Phase specific analysis of herpes zoster associated pain data: a new statistical approach.

Herpes zoster or shingles is a frequent occurrence in both elderly individuals and immunocompromised hosts. The pain associated with herpes zoster is the most debilitating complication of the disease. It can be described as acute pain and post-herpetic neuralgia or zoster associated pain (ZAP). The latter definition encompasses pain from the onset of disease through its resolution and provides a convenient analytic tool for evaluation of antiviral therapy. A heuristic examination of ZAP historical data suggests the existence of three phases of pain resolution: the acute, subacute and chronic phases. The subacute and chronic phases comprise the post-herpetic neuralgia (PHN) stage. Common analytic methods, such as a Kaplan-Meier survival function or a Cox's model, have been used to assess the pain. However, such approaches do not adequately allow for phase comparison. Notably, in the clinical trial setting the comparison of specific treatment effects on the latter stages of pain are of the greatest medical relevance since this is the most debilitating phase of the illness. In order to incorporate the phase-specific information in the modelling of time to cessation of ZAP, we assumed the hazard function was a stepwise constant. Utilizing the full likelihood function, we obtained the maximum likelihood estimate for the transition times (that is, change-points), and other parameters of medical importance. The standard error of the change-point estimates were obtained through a bootstrapping method. The asymptotic properties of the parameter estimates are also discussed. Hence, the rates of pain resolution across all phases can be examined in order to precisely define the existence of multiple phases. In addition, the covariates effect can be examined across phases and populations, thereby allowing us to translate potential efficacy of a standard therapy to different populations. These results can be utilized in the design of clinical trials or in targeting the outcome for a specific phase while controlling for the effect of other variables. Copyright 2001 John Wiley & Sons, Ltd.