Magnitude of the benefit of progression-free survival as a potential surrogate marker in phase 3 trials assessing targeted agents in molecularly selected patients with advanced non-small cell lung cancer: systematic review

Abstract

Background: In evaluation of the clinical benefit of a new targeted agent in a phase 3 trial enrolling molecularly selected patients with advanced non-small cell lung cancer (NSCLC), overall survival (OS) as an endpoint seems to be of limited use because of a high level of treatment crossover for ethical reasons. A more efficient and useful indicator for assessing efficacy is needed.

Methods and findings: We identified 18 phase 3 trials in the literature investigating EGFR-tyrosine kinase inhibitor (TKIs) or ALK-TKIs, now approved for use to treat NSCLC, compared with standard cytotoxic chemotherapy (eight trials were performed in molecularly selected patients and ten using an "all-comer" design). Receiver operating characteristic analysis was used to identify the best threshold by which to divide the groups. Although trials enrolling molecularly selected patients and all-comer trials had similar OS-hazard ratios (OS-HRs) (0.99 vs. 1.04), the former exhibited greater progression-free survival-hazard ratios (PFS-HR) (mean, 0.40 vs. 1.01; P<0.01). A PFS-HR of 0.60 successfully distinguished between the two types of trials (sensitivity 100%, specificity 100%). The odds ratio for overall response was higher in trials with molecularly selected patients than in all-comer trials (mean: 6.10 vs. 1.64; P<0.01). An odds ratio of 3.40 for response afforded a sensitivity of 88% and a specificity of 90%.

Conclusion: The notably enhanced PFS benefit was quite specific to trials with molecularly selected patients. A PFS-HR cutoff of ∼0.6 may help detect clinical benefit of molecular targeted agents in which OS is of limited use, although desired threshold might differ in an individual trial.

Fig 1. A flow chart demonstrating the selection process of the trials analyzed.

NSCLC = non-small cell lung cancer, EGFR-TKI = epidermal growth factor receptor-tyrosine kinase inhibitor, ALK-TKI = anaplastic lymphoma kinase-tyrosine kinase inhibitor.

Fig 2. Associations between the progression-free survival-hazard ratio (PFS-HR) and overall survival (OS)-HR (R-squared = 0.233) (A), and that after stratification by trial design (B) (the molecularly selected patient design [blue], R-squared = 0.002, vs.

the all-comer design [pink], R-squared = 0.409; P-value for interaction = 0.34). Associations between the odds ratio of the overall response and OS-HR (R-squared = 0.101) (C), and that after stratification by trial design (D) (the molecularly selected patient design [blue], R-squared = 0.039, vs. the all-comer design [pink], R-squared = 0.429; P-value for interaction = 0.03). All analyses were weighted by trial size.

Fig 3. Distributions of hazard ratios (HRs) for overall survival (OS) (A) and progression-free survival (PFS) (B) and odds ratios for the overall response (C), stratified according to the two types of trials.

Fig 4. Receiver operating characteristic (ROC) curve defining cutoff levels predicting outcomes in eligible trials with molecularly selected patients in terms of the progression-free survival-hazard ratio (PFS-HR) (A) and the odds ratio for the overall response (B).

The most suitable cutoff was defined as that closest to the upper left corner. A, A PFS-HR of 0.60 was the optimal cutoff for distinguishing molecularly selected patient trials from all-comer trials (sensitivity 100%, specificity 100%, and AUC [area under the receiver operating characteristic curve], 1.00). B, The odds ratio for an overall response of 3.40 was a potentially useful cutoff to distinguish trials with molecularly selected patients from all-comer trials (sensitivity 88%, specificity 90%, and AUC = 0.95).