A Hidden Markov Modelling Approach for Measuring Migraine Treatment Success

A Hidden Markov Modelling Approach for Measuring Migraine Treatment Success

We are developing a novel statistical modelling approach for measuring migraine headache treatment success. This novel statistical model will address long-standing concerns with how we evaluate the resolution of migraine attack symptoms and impacts.This work will entail what is known in the statistical literature as a Hidden Markov model approach (to be described in the subsection "Proposed Approach - Hidden Markov model" below). The novel research will include: development of models, inferential procedures, and diagnostics. It is aimed at a statistical audience that engages in applications with clinicians, neurologists, and headache specialists. Notably, the software tools developed to conduct this research will be released under a FLO (free/libre/open) license and made available to the public through a service such as Github.# Existing EndpointsMigraine therapies are evaluated using traditional endpoints of pain freedom, pain relief, and freedom from the most bothersome symptom. These traditional endpoints, whether employed in the regulatory review process to determine drug approval, or whether used in the field more broadly, are flawed measures of drug efficacy. They are flawed because they ignore relevant migraine symptoms and impacts. For example, a commonly reported metric for migraine therapy is percentage of subjects achieving pain freedom. This endpoint uses a single 4 category item (no pain, mild pain, moderate pain, severe pain) and dichotomizes it as "No pain" versus "Pain." Treatment success is defined as a subject reporting "No pain" at 2-hours post dose. This ignores other symptoms and impacts that are central to the definition of migraine, such as nausea, photophobia, phonophobia, and impairment of function. Therefore, pain freedom is an insufficient definition of treatment success. # Proposed Approach - Hidden Markov modelThis work explores an alternative approach for determining whether or not migraine treatment has been successful. The alternative approach relies on a Hidden Markov model. A Markov model is essentially a tractable way of modeling longitudinal data: in particular it allows us to include data collected at all time points during a trial rather than just the final follow-up. This allows us to evaluate how quickly a drug works and examine fluctuations in the symptoms over time. A Hidden Markov model is a type of Markov model that allows us the additional advantage of being able to make a classification of migraine using all measured symptoms and impacts. Specifically, the model uses all the primary and associated symptoms that define a migraine, namely pain, nausea, photophobia, and phonophobia, and can also incorporate the primary impact, i.e. functional disability. Neurologists' definition of migraine includes these symptoms and impacts, and the Hidden Markov model allows us to evaluate treatment efficacy holistically at various time points using those symptoms and impacts. This should yield an improved definition of migraine treatment success, one that is better aligned with clinician judgement.The purpose of this work is to both develop this novel modelling approach and demonstrate its application to particular migraine trials.

[{ "PostingID": 121, "Title": "GSK-TRX109011", "Description": "A Randomized, Double-blind, Double-dummy, Placebo-controlled, Crossover Study to Evaluate the Efficacy of TREXIMET® (Sumatriptan + Naproxen Sodium) vs. Butalbital-containing Combination Medications for the Acute Treatment of Migraine when administered during the Moderate-Severe Migraine Pain, 1 of 2" },{ "PostingID": 122, "Title": "GSK-TRX109013", "Description": "A Crossover, Randomized, Double-blind, Double-dummy, Placebo-controlled Study to Evaluate the Efficacy of TREXIMET®/Sumatriptan + Naproxen Sodium vs. Butalbital-containing Combination Medications for the Acute Treatment of Migraine when administered during Moderate-Severe Migraine Pain, 2 of 2" },{ "PostingID": 3010, "Title": "GSK-SUM30046", "Description": "A randomised, double-blind, single-attack, placebo-controlled, parallel group evaluation of the efficacy and tolerability of sumatriptan Fast Disintegrating Tablets (FDT) 50 mg and 100 mg versus placebo during the mild pain phase of a migraine attack." },{ "PostingID": 3011, "Title": "GSK-SUM30045", "Description": "A Double-Blind, Placebo-Controlled, Parallel Group Study To Evaluate Two Dose Levels (5mg And 20mg) Of Sumatriptan Nasal Spray In The Acute Treatment Of A Single Migraine Attack In Adolescent Migraineurs (12-17 Years Of Age)" }]