OBJECTIVES: In HTA, survival extrapolations using parametric models are frequently conducted to estimate long-term survival beyond observed trial data. However, these extrapolated survival probabilities are often associated with high uncertainty. NICE Technical Support Document 21 highlights that excess hazard modelling (EHM), which incorporates data from general population mortality, has the potential to reduce model uncertainty. The aim of this study is to review the use of EHM in survival analysis and assess its application in HTA.

METHODS: A search for “excess hazard” was conducted in MEDLINE, Embase, Web of Science and the Cochrane Central Register, supplemented by a pearl-growing approach using three “seed” papers. Articles were double screened for references to EHM and survival extrapolation. Data extraction was tailored to the type of article (methodology development, extrapolation of registry/randomized controlled trial [RCT] data).

RESULTS: The search yielded 352 unique articles, of which 28 were identified as relevant. The majority of articles were methodological papers (n=25) most of which used either simulation studies or registry data as examples. All of these studies demonstrated the benefit of including general population mortality into the survival model, through improved face validity or reduction in uncertainty at later time points. Only two studies presented examples using data from RCTs. One study only had access to aggregate RCT data and presented several methods of incorporating excess hazards, with different methods being more appropriate for different diseases.

CONCLUSIONS: Whilst there has been significant development and demonstration of EHM within survival analysis, its use within RCTs remains relatively low. It remains unclear how factors such as low sample sizes, immature data, and aggregate level data, all common within RCTs used in HTA, may affect survival extrapolations when using EHM. Further work on employing EHM approaches within the RCT setting is required for the robust implementation within HTA.