Background: Our objective was to determine the variability in assessment between investigators (INV) and independent review committees (IRC) for response rate (RR) and progression-free survival (PFS).
Methods: Phase III trials reporting INV and IRC assessments were identified. The difference in end point assessment (IRC - INV) across all study arms was determined. A random-effects model was used to calculate the mean difference between INV and IRC RR as well as PFS. Differences in estimated benefits of treatment (experimental - control) between IRC and INV were determined.
Results: Twenty-one trials were included (18 RR, 8 PFS). The estimated mean difference between IRC- and INV-determined RR was 4.57% [95% confidence interval (CI) 2.95% to 6.19%]. For median PFS, the estimated mean difference was -0.19 (95% CI -0.68 to 0.29) months. The difference in estimated benefits of treatment ranged from -7.0% to 7.2% for RR and -2.0 to +2.4 months for PFS; there was no evidence of systemic bias by INV (P = 0.54 for RR and 0.31 for PFS).
Conclusion: INV overestimate RR compared with IRC. Given the variability in assessing RR and PFS between INV and IRC, an IRC should be considered if the primary end point is on the basis of assessments of changes in tumor lesions.