Congenital malformations

Over the first 2 decades after the introduction of fluoxetine into practice, both dysmorphologic and neurobehavioral human studies did not detect consistent risks of congenital malformations with the drug itself or with newer SSRIs and SNRIs. Because of the relatively limited sample size of the initial studies, the common statement that described the fetal safety of these drugs was that they “do not appear to constitute major human teratogens.”

This school of thought changed with a report of Glaxo Smith Klein to the Food and Drug Administration that noted an apparent 1.5-fold excess of cardiovascular malformations, primarily ventricular septal defect (VSD) and atrial septal defect among children who were exposed in utero to their drug paroxetine, which at that time (2005) was among the most commonly prescribed antidepressants. The data presented in that report, although originally not published in a peer review journal, has led the Food and Drug Administration and Health Canada to establish warnings that implied causation and that suggested to physicians to consider discontinuation of the drug or a decrease in its dose in late pregnancy to mitigate the risk of persistent pulmonary hypertension of the newborn infant (PPHN) or withdrawal syndrome.

Over the subsequent 6 years, >30 epidemiologic studies have been published, more negative than positive, that examined an association, or lack of, between SSRIs and heart defects, mostly VSDs ( Table ). In 2010, a metaanalysis that included most of these studies, but not all, concluded that there was an increased prevalence of combined cardiac defects with first-trimester exposure to paroxetine. For cardiac defects, the prevalence odds ratio was 1.46 (95% confidence interval, 1.17–1.82); for aggregated congenital defects, the prevalence odds ratio was 1.24 (95% confidence interval, 1.08–1.43).

A large prospective study with >1100 mother-child pairs who were exposed to paroxetine, which was not included in the metaanalysis by Wurst et al, found a 0.7% rate of cardiac anomalies, which was identical to the control unexposed group and to rates that were reported in the pediatric cardiology literature.

Many of these studies used administrative databases to link the prescription of SSRI to pregnant women with pregnancy outcome; relatively few of the studies interviewed mothers or independently examined the babies. In general, database linkage studies tend to have larger numbers of exposed women than prospective cohort studies, because they draw cases from large populations of prescribers. Because they were not designed to study drug safety during pregnancy, there are important methodologic issues in the interpretation of these studies that should be considered by the practicing obstetrician.

Some studies have found higher reported rates of VSDs with paroxetine; other studies found that SSRIs were “clean.” Some studies detected these effects with fluoxetine but not with paroxetine; some studies found VSDs with citalopram but not with the previous ones. In addition, some studies have found increased risk of other kinds of malformations such as neural tube defects and gastroschisis ( Table ). Many studies did not find teratogenic risk at all. This type of heterogeneity in outcomes does not support biologic plausibility because, when a drug causes fetal damage, one expects similar effects by other drugs of the same class (eg, angiotensin-converting enzyme inhibitors), which suggests that uncontrolled sources of bias among studies may cause these fluctuations in results.

As shown recently by Kallen et al with interviews of pregnant Swedish women who received SSRIs, many of the women decided not to take the medications, although in the administrative database studies they were counted as “prescribed” and hence, as “exposed.” These findings cast serious doubts on the validity of the use of these administrative databases without independent confirmation of actual exposure to the SSRI. In general, adherence to drug regimens for chronic illnesses lies at approximately 50% and has been found to be even lower for mental illnesses such as depression. Thus, antidepressant exposure is overestimated in prescription database studies. Pedersen et al tried to overcome this limitation by selecting only women to whom ≥2 prescriptions had been dispensed during pregnancy, thereby increasing the likelihood that these drugs were actually taken. In addition, the use of the date that the antidepressant was dispensed to calculate the timing of exposure may introduce an error in the database studies.

Women with depression and anxiety have a 2- to 3-fold higher likelihood of having ultrasound scans and echocardiograms than healthy women and therefore a much higher likelihood of the detection of cardiac anomalies, which are the most common congenital malformations. Moreover, women with depression and anxiety are significantly more likely than healthy women to attend emergency departments with their infants, where pediatricians may have more opportunity to detect a heart murmur and be able to follow the case. This source of ascertainment bias is further augmented by the widely publicized news that SSRIs causes malformations, which facilitates more babies of women receiving SSRIs to be examined with no similar urge to test children of healthy women. To make things even more complex, most muscular-type VSDs tend to close spontaneously in infancy, so unexposed children, if examined later, will not exhibit the VSD, thus spuriously creating a teratogenic signal.

Although most of the studies that were based on medical birth registries stated that the reporting of malformations to the registry was mandatory, no information on the completeness of the registration of cardiac malformations was presented in any of the studies that were based on medical birth registries. Malformations should ideally be ascertained by an expert who is blinded to the mother’s status of exposure. The case-control studies and teratogen information-services studies generally adopt this procedure.

A commonly asked question is whether any particular SSRI-SNRI is safer than the other when it comes to fetal risk. Presently, there is no compelling evidence for any of these drugs to be more or less safe. As a rule of thumb, medications that have been in use for longer periods (eg, fluoxetine) have larger series and longer experience and may be preferred by some physicians for that reason. Similarly, there is no compelling evidence that any particular agent is markedly superior in terms of effectiveness in treating maternal depression, so the choice is based on the physician’s experience and preference. Often, different SSRIs-SNRIs elicit varying adverse effects in patients, which is a common reason to switch from 1 medication to another.

PPHN

A case control study suggested that gestational exposure to SSRI confers an increased relative risk for PPHN, which is a finding that was confirmed in an animal model. Additional case control and cohort studies both corroborated or failed to show this association. Of importance, although PPHN typically results in a 5-10% mortality rate because of serious cardiac anomalies, diaphragmatic hernia, and meconium aspiration, to mention a few common causes, none of the available studies adjusted in their analyses for known causes of PPHN to arrive at the attributable risk of SSRI in causing this adverse event. Sadly, similar to the paroxetine scare, regulatory agencies accepted the association as causation, which suggests that women may need to reduce or taper off their SSRI dose at term, despite the unequivocal risk of postpartum depression among women who exhibit depression in late pregnancy. Such regulatory suggestions, without a bona fide risk:benefit assessment, are not evidence-based.

Even if the original estimates by Chambers et al are correct and causation will have been proved, SSRI would confer PPHN risk in <1 of 100 babies who are exposed to SSRI. Moreover, once the known potentially fatal causes of PPHN are adjusted for, it is likely that most cases with this condition will be self-limited and successfully treated in the neonatal intensive care unit. Hence, it is critical to ensure that the mother does not abruptly discontinue her therapy, thus increasing her vulnerability to postpartum depression.

Poor neonatal adaptation syndrome

Adults who abruptly discontinue their SSRI or SNRI therapy typically experience various degrees of withdrawal, which may be exhibited as nervousness, anxiety, sleeplessness, tremor, and seizures. Not surprisingly, infants who are exposed in late pregnancy to SSRI or SRNI have a 10-30% rate of poor neonatal adaptation syndrome, which often is proved to be associated with low systemic levels of the antidepressants. Although most signs of this neonatal syndrome resemble neonatal withdrawal from opioids, benzodiazepine, and ethanol, in some cases SSRI withdrawal is associated with respiratory failure that necessitates short-term respiratory support, which is unique to the SSRI and SNRI. It has been suggested that these pulmonary symptoms may represent the “tail end” of the PPHN, although this has not yet been proved directly.

The possibility that these symptoms may represent the life-threatening serotonergic syndrome has also to be considered. At least one published case described a newborn infant with typical signs of poor neonatal adaptation syndrome, who had a toxic level of paroxetine; the symptoms subsided when levels decreased.

Long-term neurobehavioral effects

The concern that SSRIs-SNRIs may affect the normal development of fetal brain is understandable, because these drugs act at the level of central neurotransmitters. A relatively small number of studies have examined neurocognitive and behavioral outcome after in utero exposure to SSRI-SNRI. For the most part, these studies have been reassuring and have failed to show deficits in cognition, learning, or behavior. Yet, they have highlighted the confounding effects of maternal depression itself on child development. In a recently completed study, Nulman et al separated the effect of depression from its treatment by comparing children who had been exposed to venlafaxine in utero with their unexposed siblings and showed no differences in neurocognitive or behavioral achievements.