1378P - Statistical Consideration of Biases in Progression-Free Survival (PFS) Resulting from Differences in Imaging Schedules

Abstract

ABSTRACT Backgroud In the three PIII trials for metastatic colorectal cancer (mCRC) patients who are refractory to standard chemotherapy: NCIC CTG CO.17 (NEJM 2007;357:2040-8), 20020408 (JCO 2007;25:1658-65), and CORRECT (2012 ASCO-GI, #LBA 385) trial, the median PFS (mPFS) were similar and the early part of PFS curves overlapped between treatment groups regardless of significance. Generally, experimental treatments can potentially have biomarkers in such cases. However, we consider that the scheduling of imaging tests affects PFS evaluations. Material and methods We simulated PFS computationally under the assumptions for mCRC patients in first (L1), second (L2), and third line (L3) as “True Parameters” in the following table (show L3 only). In addition, we assumed that the probability of unscheduled progression was 0 and 20%, and the imaging schedules were every 6 weeks (wk) (S1), every 8 wk (S2), and 4, 8, 12 and every 8 wk thereafter (S3), sample size was 250 patients in each group. S3 is the same schedule plan as in the TAS-102 PII trial (2011 ECCO, #6005). Results The following table shows the simulation results assumed that a probability of unscheduled progression is 0%. Simulated mPFSs were similar between the schedules in each treatment group of L1 and L2, but not in the control group of L3. Simulated hazard ratios (HRs) were almost the same regardless of the schedules in all lines. The difference of mPFSs between schedules in the control group of L3 depended on the timing of first evaluation. The simulated mPFSs in L3S2 were similar to results of the above three PIII trials, and those in L3S3 were similar to results of TAS-102 PII trial. The simulation results assumed the proportion of unscheduled progression of 20% were similar to that of 0%. Setting True Paremeters Simulation Results mTTP, MST (mo) Mean of mPFS (mo) Mean of HR E C E C L3S1 2, 6 1, 4.5 1.9 [1.5, 2.6] 1.4 [1.4, 1.4] 0.58 [0.49, 0.67] L3S2 1.9 [1.9, 2.0] 1.8 [1.8, 1.9] 0.59 [0.50, 0.69] L3S3 1.9 [1.8, 1.9] 1.0 [1.0, 1.1] 0.57 [0.48, 0.65] TTP: Time To Progression, E: Experimental Group, C: Control Group, [ ]: 2.5 and 97.5% percentiles. Conclusion The HR in PFS is a more important indicator than mPFS regardless of the timing of the first evaluation of tumor response in patients with mCRC in the later line. These idea might be applicable for mCRC as well as other cancers. Disclosure T. Tanase: Employed by, and own stock in, Taiho Pharmaceutical. C. Hamada: Taiho pharmaceutical. H. Fujii: Chugai Pharmaceutical, Bristol-Myers Squibb, Sanofi-Aventis, Novartis Pharma, GlaxoSmithKline, Eisai, Yakult Honsha, Takeda Pharmaceutical, Shionogi, Taiho Pharmaceutical, Ono Pharmaceutical, Takeda Bio Development Center. N. Nakayama: merck serono, Takeda pharmaceutical. T. Denda: Taiho Pharmaceutical, Pfizer, Yakult Honsha, Daiichi Sankyo. T. Yoshino: Consulting fee from Takeda; honoraria from Chugai, Takeda, Yakult, Bristol-Myers Squibb, and MerkSerono; research funding from Daiichi Sankyo, Taiho, Bayer, and ImClone. A. Ohtsu: Employment position in Bayer, and consulting fee from Takeda, Daiichi Sankyo, Novartis, Chugai,and Taiho; honoraria from Takeda, Daiichi Sankyo, Taiho, GlaxoSmithKline, Pfizer, Yakult, MerkSerono, and Bristol-Myers Squibb. All other authors have declared no conflicts of interest.