RELAY Subgroup Analyses by EGFR Ex19del and Ex21l858r Mutations For

Home , Kyowa Hakko Kirin

Author Manuscript Published OnlineFirst on July 22, 2021; DOI: 10.1158/1078-0432.CCR-21-0273 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 RELAY Subgroup Analyses by EGFR Ex19del and Ex21L858R Mutations for

2 Ramucirumab Plus Erlotinib in Metastatic Non-small Cell Lung Cancer

4 Kazuhiko Nakagawa1, Ernest Nadal2, Edward B. Garon3, Makoto Nishio4, Takashi Seto5,

5 Nobuyuki Yamamoto6, Keunchil Park7, Jin-Yuan Shih8, Luis Paz-Ares9, Bente Frimodt-Moller10,

6 Annamaria H. Zimmermann11, Sameera Wijayawardana11, Carla Visseren-Grul12, and Martin

7 Reck13; for the RELAY study investigators*

9 1Department of Medical Oncology, Kindai University Faculty of Medicine, Osaka, Japan

10 2Department of Medical Oncology, Catalan Institute of Oncology, IDIBELL, L’Hospitalet de

11 Llobregat, Barcelona, Spain

12 3David Geffen School of Medicine at University of California Los Angeles/TRIO-US Network,

13 Los Angeles, CA, USA

14 4The Cancer Institute Hospital of JFCR, Tokyo, Japan

15 5Kyushu Cancer Center, Fukuoka, Japan

16 6Wakayama Medical University Hospital, Wakayama, Japan

17 7Samsung Medical Center, Seoul, South Korea

18 8Department of Internal Medicine, National Taiwan University Hospital, Taipei City, Taiwan

19 9Hospital Universitario 12 de Octubre, CNIO-H12o Lung Cancer Unit, Universidad

20 Complutense & CiberOnc, Madrid, Spain

21 10Eli Lilly and Company, Copenhagen, Denmark

22 11Eli Lilly and Company, Indianapolis, IN, USA

23 12Lilly Oncology, Utrecht, Netherlands

Downloaded from clincancerres.aacrjournals.org on September 27, 2021. © 2021 American Association for Cancer Research. Author Manuscript Published OnlineFirst on July 22, 2021; DOI: 10.1158/1078-0432.CCR-21-0273 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

24 13LungenClinic, Airway Research Center North (ARCN), German Center for Lung Research

25 (DZL), Grosshansdorf, Germany

27 *RELAY investigators are listed in Supplementary Material.

29 Running Title: RELAY: EGFR-Activating Mutation Subgroup Analyses

31 Keywords: (5 words/phrases allowed): NSCLC; EGFR mutation; ramucirumab; erlotinib.

33 Funding Information: Eli Lilly and Company sponsored this clinical trial, provided study

34 drugs, and participated in regulatory and ethics approval, safety monitoring, data collection, and

35 statistical analyses.

37 Corresponding Author:

38 Name: Kazuhiko Nakagawa, MD, PhD

39 Address: Kindai University Faculty of Medicine, 377-2, Ohno-higashi, Osakasayama City,

40 Osaka, 589-8511 Japan

41 Telephone: +81-90 6989 7671

42 Fax:

43 E-mail: [email protected]

44 Disclosure of Potential Conflicts of Interest:

45 KN reports grants and personal fees from Merck Sharp and Dohme (MSD), Eli Lilly Japan,

46 Bristol-Myers Squibb (BMS), Taiho Pharmaceutical, Ono Pharmaceutical, Chugai

47 Pharmaceutical, AstraZeneca, Astellas Pharma, Novartis Pharma, Nippon Boehringer Ingelheim,

48 and Pfizer Japan, during the conduct of the study; grants and personal fees from Takeda

49 Pharmaceutical, SymBio Pharmaceuticals, and Daiichi Sankyo, outside the submitted work;

50 grants from Merck Serono, ICON Japan, PAREXEL International, IQVIA Services Japan, A2

51 Healthcare, AbbVie, EP-CRSU, Linical, Otsuka Pharmaceutical, EPS International, Quintiles,

52 CMIC Shift Zero, Eisai, Kissei Pharmaceutical, Kyowa Hakko Kirin, EPS, Bayer Yakuhin,

53 inVentiv Health Japan, Gritsone Oncology, GlaxoSmithKline, Yakult Honsha, and Covance,

54 outside the submitted work; and personal fees from Kyorin Pharmaceutical, CareNet, Nichi-Iko

55 Pharmaceutical, Hisamitsu Pharmaceutical, Yodosha, Clinical Trial, Medicus Shuppan,

56 Publishers, Ayumi Pharmaceutical, Nikkei Business Publications, Thermo Fisher Scientific,

57 Nanzando, Medical Review, Yomiuri Telecasting, and Reno Medical, outside the submitted

58 work.

59 EN reports non-financial travel support from Eli Lilly and Company, and grants and consultant

60 fees from Amgen, AstraZeneca, BMS, Boehringer Ingelheim, Eli Lilly and Company, Merck

61 Serono, MSD, Pfizer, Roche, Takeda, outside the submitted work.

62 EBG reports grants to his institution from Eli Lilly and Company during the conduct of the

63 study; grants from AstraZeneca, BMS, Eli Lilly, Genentech, Merck, Novartis, Iovance, Neon,

64 Dynavax, and Mirati, outside the submitted work; advisory board personal fees from Dracen and

65 steering committee personal fees from EMD Serono outside the submitted work.

66 TS reports grants from Eli Lilly Japan during the conduct of the study; grants and personal fees

67 from Astellas Pharma, AstraZeneca, Chugai Pharmaceutical, Eli Lilly Japan, Kissei

68 Pharmaceutical, MSD, Nippon Boehringer Ingelheim, Novartis Pharma, Pfizer Japan, and

69 Takeda Pharmaceutical, outside the submitted work; personal fees from BMS, Kyowa Hakko

70 Kirin, Nippon Kayaku, Ono Pharmaceutical, Roche Singapore, Taiho Pharmaceutical, Thermo

71 Fisher Scientific, and Yakult Honsha, outside the submitted work; and grants from Bayer

72 Yakuhin, Daiichi Sankyo, Eisai, Loxo Oncology, and Merck Serono outside the submitted work.

73 MN reports grants and non-financial support from F Hoffmann-La Roche during the conduct of

74 the study; grants and personal fees from Ono Pharmaceutical, BMS, Pfizer, Chugai

75 Pharmaceutical, Eli Lilly, Taiho Pharmaceutical, AstraZeneca, Boehringer-Ingelheim, MSD, and

76 Novartis, outside the submitted work; personal fees from Sankyo Healthcare, Taiho

77 Pharmaceutical, and Merck Serono, outside the submitted work; and grants from Astellas outside

78 the submitted work.

79 J-YS reports personal fees from AstraZeneca, Roche, Boehringer Ingelheim, Eli Lilly, Pfizer,

80 Novartis, MSD, Chugai Pharma, Ono Pharmaceutical, BMS, and AbbVie outside the submitted

81 work.

82 LP-A reports medical advisor personal fees from Roche, BMS, MSD, Boehringer Ingelheim,

83 AstraZeneca, Merck, Novartis, Pfizer, Lilly, Celgene, Takeda, Amgen, Sanofi, and Phammar

84 outside the submitted work.

85 NY reports grants and personal fees from Eli Lilly Japan, during the conduct of the study; grants

86 and personal fees from AbbVie GK, Amgen, Astellas Pharma, AstraZeneca, BMS, Boehringer

87 Ingelheim, Chugai Pharmaceutical, Daiichi Sankyo, Eisai, Hisamitsu Pharmaceutical, Kyorin

88 Pharmaceutical, Life Technologies Japan, Merck Biopharma, MSD, Nippon Kayaku, Novartis

89 Pharma, Ono Pharmaceutical, Pfizer, Shionogi, Taiho Pharmaceutical, Takeda Pharmaceutical,

90 Terumo, Thermo Fisher Scientific, Toppan Printing, Tosoh, Tumura, outside the submitted work.

91 KP reports advisor fees from Amgen, Astellas BluePrint, Lilly, GlaxoSmithKline, Hanmi,

92 Kyowa Hakko Kirin, MSD, BMS, Ono Pharmaceutical, Roche, Merck KGaA, and Loxo outside

93 the submitted work; grants and advisor fees from Astra Zeneca outside the submitted work; and

94 speaker bureau and advisor fees from Boehringer Ingelheim outside the submitted work.

95 BF-M, AZ, SW, and CV-G are full-time employees of Eli Lilly and company and are Eli Lilly

96 shareholders.

97 MR reports lecture honoraria and consultancy fees from AbbVie, Amgen, AstraZeneca, BMS,

98 Boehringer Ingelheim, Celgene, Lilly, Merck, MSD, Novartis, Pfizer, and Roche outside the

99 submitted work.

100

101 Statement of Translational Relevance

102 Previous evidence suggests that in EGFR-mutated metastatic non-small cell lung cancer

103 (NSCLC), patients with ex21L858R mutation have a poorer prognosis and response to treatment

104 with EGFR-tyrosine kinase inhibitor (TKI) monotherapy than patients with ex19del mutation.

105 Results from RELAY showed that combination of the antiangiogenic agent ramucirumab with

106 EGFR-TKI targeted therapy provided significant and similar clinical benefit for both EGFR

107 ex19del and ex21L858R NSCLC, suggesting that the addition of an antiangiogenic agent may

108 abrogate the differential efficacy observed in ex21L858R patients compared with ex19del

109 patients when treated with EGFR-TKI monotherapy. These findings may help clinicians provide

110 a personalized treatment strategy and allow treatment decisions to be made on a more rational

111 basis.

112

113 ABSTRACT

114 Purpose: In EGFR-mutated metastatic non-small cell lung cancer (NSCLC), outcomes from

115 EGFR tyrosine kinase inhibitors have differed historically by mutation type present, with lower

116 benefit reported in patients with ex21L858R versus ex19del mutations. We investigated if

117 EGFR-activating mutation subtypes impact treatment outcomes in the phase 3 RELAY study.

118 Associations between EGFR mutation type and pre-existing co-occurring and treatment-

119 emergent genetic alterations were also explored.

120 Materials and Methods: Patients with metastatic NSCLC, an EGFR ex19del or ex21L858R

121 mutation, and no central nervous system metastases were randomized (1:1) to erlotinib (150

122 mg/day) with either ramucirumab (10 mg/kg) (RAM+ERL) or placebo (PBO+ERL), every 2

123 weeks, until RECIST v1.1–defined progression or unacceptable toxicity. The primary endpoint

124 was progression-free survial (PFS). Secondary and exploratory endpoints included overall

125 response rate (ORR), duration of response (DOR), PFS2, time-to-chemotherapy (TTCT), safety,

126 and next-generation sequencing analyses.

127 Results: Patients with ex19del and ex21L858R mutations had similar clinical characteristics and

128 co-mutational profiles. One-year PFS rates for ex19del patients were 74% for RAM+ERL versus

129 54% for PBO+ERL; for ex21L858R rates were 70% (RAM+ERL) versus 47% (PBO+ERL).

130 Similar treatment benefits (ORR, DOR, PFS2, and TTCT) were observed in RAM+ERL–treated

131 patients with ex19del and ex21L858R. Baseline TP53 co-mutation was associated with superior

132 outcomes for RAM+ERL in both ex19del and ex21L858R subgroups. EGFR T790M mutation

133 rate at progression was similar between treatment arms and by mutation type.

134 Conclusions: RAM+ERL provided significant clinical benefit for both EGFR ex19del and

135 ex21L858R NSCLC, supporting this regimen as suitable for patients with either of these EGFR

136 mutation types.

137

138 Introduction

139 Non-small cell lung cancer (NSCLC) is a molecularly heterogeneous disease. Approximately 40-

140 60% of East Asian patients and 10-20% of Caucasian patients with NSCLC harbor a somatic

141 driver mutation in the epidermal growth factor receptor (EGFR) gene (EGFRm) (1,2).

142 Approximately 90% of EGFR-activating mutations involve deletions in exon 19 (ex19del) or a

143 substitution mutation in exon 21, specifically Leu858Arg (ex21L858R) (1,3,4). Although

144 ex19del constitutes nearly 50% and ex21L858R 40% of EGFR-activating mutations, variation in

145 these rates is observed between regions (1,2). Ex21L858R is more prevalent in Asian countries,

146 particularly in Japan (5,6).

147

148 EGFR-activating mutations are also a major predictive factor for response to small-molecule

149 EGFR tyrosine kinase inhibitor (TKI) therapies. However, the degree of benefit may differ based

150 on the EGFR mutation type, which has been shown to be an independent prognostic marker

151 (7,8). Indeed, while both ex19del and ex21L858R mutations benefit from EGFR-TKI treatment,

152 subgroup analyses of several landmark studies in EGFRm NSCLC, indicate that patients with

153 ex21L858R derive more modest benefit with EGFR-TKI monotherapy than patients with

154 ex19del, as represented by its lower median progression-free survival (PFS), smaller delta PFS

155 and higher PFS HR (Table 1). The difference in PFS outcomes between ex19del and ex21L858R

156 is independent of EGFR-TKI generation used and is observed with first-, second-, and third-

157 generation EGFR-TKIs. As treatment benefit in a given trial can be difficult to understand by

158 analyzing results using a single measure of PFS, the totality of PFS data should be considered.

159 Treatment beneﬁt is most reflectively reported using a combination of various absolute (median

160 and delta PFS, and 1-year PFS rates) and relative (HR) PFS measures, which have

161 complementary roles in fully assessing the results of comparative time-to-event analyses and

162 together provide the most complete insights into the different outcomes between ex19del and

163 ex21L858R (9-11). Meta-analyses and real-world evidence studies have provided further

164 evidence for worse PFS outcomes and lower response rates after EGFR-TKI therapy with

165 ex21L858R compared with ex19del (12-14). The different prognostic and predictive roles for

166 ex19del and ex21L858R have led to the widespread acceptance of these mutations as an essential

167 stratification factor in EGFRm NSCLC clinical trials and thereby enhanced the design and

168 interpretation of these studies (13).

169

170 The potential to further improve patient outcomes by combining an EGFR-TKI with an

171 antiangiogenesis agent is supported by strong preclinical and, more recently, clinical trial

172 evidence (15-20). However, there is a paucity of published literature on how dual EGFR/vascular

173 endothelial growth factor (VEGF) pathway inhibition impacts outcomes by EGFR mutation

174 subtypes. In the randomized phase 3 RELAY trial for the first-line treatment of EGFRm NSCLC,

175 ramucirumab plus erlotinib (RAM+ERL) showed a consistent PFS treatment benefit compared

176 with placebo plus erlotinib (PBO+ERL) among patients with ex19del or ex21L858R (19), with

177 similar median PFS, delta PFS improvement, and HRs (Table 1), and a clear and early

178 separation of Kaplan-Meier PFS curves for both ex19del and ex21L858R patients. With a

179 median PFS of 19.4 months and an absolute PFS improvement of 8.2 months versus single-agent

180 EGFR-TKI, these are among the best reported outcomes to date for a subgroup of patients with

181 ex21L858R, albeit in a population without central nervous system (CNS) metastases.

182

183 As noted above, it is important to explore the context specificity of EGFR mutation types to

184 different therapeutic strategies. To address this, we explored more comprehensively if the EGFR-

185 activating mutation subtypes impact treatment outcomes in RELAY. In addition, potential

186 associations between EGFR mutation type and baseline characteristics, patterns of progression,

187 treatment patterns post–RELAY study treatment discontinuation, and pre-existing co-occurring

188 and treatment-emergent genetic alterations were explored.

189

190 Materials and Methods

191 Study design and patients

192 Study design and patient eligibility were previously detailed (19). Briefly, RELAY is a global,

193 double-blind, placebo-controlled, randomized phase 3 study that enrolled patients with untreated

194 metastatic NSCLC with an EGFR ex19del or ex21L858R mutation as tested locally. Random

195 assignment was stratified according to EGFR mutation subtype (ex19del vs ex21L858R), local

196 EGFR testing method (therascreen®/cobas® vs other PCR/sequencing-based methods), sex

197 (male vs female), and region (East Asia vs other). Randomized patients (1:1) received either

198 intravenous ramucirumab (10 mg/kg) or matching placebo every 2 weeks, with erlotinib (150 mg

199 oral) daily. Study treatment continued until radiographic progression as assessed by the

200 investigator according to Response Evaluation Criteria in Solid Tumours (RECIST) v1.1,

201 unacceptable toxicity, withdrawal of consent, noncompliance, or investigator decision. Use of

202 subsequent treatment was at the discretion of the investigator. The trial adhered to the

203 Declaration of Helsinki, the International Council for Harmonisation Guideline for Good Clinical

204 Practice, and applicable local regulations. All patients provided written informed consent. The

205 trial is registered at ClinicalTrials.gov (identifier: NCT02411448).

206 Assessments

207 Tumor assessments were conducted using RECIST v1.1 at baseline, every 6 weeks for 72 weeks,

208 every 12 weeks thereafter until disease progression or unacceptable toxicity, and at the 30-day

209 follow-up visit. Adverse events (AEs) were assessed at every cycle and graded according to the

210 National Cancer Institute Common Terminology Criteria for Adverse Events (v4.0).

211

212 Confirmatory central testing for ex19del and ex21L858R was conducted on archival tissue

213 samples collected at baseline and using the commercial therascreen assay. Central testing did not

214 inform eligibility or study enrollment. Plasma samples were prospectively collected at baseline,

215 Day 1 Cycle 4, and at the 30-day post–study treatment discontinuation follow-up visit, and were

216 evaluated with Guardant360 next-generation sequencing (NGS) (Guardant Health, Redwood

217 City, CA, USA).

218

219 Analysis populations and study endpoints

220 Efficacy analyses were conducted in the population of randomly assigned patients. Safety

221 analyses were evaluated in patients who received at least one dose of study drug. NGS analyses

222 were conducted in patients with valid baseline and/or 30-day follow-up samples depending on

223 whether baseline, treatment-emergent, or post-progression follow-up mutation profiles were

224 being investigated.

225

226 The primary endpoint was PFS (investigator-assessed). Secondary endpoints included safety and

227 tolerability, response rates, duration of response (DOR), and interim overall survival (iOS).

228 Prespecified PFS 2 and post hoc analyses of patterns of progression, time-to-response, time-to-

229 chemotherapy (TTCT), and NGS biomarker analyses were exploratory (see Supplementary

230 Material for definitions of clinical endpoints). Other protocol-specified endpoints are not

231 presented here (21).

232

233 Statistical analyses

234 Although EGFR mutation subtype was a prespecified subgroup, RELAY was not powered for

235 analysis of this or any other subgroup. HR and 95% confidence intervals (CI) were estimated

236 using an unstratified Cox proportional hazards model. Kaplan-Meier estimation was used to plot

237 time-to-event data, as well as to provide summary statistics. Differences between response rates

238 and associated 95% CIs for treatment arms by mutation subgroup were calculated using Fisher’s

239 exact method. Descriptive summary statistics were used for safety analyses. Sankey diagrams

240 were constructed to summarize post–study treatment discontinuation therapy across both arms

241 and to explore any treatment patterns observed. Relationships between somatic gene alterations

242 and EGFR-activating mutation subgroup on PFS were explored using interaction tests from the

243 Cox regression model. Fisher’s exact method compared mutation frequencies across treatment

244 arms within each EGFR mutation subgroup. Additional methodological details for the

245 exploratory biomarker analyses are provided in the Supplementary Material. SAS v9.4 and SAS

246 VA v7.3 (SAS Institute, Cary, NC, USA) were used for all statistical analyses.

247

248 Results

249 Patients, EGFR testing, and baseline characteristics

250 In RELAY, 449 patients (intention-to-treat population) were randomized between January 28,

251 2016 and February 1, 2018 to either RAM+ERL or PBO+ERL treatment arms (Supplementary

252 Fig. S1). Two patients were excluded from this analysis as they were subsequently shown not to

253 harbor an ex19del or ex21L858R mutation. Thus, the total population for the efficacy analyses

254 by EGFR mutation subgroups was 447 patients (243 [54%] ex19del and 204 [46%] ex21L858R).

255 The safety analyses excluded three of the 447 patients who did not receive any study drug (one

256 ex19del and two ex21L858R patients); all were assigned to treatment with RAM+ERL. As of

257 primary data cutoff (January 23, 2019), 63 ex19del (59% RAM+ERL, 41% PBO+ERL) and 44

258 ex21L858R (61% RAM+ERL, 39% PBO+ERL) patients remained on study treatment

259 (Supplementary Fig. S1).

260

261 Central tissue testing results corroborated the local EGFR testing results, with 97% of centrally

262 tested patients (305 of 316 evaluable results) having an EGFR-activating mutation detected (53%

263 ex19del and 44% ex21L858R) (19). Additionally, 324 patients had evaluable baseline plasma

264 NGS results, with ex19del and ex21L858R mutations present at equivalent rates (50%)

265 (Supplementary Fig. S1). Unless otherwise indicated, subsequent analyses for this report used

266 baseline EGFR-activating mutation status as determined by local testing.

267

268 The majority of patients in each EGFR mutation subgroup, regardless of treatment arm, were

269 female, never smokers, had similar median age, and an Eastern Cooperative Oncology Group

270 performance status of 0 (Table 2). Ex21L858R was more prevalent than ex19del among patients

271 of Asian race (83% vs 72%) and females (67% vs 60%); bone metastases were more common in

272 patients with ex21L858R than with ex19del (34% vs 26%). Other baseline patient and disease

273 characteristics were similar between EGFR mutation subgroups.

274

275 Efficacy

276 With a median follow-up of 20.7 months (range, 0.1–35.4), PFS treatment benefit with

277 RAM+ERL was similar between EGFR mutation subgroups (Table 1) (19). Improvements in

278 median PFS among patients receiving RAM+ERL versus PBO+ERL were 7.1 and 8.2 months

279 for ex19del and ex21L858R subgroups, respectively. One-year PFS rates for ex19del patients

280 were 74% (95% CI, 64.1–80.8) for RAM+ERL versus 54% (95% CI, 43.9–62.2) for PBO+ERL,

281 and for ex21L858R patients were 70% (95% CI, 58.9–77.9) for RAM+ERL versus 47% (95%

282 CI, 37.0–57.1) for PBO+ERL (Fig. 1A and Table 1) (19).

283

284 Irrespective of treatment arm, stable disease (SD) was reported more frequently in ex21L858R

285 patients (25%) than ex19del patients (15%), while partial response (PR) was more prevalent in

286 ex19del patients (79%) than in ex21L858R patients (69%). Overall response rate (ORR) was

287 lower in ex21L858R patients (70%) than ex19del patients (81%), with the lowest ORR observed

288 in the ex21L858R subgroup treated with PBO+ERL (ex19del: RAM+ERL 79%, PBO+ERL

289 83%; ex21L858R: RAM+ERL 74%, PBO+ERL 66%) (Table 3 and Supplementary Fig. S2).

290 Disease control rate was high (>90%) and similar in the ex19del and ex21L858R subgroups

291 (Table 3 and Supplementary Fig. S2). The median time-to-response for patients was similar

292 (1.4–1.5 months) across mutation subgroups and treatment arms (Table 3).

293

294 DOR analysis included fewer ex21L858R patients as SD was more prevalent in this subgroup.

295 For both mutation subgroups, DOR favored the RAM+ERL versus the PBO+ERL arm (Fig. 1B).

296 Relative to the ex21L858R subgroup, the ex19del subgroup treated with RAM+ERL exhibited a

297 numerically longer median DOR (18.2 [95% CI, 13.8–20.9] vs 16.2 [95% CI, 12.5–20.1]

298 months).

299

300 The iOS data were immature as censoring rates were high (overall censoring rate=82%) (Table

301 3). The highest rate of death was observed in the ex21L858R PBO+ERL treatment arm (ex19del:

302 RAM+ERL 17% [21 deaths], PBO+ERL 13% [15]; ex21L858R: RAM+ERL 16% [16],

303 PBO+ERL 26% [27]). Two-year OS rates were lowest for ex21L858R patients in the PBO+ERL

304 treatment arm, while RAM+ERL–treated patients with ex21L858R reported a comparable 2-year

305 OS rate to patients with ex19del (Table 3).

306

307 To further evaluate the impact of treatment by baseline EGFR mutation subgroups on other post–

308 study treatment discontinuation endpoints, PFS2 and TTCT were explored. PFS2 data were

309 immature (overall censoring rate=69%), with the lowest censoring rate observed for ex21L858R

310 patients receiving PBO+ERL (58%), indicating that most PFS2 events occurred in this subgroup.

311 Ex21L858R patients receiving RAM+ERL continued to exhibit improved benefit through PFS2

312 compared with those receiving PBO+ERL (HR, 0.600 [95% CI, 0.371–0.969]) (Fig. 1C). The

313 ex19del subgroup had similar PFS2 between treatment arms (HR, 0.926 [95% CI, 0.577–1.485]).

314 For TTCT, a treatment benefit for RAM+ERL versus PBO+ERL patients was observed for the

315 ex21L858R subgroup (HR, 0.554 [95% CI, 0.352–0.872]; censoring rate=61%) (Fig. 1D). No

316 difference between treatment arms was observed for the ex19del subgroup (HR, 1.034 [95% CI,

317 0.682–1.567]; censoring rate=63%).

318

319 Further analyses of PFS2 and TTCT both indicated a trend toward predictive effect regardless of

320 mutation type. This trend was preserved upon adjusting for other factors via a multivariate model

321 (interaction P-values: PFS2=0.2100; TTCT=0.0443).

322

323 Exposure

324 Of those patients in the ex19del subgroup, 122 received RAM+ERL and 120 PBO+ERL, and of

325 those in the ex21L858R subgroup, 97 received RAM+ERL and 105 PBO+ERL (Table 3). In the

326 ex19del subgroup, among patients receiving RAM+ERL, median (minimum–maximum)

327 duration of exposure to ramucirumab (excluding 37 patients still on treatment) was 12.4 (0.5–

328 33.8) months and to erlotinib 15.2 (0.1–33.8) months. Among patients receiving PBO+ERL,

329 median duration of exposure to placebo (excluding 26 patients still on treatment) was 11.0 (0.5–

330 35.4) months and to erlotinib 12.0 (0.4–35.5) months (Table 3). In the ex21L858R subgroup,

331 among patients receiving RAM+ERL, the median duration of exposure to ramucirumab

332 (excluding 27 patients still on treatment) was 12.7 (0.5–33.2) months and to erlotinib 15.1 (0.0–

333 33.0) months. Among patients receiving PBO+ERL, the median duration of exposure to placebo

334 (excluding 27 patients still on treatment) was 8.8 (0.5–35.0) months and to erlotinib 10.4 (0.8–

335 35.1) months (Table 3).

336

337 Safety

338 All patients reported at least one treatment-emergent AE (TEAE) (Supplementary Table S1).

339 Overall, there were no clinically meaningful differences in safety findings between the mutation

340 subgroups, although, of note, any grade hypertension was reported with a ≥10% higher incidence

341 in the ex21L858R subgroup than the ex19del subgroup (50 patients [41%] vs 50 patients [52%],

342 respectively) (Supplementary Table S2). Incidence of Grade 3 or higher TEAEs and serious

343 AEs (SAEs) was similar between ex19del and ex21L858R subgroups (Supplementary Table

344 S1). Study treatment discontinuation rates due to AEs and SAEs were also similar between

345 EGFR mutation subgroups (AEs: RAM+ERL 13% and 13% and PBO+ERL 8% and 14%,

346 ex19del vs ex21L858R, respectively; SAEs: RAM+ERL 3% and 6% and PBO+ERL 4% and 4%,

347 ex19del vs ex21L858R, respectively). Death due to an AE while on study treatment occurred in

348 two RAM+ERL–treated patients with an ex21L858R mutation (one death due to hemothorax

349 was considered related, and the other death due to encephalitis influenza was considered not

350 related to study treatment) (Supplementary Table S1).

351

352 Clinical patterns of progression

353 Overall, frequencies of disease progression were similar by treatment regardless of mutation

354 type, with rates being lower for the RAM+ERL combination (ex19del: RAM+ERL 46%,

355 PBO+ERL 65%; ex21L858R: RAM+ERL 49%, PBO+ERL 64%) (Supplementary Fig. S1).

356 The majority of sites of progression were in the lung, followed by pleura and lymph nodes,

357 regardless of treatment arm and EGFR mutation subgroup (Supplementary Fig. S3). CNS

358 metastasis developed more frequently in patients with ex21L858R (six patients in the

359 ex21L858R subgroup [two patients and four patients in the RAM+ERL and PBO+ERL arm,

360 respectively] and one patient in the ex19del subgroup [PBO+ERL arm]). The majority of patients

361 had single-site progression; this was reported more frequently in ex21L858R patients (102

362 patients, 79%) compared with ex19del patients (91 patients, 62%). This difference was mostly

363 driven by patients in the ex21L858R RAM+ERL treatment arm (Supplementary Fig. S4). Local

364 progression, defined as increase in size of a single pre-existing tumor lesion, occurred in

365 approximately 50% of patients across treatment arms and EGFR mutation subgroups. Metastatic

366 progression, defined as development of at least one new lesion, occurred in 52 patients (36%)

367 versus 53 patients (41%) in the ex19del versus ex21L858R subgroups, respectively

368 (Supplementary Fig. S4).

369

370 Subsequent therapies

371 All study treatment was to be discontinued upon RECIST v1.1–defined disease progression. At

372 time of data cutoff, 37 (30%) and 26 (22%) ex19del patients and 27 (27%) and 17 (16%)

373 ex21L858R patients (RAM+ERL vs PBO+ERL, respectively) were still on study treatment. First

374 subsequent therapy (FST) was received by 66 (54%) and 83 (69%) ex19del patients and 54

375 (55%) and 73 (70%) ex21L858R patients (RAM+ERL vs PBO+ERL, respectively)

376 (Supplementary Table S3). Sankey diagrams depict the flow and patterns of subsequent lines of

377 treatment (Supplementary Fig. S5). For both EGFR mutation groups, the most commonly used

378 FST was an EGFR-TKI, predominantly erlotinib, and to a lesser degree, osimertinib, followed by

379 platinum-based chemotherapy. Ex19del patients treated with RAM+ERL received erlotinib as

380 FST (52%) more frequently compared with PBO+ERL patients (29%); for ex21L858R patients,

381 erlotinib as FST was received at similar rates in RAM+ERL (50%) and PBO+ERL patients

382 (42%). Osimertinib was used more frequently as FST in ex19del patients treated with PBO+ERL

383 (31%) compared with RAM+ERL (14%); for ex21L858R patients, osimertinib was received at

384 similar rates in RAM+ERL (17%) and PBO+ERL patients (12%). In ex21L858R, chemotherapy

385 was used as FST at similar rates in patients treated with RAM+ERL and PBO+ERL (43% vs

386 47%), whereas more patients treated with PBO+ERL than RAM+ERL received chemotherapy as

387 second subsequent therapy (SST) (66% vs 42%).

388 At time of data cutoff, approximately 35% of patients were still on FST. Nearly one-third of

389 patients who received osimertinib as FST remained on this therapy in both ex19del and

390 ex21L858R subgroups, regardless of prior study treatment (Supplementary Table S3). As SST,

391 osimertinib was most frequently used, regardless of prior study treatment: 29 patients (39%)

392 versus 16 patients (25%) in the ex19del versus ex21L858R subgroups, respectively.

393

394 Exploratory biomarker analyses

395 NGS testing was performed to examine the somatic alteration landscape by ex19del or

396 ex21L858R subgroups to gain insights on potential biological differences at baseline and

397 mechanisms of resistance. In 320 of the patients with at least one detectable somatic gene

398 alteration, EGFR-activating mutations (ex19del: 134 [41.9%]; ex21L858R: 121 [37.8%]) were

399 detected at baseline by NGS (this was 100% concordant with local EGFR testing results); 10%

400 of ex19del and 13% of ex21L858R patients did not have any additional somatic gene alteration.

401

402 Concomitant gene alterations were widespread in both ex19del and ex21L858R mutated NSCLC

403 patients. The most frequent genes with co-occurring variants at baseline were TP53 and

404 additional EGFR variants (Supplementary Table S4), each with similar prevalence regardless

405 of EGFR-activating mutation subgroup (Fig. 2A). PIK3CA was higher in ex19del patients and

406 was the only gene with co-occurring variants having differential prevalence (>5%) by EGFR-

407 activating mutation subgroup. In patients with TP53 co-mutation, RAM+ERL demonstrated a

408 superior PFS treatment benefit versus PBO+ERL in both ex19del and ex21L858R subgroups

409 (Supplementary Fig. S6). In TP53 wild-type patients, the benefit of RAM+ERL on PFS

410 appeared more pronounced in the ex21L858R subgroup than the ex19del subgroup.

411 Treatment-emergent gene alterations co-occurring with either ex19del or ex21L858R mutations

412 are presented in Fig. 2B. EGFR T790M was the most prevalent treatment-emergent gene

413 alteration in both EGFR mutation subgroups. Other EGFR variants and TP53 were the second

414 most prevalent co-mutated genes at disease progression, and prevalence varied by EGFR

415 mutation subgroup and treatment arm. Other treatment-emergent EGFR variants were observed

416 at highest prevalence (29.2%) in ex21L858R PBO+ERL–treated patients, and TP53 variants

417 were highest (30.3%) in ex19del RAM+ERL–treated patients. The only genes with co-occurring

418 variants having differential prevalence (>10%) by EGFR-activating mutation subgroup treatment

419 arm, were KRAS (higher with RAM+ERL in ex19del patients) and PIK3CA (higher with

420 PBO+ERL in ex21L858R patients).

421

422 T790M rates at 30-day follow-up were similar across the two mutation subgroups and treatment

423 arms (Supplementary Fig. S7). Among ex19del patients, emergence of T790M appeared to be

424 delayed in those receiving RAM+ERL. T790M was only observed in ex19del patients receiving

425 RAM+ERL who progressed after completing at least 12 cycles of treatment compared with

426 PBO+ERL, where emergence was seen in patients who progressed after at least four cycles of

427 treatment. In ex21L858R patients, T790M appeared to emerge in patients after at least four

428 cycles of treatment, independent of treatment received. Among PBO+ERL–treated patients with

429 post-progression TP53 detected at the 30-day follow-up, newly emergent TP53 alterations were

430 detected as early as four cycles of treatment, independent of ex19del or ex21L858R

431 (Supplementary Fig. S8). For RAM+ERL–treated patients, emergence of TP53 was observed in

432 patients who completed at least 4 cycles for ex19del, and at least 12 cycles of treatment for

433 ex21L858R.

434

435 Discussion

436 The expanding armamentarium of effective treatments for EGFR-mutated NSCLC highlights the

437 importance of fully understanding each treatment regimen so that therapy can be tailored to

438 individual patient and disease characteristics. There is a growing body of evidence that patients

439 with an ex21L858R mutation have a poorer prognosis and response to treatment with EGFR-

440 TKIs than patients with ex19del mutation.

441

442 In RELAY, RAM+ERL showed consistent treatment benefit among patients with ex19del or

443 ex21L858R, while treatment outcomes were worse with ex21L858R than ex19del in those

444 receiving PBO+ERL. Evidence suggests that different subtypes of EGFR mutation may vary in

445 their clinical and pathological correlations (22,23). In RELAY, patient and disease characteristics

446 at baseline, including NGS determined co-mutations, were similar between EGFR mutation

447 subgroups with the exception of ex21L858R being more prevalent among females and Asian

448 patients. So, the subgroup of ex21L858R patients did not appear to be enriched with more poor

449 prognostic features than the subgroup of ex19del patients.

450

451 Patterns of progression were reported at similar rates between the two mutation subgroups, with

452 the exception of CNS metastases being more frequently reported as site of first progression in

453 patients with ex21L858R. This could indicate a more aggressive phenotype of ex21L858R

454 tumors, although the small number of events limits the strength of this observation.

455

456 The magnitude of PFS benefit with RAM+ERL observed for patients with ex21L858R was

457 similar to that of RAM+ERL–treated patients with ex19del (19). RAM+ERL–treated patients

458 with ex21L858R also had treatment benefits in ORR, DOR, PFS2, TTCT, and iOS; the worst

459 outcomes were observed in PBO+ERL–treated patients with ex21L858R. Analyses evaluating

460 TTCT showed that, despite high censoring, use of chemotherapy was significantly delayed with

461 RAM+ERL versus PBO+ERL in ex21L858R patients but not ex19del patients. Subsequent

462 therapies received and presence of treatment-emergent T790M and/or TP53 mutations may have

463 had a role in the TTCT results observed. Although a predictive effect of mutation type on PFS

464 was not apparent, further analyses of PFS2 and TTCT both indicated a trend towards predictive

465 effect regardless of mutation type on benefit of RAM+ERL. Given limited event counts and data

466 immaturity for post-progression outcomes and ex19del being more common in Western patients,

467 who were underrepresented in RELAY (28% were Causcasian), the results are to be interpreted

468 with caution. In general, there were no meaningful differences in the safety profiles of

469 RAM+ERL by activating mutation subgroups.

470

471 The underlying reasons for the differential efficacy between ex19del and ex21L858R mutation

472 subtypes being treated with EGFR-TKI monotherapy are not fully understood (12). One

473 suggestion is that ex21L858R is less efficiently inhibited by EGFR-TKIs than ex19del.

474 In vitro studies have shown that gefitinib inhibited EGFR phosphorylation, a measure of kinase

475 inhibitor potency, to a lesser degree in cancer cells with ex21L858R than with ex19del mutation

476 (24), whereas kinetic analysis has demonstrated ex21L858R to be less sensitive than ex19del to

477 erlotinib inhibition (25). These findings offer some rationale for the difference in outcomes

478 between the EGFR mutation subtypes when treated with single agent EGFR-TKI.

479

480 Interestingly, a consistent PFS benefit between ex19del and ex21L858R groups has also been

481 reported in patients receiving erlotinib and bevacizumab (Table 1) (16-18), and meta-analyses of

482 EGFR-TKI and antiangiogenic agent combinations (26,27), suggesting that addition of an

483 antiangiogenic agent appears to abrogate the differential efficacy observed in ex21L858R

484 compared with ex19del patients when treated with EGFR-TKI monotherapy. Unfortunately, a

485 comprehensive understanding of the underlying biological mechanisms that could support the

486 clinical observations with ex19del and ex21L858R mutations is currently lacking, and so it is

487 unknown if this is related to dual EGFR/VEGF pathway inhibition. Based on the observation that

488 ex21L858R patients receiving high-dose EGFR-TKI treatment have improved outcomes

489 compared to those receiving standard dose EGFR-TKI (28), it could be hypothesized that

490 addition of an antiangiogenic agent might lead to increased and sustained high concentrations

491 and uptake of EGFR-TKIs by normalizing blood flow through tumor blood vessels (25,29,30).

492 Further preclinical studies are needed to elucidate biological differences between ex19del and

493 ex21L858R mutations.

494

495 The wide range of pre-existing co-occurring and treatment-emergent genetic alterations

496 underscore the fact that EGFR-mutant lung cancer is a heterogeneous disease, which might

497 contribute to the differential sensitivity and varied responses achieved upon EGFR-TKI

498 treatment. Concurrent TP53 mutations have been shown to be a negative prognostic factor and

499 associated with poorer outcomes in patients treated with EGFR-TKIs (31). A role for TP53 in

500 angiogenesis has also been established, and multiple studies have suggested that presence of a

501 TP53 mutation is associated with VEGF pathway up-regulation and may be predictive of clinical

502 sensitivity to antiangiogenic therapies in several tumor types (32-35). In RELAY, the prevalence

503 of TP53 mutations at baseline (43% vs 43%) and at time of disease progression (17% vs 15%)

504 was similar in ex19del and ex21L858R subgroups. Baseline TP53 mutation was associated with

505 superior outcomes for RAM+ERL in both ex19del and ex21L858R subgroups, while in TP53

506 wild-type patients, the benefit of RAM+ERL appeared more pronounced in the ex21L858R than

507 the ex19del subgroup. At disease progression, treatment-emergent TP53 variants were most

508 prevalent in ex19del RAM+ERL–treated patients, and additional experimentation is needed to

509 determine the relevance and if it is related to mechanisms of acquired resistance.

510

511 Further, treatment-emergent T790M as a mechanism of resistance is reported to occur more

512 frequently in ex19del than ex21L858R patients (36,37). In RELAY, no T790M was identified in

513 circulating tumor DNA plasma samples at baseline; however, T790M was, as expected, the most

514 prevalent genetic alteration post–study treatment discontinuation. EGFR T790M mutation rate at

515 progression was similar between treatment arms and by mutation type (Supplementary Fig. S1),

516 suggesting that addition of ramucirumab to erlotinib does not alter the T790M resistance

517 mechanism pathway. In ex19del patients, the emergence of T790M may be delayed in those

518 receiving RAM+ERL, whereas this was not observed in ex21L858R patients. However, as

519 median PFS times were similar for RAM+ERL–treated ex19del and ex21L858R patients, the

520 difference in the timing of T790M emergence does not appear to contribute to the enhanced

521 benefit observed in ex21L858R patients upon addition of ramucirumab to erlotinib treatment.

522 Subsequent treatment with an agent that targets the EGFR T790M mutation, such as osimertinib,

523 could further delay disease progression and TTCT for the considerable proportion of patients

524 who acquire the EGFR T790M mutation. Indeed, in this study, osimertinib was used as post-

525 discontinuation therapy across subsequent lines of therapy (Supplementary Table S3).

526 We acknowledge a number of limitations in our study. RELAY was not sufficiently powered for

527 subgroup analyses of ex19del and ex21L858R, although these were prespecified, and a

528 randomization factor. In addition, results of the exploratory biomarker analyses should be viewed

529 with caution because of low patient numbers at varying time points. All NGS analyses were

530 performed on plasma samples only. As a consequence, some important mechanisms of

531 resistance, such as histological transformation and copy number alterations (which are more

532 difficult to capture in plasma) have not been explored optimally. Differences in prevalence and

533 timing of treatment-emergent mutations between treatment arms (eg, delay in emergence of

534 T790M or TP53) are possibly biased toward early progressors, as patients for whom no post-

535 discontinuation plasma is available yet are those still benefiting from treatment. The final NGS

536 analysis at time of OS maturity should provide a clearer understanding of the impact of

537 treatment-emergent alterations.

538

539 Analyses of the EGFR mutation subgroups in RELAY is for the purposes of hypothesis

540 generation and reflects an attempt to better understand the heterogeneity of the disease. We

541 believe that understanding the potential impact of differences between ex19del and ex21L858R

542 patients is clinically relevant and could have several potentially significant clinical and research

543 implications. Findings will be useful for counseling patients and could help inform treatment

544 decisions. Our data demonstrate that ex19del and ex21L8585R mutations have different

545 prognostic and predictive roles dependent on the therapy used and are therefore an important

546 stratification factor in clinical trials. Further drug development, particularly for tumors with

547 ex21L858R that have a poorer response to EGFR-TKI monotherapy, remains important. Further

548 prospective studies are warranted to confirm the role of combined EGFR-TKI and antiangiogenic

549 treatment in patients with ex21L858R mutations. The recently started WJOG Phase 3 study of

550 erlotinib plus ramucirumab vs osimertinib in advanced NSCLC with EGFR ex21L858R mutation

551 (REVOL858R; jRCTs051200142) will help address this question.

552

553 In summary, RELAY demonstrates that RAM+ERL has significant clinical benefit for both

554 EGFR ex19del and ex21L858R NSCLC, supporting this regimen as suitable for patients with

555 either of these EGFR mutation types. However, as ex21L858R-positive patients historically have

556 poorer outcomes relative to ex19del in response to EGFR-TKI monotherapies, this subgroup of

557 EGFR-mutant NSCLC may be a higher priority for addition of an antiangiogenic to an EGFR-

558 TKI. Our findings provide evidence that patients with EGFR ex19del and ex21L8585R

559 mutations do not have a heterogeneous response to EGFR-TKI targeted therapy when combined

560 with ramucirumab. These findings may help clinicians provide a personalized treatment strategy

561 and allow treatment decisions to be made on a more rational basis.

562

563 Authors’ Contributions

564 RELAY study design: K. Nakagawa, E. Garon, M. Reck, A.H. Zimmerman

565 Patient care, data collection: K. Nakagawa, E. Nadal, E. Garon, M. Nishio, T. Seto, N.

566 Yamamoto, K. Park, J.-Y. Shih, M. Reck

567 Statistical analyses: A.H. Zimmerman, S. Wijayawardana

568 Drafting and/or critical review of the manuscript: All authors

569 Final approval of the manuscript: All authors

570 Accountable for all aspects of the work: All authors

571

572 Acknowledgments

573 The RELAY clinical trial was sponsored and funded by Eli Lilly and Company. The authors are

574 grateful to the physicians, clinical staff, caregivers, and especially patients for their invaluable

575 contributions to this work. Medical writing assistance was provided by Susan P. Whitman, a full-

576 time employee of Eli Lilly and Company, and John Bilbruck of ProScribe – Envision Pharma

577 Group, and was funded by Eli Lilly. Shawn Estrem, a full-time employee of Eli Lilly and

578 Company, provided biomarker analyses and interpretations. Malika Mahoui, Hilary Graham,

579 Olga N. Lipkovich, and Donghui Li, full-time employees of Eli Lilly and Company, provided

580 statistical analyses and Sankey diagrams. The contributing RELAY investigators are listed in the

581 Supplementary Material.

582

583 Data Sharing Statement

584 Lilly provides access to all individual participant data collected during the trial, after

585 anonymization, with the exception of pharmacokinetic or genetic data. Data are available to

586 request 6 months after the indication studied has been approved in the US and EU and after

587 primary publication acceptance, whichever is later. No expiration date of data requests is

588 currently set once data are made available. Access is provided after a proposal has been approved

589 by an independent review committee identified for this purpose and after receipt of a signed data

590 sharing agreement. Data and documents, including the study protocol, statistical analysis plan,

591 clinical study report, blank or annotated case report forms, will be provided in a secure data

592 sharing environment. For details on submitting a request, see the instructions provided at

593 www.vivli.org.

594

595

596 References

597 1. Hsu WH, Yang JC, Mok TS, Loong HH. Overview of current systemic management of EGFR- 598 mutant NSCLC. Ann Oncol 2018;29(suppl_1):i3-i9 doi 10.1093/annonc/mdx702. 599 2. Midha A, Dearden S, McCormack R. EGFR mutation incidence in non-small-cell lung cancer of 600 adenocarcinoma histology: a systematic review and global map by ethnicity (mutMapII). Am J 601 Cancer Res 2015;5(9):2892-911. 602 3. Passaro A, Prelaj A, Bonanno L, Tiseo M, Tuzi A, Proto C, et al. Activity of EGFR TKIs in Caucasian 603 patients with NSCLC harboring potentially sensitive uncommon EGFR mutations. Clin Lung 604 Cancer 2019;20(2):e186-e94 doi 10.1016/j.cllc.2018.11.005. 605 4. Murray S, Dahabreh IJ, Linardou H, Manoloukos M, Bafaloukos D, Kosmidis P. Somatic mutations 606 of the tyrosine kinase domain of epidermal growth factor receptor and tyrosine kinase inhibitor 607 response to TKIs in non-small cell lung cancer: an analytical database. J Thorac Oncol 608 2008;3(8):832-9 doi 10.1097/JTO.0b013e31818071f3. 609 5. Zhang YL, Yuan JQ, Wang KF, Fu XH, Han XR, Threapleton D, et al. The prevalence of EGFR 610 mutation in patients with non-small cell lung cancer: a systematic review and meta-analysis. 611 Oncotarget 2016;7(48):78985-93 doi 10.18632/oncotarget.12587. 612 6. Graham RP, Treece AL, Lindeman NI, Vasalos P, Shan M, Jennings LJ, et al. Worldwide frequency 613 of commonly detected EGFR mutations. Arch Pathol Lab Med 2018;142(2):163-7 doi 614 10.5858/arpa.2016-0579-CP. 615 7. Hong W, Wu Q, Zhang J, Zhou Y. Prognostic value of EGFR 19-del and 21-L858R mutations in 616 patients with non-small cell lung cancer. Oncol Lett 2019;18(4):3887-95 doi 617 10.3892/ol.2019.10715. 618 8. Jiang H, Zhu M, Li Y, Li Q. Association between EGFR exon 19 or exon 21 mutations and survival 619 rates after first-line EGFR-TKI treatment in patients with non-small cell lung cancer. Mol Clin 620 Oncol 2019;11(3):301-8 doi 10.3892/mco.2019.1881. 621 9. Saad ED, Zalcberg JR, Peron J, Coart E, Burzykowski T, Buyse M. Understanding and 622 Communicating Measures of Treatment Effect on Survival: Can We Do Better? J Natl Cancer Inst 623 2018;110(3):232-40 doi 10.1093/jnci/djx179. 624 10. Sashegyi A, Ferry D. On the Interpretation of the Hazard Ratio and Communication of Survival 625 Benefit. Oncologist 2017;22(4):484-6 doi 10.1634/theoncologist.2016-0198. 626 11. Schnipper LE, Davidson NE, Wollins DS, Blayney DW, Dicker AP, Ganz PA, et al. Updating the 627 American Society of Clinical Oncology Value Framework: Revisions and Reflections in Response 628 to Comments Received. J Clin Oncol 2016;34(24):2925-34 doi 10.1200/JCO.2016.68.2518. 629 12. Sheng M, Wang F, Zhao Y, Li S, Wang X, Shou T, et al. Comparison of clinical outcomes of 630 patients with non-small-cell lung cancer harbouring epidermal growth factor receptor exon 19 631 or exon 21 mutations after tyrosine kinase inhibitors treatment: a meta-analysis. Eur J Clin 632 Pharmacol 2016;72(1):1-11 doi 10.1007/s00228-015-1966-0. 633 13. Zhang Y, Sheng J, Kang S, Fang W, Yan Y, Hu Z, et al. Patients with exon 19 deletion were 634 associated with longer progression-free survival compared to those with L858R mutation after 635 first-line EGFR-TKIs for advanced non-small cell lung cancer: a meta-analysis. PLoS One 636 2014;9(9):e107161 doi 10.1371/journal.pone.0107161. 637 14. Arriola E, Garcia Gómez R, Diz P, Majem M, Martinez Aguillo M, Valdivia J, et al. Clinical 638 management and outcome of patients with advanced NSCLC carrying EGFR mutations in Spain. 639 BMC Cancer 2018;18(1):106 doi 10.1186/s12885-018-4004-7. 640 15. Byers LA, Heymach JV. Dual targeting of the vascular endothelial growth factor and epidermal 641 growth factor receptor pathways: rationale and clinical applications for non-small-cell lung 642 cancer. Clin Lung Cancer 2007;8 Suppl 2:S79-85 doi 10.3816/clc.2007.s.006.

643 16. Seto T, Kato T, Nishio M, Goto K, Atagi S, Hosomi Y, et al. Erlotinib alone or with bevacizumab as 644 first-line therapy in patients with advanced non-squamous non-small-cell lung cancer 645 harbouring EGFR mutations (JO25567): an open-label, randomised, multicentre, phase 2 study. 646 Lancet Oncol 2014;15(11):1236-44 doi 10.1016/S1470-2045(14)70381-X. 647 17. Saito H, Fukuhara T, Furuya N, Watanabe K, Sugawara S, Iwasawa S, et al. Erlotinib plus 648 bevacizumab versus erlotinib alone in patients with EGFR-positive advanced non-squamous non- 649 small-cell lung cancer (NEJ026): interim analysis of an open-label, randomised, multicentre, 650 phase 3 trial. Lancet Oncol 2019;20(5):625-35 doi 10.1016/S1470-2045(19)30035-X. 651 18. Zhou Q, Wu Y-L, Cheng Y, Liu Y, Chen G, Cui J, et al. CTONG 1509: Phase III study of bevacizumab 652 with or without erlotinib in untreated Chinese patients with advanced EGFR-mutated NSCLC. 653 Ann Oncol 2019;30:v603. 654 19. Nakagawa K, Garon EB, Seto T, Nishio M, Ponce Aix S, Paz-Ares L, et al. Ramucirumab plus 655 erlotinib in patients with untreated, EGFR-mutated, advanced non-small-cell lung cancer 656 (RELAY): a randomised, double-blind, placebo-controlled, phase 3 trial. Lancet Oncol 657 2019;20(12):1655-69 doi 10.1016/S1470-2045(19)30634-5. 658 20. Le X, Nilsson M, Goldman J, Reck M, Nakagawa K, Kato T, et al. Dual EGFR-VEGF Pathway 659 Inhibition: A Promising Strategy for Patients With EGFR-Mutant NSCLC. J Thorac Oncol 2020 doi 660 10.1016/j.jtho.2020.10.006. 661 21. Thompson JC, Yee SS, Troxel AB, Savitch SL, Fan R, Balli D, et al. Detection of therapeutically 662 targetable driver and resistance mutations in lung cancer patients by next-generation 663 sequencing of cell-free circulating tumor DNA. Clin Cancer Res 2016;22(23):5772-82 doi 664 10.1158/1078-0432.CCR-16-1231. 665 22. Isaka T, Yokose T, Ito H, Nagata M, Furumoto H, Nishii T, et al. Correlations between the EGFR 666 mutation status and clinicopathological features of clinical stage I lung adenocarcinoma. 667 Medicine (Baltimore) 2015;94(42):e1784 doi 10.1097/MD.0000000000001784. 668 23. Zhou X, Cai L, Liu J, Hua X, Zhang Y, Zhao H, et al. Analyzing EGFR mutations and their association 669 with clinicopathological characteristics and prognosis of patients with lung adenocarcinoma. 670 Oncol Lett 2018;16(1):362-70 doi 10.3892/ol.2018.8681. 671 24. Zhu JQ, Zhong WZ, Zhang GC, Li R, Zhang XC, Guo AL, et al. Better survival with EGFR exon 19 672 than exon 21 mutations in gefitinib-treated non-small cell lung cancer patients is due to 673 differential inhibition of downstream signals. Cancer Lett 2008;265(2):307-17 doi 674 10.1016/j.canlet.2008.02.064. 675 25. Carey KD, Garton AJ, Romero MS, Kahler J, Thomson S, Ross S, et al. Kinetic analysis of epidermal 676 growth factor receptor somatic mutant proteins shows increased sensitivity to the epidermal 677 growth factor receptor tyrosine kinase inhibitor, erlotinib. Cancer Res 2006;66(16):8163-71 doi 678 10.1158/0008-5472.CAN-06-0453. 679 26. Deng Z, Qin Y, Liu Y, Zhang Y, Lu Y. Role of Antiangiogenic Agents Combined With EGFR Tyrosine 680 Kinase Inhibitors in Treatment-naive Lung Cancer: A Meta-Analysis. Clin Lung Cancer 681 2021;22(1):e70-e83 doi 10.1016/j.cllc.2020.08.005. 682 27. Chen F, Chen N, Yu Y, Cui J. Efficacy and Safety of Epidermal Growth Factor Receptor (EGFR) 683 Inhibitors Plus Antiangiogenic Agents as First-Line Treatments for Patients With Advanced EGFR- 684 Mutated Non-small Cell Lung Cancer: A Meta-Analysis. Front Oncol 2020;10:904 doi 685 10.3389/fonc.2020.00904. 686 28. Li X, Zhang L, Jiang D, Wang Y, Zang A, Ding C, et al. Routine-Dose and High-Dose Icotinib in 687 Patients with Advanced Non-Small Cell Lung Cancer Harboring EGFR Exon 21-L858R Mutation: 688 the Randomized, Phase II, INCREASE Trial. Clin Cancer Res 2020;26(13):3162-71 doi 689 10.1158/1078-0432.CCR-19-3064.

690 29. Chauhan VP, Stylianopoulos T, Martin JD, Popovic Z, Chen O, Kamoun WS, et al. Normalization of 691 tumour blood vessels improves the delivery of nanomedicines in a size-dependent manner. Nat 692 Nanotechnol 2012;7(6):383-8 doi 10.1038/nnano.2012.45. 693 30. Chatterjee S, Wieczorek C, Schottle J, Siobal M, Hinze Y, Franz T, et al. Transient antiangiogenic 694 treatment improves delivery of cytotoxic compounds and therapeutic outcome in lung cancer. 695 Cancer Res 2014;74(10):2816-24 doi 10.1158/0008-5472.CAN-13-2986. 696 31. Qin K, Hou H, Liang Y, Zhang X. Prognostic value of TP53 concurrent mutations for EGFR- TKIs 697 and ALK-TKIs based targeted therapy in advanced non-small cell lung cancer: a meta-analysis. 698 BMC Cancer 2020;20(1):328 doi 10.1186/s12885-020-06805-5. 699 32. Li AM, Boichard A, Kurzrock R. Mutated TP53 is a marker of increased VEGF expression: analysis 700 of 7,525 pan-cancer tissues. Cancer Biol Ther 2020;21(1):95-100 doi 701 10.1080/15384047.2019.1665956. 702 33. Schwaederlé M, Lazar V, Validire P, Hansson J, Lacroix L, Soria JC, et al. VEGF-A expression 703 correlates with TP53 mutations in non-small cell lung cancer: implications for antiangiogenesis 704 therapy. Cancer Res 2015;75(7):1187-90 doi 10.1158/0008-5472.CAN-14-2305. 705 34. Narendran A, Ganjavi H, Morson N, Connor A, Barlow JW, Keystone E, et al. Mutant p53 in bone 706 marrow stromal cells increases VEGF expression and supports leukemia cell growth. Exp 707 Hematol 2003;31(8):693-701 doi 10.1016/s0301-472x(03)00159-0. 708 35. Wheler JJ, Janku F, Naing A, Li Y, Stephen B, Zinner R, et al. TP53 Alterations Correlate with 709 Response to VEGF/VEGFR Inhibitors: Implications for Targeted Therapeutics. Mol Cancer Ther 710 2016;15(10):2475-85 doi 10.1158/1535-7163.MCT-16-0196. 711 36. Pao W, Miller VA, Politi KA, Riely GJ, Somwar R, Zakowski MF, et al. Acquired resistance of lung 712 adenocarcinomas to gefitinib or erlotinib is associated with a second mutation in the EGFR 713 kinase domain. PLoS Med 2005;2(3):e73 doi 10.1371/journal.pmed.0020073. 714 37. Kobayashi S, Boggon TJ, Dayaram T, Janne PA, Kocher O, Meyerson M, et al. EGFR mutation and 715 resistance of non-small-cell lung cancer to gefitinib. N Engl J Med 2005;352(8):786-92 doi 716 10.1056/NEJMoa044238. 717 38. Soria JC, Ohe Y, Vansteenkiste J, Reungwetwattana T, Chewaskulyong B, Lee KH, et al. 718 Osimertinib in Untreated EGFR-Mutated Advanced Non-Small-Cell Lung Cancer. N Engl J Med 719 2018;378(2):113-25 doi 10.1056/NEJMoa1713137. 720 39. Wu YL, Cheng Y, Zhou X, Lee KH, Nakagawa K, Niho S, et al. Dacomitinib versus gefitinib as first- 721 line treatment for patients with EGFR-mutation-positive non-small-cell lung cancer (ARCHER 722 1050): a randomised, open-label, phase 3 trial. Lancet Oncol 2017;18(11):1454-66 doi 723 10.1016/S1470-2045(17)30608-3. 724 40. Sequist LV, Yang JC, Yamamoto N, O'Byrne K, Hirsh V, Mok T, et al. Phase III study of afatinib or 725 cisplatin plus pemetrexed in patients with metastatic lung adenocarcinoma with EGFR 726 mutations. J Clin Oncol 2013;31(27):3327-34 doi 10.1200/JCO.2012.44.2806. 727 41. Wu YL, Zhou C, Hu CP, Feng J, Lu S, Huang Y, et al. Afatinib versus cisplatin plus gemcitabine for 728 first-line treatment of Asian patients with advanced non-small-cell lung cancer harbouring EGFR 729 mutations (LUX-Lung 6): an open-label, randomised phase 3 trial. Lancet Oncol 2014;15(2):213- 730 22 doi 10.1016/S1470-2045(13)70604-1. 731 42. Park K, Tan EH, O'Byrne K, Zhang L, Boyer M, Mok T, et al. Afatinib versus gefitinib as first-line 732 treatment of patients with EGFR mutation-positive non-small-cell lung cancer (LUX-Lung 7): a 733 phase 2B, open-label, randomised controlled trial. Lancet Oncol 2016;17(5):577-89 doi 734 10.1016/S1470-2045(16)30033-X. 735 43. Mitsudomi T, Morita S, Yatabe Y, Negoro S, Okamoto I, Tsurutani J, et al. Gefitinib versus 736 cisplatin plus docetaxel in patients with non-small-cell lung cancer harbouring mutations of the

737 epidermal growth factor receptor (WJTOG3405): an open label, randomised phase 3 trial. Lancet 738 Oncol 2010;11(2):121-8 doi 10.1016/S1470-2045(09)70364-X. 739 44. Rosell R, Carcereny E, Gervais R, Vergnenegre A, Massuti B, Felip E, et al. Erlotinib versus 740 standard chemotherapy as first-line treatment for European patients with advanced EGFR 741 mutation-positive non-small-cell lung cancer (EURTAC): a multicentre, open-label, randomised 742 phase 3 trial. Lancet Oncol 2012;13(3):239-46 doi 10.1016/S1470-2045(11)70393-X. 743

744

745 Figure Legends

746 Figure 1.

747 Clinical endpoints by baseline activating EGFR mutation subgroup. (A) progression-free

748 survival. Data from Nakagawa K, et al. Lancet Oncol 2019; 20: 1655–69; (B) duration of

749 response; (C) progression-free survival 2; and (D) time-to-chemotherapy. Abbreviations: PD,

750 progressive disease; NE, not evaluable; NR, not reached.

751

752 Figure 2.

753 Guardant360 NGS detected somatic gene alterations at baseline or treatment-emergent (TE), co-

754 occurring with ex19del or ex21L858R, and impact on outcomes. (A) Baseline somatic gene

755 alterations co-occurring with either ex19del or ex21L858R mutations as detected by

756 Guardant360 NGS. (B) Treatment-emergent somatic gene alterations co-occurring with either

757 ex19del or ex21L858R mutations in patients with Guardant360 NGS positive results and who

758 had disease progression at the post–study treatment discontinuation follow-up visit. Analysis

759 focused on those patients with at least one alteration detectable both at baseline and at disease

760 progression; this approach, in part, controls for potential differences in tumor shedding.

761 Abbreviations: CNA, copy number alterations; SNV, single nucleotide variants; wt, wild-type.

762

763

764 Table 1. Clinical trials reporting PFS outcomes data for TKI monotherapies and antiangiogenic/TKI combinations by EGFR-activating mutation. Ex19del Ex21L858R 1-year Median PFS PFS vs 1-year PFS Study Geography Size CNSc Median PFS vs HR (95% rates vs control  HR (95% rates vs (N) (%) control (mo)  mPFS CI) control (mo) mPFS CI) control TKI monotherapy FLAURAa (Ph3 osimertinib vs gefitinib Global 556 19/23 21.4 vs 11.0 10.4 0.43 (0.32– 70% vs 14.4 vs 9.5 4.9 0.51 (0.36– 59% vs or erlotinib) (38) 0.56) 43% 0.71) 38% ARCHER 1050a (Ph3 dacomitinib vs East 452 Excl 16.5 vs 9.2 7.3 0.55 (0.41– 59% vs 12.3 vs 9.8 2.5 0.63 (0.44– 51% vs gefitinib) (39) Asia/EU 0.75) 35% 0.88) 38% LUX-Lung 3 (Ph3 afatinib vs cisplatin Global 345 NR 13.8 vs 5.6 8.1 0.28 (0.18– NA 10.8 vs 8.1 2.7 0.75 (0.48– NA plus pemetrexed) (40) 0.44) 1.19) LUX-Lung 6 (Ph3 afatinib vs cisplatin East-Asia 364 NR 13.1 vs 5.6 8.1 0.20 (0.13– NA 9.6 vs 5.6 4.0 0.31 (0.19– NA plus gemcitabine) (41) 0.33) 0.52) LUX-Lung 7 (RPh2 afatinib vs International 319 16/15 12.7 vs 11.0 1.7 0.76 (0.55– 51% vs 10.9 vs 0.1 0.71 (0.47– 42% vs gefitinib) (42) 1.06) 42% 10.8 1.06) 39% WJTOG3404 (Ph3 gefitinib vs cisplatin Japan 177 NR 9.0 vs 6.0 3.0 0.45 (0.27– NA 9.6 vs 6.7 2.9 0.51 (0.29– NA plus docetaxel) (43) 0.77) 0.90) EURTACa (erlotinib vs cisplatin plus EU 173 10/13 11.0 vs 4.6 6.4 0.30 (0.18– 47% vs 8.4 vs 6.0 2.4 0.55 (0.29– 29% vs docetaxel or gemcitabine) (44) 0.50) 11% 1.02) 10% Antiangiogenics + TKI RELAYb (RPh3 ramucirumab ± erlotinib) Global 449 Excl 19.6 vs 12.5 7.1 0.65 (0.47– 74% vs 19.4 vs 8.2 0.62 (0.44– 70% vs (19) 0.90) 54% 11.2 0.87) 47%

NEJ.026a (Ph3 erlotinib ± bevacizumab; Japan 228 32/32 16.6 vs 12.4 4.2 0.69 (0.41– 71% vs 17.4 vs 3.7 0.57 (0.33– 70% vs Japan) (17) 1.16) 46% 13.7 0.97) 58%

JO25567a (RPh2 erlotinib ± bevacizumab) Japan 154 Excl 18.0 vs 10.3 7.7 0.41 (0.24– 78% vs 13.9 vs 7.1 6.8 0.67 (0.38– 75% vs (16) 0.72) 29% 1.18) 39% CTONG.1509a,b (Ph3 erlotinib ± China 311 28/30 18.0 vs 12.6 5.4 0.63 (0.44– 79% vs 18.0 vs 9.7 8.3 0.50 (0.34– 75% bevacizumab;) (18) 0.92) 60% 0.74) vs 37%

765 Abbreviations: mPFS, median progression-free survival; NA, not available. a1-year PFS rates estimated from published Kaplan-Meier survival curves. 766 bInvestigator-assessed PFS. c CNS metastasis as presented at baseline 767 Excl=excluded; NR=not reported 768

769 Table 2. Baseline patient and disease characteristics by baseline EGFR-activating mutation subgroup and by treatment arms

Total Ex19del Ex21L858R

Ex19del Ex21L858R RAM+ERL PBO+ERL RAM+ERL PBO+ERL (N=243) (N=204) (N=123) (N=120) (N=99) (N=105) Sex, n (%) Female 146 (60) 136 (67) 74 (60) 72 (60) 66 (67) 70 (67)

Age in years, median (range) 63 (23–84) 66 (41–89) 64 (27–84) 63 (23–83) 67 (44–86) 66 (41–89)

Racea Asian 174 (72) 170 (83) 89 (72) 85 (71) 81 (82) 89 (85)

Caucasian 67 (28) 33 (16) 34 (28) 33 (28) 18 (18) 15 (14)

Smoking history, n (%) Never 144 (59) 127 (62) 71 (58) 73 (61) 61 (62) 66 (63)

ECOG performance status, n (%) 0 131 (54) 103 (51) 65 (53) 66 (55) 50 (51) 53 (51)

Disease classification, n (%) Primary 207 (85) 178 (87) 103 (84) 104 (87) 91 (92) 87 (83) metastatic Recurrent 36 (15) 26 (13) 20 (16) 16 (13) 8 (8) 18 (17) metastatic Metastases sites, n (%) Lung 224 (92) 188 (92) 112 (91) 112 (93) 93 (94) 95 (91) Lymph 152 (63) 127 (62) 78 (63) 74 (62) 64 (65) 63 (60) Bone 63 (26) 69 (34) 35 (29) 28 (23) 33 (33) 36 (34) Liver 26 (11) 19 (9) 11 (9) 15 (13) 10 (10) 9 (9) Other 146 (60) 104 (51) 70 (57) 76 (63) 48 (49) 56 (53) Number of metastatic sites 1 29 (12) 23 (11) 12 (10) 17 (14) 9 (9) 14 (13) 2 90 (37) 86 (42) 50 (41) 40 (33) 46 (47) 40 (38) 3 99 (41) 74 (36) 50 (41) 49 (41) 32 (32) 42 (40) 4 20 (8) 15 (7) 11 (9) 9 (8) 9 (9) 6 (6) ≥5 5 (2) 6 (3) 0 5 (4) 3 (3) 3 (3) 770 Abbreviations: ECOG, Eastern Cooperative Oncology Group; n, number of patients per category; N, number of patients in population. 771 aPBO + ERL arm included three Other: one Black or African American, one Missing (ex19del group), one American Indian or Alaska Native (ex21L858R 772 group).

773 Table 3. Clinical outcomes by baseline EGFR-activating mutation subgroup and by treatment arms.

Ex19del Ex21L858R

RAM+ERL PBO+ERL RAM+ERL PBO+ERL Duration of therapy n 122 120 97 105 Ramucirumab/placebo

Event rates, n (%) 85 (70) 94 (78) 70 (72) 88 (84) Censoring rates, n (%) 37 (30) 26 (22) 27 (28) 17 (16) Median, mo (95% CI) 12.4 (9.4–14.8) 11.0 (8.5–12.4) 12.7 (8.0–15.1) 8.8 (6.8–11.1)

Erlotinib

Event rates, n (%) 85 (70) 94 (78) 70 (72) 88 (84) Censoring rates, n (%) 37 (30) 26 (22) 27 (28) 17 (16) Median, mo (95% CI) 15.2 (12.9–19.5) 12.0 (10.6–13.8) 15.1 (12.7–19.5) 10.4 (7.7–12.4)

Overall response rate n 123 120 99 105 ORR, % (95% CI) 79 (72-86) 83 (76-89) 74 (65, 82) 66 (57, 75)

Disease control rate n 123 120 99 105 DCR, % (95% CI) 96 (92, 99) 96 (92, 99) 95 (91, 99) 95 (91, 99)

Time-to-Response n 97 99 73 69 Median, mo (95% CI) 1.4 (1.3–2.6) 1.4 (1.3–1.7) 1.5 (1.4–1.7) 1.4 (1.3–1.6)

Duration of response n 97 99 73 69 Event rates, n (%) 53 (55) 75 (76) 48 (66) 53 (77) Censoring rates, n (%) 44 (45) 24 (24) 25 (34) 16 (23) Median, mo (95% CI) 18.2 (13.8-20.9) 11.0 (9.7-12.3) 16.2 (12.5-20.1) 11.1 (9.7-12.7) Hazard ratio (95% CI) 0.54 (0.38–0.77) 0.73 (0.49–1.08)

Interim OS

n 123 120 99 105 Event rates, n (%) 21 (17) 15 (13) 16 (16) 27 (26) Censoring rates, n (%) 102 (83) 105 (88) 83 (84) 78 (74) Median, mo NR NR NR NR HR (95% CI) 1.44 (0.74–2.80) 0.61 (0.33–1.14) 1-year OS rate, % (95% CI) 94 (88–97) 94 (88–97) 92 (84–96) 93 (86–97) 2-year OS rate, % (95% CI) 83 (74–89) 87 (78–93) 84 (74–90) 71 (58–80) 774 Abbreviations: DCR, disease control rate; DOT, duration of therapy; NR, not reached; ORR, overall response rate 775 aDuration of therapy is censored at the last date of study treatment for patients who remain on study treatment (ex19del: n=122 RAM+ERL vs 120 PBO+ERL 776 and ex21L858R: n=97 RAM+ERL vs 105 PBO+ERL patients)

Figure 1A Ex19del 1.0 N Events, n (%) Censored, n (%) Median, mo (95% CI) 0.9 RAM+ERL 123 64 (52) 59 (48) 19.6 (15.1–22.2) PBO+ERL 120 84 (70) 36 (30) 12.5 (11.1–15.3) 0.8 Unstratified HR (95% CI): 0.651 (0.469–0.903) 0.7 1-yr PFS rates: 0.6 74% vs 54%

0.5

0.4

0.3

0.2

Progression-Free Survival Probability Survival Progression-Free 0.1

0 0 3 6 9 12 15 18 21 24 27 30 33 36 Time From Randomization (mo) Patients at risk, n RAM+ERL 123 108 96 87 72 54 38 25 17 11 6 0 0 PBO+ERL 120 110 94 78 58 43 32 22 15 10 2 2 0

Ex21L858R 1.0 N Events, n (%) Censored, n (%) Median, mo (95% CI) 0.9 RAM+ERL 99 58 (59) 41 (41) 19.4 (14.1–21.9) 0.8 PBO+ERL 105 74 (70) 31 (30) 11.2 (9.6–13.8) Unstratified HR (95% CI): 0.618 (0.437–0.874) 0.7

0.6 1-yr PFS rates: 70% vs 47% 0.5

0.4

0.3

0.2

Progression-Free Survival Probability Survival Progression-Free 0.1

0 0 3 6 9 12 15 18 21 24 27 30 33 36 Time From Randomization (mo) Patients at risk, n RAM+ERL 99 87 73 66 60 49 31 24 15 9 4 1 0 PBO+ERL 105Downloaded86 from 73 clincancerres.aacrjournals.org 58 41 29 20 on September 15 1227, 2021. 5 © 2021 2 American 2 Association0 for Cancer Research. Author Manuscript Published OnlineFirst on July 22, 2021; DOI: 10.1158/1078-0432.CCR-21-0273 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Figure 1B Ex19del 1.0 N Events, n (%) Censored, n (%) Median, mo (95% CI) 0.9 RAM+ERL 97 53 (55) 44 (45) 18.2 (13.8–20.9) PBO+ERL 99 75 (76) 24 (24) 11.0 (9.7–12.3) 0.8 Unstratified HR (95% CI): 0.542 (0.380–0.772) 0.7 Continued response 0.6 at 12 mo: 69% vs 42% 0.5

0.4 Continued response at 18 mo: 0.3 50% vs 23% 0.2 Duration of Response Probability Response of Duration 0.1

0 0 3 6 9 12 15 18 21 24 27 30 33 36 Duration of Response (mo) Patients at risk, n RAM+ERL 97 90 84 74 56 39 27 13 7 6 1 0 0 PBO+ERL 99 90 77 58 37 27 16 9 7 1 1 1 0

Ex21L858R 1.0 N Events, n (%) Censored, n (%) Median, mo (95% CI) 0.9 RAM+ERL 73 48 (66) 25 (34) 16.2 (12.5–20.1) 0.8 PBO+ERL 69 53 (77) 16 (23) 11.1 (9.7–12.7) Unstratified HR (95% CI): 0.731 (0.493–1.083) 0.7

0.6 Continued response at 12 mo: 0.5 63% vs 45%

0.4 Continued response at 18 mo: 0.3 47% vs 26%

0.2 Duration of Response Probability Response of Duration 0.1

0 0 3 6 9 12 15 18 21 24 27 30 33 36 Duration of Response (mo) Patients at risk, n RAM+ERL 73 64 57 53 43 29 21 8 5 3 1 0 PBO+ERL 69Downloaded62 from 50 clincancerres.aacrjournals.org 43 28 20 12 on September 10 27, 3 2021. 2 © 2021 2 American 0 Association for Cancer Research. Author Manuscript Published OnlineFirst on July 22, 2021; DOI: 10.1158/1078-0432.CCR-21-0273 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Figure 1C Ex19del 1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3 N Events, n (%) Censored, n (%) Median, mo (95% CI) RAM+ERL 123 34 (28) 89 (72) 33.1 (28.1–NR) 0.2 PBO+ERL 120 35 (29) 85 (71) NR Unstratified HR (95% CI): 0.926 (0.577–1.485)

Progression-Free Survival 2 Probability 2 Survival Progression-Free 0.1

0 0 3 6 9 12 15 18 21 24 27 30 33 36 Time From Randomization (mo) Patients at risk, n RAM+ERL 123 120 116 113 104 90 69 51 40 27 12 4 0 PBO+ERL 120 119 117 115 98 82 62 45 37 29 12 5 0

Ex21L858R 1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3 N Events, n (%) Censored, n (%) Median, mo (95% CI) RAM+ERL 99 27 (27) 72 (73) NR (26.6–NR) 0.2 PBO+ERL 105 44 (42) 61 (58) 26.3 (22.4–NR)

Progression-Free Survival 2 Probability 2 Survival Progression-Free 0.1 Unstratified HR (95% CI): 0.600 (0.371–0.969)

0 0 3 6 9 12 15 18 21 24 27 30 33 36 Time From Randomization (mo) Patients at risk, n RAM+ERL 99 94 91 87 82 75 57 46 31 23 9 3 0 PBO+ERL Downloaded105 104 from 101 clincancerres.aacrjournals.org 93 83 67 53on September 44 27, 29 2021. 16 © 2021 5 American 3 Association 0 for Cancer Research. Author Manuscript Published OnlineFirst on July 22, 2021; DOI: 10.1158/1078-0432.CCR-21-0273 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Figure 1D Ex19del 1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3 N Events, n (%) Censored, n (%) Median, mo (95% CI) RAM+ERL 123 45 (37) 78 (63) 28.6 (26.2–33.7) 0.2 PBO+ERL 120 44 (37) 76 (63) NR (25.3–NR) Unstratified HR (95% CI): 1.034 (0.682–1.567) 0.1 Time-to-Chemotherapy Survival Probability Survival Time-to-Chemotherapy 0 0 3 6 9 12 15 18 21 24 27 30 33 36 Time From Randomization (mo) Patients at risk, n RAM+ERL 123 120 113 106 96 82 63 48 36 25 16 6 0 PBO+ERL 120 117 112 104 93 79 66 47 37 29 15 9 0

Ex21L858R 1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3 N Events, n (%) Censored, n (%) Median, mo (95% CI) 0.2 RAM+ERL 99 30 (30) 69 (70) NR (26.0–NR) PBO+ERL 105 50 (48) 55 (52) 26.3 (18.7–30.7) 0.1 Unstratified HR (95% CI): 0.554 (0.352–0.872) Time-to-Chemotherapy Survival Probability Survival Time-to-Chemotherapy 0 0 3 6 9 12 15 18 21 24 27 30 33 36 Time From Randomization (mo) Patients at risk, n RAM+ERL 99 93 89 85 79 70 55 46 33 23 12 7 0 PBO+ERL Downloaded105 100 from 93clincancerres.aacrjournals.org 83 72 59 47on September 34 27, 28 2021. 18 © 2021 9 American 4 Association 0 for Cancer Research. Author Manuscript Published OnlineFirst on July 22, 2021; DOI: 10.1158/1078-0432.CCR-21-0273 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Figure 2A EGFR mutation subtype Ex19del Ex21L858R Ex19del Ex21L858R EGFR Co-occurring (N=204) (N=180) -activating mutation Patients with co-mutation (%) mutation TP53 42.6 43.3 EGFR other* 25.0 25.0 PIK3CA 12.7 7.2 NF1 9.3 5.6 APC 6.9 7.2 BRAF 7.4 5.0 CDK6 5.9 4.4 MET 5.4 5.0 BRCA1 5.4 3.3 CTNNB1 5.4 3.3 ERBB2 1.5 6.1 KRAS 3.9 3.3 MYC 3.9 3.3 ARID1A 3.9 2.8 RB1 3.4 3.3 SMAD4 4.4 2.2 GNAS 2.9 3.3 MTOR 2.9 3.3 PDGFRA 3.4 1.7 AR 3.4 1.1 CCNE1 2.0 2.8 KIT 2.0 2.8 BRCA2 2.5 1.7 FGFR2 2.0 2.2

Downloaded from clincancerres.aacrjournals.orgEx19del on September Ex21L858R 27, 2021. © 2021 American Association for Cancer CNA SNVResearch. INDEL Multiple Author Manuscript Published OnlineFirst on July 22, 2021; DOI: 10.1158/1078-0432.CCR-21-0273 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Figure 2B EGFR mutation subtype Ex19del Ex21L858R Ex19del Ex21L858R Co- RAM+ERL PBO+ERL RAM+ERL PBO+ERL occurring (N=33) (N=53) (N=34) (N=48) mutation Patients with TE co-mutation (%) EGFR T790M 27.3 34.0 23.5 27.1 EGFR other 18.2 13.2 14.7 29.2 TP53 30.3 9.4 11.8 16.7 MET 6.1 3.8 5.9 10.4 KRAS 15.2 1.9 2.9 4.2 APC 9.1 3.8 5.9 2.1 BRAF 6.1 3.8 0 8.3 NF1 9.1 1.9 8.8 2.1 PIK3CA 3.0 1.9 0 12.5 BRCA2 0 1.9 5.9 6.3 ERBB2 3.0 1.9 2.9 6.3 FGFR2 6.1 3.8 2.9 2.1 ARID1A 3.0 3.8 2.9 2.1 KIT 6.1 0 2.9 2.1 PTEN 3.0 0 2.9 4.2 RB1 6.1 0 0 4.2 AR 3.0 0 0 4.2 CDK4 3.0 1.9 0 2.1 CDK6 3.0 0 0 4.2 MYC 0 0 2.9 4.2 SMAD4 3.0 0 0 4.2 CCND2 0 0 2.9 2.1 DDR2 3.0 0 0 2.1 ESR1 3.0 0 0 2.1 FGFR1 3.0 0 0 2.1 GNAS 3.0 0 2.9 0 MAP2K1 3.0 0 2.9 0 PDGFRA 3.0 0 0 2.1

Ex19del Ex21L858R Treatment Downloaded from clincancerres.aacrjournals.org on September 27, 2021. © 2021 American Association for Cancer CNA SNV INDELResearch. Multiple RAM PBO Author Manuscript Published OnlineFirst on July 22, 2021; DOI: 10.1158/1078-0432.CCR-21-0273 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

RELAY Subgroup Analyses by EGFR Ex19del and Ex21L858R Mutations for Ramucirumab Plus Erlotinib in Metastatic Non-small Cell Lung Cancer

Kazuhiko Nakagawa, Ernest Nadal, Edward B. Garon, et al.

Clin Cancer Res Published OnlineFirst July 22, 2021.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-21-0273

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2021/08/19/1078-0432.CCR-21-0273.DC1

Author Author manuscripts have been peer reviewed and accepted for publication but have not yet been Manuscript edited.

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2021/07/21/1078-0432.CCR-21-0273. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.