TCPR: You and your colleagues at Motherisk have published some of the major papers on the risks of psychotropics in pregnancy. [See for example Einarson A, J Clinical Pharmacol and Pharmacoepidemiol 2008;1(1):3–8; Einearson A et al, Depress Anxiety 2010;27(1):35–38; Einarson A and Boskovic R, J Psychiatr Pract 2009;15(3):183–192.] How do you obtain this data?
Ms. Einarson: Early on, we realized that there was very little reliable information available, so we decided to create it ourselves. Our methodology is simple. A woman will call us for information about a drug she is taking, and after answering her questions, we ask her if she minds being enrolled in an observational study. If she agrees, we call her back about two to four months after the expected date of delivery. Using a standardized questionnaire, we ask about the delivery and the health of the baby, and we ask permission to call the doctor in order to confirm the information.
TCPR: How many phone calls do you get specifically from pregnant women on psychiatric drugs?
Ms. Einarson: About 12 to 15 percent of all the calls to the Motherisk Program are about psychiatric drugs in pregnancy. And one of the things that prompted my interest in pursuing this research is that in the past, many of the calls were from women who were being told to immediately stop their antidepressants when they learned they were pregnant. I knew that there is a significant risk of both depressive relapse and postpartum depression in these circumstances, so I wanted to generate some data to guide practitioners in how to make these recommendations.
TCPR: In the world of teratogen research, how is your type of study classified?
Ms. Einarson: It is called a prospective observational cohort study. A cohort just means a group of people that we can follow over time. In addition to the cohort taking the antidepressant of interest, we usually enroll two comparison groups of women; one group who are taking a different antidepressant than the one we are studying, and another group who are not taking an antidepressant, but who are taking a nonpsychiatric medication that we have already studied and we know to be safe, like Tylenol. The ideal study would have another comparison group of women who are depressed and not taking a drug, but we can’t do that, because women don’t contact us unless they are on a drug. The lack of this kind of comparison group is a big problem for this field of research, because it is hard to separate the adverse effects of the depression itself from the possible adverse effects of an antidepressant.
TCPR: Isn’t the gold standard for clinical research the placebo controlled trial?
Ms. Einarson: Yes, but obviously we cannot ask pregnant women to be in a randomized controlled trial because you would have to ask them to take a drug that might harm the fetus, and nobody is going to volunteer to do that, so the best we can do is an observational study.
TCPR: A couple of years ago you reported in a paper in the American Journal of Psychiatry that paroxetine [Paxil] has very little evidence of danger in pregnancy, and yet some researchers disagree (Einarson et al, Am J Psychiatry 2008;165(6):749–752). Can you give us some of the background of this controversial issue?
Ms. Einarson: The whole thing actually started with bupropion [Wellbutrin]. In 2005, GlaxoSmithKline, the manufacturer of Wellbutrin, found in their retrospective cases that there was an increased risk for heart defects in babies who were exposed to Wellbutrin. So they said, “Okay, let’s compare Wellbutrin to other antidepressants.” When they did that, they found that only Paxil increased the risk of heart defects, and the risk was two-fold. In the scheme of birth defects, heart defects are quite common in the population, about one in 100, so they found an average of two heart defects out of 100 births for Paxil. At the same time there were two other abstracts that also found the same increased risk for Paxil. Although none of these three studies were published in the literature, the FDA and Health Canada put the abstracts on their websites for the public to see. (You can see the FDA’s Public Health Advisory about Paxil here: http://bit.ly/bNx3fJ)
TCPR: In the years since 2005, you and others have come out with a lot more data about the risk of psychiatric drugs during pregnancy. Why is the newer data better than the old data?
Ms. Einarson: It is not necessarily that much better. GlaxoSmithKline reevaluated their study and added more cases, and they lowered the risk of heart defects to an odds ratio of 1.5 (Cole JA et al, Pharmacoepidemiol Drug Saf 2007;16(10):1075–1085). Since then, the risk has never been found to be above two. When you’re talking about odds ratios, some people don’t even consider anything less than two as a problem.
TCPR: Can you define “odds ratios” for the non-researcher?
Ms. Einarson: Odds ratio is a comparison between the two groups, and is very similar to the relative risk. If it is around one, that means the groups are pretty much the same. So any odds ratio between one and two means there is a small increased risk.
TCPR: So the odds ratio for Paxil causing heart defects has come down from two to about 1.5. What does that number mean?
Ms. Einarson: It means that the relative risk of a heart defect when taking Paxil is 50% higher in infants exposed compared to infants unexposed to Paxil. However, the absolute risk is 1.5% opposed to the 1% population baseline rate. [Ed note: it is coincidental that in this case the absolute risk and the relative risk are identical.]
TCPR: How do your research methods compare to the large retrospective database studies—particularly in terms of how accurate findings are likely to be?
Ms. Einarson: Some of the other studies come from large databases that were not set up to do research. An example would be an HMO database like Kaiser Permanente in the U.S.—they are administrative databases that are used for payment information. At some point someone will say, “Oh look at this. It’s a good opportunity to do some research.” There is nothing wrong with this, except that there are inherent limitations. For example, researchers will often go on “fishing trips,” where they do hundreds of tests on the data until they come up on something that is positive. There is a risk of stumbling over a chance “association” that is not real.
TCPR: And how might the quality of their data on the risks of heart defects differ from the data generated at Motherisk?
Ms. Einarson: One of the main differences pertains to when and how the birth defect is diagnosed. Some babies are born with heart defects that resolve by themselves. In many of the retrospective studies, these would be counted as birth defects, because their data is not detailed enough for them to know about the subsequent history of the defect. But at Motherisk, since we actually interview all of our subjects, we can discover if a defect actually affects the child, and if it does not, we wouldn’t count it.
TCPR: Recent research has shown that some other antidepressants also show a risk for birth defects. Can you talk about these?
Ms. Einarson: Recently sertraline [Zoloft] and Prozac have been implicated. [See for example Tuccori M et al, Postgrad Med 2010;122(4): 49–65; Pedersen LH et al, BMJ 2009;339:b3569; and Ellfolk M and Malm H, Reprod Toxicol 2010;30(2):249–260.] Again, these findings are based on large retrospective database studies and the validity of the findings is suspect. From our prospective database at Motherisk, we have reliable clinical details of almost 1,500 women who took a range of antidepressants during pregnancy, and we have not been able to find an increased risk for any birth defects. In my opinion, there is simply no convincing evidence that any of the antidepressants increase the risk for birth defects, including Paxil.
TCPR: So for clinicians who are wondering what kind of data is best, should we assume that prospective cohort studies are now the gold standard?
Ms. Einarson: I really don’t think there is any gold standard. All of the studies have strengths and limitations. The key point is that you should understand the methodology of studies, so that you can intelligently evaluate any conclusion. [For more on this, see Einarson A, J Clinical Pharmacol and Pharmacoepidemiol 2008;1(1):3–8.]
TCPR: Thank you, Ms. Einarson.