- Kristin Palmsten, postdoctoral research fellow1,
- Sonia Hernández-Díaz, associate professor1,
- Krista F Huybrechts, instructor2,
- Paige L Williams, senior lecturer3,
- Karin B Michels, associate professor145,
- Eric D Achtyes, assistant professor6,
- Helen Mogun, programmer2,
- Soko Setoguchi, associate professor17
- 1Department of Epidemiology, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115, USA
- 2Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, 1620 Tremont Street, Suite 3030, Boston, MA 02120, USA
- 3Department of Biostatistics, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115, USA
- 4Obstetrics and Gynecology Epidemiology Center, Department of Obstetrics, Gynecology, and Reproductive Biology, Brigham and Women’s Hospital, Harvard Medical School, 221 Longwood Avenue, Boston, MA 02115, USA
- 5Institute for Prevention and Cancer Epidemiology, University of Freiburg Medical Center, Freiburg, Germany
- 6Division of Psychiatry and Behavioral Medicine, College of Human Medicine, Michigan State University, 100 Cherry Street SE, Grand Rapids, MI 49503, USA
- 7Duke Clinical Research Institute, Duke University School of Medicine, 2400 Pratt Street, Durham, NC 27705, USA
- Correspondence to: K Palmsten
- Accepted 25 July 2013
Objective To determine whether use of serotonin or non-serotonin reuptake inhibitors near to delivery is associated with postpartum hemorrhage.
Design Cohort study.
Setting 2000-07 nationwide Medicaid data (Medicaid Analytic eXtract).
Population 106 000 pregnant women aged 12-55 with a diagnosis of mood or anxiety disorder. Women were categorized into four mutually exclusive exposure groups according to pharmacy dispensing data: current (delivery date), recent (1-30 days before delivery date), past (1-5 months before delivery date), and no exposure (reference group).
Main outcome measures Risk of postpartum hemorrhage by timing of exposure and by serotonin or non-serotonin reuptake inhibitors, classes of antidepressant, and antidepressant types. Relative risks and 95% confidence intervals adjusted for delivery year, risk factors for postpartum hemorrhage, indicators of severity of mood/anxiety disorder, other indications for antidepressants, and other drugs. High dimensional propensity score (hdPS) methods were used to empirically identify and adjust for additional factors.
Results 12 710 (12%) women had current exposure to serotonin reuptake inhibitor monotherapy, and 1495 (1.4%) women had current exposure to non-serotonin reuptake inhibitor monotherapy. The risk of postpartum hemorrhage was 2.8% among women with mood/anxiety disorders but no exposure to antidepressants, 4.0% in the current users of serotonin reuptake inhibitors, 3.8% in the current users of non-serotonin reuptake inhibitors, 3.2% in the recent users of serotonin reuptake inhibitors, 3.1% in the recent users of non-serotonin reuptake inhibitors, 2.5% in the past users of serotonin reuptake inhibitors, and 3.4% in the past users of non-serotonin reuptake inhibitors. Compared with no exposure, women with current exposure to serotonin reuptake inhibitors had a 1.47-fold increased risk of postpartum hemorrhage (95% confidence interval 1.33 to 1.62) and women with current non-serotonin reuptake inhibitor exposure had a 1.39-fold increased risk (1.07 to 1.81). Results were similar with hdPS adjustment. Women with current exposure to serotonin reuptake inhibitors had an adjusted excess risk of 1.26% (0.90% to 1.62%), with a number needed to harm of 80, and for women with current exposure to non-serotonin reuptake inhibitors the excess risk was 1.03% (0.07% to 1.99%), with a number needed to harm of 97. For exposure to serotonin reuptake inhibitors the relative risk was 1.19 (1.03 to 1.38) for recent exposure and 0.93 (0.82 to 1.06) for past exposure; for non-serotonin reuptake inhibitors the figures were 1.17 (0.80 to 1.70) and 1.26 (1.00 to 1.59), respectively. Current exposure to selective serotonin reuptake inhibitor monotherapy was also associated with postpartum hemorrhage (1.42, 1.27 to 1.57), as was current serotonin norepinephrine (noradrenaline) reuptake inhibitor (1.90, 1.37 to 2.63) and tricyclic monotherapy (1.77, 0.90 to 3.47). All types of selective serotonin reuptake inhibitors available for analysis and venlafaxine, a serotonin norepinephrine reuptake inhibitor, were significantly associated with postpartum hemorrhage.
Conclusions Exposure to serotonin and non-serotonin reuptake inhibitors, including selective serotonin reuptake inhibitors, serotonin-norepinephrine reuptake inhibitors, and tricyclics, close to the time of delivery was associated with a 1.4 to 1.9-fold increased risk for postpartum hemorrhage. While potential confounding by unmeasured factors cannot be ruled out, these findings suggest that patients treated with antidepressants during late pregnancy are more likely to experience postpartum hemorrhage.
Postpartum hemorrhage is a leading cause of maternal mortality in the United States and elsewhere1 2 and is a substantial contributor to severe maternal morbidity, blood transfusions, and admissions to intensive care.3 4 The incidence of postpartum hemorrhage has increased since the 1990s in the US (from 2.3% to 2.9% between 1994-2006)5 and in several other developed countries.6 7 8 This trend is not fully explained by temporal changes in the frequency of established risk factors for postpartum hemorrhage, including multiple pregnancy and induction and/or augmentation of labor.6 7 As postpartum hemorrhage is serious, the causes of this increase in incidence need to be identified.
Many, but not all, studies have shown that use of antidepressants that inhibit serotonin reuptake increase the risk of bleeding events, such as gastrointestinal and perioperative bleeding.9 10 11 12 13 14 Use of serotonin reuptake inhibitors is thought to increase risk of bleeding by depleting platelet serotonin.15 Between 7-13% of pregnant women in the US are treated with antidepressants.16 17 18 Only two studies, both in non-US populations, however, have investigated the association between antidepressants and postpartum hemorrhage. Salkeld and colleagues reported a 1.30-fold borderline significant increased risk of postpartum hemorrhage associated with use of a selective serotonin reuptake inhibitors in the 90 days before delivery,19 and Reis and colleagues reported a 1.45-fold increased risk for bleeding during delivery in a comparison of any versus no use of antidepressants.20 While these studies raise concerns, they had limitations including potential confounding by mood disorders or factors associated with them, exposure windows that did not focus on delivery, and limited power to assess specific antidepressants.
Medicaid is the joint state and federal health insurance program for low income individuals in the US. Using nationwide Medicaid data, we assessed the association between exposure to antidepressants at the time of delivery and postpartum hemorrhage. We hypothesized that women exposed to serotonin reuptake inhibitors at the time of delivery would have an increased risk compared with those without any exposure, while women with serotonin reuptake inhibitor exposure only well before delivery and women exposed to non-serotonin reuptake inhibitors antidepressants would not have an increased risk. We accounted for the potential effect of underlying mood/anxiety disorders, considered exposure at the time of delivery (the likely relevant exposure period), classified antidepressant use according to serotonin transporter affinity, and evaluated risk of postpartum hemorrhage with respect to specific classes and types of antidepressants.
We previously identified a cohort of pregnancies ending in live birth in 2000-07 Medicaid Analytic eXtract (MAX) data among women aged 12-55.21 22 23 To ensure claim completeness during the second half of the pregnancy, women were required to have Medicaid enrollment and meet eligibility criteria—that is, no private insurance, no restricted benefits, and appropriate type of enrollment from five months before delivery until after delivery. The eligible population included 2 759 414 pregnancies from 2 340 631 women.
Mood and anxiety disorders
We restricted the primary analyses to a subcohort of 106 000 women with diagnoses for mood or anxiety disorders (as inpatients or outpatients; ICD-9 (international classification of diseases, ninth revision) code for 296.x, 300.x, 309.x, or 311.x) between one and five months before delivery (fig 1⇓) to reduce the potential for confounding by unmeasured factors that are correlated with these disorders. The positive predictive value for depression with these codes was 77%,24 indicating that most women in this subcohort likely had depression. We restricted a secondary analysis to women with diagnostic codes specific for depression—that is, those that did not include bipolar and anxiety disorders (ICD-9 codes: 296.2x, 296.3x, 296.9, 300.4x, 309.0x, 309.1x, 309.28, 311.x).
Women with an ICD-9 code for 666.x during the admission to hospital for delivery, or within three days after the delivery date for outpatient deliveries, were classified as having postpartum hemorrhage. We also considered atonic postpartum hemorrhage only (666.1x; caused by inadequate uterine contraction)25 and inpatient postpartum hemorrhage only.
Most antidepressants are inactivated and eliminated within days to a week after use,26 yet the potential carry over effect on the risk of postpartum hemorrhage after discontinuation of is unknown. Consequently, we defined four mutually exclusive exposure groups based on the dispensing date of prescription and days’ supply dispensed relative to the delivery date:
Current exposure: women with a supply of antidepressants that overlapped with the delivery date
Recent exposure: women with a supply of antidepressants on at least one day in the month before the delivery date but not on the delivery date
Past exposure: women with a supply of antidepressants ending between five and one months before delivery
Unexposed: women who had no supply of antidepressants in the five months before delivery (reference group).
Because the degree of inhibition of serotonin reuptake is thought to play a role in the association between antidepressants and abnormal bleeding events,9 10 13 14 we classified the exposure groups by serotonin transporter affinity with previously reported serotonin transporter dissociation constants (Kd)27 28 29; a lower Kd reflects a higher affinity for the serotonin transporter. Antidepressants with serotonin transporter Kd 0-9.9 nM were classified as serotonin reuptake inhibitors and antidepressants with Kd ≥10 nM were classified as non-serotonin reuptake inhibitors (appendix 1). We considered only monotherapy with either type and therefore excluded from the primary analysis the 3278 women who were exposed to both drugs types (polytherapy) during the five months before delivery.
We considered also current, recent, and past exposure to specific antidepressants if at least 100 women with mood/anxiety disorders and current serotonin or non-serotonin reuptake inhibitor monotherapy were exposed to a given drug. Next we considered exposure to any antidepressant. We then defined exposure groups by class (appendix 1): selective serotonin reuptake inhibitor, serotonin norepinephrine (noradrenaline) inhibitor, tricyclic, bupropion, and other antidepressants, excluding from this analysis women exposed to more than one class.
Baseline information on comorbidities and healthcare use was obtained from data between one and five months before delivery, before the recent and current exposure windows. Potential confounders included risk factors for postpartum hemorrhage6 25 30 31: delivery year, age (as a quadratic spline), race, multiple pregnancy, coagulopathies, diabetes (classified as diagnosis and no dispensing of antidiabetic drugs, no diagnosis but with dispensing, diagnosis and dispensing); proxies of mood/anxiety disorder severity (number of outpatient (0, 1, 2-5, 6-9, ≥10) and inpatient (0, 1, ≥2) diagnoses of mood/anxiety disorder); other indications for antidepressants (psychotic disorder, other mental health disorder, pain related diagnoses); use of other psychotropic drugs (anticonvulsants and benzodiazepines); other drugs associated with risk of bleeding (aspirin, heparin, and warfarin); and proxies of comorbidity (number of outpatient visits (quadratic spline) and days in hospital (0, 1-2, 3-4, 5-6, ≥7)).32 Potential mediators of postpartum hemorrhage—that is, risk factors that could be affected by antidepressant use and consequently would not be confounders33—included delivery characteristics (short labor, long labor, forceps or vacuum delivery, or induced labor6 7 25 30 31) as documented during the admission for delivery or within three days after the delivery date for outpatient deliveries.
We compared the risk for postpartum hemorrhage between women with exposure to antidepressants and the unexposed group using relative risks and risk differences and their corresponding 95% confidence intervals from generalized estimating equations.34 35 Models were adjusted for potential confounders and robust variances were used to account for correlations among multiple pregnancies in the same woman.35 Furthermore, we used high dimensional propensity score (hdPS) methods to empirically identify and adjust for additional factors that might be surrogates for unmeasured confounders in the databases.36 37 We excluded 2.5% of women on both extremes of the distribution of the propensity score and adjusted logistic regression models for 10ths of the score, which was estimated from investigator defined covariates and 200 empirically identified variables. To support the validity of our outcome definition, we also assessed the association between well established risk factors for postpartum hemorrhage and occurrence of postpartum hemorrhage. Because women in this cohort were younger and more racially diverse than those in previous studies,19 20 we tested for multiplicative effect modification of current exposure by age (≥30) and race/ethnicity (white and non-white). To determine if there was an indirect effect of current exposure to serotonin or non-serotonin reuptake inhibitors on postpartum hemorrhage through any of the measured delivery characteristics that are risk factors for postpartum hemorrhage, we performed a mediation analysis38 39 using estimates from logistic regression models for postpartum hemorrhage and for the delivery characteristics. All analyses were conducted with SAS software, version 9.2 for Windows (SAS Institute, Cary, NC).
We performed eight sensitivity analyses to evaluate the robustness of the results. Because obesity is associated with postpartum hemorrhage and is poorly measured by healthcare use data, we conducted a sensitivity analysis40 41 to correct the relative risks for current use of serotonin and non-serotonin reuptake inhibitors for confounding by unmeasured obesity. Using data for women aged 18-55 with depression from the 2005-10 National Health and Nutrition Examination Survey (NHANES), we estimated the prevalence for obesity among those women who reported antidepressant use and those who reported no antidepressant use, accounting for the complex sampling design.42 We assumed the relative risks for obesity and postpartum hemorrhage was between 1.5 and 2.0 for the obesity correction.43 44 Second, we repeated the primary analysis but included all women with and without mood/anxiety disorders (n=2 759 414 pregnancies). Third, we restricted the study population to the 1 248 875 (45.3% of the overall study population) women who met requirements for Medicaid enrollment and eligibility throughout pregnancy; baseline covariates were assessed with data between one month and nine months before delivery and the past exposure window began at nine months before delivery. Fourth, we restricted the cohort with mood/anxiety disorder diagnoses to women without a diagnosis of hypertension or use of antihypertensive drugs. While hypertension is a risk factor for postpartum hemorrhage,7 31 increased blood pressure can be a consequence of using certain antidepressants45 46 so we avoided stratification by hypertension in the primary analysis.47 Fifth, in a dose analysis, we categorized women with current serotonin and non-serotonin reuptake inhibitor monotherapy according to the highest antidepressant dose that overlapped with the delivery date and compared them with women without exposure to antidepressants. Dose levels were defined according to Goodman and Gilman’s usual dose (mg/day)26: low is less than the lowest usual dose, medium is less than or equal to the midpoint of the usual dose range, high is more than the midpoint of the usual dose range (appendix 1). Because of small numbers, we combined medium and high doses for non-serotonin reuptake inhibitors. Sixth, we considered users of polytherapy in each exposure window—that is, women with current exposure to both serotonin and non-serotonin reuptake inhibitors—and determined their risk of postpartum hemorrhage compared with unexposed women. Seventh, we restricted the current exposure group to women who likely had the highest adherence—that is, those with exposure to the drug on more than 75% of days during the three months before delivery. Finally, in an exposure sensitivity analysis, we classified women with an antidepressant dispensed less than 14 days before delivery, regardless of days of supply on the delivery date, as having current exposure.
Among pregnant women with mood/anxiety disorders, 15.1% had a supply of antidepressants that overlapped their delivery date. There were 12 710 (12%) women with current exposure to serotonin reuptake inhibitor monotherapy and 1495 (1.4%) women with current exposure to non-serotonin reuptake inhibitor monotherapy. Compared with unexposed women, more women in the exposed groups were older and white and had hypertension, admission to hospital with a diagnosis of anxiety/mood disorders, other antidepressant indications, and other psychotropic drugs (table 1⇓). There were, however, fewer differences in patients between groups defined by current, recent, and past exposures to serotonin and non-serotonin reuptake inhibitors. Overall, 91% of women delivered in a hospital. Among women who delivered in a hospital, 69% delivered on the admission date and 95% delivered within one day of being admitted.
The risk of postpartum hemorrhage was 2.4% among women without diagnoses of mood/anxiety disorder and without exposure to antidepressants in the five months before delivery and 2.8% among women with mood/anxiety disorders but without exposure to antidepressants. The risk of postpartum hemorrhage was 4.0% in current users of serotonin reuptake inhibitors and 3.8% in current users of non-serotonin reuptake inhibitors; for recent users the risk was 3.2% and 3.1%, respectively, and for past users risk was 2.5% and 3.4%, respectively. Compared with women without each predisposing factor, women with pre-eclampsia (relative risk 1.43, 95% confidence interval 1.26 to 1.64), labor induction (1.56, 1.42 to 1.72), and forceps or vacuum delivery (1.39, 1.21 to 1.59) had an increased risk for postpartum hemorrhage. After adjustment for potential confounders, compared with unexposed women, current users of a serotonin reuptake inhibitor had a 1.47-fold increased risk for postpartum hemorrhage (95% confidence interval 1.33 to 1.62) and the current non-serotonin reuptake inhibitor monotherapy group had a 1.39-fold increased risk (1.07 to 1.81) (table 2⇓). In absolute terms, women with current exposure to serotonin reuptake inhibitors had an adjusted excess risk of 1.26% (0.90% to 1.62%), with a number needed to harm of 80; the figure for women with current non-serotonin reuptake inhibitor exposure was 1.03% (0.07% to 1.99%), with a number needed to harm of 97. Adjustment for additional potential confounders through hdPS analysis yielded odds ratios of 1.52 (1.35 to 1.71) for current use of serotonin reuptake inhibitors and 1.39 (1.03 to 1.89) for current use of non-serotonin reuptake inhibitors (table 3⇓). For recent use of antidepressants, the relative risk was 1.19 (1.03 to 1.38) for serotonin reuptake inhibitor and 1.17 (0.80 to 1.70) for non-serotonin reuptake inhibitor exposure. While the women with past exposure to serotonin reuptake inhibitor monotherapy did not have an increased risk for postpartum hemorrhage, in those with past exposure to non-serotonin reuptake inhibitor monotherapy the risk was slightly increased and nearly significant (table 2).⇓
The risk of postpartum hemorrhage was associated with current exposure to specific serotonin reuptake inhibitor antidepressants, the highest risk being with venlafaxine (table 4⇓). Compared with unexposed women, women with current exposure to any antidepressant had an increased risk for postpartum hemorrhage (relative risk 1.44, 1.32 to 1.58), as did women with any recent exposure (1.21, 1.06 to 1.38), while women with any past exposure did not have an increased risk (table 5⇓). As a class, current selective serotonin reuptake inhibitor monotherapy was associated with postpartum hemorrhage (1.42, 1.27 to 1.57), as was current serotonin norepinephrine reuptake inhibitor (1.90, 1.37 to 2.63) and tricyclic monotherapy (1.77, 0.90 to 3.47).
There was evidence of multiplicative effect modification by age (P=0.02 for the exposure-age interaction term); there was no association between current exposure to non-serotonin reuptake inhibitor and postpartum hemorrhage among women aged at least 30 (fig 2⇓ and appendix 2). None of the delivery characteristics (short or long labor, instrumental delivery, labor induction) were substantial mediators of the current associations between serotonin or non-serotonin reuptake inhibitor and postpartum hemorrhage.
External adjustment for obesity attenuated the relative risks to between 1.3 and 1.4 for current serotonin reuptake inhibitors and about 1.3 for current non-serotonin reuptake inhibitors. A dose response was indicated for current exposure to serotonin reuptake inhibitors but not for non-serotonin reuptake inhibitors (table 6⇓). The relative risk was 1.42 (0.97 to 2.08) for current polytherapy, 1.68 (0.94 to 3.00) for recent polytherapy, and 0.93 (0.60 to 1.44) for past polytherapy. All other sensitivity analysis results are summarized in figure 2 (numbers and risks are listed in appendix 2).⇑ The confidence intervals were wider for the non-serotonin reuptake inhibitor than the serotonin reuptake inhibitor groups. Past use of non-serotonin reuptake inhibitors was not associated with postpartum hemorrhage in the cohort of women both with and without diagnoses of mood/anxiety disorder and the cohort of women who met eligibility criteria throughout pregnancy and had mood/anxiety disorders. Women with the greatest amount of antidepressant coverage in the 90 days before delivery in the current serotonin reuptake inhibitor exposure group had a 1.54-fold increased risk of postpartum hemorrhage (95% confidence interval 1.36 to 1.76), and the corresponding relative risk for the current non-serotonin reuptake inhibitor exposure group was 1.25 (0.85 to 1.83). The relative risks ranged from 1.32 to 1.54 in current users of serotonin reuptake inhibitor monotherapy and from 1.31 to 1.69 for current users of non-serotonin reuptake inhibitor monotherapy when we changed the exposure and outcome definitions and when we restricted analysis to women with depression and without hypertension.
In this population of pregnant women, use of serotonin or non-serotonin reuptake inhibitors at the time of delivery was associated with about a 1.4 to 1.5-fold increase in risk of postpartum hemorrhage and 1.0% to 1.3% excess risk. Selective serotonin reuptake inhibitor, serotonin norepinephrine reuptake inhibitor, and tricyclic monotherapy at the time of delivery were associated with postpartum hemorrhage. Any exposure to antidepressants in the month before delivery was associated with about a 1.2-fold increased risk of postpartum hemorrhage. While exposure to serotonin reuptake inhibitors one to five months before delivery was not associated with postpartum hemorrhage, exposure to non-serotonin reuptake inhibitors during that time frame was associated with a nearly significant increased risk.
Comparison with previous studies
The association we found between selective serotonin reuptake inhibitor antidepressants and postpartum hemorrhage was slightly higher than that reported by Salkeld and colleagues from the Ontario administrative databases.19 Our exposure window covered only the delivery date, whereas their primary exposure window covered the 90 days before delivery. As suggested by our analysis of recent exposure, their exposure window could have been broader than the true exposure-risk window, possibly leading to downward bias. Our estimates were more precise than those from Ontario because of the large number of exposed cases. The association we found between any current exposure to antidepressants and postpartum hemorrhage was similar to the association between antidepressants prescribed during antenatal care and bleeding during delivery reported by Reis and colleagues.20 We attempted to reduce potential confounding by factors associated with underlying mood/anxiety disorders by restriction to women with diagnoses of mood/anxiety disorder, adjusting for proxies of severity of indication, and using high dimensional propensity score analysis whereas previous studies did not.
Possible explanations of results
We found no evidence that the delivery characteristics mediated the association. Along with cofactors, serotonin contributes to platelet aggregation, and the association between serotonin reuptake inhibitor and bleeding could be explained, at least in part, by the inhibition of serotonin reuptake by platelets, which cannot synthesize serotonin.15 Normally, excessive uterine bleeding from the site of placental implantation is prevented by uterine myometrial contraction.25 Serotonin reuptake inhibitors increase extracelluar serotonin concentrations,48 and the myometrium contracts in response to serotonin.49 Besides interfering with hemostasis, serotonin reuptake inhibitors could also interfere with serotonin mediated uterine contraction and contribute to atonic postpartum hemorrhage.
When we consider the hypothesis that the risk of bleeding increases with exposure to antidepressants that block serotonin reuptake,50 the association between non-serotonin reuptake inhibitor monotherapy and postpartum hemorrhage is unexpected. The positive associations between both serotonin and non-serotonin reuptake inhibitors and postpartum hemorrhage could indicate that all antidepressants impact on risk of postpartum hemorrhage. We cannot rule out the possibility, however, that the results are confounded by unmeasured factors associated with the underlying disorders. While we did not hypothesize it a priori, it is possible that non-serotonin reuptake inhibitors could affect postpartum hemorrhage through pathways other than those potentially mediated by serotonin reuptake inhibition. Trazodone, a non-serotonin reuptake inhibitor, is a 5-HT2A receptor antagonist26 and serotonin induced myometrial contraction is mediated by this receptor,49 consequently blockade of the 5-HT2A receptor could explain an association between trazodone use and atonic postpartum hemorrhage. It is possible that some of the positive findings are caused by chance. Overall, monotherapy with a non-serotonin reuptake inhibitor was rare in the cohort, comprising only 12.1% of antidepressant users. The results for non-serotonin reuptake inhibitors were much less precise than those for serotonin reuptake inhibitors, and some of the sensitivity analyses attenuated results toward the null, particularly for past use of non-serotonin reuptake inhibitors. It is possible that we failed to control for confounders that were particularly important among this group of women such as alcohol abuse or smoking. Bupropion, which is also indicated for smoking cessation,51 accounted for nearly 80% of current use of non-serotonin reuptake inhibitors. It is unclear, however, whether smoking is positively associated with postpartum hemorrhage.43 52 53
Though adjustment for many measured potential confounders with standard multivariable regression and high dimensional propensity score analysis did not affect the results, the associations could reflect unmeasured confounding. Because the use of many antidepressants, not just serotonin reuptake inhibitors, near the time of delivery was associated with postpartum hemorrhage, it is possible that our results could potentially reflect residual confounding by unmeasured behavioral factors associated with depression and antidepressant use. These factors include inadequate diet and the use of tobacco, alcohol, and illegal drugs,54 55 which are inadequately captured by data on healthcare use. Although mood/anxiety disorders were not strongly associated with postpartum hemorrhage, it is possible that the unmeasured factors that are associated with the severity of the disorder could have confounded the results. Based on external adjustment, our results could have been attenuated slightly had we been able to adjust for body mass index (BMI). Residual confounding by use of over the counter aspirin is probably small, given the expected low prevalence of use in the third trimester.56 Misclassification of exposure is also a potential limitation because we cannot confirm that women were taking antidepressants on the days we assumed.57 No information was available regarding antidepressant use during admission to hospital, including the admission for delivery. We expect that this had a minimal impact on the misclassification of antidepressant use near the time of delivery because 95% of women delivered within one day of being admitted to hospital. Misclassification would be non-differential and would tend to attenuate the true association. The result including only women with the greatest amount of antidepressant coverage in the 90 days before delivery in the current exposure groups is reassuring because they were less likely to be affected by misclassification as these women were likely to have been taking antidepressants near the time of delivery. Furthermore, we did not have adequate power to examine the association between antidepressants and severe postpartum hemorrhage leading to blood transfusion or mortality. Finally, outcome misclassification is another source of bias, which would likely result in an underestimation of a true association. The following support the validity of the outcome definition. The positive predictive value for postpartum hemorrhage was 85% based on ICD-9 codes 666.0x-666.2x in discharge data from California hospitals.58 The magnitudes of association between well established risk factors for postpartum hemorrhage and postpartum hemorrhage were similar to those reported in previous studies.6 7 30 31 Moreover, the risk of postpartum hemorrhage was in line with US national statistics.5
Generalizability of results
This study was conducted among women enrolled in the US Medicaid program, which covers the healthcare of 40% of all pregnant women with deliveries in the US.59 Women in the cohort were low income, tended to be younger, and were more likely to be black compared with national figures for pregnant women.60 Nonetheless, the baseline risk of postpartum hemorrhage in our population was 2.9%, which is similar to that of the general US population5 but lower than the risks reported for Canada and Australia.8 Results suggested that there was no association with non-serotonin reuptake inhibitors in older women. Because of the small numbers, however, this should be confirmed in another study. Furthermore, we found no evidence of effect modification by race or ethnicity. Based on these findings, we anticipate that the results will apply to other populations with a similar baseline risk for postpartum hemorrhage.
This study is the first to report an association between exposure to antidepressants at the time of delivery and risk of postpartum hemorrhage in a US population and in a population with a diagnosis of depression. The large study size permitted us to consider specific antidepressants, yet, despite the large cohort, there were still too few women with postpartum hemorrhage exposed to certain antidepressants, including trazodone and mirtazapine, to produce stable estimates. Also, the large size allowed us to define the primary exposure window as the delivery date, rather than using a wide window to gain enough cases for precise estimates, which can bias results downward.
Our study suggests that all classes of antidepressants are associated with an increased risk for abnormal bleeding. The magnitudes of association were in the range of those previously reported from Canadian and Swedish populations,19 20 and we found that serotonin norepinephrine reuptake inhibitors and tricyclics were also associated with increased risk. Our findings regarding non-serotonin reuptake inhibitor antidepressants were unexpected and should be confirmed. Additional studies that account for clinical measures of severity of depression and behavioral factors associated with antidepressant use near the end of pregnancy are warranted to rule out the possibility that the associations with serotonin and non-serotonin reuptake inhibitors reflect confounding. Although we cannot rule out residual confounding, our study indicates that there might be about one excess case of postpartum hemorrhage for every 80 to 100 women using antidepressants near the time of delivery, if we assume causality. The absolute increase in risk associated with antidepressant exposure in the month before delivery is small, but women and their physicians should be aware of the potential risks when making treatment decisions near the end of pregnancy.
What is already known on this topic
Use of serotonin reuptake inhibitor antidepressants is associated with an increased risk for gastrointestinal bleeding
What this study adds
Antidepressant use near the time of delivery is associated with an increased risk of postpartum hemorrhage
Additional studies that account for clinical measures of depression severity and behavioral factors associated with depression are warranted
Cite this as: BMJ 2013;347:f4877
Contributors: KP, SH-D, KFH, PLW, KBM, EDA, and SS were responsible for conception and design. SH-D and SS acquired the data which were analyzed and interpreted by KP, SH-D, KFH, PLW, KBM, HM, and SS. All authors critically revised the manuscript for important intellectual content. SH-D and SS obtaining funding. KP and HM provided administrative, technical, and material support. KP drafted the manuscript and is guarantor.
Funding: This work was supported by the Agency for Healthcare Research and Quality (AHRQ) (grant R01HS018533 to SH-D). KP was supported by training grant T32HD060454 in reproductive, perinatal, and pediatric epidemiology from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health. SS was supported by a mid-career development award grant K02-HS017731 from the AHRQ. The AHRQ had no role in the study design and the collection, analysis, and interpretation of data and writing of the article and the decision to submit it for publication. Research was conducted independently of the sponsor and funders.
Ethical approval: This project was approved by Brigham and Women’s Hospital and Harvard School of Public Health Institutional Review Boards and a data use agreement was approved by the Centers for Medicare and Medicaid Services (CMS).
Competing interests: All authors have completed the ICMJE uniform disclosure form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: the Pharmacoepidemiology Program at the Harvard School of Public Health receives funding from Pfizer and Asisa; SH-D has consulted for GSK and Novartis; EA received research funding from the National Institute on Alcohol Abuse and Alcoholism (NIAAA), the National Institute of Mental Health (NIMH), the National Institute on Drug Abuse (NIDA), Pfizer, Otsuka, Janssen, (all schizophrenia), Pine Rest Foundation (ECT/dementia), University of Chicago, Michigan State University, Pine Rest Foundation (depression screening tool, biomarkers of depression and suicidality), Novartis (antidepressant development), Pine Rest Foundation, AssureRx, Priority Health (pharmacogenomic testing in depression).
Data sharing: No additional data available.
This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 3.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/3.0/.