Are antidepressants hazardous to a neonate’s health? The FDA implied as much this past summer, when they issued a new precaution and required all antidepressant makers to include something like this text (for Effexor) in their package inserts:
“Neonates exposed to SSRIs or SNRIs late in the third trimester have developed complications requiring prolonged hospitalization, respiratory support, and tube feeding. Such complications can arise immediately upon delivery. Reported clinical findings have included respiratory distress, cyanosis, apnea, seizures, temperature instability, feeding difficulty, vomiting, hypoglycemia, hypotonia, hypertonia, hyperreflexia, tremor, jitteriness, irritability, and constant crying. These features are consistent with either a direct toxic effect of SSRIs and SNRIs or, possibly, a drug discontinuation syndrome. It should be noted that, in some cases, the clinical picture is consistent with serotonin syndrome. When treating a pregnant woman with Effexor XR during the third trimester, the physician should carefully consider the potential risks and benefits of treatment.”
This is the stuff your patients will be reading, and many pregnant women will be discontinuing their antidepressants faster than you can say, “But wait–let’s look at the data!” We all need to understand the evidence supporting this warning in order to give good advice to our patients.
How dangerous are SSRIs and SNRIs during pregnancy? How good is the evidence? And how can we best communicate the degree of risk to patients?
Risk of major malformations.
Until a week before this issue went to press, this sentence was going to read: “The good news is that there is virtually no increased risk of major malformations with SSRI use.” However, GlaxoSmithKline just sent a letter to healthcare professionals announcing the resuts of a retrospective study of malformation risk in women taking Paxil during the first trimester. Paxil incurred double the risk of malformations compared with other antidepressants – this amounted to a 4% risk for Paxil vs. a 2% risk for other ADs. The baseline rate is malformations is generally considered to be 1-3%.
It’s unclear what to make of this data, since at least 15 controlled studies have been published assessing the effects on the fetus of exposure to SSRIs, and none have shown a higher than baseline incidence of major malformations. A recent review reported that the drug with the most safety data is Prozac, with over 1600 Prozac-exposed women showing no increased risk of teratogenicity; numbers for the other common SSRIs are: Zoloft, 269; Paxil, 330; Celexa, 397 (J Clin Psychopharm 2005; 25:59-72). Given the recent announcement from GSK, you’d be wise to steer women away from Paxil and toward Prozac if they are considering an AD during pregnancy.
What about Wellbutrin? While a recent study of 136 women showed no major malformations, that same study did report a relatively high rate of spontaneous abortions (Am J Ob Gyn 2005; 192:932-936). However, since depression alone can increase the risk of miscarriage, it is not clear whether Wellbutrin was responsible.
Neonatal Complications and Third Trimester Use
Recently, an excellent literature review (JAMA 2005; 293:2372-2383) focused on the evidence linking third trimester SSRI use with neonatal symptoms. The authors systematically guide us through three different levels of evidence: case reports, case series, and controlled cohort studies.
Case Reports. Case reports make lousy statistics, but are great for painting a vivid picture of a syndrome, and the JAMA authors identified 18 published case reports of third trimester SSRI-related neonatal signs. The most common symptoms reported in these cases were tremors/jitteriness/ shivering (present in 11 of the 18 cases), increased muscle tone (11), feeding/digestive disturbances (9), irritability/agitation (9), and respiratory distress (7). Interestingly, these clinical signs sound more like serotonin syndrome than like serotonin withdrawal, even though they are usually framed as signs of withdrawal. This is controversial because many of the symptoms of SSRI discontinuation are subjective, like dizziness and shock-like sensations, and you’re not likely to find a neonate capable of describing feelings like that! Does it matter? It does, because some authorities advocate giving the baby a dose of Prozac at birth to ease discontinuation symptoms–precisely the wrong thing to do if the problem is serotonin overstimulation.
Case Series. Next up on the hierarchy of clinical evidence are published case series. Mostly, these have been derived from databases of adverse events reported by physicians to the FDA (in the U.S.) or the WHO (World Health Organization, in Europe). This spontaneous reporting makes for pretty poor statistics as well, both because adverse events tend to be under-reported (when was the last time you reported an adverse event to the FDA), and those that are reported are the most serious cases.
Nonetheless, the advantage here is that these case reports are entered into government computers to create huge databases including millions of case records. Researchers can use these databases to perform a new type of statistical maneuver called “data mining.”
Recently the WHO adverse events database was mined for information on neonatal adverse effects of antidepressants, and the results were published (Lancet 2005; 365:482-487). The researchers used data-mining techniques to sift through 3 million records, looking for an association between adverse drug reactions (ADRs) in neonates and maternal SSRI use. The statistic used to report the strength of the association is the confusingly named “information component” (IC). An IC of 0 means that the number of ADRs reported in association with a given drug is no higher than what you’d expect. A positive IC means the side effect is disproportionately common, while a negative IC means it’s unusually uncommon.
You can probably guess the punch line. The ICs of SSRIs and Effexor were all positive for neonatal symptoms, including neonatal convulsions. Most cases were with Paxil, which also had the highest IC, leading the authors to conclude: “Thus paroxetine should not be used in pregnancy or, if used, should be given at the lowest effective dose.”
Cohort Studies. In cohort studies, groups of pregnant women who have been prescribed SSRIs are enrolled, and are compared with control groups. In many of the studies reviewed, the control group consisted of women who discontinued SSRIs after the first trimester, making them ideal studies for assessing specific dangers of third trimester exposure. Don’t confuse these studies with “placebo-controlled clinical trials,” in which patients are randomly assigned to treatment or placebo. Such studies of pregnant women would never win approval from research ethics panels, meaning that cohort studies represent the best quality evidence we’ll ever see.
The JAMA researchers located nine good quality cohort studies, but chose to focus on five that were particularly good in that they systematically defined and measured a specific neonatal SSRI syndrome. They crunched the numbers of these five studies and came up with an overall estimate of the risk of neonatal complications due to late SSRI exposure.
Their summary relative risk is easy to remember: 3.0. This means that babies are three times more likely to have neonatal complications if exposed to third trimester SSRI than with early or no exposure. But this is probably not the best way to describe it to your patients.
Why not? Because “relative risk” can be a misleading statistic. You’ve certainly seen the term bandied about quite a bit in the literature, because it’s cleaner and quicker than discussing the raw numbers. But if a bad event has a large relative risk, but the baseline risk is very small, then the actual absolute risk of something occurring is still quite small. With regard to neonatal SSRI syndrome, though, both the relative and absolute risks are actually pretty substantial, and providing your patients with both quantities is informative.
The table above lists the numbers used to calculate the absolute risks reported in the JAMA article. The relative risk is easy to calculate: just divide the absolute risk of neonatal complications in the third trimester exposure group by the absolute risk in the early or no exposure group. (You’ll notice we included only four studies, because the fifth study used a very different methodology that makes it difficult to compare with the others. For this reason, our number for overall risk is slightly higher than the JAMA number.)
The overall relative risk of complications is 3.4 – a pretty intimidating number if related simply as a 3.4-fold higher risk of complications. But the absolute risk is 29%, meaning that out of 100 babies exposed to third trimester SSRIs, about 29 will have neonatal complications; out of 100 babies with either early or no SSRI exposure, 8 will have similar symptoms. Framed more positively, 71 out of every 100 late-SSRI babies will have no complications whatsoever.
Suddenly, the numbers don’t seem quite as awful, particularly when you consider that all of these babies recovered fully, and that not a single baby in any of these studies had severe or life-threatening complications. Most of the affected babies were admitted to the NICU for a day or two and then sent home with mom.
Ultimately, the decision is your patient’s. Providing her with the right numbers will make her job a little bit easier.
TCR VERDICT: SSRIs in pregnancy: the numbers tell the story