1061P - Assessing differential informative censoring in control and experimental arm in trials testing immunotherapy in metastatic cancers: A systematic review

Background

The Kaplan-Meier (K-M) method assumes that informative censoring is equally distributed in arms of clinical trials. However, the number of censored patients vary between study arms, ultimately affecting trial results. Therefore, we investigated the rates of informative censoring in randomized controlled trials (RCTs) of immunotherapy (IO) in advanced cancers.

Methods

We searched articles of RCTs testing IO in advanced cancers, published up to 12/2023 in PubMed-indexed journals. For both progression free (PFS) and overall survival (OS) K-M curves, we collected: i) The number of patients at risk; ii) The rate of censored patients at the first study interval (T1); iii) The overall rate of censoring (T2). We calculated the unweighted absolute % difference of censoring, as well as the weighted difference adjusted for study enrolment size, in control versus experimental arm at T1 and T2.

Results

Of the 141 trials reviewed, censoring data at both T1 and T2 were found for 55/141 (39.0%) and 56/141 (39.7%) trials for PFS and OS K-M curves, respectively. Censoring data in either PFS or OS were not reported in 31/42 (73.8%), 9/13 (69.2%), 12/20 (60.0%) and 8/22 (36.4%) RCTs of IO in NSCLC, melanoma, genitourinary and gastrointestinal cancers, respectively. The median unweighted proportion of censored patients control and experimental arms were: i) at T1, 2.16% and 1.15%, for OS K-M); ii) at T1, 7.47% and 4.66%, for PFS K-M; iii) at T2, 31.37% and 38.37%, for OS K-M); iv) at T2, 23.50% and 26.13%, for PFS K-M). Furthermore, analysis of the weighted differences between control and experimental arms, revealed more censoring in control arms at T1 (OS: 1.02; PFS: 2.81) and more censoring in experimental arms at T2 (OS: -6.50; PFS: -2.99).

Conclusions

Our study found that many RCTs of IO in metastatic cancers did not clearly report data about informative censoring. As previously reported in other RCTs in oncology, the rate of censoring is higher in control arms at the start of the study and increases in the experimental arm over the course of the trial. Further studies are needed to elucidate the role of censoring on final survival results reported in RCTs.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Has not received any funding.

Disclosure

All authors have declared no conflicts of interest.