Purpose: Patients with chronic non-malignant pain requiring a controlled-release opioid frequently move between available treatment options over time. This study compares the cost-effectiveness of therapeutic sequences involving the following controlled-release strong opioids: oxycodone controlled-release (CR), transdermal fentanyl, morphine CR and supportive care. Methods: Randomised trials, including any one of these therapies in malignant or non-malignant pain, were systematically identified in the literature. Reported pain scores were translated into utilities on the basis of an earlier ‘mapping' study between the EQ-5D and an 11-point pain rating scale in chronic low back pain. To enable direct comparisons between treatments, mixed comparison meta-analyses of derived utility results and adverse events from the trials were performed. These used hierarchical Bayesian models incorporating fixed effect intercepts and random treatment effects. Using the results of the evidence synthesis and treatment and adverse event cost estimates, a probabilistic decision model was developed to assess the cost-effectiveness of alternative sequences of therapies. Parameter uncertainty regarding utilities was incorporated into the model directly from the posterior distribution of the evidence synthesis; other parameters were incorporated as appropriate distributions. Results: A total of 57 trials were identified. Supportive care and morphine are the least costly therapies and, although there is considerable uncertainty surrounding the estimates, mean QALYs associated with oxycodone CR are higher than those for morphine CR. Fentanyl has lower mean QALYs than either oxycodone CR or morphine CR. All therapies have an incremental cost-effectiveness ratio, relative to supportive care, less than approximately £10,000 per QALY gained, suggesting that, as long as decision-makers are willing to pay up to this value, there will be a place for all therapies in a cost-effective sequence. The probabilistic analysis shows that, at very low willingness to pay thresholds, supportive care has the highest probability of being cost-effective as first in sequence. Over a threshold of about £10,000 oxycodone CR and fentanyl have higher probabilities of being cost-effective as first in sequence but, up to £40,000 per QALY, the probabilities for oxycodone CR and fentanyl remain appreciably below 0.5. Conclusions: As long as decision-makers are willing to pay at least £10,000 per QALY, all active therapies would be cost-effective in a sequence with the most cost-effective order being morphine CR, oxycodone CR, transdermal fentanyl and supportive care.