OBJECTIVES

The optimal treatment for patients with previously untreated advanced/metastatic renal cell carcinoma has been based on the IMDC/MSKCC prognostic risk scores. Recently, treatments with differing mechanisms of action have been approved for patients with intermediate-/poor-risk disease, for which indirect treatment comparisons in this patient population are warranted. This study aims to evaluate the impact of clinical heterogeneity on the validity of the intermediate-/poor-risk network meta-analysis for progression-free survival by following the feasibility assessment framework of Cope et al. (2014), based on the ISPOR Good Practices Task Force Report.

METHODS

A systematic literature review was performed, after which a network of evidence was constructed for the intermediate-/poor-risk population including 11 randomized clinical trials and 10 treatments of interest. First, heterogeneity was analyzed in terms of treatment and outcome characteristics, as well as study and patient characteristics, using a pre-defined set of potential treatment effect modifiers: ECOG performance status, gender, IMDC/MSKCC risk factor, liver metastasis, number of metastatic sites, prior nephrectomy, PD-L1 status, race, and region. Second, direct comparisons in terms of baseline risk and observed treatment effects were evaluated for the presence of unmeasured effect modifiers.

RESULTS

Imbalances were found in prognostic risk score (intermediate vs. poor) and their measurement (IMDC vs. MSKCC), nephrectomy, and radiation. In 2 studies, only outcomes for the intermediate-risk population were reported. Furthermore, there was a lack of reporting for most of the 9 pre-specified potential treatment effect modifiers. Only one treatment comparison in the network considered evidence from multiple studies, between which no significant differences were found.

CONCLUSIONS

Eliminating heterogeneity in the evidence caused by prognostic risk score is advised. We recommend excluding studies reporting outcomes only for the intermediate-risk population, as well as meta-regression to adjust for imbalances across all trials, where possible.