Methodology of analysis of cognitive impairment in longitudinal randomized clinical trials in oncology: a scoping review.

PURPOSE: Cancer-related cognitive impairment (CRCI) is a common consequence of cancer treatments. Longitudinal randomized controlled trials (RCTs) are essential to evaluate cognitive outcomes over time, but methodological heterogeneity limits comparability. Accordingly, we systematically review cognition assessment in longitudinal oncology RCTs, focusing on study design, statistical methods, cognitive measures, and adherence to reporting guidelines. METHODS: We reviewed longitudinal RCTs published between 2010-2026 assessing cognition in adult cancer patients. Extracted data included sample size, assessment tools, statistical methods. Methodological quality of reporting was assessed using a subset of the CONSORT 2025 checklist. RESULTS: One-hundred and thirteen studies were included. Trials involved 21% central nervous system (CNS) tumours and 50% of trials focused on breast cancer. Overall, 73% assessed the effects of interventions on cognition, and 27% the impact of cancer or its treatments. Cognition was a primary endpoint in 66% of trials. Objective measures predominated, but International Cognition and Cancer Task Force-recommended tests were underused, and heterogeneity in scoring was observed. The overall RCT median sample size was 81 participants (range: 8-867). Statistical methods varied and were not mutually exclusive: 11% used descriptive analyses only; 24% baseline-adjusted/paired regression; 57% repeated measures; 11% other models. Adherence to reporting guidelines was limited, particularly for missing data (26%) and analysis sample (22%). CONCLUSION: Longitudinal cognitive assessments in oncology RCTs displayed heterogeneity in endpoints, tools, and statistical methods. Small sample sizes limit robustness of results. Future trials should prioritize primary cognitive objective, ensure adequate sample sizes, employ standardized cognitive assessments and appropriate statistical methods to enhance reproducibility and interpretability.