OBJECTIVES: Complex survival patterns commonly arise in randomized clinical trial (RCT) data, for example in immuno-oncology therapies, which typically lead to heterogeneity in patient response and hence long-term survivorship in a proportion of the study population. Such effects are often poorly represented by standard parametric models (SPMs), which consequently tend to yield unreliable short- and long-term extrapolations that are used in clinical- and cost-effectiveness analyses for health technology assessment (HTA) submissions, especially when RCT event data is limited. Advanced parametric models (APMs) have the potential to allow for more reliable survival projections even with limited patient follow-up, and hence may facilitate accelerated evaluation of treatment efficacy.

METHODS: We have conducted a targeted literature review and applied selected APMs to RCT survival data in several immunotherapy indications. These APMs include methods that explicitly account for heterogeneous responses of patient subpopulations, using either additional trial data (landmark response models; LRMs) or latent allocation (parametric mixture models; PMMs), and methods that formally incorporate external data (e.g., Bayesian multi-parameter evidence synthesis; B-MPES).

RESULTS: APMs may allow for robust modelling and reliable forecasts of complex survival trends that cannot be accurately represented by SPMs, via increased flexibility arising from additional parameters and by leveraging supplemental data.

CONCLUSIONS: Modelling approaches featuring extra parameters that are not informed by supplemental sources (e.g., frequentist PMMs) are liable to yield high uncertainty in resulting predictions and are therefore of limited use. Hence, it is desirable to formally integrate further information into the model, for example via a Bayesian framework. When long-term external data is used to inform parametric models (e.g., B-MPES), lifetime extrapolations are comparatively stable with successive follow-up. APMs are dependent on the availability and appropriate implementation of relevant supplemental data, and should therefore be formulated carefully, with verification via justification of clinical assumptions, scenario analyses, and uncertainty quantification.