The use of first-generation antipsychotics does not appear to result in a major increase in the risk of congenital malformations. Some early retrospective studies reported cases of malformation, thus far unconfirmed, particularly with phenothiazines. Data remain rare (about 500 cases of exposed pregnancies have been published) (for review, see Galbally et al. 2014). It should be borne in mind that the occasional prescription of these drugs for nonpsychiatric problems during pregnancy (e.g. phenothiazines for pregnancy-related vomiting) (Reis and Kallen 2008) may bias the interpretation of these data.

In the last two trimesters of pregnancy, maternal hypotension can cause inadequate placental perfusion (Pinkofsky 2000). Increased hospitalization rates have been described for neonates, and paediatric monitoring is recommended (Sutter-Dallay et al. 2015). No valid long-term data about potential behavioural teratogenesis are currently available.

The use of first-generation antipsychotics does not seem to result in any noticeable metabolic disorders, although sporadic results report the appearance or aggravation of gestational diabetes (Reis and Kallen 2008).

The effects of second-generation antipsychotics on embryonic and fetal development have not yet been studied sufficiently to enable the development of clear guidelines. Reis and Kallen (2008) and Nulman (2014) examined all pharmacological classes together and found a significant global increase in major congenital malformations. Drug by drug, the data regarding olanzapine (the most frequently studied) and risperidone do not show any significant increase in the risk of congenital malformations (Ennis and Damkier 2015; Galbally et al. 2014; Gentile 2010b). Aripiprazole is not indicated for first-line use because of the risk of malformation found in animal studies and the dearth of human data (Ennis and Damkier 2015; Galbally et al. 2010). Among all antipsychotics, quetiapine has the lowest level of placental passage, and no teratogenic risk has been described, although sample sizes have been small so far (Ennis and Damkier 2015; Galbally et al. 2014). Lastly, limited published data to date concerning amisulpride are available (Gentile 2010b). The CRAT (French Reference Centre on Teratogenic Agents; www.​lecrat.​org/​) does not recommend discontinuing treatment unless an alternative can be prescribed.

Babies whose mothers were treated with atypical antipsychotics during pregnancy appear to be overweight more frequently, independently of any maternal condition (Newham et al. 2008); this is a risk factor for obesity, cardiovascular disease, and diabetes in adulthood. They are also born preterm more frequently (Nulman 2014). A few sporadic cases of neonatal complications have been reported after antenatal exposure to risperidone, notably withdrawal and extrapyramidal syndromes (Galbally et al. 2010). The literature reports “perinatal complications” in approximately 15 % of cases of antenatal exposure to olanzapine, but no data are available as to their type, severity, or outcome of these complications (Gentile 2010b).

Effects of antenatal exposure to second-generation antipsychotics on later child development remain unknown. To our knowledge, a single retrospective study has reported possible psychiatric and neurological effects in adolescents exposed to antipsychotics in utero (Aagaard and Hansen 2010).

A risk of metabolic complications during pregnancy, particularly gestational diabetes, has been described, as well as low birth weight and caesarean delivery (Galbally et al. 2010, 2014).

In women who wish to breastfeed, the need to be continuously available to the baby leads to maternal sleep deprivation; moreover, breastfeeding can be a stress factor in itself (e.g. implementation of breastfeeding is sometimes difficult, uncertainty about quantities to be given). These factors should be taken into account in the benefit/risk discussion about treatment with antipsychotics during breastfeeding, above and beyond the direct effects of these drugs on the neonate.

Breastfeeding sometimes exposes babies to the risk of overdose, because of the relatively long elimination half-life of most of these medications, together with the baby’s immature liver and kidney functions (Gentile 2004). This is true for amisulpride: passage into the milk results in relative doses to the infant of 10 % of the mother’s dose (http://​www.​lecrat.​org/​). Haloperidol, olanzapine, quetiapine, and risperidone are excreted into breast milk in much smaller quantities, and relative doses to the infant remain under 3 % (Hale and Rowe 2014; Moretti 2009).

Lastly, it is important to remember that the dopaminergic agonists generally prescribed to prevent milk production for women who choose not to breastfeed should not be used in these situations, because they can induce decompensation, especially in psychotic disorders (ANSM 2013; Misdrahi et al. 2006). The alternative is to use non-medicinal measures that do not prevent milk production but keep it to a minimum (moderate water restriction, local anti-inflammatory agents, breast compression).

If decompensation requiring antipsychotics occurs during the perinatal period, whether a first episode or a relapse, rapid treatment with medication is always indicated. Previously effective treatment, if there has been one, should be continued or restarted.

Lithium is teratogenic in invertebrates at high doses. After the alarming findings by Schou et al. in the 1970s, its effects on humans are now known to be less severe than previously feared. Exposure to lithium in utero appears to be correlated with an increased frequency of cardiac defects (Bergink and Kushner 2014; Diav-Citrin et al. 2014; Giles and Bannigan 2006), but the difference no longer appears significant once anomalies that resolve spontaneously postpartum are excluded (Diav-Citrin et al. 2014). The latest meta-analysis showed no formal evidence of a link between lithium intake during pregnancy and cardiac malformations (McKnight et al. 2012). Finally, it appears impossible to conclusively assert or refute any specific relation between lithium and Ebstein’s anomaly (Diav-Citrin et al. 2014; Giles and Bannigan 2006). It is therefore recommended that all women receiving lithium during pregnancy should undergo fetal echocardiography and a level-2 (anatomic scan) ultrasound examination (Bergink and Kushner 2014).

Lithium use during the fetal period is likely to lead to a significant increase in birth weight (Diav-Citrin et al. 2014). Paediatricians should carefully monitor babies for CNS and neuromuscular complications (Newport et al. 2005). The recommendation is therefore to refer these patients to maternity hospitals with neonatal paediatric resources. The mother’s lithium blood levels should be maintained as low as possible, based on her history. The 2007 NICE guidelines advise clinicians to monitor lithium levels in maternal blood monthly from the 20th week of gestation and weekly beginning 4 weeks before delivery.

Few studies describe risks of long-term neurological, cognitive, and behavioural teratogenesis. Results at 5 (Schou et al. 1973) and 15 years of age (Van Der Lugt et al. 2012) suggest no distinctive features of these babies as they mature.

Data about breastfeeding and lithium use are sparse. The few existing studies report high blood levels in infants, up to 17 % of maternal levels (Bogen et al. 2012). Because of the narrow therapeutic margin and the potential increase in lithium blood levels in the event of sodium depletion (heat, fever, diarrhoea, vomiting, etc.), which can be frequent in babies, breastfeeding is contraindicated.

Lithium is no longer really contraindicated during the first trimester of pregnancy and is certainly not in itself an indication for termination of pregnancy today. The risk of birth defects must also be considered in the light of the availability of ultrasound screening and the progress in paediatric cardiac surgery. Here, even more than with other psychotropic drugs, assessing the benefit/risk ratio for each patient is essential. Viguera et al. (2000, 2007) have emphasized the importance of mood stabilization during pregnancy when necessary: women who stop their lithium treatment during pregnancy have a risk of postnatal recurrence twice as high as non-pregnant patients, and this recurrence may occurs 4 times more rapidly and last 5 times longer. Discontinuation of lithium treatment must be progressive, even if the embryo is exposed, because the teratogenic risk is currently considered to be less than that of decompensation generated by abrupt cessation (Burt and Rasgon 2004). If treatment is continued throughout the pregnancy, the marked variation in blood volume and the increased rate of renal excretion during pregnancy make regular monitoring of maternal plasma and erythrocyte levels of lithium necessary. Progressive decrease in dose in the days preceding birth is recommended to avoid an overdose in the immediate postpartum period. Proper hydration must be maintained during labour to avoid neonatal overdosing. In the postpartum, daily assays of plasma and erythrocyte lithium should be performed and dosage adjusted until the appropriate balance is obtained.

Data about the use of carbamazepine during pregnancy mainly concern women with epilepsy; reports about women with psychiatric disorders are extremely rare, although this antiepileptic drug is indicated for bipolar disorder. It is not contraindicated for use during pregnancy for epileptic patients by either the Food and Drug Administration (FDA) or the European Medicine Agency (EMA). Nevertheless, studies have reported an increased risk of spina bifida in children exposed during the first trimester (Ernst and Goldberg 2002; Galbally et al. 2010) and a possible dose-effect on global major congenital malformation rates (Campbell et al. 2014). Nonetheless, a recent large population-based study found antenatal exposure to carbamazepine had no effect on this risk (Veiby et al. 2014).