OBJECTIVES: In absence of head-to-head trials and common comparators, matching-adjusted indirect comparisons (MAICs) can be used to assess the efficacy of novel therapies against other standard of care (SoC). Unanchored MAICs are susceptible to systematic errors resulting from unobserved prognostic variables/effect modifiers. Often for the novel matched therapy no other trials are available, hampering outer sample validation (OSV). In that case, OSV might be performed on the placebo arm. The objective is to quantify this error using a new out-of-sample method, where the error is estimated on the placebo arm.

METHODS: We simulated five trials. One comparing placebo and active, three placebo-controlled trials with a different active and one trial for the SoC of interest. The error was estimated by matching the placebo of the first trial to the placebos of the three other trials. Weibull individual models were fitted over three comparisons (the matched placebo and aggregated placebo). The difference per trial between intercept and scale parameters of the matched placebo and the aggregated placebo was averaged. Subsequently, the first trial active arm was matched to the SoC arm of the fifth trial. The matched Weibull coefficients were adjusted for the residual error and the between study variance (BSV).

RESULTS: The first matched placebo showed an overestimation of the mean survival of 52%, the second an underestimation of 32% and the third an overestimation of 34%. After applying the average correction to the matched active arm, overall survival (OS) was reduced by 8%. The BSV was 0.13 (intercept) and 0.02 (scale).

CONCLUSIONS: Parametric survival models fitted on matched and aggregated placebo arms of reference trials can be used to adjust for systematic error resulting from MAICs on the active-arm index trial OS data. The method can also be applied if there are more trials on the active therapy.