570P - Effect of post-protocol anti-epidermal growth factor receptor (EGFR) monoclonal antibody (mAb) therapy on survival outcomes in patients with wild-ty...

Date 01 October 2012
Event ESMO Congress 2012
Session Poster presentation III
Topics Anti-Cancer Agents & Biologic Therapy
Colon Cancer
Rectal Cancer
Presenter Jean-Yves Douillard
Authors M. Peeters1, J. Douillard2, A. Rong3, S. Siena4, S. Braun5, R. Sidhu6, T. Price7
  • 1Department Of Medical Oncology, University Hospital Antwerp, Antwerp/BE
  • 2Medical Oncology, Centre René Gauducheau (ICO) Institut de Cancerologie de l'Ouest, 44805 - Nantes/FR
  • 3Biostatistics, Amgen Inc., Thousand Oaks/US
  • 4Department Of Medical Oncology, Ospedale Niguarda Ca’ Granda, Milan/IT
  • 5Medical Development, Amgen (Europe) GmbH, Zug/CH
  • 6Medical Development, Amgen Inc., Thousand Oaks/US
  • 7Department Of Medical Oncology, Queen Elizabeth Hospital and University of Adelaide, Woodville/AU



Results from two randomized, multi-centre, phase III studies of pmab plus chemotherapy (FOLFOX4 in 1st-line [PRIME]; FOLFIRI in 2nd-line [181]) demonstrated a trend toward improved overall survival (OS) in the experimental arms vs chemotherapy alone, analysing the ITT populations. We evaluated the hypothesis that crossover to post-protocol anti-EGFR mAb-containing therapy in the control arms may have attenuated OS outcomes.


For this retrospective analysis, we used data of both PRIME and 181 study final analyses prespecified to occur at 30 months after the last patient was enrolled. We conducted sensitivity analyses to estimate OS outcomes using statistical procedures: (1) estimating the effect of randomized treatment as if no patients in the chemotherapy arm received subsequent anti-EGFR mAb-containing therapya; (2) identifying the survival differences that would have been observed had all patients stayed on protocol treatmentb; (3) modifying the Cox proportional-Hazards model, allowing for a time-dependent covariate with value 0 up to the point of subsequent therapy and a value of 1 from then on; (4) censoring patients at the time of subsequent anti-EGFR use and adjusting the informative censoring with the inverse probability-of-censoring weighted (IPCW) analysisd-f.




Patients receiving chemotherapy alone crossed over earlier and more frequently to post-protocol anti-EGFR treatment versus those receiving combination therapy. All analyses produced results indicating that the lack of a statistically significant OS benefit estimate may have been due to selective crossover bias. The results also indicate that IPCW method may be particularly suited for detecting OS benefits that otherwise would not be detected with an ITT approach that ignores selective crossover bias.


J. Douillard: AMGEN: Participation in Advisory Boards, speaker in Symposia compensated MERCK:Participation in Advisory Boards, speaker in Symposia compensated, Research support ROCHE: Participation in Advisory Boards, speaker in Symposia compensatedM. Peeters: Consultant/advisory role for Amgen and also received honoraria and research funding.

A. Rong: Employee of Amgen and holds Amgen stock.

S. Siena: Advisory role for Amgen, AstraZeneca, Merck Serono, Roche, Celgene.

S. Braun: Employee of Amgen and holds Amgen stock.

R. Sidhu: Employee of Amgen and holds Amgen stock.

T. Price: Consultant/advisory role for Amgen and received honoraria.

Table: 570P

Number of patients n = 325 n = 331 n = 303 n = 294
ITT analysis for OS
Median OS, months (95% CI) 23.9 (20.3– 27.7) 19.7 (17.6–22.7) 14.5 (13.0–16.1) 12.5 (11.2–14.2)
HR (95% CI) 0.88 (0.73–1.06) 0.92 (0.78–1.10)
Post-protocol anti-EGFR mAb therapy
Incidence, (%) 13 25 12 34
Median time to use, months 21.5 15.6 12.4 7.9
OS sensitivity analyses on influence of post-protocol anti-EGFR mAb therapy
Branson-Whitehead, 2002a 0.84 (0.68-1.05) 0.90 (0.71, 1.14)
Robins and Tsiatis, 1992b 0.83 (0.66-1.04) 0.89 (0.71, 1.13)
Allison, 1995c 0.68 (0.55-0.83) 1.03 (0.87, 1.23)
IPCWd-f 0.74 (0.56-0.97) 0.71 (0.53, 0.94)

aBranson and Whitehead, 2002; bRobins and Tsiatis, 1992; cAllison, 1995;

dRimawi & Hilsenbeck, 2012; eColleoni et al, 2011; fRobins & Finkelstein, 2000. Abbreviation: CI = confidence interval