Analytical approach:
This economic evaluation was based on a discrete event simulation model which predicted the costs and benefits of the alternative treatments based on efficacy, discontinuation rate and adverse events. The time horizon of the analysis was one year and the authors stated that the perspective of the German health insurance system was adopted.
Effectiveness data:
The clinical data appear to have been derived from a selection of known, relevant studies. For example, key clinical data on the efficacy of the two treatments were derived from a recent head-to-head phase III, open-label, randomised controlled trial (RCT) of 138 patients with chronic, moderate-to-severe osteoarthritis pain. The demographic and clinical characteristics of the eligible patient population were obtained from German sources. Adverse event data were taken from a systematic review of 15 placebo controlled trials. The clinical experience of one of the study authors and a Delphi panel of experts were also used to determine some clinical aspects of the model. The key clinical outcome was improvement in the Sleep Problems Index-II (SPI) score with the two treatments.
Monetary benefit and utility valuations:
The utility valuations were derived using the Short Form 6D questionnaire from a previously published study. SPI values found in the clinical trials were translated into utility weights using a regression.
Measure of benefit:
Quality-adjusted life-years (QALYs) were used as the summary benefit measure.
Cost data:
The health services were medications, doctor visits (both specialists and general practitioners), and treatment options for dissatisfied patients (including other drugs, steroid injections, and surgical procedures). The unit costs and quantities of resources used were presented separately. The resource use was based on guidelines and German legislation. The drug costs were derived from the Rote Liste, while other costs came from official prices. All costs were in Euros (EUR) and the price year was 2005.
Analysis of uncertainty:
Univariate sensitivity analyses were carried out on all model inputs using published or author-chosen ranges. Scenarios were also run for the proportion of opioid naive patients and prior users of an opiate. In addition, a probabilistic sensitivity analysis was carried out by varying multiple parameters simultaneously. The types of parameters investigated and the types of probabilistic distributions used were reported.