OBJECTIVES

To review existing correlation analyses evaluating candidate surrogate endpoints (SEs) for OS in advanced melanoma.

METHODS

A TLR was conducted in Embase and Medline to identify meta-analyses (MAs) reporting on SEs for OS published prior to October 2020. Study design, patient/clinical characteristics, study treatments, and correlations between SEs and OS were evaluated for evidence synthesis and comparison.

RESULTS

All seven MAs identified prior randomised clinical trials in advanced melanoma. Treatments included any immunotherapy (n=5), any systemic therapy vs dacarbazine (n=1), and immunotherapy or targeted therapy (n=1). Of four MAs reporting exclusively on unresectable disease, two reported correlation measures between the treatment effects on PFS and OS weighted for sample size and treatment effect precision (random or fixed effects modelling); Pearson’s correlation measure varied from 0.71-0.89 in one MA and coefficients of determination (R²) ranged from 0.72-0.82 in the other, depending on weighting. In both MAs, sensitivity analyses improved the correlation estimates; namely these were analyses limiting trials to phase III, trials with larger sample sizes, or removing or reducing trials allowing crossover. This TLR captured three MAs that included mixed populations (e.g., NSCLC) along with advanced melanoma; these showed more equivocal associations between PFS and OS. However, subanalyses to include only the advanced melanoma trials in one MA resulted in better correlation estimates between the treatment effects on PFS and OS as compared with the overall group (R² 0.61 vs 0.37). Beyond PFS, other candidate SEs included 1- and 2-year OS rates (n=2), and disease control rate and objective response rate (both n=2).

CONCLUSIONS

The review suggests PFS is a promising SE for OS in advanced melanoma. Additionally, the correlation estimates between PFS and OS improved with increased study sample size, restricting the evidence base to only phase III studies, or to studies with minimal or no crossover.