1. NCCTG N9831 May 2005 Update Perez EA, Suman VJ, Davidson N, Martino S, Kaufman P, on Behalf of NCCTG, ECOG, SWOG, CALGB

2. Perez E. Protocol NCCTG-N9831. H=trastuzumab (4mg/kg loading dose, followed by 2mg/kg); doxorubicin dose 60mg/m 2 ; cyclophosphamide, 600mg/m 2 ; paclitaxel, 80mg/m 2 q3w=every 3 weeks; qw=weekly NCCTG N9831 Schema RANDOMIZERANDOMIZERANDOMIZERANDOMIZE Radiation and/or hormonal therapy as indicated Paclitaxel qw x 12 Arm A: AC q3w x 4 AC q3w x 4 Paclitaxel qw x 12 Paclitaxel qw x 12 Arm B: AC q3w x 4 AC q3w x 4 H qw x 52 AC q3w x 4 Paclitaxel qw x 12 + H qw x 12 Arm C: H qw x 40

3. Statistical Plan Addition of H to AC  T Two pairwise comparisons Goal –To detect a 33% increase in median DFS from 6.3 to 8.4 years Final analysis –At 663 events for A vs C comparison –At 789 events for A vs B comparison Control: AC  T Sequential AC  T  H Concurrent AC  T + H  H Control: AC  T T=paclitaxel; DFS=disease free survival vs

4. Statistical Plan Timing of H Initiation Pairwise comparison Goal –To detect a 29% increase in median DFS from 7.3 to 9.4 years Final analysis –At 590 events for B vs C comparison Sequential AC  T  H Concurrent AC  T + H  H vs

5. Cardiac Testing Time (months) LVEF measurement 6 918–2103 Arm B: AC x 4 Paclitaxel H Arm A: AC x 4 Paclitaxel Arm C: AC x 4 Paclitaxel + H H LVEF=left ventricular ejection fraction; LLN=lower limit of normal Pre-AC Post-AC No H if symptoms or LVEF ↓ >15% or ↓ to <LLN R A N D O MI Z E     

6. Impact of Joint Analysis on N9831 April 2005 Joint analysis with B-31: Concurrent approach DMC asked for an unplanned interim analysis comparing Arm B (sequential) vs Arm C (concurrent) to assist in patient management DMC=data monitoring committee AC  T + H  H significantly improves disease-free and overall survival vs control: AC  T

7. Patient/Event Status at Time of Joint Analysis April 2005 Patients –Enrollment goals met (n: >3300)  700 patients on chemotherapy 2701 patients entered prior to 1/1/2005 –Median follow up: 1.5 years Total disease-free survival events –A and B: 220 (of 789 needed) –B and C: 147 (of 590 needed)

8. Results Disease-Free Survival *Stratified – nodal status and receptor status **for patients randomized before 1/1/2005 *Stratified – nodal status and receptor status Joint Analysis N9831 Analysis A B B C

9. Disease-Free Survival: A vs B N9831 100 90 80 70 60 50 40 30 20 10 0 01234 Years Number of patients followed A 97962935316815 B 98563740316920 AC → T Events=117 Hazard ratio=0.87 Stratified logrank 2P=0.2936 AC → T → H Events=103 %

10. Disease-Free Survival: B vs C N9831 100 90 80 70 60 50 40 30 20 10 0 01234 Years Number of patients followed B 84250128516220 C 84052028517817 AC → T → H Events=84 AC → T + H → H Events=53 % Hazard ratio=0.64 Stratified logrank 2P=0.0114

11. Overall Survival *Stratified – nodal status and receptor status Joint Analysis Results N9831 Analysis Results A B B C

12. Other Relevant Factors for Patient Management HER2 testing Cardiac tolerability comparisons based on planned analyses

13. HER2 Testing in N9831 Modest level of concordance between local and central laboratories for both IHC and FISH –With HercepTest™: 81% (78-83%) –With FISH: 87% (84-90%) High level of agreement between central and reference laboratory results for HER2 –94.5% for IHC (0, 1+, 2+) –95.1% for FISH (not amplified) Accurate HER2 testing is critical given the degree of trastuzumab benefit as a component of adjuvant therapy Updated from Perez EA, et al. ASCO 2004 (abstract 567)

14. Cardiac Monitoring Plan Monthly formal review of LVEF, clinical data Interim analyses after 100, 300, and 500 patients per arm –completed AC and followed at least 6 months ~ 9 months from registration Perez EA, et al. ASCO 2005 (abstract 556)

15. Difference in the incidence of cardiac events (CHF and cardiac deaths) between non-H and H arms is <4% 9 month analysis; 500 per arm with nl LVEF or LVEF decrease  15% from baseline (after AC) –0.0% (95% CI,0.0-0.7%) for control –2.2% (95% CI,1.1-3.8%) for control vs sequential –3.3% (95% CI,2.0-5.1%) for control vs concurrent* therapy with paclitaxel Effect of the Introduction of H on Cardiac Tolerability Perez EA, et al. ASCO 2005 (abstract 556) * at month 9, concurrent pts have received 3 additional months of H compared to sequential

16. Effect of Introduction of H on Disease Recurrence Conclusions 52% decreased recurrence with concurrent vs control treatment (P=3X10 -12 ) (joint analysis finding) 13% decreased recurrence with sequential vs control treatment (P=0.2936) 36% decreased recurrence with concurrent vs sequential treatment (P=0.0114) More follow up is needed to determine whether this trend continues

17. NCCTG N9831 Next Steps Pre-specified interim analyses at 50%, 67%, and 75% of events still planned Continued exploration of predictive factors for cardiac toxicity Continued patient follow up