PARAMETRIC SIMULATION OF HEADACHE DAY FREQUENCY USING A NEGATIVE BINOMIAL DISTRIBUTION- A CASE STUDY OF ERENUMAB IN EPISODIC MIGRAINE
Author(s)
Porter JK1, Brennan A2, Palmer S3, Sapra S4, Cristino J1
1Amgen (Europe) GmbH, Zug, Switzerland, 2University of Sheffield, Sheffield, UK, 3University of York, Heslington, York, UK, 4Amgen Inc, Thousand Oaks, CA, USA
OBJECTIVES: Therapeutic goals with migraine prophylaxis are to reduce the frequency of migraine days and associated impact on patient disability and functioning. This analysis used data from a trial comparing erenumab to placebo to validate an approach to model the distribution of patients by headache day frequency where patient level data are not available. METHODS: A total of 64 weeks of data were available from the double-blind and open-label extension phases of the trial. Erenumab patient level data were used to fit a negative binomial (NB), Poisson and binomial distribution for each time point observed. Ordinary least squares (OLS) linear regression models were tested to estimate the NB dispersion parameters from the mean event frequency. The final OLS model was selected through visual inspection and goodness of fit measures. The distributions of patients in the placebo arm were simulated using the mean headache day frequencies at each time point and the estimated dispersion parameters. Goodness of fit of the NB, Poisson and binomial distributions was compared based on the total squared error relative to the trial observations for each cycle and through visual inspection comparing the predicted with the observed values. RESULTS: The squared errors for the NB distributions were consistently lower than the Poisson and binomial fits. Across the full 64 weeks, the total errors (average errors) were 0.1506 (0.009), 0.5493 (0.034) and 0.7003 (0.044) for the NB, Poisson and binomial distributions, respectively. NB distributions also consistently provided a better visual fit to the observed values. CONCLUSIONS: The use of NB distributions with estimated dispersion parameters provided estimates with lower errors in comparison with Poisson or binomial fits to the distribution of patients by headache day frequency. This approach provides a solution where distribution parameters cannot be observed directly, and facilitates the modeling of comparisons for which patient level data are not available.
Conference/Value in Health Info
2016-10, ISPOR Europe 2016, Vienna, Austria
Value in Health, Vol. 19, No. 7 (November 2016)
Code
PRM23
Topic
Clinical Outcomes, Methodological & Statistical Research
Topic Subcategory
Clinical Outcomes Assessment, Confounding, Selection Bias Correction, Causal Inference, Modeling and simulation
Disease
Neurological Disorders