A process to validate prognostic factors for unanchored matching-adjusted indirect comparison of single-arm trials in oncology: a proof-of-concept study

<b>Aim:</b> The choice of covariates in unanchored matching-adjusted indirect comparisons (MAICs) of single-arm cancer trials with time-to-event outcomes remains a challenge. Currently, there is a lack of a systematic approach for validating the selection of covariates for bias reduction in unanchored MAIC. <b>Materials & methods:</b> This study proposes a validation framework to evaluate the appropriateness of selected prognostic factors before their use in unanchored MAIC. The process involves identifying potential prognostic factors from individual patient data and calculating risk scores using the prognostic factors with regression; artificially creating two groups that are unbalanced in risk such that a predetermined hazard ratio (HR) between the two groups is achieved; creating weights based on the prognostic factors; running a re-weighted Cox regression to assess the HR, the value of which should suggest balanced risks across groups to indicate the sufficiency of prognostic factors being included. We also conducted a proof-of-concept analysis using a simulated dataset to showcase this process. <b>Results:</b> The process successfully stratified the sample into two risk groups with a pre-determined HR of 1.8. When all covariates were included in the weighting, the HR was 0.9157 (95% CI: 0.5629-2.493), which was close to one. When one of the critical prognostic factors was omitted from the covariates, the HR became 1.671 (95% CI: 1.194-2.340), which was significantly different from one. <b>Conclusion:</b> Filling a gap in the existing evidence synthesis literature, the study introduces a structured data-driven approach for covariate prioritization in unanchored MAIC. The process may be a useful tool for quantitative covariate selection.