ESMO 2019 Update

Axel Grothey Director, GI Oncology Research West Cancer Center Research Institute Encorafenib plus Cetuximab With or Without Binimetinib for BRAF V600E–Mutant Metastatic Colorectal Cancer: Expanded Results from a Randomized, 3-Arm, Phase 3 Study vs. the Choice of Either Irinotecan or FOLFIRI plus Cetuximab (BEACON CRC)

Josep Tabernero, Axel Grothey, Eric Van Cutsem, Rona Yaeger, Harpreet Wasan, Takayuki Yoshino, Jayesh Desai, Fortunato Ciardiello, Fotios Loupakis, Yong Sang Hong, Neeltje Steeghs, Tormod Kyrre Guren, Hendrik-Tobias Arkenau, Pilar Garcia-Alfonso, Ashwin Gollerkeri, Michael Pickard, Kati Maharry, Janna Christy-Bittel, Lisa Anderson, and Scott Kopetz

BEACON CRC: Binimetinib, Encorafenib, And Cetuximab COmbiNed to Treat BRAF-mutant ColoRectal Cancer

2 Study Design

Patients with BRAFV600E mCRC with disease progression after 1 or 2 prior regimens; ECOG PS of 0 or 1; and no prior treatment with any RAF inhibitor, MEK inhibitor, or EGFR inhibitor

Phase 3 Primary Endpoints: Safety Lead-in ENCORAFENIB + Triplet therapy BINIMETINIB + ENCORAFENIB + BINIMETINIB + CETUXIMAB Triplet vs Control n = 205 CETUXIMAB N = 30 OS R Doublet therapy (All randomized Pts) Encorafenib 300 mg PO daily 1:1:1 ENCORAFENIB + CETUXIMAB Binimetinib 45 mg PO bid n = 205 Cetuximab standard weekly ORR – dosing Blinded Central Control arm Review FOLFIRI + CETUXIMAB, or (1st 331 randomized Pts) irinotecan + CETUXIMAB n = 205

Randomization was stratified by ECOG PS (0 vs. 1), prior use of irinotecan (yes vs. no), and cetuximab source (US-licensed vs. EU-approved)

Secondary Endpoints: Doublet vs Control and Triplet vs Doublet - OS & ORR, PFS, Safety QOL Assessments: EORTC QOL Questionnaire (QLQ C30), Functional Assessment of Cancer Therapy Colon Cancer, EuroQol 5D5L, and Patient Global Impression of Change). Overall Survival and Objective Response Rate

Triplet vs Control Doublet vs Control

HR (95% CI): 0.52 (0.39-0.70) Triplet HR (95% CI): 0.60 (0.45-0.79) 2-sided P<0.0001 Doublet 2-sided P=0.0003

Median OS in months (95% CI) Median OS in months (95% CI) Triplet Control Doublet Control 9.0 (8.0-11.4) 5.4 (4.8-6.6) 8.4 (7.5-11.0) 5.4 (4.8-6.6) Control Control

Objective Response Rate (First 331 Randomized Patients)

Triplet Doublet Control Confirmed Response by BICR N=111 N=113 N=107 Objective Response Rate 26% 20% 2% 95% (CI) (18, 35) (13, 29) (<1, 7) p-value vs. Control <0.0001 <0.0001 Kopetz et al. Ann Oncol 2019;30 (Supplement 4): iv137–iv151 (LBA-006). Overall Survival: Triplet vs Doublet (All Randomized Patients) Study not powered to formally compare the results of the triplet combination to the doublet combination

HR (95% CI): 0.79 (0.59-1.06)

Median OS in months (95% CI) Triplet Triplet Doublet 9.0 (8.0-11.4) 8.4 (7.5-11.0)

Median Follow up: 7.8 Months* 33% of patients alive with less than 6 months FU

Doublet

5

*all randomized patients. Overall Survival: Triplet vs Doublet (First 331 Randomized Patients*) Study not powered to formally compare the results of the triplet combination to the doublet combination

HR (95% CI): 0.74 (0.53-1.04)

Median OS in months (95% CI) Triplet Triplet Doublet 9.5 (8.1-12.0) 8.3 (6.2-10.7)

Median Follow up: 12.5 Months 1% of patients alive with less than 6 months FU

Doublet

6

*Post-hoc descriptive analysis. Waterfall Plots of Best Change in Sum of Diameters (based on central review) Triplet

Control

N=87

DoubletDoublet

N=73

N=98

*Patients whose SoD was contraindicated by assessment of PD. SoD=sum of longest diameter. Includes patients with measurable disease with a baseline and at least one post-baseline scan. 7 Combined Waterfall Plots: Triplet and Doublet

Triplet (N=87) Doublet (N=98) WilcoxonDoublet Rank Sum P-Value: 0.033

*Patients whose Sum of longest diameter was contraindicated by assessment of PD. 2-sided p-value is descriptive only for this analysis. Comparison of the two waterfall plots is suggestive of a shift towards greater tumor reduction from baseline in the Triplet. 8 Grade ≥3 Adverse Events and Laboratory Abnormalities*

Triplet Doublet Control Event N=222 N=216 N=193 Grade ≥3 Grade ≥3 Grade ≥3 Diarrhea 10% 2% 10% Abdominal pain 6% 2% 5% Nausea 5% <1% 1% Vomiting 4% 1% 3% Pulmonary embolism 4% 1% 4% Intestinal obstruction 3% 4% 3% Asthenia 3% 3% 5% Acute kidney injury 3% 2% <1% Fatigue 2% 4% 4% Dermatitis acneiform 2% <1% 3% Ileus 2% 1% 2% Urinary tract infection 1% 2% 1% Cancer pain <1% 2% <1% Laboratory Abnormality Grade ≥3 Grade ≥3 Grade ≥3 Hemoglobin (g/L), hypo 11% 4% 4% Creatinine (umol/L), hyper 5% 2% 1% Creatine Kinase (IU/L), hyper 3% 0 0 Bilirubin (umol/L), hyper 2% 2% 3%

9 *Occurring in at least 2% of patients in either triplet or doublet arms.

Other CRC Highlights • FOxTROT: • Neoadjuvant FOLFOX (x3) in locally advanced colon cancer decreases R1 resection rate • Decreased surgical morbidity • 20% of patients overstaged (unnecessary chemo) • No tumor regression in MSI-H cancer • MOUNTAINEER: • HER-2 pos CRC: Tucatinib + Trastuzumab generate 55% RR (in 22 pts) • IDEA high-risk stage II: • Same regimen-effect as seen in stage III (CAPOX vs FOLFOX) • IDEA France: • ctDNA in stage III is prognostic (we knew that), but intensified chemo can improve outcome (that’s new) ctDNA in IDEA France • First documentation that “intensification” of adjuvant chemotherapy (6 vs 3 months) can improve outcome in ctDNA pos stage III colon cancer

Taieb, ESMO 2019 KEYNOTE-062 Study Design (NCT02494583)

Key Eligibility Criteria N = 256 • Locally advanced, Pembrolizumab 200 mg Q3W unresectable or for up to 35 cyclesb metastatic gastric or gastroesophageal adenocarcinoma N = 763 Until unacceptable Pembrolizumab 200 mg Q3W toxicity, disease • HER2/neu negative, R N = 257 (to 35 cycles) progression, or PD-L1-positive disease (1:1:1) + patient/physician (CPS ≥1) Chemotherapyc • ECOG PS 0 or 1 withdrawal decision

Stratification Factors N = 250 Placebo • Regiona + Chemotherapy • Locally advanced or metastatic disease • 5-FU or Capecitabine Primary endpoints: OS and PFS Secondary endpoints: ORR, Safety aEU/North America/Australia, Asia (South Korea, Hong Kong, Taiwan, Japan), Rest of World (including South America). Tabernero ASCO 2019 bAdministration of pembrolizumab monotherapy was not blinded. cChemotherapy: Cisplatin 80 mg/m2 Q3W + 5-FU 800 mg/m2/d for 5 days Q3W or capecitabine BID d1-14 Q3W (Cisplatin may be capped at 6 cycles as per country guidelines). Overall Survival: P vs C (CPS ≥1)

100 Events HR (99.2% CI) NIa 90 Pembro 79% 0.91 1.2 Chemo 86% (0.69-1.18) 80 70 12-mo rate 47% 24-mo rate 60 46% 27% 19% b 50 Median (95% CI) 10.6 mo (7.7-13.8) OS,% 40 11.1 mo (9.2-12.8) 30 20 10 0 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42

No. at Risk Time, months 256 201 162 139 120 107 94 83 59 38 23 12 4 0 0 250 230 192 144 114 94 75 49 38 21 15 6 2 2 0 aNI, non-inferiority margin; bHR (95% CI) = 0.91 (0.74-1.10), P = 0.162 for superiority of P vs C; Data cutoff: March 26, 2019. Tabernero ASCO 2019 Shitara ESMO 2019

Genetic Landscape of Biliary Cancers

ICC ECC

FGFR1-3 fusions, mutations ERBB2/3 amplification & amplifications (11-45%) (11-17%) IDH1/2 mutation (5-36%) IDH1/2 mutation (0-7%) RNF43 mutation (9%) PIK3CA mutation (7%) PIK3CA mutations (3-9%) MET mutation (4%) BRAF mutations (3-7%) BRAF mutations (3%) ERBB3 amplification (7%) MET amplification (1%) MET amplification (2-7%) ERBB3 mutation (7%) GBCA MET mutation (5%) ERBB2/3 amplification EGFR mutation (1-2%) (10-19%) PIK3CA mutation (6- 13%) BRAF mutation (1-6%) RNF43 mutation (4%) MAP2K4 mutation (4%) EGFR mutation (4%) FGFR1-3 fusions, mutations & amplifications (3%) IDH1/2 mutation (2%) Frequencies of Actionable Mutations in BTCs

30%

25%

20%

15%

10% ICC ECC 5% GBCA

0%

Goyal et al., ESMO GI 2017 Phase II: BGJ398 in FGFR altered Cholangiocarcinomas

RR: 14.8% SD: 60.7%

2/3 patients with more than 2 lines of prior therapy Javle et al, JCO 2018 Pemigatinib – Study Cohorts Pemigatinib RR (Cohort A – FGRF Fusions) Pemigatinib – Overall Survival ClarIDHy: A global, phase 3, randomized, double-blind study of ivosidenib vs placebo in patients with advanced cholangiocarcinoma with an isocitrate dehydrogenase 1 (IDH1) mutation

Ghassan K. Abou-Alfa,1,2 Teresa Maraculla,3 Milind Javle,4 R. Kate Kelley,5 Sam Lubner,6 Jorge Adeva,7 James M. Cleary,8 Daniel V. Catenacci,9 Mitesh J. Borad,10 John Bridgewater,11 William P. Harris,12 Adrian G. Murphy,13 Do-Youn Oh,14 Jonathan Whisenant,15 Bin Wu,16 Liewen Jiang,16 Camelia Gliser,16 Shuchi S. Pandya,16 Juan W. Valle,17 Andrew X. Zhu18

1Memorial Sloan-Kettering Cancer Center, New York, NY, USA; 2Weill Medical College at Cornell University, New York, NY, USA; 3Vall d’Hebron University Hospital, Vall d’Hebron Institute of Oncology (VHIO), Barcelona, Spain; 4MD Anderson Cancer Center, Houston, TX, USA; 5University of California San Francisco, San Francisco, CA, USA; 6University of Wisconsin Carbone Cancer Center, Madison, WI, USA; 7Hospital Universitario 12 de Octubre, Madrid, Spain; 8Dana-Faber Cancer Institute, Boston, MA, USA; 9University of Chicago Medical Center, Chicago, IL, USA; 10Mayo Clinic Cancer Center, Phoenix, AZ, USA; 11UCL Cancer Institute, London, UK; 12University of Washington, Seattle, WA, USA; 13Johns Hopkins University, Baltimore, MD, USA; 14Seoul National University Hospital, Cancer Research Institute, Seoul National University College of Medicine, Seoul, Korea; 15Utah Cancer Specialists, Salt Lake City, UT, USA; 16Agios Pharmaceuticals, Inc., Cambridge, MA, USA; 17University of Manchester, The Christie NHS Foundation Trust, Manchester, UK; 18Massachusetts General Hospital Cancer Center, Harvard Medical School, Boston, MA, USA

Presented at the European Society for Medical Oncology Congress, September 27– October 1, 2019, Barcelona, Spain 1 IDH1 mutations in advanced cholangiocarcinoma

Cytoplasm

Citrate

▪ Advanced cholangiocarcinoma is an aggressive rare cancer with treatment options Isocitrate limited primarily to chemotherapy1 2-HG 2-HG mIDH1 IDH1 ▪ IDH1 mutations occur in up to 20% of cholangiocarcinoma and do not confer a a- KG favorable prognosis1 2-HG NADPH

▪ Ivosidenib (AG-120) is a first-in-class, oral, targeted, small-molecule inhibitor of the Nucleus mutant IDH1 (mIDH1) protein,2 and is FDA-approved for mIDH1 R/R AML and ND AML Me not eligible for intensive chemotherapy3 a-KG-dependent dioxygenases Me ▪ A phase 1 study of ivosidenib included 73 previously treated mIDH1 cholangiocarcinoma Metabolic patients and was associated with: median PFS, 3.8 months; 6- and 12-month PFS rates, dysregulation Me 40.1% and 21.8%, respectively; and median OS 13.8 months4

Epigenetic changes Impaired cellular differentiation

2-HG=D-2-hydroxyglutarate; a-KG=alpha-ketoglutarate; AML=acute myeloid leukemia; FDA=Food and Drug Administration; Me=methyl groups; ND=newly-diagnosed; OS=overall survival; PFS=progression-free survival; R/R=relapsed/refractory. 1. Boscoe AN, et al. J Gastrointest Oncol. 2019;10:751-765. 2. Popovici-Muller J, et al. ACS Med Chem Lett. 2018;9:300-305. 3. TIBSOVO highlights of prescribing information. https://www.accessdata.fda.gov/drugsatfda_docs/label/2019/211192s001lbl.pdf. Accessed August 5, 2019. 4. Lowery MA, et al. Lancet Gastroenterol Hepatol. 2019;4:711-720. 3 ClarIDHy: Study design and endpoints

Ivosidenib 500 mg QD orally Key eligibility criteria in continuous 28-day r • ≥18 years of age or n e n

s ± b o

o ( 2 days) li nd e i • Histologically confirmed diagnosis of cholangiocarcinoma g f m t ti b pi n - i a cycles (n=124) • Centrally confirmed mIDH1* status by NGS a nu r t e z a Crossover permitted e i u 185 )

• ECOG PS score 0 or 1 by at radiographic een = • 1-2 prior therapies (at least 1 gemcitabine- or 5-FU- d disease progression n e ior th ( r doub l sc r

containing regimen) p tifi - a ando m e DH 1 m of :1 I • Measurable lesion as defined by RECIST v1.1 r r S t r 2 • Adequate hematologic, hepatic, and renal function P Placebo (n=61) NCT02989857 An independent data monitoring committee monitored the safety data throughout the study

▪ Primary endpoint: PFS by blinded independent radiology center (IRC) ▪ Secondary endpoints included: safety and tolerability; PFS by local review; OS; objective response rate; quality of life (QoL)†; pharmacokinetics/pharmacodynamics ▪ Sample size of ~186 patients based on hazard ratio (HR)=0.5, 96% power, 1-sided alpha=0.025 ▪ 780 patients were screened for IDH1 mutations across 49 sites and 6 countries

*IDH1 mutation status prospectively confirmed by NGS-based Oncomine™ Focus Assay on formalin-fixed, paraffin-embedded tumor tissue in a Clinical Laboratory Improvement Amendments-certified laboratory. †Assessed using EQ-5D-5L, EORTC QLQ-C30, EORTC QLQ-BIL21, and PGI questions. ECOG PS=Eastern Cooperative Oncology Group Performance Status; EORTC=European Organisation for Research and Treatment of Cancer; EQ-5D-5L=5-level EuroQoL-5 Dimension questionnaire; FU=fluorouracil; NGS=next-generation sequencing; PGI=Patient Global Impression; QD=once daily; QLQ-BIL21=Cholangiocarcinoma and Gallbladder Cancer module; QLQ-C30=Quality of Life Questionnaire Core 30; RECIST=Response Evaluation Criteria in Solid Tumors. 4 ClarIDHy: Patient disposition

Ivosidenib Placebo (n=124) (n=61) Treated, n (%) 121 (97.6) 59 (96.7) On treatment 38 (31.4) 8 (13.6) Discontinued treatment 83 (68.6) 51 (86.4) Progressive disease 65 (53.7) 44 (74.6) Adverse events 6 (5.0) 4 (6.8) Death 4 (3.3) 0 Withdrawal by patient 6 (5.0) 2 (3.4) Withdrawal of consent 1 (0.8) 1 (1.7) Other 1 (0.8) 0 Not treated, n (%) 3 (2.4) 2 (3.3) On study, n (%) 71 (57.3) 27 (44.3)

▪ As of the January 31, 2019 data cut, 35 placebo-treated patients (57.4%) crossed over to open-label ivosidenib upon radiographic disease progression and unblinding ▪ 26 placebo-treated patients (42.6%) did not cross over due to the following reasons: death (n=13), still on placebo treatment (n=8), never dosed (n=2), withdrawal of consent (n=2), received another treatment (n=1)

5 ClarIDHy: PFS by IRC

1.0 Ivosidenib Placebo + Censored Ivosidenib Placebo 0.9 PFS 0.8 HR=0.37 (95% CI 0.25, 0.54) Median, months 2.7 1.4 P<0.001 0.7

y 6-month rate 32% NE

ilit 0.6 b 12-month rate 22% NE 0.5 ob a r Disease control rate 53% 28% p 0.4 S S

F (PR+SD) (2% PR, 51% SD) (0% PR, 28% SD)

P 0.3

0.2

0.1

0.0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Number of patients at risk: 124 105 54 40 36 28 22 16 14 10 9 6 5 4 3 3 2 1 1 Ivoside nib 61 46 11 6 4 1 Placebo Survival (months)

NE=not estimable; PR=partial response; SD=stable disease. 7 ClarIDHy: Ivosidenib efficacy consistent across subgroups* PFS by IRC

Events/N Hazard ratio (HR) HR Lower 95% CI Upper 95% CI Overall 126/185 0.37 0.252 0.543 Prior lines of therapy 1 66/106 0.37 0.219 0.612 ≥2 60/79 0.41 0.234 0.730 Gender Female 74/117 0.36 0.220 0.589 Male 52/68 0.45 0.249 0.811 Extent of disease at screening Locally advanced 7/14 0.20 0.035 1.111 Metastatic 119/171 0.41 0.277 0.601 Cancer type at initial diagnosis Intrahepatic cholangiocarcinoma 114/169 0.38 0.257 0.567 extrahepatic cholangiocarcinoma 3/6 unknown 9/10 ECOG PS score at baseline 0 41/68 0.26 0.124 0.540 ≥1 85/117 0.52 0.332 0.803 Regions North America 83/124 0.40 0.249 0.631 Europe 34/49 0.39 0.188 0.830 Asia 9/12 0.42 0.110 1.597

0 1 Favors placebo 2 *Subgroups with events number ≤10 were not plotted. Favors ivosidenib 8 ClarIDHy: OS by intent-to-treat (ITT)

▪ Median OS based on 78 events was numerically 1.0 + Censored* Ivosidenib Placebo longer with ivosidenib than placebo (10.8 vs. Placebo (RPSFT-adjusted) 0.9 9.7 months) 0.8 − OS rates at 6 and 12 months for ivosidenib: 0.7 67% and 48% vs. 59% and 38% for placebo y 0.6 ilit

b ▪ 0.5 Rank-preserving structural failure time 1,2 ob a (RPSFT) method used to reconstruct the r p 0.4 survival curve for the placebo subjects as if S S

O 0.3 they had never crossed over to ivosidenib 0.2 HR=0.69 (95% CI 0.44, 1.10); P=0.06 0.1 ▪ With the RPSFT method, the median OS HR=0.46 (95% CI 0.28, 0.75); P<0.001 (RPSFT-adjusted) 0.0 with placebo adjusts to 6 months 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Number of patients at risk:

124 11 101 88 75 64 52 49 39 34 30 23 19 16 15 10 9 7 4 3 1 1 1 Ivosidenib 7 61 55 45 39 34 25 22 19 17 17 14 12 5 4 4 3 2 2 1 1 1 Placebo 61 55 42 32 22 16 10 4 1 1 Placebo (RPSFT- Survival (months) adjusted)

*Patients without documentation of death at the data cutoff date were censored at the date the patient was last known to be alive or the data cutoff date, whichever was earlier. 1. Watkins C, et al. Pharm Stat. 2013;12:348-357. 2. Robins JM, Tsiatis AA. Commun Stat Theory Methods. 1991;20:2609-2631. 9 ClarIDHy: Treatment-emergent adverse events (TEAEs)

Total Placebo Ivosidenib ivosidenib ▪ (n=59) (n=121) (n=156)* Grade >3 TEAE: 35.6% for placebo vs. 46.2% for total ivosidenib. Most common (placebo vs. total ivosidenib): Any TEAE, n (%) 57 (96.6) 115 (95.0) 146 (93.6) ascites (6.8% vs. 7.7%), bilirubin increase (1.7% vs. 5.8%), Most common TEAEs, n (%) anemia (0% vs. 5.1%), AST increase (1.7% vs. 5.1%) Nausea 15 (25.4) 43 (35.5) 50 (32.1) ▪ TEAEs leading to discontinuation were Diarrhea 9 (15.3) 37 (30.6) 45 (28.8) more common for Fatigue 10 (16.9) 32 (26.4) 37 (23.7) placebo (8.5% vs. 5.8%) than total Cough 5 (8.5) 25 (20.7) 30 (19.2) ivosidenib Abdominal pain 8 (13.6) 26 (21.5) 29 (18.6) ▪ TEAEs leading to dose Ascites 9 (15.3) 25 (20.7) 29 (18.6) reductions (2.6% vs. 0%) and interruptions Decreased appetite 11 (18.6) 23 (19.0) 27 (17.3) (26.3% vs. 16.9%) were Anemia 3 (5.1) 18 (14.9) 25 (16.0) more common for total ivosidenib relative to Vomiting 10 (16.9) 23 (19.0) 25 (16.0) placebo

*Total ivosidenib includes 35 patients initially assigned to placebo who had crossed over to ivosidenib upon radiographic disease progression and unblinding. >15% TEAEs based on total ivosidenib 10