Dr. Solomon has received research support from Abbott, Amgen, and Lilly.
Brigham and Women's Hospital, Harvard Medical School, and Harvard School of Public Health, Boston, Massachusetts, and Duke Clinical Research Institute, Duke University School of Medicine, Durham, North Carolina
Dr. Setoguchi has received honoraria (less than $10,000) from Sanofi-Aventis and has received research support from Johnson & Johnson. The Pharmacoepidemiology Program at the Harvard School of Public Health receives funding from Pfizer and Asisa.
To describe patterns of disease-modifying antirheumatic drug (DMARD) use during pregnancy in a population-based cohort, and to evaluate the association between autoimmune disease, DMARDs, corticosteroids, and nonsteroidal antiinflammatory drugs (NSAIDs) and preeclampsia.
Using health care utilization databases from British Columbia (1997–2006), we compared the risk for preeclampsia among 44,786 women with and without autoimmune disease with study drug dispensings before pregnancy (past users) and before and during the first 20 gestational weeks (continuous users). Relative risks (RRs) and 95% confidence intervals (95% CIs) were estimated.
Only 414 women (0.1%) had a DMARD dispensing during pregnancy. The incidence of preeclampsia was 2.3% for past DMARD users, 2.7% for past corticosteroid users, and 2.9% for past NSAID users. Compared to past users, the continuous DMARD user RR was 2.29 (95% CI 0.81–6.44), and was 0.89 (95% CI 0.51–1.56) for corticosteroid and 0.84 (95% CI 0.63–1.10) for NSAID users. Compared to women without autoimmune disease, the delivery year–adjusted RR was 2.02 (95% CI 1.11–3.64) for women with systemic lupus erythematosus (SLE). The DMARD results were attenuated when antimalarials were excluded, and the delivery year–adjusted RR was 0.95 (95% CI 0.25–3.55) when the DMARD analysis was restricted to women with autoimmune disease.
Few women were exposed to DMARDs during pregnancy. We observed a 2-fold increased risk of preeclampsia among women with SLE and a nonsignificant increase in risk in DMARD users. The DMARD and preeclampsia association was attenuated when antimalarials were excluded and null when restricted to women with autoimmune disease, which suggests the association is likely due to greater autoimmune disease severity in DMARD users.
Preeclampsia is a serious complication of pregnancy that can compromise the health of pregnant women and neonates (1). Studies have consistently shown that women with systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA) have an increased risk for preeclampsia; SLE is associated with a 2–6-fold increase in risk and RA is associated with a 1.4–2-fold increase in risk (2–6). Limited data exist on the association of other autoimmune diseases, such as multiple sclerosis and inflammatory bowel disease (IBD), and preeclampsia (7, 8).
Disease-modifying antirheumatic drugs (DMARDs) are a part of standard treatments for autoimmune disease, and their use during pregnancy may be necessary to avoid flares of disease that may compromise the health of the expectant mother and fetus (9, 10). It is possible that DMARDs may decrease the risk of preeclampsia by diminishing the maternal immune response to the placenta, which may play a role in the development of preeclampsia (11). However, it is critical to account for underlying autoimmune disease when assessing the relationship between DMARD use and preeclampsia, given the well-reported association between various autoimmune diseases and preeclampsia (2–6).
Corticosteroids and nonsteroidal antiinflammatory drugs (NSAIDs) are commonly prescribed for pregnant women with autoimmune disease (9, 10, 12). Corticosteroids may mitigate the systemic inflammatory response associated with preeclampsia (13), and aspirin, a nonselective NSAID, modestly decreases the risk for preeclampsia (14).
In this study, we first described DMARD, oral corticosteroid, and NSAID use during pregnancy in a population-based cohort. We then assessed the impact of DMARDs, oral corticosteroids, and NSAIDs on the risk for preeclampsia, adjusting for autoimmune diseases and factors related to autoimmune disease severity.
Significance & Innovations
In this population-based cohort, disease-modifying antirheumatic drug (DMARD) therapy was rare during pregnancy, and DMARD use decreased as pregnancy advanced.
Antimalarials were the most commonly dispensed DMARD during pregnancy.
Women who used DMARDs during pregnancy had a nonsignificant increased risk for preeclampsia, which was attenuated when antimalarial users were excluded.
Systemic lupus erythematosus was associated with preeclampsia, whereas inflammatory bowel disease was not.
MATERIALS AND METHODS
Comprehensive health insurance is provided to nearly all residents in the province of British Columbia (BC), Canada (15). BC Ministry of Health administrative databases contain diagnostic and procedure information from all physician services and hospitalizations. These population-based health care utilization databases were used to identify a cohort of pregnant women and infants. Databases were linked to the PharmaNet database, which contains all nonhospital pharmacy dispensing claims in BC. This study was approved by the Brigham and Women's Hospital Institutional Review Board and signed data use agreements were in place.
Woman–infant linkage and gestational timing.
Women were linked to infants by family insurance number and by delivery month and year. The delivery date was estimated from inpatient delivery discharges. Women who had more than 1 infant with the same delivery date were identified as having a multifetal gestation. Of all identified deliveries, 94% were successfully linked to infants. Because neither gestational length nor date of the last menstrual period (LMP) is available in health care utilization databases, the LMP was assigned as 245 days prior to delivery for preterm deliveries, identified from International Classification of Diseases, Ninth Revision (ICD-9) or International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) codes 644.0x, 644.2x, 765.x, P05.x, P07.x, or O60.1. For all other deliveries, we assigned the LMP as 280 days prior to the date of delivery; 280 days is the mean duration of human gestation (16). This method accurately classifies gestational age within 2 weeks for nearly 70% of preterm and 93% of term deliveries in this database (17). The baseline period was the year before the estimated LMP (Figure 1).
Only pregnancies in which women had a live birth between October 1, 1997 and January 31, 2006 and had continuous health care enrollment from 1 year prior to their LMP until 2 months after their delivery date were eligible for the study. We described baseline characteristics and patterns of DMARD, corticosteroid, and NSAID use for the 306,831 pregnancies from 224,827 women that met these criteria. To minimize the number of women who were taking DMARDs for non–autoimmune disease indications, women with an ICD-9 or ICD-10 code for organ transplant (V42.x or Z94.x–Z95.x), malaria (084.x or B50.x–B64.x), or cancer (140.x–208.x or C00.x–C97.x) were excluded from subsequent analyses that assessed the association between study drugs and preeclampsia (18–20).
Definitions of autoimmune disease.
Autoimmune disease was defined by the presence of an inpatient or outpatient diagnosis for a specific autoimmune disease on at least 2 dates during the baseline period. We used ICD-9 or ICD-10 codes 714.x or M05.x–M09.x for RA; 696.x or L40.x–L44.x for psoriasis; 710.x or M32.x, M34.x, or M35.x for SLE; 555.x or K50.x for Crohn's disease; 556.x or K51.x for ulcerative colitis; and 340.x or G35.x for multiple sclerosis. Previous studies suggest that the risk for preeclampsia in RA and psoriasis is similar (2, 3), and therefore they were grouped together as RA/psoriasis. Similarly, Crohn's disease and ulcerative colitis were grouped together as IBD.
Preeclampsia is clinically defined as hypertension and proteinuria after the 20th week of gestation (21). We identified preeclampsia using inpatient or outpatient ICD-9 codes 642.4x, 642.5x, 642.6x, or 642.7x or ICD-10 codes O11.x, O14.x, or O15.x between gestational week 20 and 1 month after delivery. Most cases of preeclampsia were identified with at least 1 inpatient code for the condition (94.2%). The positive predictive value for specific forms of preeclampsia and eclampsia ranged from 41.7–84.8% in a validation study of ICD-9 hospital discharge codes (22), and was reported to be 84% for superimposed and severe preeclampsia (23). The positive predictive value for preeclampsia based on ICD-9 codes for mild and severe disease was 93% in the Swedish Medical Birth Register (24).
Drug exposure definition.
DMARD use, including biologic agents, and oral corticosteroid and NSAID use (the generic names are listed in Supplementary Appendix A, available in the online version of this article at http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)2151-4658) were determined by the presence of a pharmacy dispensing record during the baseline period and between gestational weeks 0 and 20. Minocycline is an antibiotic that is also used as a DMARD (25); the drug was not considered as a DMARD in this study because minocycline was often prescribed in conjunction with a diagnosis of acne. Infusible DMARDs were not considered. We hypothesized that DMARDs affect the maternal immune response to implantation and placentation; therefore, the exposure window covered these stages of pregnancy (26). The exposure window ended just before 20 gestational weeks when preeclampsia, by definition, can be diagnosed. We anticipated that autoimmune disease would be less common and less severe among women who never used the study drugs than women who did. Therefore, we compared the risk for preeclampsia in 3 groups of users categorized by the timing of use rather than by use/nonuse to reduce confounding by the presence of autoimmune disease and disease severity. Women who were exposed to a study drug only during baseline were classified as past users and served as the reference group for these analyses. Continuous users were women who were exposed to the study drugs during baseline and between gestational weeks 0 and 20, whereas first-time users were women who were exposed to the study drugs only between gestational weeks 0 and 20. Based on the timing of treatment, disease activity is likely to be lowest in past users (27, 28).
Potential confounders included known risk factors for preeclampsia (29) and factors related to autoimmune disease severity (30) during baseline: age (<24, 24–29, 30–34, ≥35 years), primiparity (no prior deliveries recorded in the database for a woman; multiparas are only identified when prior deliveries are recorded), multifetal gestation, hypertension defined by the presence of diagnosis codes and/or dispensing for antihypertensive medications, pregestational diabetes mellitus, obesity, asthma, renal disease, type of autoimmune disease, ≥2 rheumatology visits, ≥1 joint radiograph and ≥1 platelet count ordered (as markers of RA severity), type of autoimmune disease, and total baseline days' supply of study drug prescriptions: DMARDs (a continuous, linear variable), corticosteroids (0, 1–6, 7–89, ≥90), or NSAIDs (0, 1–6, 7–89, ≥90). Number of distinct prescription drugs dispensed excluding the study drugs (<2, 2–3, ≥4) and physician visits (0–8, 9–14, 15–24, ≥25) during baseline were considered as markers of comorbidity (31).
First, we described the number of women who were dispensed DMARDs in 3-month intervals before pregnancy and in each trimester. We also described the number of women who were dispensed study drugs during pregnancy and the median days' supply of study drugs dispensed during pregnancy.
We used modified Poisson regression models to estimate relative risks (RRs) and 95% confidence intervals (95% CIs); this method directly estimates the RR for binary outcomes, whereas logistic regression, an alternative, only approximates the RR (32). All models were adjusted for delivery year and robust variances accounted for correlations among women who had more than 1 pregnancy in the database (33). To support the validity of our outcome definition, we assessed the association between well-established risk factors for preeclampsia and preeclampsia. The primary analysis compared the risk of preeclampsia among continuous DMARD, corticosteroid, and NSAID users to the risk among past users, separately. We also compared first-time users with past users; because only 1 first-time DMARD user case was identified, the RR for DMARD first-time versus past use was not reported. Models were additionally adjusted for preeclampsia risk factors. The fully-adjusted models also were adjusted for type of autoimmune disease and for factors related to autoimmune disease and health care utilization. We conducted multiple sensitivity analyses. First, we considered baseline drug use as part of the exposure and consequently did not adjust for baseline days' supply of the study drug of interest in the continuer versus past user analysis. Second, we excluded women who used only antimalarial DMARDs and were therefore likely to have less severe autoimmune disease. Finally, we assumed that all pregnancies had an error in gestational age of ±2 weeks; therefore, we shortened the exposure window to between 2 and 18 gestational weeks and we set the past use window to end 2 weeks prior to the LMP.
To understand the role of autoimmune disease in the association between DMARDs and preeclampsia, we first compared the delivery year–adjusted risk for preeclampsia among women with specific types of autoimmune disease to women without autoimmune disease. In an attempt to disentangle the association between autoimmune disease and study drugs with preeclampsia, we assessed the association between the study drugs dispensed between the LMP and gestational week 20 and preeclampsia only among women with autoimmune disease, adjusting for preeclampsia risk factors, type of autoimmune disease, factors related to autoimmune disease, and health care utilization. In this analysis, the reference group was comprised of women who did not use the study drugs between the LMP and week 20. All analyses were conducted with SAS for Windows, version 9.2.
Among 306,831 pregnancies, the median age was 30 years and chronic conditions such as pregestational hypertension and diabetes mellitus were rare (Table 1). Only 0.3% of women had RA or psoriasis. Within this cohort, 1,226 women (0.4%) used a DMARD in the year before pregnancy, and 414 women (0.1%) were dispensed a DMARD during pregnancy. When accounting for days' supply and the dispensing date, a maximum of 555 women (0.2%) were exposed to DMARDs during pregnancy. The number of women dispensed DMARDs decreased from periods before pregnancy through the third trimester (Figure 2). The most common DMARDs used during pregnancy were hydroxychloroquine (n = 114), chloroquine phosphate (n = 109), azathioprine (n = 91), and sulfasalazine (n = 60). Only 4 women had dispensing records for biologic DMARDs during pregnancy (all for etanercept). There were 2,444 women (0.8%) with at least 1 dispensing for corticosteroids during pregnancy and 6,315 women (2.1%) with at least 1 dispensing for NSAIDs during pregnancy. The median days' supply for a DMARD prescription during pregnancy was 42 days, whereas it was 12 days for corticosteroids and 10 days for NSAIDs. Compared with multiparous women, primiparous women had an RR for preeclampsia of 2.03 (95% CI 1.90–2.17). Compared with singleton pregnancies, women with multifetal pregnancies had a 3.97-fold increased risk of preeclampsia (95% CI 3.56–4.43). Women with pregestational diabetes mellitus had a 3.64-fold increased risk of preeclampsia (95% CI 3.09–4.29) compared with women without diabetes mellitus. Women ages ≥40 years also had an increased risk for preeclampsia.
Table 1. Baseline characteristics among all pregnancies identified from the British Columbia health care utilization databases between 1997 and 2006*
Pregnancies (n = 306,831)
Values are the number (percentage) unless otherwise indicated. IQR = interquartile range; RA = rheumatoid arthritis; SLE = systemic lupus erythematosus; IBD = inflammatory bowel disease; MS = multiple sclerosis.
Age, median (IQR)
Hypertension (diagnosis or antihypertensive)
In the subsequent analyses, 18,867 pregnancies were excluded because of a record for organ transplant (n = 76), malaria (n = 406), or cancer (n = 18,385), leaving 287,964 pregnancies. After restriction, hydroxychloroquine (n = 100), chloroquine phosphate (n = 91), azathioprine (n = 71), and sulfasalazine (n = 55) were still the most commonly used DMARDs during pregnancy. Between the LMP and gestational week 20, 302 pregnancies were exposed to a DMARD, 1,342 were exposed to a corticosteroid, and 4,728 were exposed to an NSAID. Among these women, 213 used multiple study drugs: 59 pregnancies were exposed to a DMARD and corticosteroid, 16 were exposed to a DMARD and NSAID, 129 were exposed to a corticosteroid and NSAID, and 9 were exposed to all 3 drug categories.
Few women taking corticosteroids and NSAIDs had autoimmune disease diagnoses, but many had other indications, such as osteoarthritis, back/neck disorders, soft tissue disorders, sprains/strains, headache/migraine, and premenstrual tension syndrome/other menstruation disorders (Table 2).
Table 2. Baseline characteristics by study drug and timing of use among pregnancies identified from the British Columbia health care utilization databases between 1997 and 2006 in which women had DMARD, corticosteroid, or NSAID use, but no diagnostic codes for organ transplant, malaria, or cancer*
DMARD use (n = 1,220)
Corticosteroid use (n = 7,282)
NSAID use (n = 36,284)
Past (n = 918)
Continuous (n = 183)
First time (n = 119)
Past (n = 5,940)
Continuous (n = 456)
First time (n = 886)
Past (n = 31,556)
Continuous (n = 2,036)
First time (n = 2,692)
Values are the number (percentage) unless otherwise indicated. DMARD = disease-modifying antirheumatic drug; NSAID = nonsteroidal antiinflammatory drug; IQR = interquartile range; RA = rheumatoid arthritis; SLE = systemic lupus erythematosus; IBD = inflammatory bowel disease; MS = multiple sclerosis; PMT = premenstrual tension syndrome.
Age, median (IQR)
Hypertension (diagnosis or antihypertensive)
Unspecified joint disorders
Soft tissue disorders
≥2 rheumatology visits
No. of physician visits, median (IQR)
No. of non–study drugs, median (IQR)
Total days' supply of baseline DMARDs, median (IQR)
Total days' supply of baseline corticosteroids
Total days' supply of baseline NSAIDs
The incidence of preeclampsia among women who continued DMARD use after the LMP was 7.7%, whereas it was 4.4% for corticosteroid users and 2.6% for NSAID users. The corresponding risks for past users of these medications were 2.3% for DMARDs, 2.7% for corticosteroids, and 2.9% for NSAIDs. The incidence of preeclampsia among first-time corticosteroid users was 3.7%, and it was 3.2% among first-time NSAID users. The risk was 2.4% among nonusers of these medications.
Known risk factors for preeclampsia, such as older age, multifetal gestation, and hypertension, were more common among continuous users compared to past and first-time study drug users (Table 2). Of the 302 women exposed to a DMARD during the first 20 weeks of gestation, 34.4% had at least 2 diagnoses for an autoimmune disease. RA/psoriasis, SLE, and IBD were more common among continuous DMARD and corticosteroid users compared to past and first-time users. Rheumatology visits occurred more frequently and health care utilization was the highest among the continuous users of study drugs.
Compared to past DMARD users, the delivery year–adjusted RR for preeclampsia among those who continued DMARD use was 3.22 (95% CI 1.56–6.66) and the fully-adjusted RR was 2.29 (95% CI 0.81–6.44); adjustment for autoimmune disease/factors related to autoimmune disease, particularly SLE, renal disease, and baseline corticosteroid use, attenuated the RR. When women who used only antimalarial DMARDs were excluded, the fully-adjusted RR was 1.23 (95% CI 0.30–5.02). We observed nonsignificant decreases in the risk for preeclampsia when comparing continuous corticosteroid and NSAID users to past users after adjusting for potential confounders (Table 3). If prior study drug use affects the risk of preeclampsia, it would be inappropriate to adjust for baseline use; the results did not change greatly when we did not adjust for baseline days' supply of the study drug of interest. Comparing first-time to past users, the fully-adjusted RR for preeclampsia among corticosteroid users was 1.35 (95% CI 0.94–1.94), and it was 1.17 (95% CI 0.94–1.45) for NSAID users. The results from the alternative exposure window sensitivity analysis were not very different from the original analysis with the exception of that from the DMARD continuer versus past user analysis, which was greatly attenuated (RR 0.93, 95% CI 0.31–2.80).
Table 3. RRs for preeclampsia comparing continuous, first-time, and past users by study drug group, restricted to women without diagnostic codes for organ transplant, malaria, or cancer, in the British Columbia health care utilization databases, 1997–2006*
Delivery year adjustment: 1997–2000, 2001–2003, 2004–2006.
Preeclampsia risk factor adjustment: delivery year, age (<24, 24–29, 30–34, ≥35 years), multifetal gestation, primiparity, hypertension (no hypertension, diagnosis but no antihypertensive, antihypertensive but no diagnosis, diagnosis and antihypertensive), diabetes mellitus, and obesity.
Full adjustment: preeclampsia risk factor adjustment plus asthma, renal disease, rheumatoid arthritis/psoriasis, systemic lupus erythematosus, inflammatory bowel disease, joint radiograph, ≥2 rheumatology visits, platelet count, physician visits (0–8, 9–14, 15–24, ≥25), number of non–study drugs (0–1, 2–3, ≥4), baseline days' supply of DMARDs (linear term), baseline days' supply of corticosteroids (0, 1–6, 7–89, ≥90), and baseline days' supply of NSAIDs (0, 1–6, 7–89, ≥90).
Not adjusted for diabetes mellitus, obesity, or asthma; adjusted for antihypertensive prescription or hypertension diagnosis.
Baseline days' supply of NSAIDs (0, 1–6, ≥7).
Not adjusted for baseline days' supply of corticosteroids.
Of the 1,744 women with at least 2 diagnoses of autoimmune disease, 49.8% had RA/psoriasis, 29.4% had IBD, 11.2% had SLE, and 10.6% had multiple sclerosis. Compared to women with no autoimmune disease, women with SLE were more likely to have preeclampsia (delivery year–adjusted RR 2.02, 95% CI 1.11–3.64), but women with IBD did not have a higher risk for preeclampsia (Table 4). The delivery year–adjusted RR for RA/psoriasis was 1.26 (95% CI 0.87–1.81).
Table 4. Delivery year–adjusted RRs for preeclampsia comparing women with specific autoimmune disease diagnoses to women without any autoimmune disease diagnoses, restricted to women without diagnostic codes for organ transplant, malaria, or cancer, in the British Columbia health care utilization databases, 1997–2006*
RR = relative risk; 95% CI = 95% confidence interval.
Delivery year adjustment: 1997–2000, 2001–2003, 2004–2006.
Systemic lupus erythematosus
Inflammatory bowel disease
No autoimmune disease
In the autoimmune disease cohort, 6.0% were exposed to DMARDs, 7.1% were exposed to corticosteroids, and 4.1% were exposed to NSAIDs between the LMP and week 20. The delivery year–adjusted RR comparing DMARD exposure between weeks 0 and 20 to no DMARD exposure during this period was 0.95 (95% CI 0.25–3.55) and the fully-adjusted RR was 0.62 (95% CI 0.10–3.77); there were only 3 exposed cases. The fully-adjusted RR for corticosteroid exposure was 1.25 (95% CI 0.44–3.56), and there was only 1 NSAID-exposed case among women with autoimmune disease.
In this population-based cohort, we observed that use of DMARDs was rare during pregnancy and that DMARD use decreased as pregnancy advanced. Antimalarials were the most commonly used DMARD during pregnancy. Preeclampsia occurred in 2.4% of pregnancies, which is similar to the incidence in other cohorts (34–36). Compared to women who used DMARDs only before pregnancy (past users), women who used DMARDs before pregnancy and between the LMP and week 20 (continuous users) had a nonsignificant increase in the risk for preeclampsia. SLE was associated with a 2-fold increased risk for preeclampsia, after adjusting for delivery year. The DMARD and preeclampsia association was attenuated when women who used only antimalarials were excluded from the analysis and when the population was restricted to women with autoimmune disease, suggesting that the observed risk for preeclampsia in DMARD users was likely due to underlying autoimmune disease.
We observed some patterns of DMARD utilization in our population-based cohort from BC that were similar to those among a cohort of pregnant women with RA from the US (12). Both cohorts had a decrease in DMARD exposure from the months prior to pregnancy through the third trimester. Hydroxychloroquine was the most commonly used DMARD during pregnancy in our cohort; it was the second most commonly used DMARD during pregnancy behind corticosteroids and biologic agents in the study by Kuriya et al (12). In contrast, biologic agents were only dispensed to 4 pregnant women in our cohort.
There appeared to be an increased risk for preeclampsia among corticosteroid and NSAID first-time users compared to past users. Inspection of the baseline characteristics by timing of study drug use indicated that the comparison groups were dissimilar with respect to factors related to autoimmune disease. Although we adjusted for factors related to autoimmune disease, clinical measures of disease severity were not available. Because first-time user groups may contain women with new-onset autoimmune disease, proxies for severity may not be captured during baseline. Therefore, the increased risk for preeclampsia among first-time corticosteroid and NSAID users may reflect residual confounding by postconception disease severity.
Prior studies have reported a 1.4–2-fold increased risk for preeclampsia among women with RA (2–6). Increased RA disease activity during pregnancy is also associated with lower birth weight and cesarean delivery (37). If preeclampsia risk is dependent on disease severity, an association between RA and preeclampsia in this study may have been diluted by unintentionally combining active and less active cases of RA. We did not observe a uniform increase in the risk for preeclampsia among all autoimmune disease subtypes. We found no association between IBD and preeclampsia and there was a small but nonsignificant increase in preeclampsia in RA/psoriasis patients, whereas we observed a strong association with SLE. Future studies that evaluate the association between DMARDs and preeclampsia should consider interactions with autoimmune disease subtype and severity.
We observed no difference in risk for preeclampsia between continuous and past corticosteroid and NSAID users. These results contrast with a nonsignificant increased risk for preeclampsia in the DMARD continuous/past user analysis. However, the association between DMARDs and preeclampsia was attenuated somewhat after adjustment for autoimmune disease, and the point estimate became null when the cohort was restricted to those with autoimmune disease. The difference between the results from adjustment and restriction by autoimmune disease is explained by low sensitivity but high specificity of ICD-9 and ICD-10 diagnosis codes for autoimmune disease. Autoimmune disease diagnoses were likely underrecorded because the physician services file only allows 1 diagnosis, and only one-third of women who used DMARDs during pregnancy were classified as having autoimmune disease, even after excluding non–autoimmune disease DMARD indications. Because of the low sensitivity for the ICD-9 and ICD-10 codes, power was low in the restricted analysis. However, due to the high specificity, those who were included in the restricted analysis are likely to have true disease and probably have more severe disease than those who were misclassified as not having disease. Results from the restricted analysis likely reflect less residual confounding by autoimmune disease, whereas results adjusted for misclassified autoimmune disease likely reflect residual confounding. These analyses indicated that the nonsignificant positive association between DMARDs and preeclampsia is likely due to underlying autoimmune disease or its severity.
Maternal decidual natural killer cells may impede early trophoblast invasion, leading to reduced placental profusion (38), which may set the stage for preeclampsia. A proinflammatory cytokine profile triggered by placental hypoxia may contribute to systemic inflammation observed in women with preeclampsia (38). Furthermore, Treg cells down-regulate self-reactive T cells and protect against autoimmune disease (39), and a paucity of Treg cells may trigger autoimmune disease and has also been proposed to play a role in the development of preeclampsia (40). Therefore, autoimmune disease and disease severity is a theoretical confounder for the relationship between DMARD use and preeclampsia. It is plausible that DMARDs may actually decrease the risk for preeclampsia among women with autoimmune disease by mitigating immune responses. Although our results did not show beneficial effects of DMARDs, a small beneficial effect cannot be ruled out.
Exposure assessment was based on prescription fills; therefore, we could not verify that cohort members were taking the prescriptions as assumed (41–43). Because some of the study drugs are contraindicated during pregnancy (9, 10) and women with RA tend to decrease their use during pregnancy (12, 37), we anticipate that some women who were classified as study drug users during pregnancy actually discontinued use. Underrecording of over-the-counter NSAID use and intermittent corticosteroid use likely resulted in exposure misclassification. These sources of misclassification would likely attenuate the relative risks. Alternatively, case–control surveillance studies rely on patients' recall to assess exposure; these studies experience differential misclassification where the direction and magnitude can be difficult to predict. Furthermore, misidentification of the LMP due to preterm deliveries would also result in exposure misclassification (44). This source of misclassification is more likely among women with preeclampsia because preterm delivery is more common among preeclamptic pregnancies (45, 46). However, the estimated date of the LMP was within 2 weeks of the true LMP for most preterm deliveries in our study; therefore, this source of exposure misclassification was likely minor. With the exception of the DMARD analysis, the results were not very sensitive to the alternative exposure definition that assumed a 2-week error in gestational age.
The use of population-based data is a major strength of our study; it allowed us to identify women who used DMARDs before or during pregnancy. The size of the population enabled us to assess the association between DMARDs and a maternal outcome, and to study the relationship between specific autoimmune diseases and preeclampsia. We observed the well-established associations between age, parity, multifetal gestations, and diabetes mellitus and preeclampsia (29), which support the validity of our outcome definition.
In conclusion, DMARD use during pregnancy was rare in this population-based cohort. DMARD continuation during pregnancy was associated with a nonsignificant increase in risk for preeclampsia; however, corticosteroid and NSAID continuation were not. The nonsignificant risk was attenuated when women with antimalarial use only, i.e., those likely to have less severe disease, were excluded, suggesting that the relationship between DMARD use and preeclampsia is likely due to underlying disease or severity of autoimmune disease. Larger studies with detailed clinical data are needed to confirm our findings and should investigate the relationship between DMARD use and preeclampsia among women with specific autoimmune diseases and should account for clinical measures of disease severity.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Ms Palmsten had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Palmsten, Hernández-Díaz, Kuriya, Solomon, Setoguchi.
Acquisition of data. Setoguchi.
Analysis and interpretation of data. Palmsten, Setoguchi.
The authors would like to thank Claire F. Canning, MA, for compiling the analytical data sets.