OBJECTIVES

HSUVs are a key component of economic modeling and health technology assessment (HTA). In oncology, HSUVs are typically linked to disease progression. However, in recent HTAs of IOs, the analysis of utility relative to a patient’s death (TTD) has gained prominence. We undertook a study by simulating IO trial data to quantify the impact of progression- and TTD-based approaches to assess HSUVs.

METHODS

Survival curves, follow-up duration, and patient numbers similar to those observed for IOs in lung cancer, melanoma, and renal cell carcinoma trials were simulated. As observed in these trials, utilities were assumed to decrease on progression or close to death (“alternative frameworks”). Simulated data were analyzed via general estimating equation regressions using TTD- or progression-based approaches. The mean error (ME) and mean absolute error (MAE) for intervention quality-adjusted life years (QALYs) using each approach were compared with the “true” values (which are known as a consequence of using simulated data).

RESULTS

With long-term data available (>5 years), both progression- and TTD-based approaches performed well when used in the corresponding framework (ME & MAE < 1%). Considering study designs typically seen with IOs (patient numbers of 80-400, study lengths of 12-54 months), the progression-based approach often had a high MAE and nonzero ME (even under the assumption of a progression-based framework). When a TTD-based analysis was specified in a progression-based framework, MAE increased by 7% to 17%, whereas when a progression-based analysis was specified in a TTD-based framework, MAE increased by 44% to 81%. In both cases, QALYs were underestimated.

CONCLUSIONS

HSUV estimation methods require justification based on the study designs seen with IOs as an inappropriate analysis framework results in bias and error. Even with a correctly-specified framework, this bias and error stem from IO trial features, such as relatively short follow-up times and limited observations beyond progression.