Determinants of trial completion, termination, and efficacy in solid tumor immunotherapy: Mapping IO clinical trial evolution from 2015-2022.

Abstract

e23326 Background: Clinical trial success depends not only on scientific rationale but also on trial design and strategic planning. We performed a large-scale landscape analysis of IO trials to identify factors associated with trial termination, endpoint success, and strategic shifts over time, aiming to inform more predictive and efficient trial design. Methods: We analyzed interventional solid tumor IO trials from ClinicalTrials.gov finished from 2015–2022. Trials were categorized by investigational treatment, tumor type, disease stage, treatment setting, phase, sponsor type, year of initiation and termination reason. Among completed trials, primary endpoint success or failure was assessed when outcomes were available. Statistical analyses included chi-square or Fisher exact tests and logistic regression. Results: Among 1,430 trials, 960 (67.1%) were completed and 470 (32.9%) were terminated. Termination increased over time from 23.4% (2015–2017) to 36.3% (2018–2019) and 59.1% (2020–2022) (aOR 4.39, 95% CI 3.14–6.13) p < 0.001). Investigational treatment affected termination rate (p < 0.01): PD-(L)1 + chemotherapy and PD-(L)1 monotherapy had the lowest (17.4% and 19.7%), compared with PD-(L)1 + targeted combinations, bispecific antibodies (bsAb), cell- and vaccine-based therapies combined (43.2%; p < 0.001, aOR 3.81). Termination and its reasons were associated with sponsors, being lower in industry-led trials (p = 0.04). Recruitment-related termination was more common in non-industry (41.2% vs 24.6%), and business for industry trials (38.5% vs 21.9%). Recruitment was the dominant reason for all studies ( > 60%) except for PD-(L)1 + targeted, where futility and business-related factors predominated (57%). Phase 1-2 trials had higher rate of termination than phase 3 (32.5% vs 22.4%; p = 0.009). Termination was higher in locally advanced/metastatic than in locoregional disease (37% vs 25%; p = 0.02). IO strategy mix shifted over time (p = 0.01). PD-(L)1 monotherapy declined from 13.3% to 4.7%, while targeted, bsAb, and other combinations increased from 20.6% to 32.3%. BsAb trials were heavily industry-sponsored (90.7%), while RT-IO and PD-(L)1 mono- were entirely non-industry driven (97.7% & 69.7%; p < 0.001). Among assessable completed (49.1% 471/960) trials, primary endpoint was met in 56.1% varying by tumor type (p = 0.002) (higher success in upper GI cancers 75.0%, basket trials 69.8%, NSCLC 67.2%, melanoma 62.7%, lower in CRC 34.5%, and non-TNBC breast cancer 28.0%). Conclusions: Since 2015 IO trials have shifted from late line monotherapy to earlier settings combination strategies with clusters of success in select tumors. Termination increased over time, strongly influenced by recruitment issues, IO strategy and sponsor. We provide a quantitative base for developing predictive models to guide future IO trial design and improve efficiency.