Effect of the quality of the methodology on the relationship between progression-free survival (PFS) and overall survival (OS) in advanced colorectal cancer clinical trials.

Background: Our aim was to create a quality-index assuming that high-quality trials would have a proportional relationship between PFS and OS.
Methods: Randomized clinical trials (RCT) with new compounds (oxaliplatin, irinotecan, cetuximab, panitumumab and bevacizumab) in at least one of the arms of the RCT in first-line therapy for advanced colorectal cancer were analyzed by two independent researchers. A quality score ranging from 0 to 7 based on the following variables was built: 3 aspects affecting PFS or/and OS; final planned-sub-analysis (1 point), multivariate validation of stratification factors (2 points), double blinding (1 point) or in the quality of the primary end-point (PFS) radiological central review (1 point), clear definition of events and censoring patients (1 point) and statement of protocol deviations in events and censored patients (1 point). The relationship between OS and PFS was investigated by means of meta-regression models following a random-effects approach using the ln(RR) as the dependent variable and weighting by the inverse of each study variance. RCT quality (indicator variable) and median PFS effect (ln of the median ratio) were the main factors assessed in this study. The analysis was performed using SAS version 9.1.3 software.
Results: Forty-four RCT in advanced colorectal cancer published between 2000 and 2009 were included in the study. Five RCT were excluded because full data has not been published. Thirty-one RCT were classified as “low-quality” trials (LQT, score < 3) and eight RCT fell in the “high-quality” category (HQTscore > = 3). Taking the 39 RCT selected together, the association of OS and PFS was moderate and significant: the increment in ln(RR) for the relative change in median PFS was 0.21 (p = 0.0008; R = 0.45). The increment was higher in HQT (B = 0.44, p = 0.033; R = 0.62) than in LQT (B = 0.13, p = 0.013; R = 0.32).
Conclusions: PFS explains a 27.8% more of variability of OS in those studies classified as “high-quality” RCT by the proposed score than those classified as “low-quality" RCT. In HQT, the relationship between the increase in OS and the increase in PFS is closer and stronger.