Big data is changing the face of medicine, yielding information that may drastically change how we practice medicine.  In the field of perinatal psychiatry, we are now seeing more and more studies which rely on large databases to answer questions regarding the safety of psychotropic medications during pregnancy.

The most recent example of this trend is the article published last week on the link between prenatal antidepressant exposure and increased risk of autism spectrum disorders. The media enthusiastically latches onto these reports of negative findings, satisfying our thirst for bad news.  However, most of these articles in the lay press do not highlight the limitations of these large studies.   Limited by word counts and medical expertise in a particular area, it may be unreasonable to ask journalists to be appreciate and to make readers aware of all the limitations of a particular study.

Garbage In, Garbage Out

Garbage in, garbage out (also known as GIGO) is a phrase derived from the field of computer science which refers to the fact that computers, since they operate by logical processes, will unquestioningly process unintended, even nonsensical, input data (“garbage in”) and produce undesired, often nonsensical, output (“garbage out”).

This principle extends to other fields as well, and we must heed the GIGO principle as we evaluate the findings of studies which rely on large databases.  We need to question where the data comes from.  How was it collected?  What is the quality of the data? This information is not always included the abstract of a medical article and sometimes requires some digging around to get the answer.

When the studies assess for behavioral problems, how is that information gathered?  Many studies rely on mothers’ reports, which may be a problem because various studies have demonstrated that mothers with depression are more likely to report negative behaviors in their children than non-depressed mothers.  More reliable information can be derived from face-to-face evaluations, structured questionnaires or pediatricians’ records; however, the time and effort needed to retrieve this type of data is substantial and may not be feasible in larger studies.

When the studies assess for risk for autism spectrum disorders in children, how was that diagnosis made?  Some studies rely on mothers’ reports.  Others rely on a diagnosis made by a medical professional.  When are the children evaluated?  In the United States, the average age of diagnosis with an autism spectrum disorder (ASD) is around 4 years of age. If we assess the children too early, we may miss potential cases.

Do Pregnant Women Actually Take Their Medications?

In these large studies, the most important variable is prenatal exposure to a particular medication.  In other words, did the mother take Medication X during pregnancy? And was this exposure to  Medication X associated with a particular outcome?  Sometimes the researchers get this information on medication usage directly from the mother, sometimes the data comes from medical or pharmacy records. But how accurate is this information?  Do women actually take the medications they are prescribed?

Several studies indicate that adherence to psychotropic medications is less than perfect. In a recent study, researchers analyzed blood levels of antidepressants in a group of 109 patients treated for major depressive disorder.  They found that samples drawn from 17% of the patients showed no detectable levels of antidepressants.

Angela Lupattelli and her colleagues at the University of Oslo have focused on the issue of adherence to medications during pregnancy.  In a multinational web-based study was performed in 18 countries in Europe, North America, and Australia. Medication adherence was measured using the 8-item Morisky Medication Adherence Scale (MMAS-8).

On the basis of the MMAS-8, 78 of 160 women (48.8%) demonstrated low adherence during pregnancy. The rates of low adherence were 51.3% for medications for anxiety, 47.2% for antidepressants, and 42.9% for other psychiatric disorders. The authors determined that the following factors were associated with lower adherence: smoking during pregnancy (3.9-fold more likely to report low adherence), increased perception of antidepressant risk (2.3-fold increase) , and depressive symptoms (2.5-fold increase).  The belief that the benefit of pharmacotherapy outweighed the risks was positively correlated with medication adherence.

So only about half of women are regularly taking their psychotropic medications during pregnancy.  

This is a problem for so many different reasons.  We are so eager to identify and treat women with anxiety and depression during pregnancy, yet these findings suggest that many women may not feel totally comfortable with taking psychotropic medications during pregnancy.  This is probably not such a big surprise given the stigma associated with taking medication for psychiatric illness, especially during pregnancy, and the sometimes erroneous perceptions of risk associated with various medications.

Given these extremely high levels of medication non-adherence, how can we interpret data from these large studies? With caution and with the understanding that no single study can answer the questions we are asking.  When we document an association between medication exposure and a particular outcome, we must be aware of other potential confounding factors.  Most studies do not take into consideration exposure to smoking, alcohol, recreational drugs, and other medications, factors which may negatively affect outcomes.  Anxiety and depressive symptoms may actually be more severe in women who are taking their medications inconsistently.  We have a growing body of literature to suggest that these symptoms – when left untreated or incompletely treated during pregnancy — may have long-term effects on the developing fetal brain.  So we must always ask: Is it really the medication? Or is there some other variable that is more common in women who choose to take that medication during pregnancy?

This is an exciting but challenging time for the field of perinatal psychiatry.  We have an increasing number of studies which are very reassuring and support the reproductive safety of various psychotropic medications, yet it is the studies with negative findings that get the most attention, particularly in the media.  

Ruta Nonacs, MD PhD

Lupattelli A, Spigset O, Bjornsdottir I, Hameen-Anttila K, Mardby AC, Panchaud A et al. Patterns and factors associated with low adherence to psychotropic medications during pregnancy-a cross-sectional, multinational web-based study. Depress Anxiety 2015 [Epub ahead of print]

Roberson AM, Castro VM, Cagan A, Perlis RH.  Antidepressant nonadherence in routine clinical settings determined from discarded blood samples.  J Clin Psychiatry. 2015 Nov 24. [Epub ahead of print]