Over the past few years, there have been a large number of studies which have relied on large administrative databases to generate information on the reproductive safety of various medications. One of the strengths of this type of approach is that it provides an opportunity to observe outcomes in a large number of subjects; however, there are certain, very important limitations. A recent study published in the American Journal of Obstetrics and Gynecology helps to understand some of the challenges in interpreting the data generated from these studies.
Researchers from Vanderbilt University conducted a retrospective cohort study of 228,876 singleton pregnancies enrolled in the Tennessee Medicaid program between 1995 and 2007. The study included 23,280 pregnant women who had received at least one antidepressant prescription prior to pregnancy and 6340 women who initiated antidepressant therapy during pregnancy. About 75% of the women discontinued the antidepressant before or during the first trimester. 10.7% of the women used antidepressants throughout pregnancy; however, antidepressant use was inconsistent. Only 2161 women (0.9%) consistently filled prescriptions before and throughout pregnancy.
This study found that antidepressant use (SSRIs and non-SSRIs) in the second trimester was associated with shorter gestation. Filling 1, 2, and ? 3 antidepressant prescriptions during the second trimester was associated with shortened gestational age by 1.7, 3.7, and 4.9 days, respectively.
In addition, it was observed that third trimester SSRI use was associated with infant convulsions. The risk appeared to increase with the number of prescriptions filled. Adjusted odds ratios were 1.4, 2.9, and 4.9 for filling 1, 2, and ? 3 prescriptions, respectively.
In the abstract, the authors state: “Second-trimester antidepressant use is associated with preterm birth, and third-trimester selective serotonin reuptake inhibitor use is associated with infant convulsions.” If you just read that sentence, you might never what to use antidepressants during pregnancy. However, a more careful reading of this paper actually suggests that the use of antidepressants during pregnancy may not pose serious risks.
Just because a study is large doesn’t mean it is better than a smaller study.
Here are some questions you might ask when you are reviewing similar studies:
What is the quality of the data analyzed? One of the problems with this study (and many of the studies relying upon administrative databases) is the poor quality of the data used to make its conclusions. For example, in this study much of the demographic and medical information was derived from the birth certificate. In 85% of the cases, the date of the last menstrual period (LMP) was derived from the birth certificate and used to calculate the gestational age. It is well known that ultrasound dating is much more accurate in terms of determining gestational age because most studies suggest that less than half of women can recall their LMP. Can we therefore believe the finding of a shortened gestational age (particularly when it is only a shift of 2-5 days)?
Also of concern is that in these studies, antidepressant exposure is defined as the filling of an antidepressant prescription. The authors acknowledge that only 0.9% of the women consistently filled their prescriptions. Can we thus assume that filling the prescription is the same as actually taking the drug? Are we measuring exposure to antidepressant or merely the presence of depressive symptoms?
Is the study able to control for confounding variables? The authors also admit that with an administrative database it is often difficult to control for confounding variables, such as the severity of maternal depression. With a significant number of studies indicating that untreated maternal depression is associated with shorter gestation and that this effect increases with the severity of the depression, we are again left wondering if we are looking at the negative effects of exposure to antidepressants or to untreated illness. Nor were they able to control for other variables which may negatively affect pregnancy outcomes, such as use of recreational drugs, tobacco, and alcohol.
How are the outcome variables defined and assessed? Many of the studies using administrative databases rely upon the ICD codes listed on the medical claims. The first problem is that these assessments may be influenced by the treaters’ awareness of the mother’s medication status; in other words, they are not unbiased assessments.
In addition, does a billing code actually reflect what is going on clinically? For example, what exactly is an infant convulsion? In the comments, they consider an infant convulsion “as an extreme component of an SSRI withdrawal syndrome”. They note that convulsions are rare, estimating only 1 additional case of infant convulsions for every 117 women who use SSRI medications in the third trimester.
The Bottom Line
When I first saw this article, I was alarmed by the abstract. But after the reading the entire article, it seems to add very little to what we already know about antidepressant use during pregnancy. We have seen other reports that suggest that antidepressant use may decrease the length of gestation, although this effect appears to be relatively modest and may be modulated by the severity of depression in the mother. With regard to the “convulsions”, the authors refer to them as a more severe form of antidepressant withdrawal, as opposed to a persistent neurologic issue, and emphasize that their frequency is low (<1%). We are familiar with neonatal syndromes; this is nothing new.
It is difficult for the clinician to digest this data and to use it to make well-educated and meaningful clinical recommendations. These larger studies are often published in widely read journals are thus widely disseminated. While smaller studies seem to offer their findings more cautiously and to label them as preliminary, the larger studies simply state their findings and it takes a careful reading of the entire report to fully appreciate its strengths and limitations.
Ruta Nonacs, MD PhD
Hayes RM, Wu P, Shelton RC, et al. Maternal antidepressant use and adverse outcomes: a cohort study of 228,876 pregnancies. Am J Obstet Gynecol. 2012 Jul; 207(1): 49.e1-9.
Bravo! Well said. There are many clinical factors that have to be considered in each woman who has to make decisions regarding antidepressant use in pregnancy. It is always a challenge. I would love to see a longitudinal study of development of children who’s mothers took SSRI’s during pregnancy to see if there are long-term implications. I have not seen that yet anywhere.
Thank you for this essay Dr. Nonacs.
It seems that some authors treat the abstract as a kind of advertisement for their study employing sensationalistic headlines, unfounded generalizations, and lies by omission.
Take for example the Croen Autism article and abstact:
Title: Antidepressant Use During Pregnancy and Childhood a
Autism Spectrum Disorders. One would assume from the title that there was confirmation that the medications were actually taken – not so.
The abstract refers to the study as “prospective” – not so.
and that there were 20 case children – not so, because all conslusions were derived only from SSRI antidepressants which brings the number down to 15 or only 13 if you limit to only single “exposure”.
Further, they controlled for many variables, but not for 1st depree relative with autism which is the second highest risk factor after male sex – why not?
Thank you very much for this article. Being pregnant made me so anxious and the thought of doing something that was so (seemingly) risky just made me more sad and anxious. As a non-academic, I usually only have access to the abstracts (and all the BS on the internet) and so I decided against SSRIs during pregnancy which could have been immensely helpful. I agree that abstracts are often written so that they draw the most attention (EARLY BIRTH! CONVULSIONS!) And then we don’t read the whole thing because we think we got all the information that we need! The truth is, 3 days earlier birth still would have had my baby being born 1.4 weeks late! I wish that the public had access to full articles and the tools that we need to be informed consumers. It would help us to trust our doctors and to make better decisions.