OBJECTIVES To investigate the clinical effectiveness and cost-effectiveness of naltrexone for relapse prevention in detoxified formerly opioid-dependent individuals compared with any strategy that does not use naltrexone, including treatment with placebo, other pharmacological treatments, psychosocial interventions or no treatment. DATA SOURCES Major electronic databases were searched from inception to September 2005. REVIEW METHODS Selected studies were screened and quality assessed. Meta-analyses were carried out as appropriate. A decision-analytic model using Monte Carlo simulation was developed that compared naltrexone as an adjunctive therapy to no naltrexone. It assumed compliance rates that were not enhanced by contingent management rewards (because this is current UK practice). Utility values could not be identified from the literature and so were obtained by research specially commissioned from the Value of Health Panel. RESULTS The methodological quality of the 26 randomised controlled trials (RCTs) that met the inclusion criteria was poor to moderate. The results suggest that naltrexone as maintenance therapy may be better than placebo in terms of retention in treatment, but this was not statistically significant. A meta-analysis of seven included RCTs gave the relative risk (RR) of loss of retention in treatment in the naltrexone arm as 0.94. The pooled hazard ratio (HR) reported in five of the RCTs for retention in treatment data followed up to 35 weeks was calculated as 0.90 in favour of naltrexone and also did not reach statistical significance. The risk of drug abuse in naltrexone versus placebo, with or without psychological support given in both arms, gave a pooled RR of 0.72, which was a statistically significant difference in favour of naltrexone. The pooled HR from three RCTs for opioid relapse-free rates was significantly different from placebo in favour of naltrexone 0.53; however, this fell off over time and may be of limited clinical significance. The RR of reimprisonment while on naltrexone therapy showed results in favour of naltrexone in the combined two studies of parolees or people on probation, but the number of participants was small. One study of 52 participants found that the difference in improvement score for risky sexual behaviour in the naltrexone group compared with the placebo group was not statistically significant. The adverse events data reported showed no significant difference between the naltrexone and placebo arms. The quality of the nine RCTs of interventions designed to increase retention with naltrexone was poor to moderate; however, all three different modalities of enhanced care showed some evidence of effectiveness. All of the contingency management programmes used incentive vouchers; the mean duration of treatment retention was 7.4 weeks for the contingency management intervention compared with 2.3-5.6 weeks for the naltrexone treatment alone. The mean length of time for which patients stayed on naltrexone was 84-103 days with additional psychosocial therapy compared with 43-64 days for the control group. In trials with added pharmacological agents the RRs of stopping treatment were 1.63 at 6 months and 1.31 at 12 months (in favour of naltrexone plus fluoxetine). It became statistically significant at 6 months, but not at 12 months. A meta-analysis of the RR of stopping treatment at week 12 (the minimum follow-up period) was carried out using six of the nine studies. The pooled RR of stopping treatment was 0.81. The results indicated that overall the intervention groups had 19% fewer patients who stopped treatment compared with the control group, but there was only a small number of studies and their quality was relatively poor. No existing economic evaluations were identified. The point estimate for the cost-effectiveness of naltrexone was pound42,500 per quality-adjusted life-year (QALY). Sensitivity analysis was carried out and the incremental cost-effectiveness ratio varied between pound34,600 and pound42,500 per QALY gained. CONCLUSIONS Following successful withdrawal from opioids, naltrexone may be administered on a chronic basis to block any future effects of opioids. Naltrexone appears to have some limited benefit in helping formerly opioid-dependent individuals to remain abstinent, although the quality of the evidence is relatively poor and heterogeneous. The limited quality and extent of the studies precluded an analysis of subgroups likely to benefit from naltrexone prescribing. Oral naltrexone is used infrequently in current UK practice, and this review suggests that this is appropriate as there is little evidence to support its wider implementation. There is an important deficit in information about the quality of life of people who use illicit opioids and this would perhaps be a worthwhile area of research in informing policy questions about the cost-effectiveness of different programmes and interventions.