Thiopurines and Intrahepatic Cholestasis of Pregnancy in Systemic Lupus Erythematosus and Inflammatory Bowel Disease Pregnancies

Abstract

INTRODUCTION:

The objective of this study was to examine the effects of thiopurine exposure during pregnancy on intrahepatic cholestasis of pregnancy (ICP) in systemic lupus erythematosus (SLE) and inflammatory bowel disease (IBD) pregnancies using nationwide data from Swedish healthcare registers.

METHODS:

This register-based cohort study included all singleton pregnancies with prevalent SLE or IBD in Sweden, 2007–2022. Exposure was ≥1 dispensation for azathioprine or 6-mercaptopurine from last menstrual period date to 1 day before delivery or before ICP diagnosis vs no dispensation. ICP diagnosis was identified with ICD-10 code O26.6 in inpatient or outpatient care. Propensity score matching (1 thiopurine-exposed and 2 thiopurine-unexposed) controlled for confounding factors (e.g., maternal smoking, body mass index, parity, IBD subtypes, glucocorticoids, and advanced therapies). Modified Poisson models were used to estimate risk ratios (RRs) and 95% confidence intervals (CIs).

RESULTS:

In SLE, ICP occurred in 14 of 297 (4.7%) thiopurine-exposed pregnancies vs 11 of 594 (1.9%) matched unexposed pregnancies (adjusted RR 3.06, 95% CI 1.36–6.90; adjusted risk difference 4%, 95% CI 1%–6%). In IBD, ICP developed in 140 of 1,924 (7.3%) exposed pregnancies vs 36 of 3,848 (0.9%) matched unexposed pregnancies (adjusted RR 7.78, 95% CI 5.41–11.17; adjusted risk difference 6%, 95% CI 5%–8%). Sensitivity and subgroup analyses by ICP history, pregestational hypertension, renal diseases, liver diseases, glucocorticoid use, IBD subtypes, and thiopurine substance, and comparing thiopurines with tumor necrosis factor inhibitors exposure during pregnancy provided consistent results.

DISCUSSION:

In SLE and IBD, thiopurine exposure during pregnancy was associated with an increased risk of ICP, regardless of ICP history, hypertension, renal diseases, liver diseases, and glucocorticoid use.

INTRODUCTION

Thiopurines (azathioprine and 6-mercaptopurine) are generally considered low-risk for pregnancy and can be used to manage systemic lupus erythematosus (SLE) or inflammatory bowel disease (IBD) as maintenance therapy throughout pregnancy (1,2). Most previous studies did not report an increased risk of adverse pregnancy and fetal outcomes in thiopurine-treated patients vs untreated (3,4). However, little is known about intrahepatic cholestasis of pregnancy (ICP) associated with thiopurines. ICP is the most common pregnancy-related liver disorder. Severe ICP (bile acid level ≥40 μmol/L) can increase the risk of preterm delivery, meconium-stained amniotic fluid, and stillbirth (5–7). Several case reports of ICP in patients with thiopurine-treated IBD have been published (8–10). Findings from cohort studies are limited, especially for SLE. Three cohort studies on IBD pregnancies found higher risks of ICP in thiopurine-treated pregnancies vs untreated but were based on few ICP cases (16/1,000, 15/386, and 7/217 pregnancies) and with minimal confounding control (11–13). Correspondingly, a meta-analysis of these studies showed a significant association between thiopurines and ICP (14). None of the previous studies examined early-onset ICP.

In April 2024, the US Food and Drug Administration issued an alert regarding a potential link between ICP and thiopurines based on published case reports, recommending drug discontinuation on ICP development, which seemed effective in the case reports (8,9,15). The alert may create challenges in clinical management for IBD and SLE pregnancy (limited alternatives in SLE and constraints in cost/access to advanced therapies in IBD in certain settings). To facilitate clinical communication, reliable and representative data on whether or not and to what extent thiopurines increase ICP risk are warranted. We aimed to assess the effect of thiopurines on ICP in SLE and IBD pregnancies. In addition, we described spontaneous and iatrogenic preterm delivery occurrence after ICP.

METHODS

Study design and population

This cohort study included all singleton pregnancies of individuals with prevalent SLE or IBD in Sweden with delivery dates from August 1, 2007, to December 31, 2022 (Supplementary Figure S1, Supplementary Digital Content 1, http://links.lww.com/AJG/D715). We used the National Patient Register (NPR), which includes nationwide data on hospitalizations since 1987 and outpatient specialist visits since 2001, to identify patients with SLE (16). We required ≥2 separate outpatient or inpatient visits listing an International Classification of Disease (ICD) code for SLE and ≥1 of these 2 visits had to be with a specialist who typically treats SLE (rheumatology, dermatology, nephrology, internal medicine, or pediatrics). IBD was defined as ≥2 separate NPR visits or ≥1 NPR visit plus ≥1 different visit recorded in the Swedish Quality Register for IBD (SWIBREG), established in 2005 (17). SLE and IBD ICD codes are reported in Supplementary Table S1 (Supplementary Digital Content 1, http://links.lww.com/AJG/D715). These SLE and IBD definitions have shown good positive predictive values (18–20). We focused on IBD and SLE because of data availability and because they are among the most common diseases treated with thiopurines, particularly among women of reproductive age.

To identify pregnancies with prevalent SLE or IBD, we linked patients with SLE and IBD to the Swedish Medical Birth Register using the unique maternal personal identity number, requiring the 2 disease-defining visits to occur before the last menstrual period date (LMP). The Medical Birth Register contains data on antenatal, obstetric, and neonatal care for ≥98% of deliveries in Sweden from 1973. The LMP was derived as delivery date minus gestational age at delivery. This study was approved by the Swedish Ethical Review Authority (SLE: Dnr 2021-01148; IBD: Dnr 2021-06209-01, 2022-04384-02, and 2023-04868-02).

Thiopurine exposure

Data on drug exposure were obtained from the Prescribed Drug Register, which contains information on all dispensations of prescribed drugs in pharmacies in Sweden since July 2005 (21). Thiopurine exposure (yes/no) was defined as ≥1 dispensation of azathioprine (Anatomical Therapeutic Chemical code L04AX01) or 6-mercaptopurine (L01BB02) from LMP to 1 day before delivery or ICP. In sensitivity analyses, we assessed other exposure definitions: (i) ≥1 thiopurine dispensation during first trimester (LMP to LMP + 93) and (ii) ≥2 thiopurine dispensations during 3 months prepregnancy and first trimester (LMP ± 93). Unexposed pregnancies had no thiopurine dispensations during the respective periods.

Intrahepatic cholestasis of pregnancy

ICP was defined as either ≥1 maternal ICD-10 code O26.6 in the Medical Birth Register or ≥1 NPR outpatient or inpatient visit from LMP to delivery date. ICP date was set at the first ICP visit during pregnancy. We defined overall ICP as the primary outcome and early-onset ICP (before gestational week 34 + 0) as secondary (22,23). In addition, we assessed preterm delivery (delivery before gestational week 37 + 0) by overall and early-onset ICP, subtyped by onset as spontaneous or iatrogenic. We also retrieved information on ursodeoxycholic acid dispensation (only available for the IBD cohort) from ICP until 1 month postdelivery.

Covariates

We included maternal sociodemographic characteristics (age at delivery, education, income in pre-LMP year, and country of birth) along with first trimester smoking and body mass index, calendar year of delivery, and disease duration (time from SLE/IBD to LMP). Reproductive factors (parity, history of miscarriage, history of ICP) and number of healthcare visits within 2 years before LMP were also examined.

Comorbidities recorded within 5 years before LMP included hypertension, type 2 diabetes, and diseases of the liver, gallstones, and biliary tract (hereafter termed liver diseases), including primary sclerosing cholangitis and viral hepatitis. In the SLE cohort, antiphospholipid syndrome and renal diseases were additionally examined. Medication use within 1 year before LMP was evaluated, including glucocorticoids, hydroxychloroquine, and warfarin and low-molecular weight heparin as antiphospholipid syndrome proxies in SLE, and aminosalicylates, glucocorticoids, and advanced therapies in IBD. Data on advanced therapies for IBD were obtained from the Prescribed Drug Register, NPR, and SWIBREG, including tumor necrosis factor inhibitors (TNFis), anti-integrin, anti-interleukin (IL)-12 and IL-23, Janus kinase inhibitors, and Sphingosine-1-phosphate receptor modulators. Information on IBD subtypes (Crohn's disease, ulcerative colitis, and IBD unclassified) was identified based on the last healthcare visit before pregnancy. As disease activity information was not available in register data, some of the included confounders (glucocorticoids, advanced therapies, healthcare visits, and comorbidities) served as possible proxies. Supplementary Table S1 (Supplementary Digital Content 1, http://links.lww.com/AJG/D715) provides detailed definitions and assessments for all variables.

Statistical analysis

SLE and IBD pregnancy cohorts were analyzed separately. We estimated overall and early-onset ICP risks in the source pregnancy cohorts to provide Swedish background risks. From these source cohorts, propensity score-matched cohorts were derived to control for confounding, especially confounding by indication, and to estimate thiopurine effect on ICP. The propensity score was computed through logistic regression models with all covariates described above as independent variables to predict the probability of thiopurine exposure. Missing smoking, body mass index, and socioeconomic status data were treated as separate categories. We matched 2 thiopurine-unexposed pregnancies to 1 thiopurine-exposed pregnancy using nearest neighbor matching without replacement. Covariate balance was defined as standardized mean difference of <0.1. Imbalanced variables were adjusted in the final outcome models. Modified Poisson models with robust variance estimated risk ratios (RRs) and 95% confidence intervals (CIs). Risk differences were derived using linear regression.

Sensitivity and subgroup analyses

We analyzed different exposure definitions described above. In addition, we repeated the analyses excluding parous pregnancies with a history of ICP. We stratified by pregestational glucocorticoid use and liver diseases (both cohorts), by pregestational hypertension and renal diseases (SLE), and by thiopurine substance and IBD subtypes (IBD).

To further address confounding by indication in IBD pregnancies, we performed an active comparator analysis comparing those with ≥1 thiopurine dispensation with those with ≥1 recorded use of TNFi during pregnancy (infliximab, adalimumab, or golimumab). Pregnancies with dual exposure during pregnancy were excluded to isolate the effect of thiopurines. One TNFi-exposed pregnancy was propensity score-matched to 2 thiopurine-exposed pregnancies using a similar approach described above.

The ICD-10 code O26.6 can also be coded for liver and biliary tract disorders during pregnancy other than ICP, which may be accompanied by code O99.6 (Diseases of the digestive system complicating pregnancy, childbirth, and the puerperium). Therefore, we identified ICP cases that also had O99.6 during pregnancy to examine outcome misclassification. However, this was done only in the SLE cohort because O99.6 was applied frequently to pregnant women with IBD.

We observed a higher crude risk of spontaneous and iatrogenic preterm delivery in ICP vs non-ICP cases and also in thiopurine-exposed vs unexposed pregnancies. Therefore, in a post hoc analysis, we used causal mediation analysis to explore the mediating effect of ICP in the association between thiopurines and preterm delivery, stratified by onset. We hypothesized that thiopurines might increase ICP risk, subsequently increasing preterm delivery risk. However, this was performed in IBD only due to small numbers of ICP and preterm deliveries in the SLE cohort. Total effect, natural indirect effect (mediated through ICP), and direct effect were derived using the R ExactMed package. All analyses were performed in R version 14.2.

RESULTS

We included 1,200 SLE and 11,661 IBD singleton pregnancies in Sweden with delivery dates from August 1, 2007, to December 31, 2022. Of these, 297 (24.8%) and 1,924 (16.5%) were exposed to thiopurines during pregnancy and were propensity score-matched to 594 SLE and 3,848 IBD unexposed pregnancies, respectively (Supplementary Figure S1, Supplementary Digital Content 1, http://links.lww.com/AJG/D715). Among exposed pregnancies, 81% (SLE) and 91% (IBD) had ≥1 thiopurine dispensation within 6-month prepregnancy, and 89% (SLE) and 87% (IBD) had ≥1 dispensation during pregnancy after the exposure-defining dispensation.

Table 1 describes maternal characteristics of matched SLE and IBD pregnancies by thiopurine exposure. Good covariate balance was achieved in the matched IBD cohort. In SLE, several variables, including renal diseases and glucocorticoid use, were imbalanced after matching and subsequently adjusted in the outcome models.

Table 1.

Characteristics of SLE and IBD pregnancies by thiopurine exposure in the propensity score-matched cohorts, Sweden, 2007–2022

ICP characteristics

In the source pregnancy cohorts, ICP occurred in 29 of 1,200 SLE pregnancies (2.4%), including 10 (0.8%) cases of early-onset ICP, and in 229 of 11,661 IBD pregnancies (2.0%), including 56 (0.5%) cases of early-onset ICP. Median time from the index thiopurine dispensation to ICP diagnosis was 28 and 29 weeks in SLE and IBD, respectively. Among 229 IBD ICP pregnancies, 14% had a ursodeoxycholic acid dispensation between ICP and 30-day postdelivery, without a prior dispensation during pregnancy. The proportions of overall, spontaneous, and iatrogenic preterm delivery were highest among pregnancies with early-onset ICP, followed by those with later-onset ICP, and lowest among pregnancies without ICP (Supplementary Table S2, Supplementary Digital Content 1, http://links.lww.com/AJG/D715). For instance, in IBD pregnancies, spontaneous preterm delivery occurred in 19.6%, 8.1%, and 4.0% of these groups, respectively.

Associations between thiopurines and ICP

Thiopurine exposure was associated with a significantly higher risk of ICP in main, sensitivity, and subgroup analyses for both SLE and IBD pregnancies (Table 2). In SLE (main analysis), ICP risks were 14 of 297 (4.7%) for thiopurine-exposed pregnancies vs 11 of 594 (1.9%) for matched unexposed pregnancies (adjusted RR 3.06, 95% CI 1.36–6.90). In IBD, ICP developed in 140 of 1,924 (7.3%) exposed pregnancies vs 36 of 3,848 (0.9%) matched unexposed pregnancies (adjusted RR 7.78, 95% CI 5.41–11.17). The adjusted risk difference (95% CI) was 4% (1%–6%) and 6% (5%–8%) in SLE and IBD, respectively. Early-onset ICP was not analyzed in SLE due to few events (n < 10). In IBD, early-onset ICP occurred in 36 of 1,924 (1.9%) thiopurine-exposed pregnancies vs 7 of 3,848 (0.2%) matched unexposed pregnancies, resulting in an adjusted RR of 10.29 (4.58–23.08) and an adjusted risk difference of 2% (1%–2%).

Table 2.

Adjusted associations between thiopurines and intrahepatic cholestasis of pregnancy in pregnancies with systemic lupus erythematosus and inflammatory bowel disease

Stratified analyses showed similar results (Figure 1, Supplementary Tables S3 and S4, Supplementary Digital Content 1, http://links.lww.com/AJG/D715), indicating that the association was independent of ICP history, pregestational hypertension, renal diseases, liver diseases, glucocorticoid use, IBD subtypes, and thiopurine substance. Several SLE subgroup analyses were not feasible (ICP events <5).

Figure 1.

Adjusted RRs and 95% CIs in main and subgroup analyses for the association between thiopurines and intrahepatic cholestasis of pregnancy in pregnancies with systemic lupus erythematosus and inflammatory bowel disease. Thiopurine exposure: at least 1 dispensation during pregnancy. CI, confidence interval; IBD, inflammatory bowel disease; ICP, intrahepatic cholestasis of pregnancy; RR, risk ratio; SLE, systemic lupus erythematosus.

The adjusted RR (95% CI) comparing 714 thiopurine-exposed with 357 matched TNFi-exposed IBD pregnancies was 11.33 (3.58–35.88) (3 ICP events in the TNFi group) (Supplementary Table S5, Supplementary Digital Content 1, http://links.lww.com/AJG/D715). In SLE, 2 of 29 ICP cases (all in the unexposed) had the code O99.6 during pregnancy. Excluding these 2 potential ICP misclassification cases provided similar results (RR 3.89 [1.68–8.99]).

In the post hoc causal mediation analyses in IBD pregnancies, the total effect and ICP-mediated effect (RR 95% CI) of thiopurines on spontaneous preterm delivery were 1.84 (1.43–2.30) and 1.11 (1.04–1.19), respectively. The proportion mediated was 20%. For iatrogenic preterm delivery, the total effect was smaller (1.33 [0.99–1.78]) and the ICP-mediated effect was 1.37 (1.21–1.59) (Supplementary Table S6, Supplementary Digital Content 1, http://links.lww.com/AJG/D715).

DISCUSSION

In these 2 large propensity score-matched cohorts of SLE and IBD pregnancies in Sweden, thiopurine exposure during pregnancy was associated with a significantly higher risk of ICP, which remained consistent across sensitivity analyses varying timing and number of thiopurine dispensations and subgroup analyses by ICP history, IBD subtype, thiopurine substance, pregestational comorbidities, and glucocorticoid use. The adjusted risk difference was 4% for SLE and 6% for IBD. In addition, thiopurine exposure was associated with early-onset ICP. IBD and SLE pregnancies complicated by ICP, especially early-onset ICP, had higher risks of spontaneous and iatrogenic preterm delivery than those without ICP.

Extremely limited literature is available regarding thiopurine-ICP association in SLE pregnancies. In IBD, previous data are also scarce and based on small numbers of ICP cases (≤16 in all studies) with minimal or no confounding control (11–13). One cohort study reported a higher risk of ICP in 111 thiopurine-exposed vs 275 unexposed IBD pregnancies (age-adjusted odds ratio 5.34 [1.78–16.02]), similar to our results (9). Another cohort study showed a higher unadjusted risk of ICP in thiopurine-treated vs untreated IBD pregnancies (4.0% vs 0.6%, P < 0.01) (10). Our analyses, based on almost 10 times as many ICP events as previous studies combined, revealed a 3–5 times higher and a 3–9 times higher risk of ICP associated with thiopurine use during pregnancy in SLE and IBD, respectively, after controlling for multiple important confounders. The adjusted risk difference suggested that for every 100 SLE or IBD pregnancies, there are 4 or 6 more ICP cases, respectively, in the thiopurine-treated vs thiopurine-untreated groups. The differences in results between the 2 cohorts may be explained by differences in dosing strategies and thiopurine use patterns.

ICP etiology remains poorly understood and is believed to be multifactorial, involving environmental factors, genetic susceptibility, and hormonal changes during pregnancy (3). Hormonal changes are believed to play a key role in thiopurine-induced ICP. These changes can elevate thiopurine S-methyl transferase, the key enzyme in thiopurine metabolism, resulting in increased synthesis of 6-methylmercaptopurine (associated with hepatotoxicity) and decreased formation of 6-thioguanine nucleotides (the active metabolite) (24). A previous study reported 7 ICP cases among 105 thiopurine-treated IBD pregnancies, and all cases were shunting during pregnancy (shunting definition: 6-methylmercaptopurine/6-thioguanine nucleotides ratio >11) (13).

We also observed a higher risk of both spontaneous and iatrogenic preterm delivery in pregnancies with vs without ICP, especially with early-onset ICP. This is consistent with the literature (25) and aligns with management recommendations for ICP, which often include early delivery, especially for severe cases (5,26). However, the increased spontaneous preterm delivery risk warrants closer monitoring of those pregnancies with ICP. Furthermore, our post hoc mediation analysis suggested a higher risk of spontaneous and iatrogenic preterm delivery in thiopurine-exposed pregnancies vs unexposed, a finding reported in several, but not all, studies and warranting further investigation (27,28). Mediation analysis also showed that while ICP accounted for most of the increased risk of iatrogenic preterm delivery associated with thiopurines, it was less likely to be the primary mechanism underlying the association between thiopurines and spontaneous preterm delivery (proportion mediated 20%). Ursodeoxycholic acid use after ICP, without prior use, was observed in 14% of ICP cases in IBD and could be a proxy for more severe ICP. Although this treatment may alleviate maternal itch, its benefit in improving perinatal outcomes has not been demonstrated (29).

Thiopurines have long been used effectively and at low cost for IBD and SLE in general and also during pregnancy. Despite the observed higher risk of ICP linked to thiopurines, ICP's rarity makes it unlikely to outweigh the benefits of thiopurines during pregnancy. Indeed, the absolute risk difference was small. The FDA alert did not recommend thiopurine contraindication during pregnancy but only asked for labeling updates, more careful monitoring in clinical practice, and thiopurine discontinuation on ICP development. Based on our results, we recommend promoting awareness of the thiopurine-ICP association in IBD and SLE pregnancies among clinicians and patients to facilitate timely diagnosis and management. The observed association with early-onset ICP suggests that earlier monitoring of liver function tests in pregnancy may be needed for patients on thiopurine treatment. Attention should also be given to strategies that mitigate shunting, such as split dosing, in patients requiring continued thiopurine treatment during pregnancy. In addition, closer monitoring of ICP pregnancies, especially those with early onset, regarding preterm delivery should be practiced. Future studies may assess effects of different thiopurine dosing on ICP risk and effectiveness of discontinuation or tapering if ICP occurs.

Our study has several limitations. The Medical Birth Register only captures live births and stillbirths; hence, we did not include miscarriages (<gestational week 22), which may introduce selection bias if both thiopurine use and ICP increase miscarriages. However, ICP typically presents during third trimester, unlikely to affect miscarriage. We defined thiopurine exposure based on dispensation, which may not accurately capture adherence, especially in pregnant patients, because patients could fill prescriptions without taking the drug (30). We minimized this potential exposure misclassification in a sensitivity analysis by requiring ≥2 dispensations and obtained similar results. Moreover, most exposed patients had another thiopurine dispensation within 6-month prepregnancy and during pregnancy after the index dispensation, suggesting minimal nonadherence. We also lacked data on exact thiopurine dosage, dosage change during pregnancy, and shunting, which may affect the risk of ICP. Furthermore, our database may incompletely capture advanced therapies because some drugs are used intravenously in hospital settings, which is not recorded in nationwide registers. We included data from SWIBREG, which captures IBD intravenous drug administration but is not nationwide (coverage among patients treated with advanced therapy was ∼80% in 2022) (31). Therefore, when comparing thiopurines with TNFi, some thiopurine-exposed pregnancies might have also used TNFi, leading to exposure misclassification.

SLE and IBD disease activity was not adjusted for due to data unavailability. However, it remains unclear whether disease activity influences ICP risk for it to be a meaningful confounder. Nevertheless, we adjusted for possible disease activity proxies through propensity score matching, including glucocorticoid and other drug use, comorbidities, and healthcare utilization, in all analyses and also compared thiopurines with TNFi in a sensitivity analysis for IBD, minimizing confounding by indication. The analyses in SLE might be more susceptible to confounding by indication than in IBD because thiopurine indication is more channeled in SLE. Although good covariate balance was achieved in all IBD analyses, renal diseases and glucocorticoid use remained consistently imbalanced in SLE analyses, which likely reflects that azathioprine is preferred for SLE patients with renal involvement when they transition to pregnancy as an alternative to mycophenolate mofetil. We accordingly adjusted for imbalanced variables in the outcome models and also stratified by renal diseases and glucocorticoid use. In addition, ethnicity and family history of ICP are risk factors for ICP, but we did not have their data. Although we adjusted for country of birth, this does not fully account for ethnic background. Since a positive family history of ICP is likely negatively associated with thiopurine prescription and positively associated with ICP, unmeasured confounding by a family history of ICP would underestimate the association estimates. Therefore, the family history-adjusted effect estimate would even be higher than what we observed. Nevertheless, family history of ICP is expected to be rare; thus, its impact on the estimates should be small.

Furthermore, using ICD-10 code O26.6 to identify ICP could overestimate ICP risk. However, ICP is the most common condition within this group, and clinical practice in Sweden has rarely seen O26.6 coded for other conditions. We additionally assessed O99.6 and showed a minimal impact of this potential outcome misclassification. Information on ICP severity was unavailable, making it challenging to fully assess the clinical implications of our results. Future research should examine bile acid levels because levels <40 mmol/L are not associated with adverse perinatal outcomes (6). Finally, our results' generalizability to populations without universal healthcare coverage is uncertain.

Our study is the first to simultaneously assess SLE and IBD pregnancies and by far the largest to thoroughly investigate the associations between thiopurines and ICP. We also analyzed the association with early-onset ICP, which has not been investigated before. We used rigorous methods to control for a wide range of confounders and stratified by clinically important factors. These results provide robust estimates on both relative and absolute scales, which can be used in clinical communication for different patient populations. The use of nationwide comprehensive registers minimized selection bias and ensured the representativeness of our findings to the Swedish population and other similar settings. Our nationwide risk estimates of ICP and preterm delivery by ICP may also contribute to future global epidemiological research in ICP.

In conclusion, thiopurine exposure during pregnancy was associated with a higher risk of ICP in pregnancies with SLE and IBD, regardless of ICP history, IBD subtype, thiopurine substance, pregestational comorbidities, and glucocorticoid use. IBD and SLE pregnancies complicated by ICP had higher risks of spontaneous and iatrogenic preterm delivery than those without ICP.

CONFLICTS OF INTEREST

Guarantor of the article: Ngoc V. Nguyen, MMSc, and Elizabeth V. Arkema, ScD.

Specific author contributions: N.V.N. contributed to the conceptualization, methodology, data management and analysis, interpretation of results, drafting of the original manuscript, and critical review and editing. A.S., A.D., E.S., G.B., and J.F.S. contributed to the methodology, interpretation of results, and critical review and editing of the manuscript. A.M. contributed to the conceptualization, methodology, interpretation of results, and critical review and editing. O.O. contributed to data sourcing, methodology, interpretation of results, and critical review and editing. E.V.A. contributed to funding acquisition, data sourcing, supervision, conceptualization, methodology, interpretation of results, and critical review and editing of the manuscript.

Financial support: This project was supported by a grant from the Ingegerd Johansson Donation (SLS-714651). The funding source was not involved in the study's design, conduct, and reporting.

Potential competing interests: O.O. has been P.I. on projects at Karolinska Institutet financed by grants from Janssen, Pfizer, AbbVie, Takeda, Ferring, and Bristol Myers Squibb. Karolinska Institutet has also received fees for lectures and participation by O.O. on advisory boards from Janssen, Ferring, Galapagos, Bristol Myer Squibb, Takeda, and Pfizer. O.O. also reports grants from Pfizer, AbbVie, Alfasigma, and Janssen in the context of national safety monitoring programs. The other authors report no conflict of interest.

Study Highlights.

WHAT IS KNOWN

• ✓ Thiopurines can be used as maintenance therapy in pregnant women with inflammatory bowel disease (IBD) and systemic lupus erythematosus (SLE).

• ✓ Thiopurines-associated intrahepatic cholestasis of pregnancy (ICP) has been reported; however, prospective data are limited, based on small numbers of cases, and lack confounding control.

WHAT IS NEW HERE

• ✓ In 2 large nationwide cohorts of IBD and SLE pregnancies, thiopurine exposure during pregnancy was associated with a significantly increased risk of overall ICP and early-onset ICP, after controlling for important confounders.

• ✓ The association remains significant across subgroups defined by ICP history, hypertension, renal diseases, liver diseases, glucocorticoid use, IBD subtypes, and thiopurine substances.

• ✓ Varying timing and number of thiopurine dispensations during pregnancy and comparing thiopurine exposure with tumor necrosis factor inhibitor exposure during pregnancy provided consistent results.

• ✓ ICP may mediate the possible association between thiopurines and preterm delivery in IBD pregnancies, which should be examined in future studies.

Supplementary Material

acg-121-1183-s001.docx ^{(303.6KB, docx)}

ABBREVIATIONS:

CD

Crohn's disease

IBD

inflammatory bowel disease

ICD

International Classification of Diseases

ICP

intrahepatic cholestasis of pregnancy

LMP

last menstrual period date

NPR

National Patient Register

SLE

systemic lupus erythematosus

SWIBREG

Swedish Quality Register for Inflammatory Bowel Disease

UC

ulcerative colitis