Thyroid disease in pregnancy and the risk of venous thromboembolism: a Danish nationwide cohort study

Abstract

Objectives

Thyroid hormones have widespread regulatory effects, including an effect on haemostasis. It has been hypothesised that abnormal levels of thyroid hormones can influence the risk of developing thrombosis, and the association is yet to be investigated during pregnancy and the postpartum period.

Methods

A retrospective cohort study utilising nationwide health registers on diagnoses and medications was performed examining pregnancies in Denmark from 1999 to 2015 exposed to maternal hyperthyroidism or hypothyroidism. Cases of venous thromboembolism (VTE) during pregnancy and up to one year after the pregnancy had ended were identified. Multivariate logistic regression was used to evaluate the association between maternal thyroid disease and venous thromboses, reported as adjusted odds ratio (aOR) with a 95% confidence interval (CI).

Results

Among 1,149,871 pregnancies studied, 12,633 (1.1%) were exposed to maternal hyperthyroidism and 17,940 (1.6%) to maternal hypothyroidism. Overall, VTE was diagnosed more often among women with hyperthyroidism (0.41%; aOR: 1.39, 95% CI: 1.06–1.83) and hypothyroidism (0.44%; aOR: 1.43, 95% CI: 1.14–1.79) than among non-exposed women (0.28%). However, when adjusting for maternal pre-pregnancy body mass index, a higher risk of thrombosis was found among women with hyperthyroidism (aOR: 1.47, 95% CI: 1.08–2.01), whereas the association with hypothyroidism levelled out (aOR: 1.25, 95% CI: 0.96–1.62).

Conclusion

In a large, nationwide cohort, an association between maternal thyroid disease and VTE diagnosed in relation to pregnancy was found. Results point towards maternal hyperthyroidism as the predominant risk factor for thrombosis and draw attention to confounding factors in the association with hypothyroidism.

Introduction

Thyroid hormones are involved in the regulation of vital functions throughout the body (1), and thyroid dysfunction can cause symptoms from different organ systems (2). More specifically, a role of thyroid hormones in the regulation of haemostasis has been proposed (3). Haemostasis relies on a balanced activation of thrombocytes and procoagulant factors while being simultaneously regulated through natural anticoagulants (4). Disruptions of normal haemostasis can result in adverse outcomes, including thrombosis or bleeding that can be fatal (4). Previous studies have shown alterations in coagulation factors, von Willebrand factor, fibrinogen and d-dimer among patients with thyroid disease (5). Altogether, it is theorised that these changes introduce a shift in the haemostatic balance, and overall, the hypothesis is that hypothyroidism reduces the activity of the haemostatic system (6), while hyperthyroidism may increase haemostatic activity (7). Consequently, hyperthyroidism has predominantly been associated with an increased risk of thrombosis in non-pregnant adults, and increased levels of d-dimer have been reported in this patient group (5).

Venous thromboembolism (VTE) is a common cause of acute patient referral to the medical ward and often affects the lower extremities, potentially resulting in a pulmonary embolism (8). The diagnosis relies on patient symptoms, clinical findings along with biochemical measurement of d-dimer and imaging scans. Many risk factors have been described, and pregnant and postpartum women are at risk of developing venous thrombosis due to changes in hormone levels that introduce a procoagulant state (9). Notably, pregnancy and the postpartum period are characterised by physiological alterations affecting thyroid hormones and haemostasis (10), and it may be speculated how the interplay between thyroid disease and pregnancy influences the risk of VTE. Existing studies in the literature on the association between thyroid disease and the risk of thrombosis have mainly been conducted among non-pregnant adults, and evidence to substantiate an association in pregnant and postpartum women is limited to case reports or case series.

This led us to investigate the association between maternal thyroid disease and the occurrence of VTE in pregnancy and postpartum in a Danish nationwide cohort study. The systematic collection of health data in Danish nationwide registers enabled an observational study design of all pregnant women in Denmark across decades. Furthermore, a secondary aim was to evaluate the biochemical association between maternal thyroid hormones, thyroid autoantibodies and levels of d-dimer in pregnancy in a regional subcohort enabled by maternal blood samples from a regional pregnancy biobank.

Methods

Study cohorts

This was a retrospective cohort study utilising Danish nationwide registers and regional biobank samples from the North Denmark Region Pregnancy Cohort (11). The North Denmark Region Ethics Committee on Health Research Ethics approved the study (N-20150015), which was registered according to the General Data Protection Regulation in the North Denmark Region (2015-34; 2016-76).

The study included all pregnancies with an outcome of singleton birth or spontaneous abortion in Denmark from January 1, 1999 to December 31, 2015. These pregnancies formed a full nationwide cohort (Fig. 1). Singleton live births (n = 1,008,740) and stillbirths (n = 2,538), defined as the birth of a child with no sign of life in or after pregnancy week 22 (12), were identified in the Danish Medical Birth Register (MBR), which holds registrations of all births in Denmark and information on maternal and fetal parameters (13). Spontaneous abortions (n = 138,593), defined as an involuntary loss of pregnancy before pregnancy week 22 (14), were identified from the Danish National Hospital Register (DNHR), which holds information on registered diagnoses for all in- and outpatients in Danish hospitals according to the 10th International Classification of Diseases (ICD-10) from 1994 (15). Extracted diagnoses included the ICD-10 subgroups defined by O02 and O03 along with a registration of gestational age at the time of diagnosis.

Figure 1.

Flow chart illustrating the selection of the study cohorts.

Information on maternal age at the time of pregnancy, parity and origin (born in Denmark/not born in Denmark) was available for all pregnancies. Information on mode of delivery (vaginal or caesarean section), smoking in pregnancy (from 1999 onwards) and body mass index (BMI) (from 2005 onwards) was only available for pregnancies ending with birth and was extracted from the MBR. The registration of these variables was introduced in the MBR at different time points and formed the rationale for the definition of subcohorts (Fig. 1). Hence, subcohort 1 had information on maternal smoking and included pregnancies resulting in livebirths or stillbirths from 1999 to 2015, whereas subcohort 2 had information on maternal smoking and pre-pregnancy BMI from 2005 to 2015.

In addition to the register-based cohorts, a regional subcohort was used and embedded within the frame of subcohort 2 (Fig. 1). The North Denmark Region subcohort included livebirths or stillbirths from women in the North Denmark Region from 2011–2015 (Fig. 1). These women were part of the North Denmark Region Pregnancy Cohort and had a blood sample drawn as part of the Danish prenatal screening programme for chromosomal abnormalities in median pregnancy week 10 (16). The blood sample was stored in a biobank and used for later assessment of thyroid-stimulating hormone (TSH), free thyroxine (fT4), thyroid peroxidase antibodies (TPO-Ab) and thyroglobulin antibodies (Tg-Ab). Thyroid autoantibodies were evaluated according to the previously established, pregnancy week-specific cut-offs (TPO-Ab above 59 U/mL and Tg-Ab above 33 U/mL). The biochemical measurements were performed using an ADVIA Centaur XPT (Siemens Healthineers, Erlangen, Germany) (16). In the North Denmark Regional subcohort, pregnancies from women treated with antithyroid drug (ATD) or levothyroxine (LT4) were excluded.

Exposure

Pregnancies exposed to maternal thyroid disease were identified from hospital diagnoses in the DNHR in combination with redeemed prescriptions of thyroid medications in the Danish National Prescription Register (DNPR). The DNPR holds information on all drugs redeemed from Danish pharmacies since 1995, and drugs are registered according to the Anatomical Therapeutic Chemical Classification System (ATC) (17). Thus, hyperthyroidism was defined as a hospital diagnosis of hyperthyroidism (ICD-10: E05) and/or a minimum of two redeemed prescriptions of an ATD (ATC group: H03B). Similarly, hypothyroidism was defined among women without registrations of hyperthyroidism as a hospital diagnosis of hypothyroidism (ICD-10: E03) and/or a minimum of two redeemed prescriptions of LT4 (ATC group: H03A). The registered date of the first hospital diagnosis or the first redeemed prescription, whichever came first, was defined as the onset of the thyroid disease. The study included all cases of thyroid disease with an onset before, during or up to 1 year after the pregnancy had ended.

Outcome

The outcome of VTE was identified through hospital diagnoses registered in relation to in- and outpatient contacts in the DNHR from 1998 to 2016. The onset of thrombosis was defined as the date of hospital entry. Diagnoses of thrombosis included pregnancy-specific (ICD-10: O222, O223, O228, O229, O870, O871, O878, O879, O882) and non-pregnancy-specific diagnoses (ICD-10: I26, I80, I81, I82). In addition, cases of thrombosis were categorised according to the timing in or after pregnancy. Thus, we examined diagnoses of thrombosis throughout the pregnancy and the following 365 days; during early pregnancy, defined as a registration of thrombosis before pregnancy week 22; during late pregnancy, defined as a registration of thrombosis during pregnancy week 22 or after; and after pregnancy, defined as a registration of thrombosis after the pregnancy outcome date and up to 365 days after.

In the analyses, we considered related maternal treatment in pregnancy as a possible confounder in the associations investigated. Thus, information on maternal treatment with low-molecular-weight heparin (LMWH) (ATC group: B01AB) and/or acetylsalicylic acid (ASA) (ATC: B01AC06) during pregnancy was identified from redeemed prescriptions registered in the DNPR.

In subanalyses, we identified pregnancy complications that could potentially impact the associations investigated as a confounder or intermediate for VTE in late pregnancy and after pregnancy. Thus, information on gestational hypertension or pre-eclampsia (ICD-10: O10–O16), placenta praevia (ICD-10: O44) and placental abruption (ICD-10: O45) was identified from diagnoses registered in the DNHR.

In the North Denmark Region subcohort, the first evaluation of d-dimer in pregnancy was included. d-dimer measurements were performed using ACL-TOP (Werfen, United Kingdom), and results were extracted from the clinical laboratory information system (LABKA-II), which included biochemical measurements performed in hospital laboratories in the North Denmark Region from 2006 onwards.

Statistical considerations

Descriptive results were reported as numbers (n), frequencies (%), medians and interquartile range (IQR). VTE during and after pregnancy was evaluated in the full nationwide cohort, in subcohort 1 and in subcohort 2 using multivariate logistic regression and reported with crude and adjusted odds ratios (OR) along with a 95% CI. Maternal age, parity, ethnicity and pregnancy outcome were included in the adjusted model for all pregnancies along with information on maternal treatment with LMWH and/or ASA during pregnancy. In subcohort 1 and subcohort 2, the adjusted model also included maternal smoking habits in pregnancy, and in subcohort 2, the model additionally included pre-pregnancy BMI. Furthermore, the role of mode of delivery and other pregnancy complications was investigated in a subanalysis within subcohort 2 considering VTE in late pregnancy and after pregnancy.

In the North Denmark Region subcohort, the level of d-dimer was evaluated by the median according to maternal TSH groups and thyroid autoantibody positivity and compared using the Kruskal–Wallis test or Mann–Whitney U test, as appropriate. Inspection of the data did not indicate non-linearity, and multivariate linear regression was subsequently performed to evaluate the association between maternal d-dimer and TSH and between d-dimer and thyroid autoantibody positivity, respectively. The adjusted linear model included maternal age, parity, maternal smoking habits during pregnancy and pre-pregnancy BMI. In addition, the model was adjusted for pregnancy week at the time of blood sampling, thyroid autoantibody status and TSH.

Statistical analyses were performed using STATA, version 18.0 (StataCorp LLC, USA).

Results

Overall, the full nationwide cohort included 1,149,871 pregnancies resulting in live birth, stillbirth or spontaneous abortion (Table 1). The different time spans of the subcohorts were reflected in some of the maternal characteristics, e.g. smoking was less common in the most recent subcohort 2 (Table 1). Furthermore, a diagnosis of hypothyroidism was more common in subcohort 2 (Table 1).

Table 1.

Maternal characteristics in the full nationwide cohort and in the subcohorts. The subcohorts included pregnancies with an outcome of birth only, and information on maternal smoking in pregnancy and pre-pregnancy body mass index (BMI) was available in one or both subcohorts.

	Full nationwide cohort		Subcohort 1		Subcohort 2
	n	%	n	%	n	%
Pregnancies	1,149,871	-	990,003	-	661,804	-
Pregnancy outcome
Live birth	1,008,740	87.7	987,601	99.8	660,280	99.8
Stillbirth	2,538	0.2	2,402	0.2	1,524	0.2
Spontaneous abortion	138,593	12.1	-	-	-	-
Maternal age
<25 years	183,554	16.1	160,667	16.2	102,594	15.5
25–35 years	799,372	69.9	701,808	70.9	467,882	70.7
>35 years	160,010	14.0	127,528	12.9	91,328	13.8
Missing*	6,935
Parity
Nulliparous	544,431	47.35	430,913	43.5	330,376	49.9
Multiparous	605,440	52.65	559,090	56.5	331,428	50.1
Origin
Born in Denmark	974,967	84.8	837,960	84.7	558,235	84.4
Not born in Denmark	174.628	15.2	151,826	15.3	103,390	15.6
Missing*	276		217		179
Thyroid disease
Hyperthyroidism	12,633	1.1	10,762	1.1	8,038	1.2
Hypothyroidism	17,940	1.6	15,248	1.5	12,403	1.9
LMWH/ASA in pregnancy
Yes	7,420	0.7	6,781	0.7	5,780	0.9
No	1,142,451	99.3	983,222	99.3	656,024	99.1
Smoking in pregnancy
Smoking	-	-	157,861	16.0	89,697	13.6
Nonsmoking	-	-	832,142	84.0	572,107	86.4
Mode of delivery
Vaginal birth	-	-	800,306	80.9	525,915	79.5
Caesarean section	-	-	188,804	19.1	135,300	20.5
Missing*			893		589
Pre-pregnancy BMI
<18.5 kg/m2	-	-	-	-	28,727	4.3
18.5–25 kg/m2	-	-	-	-	414,621	62.7
>25 kg/m2	-	-	-	-	218,456	33.0

LMWH, low-molecular-weight heparin; ASA, acetylsalicylic acid.

^*Missing values not included.

Overall, VTE diagnosed in pregnancy or during the first year postpartum was more frequently found among women with hyperthyroidism and hypothyroidism than among non-exposed women (Fig. 2). These overall associations were found in the full nationwide cohort and in subcohort 1, which included adjustment for maternal smoking in pregnancy (Fig. 2). When analyses were restricted to women in subcohort 2, which included maternal pre-pregnancy BMI in the adjusted model, hyperthyroidism was associated with VTE, while the association with maternal hypothyroidism was not statistically significant (Fig. 2).

Figure 2.

Bar chart showing the prevalence of venous thromboembolism diagnosed in pregnancy or the following postpartum year in the full nationwide cohort, subcohort 1 and subcohort 2 when stratified by maternal thyroid disease. The illustration includes results of multivariate logistic regression with no thyroid disease as the reference group (Ref.) and reported as adjusted odds ratios (aOR) with 95% confidence interval (CI). See the text for the definition of cohorts and variables included in the adjusted model.

The overall findings were examined in stratified analyses considering VTE diagnosed in early pregnancy (Table 2), in late pregnancy or after pregnancy (Table 3). In these analyses, no association was found between maternal thyroid disease and VTE in early (Table 2) or late pregnancy (Table 3). On the other hand, maternal hyperthyroidism and hypothyroidism were associated with an increased risk of VTE diagnosed in the year following pregnancy (Table 3), but associations differed in the adjusted models. Maternal hypothyroidism was not associated with VTE when adjusting for maternal BMI in subcohort 2, and this association was robust when additionally adjusting for other pregnancy complications and mode of delivery (Table 3). Maternal hyperthyroidism was associated with VTE after pregnancy, and the risk estimates were higher compared with hypothyroidism but at the border of statistical significance in the fully adjusted models (Table 3).

Table 2.

Association between maternal thyroid disease and venous thromboembolism (VTE) in early pregnancy evaluated using logistic regression to estimate crude and adjusted odds ratio (OR) with 95% confidence interval (CI).

	All, n	VTE		Crude		Adjusted
		n	%	OR	95% CI	OR	95% CI
Full nationwide cohort*	1,149,871
No thyroid disease	1,119,298	780	0.06	Ref.	Ref.	Ref.	Ref.
Hyperthyroidism	12,633	14	0.11	1.59	0.94–2.70	1.18	0.69–2.02
Hypothyroidism	17,940	18	0.10	1.44	0.90–2.30	0.89	0.55–1.43
Subcohort 1†	990,003
No thyroid disease	963,993	672	0.07	Ref.	Ref.	Ref.	Ref.
Hyperthyroidism	10,762	11	0.10	1.47	0.81–2.66	1.06	0.58–1.95
Hypothyroidism	15,248	16	0.10	1.51	0.92–2.47	0.89	0.53–1.48
Subcohort 2‡	661,804
No thyroid disease	641,363	526	0.08	Ref.	Ref.	Ref.	Ref.
Hyperthyroidism	8,038	11	0.14	1.67	0.92–3.03	1.21	0.66–2.24
Hypothyroidism	12,403	14	0.11	1.38	0.81–2.34	0.87	0.51–1.50

Ref., reference group.

^*Adjusted model included maternal age, parity, origin, outcome year, outcome of pregnancy and treatment with low-molecular-weight heparin (LMWH) and/or acetylsalicylic acid (ASA) in the pregnancy.

^† Adjusted model included maternal age, parity, origin, outcome year, outcome at birth, treatment with LMWH and/or ASA in the pregnancy and smoking.

^‡ Adjusted model included maternal age, parity, origin, outcome year, outcome at birth, treatment with LMWH and/or ASA in the pregnancy, smoking and body mass index.

Table 3.

Association between maternal thyroid disease and venous thromboembolism (VTE) in late pregnancy and after pregnancy evaluated using logistic regression to estimate crude and adjusted odds ratio (OR) with 95% confidence interval (CI).

	All, n	Outcome in late pregnancy						Outcome after pregnancy
		VTE		Crude		Adjusted		VTE		Crude		Adjusted
		n	%	OR	95% CI	OR	95% CI	n	%	OR	95% CI	OR	95% CI
Full nationwide cohort*	1,011,278
No thyroid disease	984,689	1,297	0.13	Ref.	Ref.	Ref.	Ref.	1,043	0.11	Ref.	Ref.	Ref.	Ref.
Hyperthyroidism	11,044	15	0.14	1.03	0.62–1.72	0.86	0.52–1.45	23	0.21	1.97	1.30–2.98	1.77	1.17–2.69
Hypothyroidism	15,545	32	0.21	1.56	1.10–2.22	1.19	0.83–1.70	29	0.19	1.76	1.22–2.55	1.49	1.02–2.16
Subcohort 1†	990,003
No thyroid disease	963,993	1,253	0.13	Ref.	Ref.	Ref.	Ref.	1,011	0.10	Ref.	Ref.	Ref.	Ref.
Hyperthyroidism	10,762	15	0.14	1.07	0.64–1.78	0.88	0.53–1.48	21	0.20	1.86	1.21–2.87	1.66	1.07–2.56
Hypothyroidism	15,248	30	0.20	1.51	1.05–2.18	1.15	0.79–1.66	29	0.19	1.82	1.25–2.63	1.54	1.06–2.24
Subcohort 2‡	661,804
No thyroid disease	641,363	879	0.14	Ref.	Ref.	Ref.	Ref.	705	0.10	Ref.	Ref.	Ref.	Ref.
Hyperthyroidism	8,038	14	0.17	1.27	0.75–2.16	1.04	0.61–1.78	16	0.20	1.81	1.10–2.98	1.61	0.98–2.65
Hypothyroidism	12,403	21	0.17	1.24	0.80–1.91	0.95	0.61–1.47	24	0.19	1.76	1.17–2.65	1.39	0.92–2.09
Subcohort 2§	661,804
No thyroid disease	641,363	879	0.14	Ref.	Ref.	Ref.	Ref.	705	0.10	Ref.	Ref.	Ref.	Ref.
Hyperthyroidism	8,038	14	0.17	1.27	0.75–2.16	1.04	0.61–1.78	16	0.20	1.81	1.10–2.98	1.60	0.97–2.63
Hypothyroidism	12,403	21	0.17	1.24	0.80–1.91	0.96	0.62–1.49	24	0.19	1.76	1.17–2.65	1.35	0.89–2.04

Ref., reference group.

^*Adjusted model included maternal age, parity, origin, outcome year, outcome of pregnancy and treatment with low-molecular-weight heparin (LMWH) and/or acetylsalicylic acid (ASA) in the pregnancy.

^† Adjusted model included maternal age, parity, origin, outcome year, outcome at birth, treatment with LMWH and/or ASA in the pregnancy and smoking.

^‡ Adjusted model included maternal age, parity, origin, outcome year, outcome at birth, treatment with LMWH and/or ASA in the pregnancy, smoking and body mass index.

^§ Adjusted model included maternal age, parity, origin, outcome year, outcome at birth, treatment with LMWH and/or ASA in the pregnancy, smoking, body mass index, mode of delivery and other pregnancy complications.

In the North Denmark Region subcohort, 1,590 of the 13,424 women (11.8%) had no known thyroid disease at the time of pregnancy along with an assessment of d-dimer in pregnancy (Table 4). In comparison, 218,339 women gave birth during the same period nationally and had a median age of 30 years (IQR: 20–40 years), a median BMI of 23.1 kg/m² (IQR: 18.0–37.5 kg/m²) and 11.7% reported smoking during pregnancy. Most d-dimers were measured in the third trimester (Table 4), and overall, levels of d-dimer increased with increasing week of pregnancy (median d-dimer in the first trimester being 0.11 mg/L FEU, 0.24 mg/L FEU in the second trimester and 0.51 mg/L FEU in the third trimester).

Table 4.

Characteristics of the 1,590 pregnant women in the North Denmark Region subcohort with biochemical assessment of thyroid function, thyroid autoantibodies and d-dimer in pregnancy. Values are presented as n (%) or as median (IQR).

Characteristics	Values
Maternal characteristics
Age (years)	30.1 (26.5–33.6)
Nulliparous	966 (60.1)
Body mass index (kg/m2)	24.5 (21.6–29.1)
Smoking in pregnancy	171 (11.1)
Thyroid assessment
Pregnancy week of thyroid assessment	10 (9–11)
TSH (mIU/L)	1.09 (0.65–1.7)
<0.1 mIU/L	40 (2.5)
>5 mIU/L	17 (1.1)
TPO-Ab positive*	171 (10.8)
Tg-Ab positive†	206 (13.0)
d-dimer assessment
Pregnancy week of d-dimer assessment	38 (34–40)
Maternal d-dimer (mg/L FEU)	0.48 (0.30–0.80)
d-dimer >5 mg/L FEU	55 (3.5)

FEU, fibrinogen equivalent unit; IQR, interquartile range; TSH, thyroid-stimulating hormone; TPO-Ab, thyroid peroxidase antibodies; Tg-Ab, thyroglobulin antibodies.

^*Levels of TPO-Ab above the pregnancy- and method-specific cut-off of 59 U/mL.

^† Levels of Tg-Ab above the pregnancy- and method-specific cut-off of 33 U/mL.

Overall, levels of d-dimer did not significantly differ by maternal TSH in early pregnancy (TSH <0.1 mIU/L (n = 37): median d-dimer (IQR) 0.51 (0.34–0.72 mg/L FEU), TSH 0.1–2.5 mIU/L (n = 1,414): 0.48 (0.29–0.81 mg/L FEU), TSH 2.5–5.0 mIU/L (n = 122): 0.49 (0.33–0.66 mg/L FEU), TSH > 5.0 mIU/L (n = 17): 0.37 (0.25–0.77 mg/L FEU), P = 0.81). Similarly, when stratified by maternal thyroid autoantibody status (TPO-Ab positive (n = 165): median d-dimer (IQR) 0.43 (0.27–0.81 mg/L FEU), TPO-Ab negative (n = 1,425): 0.48 (0.46–0.50 mg/L FEU), P = 0.28, Tg-Ab positive (n = 201): 0.50 (0.41–0.53 mg/L FEU), TPO-Ab negative (n = 1,389): 0.48 (0.46–0.50 mg/L FEU), P = 0.86).

The crude findings were corroborated in multivariate linear regression in which neither maternal TSH (adjusted beta: −0.04, 95% CI: −0.18; 0.094, P = 0.54) nor TPO-Ab positivity (adjusted beta: −0.11, 95% CI: −0.057; 0.34, P = 0.63) or Tg-Ab positivity (adjusted beta: 0.13, 95% CI: −0.29; 0.55, P = 0.56) were associated with the level of d-dimer.

Discussion

Principal findings

This observational cohort study investigated the association between maternal thyroid disease and VTE during pregnancy and postpartum over a 16-year period in Denmark. Results indicate an increased risk of being diagnosed with VTE among women with thyroid disease and suggest that the association differs by the subtype of maternal thyroid dysfunction. Hyperthyroidism appeared to be associated with an increased risk of VTE in pregnancy and postpartum, whereas for hypothyroidism the association levelled out after adjustment for potential confounders. At a biochemical level, no association was found between maternal thyroid disease or thyroid autoantibody status in pregnancy and levels of d-dimer. The study extends the hypothesis from non-pregnant adults to pregnancy that primarily elevated levels of thyroid hormones are associated with an increased risk of VTE and encourages further studies with actual measurement of biochemical markers in pregnancy and postpartum to scrutinise the underlying mechanisms of the association observed.

Interpretation

Thyroid diseases are common endocrine disorders and constitute a part of the chronic disease burden in pregnant and postpartum women. VTE is a feared complication and a high-risk situation when encountered in a young woman who is pregnant or has recently given birth. Thus, there is a clinical rationale for considering the interplay between thyroid disease and VTE in and around pregnancy to inform the management of the disorders and prevent adverse outcomes. Even if the interrelationship between thyroid disease and VTE in relation to pregnancy is of clinical relevance, the topic is hitherto sparsely investigated (10). Previous observational studies have primarily focused on the association between thyroid disease and thrombosis in non-pregnant adults, and a high prevalence of thyroid dysfunction has been found in patients diagnosed with cerebral VTE, including both hyperthyroidism and hypothyroidism (18). Some studies have suggested that hypothyroidism is associated with VTE, even if this contradicts the overall hypothesis that hypothyroidism introduces a lowering of activity in the haemostatic system and thereby an increased risk of excessive bleeding (19). Although hypothyroidism has mainly been associated with a tendency for bleeding, a meta-analysis from 2023 suggested that hypothyroidism may affect haemostasis in a bidirectional manner and that especially subclinical hypothyroidism is associated with an increased risk of thrombosis (20). Similarly, we found hypothyroidism to be associated with VTE in pregnancy, but it was an important finding that the associations levelled out after adjustment for potential confounders such as maternal BMI.

It is a characteristic of autoimmune thyroid disease and VTE that a series of environmental factors are associated with the disorders, and this raises a focus on the risk of confounding. Our findings regarding hypothyroidism and the risk of VTE in pregnancy and postpartum reflect the importance of considering covariates as potential confounders when evaluating the association (21). This applies not only in pregnancy, but also in non-pregnant adults, and the proposed association specifically with subclinical, but not overt, hypothyroidism in non-pregnant adults may emphasise a focus on other determinants. A strength of our study was the available information on smoking and BMI, and even if registration of these covariates was initiated stepwise in Denmark and led us to define separate study populations, each subcohort was nationwide and large enough to perform robust analyses. Furthermore, the registration of smoking and BMI is unique to the pregnancy cohorts because these variables are registered in the Danish MBR. On the other hand, register-based studies on non-pregnant adults often lack information on these covariates, with a risk of unmeasured or residual confounding.

In addition to maternal smoking and BMI, we considered treatment with LMWH and/or ASA in pregnancy in the analyses. Thyroid disease has been associated with pre-eclampsia (22) and other pregnancy disorders resulting in antithrombotic treatment in pregnancy. Thus, we considered LMWH and ASA treatment as potential confounders in the association between thyroid disease and VTE, but our findings were robust even with this adjustment. We gathered information on other pregnancy complications and mode of delivery, which may also increase the risk of VTE. Overall, our findings were consistent after adjustment for these variables, but the association between hyperthyroidism and VTE was at the border of statistical significance. Thus, we speculate whether these variables may act as intermediates on the causal pathway, and more evidence is encouraged to extend knowledge on the underlying mechanisms and the interplay between maternal thyroid disease, VTE and other risk factors.

Turning to hyperthyroidism, existing literature focused on non-pregnant adults and consistently pointed towards a procoagulant state and an increased risk of VTE. Previously, an increasing risk of VTE in non-pregnant individuals with higher levels of fT4 has been found (23, 24) along with increasing levels of procoagulant factors of haemostasis (25). In line with our findings in the North Denmark Region subcohort, TPO-Ab was previously not found to affect the risk of developing VTE among non-pregnant adults (24). In addition, introducing LT4 to healthy individuals has been shown to introduce an elevation of procoagulant factors (26), indicating that the underlying mechanisms are the thyroid hormones per se and not the associated thyroid autoimmunity. However, studies examining the role of autoimmunity are warranted, as the current study design does not differentiate between exposure to high or low levels of thyroid hormones and autoimmunity. From a clinical perspective, hyperthyroidism has been shown to increase the risk of pulmonary embolisms (27), but the absolute risk of VTE for patients with newly diagnosed hyperthyroidism appeared to be low (28). In addition, several case studies across age groups have been published describing thrombosis of the cerebral sinuses (29), in the deep veins with subsequent development of pulmonary embolisms (30), during the puerperal period (31) and recurrent cases of thrombosis during anticoagulation therapy in individuals with hyperthyroidism (32). Most cases were reported in relation to severely elevated levels of thyroid hormones or even during ‘thyroid storm’ and associated with high morbidity (29, 30).

When considering our results in relation to the existing evidence, we were unable to define the severity of maternal thyroid disease in a register-based design. We defined exposure to hyperthyroidism and hypothyroidism among all women diagnosed with thyroid disease before, during or after the pregnancy. This definition was made because previous studies have found that a substantial number of women with known thyroid disease encounter abnormal thyroid hormone levels in early pregnancy and that women diagnosed after a pregnancy often had abnormal thyroid hormone levels during the pregnancy, reflecting undiagnosed and untreated thyroid disease (33). This could be the case in this study, as the findings overall corroborate an association between hyperthyroidism and VTE and that this increased risk could be carried into pregnancy. The clinical impact might depend on the severity of hyperthyroidism, and clinical studies are warranted to inform guidance, especially among patients who in parallel hold other risk factors for VTE.

Methodological considerations

The advantage of using nationwide health registers is the possibility of including large, unselected cohorts, and the validity and completeness of the information in Danish registers are high (15). Systematic registrations reduce the risk of selection bias, and the combined use of hospital diagnoses and redeemed prescriptions of drugs ensures that both patients managed in hospital and in general practice are included. The validity of the ICD-10 diagnoses used for identification of thyroid disease is high, especially when combined with information on redeemed prescriptions of thyroid medication (34). However, we were unable to distinguish and evaluate the clinical subtypes of maternal hyperthyroidism and hypothyroidism in the register-based design. The validity of ICD-10 diagnoses of VTE during pregnancy and postpartum is considered high (35), but some cases of VTE may go undiagnosed, especially during the last trimester of pregnancy, as symptoms can be confused with common conditions in pregnancy. However, we find it most likely that any misclassification of the outcome would be non-differential according to exposure to maternal thyroid disease. Our study conditioned on any diagnosis of VTE in the pregnancy or postpartum period studied, and our analyses did not evaluate the risk of recurrent VTE in individual women. In addition, in our study design, the outcome could be registered before the actual registration of the exposure; however, we find the risk of immortal time bias negligible among women of young age.

We a priori decided to examine the risk of thrombosis up to one year after the pregnancy had ended to ensure that all postpartum cases of thrombosis were included and to accommodate any delays from onset of symptoms to diagnosis of VTE and thyroid disease. This follow-up window postpartum captured the typical onset of VTE, which is predominant around delivery and starts to decrease in the following weeks to months (36).

Conclusion

This was the first large and nationwide cohort study to examine the occurrence of VTE during pregnancy and postpartum in women with thyroid disease. Results aligned with the hypothesis that hyperthyroidism affects haemostasis in a procoagulant direction and increases the risk of thrombosis, whereas hypothyroidism did not increase the risk significantly. The study emphasised the importance of considering determinants that are associated with thyroid disease and VTE as potential confounders in the associations. Clinical studies with biochemical assessment are warranted to extend the register-based findings and to inform patient management.

Declaration of interest

The authors have no conflicts of interest to declare.

Funding

This study was supported by the Novo Nordisk Foundation (Grant No. NNF20OC00059465).

Author contribution statement

IMNW and SLA conceptualised the study and performed data analyses. IMNW wrote the first draft of the manuscript. All authors contributed to the interpretation and critical review of the manuscript and approved the final manuscript.