PARAMETRIC MODELLING OF MIGRAINE DAY FREQUENCY IN MIGRAINE PREVENTION- A CASE STUDY OF ERENUMAB CLINICAL TRIAL DATA
Author(s)
Porter JK1, Di Tanna GL1, Villa G1, Brennan A2, Palmer S3, Lipton RB4, Jansen JP5, Shah N6, Desai P6, Sapra S6
1Amgen (Europe) GmbH, Zug, Switzerland, 2University of Sheffield, Sheffield, UK, 3University of York, Heslington, York, UK, 4Albert Einstein College of Medicine, Bronx, NY, USA, 5Precision Health Economics, Los Angeles, CA, USA, 6Amgen Inc, Thousand Oaks, CA, USA
OBJECTIVES: The primary efficacy outcome of most clinical trials of migraine preventive drugs is change in migraine day frequency per 28 days (monthly migraine days [MMD]) relative to baseline for active treatment versus placebo. This analysis assesses parametric models of change in MMD for migraine preventives using data from erenumab clinical studies, with the objective of capturing intra- and inter-patient variability so trial observations can be modelled parametrically for the purposes of economic evaluation of migraine prophylactics. METHODS: MMD observations from the double-blind phases of two studies of erenumab were used; one in episodic migraine (EM; NCT02456740), and one in chronic migraine (CM; NCT02066415). Data from the placebo groups were used to fit non-linear mixed-effects multilevel regression models to the changes in MMD over time, using the statistical software Stata. Negative binomial and beta-binomial models were fitted in each population using count-panel data, which allows better estimation of changes in MMD. These models have previously been shown to be suitable models of MMD frequency, and provide estimates of the distribution parameters that quantify the dispersion of the data. These parameters are assumed constant over time. Goodness-of-fit was assessed by the root mean squared error (RMSE) of the estimated values compared to trial observations, and visual inspection of the predicted distributions. RESULTS: Compared to the MMD distributions observed in the trials, the negative binomial and beta-binomial regressions generate RMSEs of 0.075 and 0.102, respectively, for EM, and 0.082 and 0.081 for CM. Predicted values showed a good visual fit to the trial observations through all the time points considered. CONCLUSIONS: Modelling MMD with regression models that accommodate over-dispersion in a longitudinal framework is a statistically valid method to model the variation in MMD both within and between individual patients. This has important applications in the economic evaluation of preventive medications and policy decisions in migraine.
Conference/Value in Health Info
2017-11, ISPOR Europe 2017, Glasgow, Scotland
Value in Health, Vol. 20, No. 9 (October 2017)
Code
PRM19
Topic
Clinical Outcomes, Methodological & Statistical Research
Topic Subcategory
Clinical Outcomes Assessment, Confounding, Selection Bias Correction, Causal Inference, Modeling and simulation
Disease
Neurological Disorders