BACKGROUND Neonatal abstinence syndrome (NAS) due to opiate withdrawal may result in disruption of the mother-infant relationship, sleep-wake abnormalities, feeding difficulties, weight loss and seizures. Treatments used to ameliorate symptoms and reduce morbidity include opiates, sedatives and non-pharmacological treatments. OBJECTIVE To assess the effectiveness and safety of using an opiate, compared to a sedative or non-pharmacological treatment, for treatment of NAS due to withdrawal from opiates. METHODS The previous review was updated with additional searches of the Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library, Issue 1, 2005), MEDLINE (1966-December 2004) and EMBASE (1980-December 2004) supplemented by searches of conference abstracts and citation lists of published articles. METHODS Trials enrolling infants with NAS born to mothers with an opiate dependence, with > 80% follow up and using random or quasi-random allocation to opiate or control. Control could include an opiate, sedative or non-pharmacological treatment. METHODS Each author assessed study quality and extracted data independently. Primary outcomes included control of symptoms, seizure occurrence, mortality and neurodevelopment. Treatment effect was expressed using relative risk (RR), risk difference (RD), mean difference (MD) and weighted mean difference (WMD). Meta-analysis was performed using a fixed effect model. RESULTS Seven studies enrolling a total of 585 infants met inclusion criteria (Carin 1983; Finnegan 1984; Jackson 2004; Kaltenbach 1986; Kandall 1983; Khoo 1995; Madden 1977); however, two (Finnegan 1984; Kaltenbach 1986) may be sequential reports that include some identical patients. The studies enrolled infants of mothers who had used opiates with or without other drugs during pregnancy. Methodological concerns included the use of quasi-random rather than random patient allocation methods in three studies; sizeable, largely unexplained differences in reported numbers allocated to each group in four studies; and imbalances in group characteristics after randomisation in one study. Opiate (morphine) vs supportive care only: One study (Khoo 1995) found no significant effect on treatment failure (RR 1.29, 95% CI 0.41, 4.07), a significant increase in hospital stay (MD 15.0 days, 95% CI 8.9, 21.1) and significant reductions in time to regain birthweight (MD -2.8 days, 95% -5.3, -0.3) and duration of supportive care (MD -197.2 minutes/day, 95% CI -274.2, -120.3). Opiate vs phenobarbitone: Meta-analysis of four studies found no significant difference in treatment failure (typical RR 0.76, 95% CI 0.51, 1.11). One of these studies (Finnegan 1984) reported that opiate treatment resulted in a significant reduction in treatment failure among infants of mothers who had used only opiates; however, as this was a post-hoc analysis, this result should be interpreted with caution. One study (Jackson 2004) reported a significant reduction in duration of treatment and admission to the nursery for infants receiving morphine compared to phenobarbitone. One study (Kandall 1983) reported a reduction in seizures, of borderline statistical significance, with the use of opiate. Opiate vs diazepam: Meta-analysis of two studies found a significant reduction in treatment failure (RR 0.43, 95% CI 0.23, 0.80) with the use of opiate. No study reported neurodevelopment by allocated treatment group. CONCLUSIONS Opiates, as compared to supportive care only, appear to reduce the time to regain birth weight and reduce the duration of supportive care, but increase the duration of hospital stay; there is no evidence of effect on treatment failure. When compared to phenobarbitone, opiates may reduce the incidence of seizures but, overall, there is no evidence of effect on treatment failure. One study reported a reduction in duration of treatment and nursery admission for infants on morphine. When compared to diazepam, opiates reduce the incidence of treatment failure. A post-hoc analysis generates the hypothesis that treatment effects may vary according to whether the population includes infants born to all opiate users (i.e. with or without other drug exposure) or is restricted to infants of mothers who used opiates only. In view of the methodologic limitations of the included studies the conclusions of this review should be treated with caution.