On May 28, 2008, The Food and Drug Administration announced a major revision in the labeling of the pregnancy and breast-feeding risk of drugs. The current labeling system was devised in 1979 and grades drugs from A (controlled studies show no risk) to X (contraindicated in pregnancy). It has been criticized over the years for being confusing and for failing to incorporate the most recent data.

In the new system, package inserts will carry a short monograph on the pregnancy and lactation risks of each drug. The pregnancy monographs will include four sections: 1. An initial statement about the background risk of congenital malformations; 2. A summary statement of the risk; 3. Clinical considerations, such as the risks of the untreated disease and the concept of risk/benefit ratio; 4. A summary of the latest human and animal data. I suggest you visit the FDA’s website (http://www.fda.gov/ cder/regulatory/ pregnancy_labeling/default.htm), where you can read fictitious examples of the labels to give you a better idea what they will look like.

For an excellent summary of the latest data on psychiatric medications during pregnancy and lactation, read the American College of American College of Obstetricians and Gynecologists latest practice bulletin on the issue (ACOG Practice Bulletin, Obstet & Gyn 2008;111:1001-1020). (We have provided the current pregnancy and lactation categories on page 2 for reference.)

Is Paxil really dangerous?

In December of 2005, the FDA issued a public health advisory about the risk of cardiac anomalies in infants whose mothers have taken Paxil (paroxetine), and required the manufacturer, GlaxoSmithKline (GSK), to change the pregnancy category from C to D, a stronger warning (http://www.fda.gov/ bbs/topics/NEWS/2005/NEW01270.html). The agency based this on results from two databases of paroxetine-exposed fetuses, both of which reported rates of cardiac defects of 2%, versus the generally accepted baseline rate of about 1%.

Suddenly, Paxil was the antidepressant to avoid in pregnancy, and while this made little difference in a world filled with me-too antidepressants, the more worrisome issue was whether this meant that the other SSRIs also conferred a similar risk that had not yet been discovered. A recent paper in the American Journal of Psychiatry comes close to exonerating paroxetine (Einarson A et al., Am J Psychiatry 2008;165:749-752). It’s a paper worth describing partly because of its results, and partly because of what it teaches us about how to interpret the confusing literature on medications in pregnancy.

The problem with studying the safety of drugs in pregnancy is that the gold standard study will never occur (for a good discussion of these methodological issues, see Einarson A, J Clin Pharmacol Pharmacoepid 2008;1:3-8). It will never be ethically permissible to enroll pregnant women into a randomized, placebo controlled trial designed to determine if a drug causes birth defects. For this reason, we are left with less than ideal methods of determining risk. Most pregnancy studies are either prospective cohort studies or retrospective database studies. A prospective cohort study works like this: A woman finds out that she is pregnant, and is already on, say, paroxetine. Concerned, she calls a teratology information service, usually run by a university, to ask an expert what she should do. The expert gives advice, and then asks the mother-to-be if she would mind enrolling in a prospective study in which they will give her a call after her baby is born to find out what happened. It is a “prospective” study because the subjects are identified before the outcome of interest (the birth) has occurred.

In a retrospective database study, on the other hand, researchers use existing data that includes pregnancy and medication information (for example, city health records or pregnancy registries) to determine the rate of various birth defects in babies of exposed mothers.

The prospective studies are more reliable, because all women are interviewed personally, allowing for confirmation of drug exposure and dose, and eliminating the possibility of recall bias. Retrospective studies, on the other hand, can have the advantage of very large sample sizes, allowing researchers to examine risks of very rare birth defects.

The recent AJP study combined both methods to create a very robust estimate of the risk of paroxetine exposure. First, they amassed data from teratology information services all over the world to ascertain the outcomes of 1,174 women who had been exposed to paroxetine, and compared these women with an equal number of women who had been exposed to nonteratogens such as Tylenol. They found that the rate of cardiac defects was virtually the same in both groups, 0.7%, slightly lower than the 1% population baseline rate. Then they pooled data on 2,061 paroxetine-exposed women from various retrospective database studies; in this group, the cardiac defect rate was 1.5%. Combining the two data sets, the mean rate of cardiac defects was 1.2%, not significantly higher than baseline.

The authors conclude with the following statement, which contradicts the current FDA opinion: “With nearly 1,200 cases from teratology information services and over 2,200 cases from previously published database studies, our findings may be considered sufficient evidence to suggest that there is no association between the use of paroxetine in pregnancy and risk of cardiovascular defects in exposed infants.” And no, the study was not funded by GlaxoSmithKline, the maker of Paxil, or by any other drug company. Seven of the nine authors reported no competing interests, while the other two have received funding for research from various drug makers, but no money from GSK.

Not sure what to do with all this conflicting data? Let’s face it, with a pregnancy rating of D, the smart thing to do from a medico-legal perspective is to avoid paroxetine. But the data does suggest that paroxetine – and perhaps, by extension, all SSRIs – may be safer than what has been suggested by other smaller studies.

Benzodiazepines. In a huge study of nearly 2000 women who had been exposed to either benzodiazepines or related hypnotics (such as Ambien and Sonata), the rate of orofacial clefts was not above baseline, although there was a higher than expected rate of preterm birth and low birth rate (Wilkner BN et al., Pharmacoepid Drug Saf 2007;16:1203-1210).

Lithium. The risk for significant cardiovascular abnormalities continues to be revised downward, with the latest figures indicating 1.5 to 3 times the baseline risk. But there are various other risks, including a neonatal lithium toxicity syndrome characterized by flaccidity and lethargy. The latest guidelines include these three recommendations: 1) Taper off lithium before conception in women with mild and infrequent episodes of bipolar illness; 2) In women with moderately severe illness, taper before conception but restart after the first trimester; 3) In women with severe illness, continue lithium throughout (ACOG Practice Bulletin, Obstet & Gyn 2008;111:1001-1020).

Valproate and carbamazepine. Avoid both, especially during the first trimester. Valproate causes neural tube defects and carbamazepine causes a syndrome of facial dysmorphism and fingernail hypoplasia.

Lamotrigine has a good and growing reproductive safety profile – a good alternative to other mood stabilizers.

Antipsychotics. Forty years of experience with typical antipsychotics have endorsed both perphenazine (Trilafon) and trifluoperazine (Stelazine) as probably safe. There is not enough data on atypicals to recommend their use, although existing studies have not shown any increased risk of birth defects.