Most women using heroin are of reproductive age with major risks for their infants. We review clinical and experimental data on fetal, neonatal and postnatal complications associated with methadone, the current "gold standard", and compare these with more recent, but limited, data on developmental effects of buprenorphine, and naltrexone. Methadone is a mu-opioid receptor agonist and is commonly recommended for treatment of opioid dependence during pregnancy. However, it has undesired outcomes including neonatal abstinence syndrome (NAS). Animal studies also indicate detrimental effects on growth, behaviour, neuroanatomy and biochemistry, and increased perinatal mortality. Buprenorphine is a partial mu-opioid receptor agonist and a kappa-opioid receptor antagonist. Clinical observations suggest that buprenorphine during pregnancy is similar to methadone on developmental measures but is potentially superior in reducing the incidence and prognosis of NAS. However, small animal studies demonstrate that low doses of buprenorphine during pregnancy and lactation lead to changes in offspring behaviour, neuroanatomy and biochemistry. Naltrexone is a non-selective opioid receptor antagonist. Although data are limited, humans treated with oral or sustained-release implantable naltrexone suggest outcomes potentially superior to those with methadone or buprenorphine. However, animal studies using oral or injectable naltrexone have shown developmental changes following exposure during pregnancy and lactation, raising concerns about its use in humans. Animal studies using chronic exposure, equivalent to clinical depot formulations, are required to evaluate safety. While each treatment is likely to have maternal advantages and disadvantages, studies are urgently required to determine which is optimal for offspring in the short and long term.