Modeling Monthly Migraine-Day Distributions Using Longitudinal Regression Models and Linking Quality of Life to Inform Cost-Effectiveness Analysis: A Case Study of Fremanezumab in Japanese-Korean Migraine Patients.

Abstract

As regression approaches have been used more recently to model the effectiveness and health-related quality of life (HRQOL) of novel migraine treatments, an example is provided for fremanezumab. The objective is to estimate the distribution of mean monthly migraine days (MMD) as a continuous variable and corresponding migraine-specific utility values as a function of the MMD, to inform health states in a cost-effectiveness model (CEM). Three longitudinal regression models (zero-adjusted gamma [ZAGA], zero-inflated beta-binomial [ZIBB], and zero-inflated negative binomial [ZINBI]) were fitted to Japanese-Korean clinical trial data of episodic (EM) and chronic migraine (CM) patients treated with fremanezumab or placebo, to estimate MMD over a period of 12 months. The EQ-5D-5L and the migraine-specific quality-of-life (MSQ), mapped to the EQ-5D-3L, questionnaires were used to measure HRQOL. Migraine-specific utility values were estimated as a function of MMD using a linear mixed effects model. The ZIBB models fitted the data best in estimating the distribution of mean MMD over time. MSQ-derived values were more sensitive than the EQ-5D-5L values for the effect of the number of MMD on HRQOL, with higher values for less MMD and more time on treatment. Using longitudinal regression models to estimate MMD distributions and linking utility values as a function is an appropriate method to inform CEMs and capture inter-patient heterogeneity. The observed distribution shifts demonstrated fremanezumab's effect at reducing MMD for both EM and CM patients, while treatment effect on HRQOL was captured by MMD and time on treatment.