Analytical approach:
A Markov model was constructed to assess the costs and benefits of tapentadol or oxycodone as the second treatment for chronic pain, over one year, using a 28-day cycle. Severe adverse events or lack of efficacy led to third and fourth treatments. The authors stated that the perspective was that of the UK NHS.
Effectiveness data:
The main effectiveness data were the transition probabilities: withdrawal due to adverse events or lack of efficacy, adverse events without withdrawal, and no withdrawal and no adverse events. With effective treatment, the reduction in pain was assumed to be the same for each drug. For tapentadol and oxycodone, the probabilities were from a pooled analysis of data from three randomised controlled trials of almost 3,000 patients, most of whom (87%) had severe pain at the start. The trials lasted 112 days, so the average probabilities in the fourth cycle were extrapolated to the remaining cycles. The probabilities for third and fourth treatments were from open-label studies identified by a systematic review; the probabilities were only available for the whole study period, so it was assumed that they remained constant.
Monetary benefit and utility valuations:
Utilities were applied to each health state based on adverse events and withdrawal status. EQ-5D utilities were obtained from the three trials of tapentadol and oxycodone. It was assumed that the utility values did not depend on the opoid administered, due to insignificant differences between tapentadol and oxycodone in these trials. The weighted average for the utility for each state was calculated at the end of the trials.
Measure of benefit:
The measure of benefit was quality-adjusted life-years (QALYs) gained.
Cost data:
The costs included opioid therapy, physician visits, co-medication and treatment for adverse events. The costs of physician visits, adverse events and co-medication were from a UK medical database and a cohort study. As the severity of adverse events could not be distinguished, the average cost for all adverse events was used. Administration and co-medication costs were assumed to be the same for all opioids. Due to lack of data on fourth-line treatments, these costs were assumed to be the same as the average costs of third-line therapy. Drug costs were from the British National Formulary and the Monthly Index of Medical Specialities. The costs were reported in 2009 to 2010 UK £ and were appropriately inflated.
Analysis of uncertainty:
Probabilistic sensitivity analysis was conducted to evaluate the impact of uncertainty in all of the model parameters on the results. The results were presented on a cost-effectiveness plane and as a cost-effectiveness acceptability curve of the likelihood that the intervention was cost-effective at different willingness-to-pay thresholds. Deterministic sensitivity analyses were conducted to evaluate the impact on the results of varying individual model parameters. The costs were varied by ±50%, and utilities and probabilities were varied by ±20%. Different methods of extrapolating the probabilities beyond the end of the trial were evaluated. The subgroup of patients who had severe pain at the start, was analysed.