Selective Serotonin Reuptake Inhibitor Use and Risk of Gestational Hypertension

Data from the Slone Epidemiology Center Birth Defects Study, a multicenter case control surveillance program of birth defects in relation to environmental exposures (particularly medications [ 13, 17 ]), were analyzed. More than 35,000 mothers of babies with and without birth defects from the greater metropolitan areas of Philadelphia, San Diego, and Toronto as well as selected regions in Iowa, Massachusetts, and New York State have been interviewed via the study program since 1976. Participants are identified through review of admissions and discharges at major birth hospitals and pediatric referral hospitals and clinics, logbooks at perinatal intensive care units, weekly telephone contact with collaborators at newborn nurseries at community hospitals, and collaborations with state birth defects registries. Since 1998, the study has also included a random sample of births in Massachusetts. The mothers of nonmalformed infants are recruited independently from any exposure and therefore provide an estimate of the distribution of exposures (including the use of medications) within the study population. Institutional Review Board approval was obtained from each of the participating institutions, and mothers provide informed consent before participation in the study.

We restricted our analyses to a retrospective cohort of women who gave birth to nonmalformed live-born babies between 1998 and 2007 and were ascertained either at the hospital-based centers or through the Massachusetts birth registry (N=5,912). Women with elective terminations, miscarriages, or stillbirths were not included, since most of these women were not at risk for the outcomes of interest, which occurred after 20 weeks of gestation.

Within 6 months of delivery, trained nurses who were unaware of the study hypotheses conducted a 45- to 60-minute telephone interview with the participating mothers. The purpose of the interview was to collect data on demographic, reproductive, and medical factors as well as cigarette smoking, alcohol consumption, occupational exposures, and dietary intake. Additionally, the interview included a series of increasingly detailed questions designed to collect data regarding medications (prescription, over-the-counter, vitamins/minerals, and herbal products) taken any time from 2 months prior to conception throughout pregnancy (31) . Interview questions were standardized and included a list of indications and specific conditions (e.g., depression) and specific drugs (e.g., fluoxetine). Whenever possible, reported medications were verified by asking the subject to read information from the medication container. Identification of the timing of drug exposure was facilitated using a 12-month calendar covering periods before and after pregnancy. Significant dates (e.g., last menstrual period, holidays) were marked to help enhance recall. Data on medication were collected for starting and discontinuation dates, duration, frequency, indication, form, and number of pills per day. The interview also elicited information regarding subjects’ last menstrual period, based on either maternal recall or ultrasound exam. The last menstrual period allows for an estimation of the due date, approximate conception date, and gestational age at birth. We identified women who were receiving SSRI treatment 2 months before pregnancy and then categorized them into the following two groups: 1) those women who discontinued their treatment before the end of the first trimester and 2) those women who continued treatment after the first trimester. Since very few women initiated SSRI treatment during pregnancy (N=36), assessing the effect of starting treatment during pregnancy would have been less clinically relevant and statistically underpowered, and these women were not analyzed separately (no cases of preeclampsia were observed among these women). Treatment with non-SSRI antidepressants was classified in a similar manner.

Women were specifically asked whether a healthcare provider had diagnosed them with high blood pressure or preeclampsia/toxemia during their pregnancy, and if so, the dates when the condition started and ended and whether medication was used to treat the condition. Given the potential misclassification of pregnancy-induced hypertension with and without preeclampsia, we combined the two diagnoses in our main analyses. Thus, gestational hypertension was defined as incident hypertension during pregnancy with and without proteinuria (i.e., with and without preeclampsia/toxemia). To exclude underlying hypertension as a potential source of both confounding and outcome misclassification bias, we restricted the definition of gestational hypertension to that first diagnosed after week 20 of pregnancy. Consequently, we excluded 145 women with an early diagnosis of hypertension from all analyses. For each woman with gestational hypertension, the onset date was defined as the gestational day when gestational hypertension was first diagnosed.

We consider our study to be a retrospective cohort design. Using women who did not receive SSRI treatment during pregnancy as the reference group, we estimated the relative risk of gestational hypertension for those women who discontinued SSRI treatment before the end of the first trimester and those women who continued treatment beyond the first trimester. Characteristics of the two latter groups were compared using chi-square analyses and Fisher’s exact tests. Relative risks and 95% confidence intervals (CIs) were estimated for each outcome using Cox proportional hazards model, with time measured in days since week 20 of gestation. The following potential confounders were considered: region, birth year, maternal age, race/ethnicity, education, family income, gravidity, number of fetuses, prepregnancy body mass index (kg/m ² ), age at menarche, diabetes mellitus, infertility treatment, cigarette smoking, coffee and alcohol intake, and use of illicit drugs or other psychotherapeutic medications during pregnancy (24, 32 – 36) . In the Cox proportional hazards model, we retained those factors associated with the outcome in our study population (even if the association was not statistically significant). The analyses were repeated separately for preeclampsia and gestational hypertension without preeclampsia. Several secondary analyses were also conducted. Because of the different heritability, clinical manifestations, and prognosis of early- and late-onset gestational hypertension, these conditions were distinguished according to the timing of onset (33, 34, 37, 38) . Gestational hypertension was considered early-onset when it began between 20 and 32 weeks after the last menstrual period, and gestational hypertension that began after 32 weeks following the last menstrual period was considered late-onset. We analyzed primigravidae women separately from multigravidae women and women who smoked cigarettes separately from women who did not smoke cigarettes. In addition, our analyses were restricted to singleton births (39) . Effect measure modification on a multiplicative scale was formally tested using the Wald test comparing whether two stratum-specific relative risks were equivalent. We also verified the proportional hazards assumption by including in the models an interaction term for time and treatment with SSRIs.

Treatment with SSRIs exclusively during early pregnancy, when placentation occurs, was not associated with a higher risk for preeclampsia. Thus, if real, the effect of SSRIs on preeclampsia might not be mediated via placentation but rather through hypertension or vascular effects later during pregnancy. Serotonin may play a role in the pathophysiology of preeclampsia through its effects on hemostasis and vascular tone in uteroplacental tissues (29, 40, 41) . In addition, SSRIs inhibit synthesis of nitric oxide, a vasodilator that appears to play a role in vascular tone and reactivity both in utero and during postnatal life (42 – 44) . The more recently introduced SNRIs can cause elevations of diastolic blood pressure, probably as a result of their noradrenergic effects (45), but we did not have sufficient numbers of exposed women to examine the outcomes associated with these medications.

The association between SSRIs and preeclampsia might stem from underlying mood disorders (e.g., depression, anxiety), which could reflect both the indication for treatment and the ultimate cause of preeclampsia, or it could represent a marker for a higher risk (e.g., a genetic-linked predisposition for both depression or anxiety and preeclampsia). For example, depression or anxiety could induce vasoconstriction and uterine artery resistance through an altered excretion of vasoactive hormones and other neuroendocrine transmitters (5) . In a Finnish study (27), anxiety and depression during pregnancy were associated with a threefold increased risk for preeclampsia, and a similar association was found in a recent case-control study (28) . However, these studies did not examine the effects of these conditions independent of treatment. In the present study, the greater occurrences of gestational hypertension and preeclampsia found among women who were treated with non-SSRI antidepressants (e.g., SNRIs, tricyclic antidepressants) relative to women who were not treated with non-SSRIs, albeit nonstatistically significant, suggest that women who need antidepressive agents might be at an increased risk for these outcomes and that the increased risk might not be related to SSRIs themselves. The small number of outcomes among women who received treatment with non-SSRI antidepressants limited our ability to examine the effect of these medications compared with the effect of SSRIs. Unfortunately, among women with depression, we were unable to compare those who received SSRI treatment with those who did not receive treatment, since we did not identify mood disorders among those women who did not report drug treatment. However, although it might appear that such comparisons would have been ideal in order to adjust for confounding by the underlying condition, depression severity and other potential risk factors for preeclampsia among women with depression may differ between individuals who are and are not treated with specific antidepressants. For example, it has been observed that among pregnant women with a diagnosis of depression, those who receive treatment with SSRIs have clinical indicators suggesting that they have a greater severity of depression when compared with women who do not receive SSRI treatment (22) .

Therefore, in addition to comparing women who received antidepressant treatment with women who did not receive antidepressant treatment, we compared those women who discontinued SSRI treatment before the end of the first trimester with those who continued SSRI treatment after the first trimester. Although the latter two groups did not differ significantly in most baseline variables ( Table 1 ), these women might differ in other unmeasured characteristics that could potentially explain our findings. In particular, we did not have a measure of depression severity for women who discontinued or continued SSRI treatment. Those women who continued to receive SSRI treatment might have had better control of their symptoms than those who discontinued their treatment, or they might have had a more severe condition that required pharmacological treatment. Thus, the higher risk for preeclampsia associated with continuation of treatment after the first trimester might simply reflect a more severe mood disorder. In the absence of treatment randomization, comparing these two groups might have reduced, but not eliminated, the possibility of confounding by depression severity.

Characteristics such as high prepregnancy body mass index are positively associated with both SSRI treatment and preeclampsia (46) . Failure to adjust for these confounders could overestimate the risks associated with drug treatment. In our analyses, estimates adjusted for these confounders did not differ substantially from the unadjusted estimates. Other important socioeconomic and lifestyle factors, such as family income and maternal education level, were also considered but did not appear to be major confounders in our study population. For an unmeasured confounder to have an appreciable impact, it would have had to be reasonably prevalent and strongly associated with both SSRI use during pregnancy and the risk for gestational hypertension.

A potential limitation to the study lies in the outcome data, which were based on self-report, since we did not have routine access to obstetric notes in the mothers’ medical records. Thus, under- and overreporting of events, misclassification of the exact date of onset, and cross-classification of preeclampsia and transient hypertension during pregnancy were possible. Underdiagnosis is unlikely, since more than 99% of the mothers in our study population had prenatal care, in which screening for gestational hypertension and proteinuria is standard practice. Data errors are likely to have been minimized by the use of a carefully designed questionnaire, with specific questions pertaining to hypertension onset and preeclampsia, as diagnosed by a healthcare provider, administered within 6 months of delivery by trained nurses. Further, the data offer evidence of the general validity of the outcome classification. First, the reported occurrences of gestational hypertension among primiparous women (12.6%) and the overall occurrences of preeclampsia (2.7%) in our study population are similar to those found in population-based studies (47, 48) . Second, the frequency and effect of other known risk factors (e.g., gravidity, number of fetuses, and maternal weight) are similar to those consistently reported in the literature (24, 32 – 36) . Other previously suggested risk factors, such as advanced maternal age, were not associated with a higher frequency of gestational hypertension in our study, most likely because we excluded women with pregestational hypertension (i.e., these factors were associated with hypertension diagnosed before 20 weeks of gestation). Our failure to find the previously suggested protective effect of smoking (49) might be a result of various factors, including differences among characteristics of women who smoked over time and methodological factors. In addition, if the reported outcomes were misclassified similarly for women who were and were not treated with SSRIs, the effect would then be to bias our results toward a null effect of SSRIs. Of greater concern would be a differential misclassification of outcome among women who were and were not treated with SSRIs. For example, if women who received treatments, such as antidepressant therapy, were diagnosed with gestational hypertension or preeclampsia more readily, this situation would indeed tend to bias the association between SSRIs and gestational hypertension/preeclampsia toward an increased risk among women who were treated with SSRIs and for women who continued versus discontinued their treatment. However, although more intensive monitoring and diagnosis would likely produce stronger associations for gestational hypertension, which may be variably detected, such monitoring and diagnosis would be less likely to do so for preeclampsia, which is a clinically significant event that is difficult to overlook.

Studies that collect medication prescribing or dispensing data might not capture actual use and time-of-use during pregnancy, leading to potential misclassification of exposure (50) . Our study collected data on medications actually taken by the mother, and timing-of-use was systematically collected. Although data based on maternal recall might be subject to misclassification, we do not feel that recall bias was a major concern in the present study, since SSRIs are regularly used for treatment—and for nontrivial reasons—and the study interviewers specifically probed for the use of antidepressants and were unaware of the study hypothesis. Although underreporting of antidepressant use is possible, since women might be reluctant to disclose such information, it is reassuring that the prevalence of SSRI use in our study population was consistent with that reported by prospective studies (4, 22) . Further, if women with the study outcomes were less likely to report their treatment with SSRIs, such underreporting would lead to an underestimation of the true risk.

Several (18 – 22), but not all (8 – 12), studies have suggested that SSRIs may be associated with a greater risk for prematurity and/or fetal growth restriction. Since preeclampsia is a known risk factor for these outcomes (51), it is possible that the suggested associations between SSRIs and prematurity or fetal growth restriction may be mediated through an elevated risk of preeclampsia among women who are treated with SSRIs.