ASSESSING THE CONTRIBUTION OF PATIENTS' PROXIMITY TO DEATH IN ECONOMIC EVALUATIONS IN PATIENTS WITH HER2+ METASTATIC BREAST CANCER: AN EVENT BASED ANALYSIS

Purpose:

   In the economic evaluation of health technologies, health-related quality of life (HRQoL) is commonly expressed as health state utility values to derive quality adjusted life years. The objective of this study was to estimate and compare health state utility values using a standard disease progression model and those based on the time spent between the EQ-5D measurement and the patient’s death.

Method(s):

   We analysed data from a large (n=906), repeated measure (11,451 observations), EQ-5D dataset from the MARIANNE trial to estimate health state utility values. The EQ-5D-3L was converted into utility value using the UK tariff. Two mixed (random-coefficient) models using unstructured covariance structure were fitted to predict utility values according to baseline patient characteristics and key clinical outcomes.

   The set of variables considered for the multivariable mixed regression models included: baseline age, baseline ECOG performance status (ECOG PS), baseline disease stage, treatment, gender, visceral disease, time (visit), disease progression, hospitalisation due to adverse event (AE) and time leading to death. Final set of variables were selected using likelihood ratio test (p<0.05).

Result(s):

   The median follow up visit was 9 months (range: 0-48 months). Time was included as a random effect. Included variables demonstrated evidence of an important association with HRQoL outcomes based on magnitude and significance of effect (p<0.05). Disease progression was forced into the model for comparison purposes.

   Lowess smoother showed no trend of utility values leading to disease progression as opposed to a decline observed in time leading to death. Model 1 showed disease progression was associated with a statistically non-significant utility gain of 0.02. In contrast, Model 2 showed a substantial statistically significant utility decrement of 0.27 in the 6 months leading to patient’s death. Lastly, hospitalisation due to AEs and ECOG PS was associated with utility decrement of 0.03 and 0.1, respectively.

Conclusion(s):

   An event based analysis using a mixed model better explained variation in EQ-5D data according to key clinical outcomes and patient characteristics. Our analysis suggests that for patients with HER2+ mBC, time leading to death as opposed to disease progression may be a considerable determinant for HRQoL and should be accounted for within economic evaluations. This method also isolated the temporary loss in utility associated with hospitalisations due to AEs.