1232P - Assessing uncertainty and vulnerability of clinical trials with immune-checkpoint inhibitors and chemotherapy (ICI+CT) in early stage non-small cell lung cancer (NSCLC) by informative censoring and survival inferred fragility index (SIFI)

Background

Randomized controlled trials (RCTs) showed survival improvement with neoadjuvant/perioperative ICI + CT in early NSCLC. However, interpretation of the results can be affected by a variable level of uncertainty due to censoring imbalance and vulnerability of statistical conclusions.

Methods

PubMed, Embase and Cochrane were searched until April 2024 for RCTs comparing neoadjuvant/perioperative ICI + CT with neoadjuvant CT in patients with early-stage NSCLC. Individual Patient Data (IPD) curves and the corresponding hazard ratios (HR) were reconstructed from published EFS Kaplan-Meier data from each trial and treatment arm by using an IPD software. Reverse HR (R-HR) was obtained by flipping the status of time-dependent outcome for individual patients (reverse Kaplan Meier method) and were used to estimate censoring imbalance. SIFI was determined by iteratively transferring the best survivors from the experimental arm to the control arm until p≥0.05; conversely, negative SIFI was calculated until p<0.05.

Results

Eight RCTs (n=3387) were included. R-HR analysis showed a trend in increased censoring imbalance in the experimental arm in 1 RCT (NADIM II) and in the control arm in 2 RCTs (NeoTORCH, TD-FOREKNOW), in 1 RCT (RATIONALE-315) censoring imbalance was significantly higher in the control arm (R-HR 1.30; 95% CI 1.02-1.65). Median SIFI was 12 (IQR, 4.25 - 20.25), with absolute weighted SIFI ≤ 2% in 3 RCTs (NADIM II, AEGEAN, Checkmate 816), suggesting a potential vulnerability affecting the significance of results. Table: 1232P

Conclusions

Significant censoring imbalance in the control arm and high survival fragility were reported in 1 and 3 RCTs, respectively, suggesting that informative censoring and survival vulnerability are useful tools to ensure a reliable interpretation of data.

Clinical trial identification

Legal entity responsible for the study

The authors.

Funding

Has not received any funding.

Disclosure

All authors have declared no conflicts of interest.