In this prospective, observational comparative cohort study, 183 lithium-exposed pregnancies were followed up by the Israeli Teratology Information Service. First-trimester lithium exposure was associated with an increased risk (adjusted odds ratio=4.75 [95% CI=1.11–20.36]) of cardiovascular anomalies compared with nonteratogenic exposure. However, there were results that fell short of statistical significance for higher overall risk of major anomalies, higher risk of noncardiovascular anomalies, and higher risk of cardiovascular anomalies after excluding those anomalies that spontaneously resolved.
In the multicenter part of the study, 219 lithium-exposed pregnancies were followed up. Based on this multicenter analysis, first-trimester lithium treatment seems to be associated with an increased risk of major anomalies. The relative contribution of the three centers participating in the multicenter analysis (from Israel, Canada, and Australia) was not proportional to the country’s population. Both Canada and Australia have much larger populations than Israel. However, the time frame of data collection differed among the three centers, with the Israeli Teratology Information Service covering 16 years, the Australian MotherSafe service covering 11 years, and the Canadian Motherisk service covering 4 years. These countries also differ in birth rate, with Canada and Australia having lower annual birth rates compared with Israel. Additionally, there may be differences between the three countries in the prevalence of bipolar disorder, prescription practices, and the medico-legal environment, as well as cultural differences in taking medications, partially explaining the different input. The regression analysis of the multicenter part of the study showed that the Australian service was associated with a higher risk of major anomalies. This finding raises concern regarding the possibility of selection bias. The relative contribution of abnormal findings in a small group may result in overestimation of risk.
The risk of cardiovascular anomalies appears to be higher than reported in previous prospective studies (20–22) but still lower than reported in the retrospective Lithium Baby Register (16–18). The association between lithium exposure during pregnancy and Ebstein’s anomaly, previously reported in the literature, is supported by one prospective case and two retrospective cases in the present cohort. In the single-center part of the study, the absolute risk of overall major anomalies after lithium exposure during the first trimester of pregnancy was 6.5%, and the risk was 2.4% for cardiovascular anomalies that did not resolve spontaneously.
In the multicenter analysis, one fetus in the lithium-exposed group who was prenatally diagnosed with anencephaly and multiple anomalies also had been exposed to valproic acid (1,500 mg/day) during the first trimester, as well as venlafaxine and olanzapine. The anencephaly could potentially be explained by the exposure to valproic acid, although in human studies valproic acid has been associated mainly with lumbosacral neural tube defects (30–32). However, it is noteworthy that in a previous prospective study (21), two children in the lithium group had neural tube defects, one had hydrocephalus and meningomyelocele and also had been exposed to carbamazepine during the first trimester, and the other had spina bifida and tethered cord and had no additional drug exposures. A child with lumbar myelomeningocele after lithium exposure was reported in another prospective cohort study (20). Furthermore, there is a reported case of spina bifida with meningoyelocele published in the 1970s (16, 32,33). Some animal studies in rodents suggest that lithium salts are teratogenic and result in neural tube defects (9–11). Thus, five cases of neural tube defects (from the literature and the present study) in a relatively small group, although two with concurrent administration of other known teratogens associated with neural tube defects, raise the question of whether lithium exposure is also associated with neural tube defects or whether it potentiates the teratogenic effect induced by coadministration of valproic acid or carbamazepine.
The finding of a higher rate of cardiovascular anomalies in the bipolar group, relative to the lithium and nonteratogenic groups, that fell short of statistical significance may be a result of insufficient power of the relatively small sample size of the bipolar group or may be related to exposure to potential teratogens, such as valproic acid, as in the case of atrioventricular canal and clubfoot. Alternatively, this suggests that the risk might partly be attributed to the underlying bipolar disorder.
Further studies are needed to determine whether lithium is also associated with an increased risk for noncardiovascular anomalies. According to the logistic regressions, gestational age at initial contact with the information service center, maternal age, previous miscarriage, and exposure to lithium significantly predicted the risk of miscarriage. However, lithium exposure was the only significant predictor of cardiovascular anomalies.
Increased risk of preterm delivery among infants of women treated with lithium during pregnancy has been previously reported (34). The increased risk of preterm deliveries in lithium-exposed offspring may be associated with the underlying disorder. In another study, pregnant women with bipolar disorder had a twofold increase in the rate of preterm delivery compared with pregnant women with no history of mental illness (35). In the present study, birth weight was not higher in the lithium group, as suggested in a previous prospective study (21).
Our study has certain limitations and advantages. The study is based primarily on pregnancies of women who made contact with the Israeli Teratology Information Service, which may not represent the general population. However, we included two comparison groups, one unexposed to lithium but with psychiatric illness similar to that of the lithium group (although this comparison group was relatively small and occasionally exposed to potential teratogens) and also counseled by the Israeli Teratology Information Service and a second group counseled by the same service for nonteratogenic exposure using the same methodology of data collection. Detection bias is especially common when a concern about drug safety has been previously reported, as in our study. For example, offspring exposed in utero to lithium are more likely to be assessed for cardiovascular anomalies than unexposed offspring. To minimize this bias, standardized questionnaires were used in all groups; medical records were used for cases with congenital anomalies, when available; and the classification of anomalies was done by a pediatrician who was blinded to the exposure group. In addition, routine prenatal diagnosis is performed in Israel, Australia, and Canada. Detection bias is probably a more significant problem with congenital anomalies of lesser clinical significance than with severe anomalies resulting in serious clinical consequences. Unfortunately, the rate of general incidence of major anomalies by country statistics was not available for comparison. The study relies on maternal interview as the primary source for outcome data and lacks medical records in most cases. Physical examinations of neonates are routinely performed in the participating countries; however, the study lacks direct physical examination by an investigator as part of the protocol. Additionally, there is variation in timing of follow-up. Data are primarily from a single center. There is a multicenter part in which data were combined from three services, one in Israel, one in Australia, and one in Canada. The study design is nonrandomized, and it has limited power to detect specific rare defects. Randomized controlled trials are often not feasible in pregnancy because of ethical considerations. However, applying the same procedure to all arms of the study and its prospective nature minimizes these potential biases. Another advantage of the study is that data were available on elective terminations of pregnancies and stillbirths and were included in the analysis. Finally, the relatively large number of lithium-exposed cases gives reasonable power.
The conclusions drawn from observational epidemiological studies are often hampered by the problem of confounding. Such confounders can be found in maternal characteristics, such as age, parity, smoking, alcohol use, body mass index, ethnic background, concomitant drug exposure, and the underlying disorder (36). Selection and reporting biases, along with cognitive biases, are additional potential problems. In our study, logistic regression analysis was performed to assess the relative contribution of various available predictors to the differences in the rate of miscarriages and major anomalies. Maternal body mass index was only partially available and therefore not included as a predictor. In an attempt to address the problem of confounding by indication, comparison including a group of pregnant women with bipolar disorder and no exposure to lithium was conducted. Further studies are needed to accurately assess the risk associated with lithium treatment in pregnancy, as well as the risks associated with the underlying bipolar disorder.
When evaluating the risk-benefit ratio of lithium treatment in pregnancy, the risks associated with treatment discontinuation, as well as with alternative mood stabilizers, should also be considered. Discontinuation of mood stabilizer treatment, particularly abruptly, during pregnancy carries a high risk of morbidity in women with bipolar disorder (1, 37). Untreated or undertreated bipolar disorder during pregnancy may also increase the risk of poor pregnancy outcome and perinatal complications (38). In a prospective observational clinical study, the overall risk of at least one relapse of bipolar disorder in pregnancy was 71% (1). Women who discontinued mood stabilizer treatment, compared with those who continued treatment, had a twofold increased risk of recurrence, shorter time to first recurrence, and longer duration of recurrence. In a Swedish population-based cohort study, women with bipolar disorder, regardless of treatment with mood stabilizers, had increased risk of adverse pregnancy outcome, such as delivering a preterm infant (36). Infants of women with untreated bipolar disorder had increased risk of microcephaly and neonatal hypoglycemia. The risk of any congenital malformation in offspring of women with bipolar disorder in pregnancy in the Swedish study was 3.4% among treated women, compared with 1.9% among untreated women.
Lithium remains one of the mainstays for treatment of bipolar disorder, even during pregnancy. A clinical alternative, valproic acid, is both a human teratogen and a neurobehavioral teratogen and therefore is not recommended during pregnancy. Carbamazepine, another optional mood stabilizer, also carries a teratogenic risk. Human pregnancy experience with lamotrigine is generally reassuring; however, its effectiveness, especially in the treatment of mania, is questionable. In conclusion, the decision whether to continue lithium therapy during pregnancy should be made on an individual basis. The risks of lithium exposure in pregnancy need to be weighed against the risks of untreated bipolar disorder. Folate supplementation, with 5 mg daily before and during pregnancy, may be advisable for women who continue lithium during pregnancy to prevent neural tube and cardiovascular defects. Women who are treated with lithium during organogenesis should undergo fetal echocardiography in addition to level-2 ultrasound.