Premium
Cusums to Measure Chronic Daily Headache
Author(s) -
FreidankMueschenborn Edda,
Fox Anthony W.
Publication year - 2006
Publication title -
headache: the journal of head and face pain
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.14
H-Index - 119
eISSN - 1526-4610
pISSN - 0017-8748
DOI - 10.1111/j.1526-4610.2006.00299.x
Subject(s) - medicine , cusum , psychological intervention , confidence interval , physical therapy , statistics , psychiatry , mathematics
Objective.—To investigate whether cumulated summed differences (cusums) can be used as a method for detecting effective interventions in chronic daily headache (CDH). If so, then can such interventions be detected sooner than 28 days? Background.—CDH probably represents the greatest current challenge in the field of headache treatment. Clinical trial methodologies for its study are not standardized but usually use fixed 28‐day observation periods. Similarly, 28‐day observation periods are common review intervals for stepped‐care patient management strategies. Methods.—A theoretical patient was modeled, with daily headache scores that had maximal variation for 28 consecutive daily measures. The cusums for this “norm” or control data set were plotted and correlated with time using a simple spreadsheet. Departure of a postintervention cusum from the 95% confidence interval of the control data set was chosen as an indicator of sensitivity to various model perturbations that had clinical correlates. Results.—When 1 to 17 randomly distributed headache‐free days were used to perturb the 28‐day model, cusums consistently diverged from the norm. If at least 3 headache‐free days occurred, then the cusum detected the intervention in less than 28 days of observation. Reanalysis of a previously published study of magnesium oxide suggests that its negative conclusions should be reconsidered. Conclusion.—Cusums can be useful tools for detecting change in CDH, whether at the level of the clinical trial or when managing patients using stepped‐care treatment algorithms.