Y-K Kang, A Ohtsu, E Van Cutsem, SY Rha, A Sawaki SR Park, H-Y Lim, J Wu, B Langer, MA Shah on behalf of AVAGAST investigators AVAGAST: a randomized, double-blind placebo-controlled, phase III study of first-line capecitabine and cisplatin + bevacizumab or placebo in patients with advanced gastric cancer (AGC)

Rationale for Bevacizumab in AGC Angiogenesis important for tumor growth, progression and metastases Vascular endothelial growth factor (VEGF): – Critical growth factor for tumor angiogenesis – Over-expressed and prognostic for many human tumors Bevacizumab: – Humanized monoclonal antibody to VEGF – Effective and safe in mCRC and other tumor types – Promising results in Phase II studies in AGC 1 1 Shah et al. J Clin Oncol 2006;23:2574–2576

AVAGAST: A Randomized Double-Blind Placebo- Controlled Phase III Study Starting dose of bev/placebo: 30 minutes, subsequent doses: 15 minutes Capecitabine*/Cisplatin (XP) + Placebo q3w Capecitabine*/Cisplatin (XP) + Bevacizumab q3w Locally advanced or metastatic gastric cancer R *5-FU also allowed if cape contraindicated Cape 1000 mg/m 2 oral bid, d1–14, 1-week rest Cisplatin 80 mg/m 2 d1 Bevacizumab 7.5 mg/kg d1 Maximum of 6 cycles of cisplatin Cape and bevacizumab/placebo until PD Stratification factors: 1. Geographic region 2. Fluoropirimidine backbone 3. Disease status

Endpoints and Statistical Assumptions Primary: overall survival Secondary: PFS, TTP, ORR, duration of response, safety, QoL, biomarkers Statistical assumptions  Median overall survival improvement from 10.0 to 12.8 months (HR 0.78)  Two-sided α-level = 0.05, 80% power  Required sample size: 760 patients for 517 deaths (with interim analysis)

Main Eligibility Criteria Metastatic or inoperable, locally advanced adenocarcinoma of the stomach or gastro-esophageal junction (GEJ) Measurable or evaluable disease ECOG performance status 0–2 No previous chemotherapy for metastatic/locally advanced gastric cancer If adjuvant chemotherapy, completed at least 6 months prior to randomization No previous platinum or antiangiogenic therapy No history of other malignancies

Trial Conduct From September 2007 to December 2008, 774 patients were enrolled A total of 93 centers in 17 countries were involved Interim analysis –Planned at 345 events, but not performed according to protocol as analysis date too close to anticipated final analysis Data cutoff for final analysis –November 2009 –After 509 events

Patient Characteristics (I) Number of patients N=774 (%) XP + Placebo N=387 XP + Bev N=387 GenderMale258 (67)257 (66) Age, yearsMedian (range)59 (22–82)58 (22–81) ECOG PS0–1 ≥2 367 (95) 20 (5) 365 (94) 22 * (6) RegionAsia Europe Pan-America 188 (49) 124 (32) 75 (19) 188 (49) 125 (32) 74 (19) FluoropyrimidineCapecitabine 5-FU 365 (94) 22 (6) 364 (94) 23 (6) Disease statusLocally advanced Metastatic 9 (2) 378 (98) 20 (5) 367 (95) *1 additional patient had an ECOG PS of 4

Patient Characteristics (II) Number of patients N=774 (%) XP + Placebo N=387 XP + Bev N=387 Primary site Stomach GEJ 338 (87) 49 (13) 333 (86) 54 (14) Histologic type Intestinal Diffuse Mixed 135 (35) 206 (53) 26 (7) 155 (40) 176 (46) 35 (9) Disease measurability Measurable Evaluable 297 (77) 90 (23) 311 (80) 76 (20) Metastatic sites, n 0 1 ≥2 8 (2) 131 (34) 247 (64) 8 (2) 131 (34) 247 (64) Prior gastrectomyYes107 (28)110 (28) Liver metastasisYes126 (33)130 (34)

Overall Survival XP + Placebo XP + Bev Number at risk XP + Placebo XP + Bev HR = % CI 0.73–1.03 p = Survival rate Study month

Progression-Free Survival XP + Placebo XP + Bev Number at risk XP + Placebo XP + Bev HR = % CI 0.68–0.93 p = Progression-free survival rate Study month

Best Overall Response: Measurable Disease Population XP + Placebo N=387 XP + Bev N=387 Patients with measurable disease Overall response111 (37%)143 (46%) 95% CI31.9– –51.7 Difference9% 95% CI0.6–16.6 P value (  2 ) Complete response3 (1%)5 (2%) Partial response108 (36%)138 (44%) Stable disease90 (30%)93 (30%) Progressive disease63 (21%)44 (14%) Not assessable33 (11%)31 (10%)

Overall Survival: Subgroup Analysis Pan-America * 29 patients with locally advanced disease only

Regional Differences in Efficacy Region XP + Placebo Median, mo XP + Bev Median, mo Delta, mo Hazard Ratio95% CI OSAsia –1.25 Europe –1.14 America –0.94 PFSAsia –1.14 Europe –0.93 America –0.93

Patient Characteristics by Region % of patientsAsiaEuropePan-America Age< ≥ ECOG PS0– *94 Primary siteStomach GEJ62216 Extent of diseaseMetastatic Locally advanced158 Prior gastrectomyyes no Measurable lesionyes no Liver metastasisyes no *1 additional patient had an ECOG PS of 4

Second-Line Therapy by Region Region Patients entered Patients receiving second-line treatment% Asia Europe Pan-America

Most Frequent Grade 3–5 AEs (≥5%) % of patients XP + Placebo N=381 XP + Bev N=386 Neutropenia3735 Febrile neutropenia45 Anemia1410 Decreased appetite118 Nausea107 Vomiting96 Diarrhea48 Hypokalemia63 Asthenia65 Hand-foot syndrome36 Hypertension<16 Pulmonary embolism53 Fatigue45

AEs of Special Interest to Bevacizumab % of patients XP + Placebo (N=381)XP + Bev (N=386) TotalG1G2G3G4G5TotalG1G2G3G4G5 Patients with ≥1 AE (all body systems) VTEs12< <1343– ATEs2<1–11–2 – Bleeding151123< <1 Hypertension1367<1––21796–– Proteinuria623–––734<1–– Wound complications <1 –––2 –– GI perforations<1– –– 2––2– CHF<1–– –– – – Fistula/abscess in 2 patients on XP + Bev Reversible posterior leukoencephalopathy syndrome in 2 patients on XP + Bev

AVAGAST Summary & Conclusions Primary endpoint of OS not met Secondary efficacy endpoints (PFS, best ORR) significantly improved, indicating clinical activity of bev + chemo in AGC Heterogeneous efficacy results in both treatment arms across geographic regions  Hypothesis generating with regard to tumor burden, patient status, practice patterns, genetics? No unexpected / new safety signals for bev Further analysis ongoing, including preplanned biomarker analysis

Acknowledgments Patients and their families Investigators, study coordinators and nurses at 93 centers in 17 countries AVAGAST study team at Genentech, Roche & Chugai