Skip to main content

Design-Outcome Concordance in Targeted-Therapy trials in Non-Small cell lung cancer: a Meta-Research study

In brief

NSCLC targeted-therapy trials showed a 0.47 hazard ratio for progression-free survival, about 25% stronger than the

A review of 45 phase-III trials found the median observed hazard ratio of 0.47, markedly better than the planned 0.64, indicating a larger than anticipated benefit. Most studies (84%) reached statistical significance, but many relied on optimistic assumptions and violated proportional-hazards rules, highlighting the need for transparent retrodesign metrics and use of restricted mean survival time.

Journal
Journal of the National Cancer Institute (Q1)
Published
20 July 2026
Study design
Randomized controlled trial
Evidence level
Level 1, High (CEBM 1b)
Authors
Luca Mastrantoni, Antonio Vitale, Jacopo Russo, Giulia Giordano, Emanuele Vita, Massimo Di Maio, et al.
PMID
42477876
DOI
10.1093/jnci/djag241

Why clinicians should know about it

Abstract

BACKGROUND: In biomarker-selected non-small cell lung cancer (NSCLC) targeted therapies often produce substantial improvements in survival outcomes. We conducted a systematic review of phase III randomized controlled trials (RCTs) in oncogene-driven NSCLC to evaluate the concordance between design assumptions and observed treatment effects. METHODS: MEDLINE, Embase, and the Cochrane Central Register of Controlled Trials were searched for phase III RCTs published between January 1, 2005, and February 1, 2025, enrolling patients with metastatic or locally advanced, biomarker-selected NSCLC, comparing targeted agents versus standard regimens in a superiority design. Hazard ratios (HRs) for progression-free survival were considered. Statistical analyses included meta-analyses with meta-regression; retrodesign analysis of power, type S error, and exaggeration ratios; design-outcome concordance; and restricted mean survival time (RMST). RESULTS: 1987 records were screened, and 45 studies included. Progression-free survival was the primary outcome in all RCTs. Median expected HR was 0.64 (IQR 0.61-0.67). Pooled observed HR was 0.47 (95% CI 0.40-0.54), studies with higher standard errors were associated with larger treatment effects (P<.001). Statistically significant results were reported in 38 (84%) RCTs. Under original assumptions, median retrodesign power was 0.78 (IQR 0.72-0.88), and using the observed HR, median power was 1.00 (IQR 0.98-1.00). Pooled observed-to-expected HR ratio was 0.76 (95%CI 0.66-0.87). In Bayesian analyses, 17 (41%) trials had <1% probability of exceeding planned HR. Proportional hazards (PH) assumption was violated in 8 trials (19%). Pooled HR/RMST ratio was 1.40 (95% CI 1.29-1.52). CONCLUSIONS: Our work supports reporting retrodesign metrics, the minimal clinically important difference alongside the anticipated effect size, and systematic inclusion of RMST.

Abstract as published, via PubMed.

View on PubMedFull text at the publisherOpen in the app

For healthcare professionals. The summary is generated by AI from the published abstract, and the evidence level is assigned automatically from the study design on the Oxford CEBM hierarchy. Neither is medical advice. Read the full paper before changing practice.