Effects of Autoimmune Thyroiditis on Pregnancy Outcomes: Analysis of the Nationwide Inpatient Sample, 2016-2020.

All procedures were conducted in accordance with the ethical standards of the relevant institutional and national committees on human experimentation, as well as the Helsinki Declaration of 1964 and its subsequent amendments. This analysis also complied with the HCUP data‐use agreement. Since the dataset consists solely of de‐identified records, institutional review board approval and informed consent were not required. Furthermore, no patients or members of the public were directly involved in the conduct of this study.

Grants from Far Eastern Memorial Hospital, New Taipei City, Taiwan (FEMH‐2026‐C‐062), and the National Science and Technology Council, Taipei, Taiwan (MOST 105‐2314‐B‐418‐003‐MY3).

Hospitalization records were obtained from the Nationwide Inpatient Sample (NIS), a nationally representative database compiled by the Healthcare Cost and Utilization Project (HCUP). Covering approximately 8 million discharges annually from community hospitals across the United States (US), the NIS provides comprehensive information for population‐based research. It aggregates discharge data from approximately 1050 hospitals across 44 participating states, designed to represent about 20% of all community hospitals nationwide as defined by the American Hospital Association. The dataset contains information on patient demographics, hospitalization details (including admission type, discharge disposition, and length of stay), primary and secondary diagnoses and procedures, expected payer, and hospital‐level attributes such as size, location, teaching status, and geographic region. The NIS data were obtained through a formal request to the HCUP Central Distributor ( https://www.distributor.hcup‐us.ahrq.gov/ ), which manages data access and user licensing. The present analysis was performed in accordance with the ethical principles outlined in the Declaration of Helsinki and the HCUP data‐use agreement. As the dataset contains only de‐identified records, patient consent and institutional review board approval were not required, and no patients or members of the public were directly involved in the conduct of this study. This study included females aged ≥ 18 years who had delivery‐related discharge diagnoses or procedures confirmed by International Classification of Diseases, Tenth edition (ICD‐10) codes. All patient data were extracted from the NIS database, 2016–2020. Patients with multiple pregnancies and without NIS data on age, LOS, mortality status, or sample weight were excluded. Subjects in the final cohort were further divided into two groups on the basis of whether autoimmune thyroiditis had been diagnosed and recorded or not during the delivery hospitalization. Those without thyroiditis formed the control group. Select obstetric, perinatal, and fetal outcomes were accessed from the NIS database for all included inpatients. Obstetric outcomes analyzed were pregnancy‐induced hypertension (PIH), gestational diabetes mellitus (GDM), preeclampsia, antepartum hemorrhage, placenta previa, placenta accreta, chorioamnionitis (CAM), cesarean delivery, induction of labor, failed induction of labor, and assisted vaginal delivery. Perinatal outcomes included hospital stays ≥ 6 days, disseminated intravascular coagulation (DIC), venous thromboembolism (VTE), hemoperitoneum, and small for gestational age (SGA). Fetal outcomes included abnormal fetal heart rate or rhythm, shoulder dystocia, stillbirth, and intrauterine fetal death (IUFD), FGR, and preterm birth. Table  S1 summarizes the specific ICD codes for identifying the analyzed conditions listed above. Demographic covariates included age, race/ethnicity, household income, primary payer, smoking, obesity status, and history of polycystic ovary syndrome (PCOS), endometriosis, or GDM. Comorbidities included hypertension, pre‐existing diabetes mellitus (DM), chronic pulmonary disease, chronic kidney disease (CKD), and rheumatic disease. Whether the pregnancy was conceived through assisted reproductive technology (ART), including in vitro fertilization (IVF) or intracytoplasmic sperm injection (ICSI), was also considered and included. These conditions were also identified through relevant ICD codes documented in Table  S1 . Hospital‐related characteristics such as bed size, location/teaching status, and hospital region were also extracted from the database. Descriptive analyses were applied to summarize patient demographic and clinical characteristics. Since the NIS employs a stratified, cluster sampling design with unequal probabilities of selection, all descriptive statistics incorporated the discharge‐level sampling weights provided by HCUP to generate nationally representative estimates. Accordingly, categorical variables are presented as unweighted counts with survey‐weighted percentages, and continuous variables as survey‐weighted means with corresponding standard errors (SE). Between‐group differences in categorical data were assessed using the Rao–Scott chi‐square test, and comparisons of continuous measures were performed through survey‐specific statistical procedures that incorporated sampling weights, clustering, and stratification to yield nationally representative and reliable estimates. To minimize imbalance between exposure groups, propensity score matching (PSM) was implemented in a 1:1 ratio. Variables that demonstrated statistically significant differences between groups before matching ( p  < 0.001) were included in the propensity score model. These variables included age, race/ethnicity, household income, insurance status, obesity, PCOS, endometriosis, history of GDM, pre‐existing DM, chronic pulmonary disease, CKD, rheumatic disease, ART conception, hospital bed size, location/teaching status, and region. Matching was conducted using an iterative algorithm that selected the closest available matches sequentially until no further pairing was possible. Because this matching procedure did not retain fixed matched pair identifiers, analyses were not conducted using pair‐specific statistical methods. Propensity scores were derived through a survey‐weighted logistic regression model that fully accounted for the complex survey structure of the NIS. Covariate balance before and after matching was evaluated using standardized mean differences (SMDs), with an SMD > 0.1 indicating meaningful imbalance. Subsequent outcome analyses employed survey‐weighted unconditional multivariable logistic regression to calculate odds ratios (ORs) and 95% confidence intervals (CIs) for binary outcomes, adjusting for covariates that remained statistically significant after matching, including age in years, race, and insurance status. Statistical significance was determined using a two‐sided p ‐value threshold of < 0.05. All analyses incorporated the NIS survey design to ensure nationally valid inference and were executed using SAS software, version 9.4 (SAS Institute Inc., Cary, NC, USA).

The study population selection process is depicted in Figure  1 . A total of 3 837 735 women aged 18 years or older with delivery‐related discharge diagnoses or procedures from the NIS database from 2016 to 2020 were included. Patients with multiple pregnancies, missing data on age, LOS, mortality status, or weight values ( n  = 75 430) in the dataset were excluded. Finally, 3 762 305 patients were included in the study. After 1:1 PSM, 17 716 patients remained as the analytic sample, consisting of 8858 cases for “Patients with autoimmune thyroiditis” and 8858 cases for “Patients without autoimmune thyroiditis.” This sample represents 88 580 hospitalizations in the US (Figure  1 ). Flow diagram of study population selection. Statistical results of the study outcomes, demographics, comorbidities, and hospital‐related information of the study population are summarized in Tables  1 and 2 , Tables  S2 and S3 . The mean age of all patients was 29.1 years, 52.2% were White (using NIS race/ethnicity designations), 51.0% had insurance covered by private health insurance, 10.1% were obese, and the proportion of smokers was 11.6%. The most common comorbidity was chronic pulmonary disease (5.7%). Significant differences were observed between the two groups in age, race/ethnicity, household income, insurance status, obesity, PCOS, endometriosis, smoking, GDM history, comorbidities (except for hypertension), ART conception, and all hospital‐related information (Tables  S2 and S3 ). Characteristics of the study population by autoimmune thyroiditis status after PSM. Note: p ‐values < 0.05 are shown in bold. Continuous variables are presented as weighted mean ± SE; categorical variables are presented as unweighted counts (weighted percentage). Abbreviations: CAM, chorioamnionitis; DIC, disseminated intravascular coagulation; GDM, gestational diabetes mellitus; IUFD, intrauterine fetal death; IUGR, intrauterine growth restriction; LGA, large for gestational age; PIH, pregnancy‐induced hypertension; SGA, small for gestational age; VTE, venous thromboembolism. Statistics of pregnancy outcomes by autoimmune thyroiditis status after PSM. Note: p ‐values < 0.05 are shown in bold. Categorical variables are presented as unweighted counts (weighted percentage). Abbreviations: CAM, chorioamnionitis; DIC, disseminated intravascular coagulation; GDM, gestational diabetes mellitus; IUFD, intrauterine fetal death; IUGR, intrauterine growth restriction; LGA, large for gestational age; PIH, pregnancy‐induced hypertension; SGA, small for gestational age; VTE, venous thromboembolism. After PSM, most characteristics were well‐balanced between the case and control groups (all SMD ≤ 0.10). However, significant differences remained in certain demographic characteristics, including age (continuous), race/ethnicity, and insurance status (all p  < 0.05). In the matched cohort, outcome frequency differences were also observed between the case and control groups in several obstetric outcomes (preeclampsia, induction of labor, and failed induction of labor), one perinatal outcome (VTE), and fetal outcomes (abnormal fetal heart rate or rhythm, and FGR) (Tables  1 and 2 ). Associations between autoimmune thyroiditis and study outcomes in patients with delivery‐related discharge diagnoses/procedures are summarized in Table  3 . After adjusting for relevant variables in multivariable analysis, patients with autoimmune thyroiditis had significantly higher risk of GDM (adjusted OR [aOR] = 1.10, 95% CI: 1.01–1.20), preeclampsia (aOR = 1.22, 95% CI: 1.08–1.38), induction of labor (aOR = 1.18, 95% CI: 1.09–1.26), failed induction of labor (aOR = 1.36, 95% CI: 1.07–1.72), abnormal fetal heart rate or rhythm (aOR = 1.15, 95% CI: 1.06–1.25), and FGR (aOR = 1.33, 95% CI: 1.12–1.59) compared to those without autoimmune thyroiditis (Table  3 ). Associations between autoimmune thyroiditis and pregnancy outcomes. Note: p‐ values < 0.05 are shown in bold. Abbreviations: AOR, adjusted odds ratio; CAM, chorioamnionitis; DIC, disseminated intravascular coagulation; FGR, fetal growth restriction; GDM, gestational diabetes mellitus; IUFD, intrauterine fetal death; OR, odds ratio; VTE, venous thromboembolism. All models were adjusted for related variables of p ‐value < 0.05 after PSM, including age in years, race, and insurance status. The results of the analysis stratified by age groups ( 35 years) are depicted in Table  4 . After adjusting for relevant variables (including age, race/ethnicity, and insurance status), patients with autoimmune thyroiditis were found to have a significantly higher risk of preeclampsia, induction of labor, failed induction of labor, abnormal fetal heart rate or rhythm, and FGR compared to those without autoimmune thyroiditis in patients younger than 25 years. Similar results were observed in patients aged 25 to 35 years. However, no significant associations were found between autoimmune thyroiditis and the study outcomes in patients older than 35 years (Table  4 ). Age‐stratified associations between autoimmune thyroiditis and pregnancy outcomes. Note: p ‐values < 0.05 are shown in bold. Abbreviations: AOR, adjusted odds ratio; FGR, fetal growth restriction; GDM, gestational diabetes mellitus. All models were adjusted for related variables of p ‐value < 0.05 after PSM (except for the stratified variables), including race/ethnicity and insurance status.

Results of this large, population‐based study demonstrate that autoimmune thyroiditis is independently associated with an increased risk of adverse pregnancy outcomes as well as risks to fetal health if the mother has autoimmune thyroiditis. After adjusting for relevant demographic, clinical, and hospital‐level variables, pregnant women with autoimmune thyroiditis had higher odds of developing gestational diabetes, preeclampsia, and requiring induction of labor, with a greater likelihood of failed induction. Moreover, fetal complications such as abnormal fetal heart rate patterns and FGR were significantly more prevalent among individuals with autoimmune thyroiditis compared to those without. Stratified analysis further revealed that these associations were most pronounced in younger patients, particularly those younger than 35 years, while no significant associations were found in the subgroup of patients older than 35 years. These findings emphasize the importance of recognizing and managing autoimmune thyroiditis during pregnancy to mitigate risks for both maternal and fetal health. Results of previous studies over the years are generally complementary, reaching similar conclusions [ 13 , 16 , 17 , 18 , 19 , 20 , 21 ]. An umbrella review of autoimmune diseases and adverse pregnancy outcomes conducted in 2023 by Singh et al. [ 8 ], showed the associations between autoimmune diseases and increased risk of adverse obstetric outcomes, including (as reported more recently) gestational diabetes, preeclampsia, and FGR. Another review 1 year later (2024) showed similar results for the negative effects of pregnancy‐associated autoimmune thyroiditis on obstetric and fetal outcomes, and the authors stressed the urgency of gaining a deeper understanding of the underlying mechanisms to help improve diagnostic tools, treatment options, and preventive measures that may enhance the health and well‐being of both mothers and newborns [ 7 ]. Another large‐scale study by Karki et al. [ 16 ], perhaps most similar to the aims and methods of the present study, reported that hypothyroidism is associated with a higher risk of preeclampsia, gestational diabetes, and intrapartum hemorrhage during pregnancy and increased hospital LOS on admission for delivery, but those authors did not focus on autoimmune thyroiditis. A smaller multicenter retrospective study that did address autoimmune thyroiditis found that thyroid autoimmunity was independently associated with adverse maternal and fetal‐neonatal outcomes, with dose‐dependent effects; specifically, PIH, GDM, and neonatal lower birthweight and admission to NICU [ 13 ]. Wang et al. [ 17 ] found that thyroid dysfunction during pregnancy was associated with increased risk of preeclampsia and gestational diabetes, and these associations varied by gestational age and thyroid autoantibody status. Another fairly recent (2023) case–control study identified maternal age ≥ 30 years and positive TPO antibodies as independent risk factors for adverse outcomes, for example, preterm delivery and preeclampsia, in women with gestational hypothyroidism [ 18 ]. A novel aspect of the present study is the analysis stratified by maternal age, which revealed that the observed associations between autoimmune thyroiditis and adverse pregnancy outcomes were particularly pronounced in pregnant women younger than 35 years, while no significant associations were found in those aged 35 years or older. These age‐specific findings add a layer of complexity to the existing understanding of the effects of autoimmune thyroiditis on pregnancy. Several previous studies show similar age‐specific findings. In a multicenter, retrospective cohort study, Xu et al. [ 13 ] found that thyroid autoimmunity was independently associated with increased risks of adverse pregnancy outcomes, especially in younger women. Cai et al. [ 18 ] identified that maternal age ≥ 30 years and positive TPO antibodies were independent risk factors for adverse outcomes such as preterm delivery and preeclampsia in women with gestational hypothyroidism. One plausible explanation is that women of advanced maternal age already have a higher baseline risk of complications such as gestational diabetes, PIH, and FGR, and often receive more intensive antenatal surveillance and management for comorbidities, so the incremental impact of autoimmune thyroiditis may be attenuated in this group. By contrast, younger women with autoimmune thyroiditis may include a larger proportion of patients with unrecognized or undertreated disease, in whom thyroid autoimmunity may exert a relatively greater influence on pregnancy outcomes. From a clinical perspective, these observations suggest that thyroid autoimmunity should also be considered in the risk stratification of younger pregnant women. However, the age‐specific patterns observed in our study should be interpreted with caution and confirmed in prospective cohorts. Potential biological mechanisms, including immune‐mediated placental destruction, impaired trophoblast invasion, and endothelial dysfunction, may help to explain associations between autoimmune thyroiditis and adverse pregnancy outcomes. The presence of thyroid autoantibodies may interfere with the delicate immune balance required for successful placentation and pregnancy maintenance. In a recent study, Imbroane et al. [ 21 ] demonstrated that recurrent pregnancy loss was associated with an increased risk of a subsequent cell‐mediated autoimmune condition, occurring from 1 to 10 years after the initial pregnancy loss. In another fairly recent study in Denmark, Knøsgaard and colleagues [ 22 ] found a high frequency of adverse pregnancy outcomes among pregnancies exposed to maternal TSH above 10 m IU/L, whereas no association was found with thyroid autoantibodies; nevertheless, those authors also reported that maternal hypothyroidism and thyroid autoimmunity were associated with adverse pregnancy outcomes. Impaired trophoblastic invasion is another proposed underlying mechanism. Proper invasion of trophoblast cells into the uterine wall is crucial for establishing an adequate blood supply to the developing fetus. Thyroid autoimmunity may slow or even preclude this process, potentially leading to complications such as preeclampsia and FGR [ 22 , 23 ]. More specifically, Varberg et al. [ 23 ] examined invasive trophoblast cell development, showing that impaired trophoblast cell invasion may lead to preeclampsia or IUGR. Other investigators showed, similarly, that abnormal trophoblast differentiation contributed to placental dysfunction and complications such as preeclampsia and FGR [ 24 ]. Authors of another study suggested that endothelial dysfunction and broader maternal immune dysregulation at the maternal–fetal interface may also be potential pathways [ 25 ]. Meanwhile, Deer et al. [ 26 ] looked deeper into the role of immune cells and mediators in preeclampsia, suggesting that immune dysregulation contributes to the pathogenesis of preeclampsia through oxidative stress and endothelial dysfunction. Autoimmune processes affect blood vessel function and lead to systemic inflammation, which can negatively affect both maternal and fetal health, contributing to outcomes such as preeclampsia and abnormal fetal heart rate patterns [ 27 ]. Thyroid autoantibodies, particularly anti‐TPO, may impair endothelial function by promoting vascular inflammation and oxidative stress. Even in euthyroid individuals, elevated anti‐TPO levels have been independently associated with reduced endothelial responsiveness [ 27 ], suggesting a direct pathogenic role in vascular dysfunction at the maternal–fetal interface. It should be noted that the use of retrospective data in the present study did not allow direct confirmation of these mechanisms, and further prospective studies are needed. Findings of the present study align with and extend the results of previous studies, suggesting associations between thyroid dysfunction or autoimmune conditions and adverse obstetric outcomes. Specifically, autoimmune thyroiditis, distinct from general hypothyroidism or other autoimmune diseases, is independently associated with modest but statistically significant increases in the risk of GDM, preeclampsia, induction of labor, failed induction of labor, abnormal fetal heart rate or rhythm, and FGR. Although the magnitude of these effect estimates is small at the individual level and should be interpreted with caution, they may still be relevant for population‐level risk stratification. These findings support consideration of the need for heightened surveillance and strategies for pregnant women with thyroid autoimmunity. Such strategies may involve closer monitoring for complications such as GDM, preeclampsia, and FGR, especially in younger pregnant women with autoimmune thyroiditis. Further studies are needed to determine the optimal management strategies and whether timely treatment may help to reduce these risks. The present study has several strengths, including, first, using the large, nationally representative dataset of the NIS, which assures generalizability of the findings across diverse populations and hospital settings in the United States. Also, the use of a robust sample size allowed for sufficient statistical power to detect associations between autoimmune thyroiditis and a wide spectrum of obstetric and perinatal outcomes, including relatively uncommon complications. Second, the study incorporated a comprehensive set of covariates—including demographic, socioeconomic, clinical, and hospital‐level factors—along with rigorous statistical adjustments and PSM to minimize confounding. Third, this is one of the few population‐based studies to systematically evaluate both maternal and fetal outcomes in women with autoimmune thyroiditis, providing a more holistic view of its clinical impact. Additionally, subgroup analyses were performed stratified by maternal age. Also, while previous studies reported associations between thyroid dysfunction and risk of preeclampsia and GDM [ 17 ] and between autoimmune diseases and adverse pregnancy outcomes [ 8 ], results of the present study add novel evidence by isolating autoimmune thyroiditis as a specific contributor to these obstetric complications in a large, nationally representative cohort. Nevertheless, this study has several limitations. The diagnosis of autoimmune thyroiditis was based on administrative ICD‐10 codes, which may be subject to coding errors or misclassification bias. The NIS database does not include laboratory values such as thyroid function tests or autoantibody titers, which limits the ability to distinguish between overt, subclinical, or euthyroid autoimmune thyroiditis. Moreover, data on disease severity were not available in the NIS, precluding assessment of outcome variations by disease stage. Moreover, medication use, such as levothyroxine treatment or information on thyroid hormone replacement regimens, was not captured in the NIS database, precluding assessing how optimal disease management may have influenced outcomes. The database also does not provide longitudinal follow‐up or the timing of diagnosis, which means that long‐term maternal or neonatal outcomes cannot be evaluated beyond the index hospitalization. Additionally, while PSM improved baseline comparability between the two groups, residual confounding due to unmeasured factors cannot be completely ruled out. Several outcome prevalences, including FGR, were lower than anticipated from prior knowledge, likely reflecting underestimation associated with the claims‐based approach. Finally, the observed effect sizes were relatively small (odds ratios < 1.5), implying that while the associations achieved statistical significance, their clinical relevance may be limited. The results should thus be interpreted with caution, particularly in light of potential residual confounding and the large sample size that may have amplified statistical power.

Signed informed consent was waived due to no patients or members of the public being directly involved in the conduct of this study.

Autoimmune thyroiditis, most commonly known as Hashimoto's thyroiditis, is the most prevalent organ‐specific autoimmune disorder and the leading cause of hypothyroidism in iodine‐sufficient regions [ 1 ]. It is characterized by chronic lymphocytic infiltration of the thyroid gland and the presence of thyroid autoantibodies, such as anti‐thyroid peroxidase (anti‐TPO) and anti‐thyroglobulin antibodies [ 2 , 3 ]. While the condition frequently results in overt or subclinical hypothyroidism, thyroid autoimmunity can also exist in a euthyroid state [ 4 ]. Epidemiologic studies estimate that 5%–15% of women of childbearing age have detectable thyroid autoantibodies, and approximately 10% may develop overt hypothyroidism during their reproductive years [ 5 ]. Pregnancy is a state of complex immunological and hormonal adaptation that can unmask or exacerbate underlying thyroid dysfunction [ 6 , 7 ]. Even in the absence of overt hypothyroidism, thyroid autoimmunity has been associated with various adverse obstetric outcomes, including early pregnancy loss, recurrent miscarriage, preterm birth, gestational hypertension, preeclampsia, placental abruption, intrauterine growth restriction (IUGR), and low birth weight [ 8 , 9 ]. Several biological mechanisms may explain these associations, such as immune‐mediated disruption of placental development, impaired trophoblastic invasion, endothelial dysfunction, and broader maternal immune dysregulation at the maternal–fetal interface [ 10 , 11 , 12 ]. Several cohort studies have evaluated the impact of thyroid autoimmunity on pregnancy outcomes by measuring serum TSH, free thyroid hormones, and thyroid autoantibodies. More recently, a multicenter retrospective cohort of over 27 000 women showed that thyroid autoimmunity is independently associated with several adverse pregnancy and neonatal outcomes, with higher anti‐TPO titers conferring greater risk [ 13 ]. However, most existing studies have either focused primarily on biochemical markers in selected clinical populations or have been limited by small sample sizes, single‐center designs, or incomplete adjustment for potential confounding factors [ 13 , 14 , 15 ]. Consequently, it remains unclear how these findings translate to real‐world, nationwide delivery populations and how clinically diagnosed autoimmune thyroiditis relates to a broader spectrum of obstetric, perinatal, and fetal complications at the time of delivery. Therefore, this study aimed to evaluate the association between autoimmune thyroiditis and a range of pregnancy‐related outcomes, including maternal complications, delivery interventions, and fetal morbidities, among delivery hospitalizations in a nationally representative inpatient cohort. We hypothesized that pregnant women with autoimmune thyroiditis would be at increased risk of adverse outcomes, such as preeclampsia, Cesarean delivery, postpartum hemorrhage, and fetal growth restriction (FGR), compared with those without autoimmune thyroiditis. By focusing on a real‐world inpatient population and leveraging the large sample size of the NIS, our study is intended to complement prior biomarker‐based cohorts by providing population‐level risk estimates for clinically recognized autoimmune thyroiditis.

The authors declare no conflicts of interest.

Table S1: ICD codes used to define diagnoses and procedures. Table S2: Characteristics of the study population categorized by autoimmune thyroiditis status, before matching. Table S3: Statistics of pregnancy outcomes by autoimmune thyroiditis status before matching.