Impact of a Biomarker-Based Strategy on Oncology Drug Development: A Meta-analysis of Clinical Trials Leading to FDA Approval

Abstract

Background: In order to ascertain the impact of a biomarker-based (personalized) strategy, we compared outcomes between US Food and Drug Administration (FDA)-approved cancer treatments that were studied with and without such a selection rationale.

Methods: Anticancer agents newly approved (September 1998 to June 2013) were identified at the Drugs@FDA website. Efficacy, treatment-related mortality, and hazard ratios (HRs) for time-to-event endpoints were analyzed and compared in registration trials for these agents. All statistical tests were two-sided.

Results: Fifty-eight drugs were included (leading to 57 randomized [32% personalized] and 55 nonrandomized trials [47% personalized], n = 38 104 patients). Trials adopting a personalized strategy more often included targeted (100% vs 65%, P < .001), oral (68% vs 35%, P = .001), and single agents (89% vs 71%, P = .04) and more frequently permitted crossover to experimental treatment (67% vs 28%, P = .009). In randomized registration trials (using a random-effects meta-analysis), personalized therapy arms were associated with higher relative response rate ratios (RRRs, compared with their corresponding control arms) (RRRs = 3.82, 95% confidence interval [CI] = 2.51 to 5.82, vs RRRs = 2.08, 95% CI = 1.76 to 2.47, adjusted P = .03), longer PFS (hazard ratio [HR] = 0.41, 95% CI = 0.33 to 0.51, vs HR = 0.59, 95% CI = 0.53 to 0.65, adjusted P < .001) and a non-statistically significantly longer OS (HR = 0.71, 95% CI = 0.61 to 0.83, vs HR = 0.81, 95% CI = 0.77 to 0.85, adjusted P = .07) compared with nonpersonalized trials. Analysis of experimental arms in all 112 registration trials (randomized and nonrandomized) demonstrated that personalized therapy was associated with higher response rate (48%, 95% CI = 42% to 55%, vs 23%, 95% CI = 20% to 27%, P < .001) and longer PFS (median = 8.3, interquartile range [IQR] = 5 vs 5.5 months, IQR = 5, adjusted P = .002) and OS (median = 19.3, IQR = 17 vs 13.5 months, IQR = 8, Adjusted P = .04). A personalized strategy was an independent predictor of better RR, PFS, and OS, as demonstrated by multilinear regression analysis. Treatment-related mortality rate was similar for personalized and nonpersonalized trials.

Conclusions: A biomarker-based approach was safe and associated with improved efficacy outcomes in FDA-approved anticancer agents.

Figure 1.

Forest-plot representing the relative response rate ratio (A), hazard ratios for progression-free survival (PFS) (B), and overall survival (OS) (C) between experimental and control arms by personalized therapy status in randomized registration trials. Studies are labeled by first author’s last name, and references were numbered according to supplementary references. In (A), RRR is shown and lines to the right of the vertical line indicate improvement in the experimental arm. In (B and C), the plots show hazard ratios (HRs) and, therefore, lines to the left of the vertical line indicate improvement (ie, lower HR for PFS or OS) for the experimental arm. P values were tested for subgroup differences (chi-square) and are two-sided. df = degree of freedom; Q = Cochran’s Q test.

Figure 1.

Forest-plot representing the relative response rate ratio (A), hazard ratios for progression-free survival (PFS) (B), and overall survival (OS) (C) between experimental and control arms by personalized therapy status in randomized registration trials. Studies are labeled by first author’s last name, and references were numbered according to supplementary references. In (A), RRR is shown and lines to the right of the vertical line indicate improvement in the experimental arm. In (B and C), the plots show hazard ratios (HRs) and, therefore, lines to the left of the vertical line indicate improvement (ie, lower HR for PFS or OS) for the experimental arm. P values were tested for subgroup differences (chi-square) and are two-sided. df = degree of freedom; Q = Cochran’s Q test.

Figure 1.

Forest-plot representing the relative response rate ratio (A), hazard ratios for progression-free survival (PFS) (B), and overall survival (OS) (C) between experimental and control arms by personalized therapy status in randomized registration trials. Studies are labeled by first author’s last name, and references were numbered according to supplementary references. In (A), RRR is shown and lines to the right of the vertical line indicate improvement in the experimental arm. In (B and C), the plots show hazard ratios (HRs) and, therefore, lines to the left of the vertical line indicate improvement (ie, lower HR for PFS or OS) for the experimental arm. P values were tested for subgroup differences (chi-square) and are two-sided. df = degree of freedom; Q = Cochran’s Q test.