 5 as the highest grade. Upward-pointing arrows indicate an increase. b, The number of patients who had any grade 3 or higher, grade 4 or higher, or grade 5 or higher adverse event as their highest-grade adverse event is displayed. The denominator of the percentages is the total number of patients who started study treatment (n = 30 patients). Given that every patient could be counted only once per column in b, the numbers in b are not a summation of the numbers in a. RESEARCH Letter

Extended Data Table 6 | Treatment with pembrolizumab in the cohort of patients with MSS colorectal cancer with a high mutational load (between 140 and 290)

Overview of baseline characteristics and clinical beneft in ten patients who were treated within the cohort ‘pembrolizumab for patients with colorectal cancer with a high mutational load (between 140 and 290)’. Clinical beneft is defned as complete or partial response or absence of disease progression at ≥16 weeks. The reasons for early withdrawal from the study (other than progressive disease) included death (n = 1 patient) or adverse event (n = 1 patient), and were classifed as having no clinical beneft. SPECIAL ARTICLE

JSCOdESMOdASCOdJSMOdTOS: international expert consensus recommendations for tumour-agnostic treatments in patients with solid tumours with microsatellite instability or NTRK fusions

T. Yoshino1*, G. Pentheroudakis2, S. Mishima1, M. J. Overman3, K.-H. Yeh4, E. Baba5, Y. Naito6, F. Calvo7, A. Saxena8, L.-T. Chen9, M. Takeda10, A. Cervantes11, H. Taniguchi1, K. Yoshida12, Y. Kodera13, Y. Kitagawa14, J. Tabernero15, H. Burris16 & J.-Y. Douillard17

1Department of Gastroenterology and Gastrointestinal Oncology, National Cancer Center Hospital East, Kashiwa, Japan; 2Department of Medical Oncology, University of Ioannina, Ioannina, Greece; 3Department of Gastrointestinal Medical Oncology, Division of Cancer Medicine, The University of Texas MD Anderson Cancer Center, Houston, USA; 4Department of Medical Oncology, National Taiwan University Cancer Center and Cancer Research Center, National Taiwan University College of Medicine, Taipei, Taiwan; 5Department of Oncology and Social Medicine, Graduate School of Medical Sciences, Kyushu University, Fukuoka; 6Department of Experimental Therapeutics/Breast and Medical Oncology, National Cancer Center Hospital East, Kashiwa, Japan; 7Department of Clinical Pharmacology, University of Paris and Institute Gustave Roussy, Villejuif, France; 8Department of Medicine, Division of Hematology & Medical Oncology, Thoracic Oncology Service, Weill Cornell Medicine, New York, USA; 9National Institute of Cancer Research, National Health Research Institutes, Tainan, Taiwan; 10Department of Medical Oncology, Kindai University, Osaka, Japan; 11CIBERONC, Department of Medical Oncology, Institute of Health Research, INCLIVIA, University of Valencia, Valencia, Spain; 12Department of Surgical Oncology, Gifu University, Graduate School of Medicine, Gifu; 13Department of Gastrointestinal Surgery, Nagoya University, Nagoya; 14Department of Surgery, Keio University School of Medicine, Tokyo, Japan; 15Medical Oncology Department, Vall d’ Hebron University Hospital, Vall d’Hebron Institute of Oncology (V.H.I.O.), Barcelona, Spain; 16The Sarah Cannon Research Institute, Nashville, USA; 17ESMO, Lugano, Switzerland

Available online 6 April 2020

A Japan Society of Clinical Oncology (JSCO)-hosted expert meeting was held in Japan on 27 October 2019, which comprised experts from the JSCO, the Japanese Society of Medical Oncology (JSMO), the European Society for Medical Oncology (ESMO), the American Society of Clinical Oncology (ASCO), and the Taiwan Oncology Society (TOS). The purpose of the meeting was to focus on what we have learnt from both microsatellite instability (MSI)/ deficient mismatch repair (dMMR) biomarkers in predicting the efficacy of anti-programmed death-1 (PD-1)/ programmed death ligand-1 (PD-L1) immunotherapy, and the neurotrophic tyrosine receptor kinase (NTRK) gene fusions in predicting the efficacy of inhibitors of the tropomyosin receptor kinase (TRK) proteins across a range of solid tumour types. The recent regulatory approvals of the anti-PD-1 antibody pembrolizumab and the TRK inhibitors larotrectinib and entrectinib, based on specific tumour biomarkers rather than specific tumour type, have heralded a paradigm shift in cancer treatment approaches. The purpose of the meeting was to develop international expert consensus recommendations on the use of such tumour-agnostic treatments in patients with solid tumours. The aim was to generate a reference document for clinical practice, for pharmaceutical companies in the design of clinical trials, for ethics committees in the approval of clinical trial protocols and for regulatory authorities in relation to drug approvals, with a particular emphasis on diagnostic testing and patient selection. Key words: microsatellite instability, mismatch repair, NTRK, tumour-agnostic, recommendations

INTRODUCTION specific biomarkers rather than on tumour site, and thus The last 2 years have seen a paradigm shift in the regulatory establishing the precedent of tumour-agnostic therapies. fi The first of these agents, pembrolizumab, is a well-known approval of cancer treatments with the approval of the rst 1e3 two agents, pembrolizumab and larotrectinib, for the anti-programmed death-1 (PD-1) T-cell receptor antibody. treatment of solid tumours based on the presence of In 2015, a small investigator-initiated study (KEYNOTE-016) showed colorectal cancer (CRC) patients with deficient mismatch repair (dMMR) treated with pembrolizumab to achieve immune-related objective response (ORR) and *Correspondence to: Prof. Takayuki Yoshino, Department of Gastroenterology and Gastrointestinal Oncology, National Cancer Centre Hospital East, 6-5-1, progression-free survival (PFS) rates at 20 weeks of 40% and Kashiwanoha, Kashiwa, 277-8577, Japan. Tel: þ81-4-7134-6920; Fax: þ81-4- 78%, respectively.4 In May 2017, the US Food and Drug 7134-6928 Administration (FDA) approved pembrolizumab for the E-mail: [email protected] (T. Yoshino). treatment of adult and paediatric patients with unresectable 0923-7534/© 2020 European Society for Medical Oncology. Published by or metastatic, microsatellite instability-high (MSI-H) or dMMR Elsevier Ltd. All rights reserved.

Volume 31 - Issue 7 - 2020 https://doi.org/10.1016/j.annonc.2020.03.299 861 Annals of Oncology T. Yoshino et al. solid tumours, based on data from 149 patients from five consensus recommendations on tumour-agnostic therapies single-arm studies.5 Thus, pembrolizumab became the first based on the results of expert voting on a series of pre- drug to receive a tumour-agnostic approval.6 In December formulated recommendations focussing on patients with 2018, the Japan Pharmaceuticals and Medical Devices Agency advanced (unresectable or metastatic) MSI/dMMR and (PMDA) approved pembrolizumab for the treatment of adult NTRK fusion-positive solid tumours, as outlined below. patients with advanced MSI-H tumours.7 These approvals in turn are supported by the results of Aim an expanded proof-of-concept study which showed MSI/ The aim of the meeting was to generate a document that dMMR predicts response to PD-1 blockade across a range of could provide guidance for the use and management of the solid tumour types8 and by a review of immune checkpoint 9 currently approved tumour-agnostic therapies in patients blockade therapies in patients with MSI/dMMR tumours. with solid tumours, and to aid clinical trial design for both Additionally, another monoclonal antibody that targets these agents and those currently under development, going the PD-1 receptor, nivolumab, had previously been forward. approved by the FDA for the treatment of adults and chil- dren with MSI or dMMR metastatic CRC that had pro- Scope gressed following treatment with fluoropyrimidine, oxaliplatin, and irinotecan, as a single agent and subse- The meeting focused exclusively on the tumour-agnostic quently in combination with ipilimumab.10,11 therapies associated with MSI/dMMR and NTRK fusions. In November 2018, larotrectinib, a tyrosine kinase in- METHODOLOGY hibitor of the tropomyosin receptor kinase (TRK) proteins, TRKA, TRKB, and TRKC, encoded for by the neurotrophic Composition of the expert panel and aims tyrosine receptor kinase genes NTRK1, NTRK2 and NTRK3 This manuscript represents the opinion of 19 experts in respectively, became the second drug to receive tumour- oncology, representing JSCO and JSMO, ESMO, ASCO and agnostic FDA approval for the treatment of adult and pae- TOS, who took part in a survey of clinical questions (CQs) diatric patients with solid tumours with NTRK gene devised to test our thinking on the management of patients fusions.12,13 In 2019, larotrectinib became the first tumour- with MSI/dMMR and NRK fusion-positive tumours in the agnostic cancer treatment to be approved in the European era of tumour-agnostic drug approvals. Union. Following on from these first approvals, in 2019, Japan Clinical questions and proposed recommendations and subsequently the FDA approved entrectinib, a selective tyrosine kinase inhibitor that targets TRKA, TRKB and TRKC, In preparation for the meeting, six identical CQs relating to and the ROS1 and ALK proteins14 for patients with NTRK the MSI/dMMR and NTRK precision agnostic therapy ap- fusion-positive advanced, recurrent solid tumours.15 These proaches were formulated by Drs T. Yoshino, S. Mishima, Y. tumour-agnostic agent approvals pose several clinical Naito, H. Taniguchi and J.-Y. Douillard and approved by all questions regarding not only MSI/MMR/NTRK testing but the experts (Table 1). The evidence to support the two sets also the sequence of administration of these agents in the of recommendations proposed in response to these CQs treatment pathways of patients with MSI/dMMR or NTRK was provided by searching the PubMed and Cochrane da- fusion-positive solid tumours. Also, going forward, should tabases using the search terms listed in supplementary all cancer patients be tested, and if so, when, and using which test(s)? The European Society for Medical Oncology (ESMO) Table 1. The six identical clinical questions (CQs) formulated for the treatment and management of patients with MSI/dMMR or NTRK fusion- recommendations on MSI testing for immunotherapy, and positive tumours from which two separate series of recommendations for the detection of patients with tumours with NTRK fu- were developed, i.e. one series of clinical recommendations for each sions, were published in May 201916 and July 2019,17 clinical situation respectively. Also, the Japan Society of Clinical Oncology CQ no. CQs ‘ ’ (JSCO) published provisional clinical opinion guidelines for CQ1 Should all patients with solid tumours be tested for MSI/ the diagnosis and use of immunotherapy in patients with MMR or NTRK fusions? dMMR tumours, in July 2019.18 In order to respond to the CQ2 When is the optimal timing for tests for MSI/MMR or for NTRK fusions? potential changes in clinical practice envisaged following CQ3 Which tests are recommended for determining MSI/MMR the tumour-agnostic agent approvals described above, and status or NTRK fusions? those anticipated for other agents in the future, the JSCO CQ4 What is the appropriate biospecimen for testing for MSI/ MMR or NTRK fusions? convened a face-to-face meeting in Japan, in October 2019, CQ5 Which treatment is recommended for MSI/dMMR patients of international experts in the field of oncology represent- or patients with NTRK fusions? ing the oncology societies of Europe (ESMO), the United CQ6 Where in the treatment algorithm should immunotherapy be used in the treatment of patients with MSI/dMMR solid States (ASCO), and two additional Asian societies namely, tumours or a TRK inhibitor be used in the treatment of the Japanese Society of Medical Oncology (JSMO) and the patients with NTRK fusion-positive solid tumours? Taiwan Oncology Society (TOS). The ultimate aim of the dMMR, deficient in (DNA) mismatch repair; MSI, microsatellite instability; NTRK, meeting was to develop the present international expert neurotrophic tyrosine receptor kinase; TRK, tropomyosin receptor kinase.

862 https://doi.org/10.1016/j.annonc.2020.03.299 Volume 31 - Issue 7 - 2020 T. Yoshino et al. Annals of Oncology

Tables S1 and S2 (available at Annals of Oncology online) for unqualified response of YES in the pre-meeting survey MSI/MMR and NTRK, respectively. The details of the num- equated with ‘accept completely’ in the final voting, giving ber of records identified in response to each clinical ques- an LoA of A ¼ 100%. The remaining 10 draft recommen- tion during the systematic review and the number of dations, four for MSI/MMR (supplementary Table S8, records finally used in the synthesis of the recommenda- available at Annals of Oncology online), and six for NTRK tions are presented in supplementary Tables S3 and S4 (supplementary Table S9, available at Annals of Oncology (available at Annals of Oncology online) for MSI/MMR and online), were discussed and voted upon at the face-to-face NTRK, respectively. The two sets of proposed recommen- meeting. Each of the four groups/organisations represented dations made in response to each CQ, relating to MSI/ at the face-to-face meeting (i.e. JSCO/JSMO, ESMO, ASCO, dMMR and NTRK fusion-positive tumours together with the TOS) had the right to one vote each per recommendation. proposed levels of evidence (LoE) and grades of recom- Where changes to the text of the original recommendations mendation (GoR), based on an adapted version of the ‘In- were made, these are indicated in bold both in the main fectious Diseases Society of America-United States Public text of the manuscript and in the two summary tables of Health Service Grading System’ (supplementary Table S5, the final consensus recommendations, Tables 2 and 3.In available at Annals of Oncology online),19 were then circu- addition, the final voting patterns, in terms of GoR, LoE and lated to all 19 experts to gather their acceptance or LoA, were recorded for each recommendation. otherwise of the recommendations made (see supplementary Tables S6 and S7, available at Annals of Oncology online). The responses of the experts had to represent science-based opinion assuming that all drugs, Table 2. Summary of the expert recommendations for the treatment of diagnostic and testing modalities were available to them. patients with MSI/dMMR solid tumours

CQ1. Should all patients with solid tumours be tested for MSI/MMR? Final consensus statements 1-1 Patients with advanced (unresectable or metastatic) solid tumours with a high incidence of MSI/dMMR should be Where there was full agreement between all voting parties tested for their MSI/MMR status. for the recommendations made in response to each CQ no [LoE: III, GoR for testing: A, LoA: A [ 100%] further discussion was required. Where there was an 1-2 Patients with advanced (unresectable or metastatic) solid fi tumours with a low incidence of MSI/dMMR should be absence of full agreement, however, a modi ed Delphi considered for MSI/MMR testing. process was used during the final voting process at the face- [LoE: III, GoR for testing: B, LoA: A [ 100%] to-face working meeting to develop each of the disputed 1-3 Patients with localised resectable non-colorectal tumours should not be considered for MSI/MMR testing outside of recommendations towards a consensus. The experts pre- a clinical trial, unless Lynch syndrome is clinically sent were asked to vote on their level of agreement (LoA) suspected. [ for a particular recommendation based on the evidence [LoE: V, GoR for testing: D, LoA: A 100%] CQ2. When is the optimal timing for tests for MSI/MMR? available, on a scale of A to E, where A ¼ accept completely, MSI/MMR should be tested before or during the standard B ¼ accept with some reservation, C ¼ accept with major treatment of advanced (unresectable or metastatic) solid reservation, D ¼ reject with some reservation and E ¼ tumours. [LoE: V, GoR: A, LoA: A [ 100%] reject completely (supplementary Table S5, available at CQ3. Which tests are recommended for determining MSI/MMR status? Annals of Oncology online).19 A consensus was considered 3-1 IHC is highly recommended for testing. [ to have been achieved when 80% of experts voted to [LoE: III, GoR for testing: A, LoA: A 100%] 3-2 PCR is recommended for testing either upfront or when accept completely (A) or accept with some reservation (B), a IHC is equivocal or not available. specific recommendation made in response to a particular [LoE: III, GoR for testing: B, LoA: A [ 75%, B [ 25%] CQ. A recommendation was considered to have been 3-3 Validated NGS is recommended for testing either upfront or when IHC is equivocal or not available. rejected when >80% of the voting members indicated [LoE: III, GoR for testing: B, LoA: A [ 75%, B [ 25%] ‘reject completely’ (E) or ‘reject with some reservation’ (D). CQ4. What is the appropriate biospecimen for testing for MSI/MMR? Formalin-fixed, paraffin-embedded tissue blocks are appropriate for testing. RESULTS AND MEETING OUTCOMES [LoE: V, GoR: A, LoA: A [ 100%] CQ5. Which treatment is recommended for MSI/dMMR patients? In the initial pre-meeting surveys, the 19 experts reported PD-1/PD-L1 inhibitors are strongly recommended for on the applicability of the 10 recommendations developed patients with MSI/dMMR tumours. in response to the six CQs (Table 1) in relation to MSI/ [LoE: III, GoR: A, LoA: A [ 100%] CQ6. Where in the treatment algorithm should immunotherapy be used in dMMR tumours (supplementary Table S8, available at MSI/dMMR solid tumours? Annals of Oncology online) and on the applicability of the We recommend immunotherapy for patients with MSI/ 13 recommendations developed in response to the same six dMMR during the course of their therapy when no other satisfactory treatment options exist depending on the CQs for the treatment of patients with tumours with NTRK clinical context. gene fusions (supplementary Table S9, available at Annals [LoE: III, GoR: A, LoA: A [ 100%] of Oncology online). dMMR, deficient in (DNA) mismatch repair; GoR, grade of recommendation; IHC, immunohistochemistry; LoA, level of agreement; LoE, level of evidence; MSI, mi- Of the 23 recommendations developed in response to crosatellite instability; MMR, mismatch repair; NGS, next generation sequencing; the six CQs across both biomarker categories, 13 were fully PCR, polymerase chain reaction; PD-1 programmed (cell) death protein-1; PD-L1, agreed upon during the pre-meeting surveys. An programmed death ligand-1.

Volume 31 - Issue 7 - 2020 https://doi.org/10.1016/j.annonc.2020.03.299 863 Annals of Oncology T. Yoshino et al.

Historically, tumour MSI/MMR status has been used to Table 3. Summary of the expert recommendations for the treatment of patients with solid tumours with NTRK fusions guide prognosis for patients with stage II CRC and to poten- tially predict the efficacy of chemotherapy in patients with CQ1. Should all patients with solid tumours be tested for NTRK fusion? 20 1-1 Patients with advanced (unresectable or metastatic) solid CRC. MSI/dMMR is also found to varying degrees in other 21e23 tumours without actionable and driver gene mutations/ tumour types. This together with the recent evidence fusions/amplifications should be tested for NTRK fusion. fi [ that MMR de ciency is predictive of response to immune [LoE: V, GoR: B, LoA: A 100%] 8,24 1-2 Patients with advanced (unresectable or metastatic) solid checkpoint inhibitors and the agnostic approval of pem- tumours which are highly likely to harbour NTRK fusions brolizumab, based on tumour MSI/MMR status, points to should be tested for NTRK fusion, especially ETV6-NTRK3 MSI/MMR status becoming increasingly important in the fusion. [LoE: V, GoR: A, LoA: A [ 100%] management of cancer patients in the era of precision ther- 1-3 Patients with advanced (unresectable or metastatic) solid apy. It therefore seems prescient to determine in which pa- tumours other than above (CQ1-1 and 1-2) should be tients MSI/MMR testing is appropriate, and when and which considered for testing for NTRK fusions. [LoE: V, GoR: A, LoA: A [ 100%] tests for MSI/MMR tumour status should be carried out. 1-4 Patients with locally-advanced tumours with a high incidence of NTRK fusions should be tested when considering neoadjuvant therapy before resection. Recommendations in response to the CQs for MSI/MMR [LoE: V, GoR: B, LoA: A [ 100%] Six of the 10 draft recommendations made in response to CQ2. When is the optimal timing for tests for NTRK fusion? NTRK fusion testing should be considered before or during the six CQs in relation to MSI/MMR (Table 1) were accepted the standard treatment of advanced (unresectable or completely in the pre-meeting survey, i.e. LoA A ¼ 100% metastatic) solid tumour. (supplementary Table S8, available at Annals of Oncology [LoE: V, GoR: B, LoA: A [ 100%] CQ3. Which tests are recommended for determining NTRK fusions? online). Thus, theoretically four recommendations (CQs1-1 3-1 IHC is not recommended for confirming NTRK fusion. It and 1-3, CQ3-2 and CQ6) had to be discussed at the face-to- may be used for screening to enrich patients with NTRK face meeting. In reality, some of the other recommenda- fusion. [LoE: V, GoR: B, LoA: A [ 100%] tions were revised. All 10 recommendations are discussed 3-2 In situ hybridisation (ISH, e.g. FISH) for ETV6-NTRK3 fusion in the text below and changes made to the original rec- is recommended for patients with tumours which are ommendations (supplementary Table S6, available at highly likely to harbour NTRK fusions. ISH is not recommended for patients other than the above. Annals of Oncology online) indicated in bold text. [LoE: V, GoR: B, LoA: A [ 100%] 3-3 RT-PCR for ETV6-NTRK3 fusion is recommended for CQ1: Should all patients with solid tumours be tested for patients with tumours which are highly likely to harbour MSI/MMR? NTRK fusions. [LoE: V, GoR: B, LoA: A [ 100%] Recommendation CQ1-1. Patients with advanced (unre- 3-4 NGS which detects NTRK fusion is recommended for testing sectable or metastatic) solid tumours with a high inci- for NTRK fusion. dence of MSI/dMMR should be tested for their MSI/MMR [LoE: V, GoR: C, LoA: A [ 100%] CQ4. What is the appropriate biospecimen for testing for NTRK fusions? status. Both fresh samples as well as archival tissue samples [LoE: III, GoR for testing: A, LoA: A ¼ 100%] properly fixed and preserved are appropriate for testing. Recommendation CQ1-2. Patients with advanced (unre- [LoE: V, GoR: B, LoA: A [ 100%] CQ5. Which treatment is recommended for patients with NTRK fusions? sectable or metastatic) solid tumours with a low inci- TRK inhibitors are strongly recommended for patients with dence of MSI/dMMR should be considered for MSI/MMR NTRK fusions. testing. [LoE: III, GoR: A, LoA: A [ 100%] CQ6. Where in the treatment algorithm should a TRK inhibitor be used in [LoE: III, GoR for testing: B, LoA: A ¼ 100%] the treatment of patients with NTRK fusion-positive solid tumours? Recommendation CQ1-3. Patients with localised resect- We recommend TRK inhibitors for patients with NTRK able non-colorectal tumours should not be considered for fusions during the course of therapy, when no other satisfactory treatment options exist, depending on the MSI/MMR testing outside of a clinical trial unless Lynch clinical context. syndrome is clinically suspected. [ [LoE: III, GoR: A, LoA: A 100%] [LoE: V, GoR for testing: D, LoA: A ¼ 100%] GoR, grade of recommendation; IHC, immunohistochemistry; ISH, in situ hybrid- isation; LoA, level of agreement; LoE, level of evidence; NGS, next generation sequencing; NTRK, neurotrophic tyrosine receptor kinase; PCR, polymerase chain All the experts agreed with and accepted completely reaction; RT-PCR, reverse transcriptase PCR; TRK, tropomyosin receptor kinase. ‘recommendation CQ1-2’ above in the pre-meeting survey (supplementary Table S8, available at Annals of Oncology online). However, the experts thought the tumours highly Background to development of MSI/dMMR status as a likely to harbour MSI/dMMR in ‘recommendation CQ1-1’ predictive biomarker should be better defined, together with the definition of Cancers deficient in MMR (dMMR) are associated with early disease as it applied to ‘recommendation CQ1-3’,as short tandem-repeat sequences (microsatellites) and are early disease is not included in the label. characterised by exceptionally high numbers of somatic A pooled-data analysis of four large population-based mutations due to errors in DNA MMR. Such cancers are cohorts of CRC patients has shown universal screening of classified as exhibiting MSI, which is the phenotype of CRC patients using tumour MMR testing to be more sen- dMMR. Tumour dMMR status is the consequence of mu- sitive than clinical criteria25 in diagnosing Lynch syndrome. tations in the MLH1, MSH2, MSH6, PMS2,orEPCAM genes. Thus, patients with tumours which may be MSI/dMMR, and

864 https://doi.org/10.1016/j.annonc.2020.03.299 Volume 31 - Issue 7 - 2020 T. Yoshino et al. Annals of Oncology for whom MSI/MMR testing is generally recommended, the requirement for adjuvant chemotherapy in patients should include patients clinically suspected of having Lynch with early-stage (stage II) CRC, although not in the early syndrome and elderly female CRC patients with tumours stages of any other tumour type. After the revisions high- with a mucinous component or with a BRAF p.V600E mu- lighted in bold text above, all the experts agreed with and tation.26 A summary of tumours highly likely to harbour accepted completely ‘recommendation CQ1-3’ [A ¼ 100%]. MSI/dMMR is provided in supplementary Table S10 (avail- The GoR for testing was revised to D. able at Annals of Oncology online), based on data from a study of 15 045 patients with >50 different cancer types CQ2. When is the optimal timing for tests for MSI/MMR? 22 (NCT01775072). The wording of ‘recommendation CQ1-1’ Recommendation CQ2. MSI/MMR status should be tested was revised to specify ‘advanced (unresectable or meta- before or during the standard treatment of advanced static) solid tumours with a high incidence of MSI/dMMR’ (unresectable or metastatic) solid tumours. and the GoR revised to an A. All the experts agreed with [LoE: V, GoR: A, LoA: A ¼ 100%] and accepted completely [A ¼ 100%] the revised recommendation. Since the turnaround time for MSI/MMR testing is 1e2 Also, although all the experts agreed with and accepted weeks, MSI/MMR testing should be carried out early to completely ‘recommendation CQ1-2’ above during the pre- determine a patient’s eligibility for treatment with PD-1/PD- meeting survey, at the face-to-face meeting there was L1 inhibitors. Additionally, in the case of solid tumours for considerable discussion about the cost/economic issues of which the applicability of PD-1/PD-L1 inhibitors is judged testing patients with solid tumours associated with a low appropriate based on a biomarker other than MSI/MMR incidence of MSI/dMMR. However, because the efficacy of status, such as PD-L1 expression, and that biomarker is PD-1/PD-L1 inhibitors has been clearly and consistently negative, MSI/MMR testing is recommended, because demonstrated in advanced solid tumours with MSI/ these drugs are expected to be effective if the tumour is 8,10,11,27 dMMR, the expert opinion was that MSI/MMR MSI/dMMR.18 The general feeling of the experts was that testing should be considered to determine eligibility for the ideal scenario would be to test at the time of diagnosis treatment with PD-1/PD-L1 inhibitors for all patients with and tissue availability, when there may be only one chance advanced solid tumours (‘recommendation CQ1-2’). Clearly, at biopsy. All the experts agreed with and accepted in principle, it is not necessary to perform MSI/MMR testing completely ‘recommendation CQ2’ [A¼100%]. for solid tumours for which PD-1/PD-L1 inhibitors can be used in the second- or later-line treatment settings irre- CQ3. Which tests are recommended for determining MSI/ spective of MMR functionality. MSI/MMR testing may be MMR status? considered if it provides predictive value for PD-1/PD-L1 Recommendation CQ3-1. IHC is highly recommended for inhibitors and may prompt their use earlier in the treat- testing. ment path minimising the percentage of patients who will [LoE: III, GoR for testing: A, LoA: A ¼ 100%] miss out on immunotherapy as a result of rapid clinical Recommendation CQ3-2. PCR is recommended for testing deterioration. Thus, the ‘recommendation CQ1-2’ that ‘pa- either upfront or when IHC is equivocal or not available. tients with advanced solid tumours should be tested for [LoE: III, GoR for testing:B,LoA: A ¼ 75%, B ¼ 25%] MSI/MMR’ was revised to read ‘Patients with advanced Recommendation CQ3-3. Validated NGS is recommended (unresectable or metastatic) solid tumours with a low for testing either upfront or when IHC is equivocal or not incidence of MSI/dMMR should be considered for MSI/ available. MMR testing’, making it optional depending on treatment [LoE: III, GoR for testing: B, LoA: A ¼ 75%, B ¼ 25%] location and cost concerns. The GoR for testing was revised to B, and the experts present agreed with and accepted All the experts agreed with and accepted completely completely [A ¼ 100%] the revised recommendation. In the ‘recommendations CQ3-1’ and ‘CQ3-3’ in the pre-meeting case of ‘recommendation CQ1-3’ where the recommenda- survey. However, there was a query over the suggestion tion was that patients with early-stage disease should not in ‘recommendation CQ3-2’ that polymerase chain reaction be tested outside of a clinical trial setting, the experts (PCR) is highly recommended for testing. expressed concern over the definition of early disease, and Tumour MSI/MMR status can be tested using immuno- thought that general testing needed to be separated from histochemistry (IHC), PCR and more recently by next gen- testing in situations where Lynch syndrome was suspected, eration sequencing (NGS) techniques.30 The expression of and the text was revised accordingly (see above and MMR proteins (MLH1, MSH2, MSH6 and PMS2) in tumour Table 2). The use of immune checkpoint inhibitors in MSI/ tissue is typically examined by IHC in the first instance to dMMR early-stage colon cancer is presently being evaluated evaluate whether the tumour is dMMR and is the approach in clinical trials. Furthermore, it is known that MSI/dMMR recommended in the recently published ESMO recom- status is a favourable prognostic factor for CRC, particularly mendations on MSI (MMR) testing for immunotherapy in for stage II CRC20,28,29 in which MSI/dMMR status has cancer.16 If IHC expression of at least one protein is lost, the negative implications in terms of benefitfrom5-fluorouracil tumour is considered to be dMMR. If the IHC results are (5-FU) adjuvant chemotherapy.20,29 As a consequence, it is equivocal, the ESMO recommendation is to use MSI-PCR, considered desirable to perform MSI/MMR testing to assess based on PCR amplification of microsatellite markers.16

Volume 31 - Issue 7 - 2020 https://doi.org/10.1016/j.annonc.2020.03.299 865 Annals of Oncology T. Yoshino et al.

There was considerable discussion amongst the experts at therapy39,40 and MSH6 protein expression is reported to be the face-to-face meeting about the use of PCR (‘recom- lost after neoadjuvant radiation,41 it is desirable to use mendation CQ3-2’). For example, it was agreed that con- specimens for testing that have not been exposed to ventional MSI-PCR, which was developed and validated for cisplatin or radiation therapy. A freshly frozen tissue spec- colon cancer, was an excellent approach for patients with imen may be used if it is histologically confirmed that there CRC but that its accuracy was inferior in other tumour types are sufficient tumour cells, for the specific testing method, such as endometrial and prostate cancers.21,31 A five poly-A contained in the specimen. As stated previously (CQ2), the panel comprising five poly-A mononucleotide repeats is the general feeling was that ideally testing should be done at panel recommended by ESMO for MSI-PCR testing, due to the time of diagnosis and tissue availability, when there may its higher sensitivity and specificity,32 with MSI defined as be only one chance at biopsy. ‘loss of stability in 2 of the five microsatellite markers’.16 In addition, IHC is not reimbursed in all countries and CQ5. Which treatment is recommended for MSI/dMMR MSI-PCR is the upfront test of choice, and is also generally patients? indicated for the assessment of dMMR in cancers belonging Recommendation CQ5. PD-1/PD-L1 inhibitors are strongly to the spectrum of Lynch syndrome cancer types. The MSI- recommended for patients with MSI/dMMR tumours. PCR test kit FALCO has been approved in Japan as a com- [LoE: III, GoR: A, LoA: A ¼ 100%] panion diagnostic for pembrolizumab.18 After discussion, the experts from Japan, Taiwan and ESMO agreed with and All the experts agreed with and accepted completely accepted completely [A ¼ 75%] the revised ‘recommenda- [A ¼ 100%] ‘recommendation CQ5’ and the supporting tion CQ3-2’ (see revisions in bold text above), while the evidence in the pre-meeting survey. representatives of ASCO could only accept the revised PD-1 inhibitors are strongly recommended for the treat- recommendation with some reservation [B ¼ 25%]. ment of patients with MSI/dMMR solid tumours based on the NGS represents an alternative molecular test for the evidence from the clinical trials of pembrolizumab.4,5,8,42,43 In detection of tumour MSI status21,33 and includes several addition, both nivolumab monotherapy and nivolumab/ipili- techniques.21,34,35 NGS also has the potential to determine mumab combination therapy have demonstrated activity in tumour mutation burden (TMB). Interestingly, in the clinical MSI/dMMR metastatic CRC patients10,11 and more recently trials conducted for the application to the FDA for the nivolumab has been shown to be effective in non-colorectal approval of pembrolizumab, the screening tests for MSI/ tumours that are dMMR.44 The PD-L1 inhibitor durvalumab MMR did not include NGS. However, the reported concor- has also demonstrated efficacy in two ongoing studies (a dance rates between NGS testing and MSI-PCR testing and phase II trial in MSI/dMMR CRC and a phase I/II trial in pa- between NGS and IHC are both extremely high.36 NGS tients with MSI/dMMR solid tumours).27 testing has the potential to become the test of choice going CQ6. Where in the treatment algorithm should immuno- forward for determining patient eligibility for treatment therapy be used in the treatment of patients with MSI/ with PD-1/PD-L1 inhibitors, but should only be carried out at selected specialist centres or through validated central dMMR solid tumours? Recommendation CQ6. We recommend immunotherapy laboratory methods. It might also offer the potential to 37,38 for patients with MSI/dMMR during the course of their assess tumour response during anti-PD-1 therapy. Ex- perts from three of the four groups/organisations repre- therapy when no other satisfactory treatment options sented agreed with and accepted completely [A ¼ 75%] the exist depending on the clinical context. ¼ revised ‘recommendation CQ3-3’, while those of the fourth [LoE: III, GoR: A, LoA: A 100%] could only accept the revised recommendation with All the experts except one agreed with and accepted some reservation [B ¼ 25%]. The GoR for testing was ‘ ’ revised to a B. completely recommendation CQ6 , and the supporting ev- idence in the pre-meeting survey, but eventually the CQ4. What is the appropriate biospecimen for testing for recommendation was reworded to be less prescriptive in MSI/MMR? terms of the timing of immunotherapy. Recommendation CQ4. Formalin-fixed, paraffin- PD-1 inhibitors have demonstrated efficacy in patients with embedded (FFPE) tissue blocks are appropriate for previously-treated MSI/dMMR solid tumours.4,5,8,10,42,45 testing. Thus, pembrolizumab and nivolumab can be considered for [LoE: V, GoR: A, LoA: A ¼ 100%] second- or later-line treatment in patients with MSI/dMMR solid tumours. Also, a recent case report describes dual im- All the experts agreed with and accepted completely mune checkpoint blockade with ipilimumab plus nivolumab, ‘recommendation CQ4’ and the supporting evidence in the following sequential therapy with the PD-1 and PD-L1 in- pre-meeting survey. hibitors pembrolizumab and atezolizumab, in a patient with Thus, the expert opinion was that the recommended Lynch syndrome and metastatic colon and localised urothelial specimens for MSI/MMR testing should be FFPE tissue cancers.46 This suggests that, for some patients with MSI/ blocks of surgical specimens. Also, since MLH1 and MSH6 dMMR tumours, multiple sequential immune checkpoint protein expression is possibly lost after cisplatin-containing therapies may be beneficial. The GoR was revised to A.

866 https://doi.org/10.1016/j.annonc.2020.03.299 Volume 31 - Issue 7 - 2020 T. Yoshino et al. Annals of Oncology

Background to the development of NTRK fusions as a meeting survey (supplementary Table S9, available at biomarker for TRK inhibitors Annals of Oncology online). They thought that the wording ‘ ’ Oncogenic NTRK gene fusions induce tumour cell prolifer- of recommendations CQ1-1 to CQ1-3 should be revised to ation and activate various cancer-related downstream sig- specify advanced (unresectable or metastatic) solid tu- fi nalling pathways.13,17,47 NTRK1 gene fusions were first mours, to better de ne advanced disease, and the wording ‘ ’ fi fi identified in colon cancer48,49 but have since been identified of recommendation CQ1-4 re ned to better de ne early ‘ in a range of adult and paediatric tumours together with disease. These changes are highlighted in bold text in rec- e ’ gene fusions involving the NTRK2 and NTRK3 genes.50 54 ommendations CQ1-3 and CQ1-4 above. ’ Although NTRK gene fusions are common in a small num- At the face-to-face meeting, the experts recommenda- ‘ ’ ber of rare adult and paediatric tumour types, they also tion was that the wording of recommendation CQ1-1 was ‘ occur at lower frequencies in many common tumour types revised to patients without actionable and driver gene fi ’ (supplementary Table S11, available at Annals of Oncology mutations/fusions/ampli cations should be tested , as the online).17,55 Nearly always the 30 region of the NTRK gene is original wording was felt to be confusing, as currently, there joined with the 50 region of an unrelated fusion partner are no published data showing the coexistence of an NTRK gene.13,17,55 Currently, approximately 80 different 50 fusion fusion and certain actionable drivers (EGFR, ALK and ROS1 fi in NSCLC, KIT in gastrointestinal stromal tumour, and BRAF partners have been identi ed but the best known of the 53,64 NTRK fusions is the ETV6-NTRK3 gene fusion which occurs in in NSCLC and malignant melanoma). Also, an indepen- >95% of secretory carcinomas of the breast.56 dent analysis of the available datasets for any overlap be- Larotrectinib and entrectinib are TRK inhibitors and are tween NTRK fusions and other mutations, in particular currently being investigated in patients with oncogenic NTRK oncogenes/driver gene mutations, according to tumour 12,15,57e59 type (GENIE dataset), identified an overlap with certain in- 1, 2, and 3 gene fusions. Their recent approval for 65,66 the tumour-agnostic treatment of patients with NTRK fusions frame mutations but not with key actionable mutations. means that there is a need for guidance on the diagnosis and NTRK fusions have been reported to occur with a fre- e fi treatment of patients with tumours with NTRK fusions. The quency of 75% 100% in infantile brosarcoma (congenital fi 67e71 ESMO has recently published recommendations on the brosarcoma), secretory carcinoma of the breast,56,72,73 MASC74e77 and congenital mesoblastic standard methods to detect NTRK fusions in daily practice 71 and also for clinical research. Two other key publications on nephroma, mostly as ETV6-NTRK3 fusions, and these pa- e ‘ NTRK fusion detection across multiple assays60 62 and the tients should therefore be tested ( recommendation CQ1- 63 2’). In common tumours which harbour NTRK fusions at low molecular characterisation of cancers with NTRK fusions 50e52,78 have also recently been published. It is hoped that these frequency various partner genes have been re- publications will help inform the consensus recommenda- ported. Since TRK inhibitors have been shown to have tions generated below in response to the CQs in Table 1. excellent activity in patients with NTRK fusions, with acceptable toxicity,12,13,15,79,80 all patients with unresect- CQ1: Should all patients with solid tumours be tested for able or metastatic advanced solid tumours, other than NTRK fusion? those described in ‘recommendations CQ1-1 and 1-2’ Recommendation CQ1-1. Patients with advanced (unre- above, should be considered for testing for NTRK fusions to sectable or metastatic) solid tumours without actionable avoid missing the opportunity of treatment with a TRK in- and driver gene mutations/fusions/amplifications should hibitor (‘recommendation CQ1-3’). be tested for NTRK fusion. Finally, although there is only limited evidence to support [LoE: V, GoR: B, LoA: A ¼ 100%] the clinical utility of TRK inhibitors in patients with early-stage Recommendation CQ1-2. Patients with advanced (unre- solid tumours,81 it was felt that the high response rate of TRK sectable or metastatic) solid tumours which are highly inhibitors in tumours harbouring NTRK fusions meant that likely to harbour NTRK fusions should be tested for NTRK the use of a TRK inhibitor in the neoadjuvant setting could be fusion, especially ETV6-NTRK3 fusion. considered, with complete rewording of the initial recom- [LoE: V, GoR: A, LoA: A ¼ 100%] mendation to better define early-stage solid tumours (see Recommendation CQ1-3. Patients with advanced (unre- bold text ‘recommendation CQ1-4’ above) and the GoR sectable or metastatic) solid tumours other than above revised to B. All the experts agreed with and accepted (CQ1-1 and 1-2) should be considered for testing for NTRK completely [A ¼ 100%] the revised recommendation. fusions. [LoE: V, GoR: A, LoA: A ¼ 100%] CQ2. When is the optimal timing for tests for NTRK fusion? Recommendation CQ1-4. Patients with locally-advanced Recommendation CQ2. NTRK fusion testing should be tumours with a high incidence of NTRK fusions should considered before or during the standard treatment of be tested when considering neoadjuvant therapy before advanced solid tumours. resection. [LoE: V, GoR: B, LoA: A ¼ 100%] [LoE: V, GoR: B, LoA: A ¼ 100%] The experts queried the initial recommendation in the pre- All the experts agreed with and accepted completely meeting survey. The general feeling was that testing for NTRK ‘recommendations CQ1-1 and CQ1-2’ above in the pre- fusions should be considered before or during standard first- or

Volume 31 - Issue 7 - 2020 https://doi.org/10.1016/j.annonc.2020.03.299 867 Annals of Oncology T. Yoshino et al. subsequent-line therapy for advanced solid tumours charac- an approach for the de novo detection of transcribed fusion terised by a high frequency of NTRK fusions, and otherwise only genes. Thus, validated NGS methods which cover NTRK fu- in the context of a larger NGS panel that is being conducted to sions regardless of fusion partner are recommended.88 The identify other mutations. Thus, the recommendation was application of all these techniques is described in detail in reworded (see bold text above) and the GoR revised to B and the ESMO recommendations.17 The challenge in terms of accepted completely [A ¼ 100%] by all the experts present. diagnosis is to find a method that allows the rapid, accurate testing of a large number of patients. CQ3. Which tests are recommended for determining NTRK fusions? CQ4. What is the appropriate biospecimen for testing for Recommendation CQ3-1. IHC (immunohistochemistry) is NTRK fusions? not recommended for confirming NTRK fusion. It may be Recommendation CQ4. Both fresh samples as well as used for screening to enrich for patients with NTRK archival tissue samples properly fixed and preserved are fusions. appropriate for testing. [LoE: V, GoR: B, LoA: A ¼ 100%] [LoE: V, GoR: B, LoA: A ¼ 100%] Recommendation CQ3-2. In situ hybridisation [ISH, e.g. fluorescence ISH (FISH)] for ETV6-NTRK3 fusion is rec- Three studies were included in the qualitative synthesis 62,79,89 ommended for patients with tumours which are highly of this recommendation, and all the experts agreed likely to harbour NTRK fusions. ISH is not recommended with and accepted completely [A ¼ 100%] ‘recommenda- for patients other than the above. tion CQ4’ without revision. [LoE: V, GoR: B, LoA: A ¼ 100%] Archival FFPE tissue sections are appropriate for IHC, Recommendation CQ3-3. Reverse transcriptase (RT)-PCR FISH, RT-PCR and anchored multiplex (PCR) NGS if properly 85 for ETV6-NTRK3 fusion is recommended for patients with fixed and preserved. The quality of the archival material tumours which are highly likely to harbour NTRK fusions. to be tested is crucial, and FFPE RNA in particular is known [LoE: V, GoR: B, LoA: A ¼ 100%] to be labile. In the basket study of entrectinib, both fresh 15 Recommendation CQ3-4. Next generation sequencing and archival tissue was used. It may be necessary to (NGS) which detects NTRK fusion is recommended for recommend that, when necessary, patients should be re- testing NTRK fusion. biopsied to obtain appropriate tissue for examination. ¼ [LoE: V, GoR: C, LoA: A 100%] CQ5. Which treatment is recommended for patients with NTRK fusions? All the experts agreed with and accepted completely Recommendation CQ5. TRK inhibitors are strongly rec- ¼ [A 100%] the four recommendations listed above without ommended for patients with NTRK fusion. fi revision. A fth recommendation, originally CQ3-4, [LoE: III, GoR: A, LoA: A ¼ 100%] regarding the predictive value of nanostring technology was deleted due to a paucity of data and the original Although there has been no study comparing the two ‘ ’ recommendation CQ3-5 (supplementary Table S9, avail- TRK inhibitors (larotrectinib, entrectinib) approved for ‘ able at Annals of Oncology online) became recommenda- tumour-agnostic therapy, with other standard treatment ’ tion CQ3-4 . options, they have shown high and durable re- IHC examines the expression of the TRK proteins but does sponses13,15,59,79 coupled with relatively mild toxicity pro- 82e84 Thus, negative pro- not directly detect NTRK fusions. files. Thus, based on the available evidence, TRK inhibitors tein expression determined by TRK IHC only predicts a lack are strongly recommended for patients with NTRK fusions. of NTRK fusions.85 Consequently, IHC, when positive, may be used to enrich for patients with NTRK fusions as part of a CQ6. Where in the treatment algorithm should a TRK in- two-step process for their detection. It is noted that IHC hibitor be used in the treatment of patients with NTRK shows lower sensitivity for NTRK3 fusions, and both sensi- fusion-positive solid tumours? tivity and specificity were poor in sarcomas in one report.62 Recommendation CQ6. We recommend TRK inhibitors for ISH is also not recommended for the routine detection of patients with NTRK fusions during the course of therapy NTRK fusions in all patients but can be used in patients with when no other satisfactory treatment options exist tumours which are highly likely to harbour ETV6-NTRK3 depending on the clinical context. fusions. RT-PCR77,86 is designed to identify only known [LoE: III, GoR: A, LoA: A ¼ 100%] fusion partners and breakpoints and is not recommended for routine detection of NTRK fusions in all patients, The Japanese (JSCO, JSMO), TOS and ASCO experts although it could be used for patients with tumours that are agreed with and accepted completely the initial recom- highly likely to harbour ETV6-NTRK3 fusions. DNA-based mendation (supplementary Table S9, available at Annals of NGS, on the other hand, is effective for the detection of Oncology online) in the pre-meeting survey, but the ESMO NTRK fusions.52,54 Although, not all the NTRK fusions can be experts thought that the recommendation should only identified, especially those involving NTRK2 and NTRK3 apply to patients with tumours known to frequently where large intronic regions can render DNA-based detec- harbour NTRK fusions for whom there was no other effec- tion challenging. RNA sequencing does however85,87 offer tive first-line treatment. In the case of tumours with an

868 https://doi.org/10.1016/j.annonc.2020.03.299 Volume 31 - Issue 7 - 2020 T. Yoshino et al. Annals of Oncology alternative effective first-line treatment option and an NTRK Thus, these recommendations can be considered to be fusion, some physicians may opt for the use of TRK in- international expert consensus recommendations for the hibitors in later line settings. ‘Recommendation CQ6’ was treatment of patients with either MSI/dMMR tumours or reworded to reflect this and the GoR revised to A, and all solid tumours with NTRK fusions. The ESMO Magnitude of the experts accepted [A ¼ 100%] the revision. Currently, Clinical Benefit Scale (MCBS) score for pembrolizumab and despite the efficacy of TRK inhibitors, including in the first- TRK inhibitors in the agnostic therapy setting have not been line setting, there is no study comparing a TRK inhibitor confirmed, but the preliminary scores are 3 for both, the with standard of care for patients with NTRK fusion-positive highest score attainable for efficacy evaluated on single-trial solid tumours. data. As the numbers of clinically relevant predictive bio- Implications of prevalence of MSI and NTRK fusions in markers for the treatment of solid tumours increases, it is adult and paediatric tumours on recommendations for likely that NGS will become the key diagnostic tool to testing inform our treatment decisions. Genomic profiling of tu- mours to identify other potentially targetable alterations These recommendations, particularly those developed in (such as ALK, BRAF, BRCAness, FGFR, HER2, HER3, homol- response to the CQs1 above for testing patients for both ogous recombination deficiency (HRD), KRAS, RET, ROS1 and MSI/dMMR and known/likely NTRK fusions are made in the TMB-high), which can be used in tumour-agnostic treat- knowledge that the prevalence of MSI/dMMR is low in ment approaches, is ongoing. Thus, the era of focussing on a most common solid tumours and the prevalence of known/ tumour’s molecular biology has arrived and will alter our likely NTRK fusions in most common tumour types is approach to future drug development. extremely low. We investigated the prevalence of MSI, NTRK rearrangements and high TMB (>20 mutations/Mb) in solid ACKNOWLEDGEMENTS tumours from adult (age >18 years) and paediatric (age The authors would like to thank the JSCO staff, Y.Yamamoto <18 years) patients. Comprehensive genomic profiling of and the ESMO Scientific Coordinator, Ms K. Marinoni, for >300 cancer-related genes was carried out by Foundation their work in the preparation for the meeting, and Drs M. Medicine (Cambridge, USA) as previously described in 90,91 Futamura, K. Kurimoto, N. Matsuhashi and T. Takahashi for detail. Analysis was carried out on 217 086 samples across different solid tumour types, which already had their their on-site assistance and support as JSCO observers. The MSI status and TMB score determined92,93 (supplementary authors would like to acknowledge the voluntary contri- Tables S12 and S13, available at Annals of Oncology on- butions from both Dr R. Dienstmannn MD of the Vall ’ line). To avoid overestimation of prevalence in rare cancers, d Hebron Institute of Oncology (VHIO) who released to us the figures were reported only for those tumour types with the GENIE dataset, and Foundation Medicine (FMI, Cam- data for >500 adult patients and >100 paediatric patients. bridge, USA) in analysing the prevalence of NTRK fusions, These data support the low prevalence of MSI and MSI and TMB-H status in common solid tumours. Dr A. known/likely NTRK fusions in common tumours and show Kinsella, Cancer Communications and Consultancy Ltd, that MSI is more prevalent in adult (as high as 15.09% in Knutsford, Cheshire, UK, is acknowledged for her assistance endometrial tumours, 1.65% overall in 212 704 adult pro- in the preparation of the manuscript funded by JSCO. files) than in paediatric solid tumours (as high as 0.84% in kidney tumours, 0.23% overall in 4382 paediatric profiles) FUNDING and that conversely known/likely NTRK fusions are more All costs relating to this consensus conference were covered prevalent in paediatric (as high as 4.7% in soft tissue sar- by the JSCO from central dedicated funds. There was no comas, 1.10% overall in 4382 paediatric profiles) than in external funding of the event or the manuscript production. adult (highest at 2.49% in salivary gland tumours, 0.20% overall in 212 704 adult profiles) tumours. The percentage of DISCLOSURE patients with a high TMB was much higher than for either EB has received research funding from Taiho, Chugai, MSI or known/likely NTRK rearrangements in adult tumours Astellas, Merck biopharma, Daiichi Sankyo, Ono, Kyowa- (as high as 54.60% in skin tumours, 6.32% overall in 212 704 Kirin and Takeda; HB has received fees for consultancy/ adult profiles) but was low in pediatric patients (maximum advisory roles paid to his institution from Mersana, Astra- 2.25% in gliomas, 0.91% overall in 4382 pediatric profiles). Zeneca, FORMA therapeutics, Janssen, Novartis, Roche/ Genentech, MedImmune, BMS, Celgene, Incyte, Boehringer CONCLUSION Ingelheim, Eisai and Tolero Pharmaceuticals, and research The results of the voting by the experts from Asia, Europe funding paid to his institution from AstraZeneca, Novartis, and the United States, both before (supplementary MedImmune, BMS, Celgene, Incyte, Janssen, Roche/Gen- Tables S8 and S9, available at Annals of Oncology online) entech, MacroGenics, Boehringer Ingelheim, Lilly, Seattle and after (Tables 2 and 3) the face-to-face meeting, showed Genetics, Merck, Agios, Jounce Therapeutics, Moderna high concordance across the different geographical regions Therapeutics, CytomX Therapeutics, GlaxoSmithKline, Vera- for the testing for, and treatment of, patients with either stem, Tesaro, Immunocore, Takeda, Millennium, Biomed MSI/dMMR tumours or solid tumours with NTRK fusions. Valley Discoveries, TG therapeutics, eFFECTOR Therapeutics,

Volume 31 - Issue 7 - 2020 https://doi.org/10.1016/j.annonc.2020.03.299 869 Annals of Oncology T. Yoshino et al.

Gilead Sciences, BioAtla, CicloMed, Loxo, Vertex, Harpoon advisory roles from Amgen, Boehringer Ingelheim, Bayer, Therapeutics, Jiangsu Hengrui Medicine, Arch, Kyocera, BMS, MSD, Merck Serono, Eli Lilly, Ono and Takeda; KY has Arvinas and Revolution Medicines; AC has received fees for received fees for consultancy/advisory roles from Abbott, consultancy/advisory roles from Merck Serono, Roche, AbbVie, Asa hi Kasei Pharma, Astellas, Biogen Japan, Cel- BeiGene, Bayer, Servier, Eli Lilly, Novartis, Takeda, Astellas gene, Chugai, Covidien Japan, Daiichi Sankyo, Eisai, Eli Lilly and Pierre Fabre and research funding from Genentech, Japan, GlaxoSmithKline, Johnson & Johnson, KCI, Kyowa Merck Serono, Roche, BeiGene, Bayer, Servier, Eli Lilly, Kirin, Meiji Seika Pharma, Merck Serono, MSD, Nippon Novartis, Takeda, Astellas, FibroGen, Amcure, Sierra Kayaku, Novartis, Ono Pharm., Otsuka Pharm., Sanofi, Taiho Oncology, AstraZeneca, Medimmune, BMS and MSD; FC has Pharm., Toray Medical, Tsumura and Yakult Honsha; TY has received fees for consultancy/advisory roles from Phillips; L- received research funding from Novartis Pharma K.K., MSD TC has received research funding from Novartis, Merck K.K., Sumitomo Dainippon Pharma Co., Ltd, Chugai, Sanofi Serono, TTY, Polaris, SyncorePharm, Pfizer, and BMS, hono- K.K., Daiichi Sankyo, Parexel International Inc., Ono, Glax- raria from ONO, Eli Lilly, MSD, Pharma Engine, TTY, Syn- oSmithKline K.K. and Boehringer Ingelheim Japan. JYD and corePharm, Novartis, AstraZeneca and Ipsen, patents and SM declare no conflicts of interests. royalties for ENO-1 mAb from HuniLife, and is a Scientific Advisory Board member at Pharma Engine and a board member at Sinopharm Taiwan, Ltd; YK has received fees for REFERENCES consultancy/advisory roles from Ono Asahi Kasei and BMS 1. Buchbinder EI, Desai A. CTLA-4 and PD-1 pathways: similarities, dif- research funding from Taiho, Chugai, Yakult, Daiichi-Sankyo, ferences, and implications of their inhibition. Am J Clin Oncol. 2016;39: Merck Serono, Asahi Kasei, EA Pharma, Otsuka Pharma- 98e106. 2. Hui E, Cheung J, Zhu J, et al. T cell costimulatory receptor CD28 is a pri- ceutical Co., Ltd, Otsuka Pharmaceutical Factory Inc., mary target for PD-1-mediated inhibition. Science. 2017;355:1428e1433. Takeda, Shionogi, Kaken Pharmaceuticals, Kowa Pharma- 3. Krueger J, Rudd CE. Two strings in one bow: PD-1 negatively regulates ceuticals, Astellas, Medicon, Dainippon Sumitomo Pharma- via co-receptor CD28 on t cells. Immunity. 2017;46:529e531. ceuticals, Taisho Toyama Pharmaceuticals, Kyowa Kirin, 4. Le DT, Uram JN, Wang H, et al. PD-1 blockade in tumors with Pfizer Japan, Ono, NIHON, Japan Blood Products Organiza- mismatch-repair deficiency. N Engl J Med. 2015;372:2509e2520. fi 5. Marcus L, Lemery SJ, Keegan P, Pazdur R. FDA approval summary: tion, Medtronic Japan, Sano K.K., and grants from Eisai, pembrolizumab for the treatment of microsatellite instability-high Tsumura, KCI Licensing, Inc, Abbott Japan, Fuji Film and solid tumors. Clin Cancer Res. 2019;25:3753e3758. Toyama Chemical Co.; YKo has receive research funding 6. FDA. FDA approves keytruda (pembrolizumab) as first cancer treat- from Taiho, Chugai, Takeda, MSD, Nihon Kayaku, Yakult, Lilly ment for any solid tumor with a specific genetic feature. Available at Japan, Ono, EA Pharma, Novartis, Daiichi-Sankyo, BMS and https://www.drugs.com/newdrugs/fda-approves-keytruda-pembrolizu fi fi fi mab- rst-cancer-any-solid-tumor-speci c-genetic-feature-4538.htm. Sano ; YN has received fees for consultancy/advisory roles Accessed April 17, 2020. from Eli Lilly, AstraZeneca, Chugai, Pfizer, Novartis, Eisai, 7. Merck. Merk’s KEYTRUDA® (pembrolizumab) recieves five new approvals Bayer, Fuji Film Toyama Chemistry, Shionogi, Taiho, Ono, in Japan, including in advanced non-small cell lung cancer (NSCLC), as Guardent Health, Kyowa Kirin and Mundipharma; MJO has adjuvant therapy for melanoma, and in advanced microsatellite received fees for consultancy/advisory roles from Janssen instability-high (MSI-H) tumors. Available at https://www.mrknewsroom. com/news-release/oncology/mercks-keytruda-pembrolizumab-receives Research and Development LLC, AgilVax, Takeda Pharma- -five-new-approvals-japan-including-adva. Accessed April 17, 2020. ceuticals (Japan), Acrotech Biopharma, Promega, Genentech 8. Le DT, Durham JN, Smith KN, et al. Mismatch repair deficiency predicts Inc., and Novartis Pharmaceuticals and research funding response of solid tumors to PD-1 blockade. Science. 2017;357:409e413. from Roche, BMS, Merck, AstraZeneca and Nouscom; GP 9. Zhao P, Li L, Jiang X, Li Q. Mismatch repair deficiency/microsatellite has received fees for consultancy/advisory roles from instability-high as a predictor for anti-PD-1/PD-L1 immunotherapy ef- ficacy. J Hematol Oncol. 2019;12:54. Roche, Merck and Amgen and research funding from: 10. Overman MJ, McDermott R, Leach JL, et al. Nivolumab in patients with Roche, Amgen, Novartis, MSD, BMS, Pfizer, Boehringer and metastatic DNA mismatch repair-deficient or microsatellite instability- Astra Zeneca; AS has received fees for consultancy from high colorectal cancer (CheckMate 142): an open-label, multicentre, Genentech, AstraZeneca and Medtronic and for advisory phase 2 study. Lancet Oncol. 2017;18:1182e1191. boards from AstraZeneca and Takeda; JT has received fees 11. Overman MJ, Lonardi S, Wong KYM, et al. Durable clinical benefit with nivolumab plus ipilimumab in DNA mismatch repair-deficient/ for consultancy/advisory roles from Array Biopharma, microsatellite instability-high metastatic colorectal cancer. J Clin AstraZeneca, Bayer, BeiGene, Boehringer Ingelheim, Chugai, Oncol. 2018;36:773e779. Genentech, Genmab A/S, Halozyme, Imugene Limited, In- 12. Drilon A, Laetsch TW,Kummar S, et al. Efficacyof larotrectinib in TRK fusion- flection Biosciences Limited, Ipsen, Kura Oncology, Eli Lilly, positive cancers in adults and children. N Engl J Med. 2018;378:731e739. MSD, Menarini, Merck Serono, Merrimack, Merus, Molec- 13. Laetsch TW, DuBois SG, Mascarenhas L, et al. Larotrectinib for paediatric fi solid tumours harbouring NTRK gene fusions: phase 1 results from a ular Partners, Novartis, Peptomyc, P zer, Pharmacyclics, multicentre, open-label, phase 1/2 study. Lancet Oncol. 2018;19:705e714. ProteoDesign SL, Rafael Pharmaceuticals, F. Hoffmann-La 14. Rolfo C, Ruiz R, Giovannetti E, et al. Entrectinib: a potent new TRK, ROS1, Roche Ltd, Sanofi, SeaGen, Seattle Genetics, Servier, and ALK inhibitor. Expert Opin Investig Drugs. 2015;24:1493e1500. Symphogen, Taiho, VCN Biosciences, Biocartis, Foundation 15. Demetri GD, Paz-Ares L, Farago AF, et al. Efficacy and safety of entrec- Medicine, HalioDx SAS and Roche Diagnostics; MT has tinib in patients with NTRK fusion positive tumours: pooled analysis of STARTRK-2, STARTRK-! and ALKA-372-001. Ann Oncol. 2018;29:ix175. received fees for consultancy/advisory roles from Chugai; 16. Luchini C, Bibeau F, Ligtenberg MJL, et al. ESMO recommendations on HT has received research funding from Sysmex, Takeda and microsatellite instability testing for immunotherapy in cancer, and its Daiichi-Sankyo; KHY has received fees for consultancy/ relationship with PD-1/PD-L1 expression and tumour mutational

870 https://doi.org/10.1016/j.annonc.2020.03.299 Volume 31 - Issue 7 - 2020 T. Yoshino et al. Annals of Oncology

burden: a systematic review-based approach. Ann Oncol. 2019;30(8): 37. Cabel L, Proudhon C, Romano E, et al. Clinical potential of circulating 1232e1243. tumour DNA in patients receiving anticancer immunotherapy. Nat Rev 17. Marchio C, Scaltriti M, Ladanyi M, et al. ESMO recommendations on Clin Oncol. 2018;15:639e650. the standard methods to detect NTRK fusions in daily practice and 38. Riaz N, Havel JJ, Makarov V, et al. Tumor and microenvironment evo- clinical research. Ann Oncol. 2019;30:1417e1427. lution during immunotherapy with nivolumab. Cell. 2017;171:934e949 18. Mishima S, Taniguchi H, Akagi K, et al. Japan Society of Clinical e916. Oncology provisional clinical opinion for the diagnosis and use of 39. Bao F, Panarelli NC, Rennert H, et al. Neoadjuvant therapy induces loss immunotherapy in patients with deficient DNA mismatch repair tu- of MSH6 expression in colorectal carcinoma. Am J Surg Pathol. mors, cooperated by Japanese Society of Medical Oncology, First Edi- 2010;34:1798e1804. tion. Int J Clin Oncol. 2019;25(2):201e239. 40. Watanabe Y, Koi M, Hemmi H, et al. A change in microsatellite insta- 19. Dykewicz CA, Centers for Disease Control and Prevention, Infectious bility caused by cisplatin-based chemotherapy of ovarian cancer. Br J Diseases Society of America, American Society of Blood and Marrow Cancer. 2001;85:1064e1069. Transplantation. Summary of the guidelines for preventing opportu- 41. Goldstein J, Wu W, Borras E, et al. Can microsatellite status of colo- nistic infections among hematopoietic stem cell transplant recipients. rectal cancer be reliably assessed after neoadjuvant therapy. Clin Clin Infect Dis. 2001;33:139e144. Cancer Res. 2017;23:5246e5254. 20. Ribic CM, Sargent DJ, Moore MJ, et al. Tumor microsatellite-instability 42. Marabelle A, Fakih MG, Lopez J, et al. Association of tumor mutational status as a predictor of benefit from fluorouracil-based adjuvant burden with outcomes in patients with select advanced solid tumors chemotherapy for colon cancer. N Engl J Med. 2003;349:247e257. treated with pembrolizumab in KEYNOTE-158. Ann Oncol. 21. Hause RJ, Pritchard CC, Shendure J, Salipante SJ. Classification and 2019;30(suppl 5):v475ev532. characterization of microsatellite instability across 18 cancer types. Nat 43. Diaz Jr LA, Le D, Maio M, et al. Pembrolizumab in microsatellite Med. 2016;22:1342e1350. instability high cancers: Updated analysis of phase II KEYNOTE-164 and 22. Latham A, Srinivasan P, Kemel Y, et al. Microsatellite Instability is KEYNOTE-158 studies. Ann Oncol. 2019;30:1170. associated with the presence of Lynch syndrome pan-cancer. J Clin 44. Azad NS, Gray RJ, Overman MJ, et al. Nivolumab is effective in Oncol. 2019;37:286e295. mismatch repair-deficient noncolorectal cancers: results from arm Z1D- 23. Ishida H, Yamaguchi T, Tanakaya K, et al. Japanese Society for Cancer of A subprotocol of the NCI-MATCH (EAY131) study. J Clin Oncol. the Colon and Rectum (JSCCR) Guidelines 2016 for the clinical practice 2020;38(3):201e222. of hereditary colorectal cancer (Translated version). J Anus Rectum 45. Le D, Kavan P, Kim T, et al. Safety and antitumor activity of pem- Colon. 2018;2(Suppl 1):S1eS51. brolizumab in patients with advanced microsatellite instabilityehigh 24. Dudley JC, Lin M-T, Le DT, Eshleman JR. Microsatellite instability as a (MSI-H) colorectal cancer: KEYNOTE-164. Ann Oncol. 2018;29(suppl_5): marker for PD-1 blockade. Clin Cancer Res. 2016;22:813e820. v107. 25. Moreira L, Balaguer F, Lindor N, et al. Identification of Lynch syn- 46. Winer A, Ghatalia P, Bubes N, et al. Dual checkpoint inhibition with drome among patients with colorectal cancer. JAMA. 2012;308: ipilimumab plus nivolumab after progression on sequential PD-1/PDL-1 1555e1565. inhibitors pembrolizumab and atezolizumab in a patient with Lynch 26. Aparicio T, Schischmanoff O, Poupardin C, et al. High prevalence of syndrome, metastatic colon, and localized urothelial cancer. Oncolo- defi cient mismatch repair phenotype and the V600E BRAF mutation in gist. 2019;24:1416e1419. elderly patients with colorectal cancer. J Geriatr Oncol. 2014;5:384e 47. Yan L, Zhang W. Precision medicine becomes reality-tumor type- 388. agnostic therapy. Cancer Commun (Lond). 2018;38:6. 27. Segal N, Wainberg ZA, Overman MJ, et al. Safety and clinical activity of 48. Martin-Zanca D, Hughes SH, Barbacid M. A human oncogene formed durvalumab monotherapy in patients with microsatellite instability- by the fusion of truncated tropomyosin and protein tyrosine kinase high (MSI-H) tumors. J Clin Oncol. 2019;37:670. sequences. Nature. 1986;319:743e748. 28. Hutchins G, Southward K, Handley K, et al. Value of mismatch repair, 49. Pulciani S, Santos E, Lauver AV, et al. Oncogenes in solid human tu- KRAS, and BRAF mutations in predicting recurrence and benefits from mours. Nature. 1982;300:539e542. chemotherapy in colorectal cancer. J Clin Oncol. 2011;29:1261e1270. 50. Brenca M, Rossi S, Polano M, et al. Transcriptome sequencing identifies 29. Sargent DJ, Marsoni S, Monges G, et al. Defective mismatch repair as a ETV6-NTRK3 as a gene fusion involved in GIST. J Pathol. 2016;238:543e predictive marker for lack of efficacy of fluorouracil-based adjuvant 549. therapy in colon cancer. J Clin Oncol. 2010;28:3219e3226. 51. Okamura R, Boichard A, Kato S, et al. Analysis of NTRK alterations in 30. Middha S, Zhang L, Nafa K, et al. Reliable pan-cancer microsatellite pan-cancer adult and pediatric malignancies: implications for NTRK- instability assessment by using targeted next-generation sequencing targeted therapeutics. JCO Precis Oncol. 2018;2018. data. JCO Precis Oncol. 2017;2017. 52. Shi E, Chmielecki J, Tang CM, et al. FGFR1 and NTRK3 actionable al- 31. Hempelmann JA, Lockwood CM, Konnick EQ, et al. Microsatellite terations in “wild-type” gastrointestinal stromal tumors. J Transl Med. instability in prostate cancer by PCR or next-generation sequencing. 2016;14:339. J Immunother Cancer. 2018;6:29. 53. Stransky N, Cerami E, Schalm S, et al. The landscape of kinase fusions in 32. Goel A, Nguyen TP, Leung HC, et al. De novo constitutional MLH1 cancer. Nat Commun. 2014;5:4846. epimutations confer early-onset colorectal cancer in two new sporadic 54. Zehir A, Benayed R, Shah RH, et al. Mutational landscape of metastatic Lynch syndrome cases, with derivation of the epimutation on the cancer revealed from prospective clinical sequencing of 10,000 pa- paternal allele in one. Int J Cancer. 2011;128:869e878. tients. Nat Med. 2017;23:703e713. 33. Nowak JA, Yurgelun MB, Bruce JL, et al. Detection of mismatch repair 55. Hsiao SJ, Zehir A, Sireci AN, Aisner DL. Detection of tumor NTRK gene deficiency and microsatellite instability in colorectal adenocarcinoma fusions to identify patients who may benefit from tyrosine kinase (TRK) by targeted next-generation sequencing. J Mol Diagn. 2017;19:84e91. inhibitor therapy. J Mol Diagn. 2019;21:553e571. 34. Kautto EA, Bonneville R, Miya J, et al. Performance evaluation for rapid 56. Tognon C, Knezevich SR, Huntsman D, et al. Expression of the ETV6- detection of pan-cancer microsatellite instability with MANTIS. Onco- NTRK3 gene fusion as a primary event in human secretory breast target. 2017;8:7452e7463. carcinoma. Cancer Cell. 2002;2:367e376. 35. Niu B, Ye K, Zhang Q, et al. MSIsensor: microsatellite instability 57. Amatu A, Sartore-Bianchi A, Siena S. NTRK gene fusions as novel tar- detection using paired tumor-normal sequence data. Bioinformatics. gets of cancer therapy across multiple tumour types. ESMO Open. 2014;30:1015e1016. 2016;1:e000023. 36. Vanderwalde A, Spetzler D, Xiao N, et al. Microsatellite instability 58. Drilon A, Siena S, Ou SI, et al. Safety and antitumor activity of the status determined by next-generation sequencing and compared with multitargeted pan-TRK, ROS1, and ALK inhibitor entrectinib: combined PD-L1 and tumor mutational burden in 11,348 patients. Cancer Med. results from two phase I trials (ALKA-372-001 and STARTRK-1). Cancer 2018;7:746e756. Discov. 2017;7:400e409.

Volume 31 - Issue 7 - 2020 https://doi.org/10.1016/j.annonc.2020.03.299 871 Annals of Oncology T. Yoshino et al.

59. Drilon A. TRK inhibitors in TRK fusion-positive cancers. Ann Oncol. hitherto undescribed salivary gland tumor entity. Am J Surg Pathol. 2019;30:viii23eviii30. 2010;34:599e608. 60. Solomon JP, Benayed R, Hechtman JF, Ladanyi M. Identifying patients 77. Skalova A, Vanecek T, Simpson RH, et al. Mammary analogue secretory with NTRK fusion cancer. Ann Oncol. 2019;30:viii16eviii22. carcinoma of salivary glands: molecular analysis of 25 ETV6 gene 61. Solomon JP, Hechtman JF. Detection of NTRK fusions: merits and rearranged tumors with lack of detection of classical ETV6-NTRK3 limitations of current diagnostic platforms. Cancer Res. 2019;79:3163e fusion transcript by standard RT-PCR: report of 4 cases harboring ETV6- 3168. X gene fusion. Am J Surg Pathol. 2016;40:3e13. 62. Solomon JP, Linkov I, Rosado A, et al. NTRK fusion detection across 78. Cocco E, Scaltriti M, Drilon A. NTRK fusion-positive cancers and TRK multiple assays and 33,997 cases: diagnostic implications and pitfalls. inhibitor therapy. Nat Rev Clin Oncol. 2018;15:731e747. Mod Pathol. 2020;33(1):38e46. 79. Lassen UN, Albert CM, Kummar S, et al. Larotrectinib efficacy and 63. Gatalica Z, Xiu J, Swensen J, Vranic S. Molecular characterization of safety in TRK fusion cancer: an expanded clinical dataset showing cancers with NTRK gene fusions. Mod Pathol. 2019;32:147e153. consistency in an age and tumor agnostic approach. Ann Oncol. 64. Prasad ML, Vyas M, Horne MJ, et al. NTRK fusion oncogenes in pedi- 2018;29:viii133eviii148. atric papillary thyroid carcinoma in northeast United States. Cancer. 80. Laetsch TW, Hawkins DS. Larotrectinib for the treatment of TRK fusion 2016;122:1097e1107. solid tumors. Expert Rev Anticancer Ther. 2019;19:1e10. 65. Jiao X, Lokker A, Snider J, et al. Co-occurrence of NTRK fusions with 81. DuBois SG, Laetsch TW, Federman N, et al. The use of neoadjuvant other genomic biomarkers in cancer patients. Ann Oncol. larotrectinib in the management of children with locally advanced TRK 2019;30(suppl 5):v25ev54. fusion sarcomas. Cancer. 2018;124:4241e4247. 66. Wilson TR, Sokol ES, Trabucco SE, et al. Genomic characteristics and 82. Hechtman JF, Benayed R, Hyman DM, et al. Pan-Trk immunohisto- predicted ancestry of NTRK1/2/3 and ROS1 fusion-positive tumours chemistry is an efficient and reliable screen for the detection of NTRK from >165,000 pan-solid tumours. Ann Oncol. 2019;30(suppl 5):v159e fusions. Am J Surg Pathol. 2017;41:1547e1551. v193. 83. Lezcano C, Shoushtari AN, Ariyan C, et al. Primary and metastatic 67. Bourgeois JM, Knezevich SR, Mathers JA, Sorensen PH. Molecular melanoma With NTRK fusions. Am J Surg Pathol. 2018;42:1052e1058. detection of the ETV6-NTRK3 gene fusion differentiates congenital 84. Rudzinski ER, Lockwood CM, Stohr BA, et al. Pan-Trk immunohisto- fibrosarcoma from other childhood spindle cell tumors. Am J Surg chemistry identifies NTRK rearrangements in pediatric mesenchymal Pathol. 2000;24:937e946. tumors. Am J Surg Pathol. 2018;42:927e935. 68. Fletcher CDM, Bridge JA, Hogendoorn P, Mertens F. WHO Classifica- 85. Murphy DA, Ely HA, Shoemaker R, et al. Detecting gene rearrange- tion of Tumours of Soft Tissue and Bone. 4th ed. Lyon, France: IARC; ments in patient populations through a 2-step diagnostic test 2013. comprised of rapid IHC enrichment followed by sensitive next- 69. Knezevich SR, McFadden DE, Tao W, et al. A novel ETV6-NTRK3 gene generation sequencing. Appl Immunohistochem Mol Morphol. fusion in congenital fibrosarcoma. Nat Genet. 1998;18:184e187. 2017;25:513e523. 70. Orbach D, Brennan B, De Paoli A, et al. Conservative strategy in in- 86. Milione M, Ardini E, Christiansen J, et al. Identification and charac- fantile fibrosarcoma is possible: the European paediatric Soft tissue terization of a novel SCYL3-NTRK1 rearrangement in a colorectal cancer sarcoma study group experience. Eur J Cancer. 2016;57:1e9. patient. Oncotarget. 2017;8:55353e55360. 71. Rubin BP, Chen CJ, Morgan TW, et al. Congenital mesoblastic nephroma 87. Frattini V, Trifonov V, Chan JM, et al. The integrated landscape of driver t(12;15) is associated with ETV6-NTRK3 gene fusion: cytogenetic and genomic alterations in glioblastoma. Nat Genet. 2013;45:1141e1149. molecular relationship to congenital (infantile) fibrosarcoma. Am J 88. Pfarr N, Kirchner M, Lehmann U, et al. Testing NTRK testing: wet-lab Pathol. 1998;153:1451e1458. and in silico comparison of RNA-based targeted sequencing assays. 72. Del Castillo M, Chibon F, Arnould L, et al. Secretory breast carcinoma: a Genes Chromosomes Cancer. 2020;59(3):178e188. histopathologic and genomic spectrum characterized by a joint specific 89. Doebele RC, Drilon A, Paz-Ares L, et al. Entrectinib in patients with ETV6-NTRK3 gene fusion. Am J Surg Pathol. 2015;39:1458e1467. advanced or metastatic NTRK fusion-positive solid tumours: integrated 73. Makretsov N, He M, Hayes M, et al. A fluorescence in situ hybridization analysis of three phase 1-2 trials. Lancet Oncol. 2020;21(2):271e282. study of ETV6-NTRK3 fusion gene in secretory breast carcinoma. Genes 90. Frampton GM, Fichtenholtz A, Otto GA, et al. Development and vali- Chromosomes Cancer. 2004;40:152e157. dation of a clinical cancer genomic profiling test based on massively 74. Bishop JA, Yonescu R, Batista D, et al. Utility of mammaglobin immu- parallel DNA sequencing. Nat Biotechnol. 2013;31:1023e1031. nohistochemistry as a proxy marker for the ETV6-NTRK3 translocation 91. He J, Abdel-Wahab O, Nahas MK, et al. Integrated genomic DNA/RNA in the diagnosis of salivary mammary analogue secretory carcinoma. profiling of hematologic malignancies in the clinical setting. Blood. Hum Pathol. 2013;44:1982e1988. 2016;127:3004e3014. 75. Boon E, Valstar MH, van der Graaf WTA, et al. Clinicopathological 92. Chalmers ZR, Connelly CF, Fabrizio D, et al. Analysis of 100,000 human characteristics and outcome of 31 patients with ETV6-NTRK3 fusion cancer genomes reveals the landscape of tumor mutational burden. gene confirmed (mammary analogue) secretory carcinoma of salivary Genome Med. 2017;9:34. glands. Oral Oncol. 2018;82:29e33. 93. Hartmaier RJ, Albacker LA, Chmielecki J, et al. High-throughput 76. Skalova A, Vanecek T, Sima R, et al. Mammary analogue secretory genomic profiling of adult solid tumors reveals novel insights into carcinoma of salivary glands, containing the ETV6-NTRK3 fusion gene: a cancer pathogenesis. Cancer Res. 2017;77:2464e2475.

872 https://doi.org/10.1016/j.annonc.2020.03.299 Volume 31 - Issue 7 - 2020 REVIEW

Recommendations for the use of next-generation sequencing (NGS) for patients with metastatic cancers: a report from the ESMO Precision Medicine Working Group

F. Mosele1, J. Remon2, J. Mateo3, C. B. Westphalen4, F. Barlesi1, M. P. Lolkema5, N. Normanno6, A. Scarpa7, M. Robson8, F. Meric-Bernstam9, N. Wagle10, A. Stenzinger11, J. Bonastre12,13, A. Bayle1,12,13, S. Michiels12,13, I. Bièche14, E. Rouleau15, S. Jezdic16, J-Y. Douillard16, J. S. Reis-Filho17, R. Dienstmann18 & F. André1,19,20*

1Department of Medical Oncology, Gustave Roussy, Villejuif, France; 2Department of Medical Oncology, Centro Integral Oncológico Clara Campal (HM-CIOCC), Hospital HM Delfos, HM Hospitales, Barcelona; 3Clinical Research Program, Vall Hebron Institute of Oncology (VHIO) and Vall d’Hebron University Hospital, Barcelona, Spain; 4Comprehensive Cancer Center Munich and Department of Medicine III, University Hospital, LMU Munich, Munich, Germany; 5Department of Medical Oncology, Erasmus MC Cancer Center, Rotterdam, the Netherlands; 6Cell Biology and Biotherapy Unit, Istituto Nazionale Tumori, ‘Fondazione G. Pascale’ e IRCCS, Naples; 7ARC- Net Research Centre and Department of Diagnostics and Public Health e Section of Pathology, University of Verona, Verona, Italy; 8Breast Medicine and Clinical Genetics Services, Department of Medicine, Memorial Sloan Kettering Cancer Center, New York; 9Department of Investigational Cancer Therapeutics, The University of Texas MD Anderson Cancer Center, Houston; 10Department of Medical Oncology, Dana-Farber Cancer Institute and Harvard Medical School, Boston, USA; 11Institute of Pathology, University Hospital Heidelberg, Heidelberg, Germany; 12Department of Biostatistics and Epidemiology, Gustave Roussy, University Paris-Saclay, Villejuif; 13Oncostat U1018, Inserm, University Paris-Saclay, labeled Ligue Contre le Cancer, Villejuif; 14Department of Genetics, Institut Curie, Paris Descartes University, Paris; 15Cancer Genetic Laboratories, Department of Medical Biology and Pathology, Gustave Roussy Cancer Campus, Villejuif, France; 16Scientific and Medical Division, European Society for Medical Oncology, Lugano, Switzerland; 17Department of Pathology, Memorial Sloan Kettering Cancer Center, New York, USA; 18Oncology Data Science Group, Molecular Prescreening Program, Vall dHebron Institute of Oncology, Barcelona, Spain; 19Inserm, Gustave Roussy Cancer Campus, UMR981, Villejuif; 20Paris Saclay University, Orsay, France

Available online 24 August 2020

Next-generation sequencing (NGS) allows sequencing of a high number of nucleotides in a short time frame at an affordable cost. While this technology has been widely implemented, there are no recommendations from scientific societies about its use in oncology practice. The European Society for Medical Oncology (ESMO) is proposing three levels of recommendations for the use of NGS. Based on the current evidence, ESMO recommends routine use of NGS on tumour samples in advanced non-squamous non-small-cell lung cancer (NSCLC), prostate cancers, ovarian cancers and cholangiocarcinoma. In these tumours, large multigene panels could be used if they add acceptable extra cost compared with small panels. In colon cancers, NGS could be an alternative to PCR. In addition, based on the KN158 trial and considering that patients with endometrial and small-cell lung cancers should have broad access to anti-programmed cell death 1 (anti-PD1) antibodies, it is recommended to test tumour mutational burden (TMB) in cervical cancers, well- and moderately-differentiated neuroendocrine tumours, salivary cancers, thyroid cancers and vulvar cancers, as TMB-high predicted response to pembrolizumab in these cancers. Outside the indications of multigene panels, and considering that the use of large panels of genes could lead to few clinically meaningful responders, ESMO acknowledges that a patient and a doctor could decide together to order a large panel of genes, pending no extra cost for the public health care system and if the patient is informed about the low likelihood of benefit. ESMO recommends that the use of off-label drugs matched to genomics is done only if an access programme and a procedure of decision has been developed at the national or regional level. Finally, ESMO recommends that clinical research centres develop multigene sequencing as a tool to screen patients eligible for clinical trials and to accelerate drug development, and prospectively capture the data that could further inform how to optimise the use of this technology. Key words: next-generation sequencing (NGS), genomic alterations, metastatic cancers

INTRODUCTION *Correspondence to: Prof. Fabrice André, ESMO Head Office e Scientific and Medical Division, Via Ginevra 4, Lugano CH-6900, Switzerland. Tel: þ41-91- Next-generation sequencing (NGS) allows sequencing of a þ 973-1999; Fax: 41-91-973-1902 high number of nucleotides in a short time frame and at E-mail: [email protected] (F. André). an affordable cost per patient.1e3 In this document, we 0923-7534/© 2020 European Society for Medical Oncology. Published by will discuss the clinical utility of using NGS as a Elsevier Ltd. All rights reserved.

Volume 31 - Issue 11 - 2020 https://doi.org/10.1016/j.annonc.2020.07.014 1491 Annals of Oncology F. Mosele et al. technology, and how this technology should be used (small versus large panels) in frequent diseases. The rec- What are the actionable alterations observed in ommendations will be done at three levels: from a public a cancer type? health perspective, from the perspective of academic (Breast, lung, colorectal, pancreas, prostate, stomach, clinical research centres and the level of each individual liver, cholangiocarcinoma) patient. NGS has recently moved into the clinics with the aim of sequencing long and complex genes and/or mul- tiple genes per tumour sample, in order to identify driver External review by two experts ESCAT ranking of each alteration and/or targetable alterations. Pioneering studies have and by the panel shown that NGS presents a good analytical validity to detect clonally dominant alterations.4 Basedonthis observation, several companies and academic centres have implemented NGS assays to guide treatment de- Percentage of patients presenting Percentage of patients presenting a level I alteration + number of a level II-IV alteration cisions. While this technology has been widely imple- level I alterations mented, there are no recommendations from scientific societies about their use in daily clinical practice. Several prospective trials have reported outcomes associated Value with the use of multigene sequencing. In the SHIVA trial, (NB to test to access an approved drug matched to genomic alteration) the use of multigene sequencing did not improve outcome in patients with metastatic hard-to-treat cancers 5 in comparison with unmatched therapies. In the single- Recommendation on the use of Recommendation on the use of arm MOSCATO trial, the use of multigene sequencing multigene sequencing in multigene sequencing in daily practice clinical research centers and comparative genomic hybridisation (CGH) arrays was associated with an improved progression-free survival Figure 1. Method to develop recommendation about NGS in daily practice. (PFS) in 30% of patients and an objective response rate ESCAT, ESMO Scale for Clinical Actionability of molecular Targets. (ORR) of 11%.6 Several other studies have consistently reported that ORRs ranged between 10% and 30% in patients whose tumours harboured actionable alter- alteration. ESCAT is a framework that ranks a match be- ations.7e10 One of the major issues with most of the tween drug and genomic alterations, according to their prospective trials testing multigene sequencing is the actionability.14 ESCAT level I means that the match of an exclusion of patients whose tumours present a genomic alteration and a drug has been validated in clinical trials, alteration that matches an approved drug. Aside from and should drive treatment decision in daily practice. large prospective trials, several cases have been reported ESCAT level II means that a drug that matches the to present an outlier sensitivity to a drug given based on alteration has been associated with responses in phase I/ an unforeseen, non-recurrent, somatic genomic alter- II or in retrospective analyses of randomised trials. ESCAT ation.11,12 In the present article, we present the European level III includes alterations that are validated in another Society for Medical Oncology (ESMO) recommendations cancer, but not in the disease-to-treat. ESCAT level IV about whether and how tumour multigene NGS could be includes hypothetically targetable alterations based on used to profile metastatic cancers. preclinical data. ESCAT ranking was generated for each alteration by medical oncologists with an expertise in genomics, then validated by two external experts and by METHOD the Working Group. From the ESCAT ranking and preva- The ESMO Precision Medicine Working Group has set up a lence of alterations for each tumour type, we calculated group of experts in the field of clinical cancer genomics in the number of patients to test with NGS, to identify one order to address the following questions: patient that can be matched to an effective drug in daily Should NGS be used in daily practice? practice (ESCAT level I). The main document reports these If so, should large panels of genes be used? numbers with the hypothesis that NGS has a perfect These questions should be addressed from the perspec- analytical validity, while Supplementary Tables, available tive of public health, academic clinical research centres and at https://doi.org/10.1016/j.annonc.2020.07.014, report from the perspective of the individual patients. these numbers taking a hypothesis of 99% and 95% In order to address these questions, the group devel- sensitivity/specificity.15 We assume that there is no oped the method summarised in Figure 1. The general proven impact in terms of public health of detecting level strategy was to determine whether NGS can substitute IIeIV actionable alterations. Finally, in addition to ESCAT complex or multiple testings. First, all recurrent genomic ranking, the group integrated the results of the KN158 alterations were identified in the eight cancers that are study16 in the recommendations. The KN158 study eval- associated with highest number of deaths in the world.13 uated the efficacy of pembrolizumab single agent ac- The ESMO Scale for Clinical Actionability of molecular cording to tumour mutational burden (TMB) in 10 Targets (ESCAT) ranking was then determined for each different diseases.

1492 https://doi.org/10.1016/j.annonc.2020.07.014 Volume 31 - Issue 11 - 2020 F. Mosele et al. Annals of Oncology

MULTIGENE SEQUENCING: PREREQUISITES FROM THE Table 1. Recommendations and guidelines for the standardisation of TECHNICAL SIDE multigene sequencing

In vitro diagnostic tests, such as NGS assays, can be broadly Society guidelines Author/journal separated into two main categories. On one hand, there are Joint Recommendation of the Roy S, et al. J Mol Diagn. 2018.136 manufactured products (reagents, instruments, kits) which Association for Molecular Pathology have been cleared or approved by the respective authorities and the College of American [e.g. US Food and Drug Administration (FDA)] and are sold Pathologists Canadian College of Medical Hume S, et al. J Med Genet. to clinical laboratories for subsequent use. There are Geneticists 2019.137 numerous instances where there are unmet analytical or College of American Pathologists www.cap.org 2020.138 clinical needs, not uncommonly due to the lack of approved Szymanski J, et al. J Pathol Inform. 2018.139 and commercially available assays; in these cases, Burke W, et al. Curr Protoc Hum laboratory-developed tests (LDTs) are being designed by Genet. 2014.140 141 and deployed for clinical decision-making within a single US FDA Kaul K, et al. J Mol Diag. 2001. IQN Path Deans Z, et al. Virchows Arch. clinical, often academic, laboratory. In the dynamic and fast- 2017.142 moving field of cancer precision medicine and molecular Matthijs G, et al. Eur J Hum 143 pathology, LDTs play a central role as they are often driving Genet. 2015. A Joint Consensus Recommendation of Jennings L, et al. J Mol Diagn. diagnostic innovation at times when no approved options the Association for Molecular Pathology 2017.144 exist. Regardless of the in vitro diagnostic category that is and College of American Pathologists being used in a clinical laboratory, an environment that College of American Pathologists Aziz N, et al. Arch Pathol Lab Med. 2015.145 continuously assures and monitors assay quality and per- FDA, Food and Drug Administration; IQN Path, International Quality Network for formance is critical, as inadequate validation and use of Pathology. assays could place patients at risk. Whilst the assessment of test characteristics and quality assurance schemes are 1. Recommendations for daily practice (ESCAT level I) aim governed by country-specific legislation and different reg- to reflect the impact of the use of tumour multigene ulatory models, technical parameters, including modality of NGS on public health. sequencing, sequencing depth, fraction of on-target reads, 2. Recommendations for clinical research centres aim to alignment quality, read quality, error rates, types of sources determine whether performing multigene sequencing of DNA [ctDNA, frozen, formalin-fixed paraffin-embedded could increase access to innovation, accelerate drug (FFPE)], minimal tumour cell content are essential and development and could therefore be a mission of clin- combined under the umbrella of ‘analytical validity’. Once ical research centres. the analytical validity and the robustness of the assay are 3. Patient-centric recommendations. ascertained, its clinical validity and clinical utility need to be considered. Professional groups have endeavoured to pro- vide guidelines for the standardisation of the parameters of Health economics evidence sequencing, data analysis and interpretation of the findings, From a payer perspective, evidence of the cost-effectiveness and are listed in Table 1. of the use of multigene sequencing in daily practice is In fact, a framework that includes standardised validation weak.18e21 We identified two economic studies in non-small- protocols and reflects the concepts of (i) analytical validity cell lung cancer (NSCLC). The first one has compared the per- (i.e. the ability of a test to accurately measure the analyte of formance of targeted NGS panels with traditional assays in an interest as e.g. defined by the parameters: accuracy, pre- EGFR-mutant predominant population.22 The second one has cision, sensitivity, specificity, positive and negative predic- studied the cost-effectiveness of multigene panel sequencing tive values), (ii) clinical validity (i.e. the accuracy with which compared with single-marker testing.23 These studies suggest a genetic test identifies a particular clinical condition with that multigene sequencing in NSCLC is moderately cost- respect to a diagnostic, prognostic or predictive category) effective. Moreover, implementation of multigene and (iii) clinical utility (i.e. whether the test and any sub- sequencing in daily practice requires investments that have to sequent interventions result in an improved health be considered, especially regarding sequencing and bioinfor- outcome among people with a positive test result and the matics workflows in order to deliver results to clinicians in a risks that occur as a result of the test being carried out) timely manner.24 Finally, from a public health perspective, it should be universally considered and applied. ESMO rec- must also be considered that the results of NGS panels could ommends that genomic reports include the ranking of the lead to recommend expensive drugs outside of their genomic alterations either by ESCAT or OncoKb.17 approved indications.25 There is a need to regulate the vol- umes of NGS procedures at the national level. RECOMMENDATIONS GENOMIC ALTERATIONS IN ADVANCED NON-SQUAMOUS General frame NSCLC CLASSIFIED ACCORDING TO ESCAT Recommendations for NGS (summarised in Table 2) are EGFR mutations represent the first driver alterations iden- done at three levels. tified in advanced non-squamous NSCLC.26 Most of them

Volume 31 - Issue 11 - 2020 https://doi.org/10.1016/j.annonc.2020.07.014 1493 Annals of Oncology F. Mosele et al.

Table 2. Summary recommendations

Tumour types General recommendations for daily practice Recommendation for clinical research Special considerations for patients centres Lung adenocarcinoma Tumour multigene NGS to assess level I It is highly recommended that clinical Using large panels of genes could lead to alterations. Larger panels can be used only on research centres perform multigene few clinically meaningful responders, not the basis of specific agreements with payers sequencing in the context of molecular detected by small panels or standard taking into account the overall cost of the screening programmes in order to increase testings. In this context and outside the strategy (drug includeda) and if they report access to innovative drugs and to speed up diseases where large panels of genes are accurate ranking of alterations. NGS can either clinical research. This is particularly relevant recommended, ESMO acknowledges that a be done on RNA or DNA, if it includes level I in breast, pancreatic and hepatocellular patient and a doctor could decide together fusions in the panel. cancers where level IIeIV alterations are to order a large panel of genes, pending no Squamous cell lung No current indication for tumour multigene numerous. extra cost for the public health care system, cancers NGS and if the patient is informed about the low Breast cancers No current indication for tumour multigene likelihood of benefit. NGS Colon cancers Multigene tumour NGS can be an alternative option to PCR if it does not result in additional cost. Prostate cancers Multigene tumour NGS to assess level I alterations. Larger panels can be used only on the basis of specific agreements with payers taking into account the overall cost of the strategy and if they report accurate ranking of alterations. Gastric cancers No current indication for tumour multigene NGS Pancreatic cancers No current indication for tumour multigene NGS Hepatocellular No current indication for tumour multigene carcinoma NGS Cholangiocarcinoma Multigene tumour NGS could be recommended to assess level I alterations. Larger panels can be used only on the basis of specific agreements with payers taking into account the overall cost of the strategy (drug includeda) and if they report accurate ranking of alterations. RNA-based NGS can be used. Others Tumour multigene NGS can be used in ovarian cancers to determine somatic BRCA1/2 mutations. In this latter case, larger panels can be used only on the basis of specific agreements with payers taking into account the overall cost of the strategy (drug includeda) and if they report accurate ranking of alterations. Large panel NGS can be used in carcinoma of unknown primary. It is recommended to determine TMB in cervical cancer, salivary cancer, thyroid cancers, well-to-moderately differentiated neuroendocrine tumours, vulvar cancer, pending drug access (and in TMB-high endometrial and SCL cancers if anti-PD1 antibody is not available otherwise). anti-PD1, anti-programmed cell death 1; DRUP, drug rediscovery protocol; ESMO, European Society for Medical Oncology; NGS, next-generation sequencing; SCL, small-cell lung cancer; TMB, tumour mutational burden. a ESMO recommends using off-label drugs matched to genomics only if an access programme and a procedure of decision have been developed at the national or regional level, as illustrated by the DRUP programme. are in-frame activating deletions in exon 19 and point patients with exon 20 insertions of EGFR, poziotinib (a se- hotspot activating mutations in exon 21 (L858R), followed lective TKI) presented a limited therapeutic efficacy, also by acquired resistant mutations in exon 20 (T790M). Several evaluated in prospective studies.33,34 Another predictive randomised, phase III trials have shown that EGFR tyrosine biomarker that reaches a high position in the ESCAT is ALK kinase inhibitors (TKIs) improve outcome in patients with fusion. In randomised trials, anaplastic lymphoma kinase EGFR-mutated NSCLC.27e30 Based on these data, these (ALK) inhibitors confirmed an improvement of clinical out- specific EGFR mutations reach the highest level in ESCAT. comes across patients with ALK-rearranged NSCLC.35-39 Point mutations or duplications in exons 18e21 (G719X in Some other alterations like MET exon 14 skipping, exon 18, L861Q in exon 21, S768I in exon 20) are unusual BRAFV600E mutations and ROS1 fusions have been identi- EGFR mutations. The efficacies of afatinib and osimertinib fied.40 A significant ORR and clinical meaningful benefit were assessed in prospective, non-randomised trials, have been shown in phase I/II studies in patients with reporting a high ORR and improving PFS.31,32 In addition, in NSCLC with METex14 mutations treated with MET TKIs such

1494 https://doi.org/10.1016/j.annonc.2020.07.014 Volume 31 - Issue 11 - 2020 F. Mosele et al. Annals of Oncology as crizotinib, capmatinib or tepotinib, with BRAFV600E mu- some alterations validated in other tumour types can be tations that received dabrafenib-vemurafenib and with found in patients with NSCLC, but no evidence for drug ROS1 fusions treated with crizotinib, ceritinib or entrecti- efficacy has been reported yet (Table 3A).59e63 In Table 3B, nib.41e47 No randomised trials were developed for these we have described the main molecular variations classified aberrations. Based on these results, crizotinib obtained the by ESCAT in advanced squamous NSCLC. Breakthrough Designation from the FDA for METex14- Summary of recommendations. It is recommended that mutated NSCLC, entrectinib for ROS1-positive NSCLC by the a tumour (or plasma) sample from a patient with FDA and dabrafenib-vemurafenib was approved for NSCLC advanced non-squamous NSCLC is profiled using NGS with BRAFV600E mutation by both the FDA and the European technology, in order to detect level I alterations. Consid- Medicines Agency (EMA). Fusions involving neurotrophic ering the high frequency of fusions, RNA-based NGS, or tyrosine receptor kinase genes (NTRK1-3) occur with a low DNA-based NGS designed to capture such fusions, are the prevalence across different cancer types. Tropomyosin re- preferred options. There is no evidence that panels ceptor kinase (TRK) inhibitors (larotrectinib, entrectinib) detecting genes with a lower level of evidence brings have demonstrated durable responses in NTRK fusion- additional value from a public health perspective. They positive tumours including NSCLC,48e50 leading to agnostic could be used only if the report ranks genomic alterations drug approvals by the EMA and FDA. In addition, LOXO-292 according to valid ranking systems (e.g. ESCAT, OncoKB) showed efficacy in phase I/II studies in patients with RET and on the basis of specific agreements with payers taking fusion-positive NSCLC, receiving the FDA Breakthrough into account the overall cost of the strategy (including off- Designation.51 Several other drivers with therapeutic po- label use of drugs) as compared with small panels. tential have been identified including MET amplifications, Regarding this latter point, ESMO does not recommend KRASG12C mutations (AMG510) and ERBB2 mutations and the use of off-label drugs matched to genomic alterations, amplifications.52e57 Although it has been suggested that except if an access programme and a procedure of decision TMB-high (10 mut/Mb) could be a potential predictive has been developed at the national or regional level, as biomarker for immune checkpoint inhibitors (ICIs), this data illustrated by the drug rediscovery protocol programme.64 is not mature enough to drive decisions in NSCLC.58 Finally, It is recommended that hospitals that run drug

Table 3A. List of genomic alterations level I/II/III according to ESCAT in advanced non-squamous non-small-cell lung cancer (NSCLC)

Gene Alteration Prevalence ESCAT References EGFR Common mutations (Del19, L858R) 15% (50%e60% Asian) IA Midha A, et al. Am J Cancer Res. 201526 Acquired T790M exon 20 60% of EGFR mutant IA Mok T, et al. J Clin Oncol. 201827 Uncommon EGFR mutations (G719X in exon NSCLC IB Soria J-C, et al. N Engl J Med. 201828 18, L861Q in exon 21, S768I in exon 20) 10% IIB Ramalingam S, et al. N Engl J Med. 202029 Exon 20 insertions 2% Mok T, et al. N Engl J Med. 201730 Yang JC-H, et al. Lancet Oncol. 201531 Cho J, et al. J Thorac Oncol. 201832 Cardona A, et al. Lung Cancer. 201833 Heymach J, et al. J Thorac Oncol. 201834 ALK Fusions (mutations as mechanism of resistance) 5% IA Solomon B, et al. J Clin Oncol. 201835 Soria J-C, et al. Lancet. 201736 Peters S, et al. N Engl J Med. 201737 Zhou C, et al. Ann Oncol. 201838 Camidge D, et al. N Engl J Med. 201839 MET Mutations ex 14 skipping 3% IB Tong J, et al. Clin Cancer Res. 201640 Drilon A, et al. Nat Med. 202041 Focal amplifications (acquired resistance 3% IIB Camidge D, et al. J Clin Oncol. 201852 on EGFR TKI in EGFR-mutant tumours) BRAFV600E Mutations 2% IB Planchard D, et al. Lancet Oncol. 201642 Planchard D, et al. Lancet Oncol. 201743 Planchard D, et al. J Clin Oncol. 201744 ROS1 Fusions (mutations as mechanism 1%e2% IB Shaw A, et al. N Engl J Med. 201445 of resistance) Shaw A, et al. Ann Oncol. 201946 Drilon A, et al. Lancet Oncol. 202047 NTRK Fusions 0.23%e3% IC Drilon A, et al. N Engl J Med. 201848 Hong D, et al. Lancet Oncol. 202049 Doebele RC, et al. Lancet Oncol. 202050 RET Fusions 1%e2% IC Drilon A, et. J Thorac Oncol. 201951 KRASG12C Mutations 12% IIB Barlesi F, et al. Lancet. 201653 Fakih M, et al. J Clin Oncol. 201954 ERBB2 Hotspot mutations 2%e5% IIB Hyman D, et al. Nature. 201855 Amplifications Wang Y, et al. Ann Oncol. 201856 Tsurutani J, et al. J Thorac Oncol. 201857 BRCA 1/2 Mutations 1.2% IIIA Balasubramaniam S, et al. Clin Cancer Res. 201763 PIK3CA Hotspot mutations 1.2%e7% IIIA Cancer Genome Atlas Research Network. Nature. 201460 Vansteenkiste J, et al. J Thorac Oncol. 201562 NRG1 Fusions 1.7% IIIB Duruisseaux M, et al. J Clin Oncol. 201959

Volume 31 - Issue 11 - 2020 https://doi.org/10.1016/j.annonc.2020.07.014 1495 Annals of Oncology F. Mosele et al.

Table 3B. List of genomic alterations level I/II/III according to ESCAT in advanced squamous NSCLC

Gene Alteration Prevalence ESCAT References NTRK Fusions 0.23%e3% IC Drilon A, et al. N Engl J Med. 201848 Hong D, et al. Lancet Oncol. 202049 Doebele RC, et al. Lancet Oncol. 202050 PIK3CA Hotspot mutations 16% IIIA Cancer Genome Atlas Research Network, Nature. 201261 Vansteenkiste J, et al. J Thorac Oncol. 201562 BRCA 1/2 Mutations 1.2% IIIA Balasubramaniam S, et al. Clin Cancer Res. 201763 ESCAT, European Society for Medical Oncology (ESMO) Scale for Clinical Actionability of molecular Targets. development programmes and clinical trials run multigene accurately done by immunohistochemistry (IHC) in the sequencing in the context of molecular screening pro- local centre, there is currently no need to perform tumour grammes, since lung cancer presents some level IIeIV multigene NGS for patients with mBC in the context of alterations. daily practice. From the perspective of clinical research centres, and considering the high number of level II al- GENOMIC ALTERATIONS IN METASTATIC BREAST CANCER terations, it is important to include mBC patients in mo- CLASSIFIED ACCORDING TO ESCAT lecular screening programmes and include them in trials ERBB2 amplifications are predictive of clinical benefitof testing targeted therapies matched to genomic alterations E17K anti-HER2 therapies, which yield an improvement of overall (AKT1 , PTEN, ERBB2 mutations, ESR1 and NF1 survival (OS) and PFS,65e69 while neratinib (an irreversible mutations). pan-HER TKI) has been associated with responses in pa- tients with ERBB2 mutations.55 Phase III studies reported a GENOMIC ALTERATIONS IN METASTATIC COLORECTAL significant improvement of PFS with poly ADP ribose CANCER CLASSIFIED ACCORDING TO ESCAT polymerase inhibitors (PARPi) in patients with germline Pivotal randomised trials and meta-analysis highlighted that BRCA1/2-mutated metastatic breast cancer (mBC).70,71 It is hotspot RAS mutations (K-RAS and N-RAS) predict resis- currently estimated that somatic multigene sequencing tance to EGFR monoclonal antibodies (mAbs) in the meta- cannot substitute germline testing for BRCA1/2 status. static setting.79e81 https://doi.org/10.1093/annonc/mdw235. Alpelisib, an a-selective phosphatidylinositol 3-kinase (PI3K) The addition of encorafenib (a BRAF inhibitor) to cetuximab inhibitor, improves PFS in patients with HRþ/HER2 mBC was associated with a significant survival benefit in a recent that harbours PIK3CA hotspot mutations, and is approved phase III trial in patients presenting a BRAFV600E mutation.82 in this group of patients.72 Drugs targeting rare alterations Alterations in mismatch repair proteins (MLH1, MSH2, found in different solid tumours, like microsatellite MSH6 and PMS2) can be identified by IHC and MSI-H instability-high (MSI-H) and NTRK fusions, obtained ap- by PCR to detect smaller length DNA fragments. Testing provals across tumour types.50,73 Nevertheless, NTRK for MSI-H is of great clinical interest in metastatic colorectal fusions highly correlate with secretory phenotype and MSI- cancer (mCRC) because it predicts the efficacy of pem- high tumours are enriched in triple-negative breast cancers brolizumab and nivolumab in this setting.83,84 As mentioned (TNBCs), where anti-PDL1 antibodies are approved. ESR1 before, TRK inhibitors showed high efficacy in multi- mutations occur in around 20% of patients previously histology trials in NTRK fusion-positive tumours50,85; and treated with aromatase inhibitors and are associated mCRC with ERBB2 amplifications/overexpression (detected with response to selective estrogen receptor degraders.74 with FISH or IHC) presented significant responses with dual Nevertheless, these data are preliminary and cannot be HER2 therapy in prospective studies.86,87 In Table 5 we used in daily practice. Other promising targets in mBC are mention the main driver alterations categorised according phosphatase and tensin homologue (PTEN) loss of function to ESCAT, including those with a lack of clinical data in mutations and/or homozygous deletions (TNBCs) and mCRC, but with impact in other tumours.76,88e94 AKT1E17K mutations, which in retrospective and prospective Summary of recommendations. Since most level I al- analyses, respectively, showed a clinical benefit and terations are hotspot mutations in KRAS, NRAS and BRAF, increased responsiveness to AKT inhibitors. Nevertheless, and considering that MSI status is determined by IHC or no results are available from practice changing trials PCR, there is no need to test samples using multigene NGS yet.75,76 In addition, NF1 mutations were identified as a in the context of daily practice. Nevertheless, multigene mechanism of endocrine resistance, but there is no targeted NGS can be an alternative to PCR tests only if it does not therapy available yet in this genomic segment.77 Lastly, generate extra cost compared with standard techniques there are some alterations with no major impact in mBC already implemented in routine. This would allow detec- that are validated in other malignances (Table 4).55,63,78 tion of ERBB2 amplifications, and, in some panels, detect Summary of recommendations. Considering that so- MSI status with high accuracy. If large panel NGS is carried matic sequencing cannot fully substitute germline BRCA out, it should include detection of NTRK fusions. As for testing, that PIK3CA status can be determined by PCR on mBC patients, patients with mCRC can present oncogenic the three hotspots and pending that HER2 testing is alterations for which drugs are being developed and it is

1496 https://doi.org/10.1016/j.annonc.2020.07.014 Volume 31 - Issue 11 - 2020 F. Mosele et al. Annals of Oncology

Table 4. List of genomic alterations level I/II/III according to ESCAT in Table 5. List of genomic alterations level I/II/III according to ESCAT in metastatic breast cancer (mBC) metastatic colorectal cancer (mCRC)

Gene Alteration Prevalence ESCAT References Gene Alteration Prevalence ESCAT References ERBB2 Amplifications 15%e20% IA Slamon D, et al. N Engl J KRAS Mutations 44% Not Van Cutsem E, et al. J Med. 200165 NRAS (resistance 4% applicable Clin Oncol. 201579 Swain S, et al. N Engl J biomarker) Douillard J-Y, et al. N Med. 201566 Engl J Med. 201380 Verma S, et al. N Engl J Sorich M, et al. Ann Med. 201267 Oncol. 201581 Krop I, et al. Lancet BRAFV600E Mutations 8.5% IA https://doi.org/10.1 Oncol. 201468 093/annonc/mdw235 Murthy R, et al. N Engl J Kopetz S, et al. N Engl J Med. 202069 Med. 201982 Hotspot 4% IIB Hyman D, et al. Nature. MSI-H 4%e5% IA Overman M, et al. mutations 201855 Lancet Oncol. 201783 PIK3CA Hotspot 30%e40% IA André F, et al. N Engl J Le DT, et al. J Clin Oncol. mutations Med. 201972 202084 BRCA1/ Germline 4% IA Robson M, et al. N Engl J NTRK1 Fusions 0.5% IC Demetri G, et al. Ann 2 mutations Med. 201770 Oncol. 201885 Litton J, et al. N Engl J Doebele RC, et al. Med. 201871 Lancet Oncol. 202050 Somatic 3% IIIA Balasubramaniam S, ERBB2 Amplifications 2% IIB Meric-Bernstam F, et al. mutations et al. Clin Cancer Res. Lancet Oncol. 201986 201763 Sartore-Bianchi A, et al. MSI-H 1% IC Marcus L, et al. Clin Lancet Oncol. 201687 Cancer Res. 201973 PIK3CA Hotspot 17% IIIA Juric D, et al. J Clin NTRK Fusions 1% IC Doebele RC, et al. Lancet mutations Oncol. 201890 Oncol. 202050 ATM Mutations 5% IIIA Wang C, et al. Transl ESR1 Mutations 10% IIA Fribbens C, et al. J Clin Oncol. 201792 (mechanism of Oncol. 201674 De Bono J, et al. N Engl J resistance) Med. 202093 PTEN Mutations 7% IIA Schmid P, et al. J Clin MET Amplifications 1.7% IIIA https://clinicaltrials. Oncol. 201875 gov/ct2/show/NCT035 AKT1E17K Mutations 5% IIB Hyman D, et al. J Clin 9264194 Oncol. 201776 AKT1E17K Mutations 1% IIIA Hyman D, et al. J Clin NF1 Mutations 6% Not Pearson A, et al. Clin Oncol. 201776 (resistance applicable Cancer Res. 202077 TMB-high in 1% IIIA Fabrizio D, et al. J biomarker) MSS Gastrointest Oncol. MDM2 Amplifications w1% IIIA Dembla V, et al. 201889 Oncotarget. 201878 RET Fusions 0.3% IIIA Drilon A, et al. J Clin ERBB3 Mutations 2% IIIB Hyman D, et al. Nature. Oncol. 201891 201855 ALK Fusions 0.2% IIIA Yakirevich E, et al. Clin 88 ESCAT, European Society for Medical Oncology (ESMO) Scale for Clinical Actionability Cancer Res 2016 of molecular Targets; MSI-H, microsatellite instability-high. ESCAT, European Society for Medical Oncology (ESMO) Scale for Clinical Actionability of molecular Targets; MSI-H, microsatellite instability-high; MSS, microsatellite stable. therefore recommended for clinical research centres to include patients in molecular screening programmes to propose access to innovative agents in clinical trials. studies, although in advanced prostate cancer have shown minimal activity.73,96,97 PTEN alterations are found very frequently in mCRPC,98 and AKT inhibitors in combination GENOMIC ALTERATIONS IN ADVANCED PROSTATE CANCER with abiraterone showed antitumour activity in a retro- CLASSIFIED ACCORDING TO ESCAT spective analysis of a randomised phase II trial.99 Pre- Metastatic castration-resistant prostate cancer (mCRPC) liminary results of IPATential 150, a phase III randomised presents aberrations in DNA repair genes with a high fre- trial which evaluated ipatasertib (AKT inhibitor) with abir- quency (20%e30%). PARPi improved outcomes in patients aterone and prednisone compared with standard therapy, with different DNA repair gene alterations in a randomised showed an improvement of radiographic PFS (co-primary phase III trial; however, exploratory per-gene analysis sug- end point) in patients with PTEN loss and mCRPC, but not in gested that most of the benefit was obtained in patients the overall population.100 Some alterations ranked level I/II with BRCA1/2 somatic mutations.93 This is supported by in other diseases are observed in prostate cancer, but are multiple phase II trials, where patients with BRCA1/2 al- not yet validated101 (see Table 6). terations achieved the higher response rates. Data about Summary of recommendations. In countries where PALB2, RAD50, RAD51 or BRIP1 mutations are promising PARPi are accessible for patients with prostate cancer, it is but sparse due to the low frequency of these aberra- recommended to perform NGS on tumour samples to tions.93,95 Other genes involved in DNA repair, like MLH1/ assess the mutational status of, at least, BRCA1/2. Ac- MSH2/MSH6 lead to MSI-H when mutated. Therapy with cording to the preliminary results of the phase III trial with ICIs demonstrated effectiveness in multi-histology basket AKT inhibitors in patients with PTEN alterations, this gene

Volume 31 - Issue 11 - 2020 https://doi.org/10.1016/j.annonc.2020.07.014 1497 Annals of Oncology F. Mosele et al.

Table 6. List of genomic alterations level I/II/III according to ESCAT in Table 7. List of genomic alterations level I/II/III according to ESCAT in advanced prostate cancer metastatic gastric cancer (mGC)

Gene Alteration Prevalence ESCAT References Gene Alteration Prevalence ESCAT References BRCA1/ Somatic 9% IA De Bono J, et al. N Engl J ERBB2 Amplifications 16% IA The Cancer Genome Atlas 2 mutations/ Med. 202093 Research Network. Nature. deletions 2014102 MSI-H 1% IC Cortes-Ciriano I, et al. Nat Bang Y-J, et al. Lancet. 2010103 Commun. 201796 Hotspot 3% IIIA Hyman D, et al. Nature. 201855 Abida W, et al. J Clin Oncol. mutations 201897 MSI-H 8% IC The Cancer Genome Atlas Marcus L, et al. Clin Cancer Research Network. Nature. Res. 201997 2014102 PTEN Deletions/ 40% IIAa Abida W, et al. Proc Natl Marcus L, et al. Clin Cancer Res. mutations Acad Sci. 201998 201997 De Bono J, et al. Clin Cancer NTRK Fusions 2% IC Drilon A, et al. N Engl J Med. Res. 201999 201848 NCT03072238100 EGFR Amplifications 6% IIB Maron S, et al. Cancer Discov. ATM Mutations/ 5% IIA De Bono J, et al. N Engl J 2018104 deletions Med. 202093 MET Amplifications 3% IIB Lennerz J, et al. J Clin Oncol. PALB2 Mutations 1% IIB Mateo J, et al. N Engl J 2011105 Med. 201595 Mutations 1.3% IIIA Lee J, et al. Oncotarget. 2015107 De Bono J, et al. N Engl J PIK3CA Hotspot 7% IIIA Juric D, et al. J Clin Oncol. 201890 Med. 202093 mutations PIK3CA Hotspot 3% IIIA Crumbaker M, et al. FGFR2 Amplifications 4% IIIA Van Cutsem E, et al. Ann Oncol. mutations Cancers. 2017101 2017109 AKT1E17K Mutations 1% IIIA Crumbaker M, et al. Loriot Y, et al. N Engl J Med. Cancers. 2017101 2019110 ESCAT, European Society for Medical Oncology (ESMO) Scale for Clinical Actionability ATM Mutations 3% IIIA Bang Y-J, et al. Lancet Oncol. 108 of molecular Targets; MSI-H, microsatellite instability-high; PTEN, phosphatase and 2017 tensin homologue. BRCA1/ Mutations 1%e5% IIIA Balasubramaniam S, et al. Clin a A press release suggests that AKT inhibitors could work specifically in PTEN-mutant 2 Cancer Res. 201763 prostate cancers. PTEN could be upgraded to IA depending on the magnitude of ROS1 Fusions <1% IIIA Shaw A, et al. Ann Oncol. 201946 benefit and peer review assessment of the report. RET Fusions <1% IIIA Oxnard G, et al. J Thorac Oncol. 2018106 ERBB3 Hotspot 3% IIIB Hyman D, et al. Nature. 201855 could be added to the panel. Given that they are unlikely mutations to be cost-effective in these cases, larger panels can be ESCAT, European Society for Medical Oncology (ESMO) Scale for Clinical Actionability used only on the basis of specific agreements with payers of molecular Targets; MSI-H, microsatellite instability-high. taking into account the overall cost of the strategy (including off-label use of drugs) and pending a ranking GENOMIC ALTERATIONS IN ADVANCED PANCREATIC of additional alterations using a valid ranking system. DUCTAL ADENOCARCINOMA CLASSIFIED ACCORDING TO These panels should include DNA repair genes and MSI ESCAT signature. Patients with germline BRCA1/2-mutated advanced pancreatic ductal adenocarcinoma (PDAC) presented a longer PFS with maintenance olaparib.111,112 In advanced GENOMIC ALTERATIONS IN METASTATIC GASTRIC CANCER PDAC with somatic BRCA1/2 mutations, an increased CLASSIFIED ACCORDING TO ESCAT response with PARPi has been reported in few patients 113 ERBB2 amplifications are observed in around 15% of gastric included in a prospective trial. The panel therefore cancers.102 In these patients, trastuzumab demonstrated a considered that somatic BRCA1/2 alterations are not yet significant improvement of OS in randomised trials.103 Ac- validated in advanced PDAC. As we mentioned for other cording to basket trials, patients with MSI-H and NTRK tumours, patients with MSI-H and NTRK fusion-positive tumours presented meaningful clinical benefit with fusion-positive tumours treated with ICIs and TRK inhibitors 50,97,114,115 are expected to provide benefit.48,73 Some limited re- matched therapies in multi-histology studies. fi sponses were observed in patients with EGFR- and MET- Several additional alterations are classi ed at high level amplified metastatic gastric cancer (mGC) treated with according to ESCAT in other tumours, but have not yet 104,105 shown a significant impact in pancreatic cancer like KRAS, cetuximab and crizotinib in prospective analysis. V600E fi PIK3CA, BRAF mutations, MDM2, ERBB2 amplifications These ndings require further investigation. In addition, 55,91,116e125 many other level I/II aberrations of other cancer types are and NRG1, ALK, RET, ROS1 fusions. The main fi observed in gastric cancer, but not validated in this latter drivers of PDAC and their classi cation are described in disease.46,55,63,90,106e110 All these alterations are described Table 8. in Table 7. Summary of recommendations. It is not currently rec- Summary of recommendations. There is no current need ommended to perform tumour multigene NGS in patients to perform tumour multigene NGS in patients with mGC in with advanced PDAC in daily practice. Considering the daily practice. Detection of MSI and NTRK fusions should unmet medical needs and the high number of alterations e be done using cheap standard methods. ranked as level II IV, ESMO considers it is the mission of

1498 https://doi.org/10.1016/j.annonc.2020.07.014 Volume 31 - Issue 11 - 2020 F. Mosele et al. Annals of Oncology

Table 8. List of genomic alterations level I/II/III according to ESCAT in Table 9. List of genomic alterations level I/II/III according to ESCAT in advanced pancreatic ductal adenocarcinoma (PDAC) advanced hepatocellular carcinoma (HCC)

Gene Alteration Prevalence ESCAT References Gene Alteration Prevalence ESCAT References BRCA1/2 Germline 1%e4% IA The Cancer Genome Atlas NTRK Fusions 1% IC The Cancer Genome Atlas mutations Research Network. Cancer Research Network. Cancer Cell. Cell. 2017111 2017111 Golan T, et al. N Engl J Med. Drilon A, et al. N Engl J Med. 2019112 201848 Somatic 3% IIIB Shroff R, et al. JCO Precis MSI-H 1% IC Marcus L, et al. Clin Cancer Res. mutations Oncol. 2018113 201997 MSI-H 1%e3% IC Pihlak R, et al. Cancers. PIK3CA Hotspot 4% IIIA André F, et al. N Engl J Med. 2018115 mutations 201972 Marcus L, et al. Clin Cancer MET Amplifications 2%e6% IIIA Rimassa L, et al. Lancet Oncol. Res. 201997 2018127 NTRK Fusions <1% IC Cocco E, et al. Nat Rev Clin RAS Mutations 2% IIIA Lim H, et al. Clin Cancer Res. Oncol. 2018114 2018128 Doebele RC, et al. Lancet 50 ESCAT, European Society for Medical Oncology (ESMO) Scale for Clinical Actionability Oncol. 2020 of molecular Targets; MSI-H, microsatellite instability-high. KRAS Mutations 90% IIIA Zeitouni D, et al. Cancers. 2016116 PIK3CA Hotspot 3% IIIA Heestand G, et al. mutations Oncotarget. 2015117 Payne S, et al. J Clin Oncol. Summary of recommendations. It is not currently 2015118 recommended to perform tumour multigene NGS in V600E BRAF Mutations 3% IIIA Hyman D, et al. N Engl J Med. patients with advanced hepatocellular carcinoma (HCC) 2015119 MDM2 Amplifications 2% IIIA Azmi A, et al. Eur J Cancer. in daily practice. Considering the unmet medical needs 2010120 and the number of alterations ranked as level IIeIV, ERBB2 Amplifications/ 1%e2% IIIA Waddell N, et al. Nature. ESMO considers it is the mission of clinical research mutations 2015121 Harder J, et al. Br J Cancer. centres to propose multigene sequencing to patients 2012122 with advanced HCC in the context of molecular Hyman D, et al. Nature. 201855 screening programmes. If multigene sequencing is not NRG1 Fusions 1% IIIA Jones M, et al. Clin Cancer carried out, detection of MSI and NTRK fusions should Res. 2019123 be done using cheaper standard methods, pending ALK Fusions <1% IIIA Singhi A, et al. J Natl Compr Canc Netw. 2017124 drugs are approved and reimbursed. RET Fusions <1% IIIA Drilon A, et al. J Clin Oncol. 201891 < ROS1 Fusions 1% IIIA Pishvaian M, et al. J Clin GENOMIC ALTERATIONS IN ADVANCED Oncol. 2018125 ESCAT, European Society for Medical Oncology (ESMO) Scale for Clinical Actionability CHOLANGIOCARCINOMA CLASSIFIED ACCORDING TO of molecular Targets; MSI-H, microsatellite instability-high. ESCAT IDH1 mutations are ranked level I in ESCAT (IA).129 In clinical research centres and their networks to propose addition, pemigatinib, a selective fibroblast growth multigene sequencing to patients with advanced PDAC in factor receptor (FGFR)1,2,3 inhibitor, led to a 35% ORR the context of molecular screening programmes, in order in patients with advanced FGFR2 fusion-positive chol- for patients to get access to innovative drugs. If multigene angiocarcinoma (CC) in a prospective phase II trial,130 sequencing is not carried out, detection of MSI and NTRK getting accelerated approval by the FDA. As we fusions should be done using cheaper standard methods, mentioned previously, patients with MSI-H and NTRK pending drugs are approved and reimbursed. fusion-positive tumours presented clinically meaningful benefit with ICIs and TRK inhibitors in basket studies.50,131 Finally, rapidly accelerated fibrosarcoma/ GENOMIC ALTERATIONS IN ADVANCED HEPATOCELLULAR mitogen-activated protein kinase kinase inhibitors were CARCINOMA CLASSIFIED ACCORDING TO ESCAT associated with 42% OR in patients with advanced CC While numerous aberrations are being evaluated, very and BRAFV600E mutations132 (Table 10). In Table 10 are few targets currently have impact on clinical de- also described some alterations with efficacy in other cisions.126 As we described for the majority of cancers, tumours, but not yet validated in this disease.52,72,93,133 due to their clinical benefit larotrectinib and ICIs were Summary of recommendations. Tumour multigene approved for patients with NTRK fusion-positive and NGS could be used to detect level I actionable alter- MSI-H solid tumours, respectively, who have no alter- ations in cholangiocarcinoma. Given that they are un- native treatments.48,97 There are also other alterations likely to be cost-effective in these cases, larger panels with strong benefitacrossdifferenttumourtypeslike canbeusedonlyonthebasisofspecific agreements PIK3CA, RAS mutations and MET amplifications,72,127,128 with payers taking into account the overall cost of the and no clinical evidence in this disease (Table 9). strategy (including off-label use of drugs) and pending a

Volume 31 - Issue 11 - 2020 https://doi.org/10.1016/j.annonc.2020.07.014 1499 Annals of Oncology F. Mosele et al.

needed before implementing TMB in all cancers where Table 10. List of genomic alterations level I/II/III according to ESCAT in advanced cholangiocarcinoma (CC) anti-PD(L)1 antibodies are not approved.

Gene Alteration Prevalence ESCAT References IDH1 Mutations 20% IA Abou-Alfa G. K, et al. Ann NTRK fusions. TRK inhibitors have been shown to be Oncol. 2019129 effective in a broad range of cancers. NTRK fusions occur in FGFR2 Fusions 15% IB Vogel A, et al. Ann Oncol. <1% of cancers. The incidence of NTRK fusions is very high 2019130 MSI-H 2% IC Marabelle A, et al. J Clin in mammary analogue secretory carcinoma of salivary Oncol. 2020131 glands and in secretory breast cancers. A high incidence is NTRK Fusions 2% IC Doebele RC, et al. Lancet also observed in sarcoma and thyroid cancers. Considering Oncol. 202050 BRAFV600E Mutations 5% IIB Wainberg Z, et al. J Clin Oncol. the very low incidence, the group recommends using NGS 2019132 to detect NTRK fusions only in cancers where this tech- ERBB2 Amplifications 10% IIIA Javle MM, et al. J Clin Oncol. nology is recommended otherwise. In cancers where there Mutations 2% 2017133 PIK3CA Hotspot 7% IIIA André F, et al. N Engl J Med. is no need for multigene sequencing, it was considered mutations 201972 that the detection of NTRK fusion is not an argument per BRCA 1/2 Mutations 3% IIIA De Bono J, et al. N Engl J Med. se to recommend NGS since alternative, cheaper, diag- 202093 MET Amplifications 2% IIIA Camidge D, et al. J Clin Oncol. nostic methods exist. Such alternative, cheaper methods 201852 should be prioritised to screen patients for NTRK fusions, in ESCAT, European Society for Medical Oncology (ESMO) Scale for Clinical Actionability countries where TRK inhibitors are available. of molecular Targets.

CONCLUSION ranking of additional alterations using a valid ranking ESMO recommends using tumour multigene NGS in patients system. presenting with advanced non-squamous NSCLC, prostate, ovarian cancers and cholangiocarcinoma. Large panels of genes can be used if they generate only an acceptable in- Other tumour types. While the systematic ranking of crease in the overall cost, drugs included. In addition, based genomic alterations was done exclusively for the eight on KN158, it is recommended to determine TMB in cervical more frequent killers, we also assessed the frequency cancer, salivary cancer, thyroid cancers, well-to-moderately of level I alterations in other tumour types. In ovarian differentiated NETs, vulvar cancer, pending drug access. In cancers, where BRCA1/2 somatic mutations have been colorectal cancers, NGS can be an alternative to PCR-based fi 134 associated with increased bene ttoPARPi, the use tests, if it is not associated with extra cost. ESMO strongly of multigene NGS is justified. Larger panels can be used recommends that clinical research centres perform multi- only on the basis of specific agreements with payers gene sequencing as part of their missions to accelerate taking into account the overall cost of the strategy cancer research and drug development through clinical tri- (including off-label use of drugs) and pending an als, provide access to innovation to patients and to collect appropriate method of reporting. While there is no data. In addition, economic evaluations alongside clinical level I evidence, multigene sequencing could also be 135 trials should also be implemented to foster evidence in this used in carcinoma of unknown primary. field. Outside the indications mentioned before, and considering that the use of large panels of genes could lead to identification of few exceptional responders, ESMO ac- Specific situations knowledges that a patient and a doctor could decide Tumour mutational burden and KN158 study. KN158 has together to order a large panel of genes, pending no extra evaluated the efficacy of pembrolizumab according to TMB cost for the public health care system, and if the patient is in 10 cancers (anal cancer, cervical cancer, endometrial informed about the low likelihood of benefit. cancer, small-cell lung cancer (SCLC), salivary cancer, thyroid These recommendations will need to be updated on a cancers, well-to-moderately differentiated neuroendocrine regular basis as new data emerges for novel therapies tumours (NETs), biliary cancers, vulvar cancer, mesotheli- across tumour types. oma). Response rates were 27% and 7% in patients with TMB-high (MSI-low) or TMB-low cancers, respectively. There ACKNOWLEDGEMENT was no TMB-high detected in biliary cancers, and the per- This is a project initiated by the ESMO Translational centage of response was lower in TMB-high in anal cancer Research and Precision Medicine Working Group. We would and mesothelioma. We can classify TMB as level IIA ac- also like to thank ESMO leadership for their support in this cording to ESCAT. If we consider that indications of anti- manuscript. PD(L)1 antibodies are broad in endometrial cancers and SCLC, the TMB should be determined only in cervical can- cer, NET, salivary cancers, vulvar cancers, thyroid cancers. FUNDING Considering that the study was not agnostic, but limited to This work was supported by the European Society for few cancers, the group thinks that additional studies are Medical Oncology (no grant number is applicable).

1500 https://doi.org/10.1016/j.annonc.2020.07.014 Volume 31 - Issue 11 - 2020 F. Mosele et al. Annals of Oncology

DISCLOSURE speakers bureau: Astra Zeneca, Eli Lilly, Bayer, BMS, Illu- fi JR: advisory: Merck Sharp & Dohme (MSD), Boehringer, mina, Janssen, MSD, P zer, Roche, Seattle Genetics, Thermo Bristol-Myers Squibb (BMS), AstraZeneca, Roche; speaker’s Fisher; Grants: Bayer, BMS, Chugai. JB: travel support: BMS; bureau: Pfizer; travel support: OSE Immunotherapeutics SA, consulting fees: BMS, MSD, Astellas. SM: statistical advice: BMS, AstraZeneca, Roche. JM: advisory board: Amgen, IDDI and Janssen Cilag; Independent Data Monitoring AstraZeneca, Clovis Oncology, Janssen, MSD and Roche- Committee member: Hexal, Steba, IQVIA, Roche, Sensorion, ’ Foundation Medicine; research funding: AstraZeneca and Biophytis, Servier, Yuhan. IB: speaker s fee: AstraZeneca. ER: Pfizer Oncology; principal investigator of several industry board participation: AstraZeneca, BMS, Roche; travel fund- sponsored clinical trials. CBW: personal and speakers’ fees, ing: AstraZeneca, BMS. JSR-F: paid consultant: Goldman reimbursement for travel and accommodation and hono- Sachs, REPARE Therapeutics, and Paige.AI; member of the fi raria for participance in advisory boards from Bayer, Cel- scienti c advisory board: REPARE Therapeutics, Paige.AI, gene, Ipsen, MedScape, Rafael Pharmaceuticals, RedHill, and Volition Rx; member of the Board of Directors: Group fi Roche, Servier, Shire/Baxalta and Taiho; scientific grant Oncoclinicas; ad hoc member of the scienti c advisory support by Roche. FB: personal fees from AstraZeneca, board: Roche Tissue Diagnostics, Roche, Genentech, Bayer, Bristol-Myers Squibb, BoehringereIngelheim, Eli Lilly Novartis, and Invicro; owns shares: REPARE Therapeutics. ’ Oncology, F. HoffmanneLa Roche Ltd, Novartis, Merck, RD: advisory: Roche, Boehringer Ingelheim; speaker s fee: fi MSD, Pierre Fabre, Pfizer and Takeda. MPL: research grants Roche, Ipsen, Amgen, Servier, Sano , Merck Sharp & (to hospital): MSD, Astellas, JnJ, Sanofi; advice: Roche, Dohme; research grants: Merck and Pierre Fabre. FA: Bayer, Amgen, JnJ, Sanofi, Servier, Pfizer, Incyte. NN: research grants and talks/advisory boards compensate to fi speaker’s fee and/or advisory boards: Amgen, AstraZeneca, the hospital: Roche, P zer, Novartis, AstraZeneca, Daiichi Bayer, Biocartis, BMS, Boehringer Ingelheim, Eli Lilly, Ilu- Sankyo, Lilly. All remaining authors have declared no con- fl mina, Incyte, MERCK, MSD, Qiagen, Roche, Thermofisher, icts of interest. Sanofi; institutional financial interests (financial support to research projects): AstraZeneca, Biocartis, BMS, Illumina, REFERENCES Merck, Qiagen, Roche, Sysmex, Thermofisher; non-financial 1. van Nimwegen KJM, van Soest RA, Veltman JA, et al. Is the $1000 interests: President of the International Quality Network for genome as near as we think? A cost analysis of next-generation Pathology (IQN Path); President of the Italian Cancer Society sequencing. Clin Chem. 2016;62(11):1458e1464. (SIC). ASc: speakers bureau: Ypsen, Astra Zeneca, Amgen, 2. Marino P, Touzani R, Perrier L, et al. Cost of cancer diagnosis using next-generation sequencing targeted gene panels in routine practice: MSD, GSK; consulting: INCYTE Biosciences. MR: consulting a nationwide French study. Eur J Hum Genet. 2018;26(3):314e323. or advisory: AstraZeneca (uncompensated), Change 3. Pagès A, Foulon S, Zou Z, et al. The cost of molecular-guided therapy Healthcare, Daiichi-Sankyo (uncompensated), Epic Sciences in oncology: a prospective cost study alongside the MOSCATO trial. (uncompensated), Merck (uncompensated), Pfizer (uncom- Genet Med. 2017;19(6):683e690. pensated); research funding: AbbVie (institution), AstraZe- 4. Frampton GM, Fichtenholtz A, Otto GA, et al. Development and validation of a clinical cancer genomic profiling test based on neca (Institution), Invitae (Institution, in-kind), Merck massively parallel DNA sequencing. Nat Biotechnol. 2013;31(11): (Institution), Pfizer (institution); travel, accommodation, 1023e1031. expenses: AstraZeneca, Merck; editorial services: AstraZe- 5. Tourneau CL, Delord J-P, Gonçalves A, et al. Molecularly targeted neca, Pfizer. FM-B: consulting: Aduro BioTech Inc., therapy based on tumour molecular profiling versus conventional Alkermes, DebioPharm, eFFECTOR Therapeutics, F. therapy for advanced cancer (SHIVA): a multicentre, open-label, proof-of-concept, randomised, controlled phase 2 trial. Lancet Hoffman-La Roche Ltd, Genentech Inc., IBM Watson, Jack- Oncol. 2015;16(13):1324e1334. son Laboratory, Kolon Life Science, OrigiMed, PACT Pharma, 6. Massard C, Michiels S, Ferté C, et al. High-throughput genomics and Parexel International, Pfizer Inc., Samsung Bioepis, Seattle clinical outcome in hard-to-treat advanced cancers: Results of the Genetics Inc., Tyra Biosciences, Xencor, Zymeworks; advisory MOSCATO 01 trial. Cancer Discov. 2017;7(6):586e596. committee: Immunomedics, Inflection Biosciences, Mersana 7. André F, Bachelot T, Commo F, et al. Comparative genomic hybrid- isation array and DNA sequencing to direct treatment of metastatic Therapeutics, Puma Biotechnology Inc., Seattle Genetics, breast cancer: a multicentre, prospective trial (SAFIR01/UNICANCER). Silverback Therapeutics, Spectrum Pharmaceuticals, Zen- Lancet Oncol. 2014;15(3):267e274. talis; sponsored research: Aileron Therapeutics, Inc., Astra- 8. Tsimberidou A-M, Wen S, Hong DS, et al. Personalized medicine for Zeneca, Bayer Healthcare Pharmaceutical, Calithera patients with advanced cancer in the phase I program at MD Biosciences Inc., Curis Inc., CytomX Therapeutics Inc., Daii- Anderson: validation and landmark analyses. Clin Cancer Res. 2014;20(18):4827e4836. chi Sankyo Co. Ltd, Debiopharm International, eFFECTOR 9. Priestley P, Baber J, Lolkema MP, et al. Pan-cancer whole-genome Therapeutics, Genentech Inc., Guardant Health Inc., Mil- analyses of metastatic solid tumours. Nature. 2019;575(7781): lennium Pharmaceuticals Inc., Novartis, Puma Biotech- 210e216. nology Inc., Taiho Pharmaceutical Co.; honoraria: Chugai 10. Trédan O, Wang Q, Pissaloux D, et al. Molecular screening program to Biopharmaceuticals, Mayo Clinic, Rutgers Cancer Institute of select molecular-based recommended therapies for metastatic cancer patients: analysis from the ProfiLER trial. Ann Oncol. 2019;30(5): New Jersey; other (Travel Related): Beth Israel Deaconess 757e765. Medical Center. NW: research grant from Puma Biotech- 11. Korphaisarn K, Loree JM, Nguyen V, et al. Genomic analysis of nology; scientific advisory board and stockholder for Relay exceptional responder to regorafenib in treatment-refractory meta- Therapeutics; advisor to Eli Lilly. ASt: advisory board/ static rectal cancer: a case report and review of the literature. Oncotarget. 2017;8(34):57882e57888.

Volume 31 - Issue 11 - 2020 https://doi.org/10.1016/j.annonc.2020.07.014 1501 Annals of Oncology F. Mosele et al.

12. Espinosa M, Roldán-Romero JM, Duran I, et al. Advanced sporadic uncommon EGFR mutations: a combined post-hoc analysis of LUX- renal epithelioid angiomyolipoma: case report of an extraordinary Lung 2, LUX-Lung 3, and LUX-Lung 6. Lancet Oncol. 2015;16(7): response to sirolimus linked to TSC2 mutation. BMC Cancer. 830e838. 2018;18(1):561. 32. Cho JH, Sun J, Lee S, et al. OA10.05 An open-label, multicenter, phase 13. Bray F, Ferlay J, Soerjomataram I. Global cancer statistics 2018: II single arm trial of osimertinib in NSCLC patients with uncommon GLOBOCAN estimates of incidence and mortality worldwide for EGFR mutation (KCSG-LU15-09). J Thorac Oncol. 2018;13(10):S344. 36 cancers in 185 countries. CA Cancer J Clin. 2018;68(6):394e 33. Cardona AF, Rojas L, Zatarain-Barrón ZL, et al. EGFR exon 20 insertion 424. in lung adenocarcinomas among Hispanics (geno1.2-CLICaP). Lung 14. Mateo J, Chakravarty D, Dienstmann R, et al. A framework to rank Cancer. 2018;125:265e272. genomic alterations as targets for cancer precision medicine: the 34. Heymach J, Negrao M, Robichaux J, et al. OA02.06 A phase II trial of ESMO Scale for Clinical Actionability of molecular Targets (ESCAT). poziotinib in EGFR and HER2 exon 20 mutant non-small cell lung Ann Oncol. 2018;29(9):1895e1902. cancer (NSCLC). J Thorac Oncol. 2018;13(10):S323eS324. 15. Ramsey SD, Shankaran V, Sullivan SD. Basket cases: How real-world 35. Solomon BJ, Kim D-W, Wu Y-L, et al. Final overall survival analysis testing for drugs approved based on basket trials might lead to from a study comparing first-line crizotinib versus chemotherapy in false diagnoses, patient risks, and squandered resources. J Clin Oncol. ALK-mutation-positive non-small-cell lung cancer. J Clin Oncol. 2019;37(36):3472e3474. 2018;36(22):2251e2258. 16. FDA approves pembrolizumab for adults and children with TMB-H 36. Soria J-C, Tan DSW, Chiari R, et al. First-line ceritinib versus platinum- solid tumors - the ASCO Post. Available at: https://www.ascopost. based chemotherapy in advanced ALK-rearranged non-small-cell lung com/news/june-2020/fda-approves-pembrolizumab-for-adults-and- cancer (ASCEND-4): a randomised, open-label, phase 3 study. Lancet. children-with-tmb-h-solid-tumors/. Accessed July 7, 2020. 2017;389(10072):917e929. 17. OncoKB. Available at: https://www.oncokb.org/. Accessed March 10, 37. Peters S, Camidge DR, ALEX Trial Investigators, et al. Alectinib versus 2020. crizotinib in untreated ALK-positive non-small-cell lung cancer. N Engl 18. Veenstra DL, Mandelblatt J, Neumann P, et al. Health economics tools J Med. 2017;377(9):829e838. and precision medicine: Opportunities and challenges. Forum Health 38. Zhou C, Lee SH, Wang C, et al. Primary results of ALESIA: a rando- Econ Policy. 2020;23(1). https://doi.org/10.1515/fhep-2019-0013. mised, phase III, open-label study of alectinib vs crizotinib in Asian 19. Weymann D, Pataky R, Regier DA. Economic evaluations of next- patients with treatment-naïve ALKþ advance NSCLC. Ann Oncol. generation precision oncology: a critical review. JCO Precis Oncol. 2018;29(suppl_8):ix173eix178. 2018;2. https://doi.org/10.1200/PO.17.00311. 39. Camidge DR, Kim HR, Ahn M-J, et al. Brigatinib versus crizotinib in 20. Tan O, Shrestha R, Cunich M, et al. Application of next-generation ALK-positive nonesmall-cell lung cancer. N Engl J Med. 2018;379(21): sequencing to improve cancer management: a review of the clinical 2027e2039. effectiveness and cost-effectiveness. Clin Genet. 2018;93(3): 40. Tong JH, Yeung SF, Chan AWH, et al. MET Amplification and exon 14 533e544. splice site mutation define unique molecular subgroups of non-small 21. Phillips KA, Deverka PA, Deborah A, Marshall, et al. Methodological cell lung carcinoma with poor prognosis. Clin Cancer Res. issues in assessing the economic value of next-generation sequencing 2016;22(12):3048e3056. tests: many challenges and not enough solutions. Value Health. 41. Drilon A, Clark JW, Weiss J, et al. Antitumor activity of crizotinib in 2018;21(9):1033e1042. lung cancers harboring a MET exon 14 alteration. Nat Med. 22. Tan AC, Lai GGY, Tan GS, et al. Utility of incorporating next-generation 2020;26(1):47e51. sequencing (NGS) in an Asian non-small cell lung cancer (NSCLC) 42. Planchard D, Besse B, Groen HJM, et al. Dabrafenib plus trametinib in population: incremental yield of actionable alterations and cost- patients with previously treated BRAF(V600E)-mutant metastatic effectiveness analysis. Lung Cancer. 2020;139:207e215. non-small cell lung cancer: an open-label, multicentre phase 2 trial. 23. Steuten L, Goulart B, Meropol NJ, et al. Cost effectiveness of multi- Lancet Oncol. 2016;17(7):984e993. gene panel sequencing for patients with advanced nonesmall-cell 43. Planchard D, Smit EF, Groen HJM, et al. Dabrafenib plus trametinib in lung cancer. JCO Clin Cancer Inform. 2019;3:1e10. patients with previously untreated BRAF(V600E)-mutant metastatic 24. Sboner A, Mu XJ, Greenbaum D, et al. The real cost of sequencing: non-small-cell lung cancer: an open-label, phase 2 trial. Lancet Oncol. higher than you think!. Genome Biol. 2011;12(8):125. 2017;18(10):1307e1316. 25. Legras A, Barritault M, Tallet A, et al. Validity of targeted next- 44. Planchard D, Besse B, Kim TM, et al. Updated survival of patients (pts) generation sequencing in routine care for identifying clinically with previously treated BRAF V600Eemutant advanced non-small cell relevant molecular profiles in non-small-cell lung cancer: results of a lung cancer (NSCLC) who received dabrafenib (D) or D þ trametinib 2-year experience on 1343 samples. J Mol Diagn. 2018;20(4): (T) in the phase II BRF113928 study. J Clin Oncol. 2017;35(15_suppl): 550e564. 9075. 26. Midha A, Dearden S, McCormack R. EGFR mutation incidence in non- 45. Shaw AT, Ou S-HI, Bang Y-J, et al. Crizotinib in ROS1-rearranged non- small-cell lung cancer of adenocarcinoma histology: a systematic small-cell lung cancer. N Engl J Med. 2014;371(21):1963e1971. review and global map by ethnicity (mutMapII). Am J Cancer Res. 46. Shaw AT, Riely GJ, Bang Y-J, et al. Crizotinib in ROS1-rearranged 2015;5(9):2892e2911. advanced non-small-cell lung cancer (NSCLC): updated results, 27. Mok TS, Cheng Y, Zhou X, et al. Improvement in overall survival in a including overall survival, from PROFILE 1001. Ann Oncol. 2019;30(7): randomized study that compared dacomitinib with gefitinib in pa- 1121e1126. tients with advanced non-small-cell lung cancer and EGFR-activating 47. Drilon A, Siena S, Dziadziuszko R, et al. Entrectinib in ROS1 fusion- mutations. J Clin Oncol. 2018;36(22):2244e2250. positive non-small-cell lung cancer: integrated analysis of three 28. Soria J-C, Ohe Y,Vansteenkiste J, et al. Osimertinib in untreated EGFR- phase 1-2 trials. Lancet Oncol. 2020;21(2):261e270. mutated advanced nonesmall-cell lung cancer. N Engl J Med. 48. Drilon A, Laetsch TW, Kummar S, et al. Efficacy of larotrectinib in TRK 2017;378(2):113e125. fusion-positive cancers in adults and children. N Engl J Med. 29. Ramalingam SS, Vansteenkiste J, Planchard D, et al. Overall survival 2018;378:731e739. with osimertinib in untreated, EGFR-mutated advanced NSCLC. 49. Hong DS, DuBois SG, Kummar S, et al. Larotrectinib in patients with N Engl J Med. 2020;382(1):41e50. TRK fusion-positive solid tumours: a pooled analysis of three phase 30. Mok TS, Wu Y-L, Ahn M-J, et al. Osimertinib or platinumepemetrexed 1/2 clinical trials. Lancet Oncol. 2020;21(4):531e540. in EGFR T790Mepositive lung cancer. N Engl J Med. 2017;376(7): 50. Doebele RC, Drilon A, Paz-Ares L, et al. Entrectinib in patients with 629e640. advanced or metastatic NTRK fusion-positive solid tumours: inte- 31. Yang JC-H, Sequist LV, Geater SL, et al. Clinical activity of afatinib in grated analysis of three phase 1-2 trials. Lancet Oncol. 2020;21(2): patients with advanced non-small-cell lung cancer harbouring 271e282.

1502 https://doi.org/10.1016/j.annonc.2020.07.014 Volume 31 - Issue 11 - 2020 F. Mosele et al. Annals of Oncology

51. Drilon A, Oxnard G, Wirth L, et al. PL02.08 registrational results 72. André F, Ciruelos E, Rubovszky G, et al. Alpelisib for PIK3CA-mutated, of LIBRETTO-001: a phase 1/2 trial of LOXO-292 in patients with hormone receptor-positive advanced breast cancer. N Engl J Med. RET fusion-positive lung cancers. J Thorac Oncol. 2019;14(10): 2019;380:1929e1940. S6eS7. 73. Marcus L, Lemery SJ, Keegan P, et al. FDA approval summary: pem- 52. Camidge DR, Otterson GA, Clark JW, et al. Crizotinib in patients (pts) brolizumab for the treatment of microsatellite instability-high solid with MET-amplified non-small cell lung cancer (NSCLC): updated tumors. Clin Cancer Res. 2019;25(13):3753e3758. safety and efficacy findings from a phase 1 trial. J Clin Oncol. 2018;36: 74. Fribbens C, O’Leary B, Kilburn L, et al. Plasma ESR1 mutations and the 9062. treatment of estrogen receptorepositive advanced breast cancer. 53. Barlesi F, Mazieres J, Merlio J-P, et al. Routine molecular profiling of J Clin Oncol. 2016;34(25):2961e2968. patients with advanced non-small-cell lung cancer: results of a 1-year 75. Schmid P, Abraham J, Chan S, et al. AZD5363 plus paclitaxel versus nationwide programme of the French Cooperative Thoracic Inter- placebo plus paclitaxel as first-line therapy for metastatic triple- group (IFCT). Lancet. 2016;387(10026):1415e1426. negative breast cancer (PAKT): a randomised, double-blind, pla- 54. Fakih M, O’Neil B, Price TJ, et al. Phase 1 study evaluating the safety, cebo-controlled, phase II trial. J Clin Oncol. 2018;36(15_suppl): tolerability, pharmacokinetics (PK), and efficacy of AMG 510, a novel 1007. small molecule KRASG12C inhibitor, in advanced solid tumors. J Clin 76. Hyman DM, Smyth LM, Donoghue MTA, et al. AKT inhibition in solid Oncol. 2019;37(15_suppl):3003. tumors with AKT1 mutations. J Clin Oncol. 2017;35(20):2251e2259. 55. Hyman DM, Piha-Paul SA, Won, et al. HER kinase inhibition in patients 77. Pearson A, Proszek P, Ring A, et al. Inactivating NF1 mutations are with HER2- and HER3-mutant cancers. Nature. 2018;554:189e194. enriched in advanced breast cancer and contribute to endocrine 56. Wang Y, Jiang T, Qin Z, et al. HER2 exon 20 insertions in non-small-cell therapy resistance. Clin Cancer Res. 2020;26(3):608e622. lung cancer are sensitive to the irreversible pan-HER receptor tyro- 78. Dembla V, Somaiah N, Barata P, et al. Prevalence of MDM2 amplifi- sine kinase inhibitor pyrotinib. Ann Oncol. 2019;30(3):447e455. cation and coalterations in 523 advanced cancer patients in the MD 57. Tsurutani J, Park H, Doi T, et al. OA02.07 Updated results of phase 1 Anderson phase 1 clinic. Oncotarget. 2018;9:33232e33243. study of DS-8201a in HER2-expressing or emutated advanced non- 79. Van Cutsem E, Lenz H-J, Köhne C-H, et al. Fluorouracil, leucovorin, small-cell lung cancer. J Thorac Oncol. 2018;13(10):S324. and irinotecan plus cetuximab treatment and RAS mutations in 58. Hellmann MD, Paz-Ares L, Bernabe Caro R, et al. Nivolumab plus colorectal cancer. J Clin Oncol. 2015;33(7):692e700. ipilimumab in advanced non-small-cell lung cancer. N Engl J Med. 80. Douillard J-Y, Oliner KS, Siena S, et al. Panitumumab-FOLFOX4 treat- 2019;381(21):2020e2031. ment and RAS mutations in colorectal cancer. N Engl J Med. 59. Duruisseaux M, Liu SV, Han J-Y, et al. NRG1 fusion-positive lung 2013;369(11):1023e1034. cancers: Clinicopathologic profile and treatment outcomes from a 81. Sorich MJ, Wiese MD, Rowland A, et al. Extended RAS mutations and global multicenter registry. J Clin Oncol. 2019;37(15_suppl):9081. anti-EGFR monoclonal antibody survival benefit in metastatic colo- 60. The Cancer Genome Atlas Research Network. Comprehensive mo- rectal cancer: a meta-analysis of randomized, controlled trials. Ann lecular profiling of lung adenocarcinoma. Nature. 2014;511(7511): Oncol. 2015;26(1):13e21. 543e550. 82. Kopetz S, Grothey A, Yaeger R, et al. Encorafenib, binimetinib, and 61. Cancer Genome Atlas Research Network. Comprehensive genomic cetuximab in BRAF V600E-mutated colorectal cancer. N Engl J Med. characterization of squamous cell lung cancers. Nature. 2019;381(17):1632e1643. 2012;489(7417):519e525. 83. Overman MJ, McDermott R, Leach JL, et al. Nivolumab in patients 62. Vansteenkiste JF, Canon J-L, De Braud F, et al. Safety and efficacy of with metastatic DNA mismatch repair-deficient or microsatellite buparlisib (BKM120) in patients with PI3K pathway-activated instability-high colorectal cancer (CheckMate 142): an open-label, non-small cell lung cancer: results from the phase II BASALT-1 multicentre, phase 2 study. Lancet Oncol. 2017;18(9):1182e1191. study. J Thorac Oncol. 2015;10(9):1319e1327. 84. Le DT, Kim TW, Van Cutsem E, et al. Phase II open-label study of 63. Balasubramaniam S, Beaver JA, Horton S, et al. FDA approval sum- pembrolizumab in treatment-refractory, microsatellite instability- mary: rucaparib for the treatment of patients with deleterious BRCA high/mismatch repair-deficient metastatic colorectal cancer: KEY- mutation-associated advanced ovarian cancer. Clin Cancer Res. NOTE-164. J Clin Oncol. 2020;38(1):11e19. 2017;23(23):7165e7170. 85. Demetri GD, Paz-Ares L, Multani PS, et al. Efficacy and safety of 64. Voest E, van der Velden D, Hoes L, et al. Expanding the use of entrectinib in patients with NTRK fusion-positive tumours: Pooled approved drugs: The CPCT’s Drug Rediscovery Protocol (DRUP). Ann analysis of STARTRK-2, STARTRK-1, and ALKA-372-001. Ann Oncol. Oncol. 2017;28(suppl_5):v605ev649. 2018;29(Suppl_8):viii713. 65. Slamon DJ, Leyland-Jones B, Shak S, et al. Use of chemotherapy plus a 86. Meric-Bernstam F, Hurwitz H, Raghav KPS, et al. Pertuzumab plus monoclonal antibody against HER2 for metastatic breast cancer that trastuzumab for HER2-amplified metastatic colorectal cancer overexpresses HER2. N Engl J Med. 2001;344(11):783e792. (MyPathway): an updated report from a multicentre, open-label, 66. Swain SM, Baselga J, Kim S-B, et al. Pertuzumab, trastuzumab, and phase 2a, multiple basket study. Lancet Oncol. 2019;20(4): docetaxel in HER2-positive metastatic breast cancer. N Engl J Med. 518e530. 2015;372(8):724e734. 87. Sartore-Bianchi A, Trusolino L, Martino C, et al. Dual-targeted therapy 67. Verma S, Miles D, Gianni L, et al. Trastuzumab emtansine for HER2- with trastuzumab and lapatinib in treatment-refractory, KRAS codon positive advanced breast cancer. N Engl J Med. 2012;367(19): 12/13 wild-type, HER2-positive metastatic colorectal cancer (HERA- 1783e1791. CLES): a proof-of-concept, multicentre, open-label, phase 2 trial. 68. Krop IE, Kim S-B, González-Martín A, et al. Trastuzumab emtansine Lancet Oncol. 2016;17(6):738e746. versus treatment of physician’s choice for pretreated HER2-positive 88. Yakirevich E, Resnick MB, Mangray S, et al. Oncogenic ALK fusion in advanced breast cancer (TH3RESA): a randomised, open-label, rare and aggressive subtype of colorectal adenocarcinoma as a phase 3 trial. Lancet Oncol. 2014;15(7):689e699. potential therapeutic target. Clin Cancer Res. 2016;22(15):3831e 69. Murthy RK, Loi S, Okines A, et al. Tucatinib, trastuzumab, and cape- 3840. citabine for HER2-positive metastatic breast cancer. N Engl J Med. 89. Fabrizio DA, George TJ, Dunne RF, et al. Beyond microsatellite testing: 2020;382(7):597e609. assessment of tumor mutational burden identifies subsets of colo- 70. Robson M, Im S-A, Senkus E, et al. Olaparib for metastatic breast rectal cancer who may respond to immune checkpoint inhibition. cancer in patients with a germline BRCA mutation. N Engl J Med. J Gastrointest Oncol. 2018;9(4):610e617. 2017;377(6):523e533. 90. Juric D, Rodon J, Tabernero J, et al. Phosphatidylinositol 3-kinase ae 71. Litton JK, Rugo HS, Ettl J, et al. Talazoparib in patients with advanced selective inhibition with alpelisib (BYL719) in PIK3CA-altered solid breast cancer and a germline BRCA mutation. N Engl J Med. tumors: results from the first-in-human study. J Clin Oncol. 2018;36: 2018;379(8):753e763. 1291e1299.

Volume 31 - Issue 11 - 2020 https://doi.org/10.1016/j.annonc.2020.07.014 1503 Annals of Oncology F. Mosele et al.

91. Drilon AE, Subbiah V, Oxnard GR, et al. A phase 1 study of LOXO-292, 112. Golan T, Hammel P, Reni M, et al. Maintenance olaparib for germline a potent and highly selective RET inhibitor, in patients with RET- BRCA-mutated metastatic pancreatic cancer. N Engl J Med. 2019;381: altered cancers. J Clin Oncol. 2018;36(suppl). abstr 102. 317e327. 92. Wang C, Jette N, Moussienko D, et al. ATM-deficient colorectal cancer 113. Shroff RT, Hendifar A, McWilliams RR, et al. Rucaparib monotherapy cells are sensitive to the PARP inhibitor olaparib. Transl Oncol. in patients with pancreatic cancer and a known deleterious BRCA 2017;10(2):190e196. mutation. JCO Precis Oncol. 2018;2018. https://doi.org/10.1200/PO. 93. De Bono J, Mateo J, Fizazi K, et al. Olaparib for metastatic castration- 17.00316. resistant prostate cancer. N Engl J Med. 2020;382(22):2091e2102. 114. Cocco E, Scaltriti M, Drilon A. NTRK fusion-positive cancers and TRK 94. Savolitinib in treating participants with MET amplified metastatic or inhibitor therapy. Nat Rev Clin Oncol. 2018;15:731e747. unresectable colorectal cancer - full text view - ClinicalTrials.gov. 115. Pihlak R, Weaver JMJ, Valle JW, et al. Advances in molecular profiling Available at: https://clinicaltrials.gov/ct2/show/NCT03592641. and categorisation of pancreatic adenocarcinoma and the implica- Accessed July 7, 2020. tions for therapy. Cancers. 2018;10(1):17. 95. Mateo J, Carreira S, Sandhu S, et al. DNA-repair defects and olaparib 116. Zeitouni D, Pylayeva-Gupta Y, Der CJ, et al. KRAS mutant pancreatic in metastatic prostate cancer. N Engl J Med. 2015;373:1697e1708. cancer: no lone path to an effective treatment. Cancers. 2016;8(4): 96. Cortes-Ciriano I, Lee S, Park W-Y, Kim T-M, Park PJ. A molecular 45. portrait of microsatellite instability across multiple cancers. Nat 117. Heestand GM, Kurzrock R. Molecular landscape of pancreatic cancer: Commun. 2017;8:15180. implications for current clinical trials. Oncotarget. 2015;6:4553e4561. 97. Abida W, Cheng ML, Armenia J, et al. Microsatellite instability in 118. Payne S, Maher M, Tran N, et al. Mutant PIK3CA-mediated pancreatic prostate cancer and response to immune checkpoint blockade. J Clin tumorigenesis and the response to PI3K pathway inhibition. J Clin Oncol. 2018;36(15_suppl):5020. Oncol. 2015;33:e15273. 98. Abida W, Cyrta J, Heller G, et al. Genomic correlates of clinical 119. Hyman DM, Puzanov I, Subbiah V, et al. Vemurafenib in multiple outcome in advanced prostate cancer. Proc Natl Acad Sci. nonmelanoma cancers with BRAF V600 mutations. N Engl J Med. 2019;116(23):11428e11436. 2015;373:726e736. 99. de Bono JS, De Giorgi U, Rodrigues DN, et al. Randomized phase II 120. Azmi AS, Aboukameel A, Banerjee S, et al. MDM2 inhibitor MI-319 in study evaluating akt blockade with ipatasertib, in combination with combination with cisplatin is an effective treatment for pancreatic abiraterone, in patients with metastatic prostate cancer with and cancer independent of p53 function. Eur J Cancer. 2010;46(6):1122e without PTEN loss. Clin Cancer Res. 2019;25(3):928e936. 1131. 100. Ipatasertib plus abiraterone plus prednisone/prednisolone, relative to 121. Waddell N, Pajic M, Bailey P, et al. Whole genomes redefine the placebo plus abiraterone plus prednisone/prednisolone in adult male mutational landscape of pancreatic cancer. Nature. 2015;518:495. patients with metastatic castrate-resistant prostate cancer (IPATen- 122. Harder J, Ihorst G, Heinemann V, et al. Multicentre phase II trial of tial150). Available at: https://clinicaltrials.gov/ct2/show/NCT03 trastuzumab and capecitabine in patients with HER2 over- 072238. Accessed February 23, 2020. expressing metastatic pancreatic cancer. Br J Cancer. 2012;106(6): 101. Crumbaker M, Khoja L, Joshua AM. AR signaling and the PI3K pathway 1033e1038. in prostate cancer. Cancers. 2017;9(4):34. 123. Jones MR, Williamson LM, Topham JT, et al. NRG1 gene fusions are 102. The Cancer Genome Atlas Research Network. Comprehensive recurrent, clinically actionable gene rearrangements in KRAS wild- molecular characterization of gastric adenocarcinoma. Nature. type pancreatic ductal adenocarcinoma. Clin Cancer Res. 2014;513(7517):202e209. 2019;25(15):4674e4681. 103. Bang Y-J, Van Cutsem E, Feyereislova A, et al. Trastuzumab in com- 124. Singhi AD, Ali SM, Lacy J, et al. Identification of targetable ALK bination with chemotherapy versus chemotherapy alone for treat- rearrangements in pancreatic ductal adenocarcinoma. J Natl Compr ment of HER2-positive advanced gastric or gastro-oesophageal Canc Netw. 2017;15:555e562. junction cancer (ToGA): a phase 3, open-label, randomised controlled 125. Pishvaian MJ, Rolfo CD, Liu SV, Multani PS, Chow Maneval E, Garrido- trial. Lancet. 2010;376:687e697. Laguna I. Clinical benefit of entrectinib for patients with metastatic 104. Maron SB, Alpert L, Kwak HA, et al. Targeted therapies for targeted pancreatic cancer who harbor NTRK and ROS1 fusions. J Clin Oncol. populations: anti-EGFR treatment for EGFR amplified gastroesopha- 2018;36(4_suppl):521. geal adenocarcinoma. Cancer Discov. 2018;8(6):696e713. 126. Cancer Genome Atlas Research Network. Comprehensive and inte- 105. Lennerz JK, Kwak EL, Ackerman A, et al. MET amplification identifies a grative genomic characterization of hepatocellular carcinoma. Cell. small and aggressive subgroup of esophagogastric adenocarcinoma 2017;169:1327e1341. with evidence of responsiveness to crizotinib. J Clin Oncol. 2011;29: 127. Rimassa L, Assenat E, Peck-Radosavljevic M, et al. Tivantinib for 4803e4810. second-line treatment of MET-high, advanced hepatocellular carci- 106. Oxnard G, Subbiah V, Park K, et al. Clinical activity of LOXO-292, a noma (METIV-HCC): a final analysis of a phase 3, randomised, highly selective RET inhibitor, in patients with RET fusionþ non-small placebo-controlled study. Lancet Oncol. 2018;19:682e693. cell lung cancer. J Thorac Oncol. 2018;13(10):S349eS350. 128. Lim HY, Merle P, Weiss KH, et al. Phase II studies with refametinib or 107. Lee J, Ou S-HI, Lee JM, et al. Gastrointestinal malignancies harbor refametinib plus sorafenib in patients with RAS-mutated hepatocel- actionable MET exon 14 deletions. Oncotarget. 2015;6:28211e lular carcinoma. Clin Cancer Res. 2018;24:4650e4661. 28222. 129. Abou-Alfa GK, Macarulla Mercade T, Javle M, et al. ClarIDHy: a global, 108. Bang Y-J, Xu R-H, Chin K, et al. Olaparib in combination with paclitaxel in phase 3, randomized, double-blind study of ivosidenib (IVO) vs. pla- patients with advanced gastric cancer who have progressed following cebo in patients with advanced cholangiocarcinoma (CC) with first-line therapy (GOLD): a double-blind, randomised, placebo- an isocitrate dehydrogenase 1 (IDH1) mutation. Ann Oncol. controlled, phase 3 trial. Lancet Oncol. 2017;18(12):1637e1651. 2019;30(suppl_5):v851ev934. 109. Van Cutsem E, Bang Y-J, Mansoor W, et al. A randomized, open-label 130. Vogel A, Sahai V, Hollebecque A, et al. LBA40 - FIGHT-202: A phase II study of the efficacy and safety of AZD4547 monotherapy versus study of pemigatinib in patients (pts) with previously treated locally paclitaxel for the treatment of advanced gastric adenocarcinoma with advanced or metastatic cholangiocarcinoma (CCA). Ann Oncol. FGFR2 polysomy or gene amplification. Ann Oncol. 2017;28: 2019;30:v876. 1316e1324. 131. Marabelle A, Le DT, Ascierto PA, et al. Efficacy of pembrolizumab in 110. Loriot Y, Necchi A, Park SH, et al. Erdafitinib in locally advanced or patients with noncolorectal high microsatellite instability/mismatch metastatic urothelial carcinoma. N Engl J Med. 2019;381:338e348. repair-deficient cancer: Results from the phase II KEYNOTE-158 study. 111. The Cancer Genome Atlas Research Network. Integrated genomic J Clin Oncol. 2020;38(1):1e10. characterization of pancreatic ductal adenocarcinoma. Cancer Cell. 132. Wainberg ZA, Lassen UN, Elez E, et al. Efficacy and safety of dabra- 2017;32:185e203.e13. fenib (D) and trametinib (T) in patients (pts) with BRAF V600Ee

1504 https://doi.org/10.1016/j.annonc.2020.07.014 Volume 31 - Issue 11 - 2020 F. Mosele et al. Annals of Oncology

mutated biliary tract cancer (BTC): a cohort of the ROAR basket trial. 139. Szymanski J, Duncavage E, Pfeifer J. Next-generation sequencing J Clin Oncol. 2019;37(4_suppl):187. bioinformatics: Guidance between the sequencing and sign out. 133. Javle MM, Hainsworth JD, Swanton C, et al. Pertuzumab þ trastu- J Pathol Inform. 2018;9:23. zumab for HER2-positive metastatic biliary cancer: preliminary data 140. Burke W. Genetic tests: clinical validity and clinical utility. Curr Protoc from MyPathway. J Clin Oncol. 2017;35(4_suppl):402. Hum Genet. 2014;81:9.15.1e9.15.8. 134. Moore K, Colombo N, Scambia G, et al. Maintenance olaparib in 141. Kaul KL, Leonard DGB, Gonzalez A, et al. Oversight of genetic testing: patients with newly diagnosed advanced ovarian cancer. N Engl J an update. J Mol Diagn. 2001;3(3):85e91. Med. 2018;379(26):2495e2505. 142. Deans ZC, Costa JL, Cree I, et al. Integration of next-generation 135. Clynick B, Dessauvagie B, Sterrett G, et al. Genetic characterisation of sequencing in clinical diagnostic molecular pathology laboratories molecular targets in carcinoma of unknown primary. J Transl Med. for analysis of solid tumours; an expert opinion on behalf of IQN Path 2018;16(1):185. ASBL. Virchows Arch. 2017;470(1):5e20. 136. Roy S, Coldren C, Karunamurthy A, et al. Standards and guidelines 143. Matthijs G, Souche E, Alders M, et al. Guidelines for diagnostic next- for validating next-generation sequencing bioinformatics pipelines: a generation sequencing. Eur J Hum Genet. 2016;24(1):2e5. joint recommendation of the Association for Molecular Pathology 144. Jennings LJ, Arcila ME, Corless C, et al. Guidelines for validation and the College of American Pathologists. J Mol Diagn. 2018;20(1): of next-generation sequencing-based oncology panels: a joint 4e27. consensus recommendation of the Association for Molecular 137. Hume S, Nelson TN, Speevak M, et al. CCMG practice guideline: Pathology and College of American Pathologists. J Mol Diagn. laboratory guidelines for next-generation sequencing. J Med Genet. 2017;19(3):341e365. 2019;56(12):792e800. 145. Aziz N, Zhao Q, Bry L, et al. College of American Pathologists’ labo- 138. Next Generation Sequencing (NGS) Worksheets. College of American ratory standards for next-generation sequencing clinical tests. Arch Pathologists. Available at: https://www.cap.org/member-resources/ Pathol Lab Med. 2015;139(4):481e493. precision-medicine/next-generation-sequencing-ngs-worksheets. Accessed March 10, 2020.

Volume 31 - Issue 11 - 2020 https://doi.org/10.1016/j.annonc.2020.07.014 1505 Cancer Diagnostics and Molecular Pathology

Multicenter Comparison of Molecular Tumor Boards in The Netherlands: Definition, Composition, Methods, and Targeted Therapy Recommendations

a b c,d c g h BART KOOPMAN , HARRY J.M. GROEN, MARJOLIJN J.L. LIGTENBERG, KATRIEN GRÜNBERG, KIM MONKHORST, ADRIANUS J. DE LANGEN, g i j j k g,k MIRJAM C. BOELENS, MARTHE S. PAATS, JAN H. VON DER THÜSEN, WINAND N.M. DINJENS, NIENKE SOLLEVELD, TOM VAN WEZEL, l m n n c e f HANS GELDERBLOM, LIZZA E. HENDRIKS, ERNST-JAN M. SPEEL, TOM E. THEUNISSEN, LEONIE I. KROEZE, NIVEN MEHRA, BERBER PIET, b a a a,o o o ANTHONIE J. VAN DER WEKKEN, ARJA TER ELST, WIM TIMENS, STEFAN M. WILLEMS, RUUD W.J. MEIJERS, WENDY W.J. DE LENG, p q r q a a ANNE S.R. VAN LINDERT, TEODORA RADONIC, SAYED M.S. HASHEMI, DANIËLLE A.M. HEIDEMAN, ED SCHUURING, LÉON C. VAN KEMPEN aDepartment of Pathology and Medical Biology and bDepartment of Pulmonary Diseases, University of Groningen, University Medical Center Groningen, Groningen, The Netherlands; Departments of cPathology, dHuman Genetics, eMedical Oncology, and fPulmonary Diseases, Radboud University Medical Center, Nijmegen, The Netherlands; Departments of gPathology and hThoracic Oncology, Netherlands Cancer Institute, Amsterdam, The Netherlands; Departments of iPulmonary Medicine and jPathology, Erasmus Medical Center, University Medical Center Rotterdam, Rotterdam, The Netherlands; Departments of kPathology and lMedical Oncology, Leiden University Medical Center, Leiden, The Netherlands; Departments of mPulmonary Diseases and nPathology, GROW-School for Oncology and Developmental Biology, Maastricht University Medical Center, Maastricht, The Netherlands; Departments of oPathology and pPulmonology, University Medical Center Utrecht, Utrecht, The Netherlands; Departments of qPathology and rPulmonary Diseases, Cancer Center Amsterdam, Amsterdam University Medical Center, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands Disclosures of potential conflicts of interest may be found at the end of this article. Key Words. Molecular tumor board • Rare mutations • Molecular diagnostics • Decision making • Multidisciplinary

ABSTRACT

Background. Molecular tumor boards (MTBs) provide rational, Results. Interviews with 24 members of eight MTBs rev- genomics-driven, patient-tailored treatment recommenda- ealed that all participating MTBs focused on rare or com- tions. Worldwide, MTBs differ in terms of scope, composition, plex mutational cancer profiles, operated independently of methods, and recommendations. This study aimed to assess cancer type–specific multidisciplinary teams, and consisted differences in methods and agreement in treatment recom- of at least (thoracic and/or medical) oncologists, patholo- mendations among MTBs from tertiary cancer referral centers gists, and clinical scientists in molecular pathology. Differ- in The Netherlands. ences were the types of cancer discussed and the methods Materials and Methods. MTBs from all tertiary cancer refer- used to achieve a recommendation. Nevertheless, agree- ral centers in The Netherlands were invited to participate. A ment among MTB recommendations, based on identified survey assessing scope, value, logistics, composition, decision- actionable molecular alteration(s), was high for the 10 evalu- making method, reporting, and registration of the MTBs was ated cases (86%). completed through on-site interviews with members from Conclusion. MTBs associated with tertiary cancer referral each MTB. Targeted therapy recommendations were com- centers in The Netherlands are similar in setup and pared using 10 anonymized cases. Participating MTBs were reach a high agreement in recommendations for rare or asked to provide a treatment recommendation in accordance complex mutational cancer profiles. We propose a with their own methods. Agreement was based on which “Dutch MTB model” for an optimal, collaborative, and molecular alteration(s) was considered actionable with the nationally aligned MTB workflow. The Oncologist next line of targeted therapy. 2020;25:1–12

Correspondence: Léon C. van Kempen, Ph.D., Department of Pathology and Medical Biology, University of Groningen, University Medical Center Groningen, P.O. Box 30.001, 9700RB Groningen, The Netherlands. Telephone: 31-0-50-3615129; e-mail: [email protected] Received May 26, 2020; accepted for publication September 25, 2020. http://dx.doi.org/10.1002/onco.13580 This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adapta- tions are made.

The Oncologist 2020;25:1–12 www.TheOncologist.com © 2020 The Authors. The Oncologist published by Wiley Periodicals LLC on behalf of AlphaMed Press. 2 Molecular Tumor Boards in The Netherlands

Implications for Practice: Interpretation of genomic analyses for optimal choice of target therapy for patients with cancer is becoming increasingly complex. A molecular tumor board (MTB) supports oncologists in rationalizing therapy options. How- ever, there is no consensus on the most optimal setup for an MTB, which can affect the quality of recommendations. This study reveals that the eight MTBs associated with tertiary cancer referral centers in The Netherlands are similar in setup and reach a high agreement in recommendations for rare or complex mutational profiles. The Dutch MTB model is based on a collaborative and nationally aligned workflow with interinstitutional collaboration and data sharing.

INTRODUCTION molecular pathology laboratories. MTBs from all academic medical centers and one nonacademic tertiary cancer refer- The emergence of DNA- and RNA-based molecular cancer ral center were invited to participate. profiling techniques for predictive testing in the routine diag- nostic setting has rapidly expanded the diagnostic guidance for targeted therapies [1]. Molecular tumor boards (MTB) Assessment of Similarities and Differences Between support treating physicians in understanding the increasing MTB Methods complexity of molecular testing results, providing rational, All MTBs were invited to engage in one or more on-site inter- genomics-driven, patient-tailored treatment recommenda- views. A survey to assess similarities and differences among tions with respect to the currently available targeted drugs MTBs was designed, covering scope and perceived value, [2, 3]. MTBs are hosted in cancer centers that offer extensive logistics and composition, decision-making method, reporting, molecular profiling techniques [4–18]. Although the MTBs andregistrationofMTBcases,andtheinterviewees’ view on that have thus far published their methods all focus on trans- harmonization and collaboration among MTBs (supplemental lating molecular testing results into a therapeutic recommen- online Methods). The MTBs were asked to select inter- dation, there are major differences in terms of scope, viewees representing multiple disciplines active in their MTB. composition, and methods [2, 3]. Notably, treatment recom- Interviews and attendances of MTBs were performed by a mendations provided by MTBs seem to vary widely. A recent medical researcher (B.K.) between June and September 2019. comparison of five independent MTBs from four countries All interviewees consented to participation. revealed that only two of five MTBs provided similar recom- mendations for four fictional cases with complex mutational Comparison of Targeted Therapy Recommendations profiles [19]. Regional and international differences in com- MTB recommendations were compared using 10 cases repre- position and logistics [2], method and prioritization [19], sentative of the MTBs’ setting in terms of cancer type, muta- access to targeted drugs [3, 16], and molecular diagnostic tional profile, and inquiry. The participating MTBs were asked workup [2] are potential sources of heterogeneity. to submit an anonymized case (supplemental online Methods). The incorporation of MTBs into standard-of-care cancer The submitted cases were adjusted for further anonymization diagnostics necessitates mitigation of heterogeneity in MTB and formatting. Finalized cases were prepared as two-page recommendations. Although perfect agreement in treatment documents (example provided in the supplemental online recommendations may not be achievable because they are tai- Methods). lored to the patient at hand and dependent on drug and trial From September 26 until November 21, 2019, 10 cases availability, discrepancies in target identification might be were sent to participating MTBs. The MTBs were asked to averted. For this purpose, the Predictive Analysis for Therapy handle these as routine MTB requests and provide recom- project was initiated, which aims to optimize patient access mendations according to their usual method. After the ini- to personalized cancer therapy in The Netherlands [20]. This tial round, two cases with low rates of agreement with includes optimizing MTBs by providing directives on the mini- respect to the choice of inhibitor or its timing of use were mal requirements for hosting an MTB and achieving a recom- selected. These cases were sent for a second time to the mendation, promoting the exchange of knowledge among participants along with the recommendations received in MTBs through a shared database, and ensuring accessibility the initial round for those cases, in a blinded fashion. The to MTBs for community hospitals and laboratories. participating MTBs were asked if they would revise their ini- To identify the prerequisites for reaching a well-informed tial answer when presented with the other MTBs’ recom- MTB recommendation, this study aims to assess current simi- mendations and provide arguments for their decision. larities and differences in MTB methods and, secondly, to The research ethics board of the department of Pathology determine agreement in treatment recommendations among at the University Medical Center Groningen (UMCG) approved MTBs from tertiary cancer referral centers. the use of anonymous case descriptions for this study. The study protocol was consistent with the UMCG Research Code and national ethical and professional guidelines [21, 22].

MATERIALS AND METHODS Statistical Analysis MTBs Included in Analysis Descriptive statistics are provided. The consensus for a case MTBs associated with tertiary cancer referral centers were was defined as the most frequently provided recommenda- identified through a digital survey among all Dutch tion. Agreement was measured by the percentage of MTBs

© 2020 The Authors. The Oncologist published by Wiley Periodicals LLC on behalf of AlphaMed Press. Koopman, Groen, Ligtenberg et al. 3

Table 1. MTB demographics as attained by on-site interviews Demographic MTBs, n (%) MTBs participating in the on-site interviews 8 (100%) Type of hospital that the MTB is associated with University medical center (academic) 7 (87.5) Nonacademic tertiary cancer referral hospital 1 (12.5) Year in which the MTB was established 2014 (active for 6 years) 1 (12.5) 2015 (active for 5 years) 3 (37.5) 2016 (active for 4 years) 2 (25) 2017 (active for 3 years) 2 (25) Cancer types eligible for review by the MTB Any type of cancer 5 (62.5) Thoracic oncology 3 (37.5) Frequency of MTB meetings MTB meets once every week 4 (50) MTB meets once in every 2 weeks 4 (50) Internal reporting of recommendation Report in the patient’s electronic health record 7 (87.5) Recommendation is included in the pathology report 1 (12.5) Communication of recommendation to external applicants Directly to applicant (videoconferencing, e-mail, telephone call) 5 (62.5) By means of a medical letter 2 (25) Directly to applicant by videoconferencing and through pathology report 1 (12.5) Registration of cases reviewed by the MTB Cases, recommendations, and follow-up are registered in a local database 2 (25) Only basic case information is registered in a local database 2 (25) No registration in local database 4 (50) Regional collaboration by the MTB Cases from peripheral hospitals are reviewed 7 (87.5) Cases from other tertiary cancer referral centers are reviewed 2 (25) External specialists attend MTB meetings through videoconferencing 4 (50) Experts participating in MTB meetings Clinical scientists in molecular pathology 8 (100) Pathologists 8 (100) Thoracic oncologists 8 (100) Medical oncologists 5 (62.5) Postdocs, PhD students, researchers 5 (62.5) Clinical geneticists 3 (37.5) Clinical chemists 2 (25) Laboratory technicians 2 (25) Hemato-oncologists 1 (12.5) Nurse practitioners 1 (12.5) Pharmacists 1 (12.5) Radiation oncologists 1 (12.5) Structural biologists 1 (12.5) Abbreviation: MTB, molecular tumor board. that provided a recommendation in accordance with this actionable with the next line of targeted therapy (next action- consensus. Agreement among MTB recommendations was able target). The agreement rate was calculated from the ini- based on which molecular alteration(s) was considered tial recommendations. www.TheOncologist.com © 2020 The Authors. The Oncologist published by Wiley Periodicals LLC on behalf of AlphaMed Press. 4 Molecular Tumor Boards in The Netherlands

Figure 1. Online information resource usage for MTB decision making. Bar graphs depicting the usage of online resources for deci- sion making within MTBs. A survey was filled in by representatives of seven MTBs. (A): Data- and knowledge bases for somatic vari- ant calling and/or interpretation. (B): Data- and knowledge bases for germline variant calling and/or interpretation. (C): Online resources for genomic sequences. (D): Online resources for scientific literature. (E): Online resources for oncology guidelines. (F): Trial registries. Abbreviations: CIViC, Clinical Interpretation of Variants in Cancer; COSMIC, Catalogue of Somatic Mutations in Cancer; dbSNP, Short Genetic Variations database; DGIdb, Drug Gene Interaction Database; ESMO, European Society for Medical Oncology; ExAC, Exome Aggregation Consortium; JAX CKB, the Jackson Laboratory Clinical Knowledge Base; OncoKB, Precision Oncology Knowledge Base; PCT MD Anderson, Personalized Cancer Therapy Knowledge Base; MTB, molecular tumor board.

RESULTS Institute, Radboud University Medical Center, University Medi- cal Center Groningen, and University Medical Center Utrecht. MTBs Included in the Analysis On-site interviews were held with 24 MTB members: nine clin- MTBs from all eight Dutch tertiary cancer referral centers were ical scientists in molecular pathology (CSMPs), six pathologists, included, representing Amsterdam University Medical Centers, five thoracic oncologists, two medical oncologists, and two Erasmus Medical Center, Leiden University Medical Center, (bio)medical researchers. Meetings of MTBs were attended Maastricht University Medical Center, Netherlands Cancer when this could be combined with the on-site interview.

© 2020 The Authors. The Oncologist published by Wiley Periodicals LLC on behalf of AlphaMed Press. Koopman, Groen, Ligtenberg et al. 5

Figure 2. The Dutch MTB model. Flow diagram depicting an optimal MTB workflow as recommended by the eight molecular tumor boards participating in this study. Responsibilities of each party are annotated for external (gray) or MTB-associated (black) physi- cians/oncologists (stethoscopes) and CSMPs/pathologists (microscopes). All four parties can submit molecular-oriented questions about their cases to the MTBs. The MTB-associated oncologist is responsible for the clinical case preparation, and the CSMP and pathologist are jointly responsible for characterization of the molecular profile. During review in an MTB meeting, a diagnostic and/or therapeutic recommendation is formulated, which is communicated to the requestor, recorded in the patient’s electronic health record, and registered in a local database. The requestor can then use this recommendation in their choice of (molecular) tests or (targeted) therapy. aIn case of biomarkers with both germline and somatic implications, attendance of a clinical geneticist is recommended. In case of discussing whole-exome or whole-genome sequencing results, attendance of a bioinformatician is rec- ommended. Abbreviations: CSMP, clinical scientist in molecular pathology; MTB, molecular tumor board.

www.TheOncologist.com © 2020 The Authors. The Oncologist published by Wiley Periodicals LLC on behalf of AlphaMed Press. 6 Molecular Tumor Boards in The Netherlands

Assessment of Similarities and Differences Between 1000 Genomes Browser [25], cBioPortal [26], ClinVar [27], MTB Methods COSMIC [28], dbSNP [29], Ensembl [30], JAX-CKB [31], OncoKB [32], and PubMed [33]. Trial overviews were not sys- Scope and Value of the MTBs tematically consulted prior to the MTB meetings because of The eight MTBs were founded between 2014 and 2017 time restrictions. In some MTBs, dedicated trial-coordinating (Table 1). Reciprocal improvement of expertise for attendees oncologists were consulted after evaluation of actionability was considered a core value of an MTB and an important by the MTB. Interviewees regarded awareness of the trial reason for founding MTBs. The MTBs reviewed cases with availability variable and dependent on which oncologist(s) molecular-oriented questions, such as the clinical conse- attended meetings. quences of molecular findings, the availability of therapeutic The MTBs served as a unifying platform to achieve a options, or the most appropriate test(s) to perform for a profile-based, patient-tailored consensus recommendation case. Seven MTBs reviewed only selected rare or complex based on the identification/prioritization of genomic alter- cases (two to eight cases average per meeting); one MTB ations and potential drug actionability (prepared by CSMPs/ reviewed all molecular testing results (estimated 5–15 cases pathologists) and the assessment of availability of clinical per meeting). The most common molecular findings eligible trials or compassionate use drugs for eligible patients (pre- for review included somatic mutations, copy number vari- pared by oncologists). A recommendation could be diagnos- ants, and fusion genes detected by targeted panel next- tic (such as which additional tests to perform), therapeutic, generation sequencing, fluorescence in situ hybridization, or both (Fig. 2). immunohistochemistry, or RNA-based fusion transcript analy- sis. Results from whole-genome sequencing (WGS) or whole- Reporting and Registration of MTB Cases exome sequencing (WES) were uncommon, as these For reviewed cases, MTBs created a report in the individual methods were not used routinely at the time of this study, patient’s electronic health record (seven MTBs) or in the although all MTBs were open to interpret these results. pathology report (one MTB). Recommendations to external Five MTBs reviewed any type of cancer (“cancer agnos- applicants were communicated through videoconferencing, tic”), whereas three MTBs only reviewed thoracic oncology. e-mail, an official written letter, or the pathology report. Thoracic oncology cases were most common in six MTBs Four MTBs maintained a local database for registration of because of the diversity of relevant actionable biomarkers new cases and searching prior cases. and because the Dutch national guideline for non-small cell lung cancer (NSCLC) recommends referral to an MTB in case View on Harmonization and Collaboration fi of rare mutational pro les [23]. The most common reason Between MTBs for reviewing other cancer types was to assess eligibility for Two MTBs occasionally received inquiries from other ter- trials such as the Drug Rediscovery Protocol (DRUP) [24]. tiary referral centers harboring an MTB. As personal experi- All MTBs took place as individual meetings, outside of ence with a similar case, including treatment results, was – fi conventional, cancer type speci c multidisciplinary team considered an important factor in rationalizing a recom- (MDT) meetings. Differentiation between primary treatment mendation, facilitation of access to data on comparable modalities was considered the responsibility of the conven- cases from other MTBs was acknowledged as a valuable fi tional MDT, whereas MTBs were considered more pro cient addition to their toolset. in guiding decision making on the most appropriate treat- ment when targetable alterations are present. Two thoracic Comparison of Targeted Therapy Recommendations oncology MTBs were hosted sequentially with a conventional All eight MTBs participated in a study to compare MTB rec- MDT, selecting cases for either meeting depending on the ommendations based on a selection of anonymized cases. cancer stage. Six MTBs were not tuned to conventional Each MTB submitted one case description. The composition MDTs, although attending oncologists and pathologists were of cases was based on information gathered in the inter- also involved in conventional MDTs. views: six NSCLC cases, two melanoma cases, one colorectal cancer (CRC) case, and one gastrointestinal stromal tumor Logistics and Composition of the MTBs (GIST) case (supplemental online Table 1). In all MTBs, cases could be submitted by the treating physi- As three MTBs were restricted to thoracic oncology, cian, CSMP, or pathologist. Participants in all MTBs included 68 answers were expected (six NSCLC cases with eight oncologists (thoracic, medical, or hemato-oncologists, responses each and four non-NSCLC cases with five answers depending on the cancer type reviewed), CSMPs, and each). One MTB was unable to review the last three non- pathologists. Other common attendees were (bio)medical NSCLC cases because of lack of availability of a medical fi researchers ( ve MTBs) and clinical geneticists (three MTBs, oncologist, leaving 65 responses eligible for analysis. of which two consulted geneticists on request) (Table 1). Seven MTBs received inquiries from peripheral hospitals or Agreement Between MTB Recommendations pathology laboratories, of which four facilitated attendance The agreement between MTBs in identifying the foremost of external specialists through videoconferencing. actionable target ranged between 60% and 100% (Table 2). Seven of eight MTBs (87.5%) identified BRAF/MEK as Decision-Making Method of the MTBs actionable targets in case 1A (osimertinib-resistant NSCLC, In preparing a case, a variety of (online) resources were con- resistance by BRAF p.(V600E)). The majority of MTBs rec- sulted (Fig. 1). Nine resources were used by all MTBs: the ommended osimertinib/dabrafenib/trametinib combination

© 2020 The Authors. The Oncologist published by Wiley Periodicals LLC on behalf of AlphaMed Press. www.TheOncologist.com Table 2. Agreement between MTB recommendationsa al. et Ligtenberg Groen, Koopman, Answers Next actionable Case ID Cancer type Setting Mutational profile received Recommendations targetb Agreementc 1A NSCLC Resistance to EGFR p. n = 8 BRAF/MEK inhibition (dabrafenib + trametinib) and BRAF/MEK (n =7) 87.5% osimertinib (L747_A750delinsP) osimertinib, in combination or sequential treatment No target (n =1) BRAF p.(V600E) (n =6) Switch to chemotherapy; if that is not an option, BRAF/ MEK inhibition (dabrafenib + trametinib) (n =1) Switch to chemotherapy (n =1) 1B CRC Primary stage IV KRAS p.(L19F) n = 5 Do NOT treat with anti-EGFR antibodies (n =3) No target (n =3) 60% diagnosis No contraindication for anti-EGFR antibodies in current EGFR (n =2) guidelines (n =2) 2A Melanoma Exhaustion of BRAF p.(G464E) n = 4 First exhaust treatment options with immunotherapy, then CDK4/6 (n =3) 75% standard NRAS p.(Q61K) treat with palbociclib within the DRUP trial (if biallelic MEK (n =1) treatment CDKN2A p.(R80*) CDKN2A inactivation is proven) (n =2) options Treat with palbociclib within the DRUP trial (if biallelic CDKN2A inactivation is proven) (n =1) h nooitpbihdb ie eidcl L nbhl fApae Press. AlphaMed of behalf on LLC Periodicals Wiley by published Oncologist The MEK inhibition in compassionate use (n =1) 2B NSCLC Primary stage IV EGFR p.(G719A) n = 8 Treat with agent that may inhibit both mutations, such as EGFR/ERBB2 75% diagnosis ERBB2 p.(S310F) afatinib (n =5) (n =6) Switch to chemotherapy; treat with afatinib at progression EGFR (n =1) (n =1) ERBB2 (n =1) Treat with erlotinib; relevance of ERBB2 mutation is unknown (n =1) ERBB2 mutation may induce resistance; consider lapatinib (ERBB2 inhibitor) (n =1) 3A NSCLC Resistance to EGFR p. n = 8 EGFR inhibitor (osimertinib) combined with a MEK MEK (n =7) 87.5% osimertinib (E746_A750del) inhibitor (trametinib), preferentially within a clinical trial RAF (n =1) TP53 p.(P278R) (if available) (n =6) TRIM24-BRAF fusion Consider treatment with MEK inhibitor (n =1) gene No other options than chemotherapy. Pan-RAF inhibitor such as sorafenib may be beneficial, but no options for trials (n =1) 3B Colorectal Exhaustion of MET p.(R988C) n = 4 Anti-ERBB2 antibodies (trastuzumab + pertuzumab) within ERBB2 (n = 4) 100% cancer standard TP53 p.(R158H) the DRUP trial (n =4) treatment TP53 p.(R196*) options ERBB2 amplification 4A GIST Primary stage IV BRAF p.(V600E) n = 4 BRAF/MEK inhibition (dabrafenib + trametinib or BRAF/MEK (n = 4) 100% diagnosis vemurafenib + cobimetinib) within the DRUP trial (n =4) 4B NSCLC Primary stage IV KRAS p.(A146T) n = 8 Depending on PD-L1, immunotherapy or chemotherapy No targets (n =7) 87.5% 00TeAuthors. The 2020 © diagnosis BRAF p.(G469A) combined with immunotherapy (n =7)d RAF or MEK CDKN2A p.(H83Y) No targets available. Additional diagnostics for revision of (n =1)d TERT promoter: C228T diagnosis because of unexpected combination of mutations (n =1) 5A NSCLC Resistance to EGFR p. n = 8 Continue osimertinib; combine with MET inhibitor (n =4) MET (n =7) 87.5% osimertinib (E746_A750del) MET inhibitor or chemotherapy combined with EGFR (n =1) TP53 p.(F328Ifs*18) immunotherapy (n =1) (continued) 7 8 Molecular Tumor Boards in The Netherlands

c therapy. In case 1B (CRC, KRAS p.(L19F)), all five MTBs agreed that the mutation could potentially inhibit response to anti-EGFR antibodies, but only three MTBs (60%) consid- ered the evidence sufficient to recommend against anti- Agreement EGFR antibodies. Forcase2A(melanoma,mutationsinBRAF, NRAS,and CDKN2A), three of four MTBs (75%) recognized CDK4/6 as actionable targets of palbociclib within the DRUP trial (if CDKN2A biallelic inactivation was proven). Two MTBs rec- = 8) 100%

b ommended immunotherapy first, in line with the national n ( NSCLC guideline. In case 2B (NSCLC, mutations in EGFR and ALK Next actionable target ERBB2), six of eight MTBs (75%) acknowledged the potential inhibitory effect of afatinib on cancers harboring the ERBB2 mutation—thus identifying both EGFR and ERBB2 as action- able by afatinib—although one MTB recommended chemo- =7) therapy prior to treatment with afatinib. n In case 3A (osimertinib-resistant NSCLC, resistance cau- TRIM24 BRAF

, non-small cell lung cancer. sed by a - fusion), seven of eight MTBs (87.5%) recognized MEK as a target. The majority of MTBs rec- ommended osimertinib/trametinib combination treatment. Case 3B (CRC, ERBB2 amplification) achieved 100% agree- ment: inclusion in the DRUP trial for treatment with anti- =1)

n ERBB2 antibodies. One hundred percent agreement was also achieved for BRAF

=1) case 4A (GIST, p.(V600E)): inclusion in the DRUP trial

n for treatment with BRAF/MEK inhibition. In case 4B (NSCLC, =1)

n mutations in KRAS, BRAF, CDKN2A, and the TERT promoter region), seven of eight MTBs (87.5%) did not identify a tar- get. Only one MTB commented on the potential actionability of the BRAF non-V600 mutation with a MEK inhibitor in a =1) n lorlatinib at progression ( combination of MET inhibitorprogression and ( EGFR inhibitor at ( (necitumumab) ( late line of treatment. Seven MTBs recommended immuno- Brigatinib within an available trial or chemotherapy; Chemotherapy combined with immunotherapy; treat with Chemotherapy; treat with MET inhibitor at progression Continue osimertinib; combine with anti-EGFR antibodies therapy or chemotherapy combined with immunotherapy, all noting the omitted PD-L1 status. For case 5A (osimertinib-resistant NSCLC, mutations in EGFR and TP53 and amplification of MET and EGFR), seven = 65 Average: 86% = 8 Lorlatinib (or brigatinib within an available trial) ( n n Answers received Recommendations MTBs (87.5%) acknowledged MET as the actionable target in the next or subsequent line of targeted treatment. However, the prioritization of treatment options varied, with some MTBs preferring exhaustion of treatment options with che- le

fi motherapy and disagreement on continuation of EGFR inhi- cation

cation – ALK fi bition.Incase5B(ALKinhibitor resistant NSCLC, fusion, fi ALK p.(G1202R)), all MTBs (100%) acknowledged that ALK ampli

ampli was still actionable with a third-generation ALK inhibitor. fusion gene p.(G1202R) Lorlatinib was recommended by seven of eight MTBs. ALK ALK MET EGFR Overall, among 65 MTB recommendations on the fore- most actionable target for these 10 representative cases, the overall agreement was 86% (56 of 65 responses) after ed consensus recommendation used to determine agreement is indicated in bold. fi the first round of recommendations.

Revision of Cases with a Low Rate of Agreement

alectinib Of cases with responses from all MTBs, 2B and 5A had the lowest rates of agreement with respect to the choice of inhibitor or its timing of use. These cases were sent back to ’

rst recommendation; the provided percentage does not include recommendations after revision. the MTBs along with the other (blinded) MTBs recommen- fi dations. For case 2B (NSCLC, mutations in EGFR and ERBB2), two MTBs changed their answer based on the other MTBs’

(continued) motivations, acknowledging the provided evidence for targeting ERBB2 with afatinib. For case 5A (NSCLC with muta- tions in EGFR and TP53 and amplifications of MET and EGFR), For the next actionable target, the identi One MTB suggested targeting RAF or MEK at a later line of therapy; the other MTBs did not identify a target. Full description of cases available in supplemental online Table 1. Agreement after a b c d Abbreviations: CRC, colorectal cancer; DRUP, Drug Rediscovery Protocol; GIST, gastrointestinal stromal tumor; MTB, molecular tumor board; NSCLC Total: 5B NSCLC Resistance to Table 2. Case ID Cancer type Setting Mutational pro one MTB changed its answer to include continued targeting

© 2020 The Authors. The Oncologist published by Wiley Periodicals LLC on behalf of AlphaMed Press. Koopman, Groen, Ligtenberg et al. 9

Table 3. Definition of a molecular tumor board according to observed value of an MTB, as well as its place within the the Dutch MTB model health care process, vary. Some MTBs assess eligibility for clini- A An MTB is the primary source of information for caltrials[5].Othersalsoinclude surgeons, radiation oncolo- interpreting (rare or complex) molecular diagnostic results gists, and radiologists and are effectively expanded in oncology and serves as a reciprocal educative platform conventional MDTs [8, 16]. In this setup, complex molecular for clinicians, oncologists, and molecular pathology cases are discussed in addition to the nonmolecular cases typi- specialists. cally discussed in conventional cancer type–specificMDTs. B An MTB focuses on differentiating between targeted MTBs adhering to the Dutch MTB model, however, review therapeutic options (standard of care, within a clinical trial, or off label) or other therapeutic options based on (complex) molecular cases only, differentiate between treat- molecular diagnostic results. ment options based on molecular alteration(s), and indicate C An MTB can be cancer agnostic and operates whether the provided (targeted) therapy recommendation is independently of the conventional cancer type–specific standard of care, within a clinical trial, or off label. This is simi- MDTs but features oncologists and pathologists who also lar to other previously published MTBs [4, 6, 7, 9, 10, 13, 15, participate in conventional MDTs. 18]. The main reason for not incorporating complex molecular D An MTB features at least (a) oncologists (thoracic, medical results into an “enhanced” conventional MDT is the highly or hematological, depending on the type of cancer discussed), (b) (molecular-oriented) pathologists, and (c) specialized and complex nature of the MTB. A low case load, clinical scientists in molecular pathology. A clinical enabled by selectivity in the cases that are discussed, allows geneticist is recommended in case of discussing time to elaborately discuss biological and clinical consequences biomarkers with both germline and somatic implications. of molecular results for a patient. This leads to a reciprocal E An MTB is hosted in a tertiary cancer referral center and is improvement of expertise of all parties involved on the clinical open for participation by experts from peripheral hospitals and pathology laboratories, preferably with access through consequences of molecular test results and the biological videoconferencing. It is a responsibility of the MTB’s rationalization of targeted therapy options. In addition, the network to ensure that all patients in its region have possibility to organize an MTB in a cancer-agnostic fashion access to an MTB recommendation. allows for the interdisciplinary exchange of molecular diagnos- F An MTB ensures its recommendation is accessible in the tic and therapeutic knowledge, as well as the proper assess- ’ patient s electronic health record. ment of eligibility for trials that are not limited to a single type G An MTB maintains a local registry of reviewed cases, of cancer. preferentially with systematic follow-up of cases within the The Dutch MTB model encompasses a comprehensive MTB to evaluate effectiveness of the recommendations. approach toward a case, featuring the identification/priori- Abbreviations: MDT, multidisciplinary team; MTB, molecular tumor board. tization of genomic alterations, potential drug actionability, assessment of drug availability, and the patient’s eligibility for a targeted treatment (Fig. 2). MTBs feature experts that of EGFR with osimertinib with its previous recommendation are directly involved in differentiation between targeted for targeting MET. In addition, one MTB changed its answer therapy options: oncologists (thoracic, medical, or hemato- because a trial combining osimertinib with MET inhibitor logical, depending on the cancer type discussed), tepotinib had opened in its center. molecular-oriented pathologists, and CSMPs. These are con- In total, four MTBs changed their initial answer for case sidered the minimal attendees for reaching a consensus 2B or 5A, either based on evidence provided by other MTBs expert recommendation. Treating physicians, pathologists, or because new center-specific treatment possibilities had and CSMPs can all submit cases. The CSMP has an essential become available. role with the primary responsibility to interpret genomic variants in routine molecular diagnostics [34]. The atten- dance of CSMPs is a major distinguishing factor from con- DISCUSSION ventional MDTs, which was also the case for multiple This study describes similarities and differences between previously published MTBs [9–12, 14, 16–18]. eight MTBs from tertiary cancer referral centers in The In addition to these three core members, a majority of Netherlands. Despite differences in terms of scope and previously published MTBs also incorporate geneticists or methods, MTBs had similar compositions of experts and bioinformaticians [5, 7, 8, 10–17]. In a previous effort to reached a high level of agreement in identifying actionable assess differences between MTBs in The Netherlands, targets. Based on these results and the perspectives of par- attendance of geneticists and bioinformaticians was rec- ticipating MTBs, a consensus was formed on an optimized, ommended when discussing large-scale sequencing results, collaborative workflow for an MTB hosted by a tertiary can- such as WES and WGS [2]. However, geneticists were mem- cer referral center (Fig. 2). This workflow, which we desig- bers of only three out of eight MTBs in our study, and nated the “Dutch MTB model” (Table 3), integrates the bioinformaticians were absent. This is because current similarities identified in this study and was approved by rep- MTBs rarely discuss WES/WGS, as these techniques are at resentatives of the eight participating MTBs. present not performed in routine molecular diagnostics. One of the MTBs from this study had previously published Thus, although geneticists and bioinformaticians are valu- on the treatment outcome resulting from their workflow [4]. able, we do not currently consider them mandatory. For Several individual MTBs in other countries have also reported now, consultation on demand is deemed sufficient, unless their scope, methods, recommendations, and treatment out- the MTB frequently reviews biomarkers with germline comes [5, 7–18]. Comparing these MTBs, the scope and implications, such as BRCA1 and BRCA2 (geneticists), or www.TheOncologist.com © 2020 The Authors. The Oncologist published by Wiley Periodicals LLC on behalf of AlphaMed Press. 10 Molecular Tumor Boards in The Netherlands biomarkers detected with WES/WGS such as complex geno- performance of MTBs. We consider data sharing between mic rearrangements (bioinformaticians). MTBs and regularly comparing treatment recommendations for specific complex cases key in achieving this. Assessment and Harmonization of MTB The final treatment recommendations (Table 2) were Recommendations more heterogeneous than represented by the calculated Although our results reveal some differences in the methods agreement for the identification of the next actionable tar- used to reach a recommendation, all MTBs operate in accor- get for three major reasons. First, the prioritization of dance with the Dutch MTB model presented here. To evaluate targeted therapy versus other treatment options differed: the agreement in treatment recommendations from MTBs in cases 1A, 2A, 2B, and 5A, several MTBs recommended adhering to the Dutch MTB model, we compared recommen- nontargeted treatment modalities prior to targeted therapy. dations provided for 10 typical complex MTB cases. A prior Second, there were differences in the exact inhibitor rec- study by Rieke and colleagues used a comparable setup, com- ommended. Drugs were available both within or outside of paring five MTBs from four countries based on four fictional clinical trials (for example, ALK inhibitors in case 5B), and cases [19]. The authors found a poor agreement, with compa- MTBs tended to recommend the trial available in their own rable recommendations from two out of five MTBs (40%) for centers. Finally, there were slight differences in whether or three cases and three out of five MTBs (60%) for one case. not to combine treatments when multiple alterations could The authors attributed this heterogeneity to differences in be targeted (cases 3A and 5A). Considering our proposed interpretation of tumor and germline aberrations and stan- definition of an MTB as primarily responsible for differentia- dards of prioritization. The higher overall rate of agreement in tion between targeted therapeutic options (Table 3), the our study may in part be explained by a lower number of last difference is most significant to address. A revision for alterations than was observed in the study of Rieke et al., who two cases (2B and 5A) revealed that MTBs are adaptable included more unknown alterations and had an average of when presented with new evidence: four MTBs revised 8 alterations compared with 2.6 in our study [19]. In addition, their recommendations based on evidence provided by our comparison was performed with MTBs within the same other MTBs. In case 5A, specifically, the MTB not suggesting health care system and thus subject to the same rules and combination treatment changed its recommendation when regulations with respect to the availability of drugs. Molecular presented with the recommendations of the other MTBs. pathology laboratories associated with these MTBs were Thus, the rate of agreement between MTBs may be already collaborating in a national consortium to achieve uni- improved by ensuring MTBs maintain local databases of formity in the interpretation of genomic aberrations in cancer reviewed cases, preferentially with systematic follow-up of [20] and are unified in The Netherlands Society for Pathology cases, and sharing exceptional cases with other MTBs in a [35]. Dutch thoracic oncologists and medical oncologists col- secure database to allow exchange of knowledge [20, 40]. laborate within respective professional networks [36, 37]. We used cases submitted by the MTBs to ensure a repre- These networks have long histories of collaborating in devel- sentative case mix routinely discussed by Dutch MTBs. As the oping national guidelines, organizing joint educational and MTBs had all been incorporated into the routine diagnostic consensus meetings, and shared national training programs. A setting of their respective institutions, the selected cases rep- Dutch MTB is thus effectively a meeting of local experts rep- resented cancer types with established benefit of targeted resenting these existing cooperative efforts. We consider this molecular testing. In other words, these MTBs primarily facili- combination of national connection and interpretative kinship tate expansion of “established” precision oncology programs, through close collaboration among tertiary cancer referral cen- with a greater chance of alterations that may be targeted with ters a key factor in achieving optimal diagnostic and/or thera- off-label therapy, in compassionate use or in clinical trials. In peutic recommendations with a high agreement. contrast, MTBs focusing on discovering novel treatment In calculating agreement, we considered recommenda- options with pancancer WES or WGS primarily facilitate tions comparable based on the next identified actionable “explorative” precision oncology programs. In these MTBs, the target. This is because differentiating between treatment discussed cases harbor more alterations with unknown signifi- options based on molecular alterations is the core task of cance, decreasing the chance of benefit for patients. Ideally, an MTB distinguishing it from conventional MDTs. Our com- an MTB should encompass both types of programs. parison revealed that the MTBs were generally able to inde- Further improvement in the Dutch MTB model may be pendently gather the available evidence on a molecular especially gained by achieving homogeneity in gathering aberration (Table 2). Yet, access to the same evidence does the increasing variant-level evidence to explore treatment not necessarily lead to identical recommendations: for options based on the molecular profile. This includes grad- example, three of five MTBs cited two preclinical studies on ing the actionability of variants, for which various classifica- the relatively rare KRAS p.(L19F) mutation (case 1B) [38, tion systems have been proposed [41–43], and access to 39], but one out of these three did not regard the evidence knowledge bases such as cBioPortal [26], JAX-CKB [31], and sufficient to recommend against anti-EGFR antibodies. This OncoKB [32]. Grading of actionability was not standard pro- translates to an inevitable limitation in agreement rates cedure for participating MTBs and thus not analyzed. This is between MTBs, as the role of the MTB is to interpret the a limitation of the current Dutch MTB model, and efforts available evidence and weigh this against patient factors, are ongoing to harmonize the classification of pathogenicity guidelines, rules and regulations, and the availability of and actionability in The Netherlands. Steps requiring har- therapeutic choices. Continuous access to and revision of monization include how to distinguish germline variants the new scientific evidence is imperative to an adequate from somatic variants, how to interpret variants of

© 2020 The Authors. The Oncologist published by Wiley Periodicals LLC on behalf of AlphaMed Press. Koopman, Groen, Ligtenberg et al. 11 unknown significance, which variant-level data- and knowl- Langen, Mirjam C. Boelens, Marthe S. Paats, Jan H. von der Thüsen, edge bases need to be consulted to identify potential treat- Winand N.M. Dinjens, Nienke Solleveld, Tom van Wezel, Hans Gelderblom, Lizza E. Hendriks, Ernst Jan M. Speel, Tom E. Theunissen, ment options, which classification system(s) to apply, and Leonie I. Kroeze, Niven Mehra, Berber Piet, Anthonie J. van der Wekken, exact criteria for allocating variants to the different classes. Arja ter Elst, Wim Timens, Stefan M. Willems, Ruud W.J. Meijers, Wendy W.J. de Leng, Anne S.R. van Lindert, Teodora Radonic, Sayed M.S. Hashemi, Daniëlle A.M. Heideman, Ed Schuuring, Léon C. van Kempen CONCLUSION Data analysis and interpretation: Bart Koopman, Harry J.M. Groen, Ed This study shows that the eight Dutch MTBs reach targeted Schuuring, Léon C. van Kempen Manuscript writing: Bart Koopman, Harry J.M. Groen, Marjolijn therapy recommendations in high agreement even with differ- J.L. Ligtenberg, Katrien Grünberg, Kim Monkhorst, Adrianus J. de Langen, ences in scope, logistics, and methods. A high agreement (86%) Mirjam C. Boelens, Marthe S. Paats, Jan H. von der Thüsen, Winand was achieved especially in identifying the principal actionable N.M. Dinjens, Nienke Solleveld, Tom van Wezel, Hans Gelderblom, Lizza E. Hendriks, Ernst Jan M. Speel, Tom E. Theunissen, Leonie I. Kroeze, targetsforcaseswithrareorcomplex molecular testing results. Niven Mehra, Berber Piet, Anthonie J. van der Wekken, Arja ter Elst, Wim Regional connection and data sharing among MTBs and inter- Timens, Stefan M. Willems, Ruud W.J. Meijers, Wendy W.J. de Leng, Anne pretative kinship through collaboration among pathology S.R. van Lindert, Teodora Radonic, Sayed M.S. Hashemi, Daniëlle A.M. Heideman, Ed Schuuring, Léon C. van Kempen departments were identified as key factors to achieve a high Final approval of manuscript: Bart Koopman, Harry J.M. Groen, Marjolijn rate of agreement between MTB recommendations. An MTB, J.L. Ligtenberg, Katrien Grünberg, Kim Monkhorst, Adrianus J. de Langen, Mirjam C. Boelens, Marthe S. Paats, Jan H. von der Thüsen, Winand although hosted in a tertiary cancer referral center, should be N.M. Dinjens, Nienke Solleveld, Tom van Wezel, Hans Gelderblom, Lizza accessible for all patients with cancer, which requires active par- E. Hendriks, Ernst Jan M. Speel, Tom E. Theunissen, Leonie I. Kroeze, ticipation of health care professionals from peripheral hospitals Niven Mehra, Berber Piet, Anthonie J. van der Wekken, Arja ter Elst, Wim Timens, Stefan M. Willems, Ruud W.J. Meijers, Wendy W.J. de Leng, Anne and pathology laboratories in a regional MTB network. We rec- S.R. van Lindert, Teodora Radonic, Sayed M.S. Hashemi, Daniëlle ommend using our proposed Dutch MTB model for an optimal, A.M. Heideman, Ed Schuuring, Léon C. van Kempen collaborative, and nationally aligned MTB workflow to trans- form precision medicine from retrospective anecdotal evidence to successful prospective evidence. DISCLOSURES Kim Monkhorst: Roche, AstraZeneca, PGDx (RF), Merck Sharp & ACKNOWLEDGMENTS Dohme, Roche, AstraZeneca (H), Pfizer, Bristol-Myers Squibb, Roche, Merck Sharp & Dohme, Abbvie, AstraZeneca, Diaceutics We are grateful to Anke van den Berg, Matthew Groves, (C/A), Roche, Takeda, Pfizer (other); Lizza E. Hendriks: Roche, Geke Hospers, Birgitta Hiddinga, Hilde Jalving, Lucie Boehringer Ingelheim, Bristol-Myers Squibb, Eli Lilly & Co., Pfizer, Hijmering-Kappelle, Joost Kluiver, Michel van Kruchten, Elise Takeda, Merck Sharp & Dohme (SAB), Roche, AstraZeneca (RF), Roche (other—travel, fees for interview sessions), AstraZeneca van der Logt, Maarten Niemantsverdriet, Sjoukje Oosting, (other—mentorship program), Merck Sharp & Dohme (other— Juliana Vilacha, Michel van den Heuvel, Lieneke Steeghs, speaker educational session, travel), Quadia (other—fees for Ingrid Vogelaar, Linda Bosch, Liudmila Kodach, Egbert Smit, educational webinars); Ernst Jan M. Speel: AstraZeneca, Merck Sharp & Dohme, Bristol-Myers Squibb, Novartis (RF), Bristol-Myers Annette Bijsmans, Maxime van Berge Henegouwen, Hans Squibb, Merck Sharp & Dohme, AbbVie, Pfizer, Roche, Bayer (C/A), Morreau, Lisa Hillen, Xiao Fei Li, John Hinrichs, Anne Jansen, AbbVie (other); Niven Mehra: Roche, Merck Sharp & Dohme, Joyce Radersma-van Loon, and Aryan Vink and all other Bristol-Myers Squibb, Astellas, Janssen (C/A), Astellas, Janssen, Pfizer, Roche, Sanofi, Genzyme (RF), Astellas, Merck Sharp & members of the molecular tumor boards in The Netherlands Dohme (other—travel support); Wim Timens: Roche Diagnostics- for their contributions to the conduct and successful comple- Ventana, AbbVie, Bristol-Myers Squibb (C/A), Merck Sharp & tion of this study. This work was supported by ZonMW (The Dohme (SAB); Stefan M. Willems: AstraZeneca, Roche, Merck Sharp & Dohme, Bristol-Myers Squibb, Bayer, Amgen, Nextcure Netherlands Organization for Health Research) within the (RF); Wendy W.J. de Leng: Roche, Bristol-Myers Squibb, Pfizer Personalized Medicine Program (grant number 846001001). (RF); Sayed M.S. Hashemi: AbbVie, AstraZeneca, Bristol-Myers Squibb, Eli Lilly & Co., GlaxoSmithKline, Loxo, Merck Sharp & Dohme, Novartis, Roche, Takeda, Xcovery (RF, SAB); Daniëlle A.M. Heideman: Pfizer, Bristol-Myers Squibb (SAB), Self-Screen AUTHOR CONTRIBUTIONS B.V. (OI), Qiagen (other—speakers bureau); Ed Schuuring: Conception/design: Bart Koopman, Harry J.M. Groen, Marjolijn J.L. Ligtenberg, AstraZeneca, Roche, Pfizer, Novartis, Merck Sharp & Dohme/ Katrien Grünberg, Ed Schuuring, Léon C. van Kempen Merck, Bayer, Bristol-Myers Squibb, BioCartis, Illumina, Agena Provision of study material or patients: Bart Koopman, Harry J.M. Groen, Bioscience, Janssen Cilag (Johnson&Johnson), Diaceutics (C/A), Marjolijn J.L. Ligtenberg, Katrien Grünberg, Kim Monkhorst, Adrianus J. de Abbott, Pfizer, Biocartis, Bristol-Myers Squibb, Bio-Rad, Roche, Langen, Mirjam C. Boelens, Marthe S. Paats, Jan H. von der Thüsen, Winand Agena Bioscience, CC Diagnostics, Boehringer Ingelheim, Qiagen, N.M. Dinjens, Nienke Solleveld, Tom van Wezel, Hans Gelderblom, Lizza Promega, TATAA (RF), Bio-Rad, Novartis, Roche, Biocartis, Agena E. Hendriks, Ernst Jan M. Speel, Tom E. Theunissen, Leonie I. Kroeze, Niven Bioscience, Illumina, Pfizer, AstraZeneca (H). Léon van Kempen: Mehra, Berber Piet, Anthonie J. van der Wekken, Arja ter Elst, Wim Timens, NanoString (C/A), Roche, Bayer, NanoString (RF), Merck Sharpe & Stefan M. Willems, Ruud W.J. Meijers, Wendy W.J. de Leng, Anne S.R. van Dohme, AstraZeneca, Bristol-Myers Squibb, Pfizer (H). The other Lindert, Teodora Radonic, Sayed M.S. Hashemi, Daniëlle A.M. Heideman, Ed authors indicated no financial relationships. Schuuring, Léon C. van Kempen (C/A) Consulting/advisory relationship; (RF) Research funding; (E) Employment; (ET) Expert Collection and/or assembly of data: Bart Koopman, Harry J.M. Groen, Mar- testimony; (H) Honoraria received; (OI) Ownership interests; (IP) Intellectual property rights/ jolijn J.L. Ligtenberg, Katrien Grünberg, Kim Monkhorst, Adrianus J. de inventor/patent holder; (SAB) Scientific advisory board

REFERENCES

1. Garraway LA. Genomics-driven oncology: Current practice and future needs. Ann Oncol Netherlands: Experiences, challenges, and aspira- Framework for an emerging paradigm. J Clin 2017;28:3070–3075. tions. Br J Cancer 2019;121:34–36. – Oncol 2013;31:1806 1814. 3. Willemsen AECAB, Krausz S, Ligtenberg MJL 4. Koopman B, van der Wekken AJ, ter Elst A et al. 2. van der Velden DL, van Herpen CML, van et al. Molecular tumour boards and molecular Relevance and effectiveness of molecular tumor Laarhoven HWM et al. Molecular tumor boards: diagnostics for patients with cancer in The board recommendations for patients with non– www.TheOncologist.com © 2020 The Authors. The Oncologist published by Wiley Periodicals LLC on behalf of AlphaMed Press. 12 Molecular Tumor Boards in The Netherlands small-cell lung cancer with rare or complex muta- 18. Trédan O, Wang Q, Pissaloux D et al. Molec- 30. Cunningham F, Achuthan P, Akanni W et al. tional profiles. JCO Precis Oncol 2020;4:393–410. ular screening program to select molecular- Ensembl 2019. Nucleic Acids Res 2019;47: – 5. Basse C, Morel C, Alt M et al. Relevance of a based recommended therapies for metastatic D745 D751. fi molecular tumour board (MTB) for patients’ cancer patients: Analysis from the Pro LER trial. Ann Oncol 2019;30:757–765. 31. Patterson SE, Liu R, Statz CM et al. The clini- enrolment in clinical trials: Experience of the cal trial landscape in oncology and connectivity Institut Curie. ESMO Open 2018;3:e000339. 19. Rieke DT, Lamping M, Schuh M et al. Com- of somatic mutational profiles to targeted thera- 6. Bryce AH, Egan JB, Borad MJ et al. Experience parison of treatment recommendations by pies. Hum Genomics 2016;10:4. with precision genomics and tumor board, indi- molecular tumor boards worldwide. JCO Precis cates frequent target identification, but barriers Oncol 2018;2:PO.18.00098. 32. Chakravarty D, Gao J, Phillips SM et al. to delivery. Oncotarget 2017;8:27145–27154. OncoKB: A precision oncology knowledge base. 20. Nederlandse organisatie voor gezondheid- JCO Precis Oncol 2017;2017:513–519. 7. Dalton WB, Forde PM, Kang H et al. Personal- sonderzoek en zorginnovatie (ZonMW). Predictive ized medicine in the oncology clinic: Implementa- Analysis for Therapy: PATH to Optimising Access to 33. PubMed. U.S. National Library of Medicine tion and outcomes of the Johns Hopkins molecular Personalised Cancer Therapy in The Netherlands. Web site. Available at https://pubmed.ncbi.nlm.nih. tumor board. JCO Precis Oncol 2017;1: 2016. Available at https://www.zonmw.nl/nl/ gov/. PO.16.00046. onderzoek-resultaten/geneesmiddelen/program 34. Deans ZC, Costa JL, Cree I et al. Integration of 8. Harada S, Arend R, Dai Q et al. Implementa- mas/project-detail/personalised-medicine/predicti ve-analysis-for-therapy-path-to-optimising-access-to next-generation sequencing in clinical diagnostic tion and utilization of the molecular tumor board molecular pathology laboratories for analysis of to guide precision medicine. Oncotarget 2017;8: -personalised-cancer-therapy-in-the-net/. Accessed August 8, 2019. solid tumours; an expert opinion on behalf of IQN 57845–57854. Path ASBL. Virchows Arch 2017;470:5–20. 9. Kaderbhai CG, Boidot R, Beltjens F et al. Use of 21. Universitair Medisch Centrum Groningen. 35. Nederlandse Vereniging voor Pathologie dedicated gene panel sequencing using next gener- Research Code. 2007. Available at https://www.umcg. (NVVP). Available at https://pathology.nl/. ation sequencing to improve the personalized care nl/SiteCollectionDocuments/English/Researchcode/ Accessed April 28, 2020. of lung cancer. Oncotarget 2016;7:24860–24870. umcg-research-code-2018-en.pdf. Accessed May 5, 2020. 10. Knepper TC, Bell GC, Hicks JK et al. Key les- 36. Nederlandse Vereniging van Artsen voor sons learned from Moffitt’s molecular tumor 22. Dutch Federation of Biomedical ScientificSoci- Longziekten en Tuberculose (NVALT). Available at board: The Clinical Genomics Action Committee eties. Code of Conduct for Medical Research. 2004. https://www.nvalt.nl/. Accessed April 28, 2020. experience. The Oncologist 2017;22:144–151. Available at https://www.federa.org/sites/default/ 37. Nederlandse Vereniging voor Medische files/bijlagen/coreon/code_of_conduct_for_medical 11. Lee B, Tran B, Hsu AL et al. Exploring the Oncologie (NVMO). Available at https://www. _research_1.pdf. Accessed May 5, 2020. feasibility and utility of exome-scale tumour nvmo.org/. Accessed April 28, 2020. sequencing in a clinical setting. Intern Med J 23. Nederlandse Vereniging voor Artsen voor 2018;48:786–794. Longziekten en Tuberculose. Behandeling pat- 38. Akagi K, Uchibori R, Yamaguchi K et al. Char- acterization of a novel oncogenic K-ras mutation 12. Marks LJ, Oberg JA, Pendrick D et al. Preci- iënten met een zeldzame mutatie bij NSCLC. in colon cancer. Biochem Biophys Res Commun sion medicine in children and young adults with 2020. Available at https://richtlijnendatabase.nl/ 2007;352:728–32. hematologic malignancies and blood disorders: richtlijn/niet_kleincellig_longcarcinoom/systemisc The Columbia University experience. Front he_behandeling_stadium_iv_nsclc/behandeling_ 39. Smith G, Bounds R, Wolf H et al. Activating K- Pediatr 2017;5:265. pati_nten_met_een_zeldzame_mutatie_bij_nsclc. Ras mutations outwith “hotspot” codons in spo- html. Accessed April 23, 2020. 13. Moore DA, Kushnir M, Mak G et al. Prospec- radic colorectal tumours - implications for tive analysis of 895 patients on a UK genomics 24. van der Velden DL, Hoes LR, van der personalised cancer medicine. Br J Cancer 2010; review board. ESMO Open 2019;4:e000469. Wijngaart H et al. The Drug Rediscovery protocol 102:693–703. facilitates the expanded use of existing antican- 14. Ortiz M V, Kobos R, Walsh M et al. Integrat- 40. Health RI: Enabling data driven health. cer drugs. Nature 2019;574:127–131. ing genomics into clinical pediatric oncology Available at https://www.health-ri.nl/services/ using the molecular tumor board at the Memo- 25. 1000 Genomes Project Consortium; Auton A, cbioportal. Accessed March 3, 2020. rial Sloan Kettering Cancer Center. Pediatr Blood Brooks LD, Durbin RM et al. A global reference for Cancer 2016;63:1368–1374. human genetic variation. Nature 2015;526:68–74. 41. Mateo J, Chakravarty D, Dienstmann R et al. A framework to rank genomic alterations as tar- 15. Rolfo C, Manca P, Salgado R et al. Multi- 26. Gao J, Aksoy BA, Dogrusoz U et al. Integrative gets for cancer precision medicine: The ESMO disciplinary molecular tumour board: A tool to analysis of complex cancer genomics and clinical Scale for Clinical Actionability of molecular Tar- improve clinical practice and selection accrual profiles using the cBioPortal. Sci Signal 2013;6:pl1. gets (ESCAT). Ann Oncol 2018;29:1895–1902. for clinical trials in patients with cancer. ESMO Open 2018;3:e000398. 27. Landrum MJ, Lee JM, Riley GR et al. ClinVar: 42. Li MM, Datto M, Duncavage EJ et al. Stan- Public archive of relationships among sequence 16. Schwaederle M, Parker BA, Schwab RB et al. dards and guidelines for the interpretation and variation and human phenotype. Nucleic Acids reporting of sequence variants in cancer: A joint Molecular tumor board: The University of Res 2014;42:D980–D985. California-San Diego Moores Cancer Center expe- consensus recommendation of the Association rience. The Oncologist 2014;19:631–636. 28. Forbes SA, Beare D, Boutselakis H et al. for Molecular Pathology, American Society of COSMIC: Somatic cancer genetics at high-resolu- Clinical Oncology, and College of American 17. Tafe LJ, Gorlov IP, de Abreu FB et al. Imple- tion. Nucleic Acids Res 2017;45:D777–D783. Pathologists. J Mol Diagn 2017;19:4–23. mentation of a molecular tumor board: The impact on treatment decisions for 35 patients 29. Sherry ST, Ward MH, Kholodov M et al. 43. Leichsenring J, Horak P, Kreutzfeldt S et al. evaluated at Dartmouth-Hitchcock Medical Cen- dbSNP: The NCBI database of genetic variation. Variant classification in precision oncology. Int J ter. The Oncologist 2015;20:1011–1018. Nucleic Acids Res 2001;29:308–311. Cancer 2019;145:2996–3010.

See http://www.TheOncologist.com for supplemental material available online.

© 2020 The Authors. The Oncologist published by Wiley Periodicals LLC on behalf of AlphaMed Press. The new england journal of medicine

established in 1812 February 13, 2020 vol. 382 no. 7 Tucatinib, Trastuzumab, and Capecitabine for HER2-Positive Metastatic Breast Cancer R.K. Murthy, S. Loi, A. Okines, E. Paplomata, E. Hamilton, S.A. Hurvitz, N.U. Lin, V. Borges, V. Abramson, C. Anders, P.L. Bedard, M. Oliveira, E. Jakobsen, T. Bachelot, S.S. Shachar, V. Müller, S. Braga, F.P. Duhoux, R. Greil, D. Cameron, L.A. Carey, G. Curigliano, K. Gelmon, G. Hortobagyi, I. Krop, S. Loibl, M. Pegram, D. Slamon, M.C. Palanca‑Wessels, L. Walker, W. Feng, and E.P. Winer​​

abstract

BACKGROUND Patients with human epidermal growth factor receptor 2 (HER2)–positive metastatic The authors’ full names, academic de‑ breast cancer who have disease progression after therapy with multiple HER2-targeted grees, and affiliations are listed in the Appendix. Address reprint requests to Dr. agents have limited treatment options. Tucatinib is an investigational, oral, highly Winer at the Dana–Farber Cancer Institute, selective inhibitor of the HER2 tyrosine kinase. 450 Brookline Ave., Boston, MA 02215, or at ­eric_winer@​­dfci​.­harvard​.­edu; or to Dr. METHODS Murthy at the University of Texas M.D. We randomly assigned patients with HER2-positive metastatic breast cancer previ- Anderson Cancer Center, 1515 Holcombe ously treated with trastuzumab, pertuzumab, and trastuzumab emtansine, who had Blvd., Unit 1354, Houston, TX 77030, or at ­rmurthy1@​­mdanderson​.­org. or did not have brain metastases, to receive either tucatinib or placebo, in combina- tion with trastuzumab and capecitabine. The primary end point was progression-free This article was published on December 11, 2019, and updated on December 27, 2019, survival among the first 480 patients who underwent randomization. Secondary end at NEJM.org. points, assessed in the total population (612 patients), included overall survival, pro- N Engl J Med 2020;382:597-609. gression-free survival among patients with brain metastases, confirmed objective re- DOI: 10.1056/NEJMoa1914609 sponse rate, and safety. Copyright © 2019 Massachusetts Medical Society. RESULTS Progression-free survival at 1 year was 33.1% in the tucatinib-combination group and 12.3% in the placebo-combination group (hazard ratio for disease progression or death, 0.54; 95% confidence interval [CI], 0.42 to 0.71; P<0.001), and the median duration of progression-free survival was 7.8 months and 5.6 months, respectively. Overall survival at 2 years was 44.9% in the tucatinib-combination group and 26.6% in the placebo- combination group (hazard ratio for death, 0.66; 95% CI, 0.50 to 0.88; P = 0.005), and the median overall survival was 21.9 months and 17.4 months, respectively. Among the patients with brain metastases, progression-free survival at 1 year was 24.9% in the tucatinib-combination group and 0% in the placebo-combination group (hazard ratio, 0.48; 95% CI, 0.34 to 0.69; P<0.001), and the median progression-free survival was 7.6 months and 5.4 months, respectively. Common adverse events in the tucatinib group included diarrhea, palmar–plantar erythrodysesthesia syndrome, nausea, fatigue, and vomiting. Diarrhea and elevated aminotransferase levels of grade 3 or higher were more common in the tucatinib-combination group than in the placebo-combination group. CONCLUSIONS In heavily pretreated patients with HER2-positive metastatic breast cancer, including those with brain metastases, adding tucatinib to trastuzumab and capecitabine resulted in better progression-free survival and overall survival outcomes than adding placebo; the risks of diarrhea and elevated aminotransferase levels were higher with tucatinib. (Funded by Seattle Genetics; HER2CLIMB ClinicalTrials.gov number, NCT02614794.)

n engl j med 382;7 nejm.org February 13, 2020 597 The New England Journal of Medicine Downloaded from nejm.org at Walaeus Library - Leiden University Medical Center on April 7, 2021. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved. The new england journal of medicine

pproximately 15 to 20% of breast with trastuzumab, pertuzumab, and trastuzumab cancers overexpress human epidermal emtansine. Agrowth factor receptor 2 (HER2).1,2 De- spite dramatic therapeutic advances over the Methods past 20 years, most patients with HER2-positive A Quick Take is available at metastatic breast cancer ultimately die from Trial Design and Oversight NEJM.org their disease.3,4 Moreover, as systemic therapies We conducted an international, randomized, have improved, the incidence of brain metastases, double-blind trial in which the combination of for which effective treatment options are limit- tucatinib plus trastuzumab and capecitabine was ed, has increased such that brain metastases may compared with placebo plus trastuzumab and develop in up to half of patients.5-9 Standard-of- capecitabine. The trial was conducted in accor- care treatment for patients with HER2-positive dance with regulatory requirements and Interna- metastatic breast cancer is first-line trastuzumab tional Conference on Harmonisation Good Clin- plus pertuzumab and a taxane, followed by ical Practice guidelines. The protocol, available second-line trastuzumab emtansine for patients with the full text of this article at NEJM.org, was who have disease progression.10-12 After progres- approved by institutional review boards and eth- sion during treatment with trastuzumab emtan- ics committees, according to the practice at each sine, no single regimen is considered the stan- participating trial site. All patients provided dard of care; commonly used regimens include written informed consent. A steering committee tyrosine kinase inhibitors such as lapatinib with provided scientific advice, and an independent trastuzumab or capecitabine, trastuzumab with data and safety monitoring committee oversaw chemotherapy, or participation in a clinical trial. the conduct of the trial. The steering committee Treatment for brain metastases in patients and representatives of the sponsor (Seattle Ge- with HER2-positive breast cancer includes local netics) designed the trial. The authors wrote the therapies, such as neurosurgical resection and manuscript with the assistance of a medical stereotactic (or whole-brain) radiation therapy.13 writer funded by the sponsor. All the authors Although limited antitumor activity in the brain had full access to the relevant data, vouch for the has been reported for some systemic HER2-tar- completeness and accuracy of the data and for geted agents, including tyrosine kinase inhibi- adherence of the trial to the protocol, and had tors and chemotherapy,14-18 data from random- final responsibility for the content of the man- ized trials showing better treatment outcomes in uscript and for the decision to submit the manu- these patients are lacking. script for publication. Tucatinib is an investigational, oral tyrosine kinase inhibitor that is highly selective for the Patients kinase domain of HER2 with minimal inhibition Patients 18 years of age or older were eligible to of epidermal growth factor receptor, which may participate if they had advanced breast carcinoma alter the toxicity profile.19,20 In a phase 1b dose- that had been determined to be HER2-positive escalation trial, tucatinib in combination with on the basis of immunohistochemical analysis, trastuzumab and capecitabine showed encourag- in situ hybridization, or fluorescence in situ hy- ing antitumor activity in patients with HER2- bridization and confirmed at a central location; positive metastatic breast cancer, including those had previously been treated with trastuzumab, with brain metastases. This regimen was also pertuzumab, and trastuzumab emtansine; and associated with diarrhea, nausea, palmar–plantar had an Eastern Cooperative Oncology Group erythrodysesthesia syndrome, fatigue, and vom- (ECOG) performance-status score of 0 or 1 (on a iting; however, each grade 3 or higher treatment- 5-point scale in which higher numbers reflect related toxic effect (fatigue, diarrhea, and palmar- greater disability). Patients were excluded if they plantar erythrodysesthesia syndrome) occurred in had previously received treatment for metastatic no more than 10% of patients.21 The HER2CLIMB disease with capecitabine or a HER2-targeted trial evaluated tucatinib combined with trastuzu­ tyrosine kinase inhibitor (although patients who mab and capecitabine in patients with HER2- had received lapatinib more than 12 months positive metastatic breast cancer previously treated before initiating a trial regimen were eligible for

598 n engl j med 382;7 nejm.org February 13, 2020 The New England Journal of Medicine Downloaded from nejm.org at Walaeus Library - Leiden University Medical Center on April 7, 2021. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved. Tucatinib for HER2-Positive Metastatic Breast Cancer

inclusion). Patients with brain metastases were version 4.03). Changes in vital signs and labora- included unless they were in need of immediate tory results were assessed in patients who received local intervention, in which case they could re- at least one dose of any trial drug or placebo. ceive local therapy and be enrolled subsequently. Patients with untreated brain metastases larger End Points than 2 cm in diameter could be enrolled with The primary end point was progression-free sur- approval from the medical monitor. Patients with vival (defined as the time from randomization to leptomeningeal disease were excluded. A com- documented disease progression, as assessed plete list of eligibility criteria is provided in the by means of blinded independent central review, or protocol. death from any cause, whichever occurred first) in the first 480 patients who underwent ran- Randomization and Treatment domization (primary end-point analysis popula- Patients were randomly assigned in a 2:1 ratio to tion). Multiplicity-adjusted secondary end points receive either tucatinib (300 mg orally twice were assessed in the total population and in- daily throughout the treatment period) or pla- cluded overall survival (defined as the time from cebo (orally twice daily), in combination with randomization to death from any cause); pro- trastuzumab (6 mg per kilogram of body weight gression-free survival, as assessed by means of intravenously once every 21 days, with an initial blinded independent central review, among the loading dose of 8 mg per kilogram; subcutane- patients who had brain metastases at baseline; ous administration was allowed) and capecitabine and confirmed objective response rate (defined (1000 mg per square meter of body-surface area as the percentage of patients with measurable orally twice daily on days 1 to 14 of each 21-day disease at baseline who had a confirmed com- cycle) (Fig. S1 in the Supplementary Appendix, plete response or partial response, as assessed by available at NEJM.org). Patients were stratified means of blinded independent central review). according to whether brain metastases were Safety was a secondary end point. present (yes or no), ECOG performance-status score (0 or 1), and geographic region (United Statistical Analysis States, Canada, or the rest of the world). The primary end-point analysis was to be per- formed after approximately 288 events had oc- Assessments curred in the primary end-point analysis popula- Contrast-enhanced spiral computed tomography tion, which would provide 90% power to detect (CT), positron-emission tomography–CT, or con- a hazard ratio of 0.67 with a two-sided log-rank trast-enhanced magnetic resonance imaging test at an alpha level of 0.05. If the result with (MRI) (or combinations thereof) was performed respect to the primary end point was significant, at baseline, every 6 weeks for 24 weeks, and every overall survival and progression-free survival 9 weeks thereafter. MRI of the head at baseline among the patients with brain metastases were was required for all patients; those with brain to be tested in parallel at alpha levels of 0.02 and metastases on the baseline scan underwent a 0.03, respectively, in the first interim analysis. contrast MRI of the head at the same intervals. The final analyses for these two variables were Disease response and progression were evaluated to be performed with 361 and 220 events, re- in accordance with Response Evaluation Criteria spectively. Approximately 600 patients were to in Solid Tumors (RECIST) criteria, version 1.1,22 undergo randomization in order for the required by means of blinded independent central review. number of events to be accumulated (see the Laboratory assessments were performed at least Statistical Methods section in the Supplementa- every 3 weeks throughout the treatment period ry Appendix). Overall type I error in the interim and 30 days after the last trial treatment date. and final analyses of overall survival in the total Safety was assessed on the basis of the inci- population and progression-free survival among dence of adverse events (defined according to the patients with brain metastases was con- the Medical Dictionary for Regulatory Activities, ver- trolled with the use of the group sequential sion 22.0, and the National Cancer Institute Com- Holm variable procedure23 with the Lan–DeMets mon Terminology Criteria for Adverse Events, alpha-spending function with an O’Brien–Fleming

n engl j med 382;7 nejm.org February 13, 2020 599 The New England Journal of Medicine Downloaded from nejm.org at Walaeus Library - Leiden University Medical Center on April 7, 2021. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved. The new england journal of medicine

boundary. If the results with respect to both overall survival and 0.008 for progression-free overall survival in the total population and pro- survival among the patients with brain metastases. gression-free survival among the patients with brain metastases were significant, the between- Results group difference in the percentage of patients who had a confirmed objective response was to Patient Characteristics be tested at a two-sided alpha level of 0.05. Between February 23, 2016, and May 3, 2019, a The size of the initial trial population (180 total of 612 patients were enrolled at 155 sites in patients) was increased to 480 to provide ade- 15 countries (total population); 410 were ran- quate statistical power to evaluate the end point domly assigned to the tucatinib-combination of progression-free survival and increase the group and 202 to the placebo-combination group. power of the subgroup analyses. The size of the In the primary end-point analysis population trial population was later increased to approxi- (480 patients), 320 patients were randomly as- mately 600 patients to ensure sufficient power to signed to the tucatinib-combination group and show a progression-free survival benefit among 160 to the placebo-combination group (Fig. S1 in the patients with brain metastases. At no point the Supplementary Appendix). The median dura- were these decisions based on an assessment tion of follow-up in the total population was of the ongoing results, and blinding was never 14.0 months. Patient demographic and disease broken. characteristics at baseline were well balanced The Kaplan–Meier method was used to esti- between the treatment groups (Table 1). In the mate progression-free survival and overall survival total population, 291 patients (47.5%) had brain time curves, median progression-free survival and metastases at baseline — 48.3% in the tucatinib- overall survival, and 95% confidence intervals for combination group and 46.0% in the placebo- the treatment groups. Cox proportional-hazards combination group (Table S1). models, with stratification factors taken into ac- count, were used to estimate hazard ratios and Efficacy in the Primary End-Point Population 95% confidence intervals. The treatment groups At 1 year, the estimated progression-free sur- were compared with the use of a stratified log- vival was 33.1% (95% confidence interval [CI], rank test, and P values were calculated by means 26.6 to 39.7) in the tucatinib-combination group of a rerandomization procedure to reflect the and 12.3% (95% CI, 6.0 to 20.9) in the placebo- dynamic, hierarchical randomization scheme.24 combination group, and the median duration In the analysis of progression-free survival, data of progression-free survival was 7.8 months from patients without any documented event or (95% CI, 7.5 to 9.6) and 5.6 months (95% CI, 4.2 who received anticancer therapy not specified in to 7.1), respectively. The risk of disease progres- the protocol were censored at the date the pa- sion or death, as assessed by means of blinded tient was last known to be event-free. The cen- independent central review in the primary end- soring scheme for the primary end point is de- point analysis population, was 46% lower in the scribed in the statistical analysis plan (available tucatinib-combination group than in the place- with the protocol at NEJM.org). The P value for bo-combination group (hazard ratio, 0.54; 95% the between-group comparison of the percent- CI, 0.42 to 0.71; P<0.001) (Fig. 1A). Hazard ra- age of patients who had a confirmed objective tios across all subgroups were consistent with response was calculated with the use of a strati- the hazard ratio in the overall analysis (Fig. 1B). fied Cochran–Mantel–Haenszel test. A comparison of investigator-assessed progres- At the date of data cutoff (September 4, 2019), sion-free survival between the treatment groups a total of 275 events of disease progression or yielded results consistent with those in the death had occurred in the primary end-point primary end-point analysis (Fig. S2). analysis population, 215 deaths had occurred in the total population, and 157 events of disease Efficacy in the Total Trial Population progression or death had occurred among the At 2 years, the estimated overall survival was patients with brain metastases in the total popu- 44.9% (95% CI, 36.6 to 52.8) in the tucatinib- lation. On the basis of the observed number of combination group and 26.6% (95% CI, 15.7 to events, the multiplicity-adjusted, two-sided alpha 38.7) in the placebo-combination group, and the levels at the first interim analysis were 0.007 for median duration of overall survival was 21.9

600 n engl j med 382;7 nejm.org February 13, 2020 The New England Journal of Medicine Downloaded from nejm.org at Walaeus Library - Leiden University Medical Center on April 7, 2021. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved. Tucatinib for HER2-Positive Metastatic Breast Cancer

Table 1. Patient Demographic and Disease Characteristics in the Primary End-Point Analysis Population and the Total Population.*

Primary End-Point Analysis Population Total Population Characteristic (N = 480) (N = 612)

Tucatinib Placebo Tucatinib Placebo Combination Combination Combination Combination (N = 320) (N = 160) (N = 410) (N = 202) Female sex — no. (%) 317 (99.1) 158 (98.8) 407 (99.3) 200 (99.0) Age — no. (%) <65 yr 252 (78.8) 132 (82.5) 328 (80.0) 168 (83.2) ≥65 yr 68 (21.2) 28 (17.5) 82 (20.0) 34 (16.8) Median age — yr 54.0 54.0 55.0 54.0 Race — no. (%)† Asian 17 (5.3) 3 (1.9) 18 (4.4) 5 (2.5) Black 30 (9.4) 13 (8.1) 41 (10.0) 14 (6.9) White 225 (70.3) 125 (78.1) 287 (70.0) 157 (77.7) Unknown or other 48 (15.0) 19 (11.9) 64 (15.6) 26 (12.9) Geographic region — no. (%) United States and Canada 204 (63.8) 103 (64.4) 246 (60.0) 123 (60.9) Rest of the world 116 (36.2) 57 (35.6) 164 (40.0) 79 (39.1) Hormone-receptor status — no. (%) Positive for ER or PR or both 190 (59.4) 99 (61.9) 243 (59.3) 127 (62.9) Negative for ER and PR 126 (39.4) 61 (38.1) 161 (39.3) 75 (37.1) Other 4 (1.2) 0 6 (1.5) 0 ECOG performance-status score — no. (%)‡ 0 159 (49.7) 76 (47.5) 204 (49.8) 94 (46.5) 1 161 (50.3) 84 (52.5) 206 (50.2) 108 (53.5) Stage IV at initial diagnosis — no. (%) 108 (33.8) 67 (41.9) 143 (34.9) 77 (38.1) Presence or history of brain metastases — no. (%) 148 (46.2) 71 (44.4) 198 (48.3) 93 (46.0) Location of other metastases — no. (%) Lung 160 (50.0) 82 (51.2) 200 (48.8) 100 (49.5) Liver 108 (33.8) 64 (40.0) 137 (33.4) 78 (38.6) Bone 178 (55.6) 85 (53.1) 223 (54.4) 111 (55.0) Previous lines of therapy, median no. (range) 4 (2–14) 4 (2–17) 4 (2–14) 4 (2–17) Previous lines of therapy for metastatic cancer, median 3 (1–14) 3 (1–13) 3 (1–14) 3 (1–13) no. (range) Previous therapies — no. (%) Trastuzumab 320 (100) 160 (100) 410 (100) 202 (100) Pertuzumab 320 (100) 159 (99.4) 409 (99.8) 201 (99.5) Trastuzumab emtansine 320 (100) 160 (100) 410 (100) 202 (100) Lapatinib 22 (6.9) 10 (6.2) 24 (5.9) 10 (5.0)

* The primary end-point analysis population included the first 480 patients who were randomly assigned to the tucatinib-combination group (tucatinib plus trastuzumab and capecitabine) or to the placebo-combination group (placebo plus trastuzumab and capecitabine), and the total population included 612 patients who underwent randomization. Randomization stratification factors included geographic region (United States, Canada, or the rest of the world), presence or history of brain metastases (yes or no), and Eastern Cooperative Oncology Group (ECOG) performance-status score (0 or 1). ER denotes estrogen receptor, and PR progesterone receptor. Data from the patients in the United States and Canada were combined for this analysis. † Race was determined by the local investigator. ‡ ECOG performance-status scores range from 0 to 5, with higher scores indicating greater disability.

n engl j med 382;7 nejm.org February 13, 2020 601 The New England Journal of Medicine Downloaded from nejm.org at Walaeus Library - Leiden University Medical Center on April 7, 2021. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved. The new england journal of medicine

A Kaplan–Meier Estimates of Progression-free Survival 100 Median 90 No. of Events/ Duration 80 No. of Patients (95% CI) mo 70 62.9 Tucatinib Combination 178/320 7.8 (7.5–9.6) 60 Placebo Combination 97/160 5.6 (4.2–7.1) 50 Hazard ratio for disease progression or death, 40 46.3 0.54 (95% CI, 0.42–0.71) 33.1 Tucatinib P<0.001 30 combination 20 Placebo Disease Progression (%) combination Patients Alive and Free from 10 12.3 0 0 3 6 9 12 15 18 21 24 27 30 33 36 Months since Randomization No. at Risk Tucatinib combination 320 235 152 98 40 29 15 10 8 4 2 1 0 Placebo combination 160 94 45 27 6 4 2 1 1 0 0 0 0

B Subgroup Analysis of Progression-free Survival No. of Events/ Subgroup Total No. Hazard Ratio for Disease Progression or Death (95% CI) Total 275/480 0.54 (0.42–0.71) Age ≥65 yr 51/96 0.59 (0.32–1.11) <65 yr 224/384 0.54 (0.41–0.72) Race White 206/350 0.57 (0.42–0.77) Nonwhite 69/130 0.46 (0.26–0.82) Hormone-receptor status Positive for ER, PR, or both 172/289 0.58 (0.42–0.80) Negative for ER and PR 103/191 0.54 (0.34–0.86) Baseline brain metastasis Yes 138/219 0.46 (0.31–0.67) No 136/260 0.62 (0.44–0.89) ECOG performance-status score 0 134/235 0.56 (0.39–0.80) 1 141/245 0.55 (0.38–0.79) Geographic region United States and Canada 179/307 0.57 (0.41–0.78) Rest of the world 96/173 0.51 (0.33–0.79) 0.1 1.0 10.0

Tucatinib Combination Placebo Combination Better Better

Figure 1. Kaplan–Meier Estimates of Progression-free Survival in the Primary End-Point Analysis Population and Prespecified Subgroups. Panel A shows the Kaplan–Meier estimates of progression‑free survival in the primary end‑point analysis population, which included the first 480 patients who underwent randomization. Progression‑free survival was assessed by means of blinded independent central review. The number of events is the number of events of disease progression or death. The dashed vertical lines indicate landmark time points. Panel B shows the analysis of progression‑free survival in prespecified subgroups. Scores for the Eastern Cooperative Oncology Group (ECOG) performance status range from 0 to 5, with higher scores indicating greater disability. Data from the patients in the United States and Canada were combined for this analysis. ER denotes estrogen receptor, and PR progesterone receptor.

602 n engl j med 382;7 nejm.org February 13, 2020 The New England Journal of Medicine Downloaded from nejm.org at Walaeus Library - Leiden University Medical Center on April 7, 2021. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved. Tucatinib for HER2-Positive Metastatic Breast Cancer

A Kaplan–Meier Estimates of Overall Survival 100 Median 90 No. of Deaths/ Duration No. of Patients (95% CI) 80 75.5 Tucatinib mo 70 combination Tucatinib Combination 130/410 21.9 (18.3–31.0) 60 62.4 Placebo Combination 85/202 17.4 (13.6–19.9) 50 44.9 Hazard ratio for death, 40 Placebo 0.66 (95% CI, 0.50–0.88) 30 combination P=0.005 Patients Alive (%) 20 26.6 10 0 0 3 6 9 12 15 18 21 24 27 30 33 36 Months since Randomization No. at Risk Tucatinib combination 410 388 322 245 178 123 80 51 34 20 10 4 0 Placebo combination 202 191 160 119 77 48 32 19 7 5 2 1 0

B Subgroup Analysis of Overall Survival No. of Deaths/ Subgroup Total No. Hazard Ratio for Death (95% CI) Total 215/612 0.66 (0.50–0.88) Age ≥65 yr 53/116 0.58 (0.32–1.06) <65 yr 162/496 0.69 (0.50–0.95) Race White 160/444 0.69 (0.50–0.96) Nonwhite 55/168 0.51 (0.28–0.93) Hormone-receptor status Positive for ER, PR, or both 128/370 0.85 (0.59–1.23) Negative for ER and PR 87/242 0.50 (0.31–0.80) Baseline brain metastasis Yes 114/291 0.58 (0.40–0.85) No 101/319 0.72 (0.48–1.08) ECOG performance-status score 0 81/298 0.51 (0.33–0.80) 1 134/314 0.84 (0.59–1.20) Geographic region United States and Canada 148/369 0.68 (0.48–0.95) Rest of the world 67/243 0.63 (0.39–1.03) 0.1 1.0 10.0

Tucatinib Combination Placebo Combination Better Better

Figure 2. Kaplan–Meier Estimates of Overall Survival in the Total Population and Prespecified Subgroups. Panel A shows the Kaplan–Meier estimates of overall survival in the total population, which included 612 patients who underwent randomization. The dashed vertical lines indicate landmark time points. Panel B shows the analysis of overall survival in prespecified subgroups. Data from the patients in the United States and Canada were combined for this analysis.

months (95% CI, 18.3 to 31.0) and 17.4 months ratios across all subgroups were consistent with (95% CI, 13.6 to 19.9), respectively. In the total the hazard ratio in the overall analysis (Fig. 2B). population, the risk of death was 34% lower in In addition, progression-free survival in the total the tucatinib-combination group than in the population was consistent with that in the pri- placebo-combination group (hazard ratio, 0.66; mary end-point analysis (Fig. S3). 95% CI, 0.50 to 0.88; P = 0.005) (Fig. 2A). Hazard Among the patients with brain metastases,

n engl j med 382;7 nejm.org February 13, 2020 603 The New England Journal of Medicine Downloaded from nejm.org at Walaeus Library - Leiden University Medical Center on April 7, 2021. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved. The new england journal of medicine

the estimated progression-free survival at 1 year the tucatinib-combination group and 9.1% in the was 24.9% (95% CI, 16.5 to 34.3) in the tuca- placebo-combination group). tinib-combination group and 0% in the placebo- The most common adverse events observed combination group, and the median duration of among the patients in the tucatinib-combination progression-free survival was 7.6 months (95% group were diarrhea, palmar–plantar erythro- CI, 6.2 to 9.5) and 5.4 months (95% CI, 4.1 to dysesthesia syndrome, nausea, fatigue, and vom- 5.7), respectively. The risk of disease progression iting (Table 2); most events were of grade 1 or 2. or death was 52% lower in the tucatinib-combi- The most common adverse events of grade 3 or nation group than in the placebo-combination higher observed among the patients in the tuca- group (hazard ratio, 0.48; 95% CI, 0.34 to 0.69; tinib-combination group were palmar–plantar P<0.001) (Fig. 3A). Hazard ratios across all sub- erythrodysesthesia syndrome, diarrhea, elevations groups were consistent with the hazard ratio in in alanine aminotransferase (ALT) and aspartate the overall analysis (Fig. 3B). In a prespecified aminotransferase (AST) levels, and fatigue. analysis involving the patients without brain Diarrhea was the most common adverse event metastases, the risk of disease progression or in both the tucatinib-combination group and the death was 43% lower in the tucatinib-combina- placebo-combination group, and most events of tion group than in the placebo-combination group diarrhea were of grade 1 (in 43.3% and 32.0% of (hazard ratio, 0.57; 95% CI, 0.41 to 0.80) (Fig. S4). the patients, respectively) or grade 2 (in 24.8% Among the 511 patients with measurable disease and 12.7%, respectively); diarrhea of grade 3 or at baseline, as assessed by means of blinded in- higher occurred in 12.9% and 8.6%, respectively. dependent central review, the percentage who Antidiarrheal prophylaxis was not mandated. had a confirmed objective response was 40.6% Among the patients who used antidiarrheal (95% CI, 35.3 to 46.0) in the tucatinib-combina- agents, the median duration of use was 3 days tion group and 22.8% (95% CI, 16.7 to 29.8) in the per cycle in both treatment groups. placebo-combination group (P<0.001) (Table S2). Elevations in the ALT and AST levels were mostly low-grade, transient, and reversible; ALT Treatment Duration and AST elevations that were of grade 3 or In the primary end-point analysis population, the higher occurred in 5.4% and 4.5% of the pa- median duration of exposure to tucatinib was tients, respectively, in the tucatinib-combination 7.3 months (range, <0.1 to 35.1), and the median group and in 0.5% of the patients (for each) in duration of exposure to placebo was 4.4 months the placebo-combination group. Elevations in the (range, <0.1 to 24.0); the durations of exposure bilirubin level of any grade occurred in 18.6% of to trastuzumab and capecitabine were also short- the patients in the tucatinib-combination group er in the placebo-combination group. Among and in 10.2% of those in the placebo-combina- the 601 patients who received at least one dose tion group; however, elevations of grade 3 or of any trial drug or placebo, the median duration higher occurred in fewer patients in the tuca- of exposure to tucatinib or placebo was 5.8 tinib-combination group (0.7% vs. 2.5%). A sin- months (range, <0.1 to 35.1) and 4.4 months gle patient in the tucatinib-combination group (range, <0.1 to 24.0), respectively. At the date of met Hy’s law criteria on the basis of ALT, 152 U data cutoff, 118 of 410 patients (28.8%) in the per liter (4.8 times the upper limit of normal tucatinib-combination group and 27 of 202 [ULN]); AST, 128 U per liter (3.6 × ULN); total bili- (13.4%) in the placebo-combination group were rubin, 2.6 mg per deciliter (2.2 × ULN); and alka- continuing to receive their assigned treatment. line phosphatase (ALP), 165 U per liter (1.4 × ULN). This patient had an adaptive response, with re- Safety covery of laboratory values to baseline after dose The occurrence of adverse events according to modification of both tucatinib and capecitabine treatment group is summarized in Table 2 and and continued treatment with all three drugs for Table S3. Adverse events led to the discontinua- six more cycles until disease progression. An in- tion of tucatinib in 5.7% of the patients, to the dependent expert hepatology review subsequently discontinuation of placebo in 3.0% of the pa- concluded that the mild AST and ALT elevations tients (Table S4), and to the discontinuation of and adaptation with retreatment did not qualify capecitabine in 9.8% of the patients (10.1% in this as a true Hy’s law case.

604 n engl j med 382;7 nejm.org February 13, 2020 The New England Journal of Medicine Downloaded from nejm.org at Walaeus Library - Leiden University Medical Center on April 7, 2021. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved. Tucatinib for HER2-Positive Metastatic Breast Cancer

A Kaplan–Meier Estimates of Progression-free Survival among Patients with Brain Metastases 100 Median 90 No. of Events/ Duration No. of Patients (95% CI) 80 mo 70 Tucatinib Combination 106/198 7.6 (6.2–9.5) 60.4 60 Placebo Combination 51/93 5.4 (4.1–5.7) 50 Hazard ratio for disease progression or death, 40 0.48 (95% CI, 0.34–0.69) P<0.001 30 33.9 24.9 20 Disease Progression (%) Placebo Tucatinib Patients Alive and Free from 10 combination combination 0 0 0 3 6 9 12 15 18 21 24 27 30 33 36 Months since Randomization No. at Risk Tucatinib combination 198 144 78 45 14 8 2 1 1 1 1 1 0 Placebo combination 93 49 12 4 0 0 0 0 0 0 0 0 0

B Subgroup Analysis of Progression-free Survival among Patients with Brain Metastases No. of Events/ Subgroup Total No. Hazard Ratio for Disease Progression or Death (95% CI) Total 157/291 0.48 (0.34–0.69) Age ≥65 yr 24/48 0.41 (0.17–1.01) <65 yr 133/243 0.51 (0.35–0.76) Race White 112/199 0.49 (0.32–0.75) Nonwhite 45/92 0.45 (0.23–0.90) Hormone-receptor status Positive for ER, PR, or both 92/166 0.48 (0.31–0.75) Negative for ER and PR 65/125 0.50 (0.27–0.95) ECOG performance-status score 0 76/130 0.68 (0.41–1.15) 1 81/161 0.34 (0.21–0.55) Geographic region United States and Canada 97/177 0.43 (0.27–0.67) Rest of the world 60/114 0.60 (0.33–1.08) 0.1 1.0 10.0

Tucatinib Combination Placebo Combination Better Better

Figure 3. Kaplan–Meier Estimates of Progression-free Survival among the Patients with Brain Metastases in the Total Population and Prespecified Subgroups. Panel A shows the Kaplan–Meier estimates of progression‑free survival among 291 patients with brain metastases in the total population, as assessed by means of blinded independent central review. The number of events is the number of events of disease progression or death. Panel B shows the analysis of progression‑free survival in pre‑ specified subgroups. Data from the patients in the United States and Canada were combined for this analysis.

Increases in the serum creatinine level were (see the Supplementary Appendix). Increases in observed in 13.9% of the patients in the tuca- the serum creatinine level occurred early, re- tinib-combination group and in 1.5% of those in mained clinically nonsignificant during the the placebo-combination group. Tucatinib has been treatment period with no intervention to lower shown to inhibit the multidrug and toxin extru- the level, were reversible, and were not the cause sion protein 1 and 2-K (MATE1 and MATE2-K) of discontinuation of therapy in any patient. transporters, which increases the serum creati- Of 215 deaths that occurred during the trial, nine level without affecting glomerular function the most common cause in both treatment

n engl j med 382;7 nejm.org February 13, 2020 605 The New England Journal of Medicine Downloaded from nejm.org at Walaeus Library - Leiden University Medical Center on April 7, 2021. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved. The new england journal of medicine

Table 2. Most Common Adverse Events.*

Tucatinib-Combination Group Placebo-Combination Group Event (N = 404) (N = 197)

Any Grade Grade ≥3 Any Grade Grade ≥3

number of patients (percent) Any adverse event 401 (99.3) 223 (55.2) 191 (97.0) 96 (48.7) Diarrhea 327 (80.9) 52 (12.9) 105 (53.3) 17 (8.6) PPE syndrome 256 (63.4) 53 (13.1) 104 (52.8) 18 (9.1) Nausea 236 (58.4) 15 (3.7) 86 (43.7) 6 (3.0) Fatigue 182 (45.0) 19 (4.7) 85 (43.1) 8 (4.1) Vomiting 145 (35.9) 12 (3.0) 50 (25.4) 7 (3.6) Stomatitis 103 (25.5) 10 (2.5) 28 (14.2) 1 (0.5) Decreased appetite 100 (24.8) 2 (0.5) 39 (19.8) 0 Headache 87 (21.5) 2 (0.5) 40 (20.3) 3 (1.5) Aspartate aminotransferase 86 (21.3) 18 (4.5) 22 (11.2) 1 (0.5) ­increased Alanine aminotransferase 81 (20.0) 22 (5.4) 13 (6.6) 1 (0.5) ­increased

* Listed are adverse events that were reported in at least 20% of the patients in the tucatinib-combination group. Safety analyses included all the patients who received at least one dose of any trial drug or placebo. Data are reported according to preferred terms in the Medical Dictionary for Regulatory Activities, version 22.0. PPE denotes palmar–plantar erythro‑ dysesthesia.

groups was disease progression. Adverse events among patients with metastatic HER2-positive were the cause of death in 6 of 404 patients breast cancer with brain metastases. Previous (1.5%) in the tucatinib-combination group (car- studies have shown that the combination of a diac arrest, cardiac failure, dehydration, multiple- HER2 tyrosine kinase inhibitor with capecitabine organ dysfunction syndrome, sepsis, and septic has some activity in brain metastases from shock in 1 patient each) and in 5 of 197 patients HER2-positive breast cancer.17,25 However, toxic (2.5%) in the placebo-combination group (cardiac effects, including those associated with inhibi- arrest, multiple-organ dysfunction syndrome, tion of epidermal growth factor receptor, have myocardial infarction, sepsis, and systemic in- limited the use of these combinations. flammatory response syndrome in 1 patient each). The survival benefit with tucatinib was ob- served in the total HER2CLIMB trial population Discussion and across all subgroups tested. In addition, the results showed that, in combination with Among patients with HER2-positive metastatic capecitabine, simultaneous targeting of the in- breast cancer previously treated with trastuzu­ ternal domain of HER2, as well as the external mab, pertuzumab, and trastuzumab emtansine, domain with trastuzumab, led to substantially the addition of tucatinib to trastuzumab and better survival outcomes than did targeting the capecitabine resulted in a clinically meaningful external domain alone. lower risk of disease progression or death than Patients were enrolled across 155 sites in 15 the addition of placebo. Most importantly, over- countries, and all patients had received previous all survival was longer by 4.5 months with tuca- treatment with the current standard of care tinib (median of 21.9 months vs. 17.4 months). (trastuzumab, pertuzumab, and trastuzumab The tucatinib combination was associated emtansine). The comparator group received a with a significantly lower risk of disease pro- recommended regimen for management in this gression or death than the placebo combination clinical context, as supported by a randomized,

606 n engl j med 382;7 nejm.org February 13, 2020 The New England Journal of Medicine Downloaded from nejm.org at Walaeus Library - Leiden University Medical Center on April 7, 2021. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved. Tucatinib for HER2-Positive Metastatic Breast Cancer

phase 3 trial that showed longer progression- fees and meal reimbursement from R-Pharm and Mylan, advi- free and overall survival among patients treated sory board fees, travel support, and meal reimbursement from Pfizer, and meal reimbursement from Amgen and Tesaro; Dr. with trastuzumab plus capecitabine than among Hamilton, receiving grant support, paid to her institution, and those treated with lapatinib plus capecitabine.26 consulting fees, paid to her institution, from Pfizer, Puma Bio- Survival outcomes in the placebo-combination technology, Mersana, Boehringer Ingleheim, Cascadian Thera- peutics, Silverback, and Black Diamond, grant support, paid to group in our trial were similar to those in other her institution, consulting fees, paid to her institution, and re- contemporary trials of HER2-targeted therapy.25,27 imbursement of expenses from Roche/Genentech, Lilly, Daiichi In our trial, we enrolled a large percentage of Sankyo, AstraZeneca, Eisai, and Novartis, grant support, paid to her institution, from Hutchinson MediPharma, OncoMed, Med- patients with untreated or previously treated Immune, Stem CentRx, Curis, Verastem, Zymeworks, Syndax, progressing brain metastases, a population typi- Lycera, Rgenix, Millennium, TapImmune, BerGenBio, Mediva- cally excluded from clinical trials despite this tion, H3 Biomedicine, Acerta Pharma, Takeda, Macrogenics, AbbVie, Immunomedics, Fujifilm, eFECTOR Therapeutics, Merus, condition being a common clinical problem. The Nucana, Regeneron, Leap Therapeutics, Taiho Pharmaceutical, inclusion of patients with brain metastases in Syros, CytomX, InventisBio, Deciphera, ArQule, Sermonix Phar- the initial phase 1 trials of tucatinib generated maceuticals, Sutro Biopharma, Zenith Epigenetics, Arvinas, Torque Therapeutics, Harpoon Therapeutics, Fochon Pharma- preliminary efficacy data to support their inclu- ceuticals, Orinove, Molecular Templates, Unum Therapeutics, sion in HER2CLIMB and is in line with current Aravive, Compugen, G1 Therapeutics, Karyopharm Therapeu- recommendations from the American Society of tics, and Torque Therapeutics, grant support, paid to her insti- tution, and reimbursement of expenses from Tesaro, Radius Clinical Oncology–Friends of Cancer Research Health, EMD Serono, and Clovis, reimbursement of expenses Eligibility Criteria Working Group.28 from Amgen, Bayer, Bristol-Myers Squibb, Genzyme, Helsinn Tucatinib in combination with trastuzumab Therapeutics, HERON, Lexicon, Medivation, Merck, Roche, Sys- mex, Guardant Health, and Foundation Medicine, and consult- and capecitabine was associated with toxic ef- ing fees, paid to her institution, from NanoString; Dr. Hurvitz, fects, with the majority of adverse events being receiving grant support from Amgen, Ambrx, Bayer, GSK, Immu- low grade. Safety events of note included diar- nomedics, Macrogenics, Novartis, Pieris, Radius, and Sanofi, and grant support and medical writing support from Daiichi rhea that was managed with short courses of Sankyo, Roche/Genentech, Lilly, Pfizer, OBI Pharma, Puma Bio- antidiarrheal agents and transient, reversible ele- technology, and Dignitana; Dr. Lin, receiving grant support vations in liver enzyme levels. A majority of pa- from Array Biopharma, Genentech, and Merck, consulting fees from Daiichi Sankyo, grant support, advisory board fees, con- tients in the tucatinib-combination group had sulting fees, and travel support from Pfizer/NCCN, advisory toxic effects, but less than 6% of the patients board fees and travel support from Puma Biotechnology, and discontinued treatment with tucatinib as a con- grant support, advisory board fees, and consulting fees from Seattle Genetics; Dr. Abramson, receiving advisory board fees sequence. from Daiichi Sankyo, AbbVie, and Novartis, and grant support In conclusion, tucatinib plus trastuzumab and from Genentech; Dr. Anders, receiving research funding and capecitabine is an active combination in heavily consulting fees from Puma Biotechnology and IPSEN, research funding from Lilly, Merck, Nektar, Tesaro, and G1 Therapeutics, pretreated patients with HER2-positive meta- fees for serving on a data and safety monitoring board from static breast cancer, including those with previ- Roche/Genentech, consulting fees from Eisai, and advisory ously untreated, treated and stable, or treated board fees from Seattle Genetics; Dr. Bedard, receiving grant support, paid to his institution, from and serving on advisory and progressing brain metastases. boards for Bristol-Myers Squibb, Sanofi, and Roche/Genentech, Supported by Seattle Genetics. receiving grant support, paid to his institution, from Novartis, Dr. Murthy reports receiving grant support, paid to her institu- GlaxoSmithKline, Nektar Therapeutics, Merck, Lilly, Servier, and tion, from EMD-Serono and Pfizer, grant support, paid to her in- PTC Therapeutics, and serving on advisory boards for Pfizer; Dr. stitution, consulting fees, and advisory board fees from Daiichi Oliveira, receiving grant support, consulting fees, lecture fees, Sankyo, grant support, paid to her institution, consulting fees, and travel support from Roche, grant support and consulting advisory board fees, and travel support from Roche/Genentech, fees from Genentech, GlaxoSmithKline, Puma Biotechnology, consulting fees, advisory board fees, and travel support from and AstraZeneca, grant support and lecture fees from Seattle Puma, and consulting fees, advisory board fees, travel support, Genetics, grant support from Immunomedics and Boehringer and medical writing support from Seattle Genetics; Dr. Loi, re- Ingelheim, travel support from Pierre-Fabre, Eisai, GP Pharma, ceiving research funding, paid to her institution, from and serv- and Grünenthal, and grant support, lecture fees, and travel sup- ing as a consultant for Novartis, Bristol-Myers Squibb, Roche/ port from Novartis; Dr. Jakobsen, receiving advisory board fees Genentech, and Pfizer, receiving research funding, paid to her from Pfizer, Novartis, and Roche; Dr. Bachelot, receiving advi- institution, from Puma Biotechnology and Eli Lilly, serving as a sory board fees and travel support from Roche, grant support, consultant for Seattle Genetics and Merck, and receiving con- advisory board fees, lecture fees, and travel support from Novar- sulting fees, paid to her institution, from Aduro Biotech; Dr. tis, grant support, advisory board fees, and travel support from Okines, receiving lecture fees from Roche; Dr. Paplomata, re- AstraZeneca and Pfizer, and advisory board fees from Seattle ceiving grant support, paid to her institution, and meal reim- Genetics; Dr. Shachar, receiving lecture fees from Novartis and bursement from Genentech, Merck, and Corcept, grant support, Pfizer, and lecture fees and travel support from Roche; Dr. Mül- paid to her institution, advisory board fees, travel support, and ler, receiving grant support, paid to his institution, lecture fees, meal reimbursement from Novartis, grant support, paid to her and consulting fees from Roche, Novartis, and Pfizer, consult- institution, and logistical support from AbbVie, advisory board ing fees from Amgen and AstraZeneca, lecture fees and consult-

n engl j med 382;7 nejm.org February 13, 2020 607 The New England Journal of Medicine Downloaded from nejm.org at Walaeus Library - Leiden University Medical Center on April 7, 2021. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved. The new england journal of medicine

ing fees from Daiichi Sankyo, Eisai, Teva, Tesaro, and Hexal, EP14153692.0 on a method for predicting the response to anti– and grant support, paid to his institution, and consulting fees human epidermal growth factor receptor 2 (HER2)-containing from Nektar; Dr. Duhoux, receiving consulting fees, paid to his therapy, chemotherapy, or both in patients with breast cancer; institution, and travel support from Roche, Pfizer, and Amgen, Dr. Pegram, receiving consulting fees from Seattle Genetics, consulting fees, paid to his institution, from AstraZeneca, Lilly, Roche/Genentech, Pfizer, Samsung Bioepis, AstraZeneca, and Novartis, and Mundipharma, and travel support from Teva; Dr. Puma Biotechnology; Dr. Slamon, receiving travel support from, Greil, receiving grant support, consulting fees, and travel sup- serving on the board of directors for, and owning stock in Bio- port from Celgene, Roche, Takeda, AstraZeneca, Novartis, Am- marin, receiving grant support and travel support from and gen, Bristol-Myers Squibb, MSD, AbbVie, Gilead, and Janssen, owning stock in Pfizer, receiving grant support, advisory board consulting fees and travel support from Merck and Daiichi San- fees, consulting fees, and travel support from Novartis, receiv- kyo, and grant support and consulting fees from Sandoz; Dr. ing consulting fees from Eli Lilly, and owning stock in Amgen Cameron, receiving grant support, consulting fees, and fees for and Seattle Genetics; Dr. Palanca-Wessels, being employed by independent data monitoring committee work from Roche, Seattle Genetics; Dr. Walker, being employed by Seattle Genetics, grant support and consulting fees from Novartis, fees for inde- holding pending patent 62/923,659 on methods of treating HER2- pendent data monitoring committee work from Synthon, and positive breast cancer with tucatinib in combination with capeci­ consulting fees from Daiichi Sankyo, Samsung BIOEPSIS, Puma tabine and trastuzumab, holding pending patent 62/797,854 on Biotechnology, Seattle Genetics, and Zymeworks (all support methods of treating breast cancer with tucatinib, and holding paid to his institution); Dr. Curigliano, receiving advisory board pending patent PCT/US2018/029899 on treatment of HER2-pos- fees from Seattle Genetics, Roche, Pfizer, Daiichi Sankyo, and itive cancers; Dr. Feng, being employed by Seattle Genetics; and Lilly; Dr. Gelmon, receiving grant support, paid to BC Cancer, Dr. Winer, receiving consulting fees from Carrick Therapeutics, and advisory board fees from Roche, grant support and advisory Roche/Genentech, Genomic Health, GlaxoSmithKline, Jounce, board fees from Pfizer, AstraZeneca, and Nanostring, advisory Lilly, Merck, and Seattle Genetics, and consulting fees and advi- board fees from Novartis, Mylan, Merck, Lilly, and Genomic sory board fees from Leap. No other potential conflict of inter- Health, grant support from Bristol-Myers Squibb, and fees for est relevant to this article was reported. expert testimony from Genentech; Dr. Krop, receiving grant Disclosure forms provided by the authors are available with support, advisory board fees, consulting fees, and travel support the full text of this article at NEJM.org. from Roche/Genentech and Daiichi Sankyo, grant support from A data sharing statement provided by the authors is available Pfizer, advisory board fees and travel support from Macrogen- with the full text of this article at NEJM.org. ics, Context Therapeutics, and Taiho Oncology, fees for serving We thank the patients who participated in this trial and on a data and safety monitoring board from Merck and Novar- their families, as well as the investigators and staff at all HER- tis, and lecture fees from AstraZeneca; Dr. Loibl, receiving grant 2CLIMB clinical sites; the members of the independent data support, lecture fees, and advisory board fees, paid to her insti- and safety monitoring committee and the independent review tution, from AbbVie, Amgen, AstraZeneca, Celgene, Novartis, committee; Andres Forero-Torres, M.D., and Matthew Blahna, Pfizer, and Roche, lecture fees and advisory board fees, paid to Ph.D., Seattle Genetics, for critical review and revision of the her institution, from Seattle Genetics and Samsung, lecture manuscript; Laurie LaRusso, M.S., Chestnut Medical Commu- fees, paid to her institution, from prIME/Medscape, lecture fees nications, for writing support during the development of the from Chugai, grant support from Teva and Vifor, grant support, manuscript; Jorge Ramos, D.O., JoAl Mayor, Pharm.D., paid to her institution, from Daiichi Sankyo, advisory board B.C.O.P., and Suzanne McGoldrick, M.D., for the medical mon- fees, paid to her institution, from Lilly, and consulting fees, paid itoring of the trial; and the entire HER2CLIMB trial team for to her institution, from Eirgenix, and holding pending patent their diligent data review.

Appendix The authors’ full names and academic degrees are as follows: Rashmi K. Murthy, M.D., Sherene Loi, M.D., Alicia Okines, M.D., Elisa- vet Paplomata, M.D., Erika Hamilton, M.D., Sara A. Hurvitz, M.D., Nancy U. Lin, M.D., Virginia Borges, M.D., Vandana Abramson, M.D., Carey Anders, M.D., Philippe L. Bedard, M.D., Mafalda Oliveira, M.D., Erik Jakobsen, M.D., Thomas Bachelot, M.D., Shlomit S. Shachar, M.D., Volkmar Müller, M.D., Sofia Braga, M.D., Francois P. Duhoux, M.D., Richard Greil, M.D., David Cameron, M.D., Lisa A. Carey, M.D., Giuseppe Curigliano, M.D., Ph.D., Karen Gelmon, M.D., Gabriel Hortobagyi, M.D., Ian Krop, M.D., Ph.D., Sibylle Loibl, M.D., Mark Pegram, M.D., Dennis Slamon, M.D., M. Corinna Palanca‑Wessels, M.D., Ph.D., Luke Walker, M.D., Wentao Feng, Ph.D., and Eric P. Winer, M.D. The authors’ affiliations are as follows: M.D. Anderson Cancer Center, Houston (R.K.M., G.H.); Peter MacCallum Cancer Centre, Melbourne, VIC, Australia (S. Loi); the Royal Marsden NHS Foundation Trust, London (A.O.), and Edinburgh Cancer Research Centre, Edinburgh (D.C.) — both in the United Kingdom; Winship Cancer Institute, Atlanta (E.P.); Sarah Cannon Research Institute/Tennessee Oncology–Nashville (E.H.) and Vanderbilt University Medical Center (V.A.), Nashville; University of California, Los Angeles, Medical Center–Jonsson Comprehensive Cancer Center, Los Angeles (S.A.H., D.S.), and Stanford Comprehensive Cancer Institute, Palo Alto (M.P.) — both in California; Dana–Farber Cancer Institute, Boston (N.U.L., I.K., E.P.W.); University of Colorado Cancer Center, Au- rora (V.B.); Duke Cancer Institute, Durham (C.A.), and University of North Carolina Lineberger Comprehensive Cancer Center, Chapel Hill (L.A.C.) — both in North Carolina; University Health Network, Princess Margaret Cancer Centre, Toronto (P.L.B.), and British Columbia Cancer, Vancouver (K.G.) — both in Canada; Hospital Universitario Vall D’Hebron, Barcelona (M.O.); Sygehus Lillebaelt– Vejle Sygehus, Vejle, Denmark (E.J.); Centre Léon Bérard, Lyon, France (T.B.); Rambam Health Care Campus, Haifa, Israel (S.S.S.); Universitaetsklinikum Hamburg–Eppendorf, Hamburg (V.M.), and German Breast Group, Neu-Isenburg (S. Loibl) — both in Germany; Hospital Cuf Descobertas R. Mário Botas, Lisbon, Portugal (S.B.); Cliniques Universitaires Saint-Luc, Brussels (F.P.D.); Third Medical Department, Paracelsus Medical University Salzburg, Salzburg Cancer Research Institute–Center for Clinical Cancer and Immunology Trials, and Cancer Cluster Salzburg, Salzburg, Austria (R.G.); Istituto Europeo di Oncologia, IRCCS, University of Milan, Milan (G.C.); and Seattle Genetics, Bothell, WA (M.C.P.-W., L.W., W.F.).

608 n engl j med 382;7 nejm.org February 13, 2020 The New England Journal of Medicine Downloaded from nejm.org at Walaeus Library - Leiden University Medical Center on April 7, 2021. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved. Tucatinib for HER2-Positive Metastatic Breast Cancer

References 1. Owens MA, Horten BC, Da Silva MM. 11. Giordano SH, Temin S, Davidson NE. et al. In vitro and in vivo activity of ARRY- HER2 amplification ratios by fluores- Systemic therapy for patients with ad- 380: a potent, small molecule inhibitor of cence in situ hybridization and correla- vanced human epidermal growth factor ErbB2. Cancer Res 2009;​69:​1795. abstract. tion with immunohistochemistry in a co- receptor 2-positive breast cancer: ASCO 21. Murthy R, Borges VF, Conlin A, et al. hort of 6556 breast cancer tissues. Clin clinical practice guideline update sum- Tucatinib with capecitabine and trastuzu­ Breast Cancer 2004;5:​ 63-9.​ mary. J Oncol Pract 2018;​14:​501-4. mab in advanced HER2-positive metastat- 2. Cronin KA, Harlan LC, Dodd KW, 12. Cardoso F, Senkus E, Costa A, et al. ic breast cancer with and without brain Abrams JS, Ballard-Barbash R. Popula- 4th ESO-ESMO international consensus metastases: a non-randomised, open-label, tion-based estimate of the prevalence of guidelines for advanced breast cancer phase 1b study. Lancet Oncol 2018;​19:​ HER-2 positive breast cancer tumors for (ABC 4). Ann Oncol 2018;​29:​1634-57. 880-8. early stage patients in the US. Cancer In- 13. National Comprehensive Cancer Net- 22. Eisenhauer EA, Therasse P, Bogaerts J, vest 2010;​28:​963-8. work (NCCN). Clinical practice guide- et al. New response evaluation criteria in 3. Baselga J, Cortés J, Kim S-B, et al. Per- lines in oncology:​ central nervous system solid tumours: revised RECIST guideline tuzumab plus trastuzumab plus docetaxel cancers — breast cancer. Version 3.2019. (version 1.1). Eur J Cancer 2009;​45:​228- for metastatic breast cancer. N Engl J Med October 18, 2019 (https://www​.nccn​.org/​ 47. 2012;366:​ 109-19.​ professionals/physician_gls/​ pdf/​ cns​ .pdf).​ 23. Ye Y, Li A, Liu L, Yao B. A group se- 4. Swain SM, Kim SB, Cortés J, et al. Per- 14. Lin NU, Diéras V, Paul D, et al. Multi- quential Holm procedure with multiple tuzumab, trastuzumab, and docetaxel for center phase II study of lapatinib in pa- primary endpoints. Stat Med 2013;​32:​1112- HER2-positive metastatic breast cancer tients with brain metastases from HER2- 24. (CLEOPATRA study): overall survival re- positive breast cancer. Clin Cancer Res 24. Rosenberger WF, Lachin JM. Ran- sults from a randomised, double-blind, 2009;​15:​1452-9. domization in clinical trials:​ theory and placebo-controlled, phase 3 study. Lancet 15. Sutherland S, Ashley S, Miles D, et al. practice. Hoboken, NJ:​ John Wiley and Oncol 2013;14:​ 461-71.​ Treatment of HER2-positive metastatic Sons, 2016. 5. Bendell JC, Domchek SM, Burstein breast cancer with lapatinib and capecita­ 25. Saura C, Oliveira M, Feng Y-H, et al. HJ, et al. Central nervous system metasta- bine in the lapatinib expanded access pro- Neratinib + capecitabine versus lapatinib ses in women who receive trastuzumab- gramme, including efficacy in brain metas- + capecitabine in patients with HER2+ based therapy for metastatic breast carci- tases — the UK experience. Br J Cancer metastatic breast cancer previously treat- noma. Cancer 2003;97:​ 2972-7.​ 2010;102:​ 995-1002.​ ed with ≥ 2 HER2-directed regimens: 6. Brufsky AM, Mayer M, Rugo HS, et al. 16. Bachelot T, Romieu G, Campone M, findings from the multinational, random- Central nervous system metastases in pa- et al. Lapatinib plus capecitabine in pa- ized, phase III NALA trial. J Clin Oncol tients with HER2-positive metastatic tients with previously untreated brain 2019;​37:​1002. abstract. breast cancer: incidence, treatment, and metastases from HER2-positive metastat- 26. Pivot X, Manikhas A, Żurawski B, et al. survival in patients from registHER. Clin ic breast cancer (LANDSCAPE): a single- CEREBEL (EGF111438): a phase III, ran- Cancer Res 2011;​17:​4834-43. group phase 2 study. Lancet Oncol 2013;​ domized, open-label study of lapatinib 7. Leyland-Jones B. Human epidermal 14:64-71.​ plus capecitabine versus trastuzumab plus growth factor receptor 2-positive breast 17. Freedman RA, Gelman RS, Anders capecitabine in patients with human epi- cancer and central nervous system metas- CK, et al. TBCRC 022: a phase II trial of dermal growth factor receptor 2–positive tases. J Clin Oncol 2009;​27:​5278-86. neratinib and capecitabine for patients metastatic breast cancer. J Clin Oncol 8. Olson EM, Najita JS, Sohl J, et al. Clin- with human epidermal growth factor re- 2015;33:​ 1564-73.​ ical outcomes and treatment practice pat- ceptor 2-positive breast cancer and brain 27. Rugo HS, Im S-A, Wright GLS, et al. terns of patients with HER2-positive metastases. J Clin Oncol 2019;​37:​1081-9. SOPHIA primary analysis: a phase 3 (P3) metastatic breast cancer in the post- 18. Krop IE, Lin NU, Blackwell K, et al. study of margetuximab (M) + chemother- trastuzumab era. Breast 2013;​22:​525-31. Trastuzumab emtansine (T-DM1) versus apy (C) versus trastuzumab (T) + C in pa- 9. Pestalozzi BC, Holmes E, de Azam- lapatinib plus capecitabine in patients tients (pts) with HER2+ metastatic (met) buja E, et al. CNS relapses in patients with with HER2-positive metastatic breast breast cancer (MBC) after prior anti-HER2 HER2-positive early breast cancer who have cancer and central nervous system metas- therapies (Tx). J Clin Oncol 2019;​37:​1000. and have not received adjuvant trastuzumab: tases: a retrospective, exploratory analysis abstract. a retrospective substudy of the HERA trial in EMILIA. Ann Oncol 2015;​26:​113-9. 28. Lin NU, Prowell T, Tan AR, et al. (BIG 1-01). Lancet Oncol 2013;​14:​244-8. 19. Moulder SL, Borges VF, Baetz T, et al. Modernizing clinical trial eligibility cri- 10. National Comprehensive Cancer Net- Phase I study of ONT-380, a HER2 inhibi- teria: recommendations of the American work (NCCN). Clinical practice guide- tor, in patients with HER2+-advanced solid Society of Clinical Oncology–Friends of lines in oncology: ​breast cancer. Version tumors, with an expansion cohort in HER2+ Cancer Research Brain Metastases Work- 3.2018. October 25, 2018 (http://www​ metastatic breast cancer (MBC). Clin Can- ing Group. J Clin Oncol 2017;​35:​3760-73. .nccn​.org/​professionals/​physician_gls/​pdf/​ cer Res 2017;​23:​3529-36. Copyright © 2019 Massachusetts Medical Society. breast​.pdf). 20. Pheneger T, Bouhana K, Anderson D,

n engl j med 382;7 nejm.org February 13, 2020 609 The New England Journal of Medicine Downloaded from nejm.org at Walaeus Library - Leiden University Medical Center on April 7, 2021. For personal use only. No other uses without permission. Copyright © 2020 Massachusetts Medical Society. All rights reserved. Articles

Pertuzumab, trastuzumab, and docetaxel for HER2-positive metastatic breast cancer (CLEOPATRA): end-of-study results from a double-blind, randomised, placebo-controlled, phase 3 study

Sandra M Swain, David Miles, Sung-Bae Kim, Young-Hyuck Im, Seock-Ah Im, Vladimir Semiglazov, Eva Ciruelos, Andreas Schneeweiss, Sherene Loi, Estefanía Monturus, Emma Clark, Adam Knott, Eleonora Restuccia, Mark C Benyunes, Javier Cortés, on behalf of the CLEOPATRA study group*

Summary Background CLEOPATRA was a phase 3 study comparing the efficacy and safety of pertuzumab, trastuzumab, and Lancet Oncol 2020; 21: 519–30 docetaxel with placebo, trastuzumab, and docetaxel in patients with HER2-positive metastatic breast cancer. In the Published Online primary analysis and subsequent reports, progression-free and overall survival were significantly improved in the March 12, 2020 pertuzumab group compared with the placebo group. Here, we report the end-of-study analysis of CLEOPATRA. https://doi.org/10.1016/ S1470-2045(19)30863-0 See Comment page 471 Methods This was a double-blind, randomised, placebo-controlled, phase 3 trial that was done at 204 centres in *A complete list of investigators 25 countries. Eligible patients were 18 years or older, had HER2-positive, metastatic breast cancer, had not received is provided in the previous chemotherapy or biological treatment for their metastatic disease, and had an Eastern Cooperative Oncology appendix (pp 2–9) Group performance status of 0 or 1. All study drugs were given intravenously, every 3 weeks. Patients were assigned Georgetown University to receive either pertuzumab or placebo at a loading dose of 840 mg, and 420 mg thereafter; plus trastuzumab at Medical Center, Washington 8 mg/kg loading dose and 6 mg/kg thereafter; and docetaxel at 75 mg/m², escalating to 100 mg/m² if tolerated. DC, USA (Prof S M Swain MD); Lombardi Comprehensive Pertuzumab or placebo and trastuzumab were given until disease progression; docetaxel was given for six cycles, or Cancer Center, Washington, DC, longer at the investigators’ discretion. Randomisation (1:1) was done by use of an interactive voice-response system USA (Prof S M Swain); MedStar and was stratified by geographical region (Asia, Europe, North America, or South America) and previous treatment Health, Washington, DC, USA (previous adjuvant or neoadjuvant chemotherapy vs none). The primary endpoint was independent review facility- (Prof S M Swain); Mount Vernon Cancer Centre, Northwood, UK assessed progression-free survival, which has been reported previously. Since the confirmatory overall survival (Prof D Miles MD); Department analysis had also occurred before this prespecified end-of-study analysis, analyses presented here are descriptive. of Oncology, Asan Medical Overall survival analyses were based on the intention-to-treat population with crossover patients analysed in the Center, University of Ulsan placebo group; analyses were not adjusted for crossover to the pertuzumab group and are likely to be conservative. College of Medicine, Seoul, Korea (Prof S-B Kim MD); Safety analyses were based on treatment received; crossover patients were counted in the placebo group up to the day Department of Medicine, before first pertuzumab dose. This trial is registered with ClinicalTrials.gov, number NCT00567190. Samsung Medical Center, Sungkyunkwan University Findings Between Feb 12, 2008, and July 7, 2010, 1196 patients were assessed for eligibility, of whom 808 were enrolled School of Medicine, Seoul, Korea (Prof Y-H Im MD); Seoul and randomly assigned. 402 patients were assigned to receive docetaxel plus trastuzumab plus pertuzumab, and National University Hospital, 406 patients were assigned to receive docetaxel plus trastuzumab plus placebo. Clinical cutoff for this analysis was Cancer Research Institute, Nov 23, 2018. Between July 2012 and clinical cutoff, 50 patients crossed from the placebo to the pertuzumab group. Seoul National University Median follow-up was 99·9 months in the pertuzumab group (IQR 92·9–106·4) and 98·7 months (90·9–105·7) in College of Medicine, Seoul, Korea (Prof S-A Im MD); the placebo group. Median overall survival was 57·1 months (95% CI 50–72) in the pertuzumab group and N N Petrov Research Institute 40·8 months (36–48) in the placebo group (hazard ratio 0·69, 95% CI 0·58–0·82); 8-year landmark overall survival of Oncology, St Petersburg, rates were 37% (95% CI 31–42) in the pertuzumab group and 23% (19–28) in the placebo group. The most common Russia (Prof V Semiglazov MD); grade 3–4 adverse event was neutropenia (200 [49%] of 408 patients in the pertuzumab group, 183 [46%] of 396 patients 12 de Octubre University Hospital, Medical Oncology in the placebo group). Five (1%) of 408 patients in the pertuzumab group and six (2%) of 396 patients in the placebo Department, Madrid, Spain group had treatment-related deaths. One new serious adverse event suggestive of congestive heart failure (pertuzumab (E Ciruelos MD); National Center group) and one new symptomatic left ventricular systolic dysfunction (post-crossover) occurred since the previous for Tumor Diseases, University analysis. Hospital, German Cancer Research Center, Heidelberg, Germany Interpretation Our analysis shows that the previously observed improvements in overall survival with pertuzumab, (Prof A Schneeweiss MD); trastuzumab, and docetaxel versus placebo, trastuzumab, and docetaxel were maintained after a median of more than Peter MacCallum Cancer 8 years of follow-up. The long-term safety and cardiac safety profiles of pertuzumab, trastuzumab, and docetaxel were Centre, Melbourne, VIC, Australia (Prof S Loi MD); maintained in the overall safety population and within crossover patients. HER2-targeted therapy has changed the F Hoffmann-La Roche, Basel, natural history of HER2-positive metastatic breast cancer, with the dual blockade of pertuzumab and trastuzumab, Switzerland (E Monturus PhD, with docetaxel, demonstrating an 8-year landmark overall survival rate of 37%. E Restuccia MD); Roche Products, Welwyn Garden City, UK (E Clark MSc, Funding F Hoffmann-La Roche and Genentech. www.thelancet.com/oncology Vol 21 April 2020 519 Articles

A Knott PhD); Genentech, Copyright © 2020 Elsevier Ltd. All rights reserved. South San Francisco, CA, USA (M C Benyunes MD); in the curative setting and has become available for IOB Institute of Oncology, Introduction Quirónsalud Group, Madrid Metastatic breast cancers with overexpression of HER2 the neoadjuvant and adjuvant treatment of patients with and Barcelona, Spain (known as HER2-positive breast cancers) were historically early HER2-positive breast cancer at high risk of (Prof J Cortés MD); known to be aggressive and associated with poor recurrence.3–5,7–10 and Vall d’Hebron Institute of 1 2 Oncology (VHIO), Barcelona, prognosis until Slamon and colleagues showed that the Despite the survival improvement with the pertuzumab- Spain (Prof J Cortés) anti-HER2 monoclonal antibody trastuzumab signifi­ based regimen, patients eventually progress and die. Correspondence to: cantly improved progression-free survival and overall Trastuzumab emtansine (T-DM1), an antibody–drug Prof Sandra M Swain, survival when added to chemotherapy, compared with conjugate engineered to deliver potent chemotherapy Georgetown University chemo­therapy alone. In the primary analysis of directly to HER2-positive cancer cells, is the standard of Medical Center, the Clinical Evaluation of Pertuzumab and Trastuzumab care for the second-line treatment of advanced HER2- Washington DC, 20057-1400 3 11 sandra.swain@georgetown. (CLEOPATRA) study, adding pertuzumab to trastu­ positive disease, and reduces the risk of recurrence when edu zumab and docetaxel for the first-line treatment of given in the adjuvant setting to patients with HER2- See Online for appendix HER2-positive metastatic breast cancer further improved positive early breast cancer and residual invasive disease clinical outcomes and resulted in significantly improved following standard neoadjuvant therapy and surgery.12 In independently assessed median progression-free survival later lines, there is no standard of care, and patients compared with placebo, trastu­zumab, and docetaxel might receive trastuzumab in combination with a (hazard ratio [HR] 0·62, 95% CI 0·51–0·75; p<0·001). A chemotherapy agent other than docetaxel, or with first interim analysis of overall survival showed a non- lapatinib (an oral tyrosine-kinase inhibitor that reversibly significant improvement (HR 0·64, 95% CI 0·47–0·88; inhibits HER1 or EGFR and HER2 kinases), or lapatinib p=0·005),3 and the second interim overall survival with capecitabine, or a combination of trastuzumab or analysis showed a significant improvement in the lapatinib with endocrine therapy in hormone receptor- pertuzumab group (0·66, 0·52–0·84; p=0·0008).4 In positive disease. Several novel anti-HER2 treatments with the final analysis, median overall survival was diverse mechanisms of action targeting the HER2 also significantly improved with pertuzumab (0·68, signalling pathway are being investigated in the advanced 0·56–0·84; p<0·001) to 56·5 months (15·7 months HER2-positive breast cancer setting, in attempts to longer than with placebo).5 In all analyses, the safety overcome acquired resistance to HER2-targeted therapies. profiles of the two regimens, including the cardiac safety Predictive or prognostic biomarkers might help to profiles, remained consistent and were similar.3–6 identify patients with the best chance of responding to The pertuzumab-based regimen thus became the treatment. Biomarker analyses of CLEOPATRA con­ standard of care for the first-line treatment of HER2- cluded that, although HER2 remains the only marker positive metastatic breast cancer based on CLEOPATRA. suitable for selecting patients for pertuzumab, trastu­ The regimen was also shown to deliver significant benefit zumab, and docetaxel therapy in this indication, HER2,

Research in context Evidence before this study docetaxel in HER2-positive metastatic breast cancer; Metastatic breast cancer is incurable and disease progression is providing substantiation that the benefits shown in previous unfortunately inevitable. We searched PubMed using the terms analyses are maintained with long-term follow-up (median “overall survival”, “trastuzumab”, “HER2-positive”, and follow-up of > 8 years), and valuable information about long- “metastatic breast cancer” and restricted results to those of term safety aspects (including cardiac safety). This study also first-line phase 2 and 3 prospective clinical trials published provides data on the clinical characteristics, including between Feb 19, 2015, and Aug 7, 2019 —ie, since the previous biomarkers, of patients who can be classified as long-term analysis. The search results supported that, despite other therapy responders. combinations and agents being investigated in this setting, the Implications of all the available evidence overall survival benefit of an additional 15·7 months with the The combination of pertuzumab, trastuzumab, and docetaxel addition of pertuzumab to trastuzumab and docetaxel in remains the standard of care for the first-line treatment of CLEOPATRA remained the longest in the first-line setting, and HER2-positive metastatic breast cancer, owing to its overall that longer follow-up was needed for other combinations. The survival benefits and maintained long-term overall and absence of long-term follow-up in other prospective studies also cardiac safety. Patients with specific characteristics could leaves a gap in the knowledge with regards to which patients are benefit from the addition of pertuzumab to trastuzumab and likely to continue to respond to long-term treatment. chemotherapy. Added value of this study Overall, this study adds to the evidence supporting the overall survival benefit of first-line pertuzumab, trastuzumab, and

520 www.thelancet.com/oncology Vol 21 April 2020 Articles

HER3, and PIK3CA were relevant prognostic factors for Tumour and serum samples were mandatory, and progression-free survival.13 PIK3CA showed the greatest most biopsy samples taken for biomarker analyses were prognostic effect: median progression-free survival was from the primary tumour.13 longer in both study groups for patients with PIK3CA- CLEOPATRA was conducted in accordance with the wild-type tumours compared with those with mutated guidelines for Good Clinical Practice and the Declaration PIK3CA.13 A retrospective analysis14 of CLEOPATRA also of Helsinki, with protocol approval and amendments showed that a higher tumour-infiltrating lymphocyte obtained from an independent ethics committee for each value was significantly associated with improved overall site. Written, informed consent was obtained from each survival. participant. Here, we report the end-of-study analysis with an additional 4 years of follow-up since the last report Randomisation and masking of CLEOPATRA (median follow-up 99 months, Patients were randomly assigned (1:1) to pertuzumab, IQR 92–106; maximum of 120 months). As durable trastuzumab, and docetaxel; or placebo, trastuzumab, responses have been observed in patients given anti- and docetaxel.3,4 Patients were stratified by geographical HER2 therapy for metastatic breast cancer,2 we also region (Asia, Europe, North America, or South America) compared the clinical and biomarker profiles of patients and previous treatment status (neoadjuvant or adjuvant in CLEOPATRA categorised as long-term responders chemotherapy received vs not), centrally via an interactive and non-long-term responders. voice-response system, which was also used to gather screening information. Treatment assignment was Methods balanced within each stratum by applying a complete Study design and participants block randomisation scheme (block size of four). Details of the CLEOPATRA study and its biomarker Identification numbers were allocated sequentially in the analyses have been published previously.3,13,14 Briefly, order in which patients were enrolled. CLEOPATRA was CLEOPATRA was a double-blind, randomised, placebo- a double-blind study. Placebo was matched in appearance controlled, phase 3 trial done at 204 sites in 25 countries. to the study medication. After the significant overall Eligible patients with metastatic, HER2-positive breast survival improve­ment in the pertuzumab group was cancer were aged 18 years or older, had an Eastern confirmed at the second overall survival interim analysis,4 Cooperative Oncology Group (ECOG) performance investigators were informed about study-group assign­ status of 0 or 1, left ventricular ejection fraction (LVEF) of ments and patients without disease progression in the at least 50% at baseline, and were naive to previous placebo group were offered the opportunity to cross over chemotherapy or biological therapy in terms of treat­ their study treatment from placebo to pertuzumab. This ment for metastatic disease, but could have received opportunity was offered in July, 2012, before which chemotherapy with or without trastuzumab in the investigators and patients remained masked to neoadjuvant or adjuvant settings as long as there was a randomised assignment.5 minimum 12-month interval between completion of all therapy and metastatic disease diagnosis. Patients Procedures could have received up to one hormonal treatment for Study drugs were given intravenously every 3 weeks.3 metastatic disease but had no concurrent hormonal Pertuzumab or placebo was given on day 1 of each cycle, therapy before disease progression. starting at a loading dose of 840 mg in the first cycle and Eligible patients had adequate organ function (assessed decreasing to 420 mg maintenance dose in subsequent by absolute neutrophil count <1500 cells per mm³, cycles. Trastuzumab was given on day 2 of the first cycle at platelet count <100 000 cells per mm³, haemoglobin 8 mg/kg loading dose, changing to 6 mg/kg maintenance­ <9 g/dL, total bilirubin >upper limit of normal [ULN; dose for subsequent cycles. Docetaxel was given on day 2 unless the patient had documented Gilbert’s syndrome], of the first cycle at 75 mg/m², increasing to 100 mg/m² if aspartate amino­transferase and alanine aminotransferase tolerated. If well tolerated, all three agents could be given >2·5 × ULN, aspartate aminotransferase or alanine on day 1 of subsequent cycles. If not well tolerated, the aminotransferase >1·5 × ULN with concurrent serum cycle 1 schedule was maintained. Pertuzumab or placebo alkaline phosphatase >2·5 × ULN [unless bone and trastuzumab were given until disease progression or metastases were present], serum creatinine >2·0 mg/dL unmanageable toxic effects, and docetaxel was given for at or 177 μmol/L, international normalised ratio and least 6 cycles (more than 6 cycles were given at the activated partial thromboplastin time >1·5 × ULN [unless investigators’ discretion). If chemotherapy was on therapeutic coagulation]). Exclusion criteria included discontinued due to toxic effects, pertuzumab, placebo, or treatment for metastatic breast cancer (other than that trastuzumab were continued until disease progression, already described), CNS metastases, previous exposure unacceptable toxic effects, or withdrawal of consent. to a cumulative dose of doxorubicin of more than Reduction in dose of pertuzumab, placebo, or trastuzumab 360 mg/m², and a decline in LVEF to less than 50% was not allowed. Docetaxel doses could be reduced by 25% during or after previous trastuzumab treatment. for febrile neutropenia or neutrophils less than www.thelancet.com/oncology Vol 21 April 2020 521 Articles

1196 patients assessed for eligibility

388 did not meet eligibility criteria or declined to participate

808 randomly assigned

406 randomly assigned to receive placebo, 402 randomly assigned to receive trastuzumab, and docetaxel pertuzumab, trastuzumab, and docetaxel

9 received non-randomised 1 received non-randomised treatment treatment

2 did not receive any 2 did not receive any treatment treatment

406 included in the 396 treated and included in the safety 408 treated and included in the safety 402 included in the intention-to-treat population population intention-to-treat population population

261 died 238 died 240 progressive disease 215 progressive disease 12 adverse event during treatment 8 adverse event during treatment phase phase 9 death during post-treatment phase 15 death during post-treatment phase unrelated to disease progression unrelated to disease progression

50 crossed over to the pertuzumab, 14 died trastuzuab, and docetaxel group 12 progressive disease 1 adverse event during treatment phase 1 death during post-treatment phase unrelated to disease progression

Figure 1: Trial profile

500 cells/mm³ for more than 1 week (after fully recovering Outcomes to a neutrophil count ≥1500 cells/mm³), platelet count less The primary endpoint was independent review facility- than 100 000 cells/mm³ (after recovering to a platelet assessed progression-free survival (time from ran­dom­ count ≥100 000 cells/mm³), or severe or cumulative isation to the first documentedradiographic evidence of cutaneous reactions. Pertuzumab or trastuzumab doses disease progression per Response Evaluation Criteria in could be delayed for toxic effects, with reloading doses of Solid Tumors v1.0).3 Secondary endpoints were overall the antibodies given for interruptions of one cycle. survival, investigator-assessed progression-free survival, Chest x-rays were done at screening or baseline, at each objective response rate, duration of objective response, cycle if clinically indicated, at the treatment discon­ safety, time to symptom progression, and association of tinuation visit, and at week 18 post-treatment if clinically biomarkers with clinical outcomes (eg, HER3 expression, indicated. Tumour assessments via MRI or CT were Fcγ, and serum HER2 extracellular domain and HER done at screening or baseline and every 9 weeks from ligands concentrations).3 In this end-of-study analysis, we randomisation until progressive disease confirmed by an report final overall survival, investigator-assessed independent review facility (Bioclinica). Laboratory progression-free survival, safety results, and correlation monitoring was done at screening or baseline, each cycle, of biomarkers with clinical outcomes. and at the treatment discon­tinuation visit. Adverse events were assessed at screening or baseline (serious adverse Statistical analysis events related to study-specific procedures only at this Planned enrolment was 800 patients, and the primary point) and throughout the treatment and follow-up analysis was planned to be performed after approximately periods, per National Cancer Institute’s Common 381 independently assessed progression-free survival Terminology Criteria for Adverse Events version 3.0. events had occurred to give the study 80% power to detect

522 www.thelancet.com/oncology Vol 21 April 2020 Articles

A 100 p<0·0001 90 80 70 60 50 Landmark overall survival at 8 years 37%, 235 events (58%) 40

Overall survival (%) 30 20 Pertuzumab, trastuzumab, and docetaxel Landmark overall survival at 8 years 23%, 280 events (69%) 10 Placebo, trastuzumab, and docetaxel 0 0 10 20 30 40 50 60 70 80 90 100 110 120 Number at risk (number censored) Pertuzumab 402 (0) 371 (14) 318 (23) 269 (32) 228 (41) 188 (48) 165 (50) 150 (54) 137 (56) 120 (59) 71 (102) 20 (147) 0 (167) Placebo 406 (0) 350 (19) 289 (30) 230 (36) 181 (41) 149 (48) 115 (52) 96 (53) 88 (53) 75 (57) 44 (84) 11 (115) 1 (125)

B 100 p<0·0001 90 80 70 60 50 40 survival (%) 30 Landmark progression-free survival at 8 years 16%, 304 events (76%) 20 10 Investigator-assessed progression-free Investigator-assessed Landmark progression-free survival at 8 years 10%, 329 events (81%) 0 0 10 20 30 40 50 60 70 80 90 100 110 120 Time since randomisation (months) Number at risk (number censored) Pertuzumab 402 (0) 284 (18) 179 (24) 121 (34) 93 (40) 71 (47) 60 (49) 52 (54) 43 (60) 34 (66) 21 (78) 6 (92) 0 (98) Placebo 406 (0) 223 (27) 110 (32) 76 (39) 53 (44) 43 (47) 35 (49) 30 (52) 23 (54) 21 (56) 10 (67) 4 (73) 0 (77)

Figure 2: Kaplan-Meier plots of overall survival and investigator-assessed progression-free survival (A) Overall survival in the intention-to-treat population. Crossover patients were analysed in the placebo group. (B) Investigator-assessed progression-free survival in the intention-to-treat population. Investigator-assessed progression-free survival was compared between groups by use of the log-rank test, stratified by previous treatment status and geographical region. a 33% improvement in median progression-free survival PIK3CA-mutated tumours. Subgroup analyses are in the pertuzumab group (HR 0·75) at a two-sided unstrati­fied. An exploratory post-hoc subgroup analysis significance level of 5%.3 Since the confirmatory overall within patients who had received prior trastuzumab was survival analysis had already occurred,4 analyses are defined after database lock. The method was the same as descriptive. Clinical cutoff for this analysis was applied to the predefined subgroup analyses. Investigator- Nov 23, 2018; 4 years and 9 months after the cutoff for the assessed progression-free survival was analysed per previous analysis.5 overall survival (as was the primary analysis3). Overall survival analyses were based on the intention- Safety analyses were based on treatment received; to-treat (ITT) population, with crossover patients crossover patients were counted in the placebo group up analysed in the placebo group (their randomly assigned to the day before the first dose of pertuzumab. Patients treatment group); analyses were not adjusted for randomly assigned to placebo who received at least crossover to the pertuzumab group and are likely to be one dose of pertuzumab in error (excluding crossover conservative. pertuzumab) are included in the pertuzumab group for Overall survival was compared between groups by use of safety analyses. Post-crossover safety data are reported the log-rank test, stratified by previous treatment status separately. and geographical region. The Kaplan-Meier approach was Long-term responders were defined as patients still on used to estimate median overall survival, and a stratified study treatment at 35 months, and with a time to Cox proportional hazards model was used to estimate investigator-assessed progression-free survival of at least the HR and 95% CIs. Similar methods were used to 35 months. Non-long-term responders were defined as assess subgroups of patients with PIK3CA-wild-type and patients who discontinued study treatment before www.thelancet.com/oncology Vol 21 April 2020 523 Articles

Pertuzumab, Placebo, Hazard ratio H-score, HER3 mRNA, and serum HER2 extracellular trastuzumab, trastuzumab, (95% CI) domain were chosen as biomarkers of interest for this and docetaxel and docetaxel post-hoc analysis, because the p value for the prognostic (events/N) (events/N) effect in the previous biomarker analysis was less than Previous neoadjuvant or adjuvant systemic therapy 0·05 (or very close to 0·05 in the case of HER2 membrane No 123/218 149/214 0·65 (0·51–0·82) H-score [p=0·0502]).13 Tumour-infiltrating lymphocyte Yes 112/184 131/192 0·74 (0·57–0·95) analyses were done independently of other biomarker Region analyses. Statistical analyses were done by use of SAS Europe 101/152 0·68 (0·51–0·91) 85/154 software (version 9). North America 33/67 53/68 0·46 (0·30–0·72) This trial is registered with ClinicalTrials.gov, number South America 38/56 44/58 0·65 (0·42–1·00) Asia 79/125 82/128 0·85 (0·63–1·16) NCT00567190. Age group <65 years 200/342 48/67 0·71 (0·59–0·86) Role of the funding source ≥65 years 35/60 48/67 0·55 (0·35–0·85) This study was funded and sponsored by F Hoffmann-La <75 years 230/397 269/392 0·70 (0·58–0·83) Roche and Genentech, and designed by the senior ≥75 years 5/5 11/14 0·72 (0·23–2·31) academic authors and representatives of the sponsor. Ethnic origin The sponsor provided study drugs, and was involved in White 141/245 165/235 0·65 (0·52–0·82) protocol development, regulatory and ethics approvals, Black 5/10 14/20 0·58 (0·21–1·61) and safety monitoring. The data were collected by the Asian 80/128 86/133 0·85 (0·63–1·15) sponsor and analysed and interpreted by the sponsor in Other* 9/19 15/18 0·32 (0·14–0·74) collaboration with the senior academic authors. Disease type Employees of the sponsor (EM, ER, ECl, AK, and MCB) Visceral disease 187/314 234/316 0·60 (0·50–0·73) were involved in the writing of the report and the decision Non-visceral disease 48/88 46/90 1·08 (0·72–1·62) Oestrogen and progesterone receptor status to submit for publication along with the academic Positive 114/189 134/199 0·74 (0·58–0·96) authors. All authors had full access to all study data and Negative 120/212 138/196 0·64 (0·50–0·81) had final responsibility for the decision to submit for Unknown 1/1 8/11 8·94 (0·56–143·60) publication. HER2 status Immunohistochemistry 197/350 252/371 0·68 (0·56–0·82) Results score 3+ Between Feb 12, 2008 and July 7, 2010, 1196 patients were FISH-positive 226/384 260/383 0·72 (0·60–0·86) screened and 808 patients were enrolled. 402 patients All 235/402 280/406 0·69 (0·58–0·82) were randomly assigned to the pertuzumab group and 406 to the placebo group (figure 1). Baseline characteristics 0·001 0·1 1 10 100 of the patients were similar in the two groups.3 Clinical

Favours pertuzumab, trastuzumab, Favours placebo, trastuzumab, cutoff for this analysis was Nov 23, 2018. Between and docetaxel and docetaxel July 2012 and clinical cutoff, 50 patients crossed from the placebo to the pertuzumab group. All patients who Figure 3: Forest plot for overall survival in subgroups *Other includes Native American and Alaska Native populations. Ethnic origin was defined by investigator. were receiving pertuzumab when the study closed were FISH=fluorescence in-situ hybridisation. provided access to pertuzumab in a post-study option: 25 were enrolled in the pertuzumab extension study, 35 months due to disease progression, since these PEREX (NCT02320435); 11 entered a post-trial access patients were considered to be genuine non-long-term programme; and 23 continued to receive pertuzumab as responders (otherwise, this category would have included the standard of care in this indication. Median follow-up patients who stopped treatment early for another reason was 99·9 months (IQR 92·9–106·4) in the pertuzumab while still progression-free). The choice of 35 months as group and 98·7 months (90·9–105·7) in the placebo the cutoff was driven by visual inspection of the group. investigator-assessed progression-free survival Kaplan- In the ITT population, based on randomised treatment, Meier curve, representing the time when the curve starts there were 235 (58%) overall survival events in to plateau in the pertuzumab group. The main aim of 402 patients in the pertuzumab group and 280 (69%) this post-hoc exploratory analysis was to evaluate the overall survival events in 406 patients in the placebo clinical profile of the long-term responders, and to group (figure 2A). Median overall survival was identify molecular or clinico­pathological characteristics 57·1 months (95% CI 50–72) in the pertuzumab group that help identify a subgroup of patients with favourable versus 40·8 months (36–48) in the placebo group long-term response. Long-term responders and non- (HR 0·69, 95% CI 0·58–0·82). Landmark overall survival long-term responders from each group were identified rates were 49% (95% CI 44–54) in the pertuzumab group and their baseline characteristics and key biomarkers and 35% (30–40) in the placebo group at 5 years, assessed. PIK3CA, HER2 mRNA, HER2 membrane 45% (40–50) in the pertuzumab group and 28% (24–33)

524 www.thelancet.com/oncology Vol 21 April 2020 Articles

Pertuzumab, trastuzumab, and Placebo, trastuzumab, and docetaxel Pertuzumab, trastuzumab, and docetaxel group (pre-crossover, safety group (pre-crossover, safety docetaxel group (crossover population) (n=408) population) (n=396) population) (n=50) Grade 1–2 Grade 3 Grade 4 Grade 1–2 Grade 3 Grade 4 Grade 1–2 Grade 3 Grade 4 General disorders and administration site conditions Fatigue 148 (36%) 8 (2%) 1 (<1%) 135 (34%) 13 (3%) 0 5 (10%) 0 0 Asthenia 104 (25%) 11 (3%) 0 115 (29%) 7 (2%) 0 3 (6%) 0 0 Peripheral oedema 100 (25%) 2 (<1%) 0 107 (27%) 3 (1%) 0 1 (2%) 0 0 Mucosal inflammation 105 (26%) 6 (1%) 0 75 (19%) 3 (1%) 1 (<1%) 1 (2%)0 Pyrexia 80 (20%) 4 (1%) 1 (<1%) 70 (18%) 1 (<1%) 1 (<1%) 4 (8%) 0 0 Oedema 46 (11%) 3 (1%) 0 44 (11%) 5 (1%) 0 1 (2%) 0 0 Skin and subcutaneous tissue disorders Alopecia 248 (61%) 0 0 239 (60%) 1 (<1%) 0 4 (8%) 0 0 Rash 153 (38%) 3 (1%) 0 93 (23%) 3 (1%) 0 11 (22%) 0 0 Nail disorder 91 (22%) 5 (1%) 0 91 (23%) 1 (<1%) 0 2 (4%) 0 0 Pruritus 75 (18%) 0 0 40 (10%) 0 0 6 (12%) 0 0 Dry skin 47 (12%) 0 0 25 (6%) 0 0 4 (8%) 0 0 Gastrointestinal disorders Diarrhoea 240 (59%) 39 (10%) 0 171 (43%) 19 (5%) 1 (<1%) 24 (48%) 0 1 (2%) Nausea 179 (44%) 5 (1%) 0 166 (42%) 2 (1%) 0 4 (8%) 0 0 Vomiting 105 (26%) 6 (1%) 0 91 (23%) 6 (2%) 0 5 (10%) 0 0 Constipation 69 (17%) 0 0 97 (24%) 4 (1%) 0 4 (8%) 0 0 Stomatitis 80 (20%) 2 (<1%) 0 62 (16%) 1 (<1%) 0 6 (12%) 0 0 Abdominal pain 64 (16%) 0 0 48 (12%) 3 (1%) 0 2 (4%) 0 0 Dyspepsia 55 (13%) 0 0 48 (12%) 0 0 3 (6%) 0 0 Abdominal pain upper 43 (11%) 1 (<1%) 0 43 (11%) 0 0 2 (4%) 0 0 Nervous system disorders Headache 98 (24%) 7 (2%) 1 (<1%) 72 (18%) 4 (1%) 0 7 (14%) 0 0 Neuropathy peripheral 83 (20%) 11 (3%) 0 71 (18%) 7 (2%) 0 1 (2%) 0 0 Dysgeusia 75 (18%) 0 0 62 (16%) 0 0 1 (2%) 0 0 Dizziness 64 (16%) 3 (1%) 0 53 (13%) 0 0 4 (8%) 0 0 Peripheral sensory neuropathy 50 (12%) 2 (<1%) 0 58 (15%) 1 (<1%) 0 2 (4%) 0 0 Paraesthesia 42 (10%) 1 (<1%) 0 38 (10%) 3 (1%) 0 0 0 0 Musculoskeletal and connective tissue disorders Myalgia 93 (23%) 5 (1%) 0 96 (24%) 3 (1%) 0 5 (10%) 0 0 Arthralgia 82 (20%) 1 (<1%) 0 68 (17%) 3 (1%) 0 5 (10%) 0 0 Pain in extremity 74 (18%) 2 (<1%) 0 51 (13%) 1 (<1%) 0 5 (10%) 0 0 Back pain 62 (15%) 5 (1%) 1 (<1%) 44 (11%) 4 (1%) 0 5 (10%) 1 (2%) 0 Muscle spasms 48 (12%) 2 (<1%) 0 20 (5%) 0 0 2 (4%) 0 0 Infections and infestations Upper respiratory tract infection 87 (21%) 3 (1%) 0 57 (14%) 0 0 13 (26%) 0 0 Nasopharyngitis 76 (19%) 0 0 59 (15%) 1 (<1%) 0 13 (26%) 0 0 Paronychia 32 (8%) 0 0 15 (4%) 1 (<1%) 0 6 (12%) 0 0 Cystitis 16 (4%) 0 0 6 (2%) 0 0 5 (10%) 0 0 Influenza 29 (7%) 1 (<1%) 0 22 (6%) 0 0 5 (10%) 1 (2%) 0 Respiratory, thoracic, and mediastinal disorders Cough 99 (24%) 2 (<1%) 0 78 (20%) 1 (<1%) 0 6 (12%) 0 0 Dyspnoea 64 (16%) 4 (1%) 0 55 (14%) 8 (2%) 0 1 (2%) 0 0 Epistaxis 41 (10%) 0 0 35 (9%) 0 0 2 (4%) 0 0 Metabolism and nutrition disorders Decreased appetite 114 (28%) 7 (2%) 0 100 (25%) 6 (2%) 0 2 (4%) 0 0 (Table 1 continues on next page)

www.thelancet.com/oncology Vol 21 April 2020 525 Articles

Pertuzumab, trastuzumab, and Placebo, trastuzumab, and docetaxel Pertuzumab, trastuzumab, and docetaxel group (pre-crossover, safety group (pre-crossover, safety docetaxel group (crossover population) (n=408) population) (n=396) population) (n=50) Grade 1–2 Grade 3 Grade 4 Grade 1–2 Grade 3 Grade 4 Grade 1–2 Grade 3 Grade 4 (Continued from previous page) Blood and lymphatic system disorders Anaemia 92 (23%) 9 (2%) 1 (<1%) 64 (16%) 12 (3%) 2 (1%) 4 (8%) 2 (4%) 0 Leukopenia 25 (6%) 43 (11%) 7 (2%) 23 (6%) 49 (12%) 10 (3%) 0 0 0 Neutropenia 18 (4%) 64 (16%) 136 (33%) 15 (4%) 59 (15%) 124 (31%) 1 (<1%) 0 0 Eye disorders Lacrimation increased 60 (15%) 0 0 55 (14%) 0 0 0 0 0 Psychiatric disorders Insomnia 66 (16%) 0 0 55 (14%) 0 0 2 (4%) 0 0 Vascular disorders Hypertension 43 (11%) 10 (2%) 0 25 (6%) 7 (2%) 0 3 (6%) 1 (2%) 0

Data are number of patients (%). Adverse events at grades 1–2 occurring in ≥10% of patients in any group and the corresponding grade 3–4 events, or grade 3–4 events occurring in ≥10% of patients in any group and the corresponding grade 1–2 events are shown.

Table 1: Adverse events (safety and crossover populations)

in the placebo group at 6 years, 40% (35–46) in the Pertuzumab, trastuzumab, and Placebo, trastuzumab, and docetaxel group docetaxel group pertuzumab group and 26% (21–31) in the placebo group at 7 years, and 37% (31–42) in the pertuzumab group and Long-term Non-long-term Long-term Non-long-term responders responders responders responders 23% (19–28) in the placebo group at 8 years. (n=99) (n=235) (n=53) (n=286) There were 304 (76%) investigator-assessed pro­gression- Age, years free survival events in 402 patients in the pertuzumab <65 83 (84%) 207 (88%) 47 (89%) 245 (86%) group (296 disease progressions and eight deaths with­ ≥65 16 (16%) 28 (12%) 6 (11%) 41 (14%) out previous disease progression) and 329 (81%) in 406 patients in the placebo group (312 disease progressions Sex and 17 deaths without previous disease progression; Female 99 (100%) 234 (>99%) 53 (100%) 286 (100%) figure 2B). edianM investigator-assessed progression-free Male 0 1 (<1%) 0 0 survival was 12·4 months (95% CI 10–14) in the placebo Ethnic origin group versus 18·7 months (17–22) in the pertuzumab White 56 (57%) 144 (61%) 22 (42%) 175 (61%) group (HR 0·69, 95% CI 0·59–0·81). Black 2 (2%) 4 (2%) 2 (4%) 9 (3%) Subgroup analyses of overall survival were generally Asian 34 (34%) 78 (33%) 28 (53%) 88 (31%) consistent with the results for all patients (figure 3). In Other* 7 (7%) 9 (4%) 1 (2%) 14 (5%) the subgroup of patients who previously received Region trastuzumab, median overall survival was 53·8 months Europe 33 (33%) 98 (42%) 15 (28%) 113 (40%) (95% CI 41–71) in the pertuzumab group and 46·6 months Asia 33 (33%) 76 (32%) 27 (51%) 85 (30%) (30–70) in the placebo group (HR 0·86, 95% CI North America 21 (21%) 32 (14%) 5 (9%) 49 (17%) 0·51–1·43; post-hoc analysis), which was unchanged South America 12 (12%) 29 (12%) 6 (11%) 39 (14%) from the previous analysis.5 Pertuzumab also showed a Disease type consistent overall survival benefit in patients with Measurable disease 85 (86%) 220 (94%) 48 (91%) 262 (92%) PIK3CA wild-type tumours and PIK3CA-mutant Non-measurable disease 14 (14%) 15 (6%) 5 (9%) 24 (8%) tumours (appendix p 10). Visceral or non-visceral lesions 275 patients in the pertuzumab group and 301 in the Visceral disease 72 (73%) 191 (81%) 38 (72%) 233 (81%) placebo group received breast cancer treatment after Non-visceral disease 27 (27%) 44 (19%) 15 (28%) 53 (19%) coming off study treatment (appendix p 11; pertuzumab Bone only 4 (4%) 6 (3%) 3 (6%) 16 (6%) received by 50 crossover patients was deemed study Bone and other 8 (8%) 19 (8%) 4 (8%) 17 (6%) treatment and therefore is not included in the appendix No bone 15 (15%) 19 (8%) 8 (15%) 20 (7%) table. However, post-treatment breast cancer therapy Previous neoadjuvant or adjuvant systemic therapy received by crossover patients who had subsequently No 52 (53%) 128 (54%) 27 (51%) 149 (52%) discontinued study treatment is included). Eight patients Yes 47 (47%) 107 (46%) 26 (49%) 137 (48%) in the pertuzumab group and 17 patients in the placebo (Table 2 continues on next page) group died without progression. For the remaining patients with a progression-free survival event, no details

526 www.thelancet.com/oncology Vol 21 April 2020 Articles

of subsequent breast cancer treatment were reported for Pertuzumab, trastuzumab, and Placebo, trastuzumab, and 21 (pertuzumab group) and 11 (placebo group) patients; it docetaxel group docetaxel group was not mandatory for these data to be collected. Long-term Non-long-term Long-term Non-long-term Median number of treatment cycles pre-crossover responders responders responders responders remained unchanged since the previous analysis (24·0 in (n=99) (n=235) (n=53) (n=286) 5 the pertuzumab group and 15·0 in the placebo group). (Continued from previous page) The maximum number of cycles received increased from Baseline ECOG score 96 to 167 in the pertuzumab group and remained at 67 in 0 70 (71%) 161 (69%) 39 (74%) 171 (60%) the placebo group. Post-crossover, the median number of 1 29 (29%) 73 (31%) 14 (26%) 113 (40%) pertuzumab cycles increased from 22·5 to 42·0, and the Oestrogen and progesterone receptor status maximum number increased from 28 to 105 since the Positive 48 (48%) 110 (47%) 28 (53%) 142 (50%) previous analysis.5 Updated data5 for the median number Negative 51 (52%) 124 (53%) 25 (47%) 135 (47%) of docetaxel cycles (previously eight) were not collected for this analysis. Oestrogen receptor status In the safety population (pre-crossover), based on Positive 44 (44%) 102 (43%) 26 (49%) 137 (48%) treatment received, higher incidences of diarrhoea and Negative 55 (56%) 132 (56%) 27 (51%) 141 (49%) rash were reported in the pertuzumab group than in the Progesterone receptor status placebo group, both at any grade and at grade 3 or higher Positive 35 (35%) 60 (26%) 21 (40%) 95 (33%) (table 1; appendix pp 12,13). Diarrhoea and rash were the Negative 64 (65%) 172 (73%) 32 (60%) 179 (63%) most common post-crossover adverse events, and these HER2 immunohistochemistry status were generally grade 1–2. The most common grade 3–4 0 or 1+ 1 (1%) 3 (1%) 0 2 (1%) adverse event was neutropenia (200 [49%] of 408 patients 2+ 2 (2%) 37 (16%) 1 (2%) 29 (10%) in the pertuzumab group, 183 [46%] of 396 patients in the 3+ 95 (97%) 195 (83%) 52 (98%) 255 (89%) placebo group). Histological tumour grade 39 patients withdrew from HER2-targeted treatment Well differentiated 4 (4%) 6 (3%) 2 (4%) 13 (5%) (and thus stopped all study medication) pre-crossover Moderately differentiated 37 (37%) 70 (30%) 22 (42%) 90 (32%) due to an adverse event in the pertuzumab group, while Poorly differentiated 26 (26%) 87 (37%) 14 (26%) 90 (32%) 24 withdrew in the placebo group. Post-crossover, four Anaplastic 1 (1%) 1 (<1) 1 (2%) 0 patients withdrew. Of these, 31 patients in the Unknown 31 (31%) 71 (30%) 14 (26%) 92 (32%) pertuzumab group, 21 patients in the placebo group, and Previous anthracyclines one crossover patient had events considered possibly Yes 39 (39%) 86 (37%) 22 (42%) 120 (42%) related to study treatment. No 60 (61%) 149 (63%) 31 (58%) 166 (58%) Since the previous analysis, there was only one new Previous taxanes serious adverse event suggestive of congestive heart Yes 25 (25%) 54 (23%) 16 (30%) 67 (23%) failure (pertuzumab group) and one new symptomatic No 74 (75%) 181 (77%) 37 (70%) 219 (77%) left ventricular dysfunction (pertuzumab group post- Previous radiotherapy crossover; appendix p 12). The most common treatment- Yes 39 (39%) 103 (44%) 25 (47%) 124 (43%) related serious adverse events were febrile neutropenia No 60 (61%) 132 (56%) 28 (53%) 162 (57%) in the pertuzumab group (46 [11%]), and febrile neutro­ Previous trastuzumab penia (19 [5%]) and neutropenia (19 [5%]) in the placebo Yes 9 (9%) 30 (13%) 4 (8%) 35 (12%) group (appendix p 14). No 90 (91%) 205 (87%) 49 (92%) 251 (88%) Pre-crossover, 238 (58%) of 408 patients in the safety Mean time from first 33·2 (n=86) 27·6 (n=212) 30·7 (n=46) 29·9 (n=264) population had died in the pertuzumab group, histological diagnosis to SD 36·9 SD 40·2 SD 44·2 SD 41·0 compared with 261 (66%) of 396 in the placebo group, metastatic disease, months mostly due to disease progression (215 [53%] of Data are n (%) unless otherwise stated. *Other includes Native American and Alaska Native populations. Ethnic 408 patients in the pertuzumab group and 240 [61%] of origin was defined by investigator. ECOG=Eastern Cooperative Oncology Group.

396 patients in the placebo group). Adverse events led to Table 2: Baseline characteristics of responders and non-responders deaths during the treatment phase in eight (2%) patients in the pertuzumab group and 12 (3%) patients in the placebo group (two separate adverse events were febrile neutropenia (three [1%] of 408), respiratory tract recorded as having led to death in one patient; not infection, and somnolence in the pertuzumab group shown in table 1 due to specific cutoffs). The most (one [<1%] of 408 each), and intestinal perforation common adverse events leading to death were febrile (two [1%] of 396), pneumonia, sepsis, myocardial neutropenia in the pertuzumab group (three [1%] of infarction, and cerebrovascular accident in the placebo 408 patients) and myocardial infarction in the placebo group (one [<1%] of 396 each; appendix p 16). In the group (three [1%] of 396 patients; appendix p 15). Deaths crossover population, 14 (28%) of 50 patients died: due to treatment-related adverse events were caused by 12 (24%) due to disease progression, one (<1%) due to www.thelancet.com/oncology Vol 21 April 2020 527 Articles

metastatic breast cancer during the treatment period, and subgroup analyses of overall survival were consistent and one (<1%) due to unknown cause in the post- with the overall analysis and previous results.3–5 The treatment period. investigator-assessed progression-free survival treatment 99 patients were classified as long-term responders in effect was also maintained. Results were not likely to be the pertuzumab group (235 non-long-term responders), influenced by breast cancer treatments received after and 53 in the placebo group (286 non-long-term coming off study treatment, since the types of therapies responders; appendix p 17). Baseline characteristics of were generally balanced between study groups. responders and non-responders are shown in table 2. In The safety profiles of the overall safety and crossover the pertuzumab group, more long-term responders than populations were consistent with the known safety non-long-term responders had non-measurable, non- profile of pertuzumab, and long-term cardiac safety was visceral, progesterone receptor-positive, HER2 immuno­ maintained. histochemistry 3+ disease, and had a longer mean time The largest differences in long-term responders versus from first histological diagnosis to metastatic disease non-long-term responders in the pertuzumab group (difference of ≥7%; the minimum we considered to be were for the HER2 immunohistochemistry 3+ group suggestive of a difference between long-term responders (table 2) and the PIK3CA group (appendix p 18). Overall, and non-long-term responders in this post-hoc long-term responders tended to have non-measurable exploratory analysis). In terms of bio­markers, more long- disease, non-visceral lesions, progesterone receptor- term responders in the pertuzumab group had PIK3CA- positive disease, longer time from diagnosis to metastatic wild type tumours and tumours with high HER2 mRNA disease, high HER2 mRNA expression levels, and low and HER2 membrane H-score, and low serum HER2, serum HER2 extracellular domain levels. Consistent versus the non-long-term responders (appendix p 18). with the previous biomarker analysis of progression-free Similar findings were observed in long-term responders survival,13 more long-term responders had PIK3CA-wild- in the placebo group, with the exception of a slightly type tumours than non-long-term responders, and higher proportion of patients with high HER2 mRNA. In median overall survival was longer in patients with the pertuzumab group, median tumour-infiltrating PIK3CA-wild-type tumours than those with PIK3CA- lymphocyte value was higher in long-term responders mutant tumours. Regardless of PIK3CA status, median compared with the non-long-term responders, while in overall survival was improved in the pertuzumab group the placebo group the median values were the same compared with the placebo group. (appendix p 18). Biomarker samples were not available Higher tumour-infiltrating lymphocyte values were for all of the patients in the long-term responder versus observed in patients classified as long-term responders non-long-term responder analysis; hence, these results in the pertuzumab group, supporting the previous should be interpreted with caution. analysis that showed that higher values were associated with improved overall survival with pertuzumab and Discussion trastuzumab treatment.14 This end-of-study analysis of CLEOPATRA showed that A retrospective study24 of patients with de novo, stage IV the overall survival improvement with first-line pertu­ disease given first-line trastuzumab (approximately 20% zumab, trastuzumab, and docetaxel versus placebo, also received pertuzumab) showed that patients who had trastuzumab, and docetaxel for HER2-positive, metastatic no evidence of disease after HER2-targeted therapy had breast cancer was maintained at a median follow-up of high rates of progression-free survival and overall survival 99 months. 37% of patients on continued pertuzumab at 5 years (100% and 98%, respectively) and remained the and trastuzumab-based therapy in CLEOPATRA were same at 10 years.24 It is difficult to compare these studies still alive with more than 8 years of follow-up, and due to differences in study design. 79% of patients in the 16% did not progress, suggesting that these patients retrospective study had oligometastatic disease, and some could continue to maintain clinical benefit over time. might have also received surgery or radiotherapy to those Another randomised phase 3 study (PUFFIN),15 done in sites, resulting in them having no evidence of disease. Chinese patients, showed efficacy and safety of the Survival outcomes in patients with no evidence of disease pertuzumab-based regimen consistent with those seen were not assessed in the post-hoc long-term responder in the global population of CLEOPATRA. analysis of CLEOPATRA, which focused on patients who Prospective phase 2 and 3 studies published since the were still alive and on treatment with no sign of disease previous CLEOPATRA analysis5 have investigated other progression after at least 35 months, irrespective of therapy combinations in the first-line metastatic setting tumour response. Both the retrospective study24 and the but have not shown the overall survival benefit shown in current analysis led to similar conclusions, suggesting a this analysis in a prospective, double-blind, randomised, subset of patients exists who could have good survival controlled manner.16–23 Long-term follow-up reports from outcomes after receiving HER2-targeted therapy for their these studies have not yet been published. Our ITT HER2-positive advanced disease.24 analysis further showed the robustness of the results in Limitations of this analysis include the crossover the presence of crossover from placebo to pertuzumab, from the placebo to the pertuzumab group, the

528 www.thelancet.com/oncology Vol 21 April 2020 Articles

exploratory and descriptive nature of the analyses with provided substantial contributions to the conception and design of the no formal statistical comparisons, and the small size of study. DM and S-AI collected data. S-AI enrolled patients. EM monitored some subgroups. The study was also not powered to safety and collected data. detect predictive effects of biomarkers.13 As most biopsy Declaration of interests All authors disclose third-party writing assistance from F Hoffmann-La samples taken for biomarker analyses were from the Roche. SMS reports grants (to institution) and other support (third-party 13 primary tumour rather than a metastatic sample, writing assistance) from Genentech (F Hoffmann-La Roche), during the results should be interpreted with caution because conduct of the study; grants (to institution) from F Hoffmann-La Roche, these markers can be differentially expressed in primary Genentech, and Pfizer; personal fees (for advisory services/consulting) from Athenex, AstraZeneca, Daiichi-Sankyo, Eli Lilly, F Hoffmann-La and metastatic tissues. Pragmatic efforts should be Roche, Genentech, Genomic Health, Inivata, Pieris Pharmaceuticals, made to identify possible clinical or biological pro­ and Tocagen; non-financial support (ie, travel, accommodations, or food gnostic and predictive factors that could support and beverage) from Athenex, AstraZeneca, Bristol-Myers Squibb, decision-making in clinical practice. Experimental Caris Life Sciences, Daiichi-Sankyo, Eli Lilly, F Hoffmann-La Roche, Genentech, Inivata, NanoString Technologies, Novartis, and Pieris anti-HER2 agents including novel anti-HER2 anti­ Pharmaceuticals; and other support from AstraZeneca (professional 25 26 bodies (margetuximab and trastuzumab deruxtecan), services agreement for independent data monitoring committee service) tyrosine-kinase inhibitors (neratinib,27 tucatinib28 and F Hoffmann-La Roche (third-party writing assistance), outside the and pyrotinib),29 CDK 4/6 inhibitors (palbociclib,30 submitted work. DM reports personal fees from Genentech, and F Hoffmann-La Roche for advisory boards, outside the submitted work. NCT02947685), cancer immunotherapy (trastuzumab S-BK reports grants from Novartis, Sanofi-Genzyme, and Dongkook emtansine or trastu­zumab, pertuzumab, and a taxane Pharma, outside the submitted work. S-AI reports grants and personal plus atezolizumab [NCT02924883, NCT03199885]) fees from AstraZeneca (research grant and advisory role); grants from either as new treatments or in addition to anti-HER2 Pfizer (research grant and advisory role); personal fees from Eisai (advisory role), F Hoffmann-La Roche (advisory role), and Novartis plus taxane therapy, are at different stages of (advisory role,) and Pfizer (advisory role); and travel expenses from development and are expected to provide new and Novartis, outside the submitted work. ECi reports personal fees from effective treatment sequences. New treatment options F Hoffmann-La Roche (speaker and advisory board services, congress could help to improve clinical outcomes after pro­ assistance), Lilly (speaker and advisory board meetings), Novartis (speaker and advisory board meetings), and Pfizer (speaker and advisory gression on the standard pertuzumab-based regimen board services, congress assistance), during the conduct of the study. and trastuzumab emtansine. With this promising AS reports research grants from Celgene, F Hoffmann-La Roche, perspective, the importance of long-term follow-up in AbbVie, and Molecular Partner; personal fees from F Hoffmann-La clinical trials becomes paramount, as does leveraging Roche (expert testimony), AstraZeneca (expert testimony), Celgene (travel expenses), F Hoffmann-La Roche (travel expenses), F Hoffmann- alternative sources of evidence (eg, real-world data) to La Roche (honoraria), Celgene (honoraria), Pfizer (honoraria), further complement findings from randomised phase 3 AstraZeneca (honoraria), Novartis (honoraria), Merck Sharp & Dohme studies, and to validate new methods (eg, molecular (honoraria), Tesaro (honoraria), Lilly (honoraria), and Pfizer (travel imaging, and assessment of circulating­ tumour DNA) expenses); and a medical writing grant from F Hoffmann-La Roche, outside the submitted work. SL reports research funding to institution to further understand tumour biology. and non-remunerated consultant roles for Novartis, Bristol Meyers- In conclusion, overall survival and investigator-assessed Squibb, Genentech, F Hoffmann-La Roche, and Merck; research funding progression-free survival improvements with pertu­ to institution from Puma Biotechnology and Eli Lilly; non-remunerated zumab, trastuzumab, and docetaxel versus placebo, consultant roles for Seattle Genetics and Pfizer; and consulting fees paid to institution from Aduro Biotechnology, outside the submitted work. trastuzumab, and docetaxel observed in previous analyses EM is employed by F Hoffmann-La Roche and discloses stock or were maintained after more than 8 years of median ownership in Genentech, F Hoffmann-La Roche, along with an follow-up in CLEOPATRA. To our knowledge, this is the immediate family member. ECl is employed by Roche Products and discloses stock or ownership in Genentech, F Hoffmann-La Roche; and longest follow-up of patients for first-line treatment of is named on a patent for Roche Products (uses for and article of HER2-positive metastatic breast cancer (maximum of manufacture including HER2 dimerisation inhibitor pertuzumab 120 months). The long-term safety and cardiac safety [PCT/US2012/059683]). AK discloses employment by Roche Products profiles of pertuzumab, trastuzumab, and docetaxel in and stock or ownership in Genentech, F Hoffmann-La Roche. ER discloses employment by F Hoffmann-La Roche and stock or the overall safety population, and within crossover ownership in Genentech/ F Hoffmann-La Roche. MCB discloses patients, were also maintained. Prospective identification employment by Genentech and has a patent pending for Genentech/ of patients who will be long-term responders to treatment F Hoffmann-La Roche. JC discloses stock or ownership in MedSIR; honoraria from F Hoffmann-La Roche, Novartis, Celgene, Eisai, Pfizer, is an area for future research. HER2-targeted therapy has For more on Roche’s criteria for Samsung Bioepis, Lilly, and Merck Sharp & Dohme; consulting or eligible studies see www. changed the natural history of HER2-positive metastatic advisory roles for F Hoffmann-La Roche, Celgene, Cellestia, clinicalstudydatarequest.com breast cancer, with the dual blockade of pertuzumab and AstraZeneca, Biothera Pharmaceutical, Merus, Seattle Genetics, trastuzumab, with docetaxel, demonstrating an 8-year Daiichi-Sankyo, Erytech, Athenex, Polyphor, Lilly, Servier, Merck Sharp & For more on Roche’s Global Policy on the Sharing of Clinical landmark overall survival rate of 37%. Dohme, and GlaxoSmithKline; research funding paid to institution from F Hoffmann-La Roche, Ariad Pharmaceuticals, AstraZeneca, Baxalta/ Information and how to Contributors Servier Affaires, Bayer Healthcare, Eisai, Guardanth Health, Merck request access to related All authors have reviewed the data analyses, contributed to data Sharp & Dohme, Pfizer, Piqur Therapeutics, Puma C, Queen Mary clinical study documents see interpretation, contributed to drafting the work and revising the University of London, and Seagen, outside the submitted work. https://www.roche.com/ publication for important intellectual content, approved the final version research_and_development/ Data sharing to be published, and agree to be accountable for all aspects of the work who_we_are_how_we_work/ in ensuring that questions related to the accuracy or integrity of any part Qualified researchers can request access to individual patient-level data clinical_trials/our_commitment_ of the work are appropriately investigated and resolved. DM and S-AI through the clinical study data request platform. to_data_sharing.htm

www.thelancet.com/oncology Vol 21 April 2020 529 Articles

Acknowledgments 15 Xu B, Li W, Zhang Q, et al. A phase III, randomized, double-blind, Funding for this analysis was provided by F Hoffmann-La Roche and placebo (Pla)-controlled study of pertuzumab (P) + trastuzumab (H) Genentech. We thank all the patients who participated in the trial, and + docetaxel (D) v Pla + H + D in previously untreated HER2-positive their families, the investigators, clinicians, and research staff at the locally recurrent/metastatic breast cancer (LR/MBC) (PUFFIN). 204 centres in 25 countries. We would also like to thank Robert Salgado Proc Am Soc Clin Oncol 2019; 37(suppl 15): 1026 (abstr). for his contribution to this study. Support for third-party writing 16 Gelmon KA, Boyle FM, Kaufman B, et al. Lapatinib or trastuzumab assistance for this manuscript, provided by Daniel Clyde (Health plus taxane therapy for human epidermal growth factor receptor 2-positive advanced breast cancer: final results of NCIC CTG Interactions), was provided by F Hoffmann-La Roche. MA.31. J Clin Oncol 2015; 33: 1574–83. References 17 Drooger JC, van Tinteren H, de Groot S, et al. A randomized 1 Ross JS, Slodkowska EA, Symmans WF, Pusztai L, Ravdin PM, phase 2 study exploring the role of bevacizumab and a Hortobagyi GN. The HER-2 receptor and breast cancer: ten years of chemotherapy-free approach in HER2-positive metastatic breast targeted anti-HER-2 therapy and personalized medicine. Oncologist cancer: The HAT study (BOOG 2008-2003), a Dutch Breast Cancer 2009; 14: 320–68. Research Group trial. Cancer 2016; 122: 2961–70. 2 Slamon DJ, Leyland-Jones B, Shak S, et al. Use of chemotherapy 18 Farhat F, Kattan JG, Ghosn M. Oral vinorelbine in combination plus a monoclonal antibody against HER2 for metastatic breast with trastuzumab as a first-line therapy of metastatic or locally cancer that overexpresses HER2. N Engl J Med 2001; 344: 783–92. advanced HER2-positive breast cancer. Cancer Chemother Pharmacol 3 Baselga J, Cortés J, Kim SB, et al. Pertuzumab plus trastuzumab 2016; 77: 1069–77. plus docetaxel for metastatic breast cancer. N Engl J Med 2012; 19 Perez EA, López-Vega JM, Petit T, et al. Safety and efficacy of 366: 109–19. vinorelbine in combination with pertuzumab and trastuzumab for 4 Swain SM, Kim SB, Cortés J, et al. Pertuzumab, trastuzumab, first-line treatment of patients with HER2-positive locally advanced and docetaxel for HER2-positive metastatic breast cancer or metastatic breast cancer: VELVET Cohort 1 final results. (CLEOPATRA study): overall survival results from a randomised, Breast Cancer Res 2016; 18: 126. double-blind, placebo-controlled, phase 3 study. Lancet Oncol 2013; 20 Smyth LM, Iyengar NM, Chen MF, et al. Weekly paclitaxel with 14: 461–71. trastuzumab and pertuzumab in patients with HER2- 5 Swain SM, Baselga J, Kim SB, et al. Pertuzumab, trastuzumab, overexpressing metastatic breast cancer: overall survival and and docetaxel in HER2-positive metastatic breast cancer. updated progression-free survival results from a phase II study. N Engl J Med 2015; 372: 724–34. Breast Cancer Res Treat 2016; 158: 91–97. 6 Swain SM, Ewer MS, Cortés J, et al. Cardiac tolerability of 21 Andersson M, López-Vega JM, Petit T, et al. Efficacy and safety of pertuzumab plus trastuzumab plus docetaxel in patients with pertuzumab and trastuzumab administered in a single infusion HER2-positive metastatic breast cancer in CLEOPATRA: bag, followed by vinorelbine: VELVET Cohort 2 final results. a randomized, double-blind, placebo-controlled phase III study. Oncologist 2017; 22: 1160–68. Oncologist 2013; 18: 257–64. 22 Sakaguchi K, Nakatsukasa K, Koyama H, et al. Phase II clinical trial 7 Gianni L, Pienkowski T, Im YH, et al. Efficacy and safety of of first-line eribulin plus trastuzumab for advanced or recurrent neoadjuvant pertuzumab and trastuzumab in women with locally HER2-positive breast cancer. Anticancer Res 2018; 38: 4073–81. advanced, inflammatory, or early HER2-positive breast cancer 23 Fujii T, Horiguchi J, Yanagita Y, et al. Phase II study of S-1 plus (NeoSphere): a randomised multicentre, open-label, phase 2 trial. trastuzumab for HER2-positive metastatic breast cancer Lancet Oncol 2012; 13: 25–32. (GBCCSG-01). Anticancer Res 2018; 38: 905–09. 8 von Minckwitz G, Procter M, de Azambuja E, et al. Adjuvant 24 Wong Y, Raghavendra AS, Hatzis C, et al. Long-term survival of de pertuzumab and trastuzumab in early HER2-positive breast cancer. novo stage IV human epidermal growth receptor 2 (HER2) positive N Engl J Med 2017; 377: 122–31. breast cancers treated with HER2-targeted therapy. Oncologist 2019; 9 Schneeweiss A, Chia S, Hickish T, et al. Pertuzumab plus 24: 313–18. trastuzumab in combination with standard neoadjuvant 25 Rugo HS, Im S-A, Shaw Wright GL, et al. SOPHIA primary anthracycline-containing and anthracycline-free chemotherapy analysis: a phase 3 (P3) study of margetuximab (M) + chemotherapy regimens in patients with HER2-positive early breast cancer: (C) versus trastuzumab (T) + C in patients (pts) with HER2+ a randomized phase II cardiac safety study (TRYPHAENA). metastatic (met) breast cancer (MBC) after prior anti-HER2 Ann Oncol 2013; 24: 2278–84. therapies (Tx). Proc Am Soc Clin Oncol 2019; 37 (suppl 15): 1000. 10 Swain SM, Ewer MS, Viale G, et al. Pertuzumab, trastuzumab, 26 Tamura K, Tsurutani J, Takahashi S, et al. Trastuzumab deruxtecan and standard anthracycline- and taxane-based chemotherapy for (DS-8201a) in patients with advanced HER2-positive breast cancer the neoadjuvant treatment of patients with HER2-positive localized previously treated with trastuzumab emtansine: a dose-expansion, breast cancer (BERENICE): a phase II, open-label, multicenter, phase 1 study. Lancet Oncol 2019; 20: 816–26. multinational cardiac safety study. Ann Oncol 2018; 29: 646–53. 27 Deeks ED. Neratinib: first global approval. Drugs 2017; 11 Verma S, Miles D, Gianni L, et al. Trastuzumab emtansine for 77: 1695–1704. HER2-positive advanced breast cancer. N Engl J Med 2012; 28 Borges VF, Ferrario C, Aucoin N, et al. Tucatinib combined with 367: 1783–91. ado-trastuzumab emtansine in advanced ERBB2/HER2-positive 12 von Minckwitz G, Huang CS, Mano MS, et al. Trastuzumab metastatic breast cancer: a phase 1b clinical trial. JAMA Oncol 2018; emtansine for residual invasive HER2-positive breast cancer. 4: 1214–20. N Engl J Med 2019; 380: 617–28. 29 Ma F, Ouyang Q, Li W, et al. Pyrotinib or lapatinib combined with 13 Baselga, Cortés J, Im SA, et al. Biomarker analyses in CLEOPATRA: capecitabine in HER2–positive metastatic breast cancer with prior a phase III, placebo-controlled study of pertuzumab in human taxanes, anthracyclines, and/or trastuzumab: a randomized, epidermal growth factor receptor 2-positive, first-line metastatic phase II study. J Clin Oncol 2019; 37: 2610–19. breast cancer. J Clin Oncol 2014; 32: 3753–61. 30 Gianni L, Bisagni G, Colleoni M, et al. Neoadjuvant treatment with 14 Luen SJ, Salgado R, Fox S, et al. Tumour-infiltrating lymphocytes in trastuzumab and pertuzumab plus palbociclib and fulvestrant in advanced HER2-positive breast cancer treated with pertuzumab or HER2-positive, ER-positive breast cancer (NA-PHER2): placebo in addition to trastuzumab and docetaxel: a retrospective an exploratory, open-label, phase 2 study. Lancet Oncol 2018; analysis of the CLEOPATRA study. Lancet Oncol 2017; 18: 52–62. 19: 249–56.

530 www.thelancet.com/oncology Vol 21 April 2020 Annals of Oncology Editorials 22. Jamal-Hanjani M, Wilson GA, McGranahan N et al. Tracking the evolution 26. https://www.patientpower.info/medicare-to-cover-genetic-sequencing- of non-small-cell lung cancer. N Engl J Med 2017; 376(22): 2109–2121. in-cancer-patients (13 March 2019, date last accessed). 23. Ciccarelli FD. Mutations differ in normal and cancer cells. Nature 2019; 27. McGranahan N, Furness AJ, Rosenthal R et al. Clonal neoantigens elicit 565 (7739): 301–303. T cell immunoreactivity and sensitivity to immune checkpoint blockade. 24. Marquart J, Chen EY, Prasad V. Estimation of the percentage of US Science 2016; 351(6280): 1463–1469. patients with cancer who benefit from genome-driven oncology. JAMA Oncol 2018; 4(8): 1093–1098. doi:10.1093/annonc/mdz088 25. Joyner MJ, P N. Promises, promises and precision medicine. J Clin Published online 13 March 2019 Invest 2019; 129(3): 946–948.

Personalised reimbursement: a risk-sharing Human Use (CHMP) at the EMA was considered insufficient. model for biomarker-driven treatment of Major concerns were the non-comparative design of the pivotal study, the limited number of patients (n ¼ 74), the absence of rare subgroups of cancer patients overall survival data, and high discordance between local and cen- Precision medicine in oncology is based on the premise that every tral MSI testing. In addition, CHMP had concerns regarding the tumour is unique and therefore requires a thorough molecular placing of nivolumab in second line (after prior fluoropyrimidine- analysis to identify the best possible targeted treatment. In gen- based therapy), in the absence of convincing evidence and with sev- eral, access to precision medicine, especially outside an approved eral established treatment options are available [3]. indication is challenging. There are several barriers and concerns. In the Netherlands, a non-randomised, multi-centre basket Although the paradigm of precision medicine in cancer is to tar- trial, The Drug Rediscovery Protocol (DRUP) [4], is active to get a specific genetic aberration, there is uncertainty regarding ef- specifically identify signals of clinical benefit of approved drugs fectiveness for every biomarker–tumour–drug combination. used outside their label in rare, molecularly defined subsets of Various other factors, such as post-transcriptional modifications, patients who have exhausted standard-of-care treatment options. protein expression, tissue context, heterogeneity of the tumour The trial also contains an MSI cohort in which patients with MSI- and its microenvironment, variations in patient characteristics, H tumours are treated with nivolumab, irrespective of their tu- and prior treatments also contribute to uncertainty of treatment mour type (with the exception of approved indications). The outcome. results of this cohort of 30 patients are in line with the retrospec- Collecting data and generating evidence on off-label use are tive data used for the FDA accelerated approval of pembrolizu- complex outside a clinical trial. Randomised clinical trials are dif- mab, further underlining the efficacy of checkpoint inhibitors in ficult to conduct as small numbers of patients carry a particular patients with these tumours. genetic aberration in a specific tumour type. Clinical evidence is Currently, as the MSI-H cohort of the DRUP trial has reached therefore mostly based on case-studies or small single-armed tri- target recruitment and is therefore closed, newly diagnosed als. Historical data are often not available to compare treatment patients in the Netherlands have no access to treatment. There is outcome to conventional treatment, as earlier studies have not al- also no coverage by health insurers for this biomarker-driven in- ways taken the genetic make-up of the tumour into account. dication with promising data. This poses a serious dilemma and However, regulatory agencies have developed tools (e.g. condi- it is likely that other anticancer agents with high antitumour ac- tional market authorisation or accelerated approvals) to address tivity in non-randomised studies will encounter similar barriers. this problem and facilitate timely access to the patient. As a consequence, there is a growing need for a learning health An illustrative example of the latter is the accelerated approval care model which enables early access to potentially effective ther- of the checkpoint inhibitor pembrolizumab for adult and paedi- apies, where no other established treatment options are available, atric patients with unresectable or metastatic, microsatellite without overestimating the findings that are based on small instability-high (MSI-H) or mismatch repair deficient (dMMR) cohorts of patients. Continuous monitoring to enrich a real- advanced solid tumours that have progressed following prior world database is essential for this learning model. treatment and who have no satisfactory alternative treatment op- In the Netherlands, the government determines the content of tion. The U.S. Food and Drug Administration (FDA) approval the standard health insurance package that covers necessary was tissue/site-agnostic and based on retrospective analysis of healthcare costs. Subsequently, this package is offered by all data from 149 patients, including 90 patients with metastatic co- insurers. The government is advised by the National Health Care lorectal cancer (mCRC), across 5 single-arm clinical trials [1]. Institute (Zorginstituut Nederland), an independent health tech- MSI-H tumours are rare and their prevalence varies widely nology assessment (HTA) authority that evaluates interventions, among tumour types. Bonneville et al. [2] detected MSI-H with a to ensure that the standard health insurance package is cost- prevalence >1% in 12 out of the 39 different types of cancers, effective, evidence-based, and in accordance with state of the art which severely hampers the execution of adequately powered and state of the practice. In some cases, the health insurers can de- randomised trials. cide to reimburse drugs which are not included in the package, In June 2017, the application of another checkpoint inhibitor, for instance, when a disease is extremely rare (prevalence <1/ nivolumab, for MSI-H or dMMR mCRC patients with the 150 000) and no other treatment option is available [5]. European Medicines Agency (EMA) was withdrawn as the evi- The Dutch government regularly negotiates price/volume dence presented to the EU Committee for Medicinal Products for agreements for drugs with a high budget impact. The immune

Volume 30 | Issue 5 | 2019 doi:10.1093/annonc/mdz119 | 663 Editorials Annals of Oncology

rd rd 1st stage 2nd stage 3 stage: first 16 weeks drug(s) provided 3 stage: after 16 weeks by pharma

Complete response at second response evaluation (16 weeks)

Partial response at second Reimbursed care response evaluation untill disease ≥5x patients DRUP (16 weeks) progression with clinical expansion cohort ≥ 1x patient benefit Stable disease at second clinical + 16 patients response evaluation benefit (16 weeks) 8 patients <5 patients with clinical Close cohort No clinical benefit No reimbursement for Close cohort Progressive disease benefit treatment

Figure 1. A performance-based, personalised reimbursement scheme after 16 weeks of clinical benefit at stage III, when the effectiveness is proven for an individual patient, commercial medication will be reimbursed by payers.

checkpoint inhibitors have also been subjected to confidential na- RECIST 1.1 (or iRECIST for in case of ICI) at 16 weeks or pro- tional price agreements. However, off-label use of expensive longed stable disease (at least 16 weeks but duration can vary drugs is usually not covered by payers. As the health care budget depending on tumour–drug combinations). is limited and there is a continuous rise in expenditure, the payers Although this model provides access to potentially effective are obliged to allocate their budgets reasonably and responsibly. drugs for patients without other treatment options and allows This presents a dilemma, especially when clear clinical benefit is risk-sharing between the manufacturer and payers, there are seen in a small patient population such as MSI-H patients. In the some considerations and limitations: case of MSI-H, the Health Insurers in the Netherlands and the National Health Care Institute acknowledge the medical need in 1. The manufacturers are needed to partner in this approach by patients who have exhausted other treatment options. This providing investigational medicinal product for free until a prompted us to collaborate in developing a personalised meaningful clinical response is achieved at 16 weeks. decision-making model to enable early access to potentially effec- 2. Payers and HTA authorities need to approve the model, pref- tive therapies whilst being aware of the increased pressure on the erably by embracing general rules of the proposed scheme. In health care budget. fact, the presented scheme is a result of close collaboration Here, we present a performance-based, personalised reim- among medical oncologists, National Health Care Institute, bursement scheme that enables access to precision medicine in and health insurers in the Netherlands, all of whom support rare biomarker-defined subgroups. In the Netherlands, this this model. scheme will be an integral part of the ongoing DRUP trial. In this 3. A molecular tumour board, which consists of a multidiscipli- trial, eligible patients for a particular tumour–drug combination nary team of experts, should evaluate molecular and clinical are recruited based on Simon’s two-stage design approach. Eight data and provide recommendations on inclusion in the DRUP patients are enrolled in stage I and 16 more in stage II, if more study. than 1 response is observed in the first stage. If less than five 4. The patient should be notified of the experimental nature of patients show an objective tumour response or stable disease at the treatment and provide consent, and also to allow further 16 weeks, the cohort is closed (Figure 1). However, when the sec- (translational) research. ond stage is successful with five or more patients benefitting from 5. As the magnitude of benefit on overall survival and quality of the therapy, the cohort will be expanded to a third stage, with de- life is unclear, it is important to periodically analyse the fined inclusion criteria, duration of treatment and number of results. The structure of a clinical trial with predefined num- patients needed to confirm the initial results. The first two stages ber of patients, pre-planned interim analysis, and futility of the DRUP trial are exploratory, with medication considered to assessments can save resources. be investigational medicinal products provided for free by the 6. It is important to gather biomarker data that can be used to marketing authorisation holder (MAH). The third stage is further refine patient selection in the future and hence im- designed to confirm the findings in the first and second stages prove quality of care. and can be partly reimbursed based on a pay for performance 7. Nationwide, patients need to have equal access to the treat- model. In this model, patients start on treatment with the investi- ment and treatment evaluations need to be harmonised. gational medicinal product as provided by the MAH and con- 8. The necessity of the continuation of the performance-based tinue on the regular drug product which is reimbursed in case of reimbursement scheme should regularly be evaluated based adequate individual treatment response. Adequate response is on predefined outcome criteria and availability of better defined as complete remission or partial remission based on treatment options.

664 | van Waalwijk van Doorn-Khosrovani et al. Volume 30 | Issue 5 | 2019 Annals of Oncology Editorials 9. Organisation of such personalised reimbursement schemes is Institute, Amsterdam; 13Department of Clinical Pharmacy, University complex in terms of infrastructure and administrative burden. Medical Centre Utrecht, Utrecht University, Utrecht, The Netherlands (*E-mail: [email protected]) The performance-based reimbursement scheme that we pro- pose here will run as a pilot, using the infrastructure of the DRUP trial. By integrating into the current infrastructure of the DRUP trial, through expanding the trial to a third stage, we will guaran- Funding tee careful data-management and uniform genomic and MSI The DRUP trial is supported by the Barcode for Life Foundation testing and evaluation. This stepwise approach can be used in the (grant number not applicable); the Dutch Cancer Society (grant future for other rare molecular subgroups. number 10014); and all participating pharmaceutical companies To our knowledge, this is the first time that a risk-sharing model (grant numbers not applicable): Amgen, AstraZeneca, Bayer, has been set up between pharmaceutical industry and payers for Boehringer Ingelheim, Bristol-Myers Squibb, Eisai, Novartis, biomarker-driven, tissue-independent, cancer treatment. The Merck Sharp & Dohme, Pfizer and Roche. learning health care scheme proposed here allows patients with various tumour types to have early access to potentially effective off-label drugs based on their specific molecular profile, while at Disclosure the same time real-world evidence for precision medicine is gener- ated. In the pilot, to be run in the Netherlands, the performance- EEV, HMWV and HG have through the DRUP and other stud- based reimbursement step will run alongside the national financial ies support from pharmaceutical companies participating in the agreements with manufacturers to ensure responsible use of DRUP. All remaining authors have declared no conflicts of health care resources. This model can be a step forward in deliver- interest. ing precision medicine in a sustainable and affordable manner. References S. B. van Waalwijk van Doorn-Khosrovani1, A. Pisters-van Roy1, L. van Saase2, M. van der Graaff2, J. Gijzen1, S. Sleijfer3,4, L. R. Hoes5, 1. The U.S. Food and Drug Administration (FDA) grants accelerated ap- J. M. van Berge Henegouwen6, H. van der Wijngaart7, proval to pembrolizumab for first tissue/site agnostic indication 2017; D. L. van der Velden5, E. van Werkhoven8, V. P. Retel9,10, https://www.fda.gov/drugs/informationondrugs/approveddrugs/ucm5600 W. H. van Harten9,10,11, A. D. R. Huitema12,13, L. Timmers2, 40.htm (5 October 2018, date last accessed). 2. Bonneville R, Krook MA, Kautto EA et al. Landscape of microsatellite in- H. Gelderblom6, H. M. W. Verheul6 & E. E. Voest4,5* 1 2 stability across 39 cancer types. JCO Precis Oncol 2017; (1): 1–15. CZ Health Insurance, Tilburg; National Health Care Institute 3. European Medicines Agency. Withdrawal Assessment Report EMA/ 3 (Zorginstituut Nederland), Diemen; Department of Medical Oncology, 772719/2017; http://www.ema.europa.eu/docs/en_GB/document_library/ Erasmus MC Cancer Institute, Erasmus University Medical Center, Application_withdrawal_assessment_report/human/003985/WC5002399 4 Rotterdam; Center for Personalised Cancer Treatment (CPCT); 43.pdf (5 October 2018, date last accessed). 5 Division of Molecular Oncology, The Netherlands Cancer Institute, 4. The Drug Rediscovery Protocol. ClinicalTrials.gov identifier NCT02 Amsterdam; 6Division of Medical Oncology, Leiden University Medical 9252234. Center, Leiden; 7Division of Medical Oncology, Amsterdam University 5. The Government of the Netherlands. Which medication do I get reim- Medical Center, Cancer Center Amsterdam, Amsterdam; 8Department bursed? https://www.rijksoverheid.nl/onderwerpen/geneesmiddelen/vraag- of Biometrics, Netherlands Cancer Institute, Amsterdam; 9Division of en-antwoord/welke-medicijnen-krijg-ik-vergoed (19 April 2019, date last Psychosocial Research and Epidemiology, The Netherlands Cancer accessed). Institute, Amsterdam; 10Department of Health Technology and Services Research, University of Twente, Enschede; 11Rijnstate Hospital, Arnhem; doi:10.1093/annonc/mdz119 12Department of Pharmacy & Pharmacology, Netherlands Cancer Published online 26 April 2019

Enthuse for PERUSE: when clinical the pathway has been demonstrated across further lines of anti- judgment overcomes regulatory boundaries HER2-based therapies speeding up the development of new HER2-targeted agents including lapatinib, pertuzumab and T- Contemporary treatment of human epidermal growth factor DM1 [6–8]. The appealing synergy of dual anti-HER2 targeting 2 (HER2)-positive breast cancers is a successful example of the ra- fully revealed its clinical value in the Clinical Evaluation of tionale development of molecularly targeted therapies. The ob- Pertuzumab and Trastuzumab study (CLEOPATRA). In this servation that HER2 overexpression or gene amplification was trial, the addition of pertuzumab to trastuzumab and docetaxel associated with more aggressive phenotype and poorer prognosis as first-line therapy resulted, among other things, in a 15.7- has laid the groundwork for developing agents to antagonize this month overall survival improvement [9–11]. The results of pathway [1]. the CLEOPATRA trial led to regulatory approval of pertuzumab– The humanized monoclonal anti-HER2 antibody trastuzumab trastuzumab and docetaxel as first-line therapy for HER2- combined with cytotoxic agents was rapidly established as the positive advanced breast cancer patients. standard therapy of early and advanced HER2-positive breast Despite EMA and FDA labels including docetaxel as the che- cancer in light of its unquestionable efficacy [2–5]. But this motherapy backbone, in routine oncology practice, paclitaxel is was just the beginning. Indeed, metastatic breast cancer is still often preferred in the metastatic setting because of its more favor- incurable. The importance of maintaining the inhibition of able safety profile.

Volume 30 | Issue 5 | 2019 doi:10.1093/annonc/mdz091 | 665 REVIEW

Druggable targets meet oncogenic drivers: opportunities and limitations of target-based classification of tumors and the role of Molecular Tumor Boards

R. Danesi1, S. Fogli1y, S. Indraccolo2, M. Del Re1, A. P. Dei Tos3, L. Leoncini4, L. Antonuzzo5, L. Bonanno6, V. Guarneri6,7, A. Pierini8, G. Amunni9* & P. Conte6,7

1Clinical Pharmacology and Pharmacogenetics Unit, Department of Clinical and Experimental Medicine, University of Pisa, Pisa; 2Molecular Oncology Unit, Veneto Institute of Oncology IOV - IRCCS, Padua; 3Department of Medicine, School of Medicine, University of Padua, Padua; 4Department of Medical Biotechnology, Anatomic Pathology Division, University of Siena, Siena; 5Medical Oncology Unit, Careggi University Hospital, Florence; 6Medical Oncology 2, Veneto Institute of Oncology IOV - IRCCS, Padua; 7Department of Surgery, Oncology and Gastroenterology, University of Padua, Padua; 8Integrated Access, Roche, Monza; 9Institute for the Study, Prevention and Oncology Network (ISPRO), Florence, Italy

Available online xxx

The therapeutic landscape of cancer is changing rapidly due to the growing number of approved drugs capable of targeting specific genetic alterations. This aspect, together with the development of noninvasive methods for the assessment of somatic mutations in the peripheral blood of patients, generated a growing interest toward a new tumor-agnostic classification system based on ‘predictive’ biomarkers. The current review article discusses this emerging alternative approach to the classification of cancer and its implications for the selection of treatments. It is suggested that different types of cancers sharing the same molecular profiles could benefit from the same targeted drugs. Although recent clinical trials have demonstrated that this approach cannot be generalized, there are also specific examples that demonstrate the clinical utility of this alternative vision. In this rapidly evolving scenario, a multidisciplinary approach managed by institutional Molecular Tumor Boards is fundamental to interpret the biological and clinical relevance of genetic alterations and the complexity of their relationship with treatment response. Key words: cancer, biomarker, target therapy, oncogenic drivers, Molecular Tumor Board

INTRODUCTION approved to treat neurotrophic tropomyosin-receptor The identification of therapeutic vulnerabilities based on kinase (NTRK) fusion-positive cancers regardless of their deregulated signal transduction pathways has allowed the location and histology. ‘ ’ development of highly effective targeted drugs. There is Druggable are proteins involved in cell survival/ much hope that predictive biomarkers will be used proliferation, which may or may not be mutated but in any fi extensively and change traditional tumor classification case must be deregulated, that can be targeted by speci c criteria in a new system based on specific molecular drugs [i.e. epidermal growth factor receptor (EGFR) in lung aberrations (Figure 1). Such an approach may change the cancer, vascular endothelial growth factor (VEGF)/VEGF ‘ ’ way by which oncologists will select systemic treatments.1 receptor (VEGFR) in renal cancer]. Actionable is a more The first example of this conceptual revolution regards generic label given to those genetic aberrations involved in pembrolizumab, an anti-programmed cell death protein 1 tumor cell growth which have an impact on clinical (PD-1) approved for metastatic or unresectable tumors with management of patients, a growing number of them being high microsatellite instability (MSI-H) or DNA mismatch druggable and potentially responsive to targeted therapy. repair deficiency. Furthermore, entrectinib has been Although treatment decisions are increasingly being guided by actionable mutations, several limitations in determining actionability need to be considered. For *Correspondence to: Prof. Gianni Amunni, Institute for the Study, Prevention example, the genomic make-up of a tumor may vary due to and Oncology Network (ISPRO), Via Cosimo Il Vecchio 2, 50139 Florence, Italy tumor heterogeneity, the types of biopsy (i.e. primary or E-mail: [email protected] (G. Amunni). metastatic tumor sites, solid or liquid biopsies), and the y Co-first author. time of biopsy sampling (before or after treatments). From 2059-7029/© 2020 Published by Elsevier Limited on behalf of European Society for Medical Oncology This is an open access article under the CC BY- the clinical pharmacology viewpoint, it may be conceivable NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Volume 6 - Issue 2 - 2021 https://doi.org/10.1016/j.esmoop.2020.100040 1 ESMO Open R. Danesi et al.

TARGET VALIDATION PROCESS

TARGET CELL-BASED ANIMAL PHARMACOLOGICAL TRANSLATIONAL IDENTIFICATION DISEASE MODELS MODELS STUDIES RESEARCH

Figure 1. The process of target validation including biomarker development. to classify genetic variants as druggable and undruggable in these patients. All EGFR mutations can be dynamically depending on the availability of specific drugs.2 Cancer monitored using liquid biopsy (Table 1). pharmacology is moving toward a tumor-agnostic drug In conclusion, activating and resistance mutations are classification based on biomarkers that are both predictive druggable and predictive of response to targeted drugs in and druggable, with important implications for the selection NSCLC. No evidence has been reported in other cancer of treatments. Basket and umbrella clinical trials are now types (Table 1). exploring alternative ways to accelerate drug development. In this review article, we describe the biologic role of HER2/ErbB2 drug targets and their potential clinical relevance in the context of a personalized medicine approach, which is the HER2/ErbB2 belongs to the ErbB family with complex working topic of Molecular Tumor Board (MTB). involvement in cancer biology; this type of receptor, which does not have a specific ligand, transmits downstream ONCOGENES AND SIGNAL TRANSDUCTION PATHWAYS signals by heterodimerization with other receptors in the family.17 HER2/ErbB2 is altered in 5.46% of all cancers with EGFR/ErbB1 higher prevalence in breast cancer (BrCa), NSCLC, CRC, EGFR belongs to the c-erb superfamily of receptor tyrosine bladder, gastric, and gastroesophageal junction tumors. The kinases (RTKs) that play a key role in cancer development most common alterations are amplification (3.82%) and and progression. Once bound to its ligands, EGFR can mutation (3.69%), including S310F (0.32%), Y772_A775dup activate multiple signaling pathways, including RASeRAFe (0.28%), and L755S (0.23%).5-8 MEKeERK (extracellular signal-regulated kinase), moAbs (pertuzumab, trastuzumab), antibodyedrug con- phosphatidylinositol-3 kinase (PI3K), and protein kinase B jugates (trastuzumab-emtansine, trastuzumab deruxtecan), (Akt).3 The prevalence of EGFR mutations in non-small-cell and TKIs (lapatinib, neratinib and tucatinib) are HER2/ lung cancer (NSCLC) is 32.3% (from 38.4% in Asian to ErbB2-targeting agents and are standard of care in HER2þ 14.1% in Caucasian populations).4 EGFR is altered in 6.82% cancers.18-20 (Table 1). HER2 aberrations may develop in of all cancers and the most common EGFR alterations are NSCLC after first- or second-generation EGFR TKIs14,21 and somatic mutation (7.26%) and amplification (2.83%).5-8 also as a primary event.21 Although limited by the small Point mutations include L858R; the subclonal resistance sample size, clinical trials demonstrated that NSCLC22,23 and mutations T790M and C797S are selected by previous CRC24 are also sensitive to HER2 targeting by moAbs, treatments with EGFR tyrosine-kinase inhibitors (TKIs).9 antibodyedrug conjugates, or TKIs. First-, second-, and third-generation drugs are gefitinib HER2 can be therefore classified as druggable, due to and erlotinib; afatinib; and rociletinib and osimertinib, strong clinical evidence showing a relationship between respectively, the latter displaying high activity against HER2 positivity and clinical outcomes in BrCa, gastric, and T790M. The monoclonal antibodies (moAbs) cetuximab and gastroesophageal junction cancer patients (Table 1). panitumumab target ErbB1 and are used in colorectal cancer (CRC).10 c-Met NSCLCs harboring exon 19 deletion (ex19del) have higher overall response rate (ORR) and overall survival (OS) than The mesenchymaleepithelial transition factor (c-Met) is a those with exon 21 mutation (ex21mut).11 Afatinib RTK stimulated by the hepatocyte growth factor (HGF). In improves clinical outcomes in treatment-naïve patients with cancer cells, aberrant activation of the HGFec-Met axis due L858R or ex19del, compared with gefitinib, with a to c-Met gene mutations, overexpression, or amplification manageable tolerability profile.12 Osimertinib has higher can activate several signaling pathways, including PI3K/AKT efficacy than first-generation EGFR TKIs with a similar safety and RAS/ERK. c-Met alterations occur in 2.71% of cancers profile.13 Approximately half of patients treated with first- including NSCLC, melanoma, CRC, malignant glioma, and or second-generation drugs may acquire T790M-dependent BrCa. Mutations (2.53%) include amplification (0.71%), drug resistance14; osimertinib15 and rociletinib16 are active X1010_splice (0.13%), and X963_splice (0.08%).5-8

2 https://doi.org/10.1016/j.esmoop.2020.100040 Volume 6 - Issue 2 - 2021 R. Danesi et al. ESMO Open

Table 1. Biomarkers and available drugs

Name of marker Druggable/actionable Tumor type Predictive value, LoE (e.g. available FDA-approved alterations drugs) liquid biopsy CDx test EGFR/ErbB1 Mutations (e.g. L858R, ex19del, NSCLC 1 (gefitinib, erlotinib, afatinib, Yes T790M) osimertinib, dacomitinib) HER2/ErbB2 Amplification Breast 1 (trastuzumab, T-DM1, trastuzumab þ No pertuzumab, lapatinib, neratinib) Amplification Esophagogastric 1 (trastuzumab) No Point mutations (V659E) NSCLC 3A (lapatinib) No c-Met ex14 skipping mutations, amplification NSCLC 1 (crizotinib, capmatinib, savolitinib*, No tepotinib) RET Fusion NSCLC 1 (selpercatinib, pralsetinib), 2A No (cabozantinib), 3A (vandetanib) ALK Fusion NSCLC 1 (crizotinib, alectinib, ceritinib, Yes (alectinib) lorlatinib), 3A (brigatinib) Mutations (L1196M, L1196Q) Soft tissue sarcoma 2A (crizotinib, ceritinib) No ROS1 Fusion, mutation NSCLC 1 (crizotinib, entrectinib) No NTRK Fusion All tumors 1 (larotrectinib, entrectinib) No c-Kit Mutations (e.g. 449_514mut), GIST 1 (imatinib, sunitinib, regorafenib), 2A No deletions (e.g. D419del) (sorafenib) Thymic tumors 2A (sunitinib) No Mutations (e.g. K642E) Melanoma 2A (imatinib) No PDGFR Mutations (e.g. D842V), GIST 2A (imatinib, dasatinib) No deletions (e.g. C456_N468del) Leukemia, myelodysplasia 1 (imatinib) No FGFR1 Amplification LSCC 3A (erdafitinib) No NSCLC 3A (AZD4547) No FGFR2 Fusion, mutation Bladder, cholangiocarcinoma 1 (erdafitinib, pemigatinib) No Amplification Breast 3A (dovitinib) No FGFR3 Fusion, mutation Bladder 1 (erdafitinib) No RAS Wild-type CRC 1 (cetuximab, panitumumab) No BRAF Mutations (e.g. V600E) Melanoma 1 (vemurafenib, dabrafenib, trametinib, No combo), 3A (trametinib) NSCLC 1 (dabrafenib þ trametinib) No Histiocytosis 3A (cobimetinib) No Mutation (V600E) CRC 1 (encorafenib þ cetuximab) Yes Fusions Ovarian 3A (trametinib, cobimetinib) No MEK Mutations Melanoma, NSCLC, ovarian, 3A (trametinib, cobimetinib, No histiocytic disorder selumetinib) mTOR Mutations (e.g. E2014K) Bladder, RCC 3A (everolimus, temsirolimus) No AKT Mutation (E17K) Breast, ovarian 3A (capivasertib) No PTEN Homozygous deletions, Breast 2A (capivasertib) No loss-of-function mutations PIK3CA Mutations Breast 1 (alpelisib) Yes CDK4 Amplification Soft tissue sarcoma 2A (palbociclib) No IDH1 Mutations AML, cholangiocarcinoma 1-3A (ivosidenib) No IDH2 Mutations AML 1 (enasidenib) No BRCA1/2 and ATM Mutations (somatic) Breast 1 (olaparib, talazoparib, rucaparib) No Mutations (somatic) Ovarian, prostate 1 (rucaparib, olaparib) Yes ERa Mutations (e.g. E380Q) Breast 2A (fulvestrant) No MSI-H Not applicable All 1 (pembrolizumab) Yes TML Not applicable Multiple tumor types 1 (pembrolizumab, nivolumab) No AML, acute myeloid leukemia; CRC, colorectal cancer; GIST, gastrointestinal stromal tumors; IHC, immunohistochemistry; MSI-H, microsatellite instability-high; NSCLC, non-small- cell lung cancer; LSCC, lung squamous cell carcinoma; RCC, renal cell carcinoma; CDx, Companion Diagnostics; *, MET and EGFR. LoE, level of evidence was based on the AACR Project GENIE database8 and the ESMO Scale for Clinical Actionability of molecular Targets (ESCAT).138 Classification criteria include tumors having alterations for which FDA-approved drug or standard care is available (1 or 2A), that are Level 1 or 2A in other tumor types (2B), and for which clinical evidence for response to investigational therapies in the same disease does exist (3A).

HGFec-Met can be targeted by TKIs including (i) selective c-Met alterations in cancer are associated with shorter type I inhibitors (e.g. crizotinib, savolitinib, tepotinib) that survival and response to targeted drugs; in renal cell competitively bind to the active c-Met conformation; carcinoma (RCC), NSCLC, and CRC the predictive value of (ii) nonselective type II inhibitors (e.g. cabozantinib, c-Metþ versus c-Met was demonstrated.25-28 foretinib) that bind to the c-Met inactive conformation; and Druggable aberrations with the highest predictive power (iii) nonselective type III inhibitors (e.g. tivantinib) that do are exon 14 skipping (ex14skip) mutations29; crizotinib, not compete with ATP binding. moAbs can be directed capmatinib, and tepotinib received a breakthrough Food against c-Met (e.g. emibetuzumab, onartuzumab) or HGF and Drug Administration (FDA) approval in NSCLC patients (e.g. ficlatuzumab, rilotumumab).25 harboring ex14skip due to the improvement in OS.30,31

Volume 6 - Issue 2 - 2021 https://doi.org/10.1016/j.esmoop.2020.100040 3 ESMO Open R. Danesi et al.

In conclusion, c-Met aberrations can be classified as missense mutations (3.48%), amplification (0.12%), and druggable with high predictive value in NSCLC (Table 1). point mutations F1174L (0.03%) and P367R (0.03%).5-8 First-, second-, and third-generation ALK/ROS1 inhibitors RET such as crizotinib, alectinib, ceritinib, brigatinib, lorlatinib, can be used depending on the resistance mutation on The RET proto-oncogene encodes an RTK that binds ligands which they are active, including L1196M, D1203N, F1174L, belonging to the glial cell-derived neurotrophic factor G1202R, C1156Y, and G1269A.46,47 family, including glial cell-derived neurotrophic factor, Crizotinib is superior to chemotherapy in previously neurturin, artemin, and persephin. RET is altered in 2.53% treated, advanced ALK-orROS1-rearranged NSCLC of cancers including NSCLC, CRC, BrCa, melanoma, and patients.48-50 Ceritinib,51,52 alectinib,53,54 and brigatinib55 thyroid tumors. Mutations (2.78%) and gene loss (0.18%) are superior to crizotinib in untreated ALKþ NSCLC. are commonly observed in RETþ patients; relevant genetic Lorlatinib, a third-generation inhibitor, may represent an alterations are M918T (0.13%), D567N (0.03%), and effective treatment option in first-line or subsequent lines amplifications (0.08%).5-8 Gain-of-function mutations of RET of therapy due to the central nervous system activity in cause sporadic and familial medullary thyroid carcinoma, ALKþ and ROS1þ patients.56-59 The ALKþ inflammatory and multiple endocrine neoplasia 2A syndrome, whereas myofibroblastic tumor treated with crizotinib displayed a loss-of-function aberrations occur in Hirschsprung's disease. higher percentage of objective responses than ALK.60 An RET is activated by fusion to form the chimeric oncogene optimized sequence of ALK inhibitors may increase survival RET/papillary thyroid carcinoma.32 Multikinase inhibitors in patients.61 are cabozantinib and vandetanib, whereas selective drugs In conclusion, a large amount of clinical evidence include selpercatinib and pralsetinib.33,34 RET inhibitors are demonstrates that target-based stratification is an optimal active in several tumors, including NSCLC.35 strategy to select patients who will benefit from ALK Actionability of RET alterations is demonstrated in several inhibitors (Table 1). The same approach for other cancer cancers.36 NSCLC patients with RET rearrangements (e.g. types requires further investigation. KIF5BeRET fusion) respond to cabozantinib37 and vandetanib.38 Patients with RET-mutant NSCLC and NTRK medullary thyroid carcinoma treated with pralsetinib experienced durable clinical responses with manageable NTRK1-3 activate PI3K, RAS/MEK/ERK, and phospholipase C toxicity profiles.39 Furthermore, case report studies showed (PLC)-g pathways during neuronal development and are clinical response after treatment with cabozantinib in a extremely rare drivers of a variety of cancers, with NTRK3 metastatic BrCa patient with the NCOA4eRET fusion fusions being more common in secretory BrCa, mammary progressing to HER2-targeted therapy40 and in a late-stage analog secretory carcinoma, and congenital fibrosarcomas. NSCLC patient with the KIF5BeRET gene fusion.41 Finally, The most common NTRK1 alterations are mutation (2.09%) osimertinib combined with pralsetinib induces response in and amplification (0.43%); among mutations, R214W double EGFR/RET-mutant NSCLC patients.42 (0.03%), R157H (0.02%), and T434M (0.02%) are the most In conclusion, RET aberrations are druggable and further frequent.5-8 The most common alterations of NTRK2 are investigation using well-designed clinical trials with mutation (1.39%), mainly A662T (0.02%) and V606I (0.02%); sufficient statistical power is needed to confirm the possible amplification (0.04%); and loss of function (0.03%).5-8 histology-independent classification of RET-driven tumors. Finally, NTRK3 genetic abnormalities are mutation (2.61%), amplification (0.10%), and loss (0.04%); relevant NTRK3 ALK/ROS1 mutations are K746T (0.02%) and R306H (0.02%).5-8 Potent and selective NTRK inhibitors are entrectinib and The anaplastic lymphoma kinase (ALK) is a RTK involved in larotrectinib; nonselective inhibitors are also available.62 several tumors, including NSCLC and anaplastic large Entrectinib induces durable responses in adult or pediatric cell lymphomas. ALK recognizes some different ligands, patients with advanced NTRK fusion-positive solid cancers including pleiotrophin and neurite growth-promoting factor regardless of the tumor type,63 and the same evidence was 2 (NEGF2).43 Alterations including ALK fusion, ALK provided for larotrectinib.64,65 FDA has approved NTRK TKIs copy-number gain, and activating ALK mutations are found for tumor-agnostic indication due to its recognized clinical in multiple cancers.44 ROS1 gene encodes a receptor that efficacy in multiple cancer types.66 belongs to the ALK/leukocyte TK (LTK) and insulin RTK NTRK TKIs have proven clinical efficacy in NTRK families and, once activated, induces cell proliferation by fusion-positive cancers. The availability of validated stimulating MAPK (mitogen-activated protein kinase)/ERK, methods for NTRK analysis makes this biomarker a PI3K/AKT, and JAK/STAT3 (Janus kinase/signal transducers successful example of the application of target-based and activators of transcription 3) signaling pathways. The classification of tumors (Table 1). homology to ALK accounts for the development of drugs targeting both.45 ALK is altered in 3.21% of all cancers with c-Kit NSCLC, CRC, melanoma, BrCa, and uterine corpus neoplasm having the greatest prevalence, the most common The proto-oncogene c-Kit encodes for a RTK belonging to aberrations being fusion (w5%), mutation (3.81%), the PDGF/c-Kit family that can bind stem cell factor; mutant

4 https://doi.org/10.1016/j.esmoop.2020.100040 Volume 6 - Issue 2 - 2021 R. Danesi et al. ESMO Open c-Kit has been implicated in the tumorigenesis of cholangiocarcinomas87 while FGFR3 is altered in 4.81% of all melanoma, acute myeloid leukemia (AML), and cancers including bladder, BrCa, lung, ovarian, stomach, gastrointestinal stromal tumors (GISTs).67 c-Kit is altered in gliomas, and sarcomas.88 The most common alterations in 2.99% of cancers with connective and soft tissue FGFR3 are amplifications (2.53%) and mutations (0.36%), neoplasms, NSCLC, melanoma, glioma, and CRC having the including R181H (0.01%) and E189K (0.01%).5-8 Pan-FGFR highest prevalence. The most common alterations are inhibitors include erdafitinib and rogaratinib,89 while mutations (3.38%), amplifications (0.64%), and deletions infigratinib and pemigatinib are two selective and potent (0.54%); clinically relevant missense mutation is D816H inhibitors of FGFR1-3.90 Patients harboring FGFR2 fusions or (0.12%).5-8 In GIST, c-Kit mutations occur more frequently in rearrangements had a 35.5% objective response to exon 11 (70%), and exons 9, 13, 14, and 17.68 c-Kit pemigatinib.91 Previously treated patients with locally multitargeted TKIs include imatinib, dasatinib, and nilotinib, advanced and unresectable or metastatic urothelial which also inhibit platelet-derived growth factor receptor carcinoma harboring FGFR alterations respond favorably to (PDGFR)-a, PDGFR-b, and ephrin RTKs.69 Adjuvant imatinib erdafitinib.92 has been associated with reduced recurrence rates and In conclusion, FGFR2/3 alterations may predict response improved OS in high-risk primary GIST,70 whereas dasatinib to selective FGFR inhibitors in multiple cancers (Table 1). In induces high response rates in TKI-naïve GIST.71 line with this notion, phase II and III trials are ongoing to c-Kit overexpression is a strong negative prognostic factor test safety and efficacy of pemigatinib for various in NSCLC72; imatinib has no activity in c-Kit overexpressing FGFR-driven tumors. small-cell lung cancer (SCLC) owing to the lack of c-Kit activating mutations.67 VEGF/VEGFR Imatinib is effective in c-Kit-mutated melanoma, a rare The VEGF and its receptor (VEGFR) play a major role in clinicopathological entity mainly located at acral and cancer by regulating angiogenesis (VEGF-A and VEGFR-1/2) mucosal sites,73 but not in c-Kit-amplified tumors.74 and lymphangiogenesis (VEGF-C/D and VEGFR-3). Nilotinib is active in imatinib-treated patients,75 while Angiopoietin and its receptors (TIE1-2) are also involved in dasatinib response rate is low among c-Kitþ melanoma.76 pathological neovascularization.93 The VEGFeVEGFR axis is Sunitinib shows activity in the treatment of mucosal and targeted by moAbs (e.g. bevacizumab, ramucirumab), and acral melanoma, regardless of the presence of a c-Kit VEGFR TKIs (e.g. sorafenib, sunitinib, axitinib, tivozanib, mutation.77 c-Kit is mutated in 60%-80% of AML patients78 pazopanib, regorafenib, and cediranib). and inhibitors may play a therapeutic role.79 Anti-VEGF/VEGFR agents are successfully used in multiple In conclusion, c-Kit activating mutations appear to be cancer types including RCC, CRC, hepatocellular carcinoma, druggable and predictive of response (Table 1). Further NSCLC, and cervical cancer. For others (e.g. glioblastoma, clinical investigations are needed to clearly establish the BrCa, and ovarian cancers), the increase in progression-free reliability of a target-based tumor classification. survival (PFS) was not associated with OS improvement.94 Anti-angiogenic drugs are used for diverse purposes, for PDGFRa example, chemo-potentiation in NSCLC, maintenance in The platelet-derived growth factor receptor alpha (PDGFRa) ovarian cancer, and immune system regulation in gene is mutated in GIST (5%-10%), NSCLC (6%), CRC (5%), combination with checkpoint inhibitors.7,8,94 Unfortunately, and gliomas (1%); gene amplifications and fusions are predictive biomarkers of response are lacking.95 However, a observed in glioblastoma (12%),5-8,80 hypereosinophilic recent study showed that a composite model using syndrome,81 and dermatofibrosarcoma protuberans.82 VHL, TP53, and VEGFR-1 predicted PFS on first-line Multitargeting TKIs with PDGFRa-inhibitory activity VEGF-targeted therapies in RCC.96 include imatinib, sunitinib, regorafenib, crenolanib, and In conclusion, no single actionable mutations were found avapritinib.83 in VEGF/VEGFR (Table 1); however, the investigation of Activating mutations of PDGFRa are uncommon in GISTs predictive biomarkers should be encouraged in prospective (5%-10%) and mutually exclusive with c-Kit but confer clinical trials to optimize VEGF-targeted therapies. sensitivity to imatinib.84,85 PDGFRa D842V mutation is associated with resistance to imatinib, sunitinib, and RAS regorafenib,84 but sensitive to crenolanib and avapritinib,84 RAS mutations are frequent in several cancers8,97,98;in while Y288C is a resistance mutation to PDGFR inhibitors.80 particular, KRAS (overall, 14%) is mutated in pancreatic In conclusion, GISTs must be analyzed for mutations in cancer, NSCLC, and CRC; NRAS (2.6%) in melanoma and c-Kit and PDGFRa at the time of diagnosis to guide AML; and HRAS (1%) in bladder cancer.8,99 The most treatment strategies based on targeted drugs (Table 1). common alterations in KRAS are mutations (20.6%), among which codon 12 missense (15.9%)dparticularly FGFR G12D (5.6%)dis the most common.5-8 KRAS or NRAS Fibroblast growth factors (FGFs) and their receptors mutations in exons 2-4 are found in 50% of CRC patients; (FGFR1-4) have a pathogenic role in cancer.86 FGFR2 fusions they are predictive of resistance to anti-EGFR treatments, or rearrangements occur in 10%-16% of intrahepatic thus limiting the use of panitumumab and cetuximab to

Volume 6 - Issue 2 - 2021 https://doi.org/10.1016/j.esmoop.2020.100040 5 ESMO Open R. Danesi et al. patients with RAS wild-type tumors and chemotherapy/ mutations (0.51%, 0.07% of which are E322K), amplifica- antiangiogenic drugs to RAS-mutated cancers.10,100 tions (0.26%), A7 duplication (0.12%), and gene loss Several covalent KRAS G12C inhibitors, such as sotorasib (0.07%).5-8 (AMG-510), adagrasib (MRTX849), and JNJ-74699157, are Binimetinib, trametinib, and cobimetinib are MEK in clinical development.101-103 inhibitors; their combinations with BRAF TKIs significantly In conclusion, RAS mutations are difficult to target improved OS in BRAF V600E/K melanoma patients, as because GTP binds the guanosine triphosphate/guanosine compared with single agents.108,119,120 diphosphate pocketdthe only one that can be targeted by In conclusion, MEK inhibitors are currently used in small moleculesdwith extremely high affinity103; however, combination and the clinical perspectives of these drugs are the new drugs give hope that KRAS will eventually become in BRAFeRASeERK-driven tumors. a druggable target.

BRAF PI3K/AKT/mTOR BRAF is altered in 7%-15% of all cancers including CRC, The PI3K/AKT/mTOR pathway plays a pivotal role in cancer and is negatively counterbalanced by phosphatase and tensin melanoma, NSCLC, papillary thyroid carcinoma, hairy cell 121,122 leukemia (HCL), Langerhans cell histiocytosis, and Erdheime homolog (PTEN). However, PTEN activity is frequently 8,104 lost, leading to constitutive PI3K/AKT/mTOR activation, a Chester disease. The most common alterations in BRAF 122,123 are mutations and amplifications and are mutually exclusive poor prognosis indicator. Mutations occur in the p110a with RAS aberrations.5-8 The glutamic acid substitution catalytic subunit of PI3K (PIK3CA) and are detectable in within the activation segment of the kinase domain, gastric cancers (18%), CRC (15%), BrCa (20%-50%), and head resulting from V600E missense mutation, constitutively and neck squamous cell cancers (HNSCCs; 30.5%), whereas fi 122 activates the MEKeERK pathway. Atypical, non-V600 BRAF genomic ampli cation is more frequent in NSCLC. Loss of mutants are found in NSCLC but are rare in CRC.105 PTEN is observed in CRC (20%-40%), whereas mTOR w Approximately 50% of melanomas harbor activating BRAF activation occurs in bladder and prostate cancers ( 40%); w V600 mutations with the V600E variant accounting for up to BrCa and ovarian cancers ( 40%) and prostate cancers w 123 90% of cases.106 ( 50%) show increased AKT1 activity. Selective BRAF TKIs include vemurafenib, encorafenib, Abnormalities of the PIK3CA gene copy number and and trametinib; a high response rate is obtained with missense mutations result in persistent activation of PI3K. 107 Most common point mutations are H1047R (4.3%), E545K vemurafenib in BRAF V600E/K-mutated melanomas, 7,8,124 which is further increased by combination with MEK (3.7%), and E542K (2.3%). AKT1 displays mutation fi inhibitors (i.e. dabrafenib/trametinib).108-111 [2.09%, including E17K (1.26%)], ampli cation (0.30%), and gene loss (0.07%)8 in human cancers, with actionable BRAF mutations occur in 10% of CRC and are fourfold 124-127 higher in patients with right- than left-sided cancer.112 mutations being uncommon. Activation of the PI3K/AKT pathway plays a role in Burkitt lymphomagenesis Combination of encorafenib and the MEK inhibitor 128,129 binimetinib plus cetuximab to suppress EGFR-mediated and is associated with worse outcomes. Finally, mTOR resistance improves OS in this poor-prognosis aberrations are rare; the most frequent are E1799K (0.05%) fi 8 population.113 and gene ampli cation (0.04%). Vemurafenib is effective in relapsed/refractory Pathway inhibition can be achieved by targeting mTOR 114 (temsirolimus and everolimus), PI3K (alpelisib and taselisib), BRAF-mutated HCL with an ORR of 91% at 1 year. 130 Resistance frequently occurs and the BRAF/MEK TKI and AKT (ipatasertib, capivasertib). combination is not more effective over BRAF TKI alone114; PIK3CA and/or PTEN aberrations predict response to PI3K/AKT/mTOR inhibitors in a histology-independent however, relapsed/refractory HCL patients may well 130 respond to vemurafenib/rituximab.115 BRAF mutations are fashion. PIK3CA hotspot mutations may predict response to taselisib in patients with advanced solid found in 1-2% of NSCLC and dabrafenib showed 33% ORR in 131 pretreated patients116; the BRAF/MEK TKI combination tumors. Taselisib plus fulvestrant induced a higher ORR fi 117 in patients with PIK3CA-mutated versus nonmutated shows encouraging activity in rst line setting. 132 In conclusion, BRAF V600 mutations are druggable in BrCa. Alpelisib is active in PIK3CA-mutated solid tumors,133 and in combination with fulvestrant prolongs some, but not all, cancers (Table 1). This statement is in 134 line with data obtained in V600 mutation-positive PFS in previously treated PIK3CA-mutated BrCa, although nonmelanoma patients treated with vemurafenib.118 tumors with PIK3CA mutations occasionally do not respond to these drugs.135 PI3K/AKT inhibitors are used in refractory, indolent, and aggressive B-NHLs.136,137 PIK3CA aberrations MEK are classified as tier IA (high level of evidence), according to MEK (also known as MAPK) serves as a downstream target the ESMO Scale for Clinical Actionability of molecular for both RAS and BRAF proteins. MAPK1 is altered in 0.75% Targets (ESCAT).138 of cancers including NSCLC, bladder, CRC, BrCa, and In conclusion, the PI3K/AKT/mTOR pathway may offer endometrium.8 The most common alterations are both actionable (PTEN aberrations) and druggable (PIK3CA

6 https://doi.org/10.1016/j.esmoop.2020.100040 Volume 6 - Issue 2 - 2021 R. Danesi et al. ESMO Open mutations) targets (Table 1). It is worth noting that PIK3CA In conclusion, clinical evidence indicates that IDH and KRAS mutations may co-occur, suggesting that a more mutations can be considered druggable targets in AML but comprehensive screening is advisable. not in glioma (Table 1).

CDK4/6 NUCLEAR PROTEINS The cyclin D/cyclin-dependent kinase 4/6 (CDK4/6) is a PARP molecular complex that plays a pivotal role in cell cycle progression from the G1 to S phase due to phosphorylation Poly (ADP-ribose) polymerases (PARPs) have a key role in of Rb1 and loss of repression of the E2F transcription factor. DNA repair; PARP inhibition increases tumor sensitivity to Several types of cancers display activation of the cyclin DNA-damaging agents by destabilizing replication forks and fi D/CDK4/6 pathway; cyclin D1 (CCND1) gene amplification inducing death in cells lacking pro cient homologous was reported in w15-20% of human BrCa. recombination mechanisms (e.g. BRCA1 and BRCA2 153 Palbociclib, ribociclib, and abemaciclib are third- mutants). BRCA1 is dysfunctional in 3% of cancers generation inhibitors of CDK4/6 and Rb1 aberrations are a including NSCLC, BrCa, CRC, ovarian, and melanoma, with the promising biomarker so far.139 most common alterations being mutations [both germline 8,54 Actionable mutations or deregulated expression of and somatic (3%)] and frameshifts (0.4%). Germline BRCA PIK3CA, AKT, mTOR, VEGFR1/2/3, PDGFRb, c-Kit, FGFR1/2/ mutations were reported in 4-7% of patients with pancreatic 154 3/4, MEK, JAK1/2, STAT3/5,andCDK2/7/9 is involved in cancer. BRCA2 is altered in 4.6% of cancers including CRC, resistance to CDK4/6 inhibitors7,140; CDKN2A/2B loss could NSCLC, BrCa, prostate, and bladder; the most common 8 predict sensitivity to CDK4/6 inhibition.141 alterations are mutations (5.23%) and frameshifts (0.94%). In conclusion, although preclinical studies identified some PARP inhibitors (olaparib, niraparib, rucaparib, and fi potential actionable alterations, no clinical evidence has talazoparib) have high ef cacy against BRCA-mutated 155,156 been provided so far that these alterations can predict tumors, particularly ovarian cancer. In this disease, fi response to CDK4/6 inhibitors in patients with multiple veliparib signi cantly improves PFS in BRCA mutants after 157 fi cancer types (Table 1). Nonetheless, this drug class is induction therapy, while olaparib provides a signi cant fi 158 commonly used in combination with endocrine therapy in PFS gain in homologous recombination-de cient tumors. BrCa patients. Niraparib increases PFS in patients previously treated with platinum-based schedules, regardless of BRCA mutations.159 fi IDH1/2 The same clinical bene t was observed with talazoparib in BRCA1/2þ BrCa patients160 and with olaparib in metastatic Mutant isocitrate dehydrogenases 1 and 2 (IDH1/2) produce pancreatic cancer.161 Rucaparib is effective in advanced 2-hydroxyglutarate, and are involved in chromatin ovarian cancer with deleterious germline and/or somatic remodeling. IDH1 is altered in w3% of cancers with BRCA mutations.162 glioblastoma, oligodendroglioma, astrocytoma, AML, and In conclusion, BRCA mutations in ovarian and BrCa 8,142 bile duct carcinoma having the greatest prevalence. The patients are actionable targets that predict response to most common IDH1 alterations are mutations (4%), PARP inhibitors. The strength of evidence is lower for including missense R132H/C/L (2%, 0.8%, and 0.14%, somatic than germline BRCA mutations (Table 1). For other 5-8 respectively). IDH2 is altered in 1.13% of cancers types of cancer (e.g. prostate, pancreatic, and SCLC), the including AML, BrCa, CRC, NSCLC, and myelodysplastic predictive role of BRCA status on the response to PARP syndromes; commonly observed alterations are mutation inhibitors must be consolidated. (1.35%), missense mutations in codons 140 (0.41%, R140Q 0.38%) and 172 (0.34%), and amplification (0.2%).8 ERa and ERb Owing to their role in leukemogenesis,143,144 inhibitors of IDH1/2 (ivosidenib, enasidenib) were developed for the Estrogen receptors (ERa and ERb) belong to the steroid/ treatment of AML.145 nuclear receptor superfamily and play a pivotal role in Ivosidenib induces durable remission in IDH1-mutant endocrine regulation of BrCa.163,164 ERa is mainly expressed AML,146 while enasidenib is active in AML harboring IDH2 in the uterus, prostate stroma, ovarian theca cells, Leydig mutations in codons 140 and 172147,148; however, cells in the testis, epididymis, mammary gland, and liver, resistance may develop due to second-site Q316E and whereas ERb is expressed in prostate epithelium, testes, I319M mutations.149 It is unclear whether IDH inhibitors are ovarian granulosa cells, bone marrow, and the brain. ERa effective therapeutic strategy in glioma, as their ability to promotes cell cycle progression, acting as ligand-dependent cross the bloodebrain barrier remains unclear.150 The transcription factor.164 IDH1/2 inhibitor vorasidenib crosses the bloodebrain Selective ER modulators (tamoxifen, raloxifene) or barrier and is being developed for the treatment of degrader (fulvestrant) and aromatase inhibitors low-grade IDH-mutant glioma.151 Another element of (anastrozole, exemestane)165 are available. uncertainty is the prognosis of gliomas which seems to be ERa mutations are frequently observed in metastatic better in mutated cases.152 BrCa patients previously treated with aromatase

Volume 6 - Issue 2 - 2021 https://doi.org/10.1016/j.esmoop.2020.100040 7 ESMO Open R. Danesi et al. inhibitors166 and are related to constitutive ligand- the TML cut-off predictive of drug response may vary in independent ER activation and resistance to hormonal different tumor types, suggesting the need for additional therapies. Thus, ERa mutations detected in BRCA-positive studies integrating TML with other parameters to refine BrCa may have important therapeutic implication.167 ERa predictions for improved patient selection. was found mutated in a patient with low-grade serous ovarian carcinoma who became resistant to aromatase Microsatellite instability inhibitor therapy after prolonged response.168 Genetic instability in short nucleotide repeats (MSI) is the In conclusion, ESMO ranked ERa mutations in class 2A result of abnormal DNA mismatch repair caused by a high (Table 1) although ERa actionability is limited to ER-positive mutation rate. MSI frequency is reported in CRC (13%), BrCa patients. endometrial and gastric cancer (22%), HNSCC (3%), RCC (2%), and bladder cancer (1%). These figures are lower than OTHER BIOMARKERS those of patients that may respond to PD-1/PD-L1 177 PD-1/PD-L1 inhibitors. MSI-H tumors have durable responses, high ORR, and a statistically significant improvement in OS PD-1 and its ligand (PD-L1) constitute one of the most to PD-1 inhibitors. Nivolumab was approved for treatment important nodes in tumor immune escape. PD-L1 is of MSI-H metastatic CRC patients and pembrolizumab for fi expressed in several cancers as well as in tumor-in ltrating any MSI-H or MMR-deficient solid tumor.178,179 As clinical fi immune cells; mutations in the TTK dual-speci city kinase MSI testing is mostly performed in CRC and endometrial and PIK3CA may affect PD-L1 expression.169 Blocking PD-1/ 170 tumors, the prevalence of MSI in other cancer types is less PD-L1 interaction enhances T-cell response and for this described. The ability to detect MSI by approved liquid reason anti-PD-1 moAbs pembrolizumab, nivolumab, and biopsy methods may allow a dynamic monitoring of drug cemiplimab, and anti-PD-L1 moAbs atezolizumab, avelumab, response in different cancer types.180 and durvalumab have dramatically changed the manage- In conclusion, MSI is the first approved tissue-agnostic ment of several tumors, particularly in NSCLC and melanoma. biomarker for anti-PD-1 treatments, regardless of PD-L1 Suppression of the PD-1ePD-L1 axis provided excellent re- 171 expression, and data suggest that the use of MSI can also sults in classical Hodgkin lymphoma as well. Atezolizumab be extended to other drugs of the same class (Table 1). added to nab-paclitaxel improves OS in triple-negative BrCa, more effectively in PD-L1-positive tumors.172 Similar findings THE MOLECULAR TUMOR BOARD were observed with pembrolizumab and chemotherapy; pembrolizumab improves the pathologic complete res- The concept of MTB is strongly linked to the unique ponse in the neoadjuvant setting, regardless of PD-L1 evolution of cancer treatments toward the discovery of expression.173 actionable mutations that can also be druggable. Candidate Although with limitations, it can be concluded that PD-L1 patients are those with different types of cancers that either expression represents the best predictive biomarker. failed standard therapy or are expected to fail, those with Standardization of method validation for PD-L1 expression rare tumors for which there are no/few standard options, 181 may provide a less variable definition of PD-L1 and those with tumors of unknown primary origin. expression and cut-offs for selecting PD-L1 patients to MTB is a group of oncologists, radiologists, pathologists, receive anti-PD-1/PD-L1 treatments. molecular biologists, geneticists, bioinformaticians, pharmacologists, pharmacists, bioethicists, and patient representatives (Figure 2) that provide support for the Tumor mutational load inclusion of patients in clinical trials or the use of off-label Tumor cells harbor somatic mutations that contribute to drugs.182-184 their malignant phenotype. However, these mutations can Each professional has a definite role: the clinician lead to the production of neoantigens that may become will select the patients suitable for discussion in the MTB; more easily recognized by the immune system.174 he/she will document the absence of standard lines of Tumor mutational load (TML) has been associated with therapy, evaluate the risks/benefits of an experimental better responses to immune checkpoint inhibitors in therapeutic option proposed by the MTB and its melanoma, NSCLC, and HNSCC. Not limited to these tumors, compatibility with previous therapies. The molecular high somatic TML might predict clinical benefit across pathologist will diagnose the molecular alterations, discuss diverse cancer types,175 whereas no association was found their relevance, and, if necessary, suggest which additional in patients not treated with immune checkpoint inhibitors, tests should be conducted based on available highlighting the predictive, not prognostic, value of this methodologies. The radiologist will assess the extent of the biomarker.175 A strong correlation between TML and ORR disease and its progression, provide a comparative analysis with anti-PD-1 or anti-PD-L1 was found in 27 tumor of the imaging data, and select representative images to types.176 facilitate multidisciplinary discussion. The geneticist and In conclusion, clinical evidence suggests that TML can be molecular biologist with expertise in oncogenetics will a reliable predictive biomarker of response to anti-PD-1/ evaluate the cases to be tested on germline DNA, and PD-L1 moAbs (Table 1). However, as for PD-L1 expression, discuss their role and the risk of heredity. The clinical

8 https://doi.org/10.1016/j.esmoop.2020.100040 Volume 6 - Issue 2 - 2021 R. Danesi et al. ESMO Open

MOLECULAR TUMOR BOARD (MTB): TEAM MEMBERS

Bioethicist: clearance of ethical issues Radiologist: diagnosis of initial/recurrent diseases Bioinformatician: data elaboration and interpretation Molecular geneticist: molecular validation of mutations Clinician: clinical management Pharmacist: regulatory assessments of treatments Pathologist: diagnosis of initial/recurrent diseases Pharmacologist: functional validation of druggable mutations Patient representative: patient’s participation in the process

Figure 2. Composition and function of the Molecular Tumor Board (MTB). pharmacologist will evaluate the patient's previous supporting/evaluating precision medicine trials focused on pharmacological history, perform molecular analysis of specific cancer genetic alterations. The major challenges in resistance mutations on liquid biopsy, discuss the biomarker-driven studies consist in understanding whether congruence of molecular alterations with the candidate a driver mutation is druggable or actionable. An example of drug as well as the risk of adverse reactions or drug tumor classification based on the overall frequency of interactions, the latter in collaboration with the pharmacist actionability has been reported in the AACR Project GENIE.8 who will assess the prescribability of specific treatments, Criteria adopted for classification include tumors having according to local and national regulations, particularly alterations for which approved drugs are available (1 or 2A); when it comes to off-label drugs. The bioinformatician will that are level 1 or 2A in other tumor types (2B), for which analyze the molecular data using networks and computer clinical evidence for response to investigational therapies in systems integrated with public databases. Lastly, the the same disease does exist (3A); and that are level 3A in bioethicist and patient representative are empowered to other tumor types (3B).8 Actionability rate varies across assess the ethical aspects associated with the use of different cancer types but, overall, the value stands at or experimental treatments and to communicate health care above 30%.8 OncoKB classifies oncogenic drivers into four decisions to the patient. levels of evidence based on whether genetic alterations are Hospitals with experienced MTBs in the United States able to predict response to FDA-approved targeted drugs or include the M.D. Anderson, Dana Farber Cancer Institute, investigational agents.185 Finally, actionable mutations have and the University of California at San Diego while, in been recently revised by the ESMO.138 Europe, the Center for Personalized Cancer Treatment in the In Figure 3 we graphically represent the workflow leading Netherlands and the Gustave Roussy Hospital in Paris, to MTB consultation, while in Table 1 we summarize the France, are worth mentioning. classification of druggable/actionable alterations. The need for integrated professional skills can be Liquid biopsy is widely used to monitor several actionable exemplified by taking into account the complexity of mutations; however, only few tests, including detection of

MOLECULAR CLINICAL MTB CLINICAL PROFILING INTERPRETATION MEETING DECISION APPLICATION MAKING

PATIENT BIOINFORMATIC VALIDATION OF ENROLLMENT ELABORATION ACTIONABLE TARGETS

Figure 3. Work flowdfrom patient enrollment to pharmacological treatment selection.

Volume 6 - Issue 2 - 2021 https://doi.org/10.1016/j.esmoop.2020.100040 9 ESMO Open R. Danesi et al.

EGFR mutations and ALK fusions in NSCLC, BRAF mutations employee of Roche and has integrated access to data. SI, LL, in CRC, PIK3CA mutations in breast cancer, BRCA1-2 and LA, LB, GA, and PC have nothing to disclose. ATM mutations in ovarian/prostate cancer and the MSI-H genotype (Table 1),186,187 have been approved as Companion Diagnostics (CDx) by regulatory authorities, REFERENCES such as the FDA. Genomic instability may allow tumors to 1. Hierro C, Matos I, Martin-Liberal J, et al. Agnostic-histology approval easily acquire new resistance mutations after initial of new drugs in oncology: are we already there? Clin Cancer Res. response to targeted drugs. Liquid biopsy is an efficient way 2019;25(11):3210-3219. to perform a monitoring of drug response and scientific 2. Dang CV, Reddy EP, Shokat KM, et al. Drugging the ‘undruggable’ efforts should be focused to obtain validated tests. cancer targets. Nat Rev Cancer. 2017;17(8):502-508. 3. Sigismund S, Avanzato D, Lanzetti L. Emerging functions of the EGFR Other important issues to be considered are study design in cancer. Mol Oncol. 2018;12(1):3-20. and selection of the clinical settings and outcomes. For 4. Zhang YL, Yuan JQ, Wang KF, et al. The prevalence of EGFR mutation example, translational research studies are generally in patients with non-small cell lung cancer: a systematic review and single-arm trials where the study drug is given to patients meta-analysis. Oncotarget. 2016;7(48):78985-78993. harboring specific genetic variants. The aim of these studies 5. My Cancer Genome Vanderbilt-Ingram Cancer Center. Nashville, TN: Vanderbilt University; 2017. Available at: https://www.mycancergen is to evaluate drug response in mutated versus nonmutated ome.org/. Accessed January 12, 2019. 8 population. Large-scale trials, such as the NCI-MATCH, 6. The Cancer Genome Atlas. Bethesda, MD: NCI/NIH; 2019. Available require a wide network of investigators, an extensive at: https://cancergenome.nih.gov/. Accessed January 12, 2019. collaboration with testing laboratories, and rapid 7. Bailey MH, Tokheim C, Porta-Pardo E, et al. Comprehensive charac- enrollment of a large number of cancer patients.66 terization of cancer driver genes and mutations. Cell. 2018;173(2): 371-385.e18. 8. The AACR Project GENIE Consortium. AACR Project GENIE: powering CONCLUSIONS precision medicine through an international consortium. Cancer An increasing number of new targeted drugs demonstrated Discov. 2017;7(8):818-831. fi 9. Del Re M, Petrini I, Mazzoni F, et al. Incidence of T790M in patients clinical bene ts in multiple cancer types. The availability of with NSCLC progressed to gefitinib, erlotinib, and afatinib: a study on new technologies (e.g. next generation sequencing) and circulating cell-free DNA. Clin Lung Cancer. 2019;21(3):P232-P237. approaches (e.g. liquid biopsy) allow clinicians to better select 10. Roskoski R Jr. The ErbB/HER family of protein-tyrosine kinases and patients based on their genetic make-up. Therefore, in the cancer. Pharmacol Res. 2014;79:34-74. era of precision medicine, integration of different 11. Jiang H, Zhu M, Li Y, et al. Association between EGFR exon 19 or exon 21 mutations and survival rates after first-line EGFR-TKI treatment in pa- professional skills is mandatory and the establishment of tients with non-small cell lung cancer. Mol Clin Oncol. 2019;11(3):301- MTB may represent the most important asset to support 308. clinicians in translating new scientific knowledge into daily 12. Park K, Tan EH, O'Byrne K, et al. Afatinib versus gefitinib as first-line clinical practice. Target-based classification is increasingly treatment of patients with EGFR mutation-positive non-small-cell used to integrate the histology-based classification of tumors, lung cancer (LUX-Lung 7): a phase 2B, open-label, randomised controlled trial. Lancet Oncol. 2016;17(5):577-589. which remains the backbone of cancer diagnosis and 13. Soria JC, Ohe Y, Vansteenkiste J, et al. Osimertinib in untreated EGFR- management. mutated advanced non-small-cell lung cancer. N Engl J Med. 2018;378(2):113-125. FUNDING 14. Fogli S, Polini B, Del Re M, et al. EGFR-TKIs in non-small-cell lung cancer: focus on clinical pharmacology and mechanisms of resistance. None declared. Pharmacogenomics. 2018;19(8):727-740. 15. Janne PA, Yang JC, Kim DW, et al. AZD9291 in EGFR inhibitor-resistant DISCLOSURE non-small-cell lung cancer. N Engl J Med. 2015;372(18):1689-1699. 16. Sequist LV, Soria JC, Goldman JW, et al. Rociletinib in EGFR- RD serves on the scientific advisory board and has consulting mutated non-small-cell lung cancer. N Engl J Med. 2015;372(18): relationship with Ipsen, Novartis, Pfizer, Sanofi Genzyme, 1700-1709. AstraZeneca, Janssen, Gilead, Lilly, Gilead, EUSA Pharma; and 17. Sidhanth C, Manasa P, Krishnapriya S, et al. A systematic under- reports support for travel, accommodation, and expenses standing of signaling by ErbB2 in cancer using phosphoproteomics. fi fi Biochem Cell Biol. 2018;96(3):295-305. from Ipsen, and Sano Genzyme. SF serves on the scienti c 18. Waks AG, Winer EP. Breast cancer treatment: a review. JAmMed advisory board of, has consulting relationship with, and Assoc. 2019;321(3):288-300. reports receiving support for travel expenses from Novartis, 19. Slamon DJ, Leyland-Jones B, Shak S, et al. Use of chemotherapy plus a Teva, Roche, BMS, Lilly. MDR serves on the scientificadvisory monoclonal antibody against HER2 for metastatic breast cancer that board and has consulting relationship with Ipsen, Novartis, overexpresses HER2. N Engl J Med. 2001;344(11):783-792. fi fi 20. Bang YJ, Van Cutsem E, Feyereislova A, et al. Trastuzumab in com- P zer, Sano Genzyme, AstraZeneca, Pierre-Fabre, Janssen; bination with chemotherapy versus chemotherapy alone for treat- and reports support for travel, accommodation, expenses ment of HER2-positive advanced gastric or gastro-oesophageal from Ipsen, AstraZeneca, Sanofi Genzyme. APDT serves on junction cancer (ToGA): a phase 3, open-label, randomised controlled the advisory boards of Roche, Bayer, Novartis, and reports trial. Lancet. 2010;376(9742):687-697. receiving a travel grant from PharmaMar. VG serves on the 21. Jebbink M, de Langen AJ, Boelens MC, et al. The force of HER2 e a druggable target in NSCLC? Cancer Treat Rev. 2020;86:101996. advisory board and speakers' bureau of Lilly and Novartis 22. Cortes J, Calvo E, Vivancos A, et al. New approach to cancer therapy (Advisory board and speakers' bureau), and reports receiving based on a molecularly defined cancer classification. CA Cancer J Clin. an institutional research grant from Roche. AP is an 2014;64(1):70-74.

10 https://doi.org/10.1016/j.esmoop.2020.100040 Volume 6 - Issue 2 - 2021 R. Danesi et al. ESMO Open

23. Li BT, Shen R, Buonocore D, et al. Ado-trastuzumab emtansine for 47. Chuang YC, Huang BY, Chang HW, et al. Molecular modeling of ALK patients with HER2-mutant lung cancers: results from a phase II L1198F and/or G1202R mutations to determine differential crizotinib basket trial. J Clin Oncol. 2018;36(24):2532-2537. sensitivity. Sci Rep. 2019;9(1):11390. 24. Sartore-Bianchi A, Trusolino L, Martino C, et al. Dual-targeted therapy 48. Shaw AT, Kim DW, Nakagawa K, et al. Crizotinib versus chemotherapy with trastuzumab and lapatinib in treatment-refractory, KRAS codon in advanced ALK-positive lung cancer. N Engl J Med. 2013;368(25): 12/13 wild-type, HER2-positive metastatic colorectal cancer (HERA- 2385-2394. CLES): a proof-of-concept, multicentre, open-label, phase 2 trial. 49. Solomon BJ, Mok T, Kim DW, et al. First-line crizotinib versus Lancet Oncol. 2016;17(6):738-746. chemotherapy in ALK-positive lung cancer. N Engl J Med. 25. Moosavi F, Giovannetti E, Saso L, et al. HGF/MET pathway aberrations 2014;371(23):2167-2177. as diagnostic, prognostic, and predictive biomarkers in human can- 50. Shaw AT, Ou SH, Bang YJ, et al. Crizotinib in ROS1-rearranged non- cers. Crit Rev Clin Lab Sci. 2019;56(8):533-566. small-cell lung cancer. N Engl J Med. 2014;371(21):1963-1971. 26. Zhang Y, Xia M, Jin K, et al. Function of the c-Met receptor tyrosine 51. Shaw AT, Kim TM, Crino L, et al. Ceritinib versus chemotherapy in kinase in carcinogenesis and associated therapeutic opportunities. patients with ALK-rearranged non-small-cell lung cancer previously Mol Cancer. 2018;17(1):45. given chemotherapy and crizotinib (ASCEND-5): a randomised, 27. Drilon A, Cappuzzo F, Ou SI, et al. Targeting MET in lung cancer: will controlled, open-label, phase 3 trial. Lancet Oncol. 2017;18(7): expectations finally be MET? J Thorac Oncol. 2017;12(1):15-26. 874-886. 28. Choueiri TK, Escudier B, Powles T, et al. Cabozantinib versus everolimus 52. Shaw AT, Kim DW, Mehra R, et al. Ceritinib in ALK-rearranged non- in advanced renal-cell carcinoma. N Engl J Med. 2015;373(19):1814- small-cell lung cancer. N Engl J Med. 2014;370(13):1189-1197. 1823. 53. Peters S, Camidge DR, Shaw AT, et al. Alectinib versus crizotinib in 29. Pasquini G, Giaccone G. C-MET inhibitors for advanced non-small cell untreated ALK-positive non-small-cell lung cancer. N Engl J Med. lung cancer. Expert Opin Investig Drugs. 2018;27(4):363-375. 2017;377(9):829-838. 30. Awad MM, Leonardi GC, Kravets S, et al. Impact of MET inhibitors on 54. Gilbert JA. Alectinib surpasses crizotinib for untreated ALK-positive survival among patients with non-small cell lung cancer harboring NSCLC. Lancet Oncol. 2017;18(7):e377. MET exon 14 mutations: a retrospective analysis. Lung Cancer. 55. Camidge DR, Kim HR, Ahn MJ, et al. Brigatinib versus crizotinib in ALK- 2019;133:96-102. positive non-small-cell lung cancer. N Engl J Med. 2018;379(21):2027- 31. Capmatinib could alter NSCLC treatment landscape. Cancer Discov. 2039. 2020;10(6):OF4. 56. Shaw AT, Felip E, Bauer TM, et al. Lorlatinib in non-small-cell lung 32. Plaza-Menacho I. Structure and function of RET in multiple endocrine cancer with ALK or ROS1 rearrangement: an international, multi- neoplasia type 2. Endocr Relat Cancer. 2018;25(2):T79-T90. centre, open-label, single-arm first-in-man phase 1 trial. Lancet Oncol. 33. Iams WT, Lovly CM. Stop fRETting the target: next-generation RET 2017;18(12):1590-1599. inhibitors have arrived. Cancer Discov. 2018;8(7):797-799. 57. Solomon BJ, Besse B, Bauer TM, et al. Lorlatinib in patients with ALK- 34. Kashoki M, Hanaizi Z, Yordanova S, et al. A comparison of EMA and positive non-small-cell lung cancer: results from a global phase 2 FDA decisions for new drug marketing applications 2014-2016: study. Lancet Oncol. 2018;19(12):1654-1667. concordance, discordance, and why. Clin Pharmacol Ther. 58. Shaw AT, Solomon BJ, Chiari R, et al. Lorlatinib in advanced ROS1- 2020;107(1):195-202. positive non-small-cell lung cancer: a multicentre, open-label, 35. Bronte G, Ulivi P, Verlicchi A, et al. Targeting RET-rearranged non- single-arm, phase 1-2 trial. Lancet Oncol. 2019;20(12):1691-1701. small-cell lung cancer: future prospects. Lung Cancer (Auckl). 59. Shaw AT, Friboulet L, Leshchiner I, et al. Resensitization to crizotinib 2019;10:27-36. by the lorlatinib ALK resistance mutation L1198F. N Engl J Med. 36. Subbiah V, Velcheti V, Tuch BB, et al. Selective RET kinase inhibition for 2016;374(1):54-61. patients with RET-altered cancers. Ann Oncol. 2018;29(8):1869-1876. 60. Schoffski P, Sufliarsky J, Gelderblom H, et al. Crizotinib in patients 37. Drilon A, Rekhtman N, Arcila M, et al. Cabozantinib in patients with with advanced, inoperable inflammatory myofibroblastic tumours advanced RET-rearranged non-small-cell lung cancer: an open-label, with and without anaplastic lymphoma kinase gene alterations single-centre, phase 2, single-arm trial. Lancet Oncol. 2016;17(12): (European Organisation for Research and Treatment of Cancer 90101 1653-1660. CREATE): a multicentre, single-drug, prospective, non-randomised 38. Yoh K, Seto T, Satouchi M, et al. Vandetanib in patients with previ- phase 2 trial. Lancet Respir Med. 2018;6(6):431-441. ously treated RET-rearranged advanced non-small-cell lung cancer 61. Barrows SM, Wright K, Copley-Merriman C, et al. Systematic review of (LURET): an open-label, multicentre phase 2 trial. Lancet Respir Med. sequencing of ALK inhibitors in ALK-positive non-small-cell lung 2017;5(1):42-50. cancer. Lung Cancer (Auckl). 2019;10:11-20. 39. Subbiah V, Gainor JF, Rahal R, et al. Precision targeted therapy with 62. Naito Y, Mishima S, Akagi K, et al. Japan Society of Clinical Oncology/ BLU-667 for RET-driven cancers. Cancer Discov. 2018;8(7):836-849. Japanese Society of Medical Oncology-led clinical recommendations 40. Paratala BS, Chung JH, Williams CB, et al. RET rearrangements are on the diagnosis and use of tropomyosin receptor kinase inhibitors in actionable alterations in breast cancer. Nat Commun. 2018;9(1):4821. adult and pediatric patients with neurotrophic receptor tyrosine ki- 41. Wang Y, Xu Y, Wang X, et al. RET fusion in advanced non-small-cell nase fusion-positive advanced solid tumors, cooperated by the Jap- lung cancer and response to cabozantinib: a case report. Medicine anese Society of Pediatric Hematology/Oncology. Int J Clin Oncol. (Baltimore). 2019;98(3):e14120. 2020;25(3):403-417. 42. Piotrowska Z, Isozaki H, Lennerz JK, et al. Landscape of acquired 63. Doebele RC, Drilon A, Paz-Ares L, et al. Entrectinib in patients with resistance to osimertinib in EGFR-mutant NSCLC and clinical valida- advanced or metastatic NTRK fusion-positive solid tumours: inte- tion of combined EGFR and RET inhibition with osimertinib and BLU- grated analysis of three phase 1-2 trials. Lancet Oncol. 2020;21(2): 667 for acquired RET fusion. Cancer Discov. 2018;8(12):1529-1539. 271-282. 43. Alshareef A. Novel molecular challenges in targeting anaplastic lym- 64. Laetsch TW, DuBois SG, Mascarenhas L, et al. Larotrectinib for pae- phoma kinase in ALK-expressing human cancers. Cancers (Basel). diatric solid tumours harbouring NTRK gene fusions: phase 1 results 2017;9(11):148. from a multicentre, open-label, phase 1/2 study. Lancet Oncol. 44. Holla VR, Elamin YY, Bailey AM, et al. ALK: a tyrosine kinase target for 2018;19(5):705-714. cancer therapy. Cold Spring Harb Mol Case Stud. 2017;3(1):a001115. 65. Hong DS, DuBois SG, Kummar S, et al. Larotrectinib in patients 45. Lin JJ, Shaw AT. Recent advances in targeting ROS1 in lung cancer. with TRK fusion-positive solid tumours: a pooled analysis of three J Thorac Oncol. 2017;12(11):1611-1625. phase 1/2 clinical trials. Lancet Oncol. 2020;21(4):531-540. 46. Recondo G, Mezquita L, Facchinetti F, et al. Diverse resistance 66. Chen AP, Eljanne M, Harris L, et al. National cancer institute basket/ mechanisms to the third-generation ALK inhibitor lorlatinib in ALK- umbrella clinical trials: MATCH, LungMAP, and beyond. Cancer J. rearranged lung cancer. Clin Cancer Res. 2020;26(1):242-255. 2019;25(4):272-281.

Volume 6 - Issue 2 - 2021 https://doi.org/10.1016/j.esmoop.2020.100040 11 ESMO Open R. Danesi et al.

67. Stankov K, Popovic S, Mikov M. C-KIT signaling in cancer treatment. primary resistance to front-line targeted therapy or immunotherapy. Curr Pharm Des. 2014;20(17):2849-2880. Clin Genitourin Cancer. 2020;18(4):252-257. 68. Oppelt PJ, Hirbe AC, Van Tine BA. Gastrointestinal stromal tumors 88. Roskoski R Jr. The role of fibroblast growth factor receptor (FGFR) (GISTs): point mutations matter in management, a review. protein-tyrosine kinase inhibitors in the treatment of cancers J Gastrointest Oncol. 2017;8(3):466-473. including those of the urinary bladder. Pharmacol Res. 2020;151: 69. Lindauer M, Hochhaus A. Dasatinib. Recent Results Cancer Res. 104567. 2018;212:29-68. 89. Roskoski R Jr. Properties of FDA-approved small molecule protein 70. Raut CP, Espat NJ, Maki RG, et al. Efficacy and tolerability of 5-year kinase inhibitors: a 2020 update. Pharmacol Res. 2019;152:104609. adjuvant imatinib treatment for patients with resected intermedi- 90. Casadei C, Dizman N, Schepisi G, et al. Targeted therapies for ate- or high-risk primary gastrointestinal stromal tumor: the PERSIST- advanced bladder cancer: new strategies with FGFR inhibitors. Ther 5 clinical trial. JAMA Oncol. 2018;4(12):e184060. Adv Med Oncol. 2019;11, 1758835919890285. 71. Montemurro M, Cioffi A, Domont J, et al. Long-term outcome of 91. Abou-Alfa GK, Sahai V, Hollebecque A, et al. Pemigatinib for previ- dasatinib first-line treatment in gastrointestinal stromal tumor: a ously treated, locally advanced or metastatic cholangiocarcinoma: a multicenter, 2-stage phase 2 trial (Swiss Group for Clinical Cancer multicentre, open-label, phase 2 study. Lancet Oncol. 2020;21(5):671- Research 56/07). Cancer. 2018;124(7):1449-1454. 684. 72. Xiao H, Wang J, Liu Y, et al. Relative influence of c-Kit expression and 92. Loriot Y, Necchi A, Park SH, et al. Erdafitinib in locally advanced or epidermal growth factor receptor gene amplification on survival in metastatic urothelial carcinoma. N Engl J Med. 2019;381(4):338-348. patients with non-small cell lung cancer. Oncol Lett. 2014;8(2):582- 93. Shibuya M. Vascular endothelial growth factor (VEGF) and its re- 588. ceptor (VEGFR) signaling in angiogenesis: a crucial target for anti- and 73. Ponti G, Manfredini M, Greco S, et al. BRAF, NRAS and C-KIT pro-angiogenic therapies. Genes Cancer. 2011;2(12):1097-1105. advanced melanoma: clinico-pathological features, targeted-therapy 94. Hegde PS, Wallin JJ, Mancao C. Predictive markers of anti-VEGF and strategies and survival. Anticancer Res. 2017;37(12):7043-70 emerging role of angiogenesis inhibitors as immunotherapeutics. 48. Semin Cancer Biol. 2018;52(Pt 2):117-124. 74. Hodi FS, Corless CL, Giobbie-Hurder A, et al. Imatinib for melanomas 95. Choueiri TK, Fay AP, Gagnon R, et al. The role of aberrant VHL/HIF harboring mutationally activated or amplified KIT arising on mucosal, pathway elements in predicting clinical outcome to pazopanib ther- acral, and chronically sun-damaged skin. J Clin Oncol. 2013;31(26): apy in patients with metastatic clear-cell renal cell carcinoma. Clin 3182-3190. Cancer Res. 2013;19(18):5218-5226. 75. Guo J, Carvajal RD, Dummer R, et al. Efficacy and safety of nilotinib in 96. Stenehjem DD, Hahn AW, Gill DM, et al. Predictive genomic markers patients with KIT-mutated metastatic or inoperable melanoma: final of response to VEGF targeted therapy in metastatic renal cell carci- results from the global, single-arm, phase II TEAM trial. Ann Oncol. noma. PLoS One. 2019;14(1):e0210415. 2017;28(6):1380-1387. 97. Cox AD, Fesik SW, Kimmelman AC, et al. Drugging the undruggable 76. Kalinsky K, Lee S, Rubin KM, et al. A phase 2 trial of dasatinib in RAS: mission possible? Nat Rev Drug Discov. 2014;13(11):828-851. patients with locally advanced or stage IV mucosal, acral, or vulvo- 98. McCormick F. KRAS as a therapeutic target. Clin Cancer Res. vaginal melanoma: a trial of the ECOG-ACRIN Cancer Research Group 2015;21(8):1797-1801. (E2607). Cancer. 2017;123(14):2688-2697. 99. Porru M, Pompili L, Caruso C, et al. Targeting KRAS in metastatic 77. Buchbinder EI, Sosman JA, Lawrence DP, et al. Phase 2 study of colorectal cancer: current strategies and emerging opportunities. sunitinib in patients with metastatic mucosal or acral melanoma. J Exp Clin Cancer Res. 2018;37(1):57. Cancer. 2015;121(22):4007-4015. 100. Sforza V, Martinelli E, Ciardiello F, et al. Mechanisms of resistance to 78. Malaise M, Steinbach D, Corbacioglu S. Clinical implications of c-Kit anti-epidermal growth factor receptor inhibitors in metastatic colo- mutations in acute myelogenous leukemia. Curr Hematol Malig Rep. rectal cancer. World J Gastroenterol. 2016;22(28):6345-6361. 2009;4(2):77-82. 101. Canon J, Rex K, Saiki AY, et al. The clinical KRAS(G12C) inhibitor AMG 79. Heo SK, Noh EK, Kim JY, et al. Targeting c-KIT (CD117) by dasatinib and 510 drives anti-tumour immunity. Nature. 2019;575(7781):217-223. radotinib promotes acute myeloid leukemia cell death. Sci Rep. 102. Hallin J, Engstrom LD, Hargis L, et al. The KRAS(G12C) inhibitor 2017;7(1):15278. MRTX849 provides insight toward therapeutic susceptibility of KRAS- 80. Ip CKM, Ng PKS, Jeong KJ, et al. Neomorphic PDGFRA extracellular mutant cancers in mouse models and patients. Cancer Discov. domain driver mutations are resistant to PDGFRA targeted therapies. 2020;10(1):54-71. Nat Commun. 2018;9(1):4583. 103. Nagasaka M, Li Y, Sukari A, et al. KRAS G12C Game of Thrones, which 81. Cools J, DeAngelo DJ, Gotlib J, et al. A tyrosine kinase created by direct KRAS inhibitor will claim the iron throne? Cancer Treat Rev. fusion of the PDGFRA and FIP1L1 genes as a therapeutic target of 2020;84:101974. imatinib in idiopathic hypereosinophilic syndrome. N Engl J Med. 104. Cohn AL, Day BM, Abhyankar S, et al. BRAF(V600) mutations in solid 2003;348(13):1201-1214. tumors, other than metastatic melanoma and papillary thyroid can- 82. Simon MP, Pedeutour F, Sirvent N, et al. Deregulation of the platelet- cer, or multiple myeloma: a screening study. Onco Targets Ther. derived growth factor B-chain gene via fusion with collagen gene 2017;10:965-971. COL1A1 in dermatofibrosarcoma protuberans and giant-cell fibro- 105. Dankner M, Rose AAN, Rajkumar S, et al. Classifying BRAF alterations blastoma. Nat Genet. 1997;15(1):95-98. in cancer: new rational therapeutic strategies for actionable muta- 83. Boonstra PA, Gietema JA, Suurmeijer AJH, et al. Tyrosine kinase in- tions. Oncogene. 2018;37(24):3183-3199. hibitor sensitive PDGFRAlpha mutations in GIST: two cases and re- 106. Cheng L, Lopez-Beltran A, Massari F, et al. Molecular testing for BRAF view of the literature. Oncotarget. 2017;8(65):109836-109847. mutations to inform melanoma treatment decisions: a move toward 84. Florou V, Trent JC, Wilky BA. Precision medicine in gastrointestinal precision medicine. Mod Pathol. 2018;31(1):24-38. stromal tumors. Discov Med. 2019;28(155):267-276. 107. Sosman JA, Kim KB, Schuchter L, et al. Survival in BRAF V600-mutant 85. Heinrich MC, Corless CL, Demetri GD, et al. Kinase mutations and advanced melanoma treated with vemurafenib. N Engl J Med. imatinib response in patients with metastatic gastrointestinal stromal 2012;366(8):707-714. tumor. J Clin Oncol. 2003;21(23):4342-4349. 108. Robert C, Grob JJ, Stroyakovskiy D, et al. Five-year outcomes with 86. Patani H, Bunney TD, Thiyagarajan N, et al. Landscape of activating dabrafenib plus trametinib in metastatic melanoma. N Engl J Med. cancer mutations in FGFR kinases and their differential responses to 2019;381(7):626-636. inhibitors in clinical use. Oncotarget. 2016;7(17):24252-24268. 109. Garbe C, Amaral T, Peris K, et al. European consensus-based inter- 87. Hamieh L, Beck RL, Le VH, et al. The efficacy of lenvatinib plus disciplinary guideline for melanoma. Part 2: Treatment e update everolimus in patients with metastatic renal cell carcinoma exhibiting 2019. Eur J Cancer. 2020;126:159-177.

12 https://doi.org/10.1016/j.esmoop.2020.100040 Volume 6 - Issue 2 - 2021 R. Danesi et al. ESMO Open

110. Robert C, Karaszewska B, Schachter J, et al. Improved overall survival 133. Juric D, Rodon J, Tabernero J, et al. Phosphatidylinositol 3-kinase in melanoma with combined dabrafenib and trametinib. N Engl J alpha-selective inhibition with alpelisib (BYL719) in PIK3CA-altered Med. 2015;372(1):30-39. solid tumors: results from the first-in-human study. J Clin Oncol. 111. Flaherty KT, Infante JR, Daud A, et al. Combined BRAF and MEK in- 2018;36(13):1291-1299. hibition in melanoma with BRAF V600 mutations. N Engl J Med. 134. Andre F, Ciruelos E, Rubovszky G, et al. Alpelisib for PIK3CA-mutated, 2012;367(18):1694-1703. hormone receptor-positive advanced breast cancer. N Engl J Med. 112. Ducreux M, Chamseddine A, Laurent-Puig P, et al. Molecular targeted 2019;380(20):1929-1940. therapy of BRAF-mutant colorectal cancer. Ther Adv Med Oncol. 135. Yang J, Nie J, Ma X, et al. Targeting PI3K in cancer: mechanisms and 2019;11, 1758835919856494. advances in clinical trials. Mol Cancer. 2019;18(1):26. 113. Kopetz S, Grothey A, Yaeger R, et al. Encorafenib, binimetinib, and 136. Awan FT, Gore L, Gao L, et al. Phase Ib trial of the PI3K/mTOR in- cetuximab in BRAF V600E-mutated colorectal cancer. N Engl J Med. hibitor voxtalisib (SAR245409) in combination with chemo- 2019;381(17):1632-1643. immunotherapy in patients with relapsed or refractory B-cell 114. Tiacci E, Park JH, De Carolis L, et al. Targeting mutant BRAF in relapsed malignancies. Br J Haematol. 2016;175(1):55-65. or refractory hairy-cell leukemia. N Engl J Med. 2015;373(18):1733- 137. Bhatti M, Ippolito T, Mavis C, et al. Pre-clinical activity of targeting the 1747. PI3K/Akt/mTOR pathway in Burkitt lymphoma. Oncotarget. 115. Falini B, Tiacci E. New treatment options in hairy cell leukemia with 2018;9(31):21820-21830. focus on BRAF inhibitors. Hematol Oncol. 2019;37(suppl 1):30-37. 138. Condorelli R, Mosele F, Verret B, et al. Genomic alterations in breast 116. Planchard D, Kim TM, Mazieres J, et al. Dabrafenib in patients with cancer: level of evidence for actionability according to ESMO Scale for BRAF(V600E)-positive advanced non-small-cell lung cancer: a single- Clinical Actionability of molecular Targets (ESCAT). Ann Oncol. arm, multicentre, open-label, phase 2 trial. Lancet Oncol. 2019;30(3):365-373. 2016;17(5):642-650. 139. Condorelli R, Spring L, O'Shaughnessy J, et al. Polyclonal RB1 muta- 117. Leonetti A, Facchinetti F, Rossi G, et al. BRAF in non-small cell lung tions and acquired resistance to CDK 4/6 inhibitors in patients with cancer (NSCLC): pickaxing another brick in the wall. Cancer Treat Rev. metastatic breast cancer. Ann Oncol. 2018;29(3):640-645. 2018;66:82-94. 140. Tong Z, Sathe A, Ebner B, et al. Functional genomics identifies pre- 118. Hyman DM, Puzanov I, Subbiah V, et al. Vemurafenib in multiple dictive markers and clinically actionable resistance mechanisms to nonmelanoma cancers with BRAF V600 mutations. N Engl J Med. CDK4/6 inhibition in bladder cancer. J Exp Clin Cancer Res. 2019;38(1): 2015;373(8):726-736. 322. 119. Neuzillet C, Tijeras-Raballand A, de Mestier L, et al. MEK in cancer and 141. Ozeki T, Nagahama M, Fujita K, et al. Influence of CYP3A4/5 and ABC cancer therapy. Pharmacol Ther. 2014;141(2):160-171. transporter polymorphisms on lenvatinib plasma trough concentra- 120. Long GV, Flaherty KT, Stroyakovskiy D, et al. Dabrafenib plus trame- tions in Japanese patients with thyroid cancer. Sci Rep. 2019;9(1): tinib versus dabrafenib monotherapy in patients with metastatic 5404. BRAF V600E/K-mutant melanoma: long-term survival and safety 142. Rohle D, Popovici-Muller J, Palaskas N, et al. An inhibitor of mutant analysis of a phase 3 study. Ann Oncol. 2017;28(7):1631-1639. IDH1 delays growth and promotes differentiation of glioma cells. 121. Chalhoub N, Baker SJ. PTEN and the PI3-kinase pathway in cancer. Science. 2013;340(6132):626-630. Annu Rev Pathol. 2009;4:127-150. 143. Kats LM, Reschke M, Taulli R, et al. Proto-oncogenic role of mutant 122. Tian T, Li X, Zhang J. mTOR signaling in cancer and mTOR inhibitors in IDH2 in leukemia initiation and maintenance. Cell Stem Cell. solid tumor targeting therapy. Int J Mol Sci. 2019;20(3):755. 2014;14(3):329-341. 123. Altomare DA, Testa JR. Perturbations of the AKT signaling pathway in 144. Amatangelo MD, Quek L, Shih A, et al. Enasidenib induces acute human cancer. Oncogene. 2005;24(50):7455-7464. myeloid leukemia cell differentiation to promote clinical response. 124. McGranahan N, Favero F, de Bruin EC, et al. Clonal status of action- Blood. 2017;130(6):732-741. able driver events and the timing of mutational processes in cancer 145. Golub D, Iyengar N, Dogra S, et al. Mutant isocitrate dehydrogenase evolution. Sci Transl Med. 2015;7(283):283ra54. inhibitors as targeted cancer therapeutics. Front Oncol. 2019;9: 125. Brastianos PK, Carter SL, Santagata S, et al. Genomic characterization 417. of brain metastases reveals branched evolution and potential thera- 146. DiNardo CD, Stein EM, de Botton S, et al. Durable remissions with peutic targets. Cancer Discov. 2015;5(11):1164-1177. ivosidenib in IDH1-mutated relapsed or refractory AML. N Engl J Med. 126. Vandekerkhove G, Todenhofer T, Annala M, et al. Circulating tumor 2018;378(25):2386-2398. DNA reveals clinically actionable somatic genome of metastatic 147. Stein EM, DiNardo CD, Pollyea DA, et al. Enasidenib in mutant IDH2 bladder cancer. Clin Cancer Res. 2017;23(21):6487-6497. relapsed or refractory acute myeloid leukemia. Blood. 2017;130(6): 127. Yi KH, Lauring J. Recurrent AKT mutations in human cancers: func- 722-731. tional consequences and effects on drug sensitivity. Oncotarget. 148. Stein EM, DiNardo CD, Fathi AT, et al. Molecular remission and 2016;7(4):4241-4251. response patterns in patients with mutant-IDH2 acute myeloid leu- 128. Schmitz R, Young RM, Ceribelli M, et al. Burkitt lymphoma patho- kemia treated with enasidenib. Blood. 2019;133(7):676-687. genesis and therapeutic targets from structural and functional ge- 149. Intlekofer AM, Shih AH, Wang B, et al. Acquired resistance to IDH nomics. Nature. 2012;490(7418):116-120. inhibition through trans or cis dimer-interface mutations. Nature. 129. Patnaik A, Appleman LJ, Tolcher AW, et al. First-in-human phase I 2018;559(7712):125-129. study of copanlisib (BAY 80-6946), an intravenous pan-class I phos- 150. Karpel-Massler G, Nguyen TTT, Shang E, et al. Novel IDH1-targeted phatidylinositol 3-kinase inhibitor, in patients with advanced solid glioma therapies. CNS Drugs. 2019;33(12):1155-1166. tumors and non-Hodgkin's lymphomas. Ann Oncol. 2016;27(10):1928- 151. Konteatis Z, Artin E, Nicolay B, et al. Vorasidenib (AG-881): a first-in- 1940. class, brain-penetrant dual inhibitor of mutant IDH1 and 2 for treat- 130. Janku F, Hong DS, Fu S, et al. Assessing PIK3CA and PTEN in early- ment of glioma. ACS Med Chem Lett. 2020;11(2):101-107. phase trials with PI3K/AKT/mTOR inhibitors. Cell Rep. 2014;6(2): 152. IDH inhibitors target common glioma mutation. Cancer Discov. 377-387. 2019;9(8):992. 131. Juric D, Krop I, Ramanathan RK, et al. Phase I dose-escalation study of 153. Farmer H, McCabe N, Lord CJ, et al. Targeting the DNA repair defect in taselisib, an oral PI3K inhibitor, in patients with advanced solid tu- BRCA mutant cells as a therapeutic strategy. Nature. 2005;434(7035): mors. Cancer Discov. 2017;7(7):704-715. 917-921. 132. Dickler MN, Saura C, Richards DA, et al. Phase II study of taselisib 154. Friedenson B. BRCA1 and BRCA2 pathways and the risk of cancers (GDC-0032) in combination with fulvestrant in patients with HER2- other than breast or ovarian. MedGenMed. 2005;7(2):60. negative, hormone receptor-positive advanced breast cancer. Clin 155. Slade D. PARP and PARG inhibitors in cancer treatment. Genes Dev. Cancer Res. 2018;24(18):4380-4387. 2020;34(5-6):360-394.

Volume 6 - Issue 2 - 2021 https://doi.org/10.1016/j.esmoop.2020.100040 13 ESMO Open R. Danesi et al.

156. Haddad G, Saade MC, Eid R, et al. PARP inhibitors: a tsunami of in- 171. Matsuki E, Younes A. Checkpoint inhibitors and other immune ther- dications in different malignancies. Pharmacogenomics. 2020;21(3): apies for Hodgkin and non-Hodgkin lymphoma. Curr Treat Options 221-230. Oncol. 2016;17(6):31. 157. Coleman RL, Fleming GF, Brady MF, et al. Veliparib with first-line 172. Schmid P, Adams S, Rugo HS, et al. Atezolizumab and nab-paclitaxel in chemotherapy and as maintenance therapy in ovarian cancer. advanced triple-negative breast cancer. N Engl J Med. 2018;379(22): N Engl J Med. 2019;381(25):2403-2415. 2108-2121. 158. Ray-Coquard I, Pautier P, Pignata S, et al. Olaparib plus bevacizumab 173. Schmid P, Cortes J, Pusztai L, et al. Pembrolizumab for early triple- as first-line maintenance in ovarian cancer. N Engl J Med. negative breast cancer. N Engl J Med. 2020;382(9):810-821. 2019;381(25):2416-2428. 174. Galuppini F, Dal Pozzo CA, Deckert J, et al. Tumor mutation burden: 159. Gonzalez-Martin A, Pothuri B, Vergote I, et al. Niraparib in patients from comprehensive mutational screening to the clinic. Cancer Cell with newly diagnosed advanced ovarian cancer. N Engl J Med. Int. 2019;19:209. 2019;381(25):2391-2402. 175. Samstein RM, Lee CH, Shoushtari AN, et al. Tumor mutational load 160. Litton JK, Rugo HS, Ettl J, et al. Talazoparib in patients with advanced predicts survival after immunotherapy across multiple cancer types. breast cancer and a germline BRCA mutation. N Engl J Med. Nat Genet. 2019;51(2):202-206. 2018;379(8):753-763. 176. Yarchoan M, Hopkins A, Jaffee EM. Tumor mutational burden and 161. Golan T, Hammel P, Reni M, et al. Maintenance olaparib for germline response rate to PD-1 inhibition. N Engl J Med. 2017;377(25):2500- BRCA-mutated metastatic pancreatic cancer. N Engl J Med. 2501. 2019;381(4):317-327. 177. Chen Q, Li T, Yue W. Drug response to PD-1/PD-L1 blockade: based on 162. Balasubramaniam S, Beaver JA, Horton S, et al. FDA approval sum- biomarkers. Onco Targets Ther. 2018;11:4673-4683. mary: rucaparib for the treatment of patients with deleterious BRCA 178. Marcus L, Lemery SJ, Keegan P, et al. FDA approval summary: pem- mutation-associated advanced ovarian cancer. Clin Cancer Res. brolizumab for the treatment of microsatellite instability-high solid 2017;23(23):7165-7170. tumors. Clin Cancer Res. 2019;25(13):3753-3758. 163. Brufsky AM, Dickler MN. Estrogen receptor-positive breast cancer: 179. Mehrvarz Sarshekeh A, Overman MJ, Kopetz S. Nivolumab in the exploiting signaling pathways implicated in endocrine resistance. treatment of microsatellite instability high metastatic colorectal Oncologist. 2018;23(5):528-539. cancer. Future Oncol. 2018;14(18):1869-1874. 164. Jameera Begam A, Jubie S, Nanjan MJ. Estrogen receptor agonists/ 180. Bonneville R, Krook MA, Kautto EA, et al. Landscape of microsatellite antagonists in breast cancer therapy: a critical review. Bioorg Chem. instability across 39 cancer types. JCO Precis Oncol. 2017;2017. PO.17. 2017;71:257-274. 00073. 165. Gombos A. Selective oestrogen receptor degraders in breast cancer: a 181. Schwartzberg L, Kim ES, Liu D, et al. Precision oncology: who, how, review and perspectives. Curr Opin Oncol. 2019;31(5):424-429. what, when, and when not? Am Soc Clin Oncol Educ Book. 2017;37: 166. Fribbens C, O'Leary B, Kilburn L, et al. Plasma ESR1 mutations and the 160-169. treatment of estrogen receptor-positive advanced breast cancer. J Clin 182. Stoekle HC, Mamzer-Bruneel MF, Frouart CH, et al. Molecular tumor Oncol. 2016;34(25):2961-2968. boards: ethical issues in the new era of data medicine. Sci Eng Ethics. 167. Vidula N, Rich TA, Sartor O, et al. Routine plasma-based genotyping to 2018;24(1):307-322. comprehensively detect germline, somatic, and reversion BRCA mu- 183. Rolfo C, Manca P, Salgado R, et al. Multidisciplinary molecular tumour tations among patients with advanced solid tumors. Clin Cancer Res. board: a tool to improve clinical practice and selection accrual for 2020;26(11):2546-2555. clinical trials in patients with cancer. ESMO Open. 2018;3(5):e00 168. Stover EH, Feltmate C, Berkowitz RS, et al. Targeted next-generation 0398. sequencing reveals clinically actionable BRAF and ESR1 mutations in 184. Harada S, Arend R, Dai Q, et al. Implementation and utilization of the low-grade serous ovarian carcinoma. JCO Precis Oncol. 2018;2018(2): molecular tumor board to guide precision medicine. Oncotarget. 1-8. 2017;8(34):57845-57854. 169. Menyhart O, Pongor LS, Gyorffy B. Mutations defining patient cohorts 185. Chakravarty D, Gao J, Phillips SM, et al. OncoKB: a precision oncology with elevated PD-L1 expression in gastric cancer. Front Pharmacol. knowledge base. JCO Precis Oncol. 2017;2017. PO.17.00011. 2018;9:1522. 186. Ou SI, Nagasaka M, Zhu VW. Liquid biopsy to identify actionable 170. Dermani FK, Samadi P, Rahmani G, et al. PD-1/PD-L1 immune genomic alterations. Am Soc Clin Oncol Educ Book. 2018;38:978-997. checkpoint: potential target for cancer therapy. J Cell Physiol. 187. Malone ER, Oliva M, Sabatini PJB, et al. Molecular profiling for pre- 2019;234(2):1313-1325. cision cancer therapies. Genome Med. 2020;12(1):8.

14 https://doi.org/10.1016/j.esmoop.2020.100040 Volume 6 - Issue 2 - 2021