JR Skillings Genentech, Inc, South San Francisco, CA

1. ARTERIAL THROMBOEMBOLIC EVENTS IN A POOLED ANALYSIS OF 5 RANDOMIZED, CONTROLLED TRIALS OF BEVACIZUMAB WITH CHEMOTHERAPY JR Skillings Genentech, Inc, South San Francisco, CA

2. BACKGROUND • Bevacizumab (BV; AvastinTM) is a recombinant, humanized, anti-vascular endothelial growth factor (VEGF) monoclonal antibody • BV combined with first-line irinotecan, 5-fluorouracil (5-FU), and leucovorin (IFL) chemotherapy increases survival in patients with metastatic colorectal cancer (mCRC) • Safety reports from several randomized, controlled trials suggested that adding BV to chemotherapy may increase the risk of arterial thromboembolic events (ATEs)

3. OBJECTIVE • To evaluate the risk of ATEs in patients with metastatic cancer receiving BV with chemotherapy using data pooled from 5 randomized, controlled trials

4. METHODS: STUDY DESIGN CAP=capecitabine; C=carboplatin; P=paclitaxel MBC= metastatic breast cancer; NSCLC=non-small-cell lung cancer 1Hurwitz H, et al. N Engl J Med. 2004;350:2335-2342; 2Kabbinavar F, et al. J Clin Oncol. 2005. In Press; 3Kabbinavar F, et al. J Clin Oncol. 2003;21:60-65; 4Miller K, et al. J Clin Oncol. 2005; 23:792-799; 5Johnson D, et al. J Clin Oncol. 2004;22:2184-2191

5. METHODS: IDENTIFICATION OF ATE CASES • The pooled database was broadly queried for adverse event terms that might be associated with the clinical consequences of ATEs • The following COSTART terms were used for the initial search: cerebrovascular accident, cerebral ischemia, subarachnoid hemorrhage, cerebral infarction, angina pectoris, myocardial infarction, myocardial ischemia, and arterial thrombosis. Thrombosis and verbatim terms including left ventricle were added • Death due to causes other than progressive disease were clinically reviewed for text fields including cerebrovascular accident, myocardial infarction or ATE. A case of ischemic bowel was added • The safety database was reviewed using broad search criteria and one additional case of right leg arterial occlusion was added • Cases of subarachnoid hemorrhage were removed because these were interpreted as bleeding events • Cases of cardiac arrest were removed when review of narratives suggested that the cases were related to progressive disease • A combined list of cases was identified

6. METHODS: IDENTIFIED TREATMENT-EMERGENT ATEs

7. METHODS: STATISTICAL ANALYSIS • Incidence of ATEs was tabulated for each treatment group • Kaplan-Meier plots of time-to-ATE were generated for each treatment group and median values estimated. A Cox proportional hazards regression was used to calculate hazard ratios, and P-values were calculated by the log-rank test • ATE rates per 100 person-years with 95% confidence intervals were computed by standard methods. The number of person-years of observation was defined as the sum of the time-to-ATE for all patients (for patients without an ATE, the observation time was defined as the last date of treatment plus 30 days) • Poisson regression was used to formally compare rates per 100 person-years between BV-treated patients and controls

8. METHODS: BASELINE VARIABLES USED FOR POTENTIAL RISK FACTOR ANALYSIS DBP=diastolic blood pressure; HTN=hypertension; SBP=systolic blood pressure

9. METHODS: OTHER BASELINE VARIABLES USED FOR POTENTIAL RISK FACTOR ANALYSIS TIA=transient ischemic attack

10. METHODS: POTENTIAL RISK FACTOR ANALYSIS • For each risk factor, numbers of patients with and without the factor were tabulated • Fisher’s exact test was used to compare the frequency of ATEs between BV-treated and control patients for each risk factor • Cox’s proportional hazards regression was used to compute hazard ratios and P-values for the effect of each factor on the hazard of ATEs • Backward elimination in a multivariate Cox model was used to assess the significance of each baseline risk factor that was significant in univariate analysis • Pooled population patients were assigned to subgroups based on significant baseline risk factors, and ATE rates per 100 person-years were calculated as described above within each subgroup by treatment • Irinotecan-treated patients from the pivotal trial in mCRC (AVF2107g)1 were assigned to subgroups based on these risk factors. Median PFS and overall survival were estimated for each subgroup using Kaplan-Meier methods, and hazard ratios were calculated from a Cox proportional hazards model 1Hurwitz H, et al. N Engl J Med. 2004;350:2335-2342

11. RESULTS: INCIDENCE OF ATEs FOR EACH TREATMENT GROUP MBC= metastatic breast cancer; NSCLC=non-small-cell lung cancer 1Hurwitz H, et al. N Engl J Med. 2004;350:2335-2342; 2Kabbinavar F, et al. J Clin Oncol. 2005. In Press; 3Kabbinavar F, et al. J Clin Oncol. 2003;21:60-65; 4Miller K, et al. J Clin Oncol. 2005; 23:792-799; 5Johnson D, et al. J Clin Oncol. 2004;22:2184-2191

12. RESULTS: ATE ANALYSIS CI=confidence interval *P=0.076; **P=0.03

13. RESULTS: KAPLAN-MEIER TIME-TO-ATE Patientsat risk by time point 0 5 10 15 20 25 Months 782 405 173 64 17 0 control 963683 421 260 111 15 chemo/AVF Uncorrected for uneven time on treatment and shorter follow up for control patients

14. RESULTS: ATE POTENTIAL RISK FACTORS IN UNIVARIANT ANALYSIS DBP=diastolic blood pressure; HTN=hypertension; SBP=systolic blood pressure *On study indicates occurrence during treatment (and prior to an ATE if one occurred) †Collected in selected trials1-3 1Hurwitz H, et al. N Engl J Med. 2004;350:2335-2342; 2Kabbinavar F, et al. J Clin Oncol. 2005. In Press; 3Miller K, et al. J Clin Oncol. 2005; 23:792-799

15. RESULTS: RELATIONSHIP BETWEEN ATEs AND PROTEINURIA • Patients with proteinuria (>500mg/24 hours) are at risk of ATEs regardless of treatment. The explanation for this finding is unclear Proteinuria (>500mg/24 hours) n=subset; N=total number

16. RESULTS: BASELINE ATE RISK FACTORS SIGNIFICANT IN MULTIVARIATE ANALYSIS • In a multivariate analysis with both baseline and on-study factors, no on-study factors were significant except presence of proteinuria >500mg/24 hours • In a separate multivariate analysis with only on-study factors, presence of proteinuria alone was significant

17. RESULTS: RISK/BENEFIT OF BV in mCRC • To understand the risk/benefit of BV in mCRC, data from the pivotal trial AVF21071 were reanalyzed* for survival based on subgroups created using risk factors for ATEs *The dataset used to support the preplanned final analysis for AVF2107g was used for this analysis 1Hurwitz H, et al. N Engl J Med. 2004;350:2335-2342

18. RESULTS: ATE INCIDENCE BY RISK GROUP n=subgroup; N=total number *These groups are not mutually exclusive

19. RESULTS: SURVIVAL HAZARD RATIOS IN mCRC BY RISK GROUP CI=confidence interval; PFS=progression-free survival *These groups are not mutually exclusive 1Hurwitz H, et al. N Engl J Med. 2004;350:2335-2342

20. CONCLUSIONS • Bevacizumab is associated with an approximately 2-fold increased risk of ATEs in patients with metastatic cancer receiving chemotherapy • Age ≥65 years and a history of ATEs are independent baseline risk factors for these events

21. CONCLUSIONS (cont’d) • Despite this risk, analysis of data from the pivotal trial1 indicate that bevacizumab confers a consistent survival benefit in mCRC patients overall, in all prespecified subgroups, and in these ATE risk subgroups • Oncologists must use their own clinical judgment in assessing the overall risk/benefit of adding bevacizumab to chemotherapy in patients at risk of ATEs 1Hurwitz H, et al. N Engl J Med. 2004;350:2335-2342