Smoking may influence on the occurrence of thyroid disease, but studies have led to inconsistent results. In Denmark, information on maternal smoking during pregnancy is registered by midwives, and we investigated the association between maternal smoking as reported during pregnancy and the subsequent maternal risk of having hyper- or hypothyroidism diagnosed.
Population-based cohort study.
Using Danish nationwide registers, we identified mothers giving birth in Denmark, 1996–2008, and studied their first pregnancy in the study period.
Information on maternal smoking during the pregnancy and maternal diagnosis of hyper- or hypothyroidism was obtained from the Danish National Hospital Register (DNHR) and prescription of thyroid medication from the Danish National Prescription Register (DNPR). Cox proportional hazards model was used to estimate hazard ratio (HR) with 95% confidence interval (95% CI) for onset of maternal hyper- or hypothyroidism after birth of the child in multivariate analyses adjusting for potential confounders.
Among mothers included (n = 450 842), altogether 89 022 (19·7%) reported that they were smokers during the first pregnancy in the study period, and 8905 (2·0%) developed hyper (n = 3389)- or hypothyroidism (n = 5516) after birth of the child. Maternal smoking was associated with a subsequent decreased risk of developing hypothyroidism (adjusted HR 0·75 (95% CI 0·70–0·81)) and an increased risk of hyperthyroidism (1·38 (1·27–1·49)).
Danish nationwide registration of maternal smoking during pregnancy adds further evidence to an association between smoking and thyroid dysfunction; smoking reduced the risk of hypothyroidism and increased the risk of hyperthyroidism.
Thyroid disorders are common in women of reproductive age and are often of autoimmune origin.[1, 2] Development of disease seems to depend on both genetic and environmental factors, and smoking is one of a number of environmental factors influencing on the occurrence of thyroid disease.[3, 4]
Previous studies on the association between smoking and autoimmune thyroid disease consistently showed an increased risk of Graves' disease and especially severe Graves' orbitopathy in smokers.[5-9] For autoimmune hypothyroidism, results have been less consistent,[5, 6, 9, 10] but a number of cross-sectional studies[11-13] have shown a lower prevalence of thyroid autoantibodies in smokers. In an animal model, the tobacco alkaloid anatabine protected against thyroid autoimmunity, and in a recent case–control study by Carlé et al., smoking cessation was associated with a transient increase in the risk of having diagnosed overt autoimmune hypothyroidism, but no significant association with current smoking was observed.
In previous studies on the association between smoking and the risk of thyroid dysfunction, the number of exposed cases was often limited. In Denmark, information on smoking has been registered in a large number of individuals, because midwives have registered maternal smoking habits during pregnancy for every birth since 1996. Using Danish nationwide registers, we aimed to investigate the association between maternal smoking as reported during pregnancy and the subsequent risk of maternal hyper- and hypothyroidism.
Methods and materials
Study design and study population
We conducted a population-based cohort study. All Danish citizens are assigned a unique ten-digit personal identification number, which enables linkage between the nationwide registers. In the Danish Civil Registration System, we identified all live-born children in Denmark between 1 January 1996 and 31 December 2008, and in the Medical Birth Registry, we identified their mothers and information on maternal age and parity at the time of child's birth as well as information on whether it was a singleton or multiple pregnancy.
Maternal hospital diagnosis of thyroid dysfunction
Information on maternal diagnosis of thyroid dysfunction was obtained from the Danish National Hospital Register (DNHR), which holds data on all admissions to any Danish hospital since 1977 and in addition all hospital outpatient visits since 1995. For every admission, the register contains date of admission and discharge, and diagnoses classified according to the 8th revision of the International Classification of Disease (ICD-8) from 1977 to 1993 and the 10th revision (ICD-10) from 1994 onwards.
We included all in- and outpatient visits (except emergency room visits) with a main or additional first-time diagnosis of thyroid dysfunction after 1 January 1977 and before 1 January 2009. Hyperthyroidism was defined as ICD-8: 242·00–242·29 and ICD-10: E05-E05.9 (excluding thyrotoxicosis factitia (E05.4), overproduction of thyroid-stimulating hormone (E05.8A) and thyrotoxic heart disease (E05.9A)). Hypothyroidism was defined as ICD-8: 243·99, 244·00–244·09 (excluding secondary hypothyroidism 244·02) and ICD-10: E03-E03.9 and E89.0 (excluding unspecified congenital goitre (E03.0A) and atrophy of the thyroid (congenital E03.1B, acquired E03.4)).
Maternal prescription of thyroid medication
Information on prescription of thyroid medication was obtained from the Danish National Prescription Register (DNPR), which holds data on all prescription drugs dispensed from Danish pharmacies since 1995. Prescription information including the patient's personal identification number, the type and amount of drug prescribed according to the Anatomical Therapeutic Chemical Classification System (ATC) and the date of sale is transferred from the pharmacies to the register. Thyroid hormones (ATC H03A) and antithyroid medication (ATC H03B) are sold solely as prescription drugs in Denmark, and we identified prescriptions of thyroid medication dispensed between 1 January 1995 and 31 December 2008.
Onset of disease and classification of maternal thyroid dysfunction
The ‘onset’ of maternal thyroid disease was defined as the day the first prescription of thyroid medication was dispensed. If a hospital diagnosis was registered before 1 January 1996, the day of first admission to hospital defined the onset of disease. Onset of disease was categorized as before, during and after the first pregnancy in the study period, and pregnancy period was estimated by subtracting gestational age at birth from the date the child was born.
Hyperthyroidism was defined by (i) a first-time hospital diagnosis of hyperthyroidism and at least one prescription of antithyroid medication or (ii) if no hospital diagnosis was registered, at least two prescriptions of antithyroid medication. Hypothyroidism was defined by (i) a first-time hospital diagnosis of hypothyroidism, at least one prescription of thyroid hormones and no prescription of antithyroid medication or (ii) if no hospital diagnosis was registered, at least two prescriptions of thyroid hormones and no prescription of antithyroid medication.
Mothers with a hospital diagnosis before 1 January 1995 and no prescription of thyroid medication registered (n = 231) were included as treatment may have ended before prescription registration was initiated. Similarly, we included mothers with the first prescription dispensed in the year 2008 (n = 431) and only one prescription registered as the time from first prescription to the end of the study period may have been too short to include the subsequent prescription(s). Mothers with inconsistent registration of thyroid dysfunction were excluded from the main analyses (Fig. 1).
Classification of maternal smoking
Information on maternal smoking was obtained from DNHR, which includes nationwide information on maternal smoking during pregnancy since 1996: DUT00 ‘no smoking’, DUT20-23 and DUT29 ‘smoking’ (≤5, 6–10, 11–20, >20 or unknown number of cigarettes per day), DUT10-11 ‘smoking cessation during the pregnancy’ and DUT99 ‘smoking status unknown’. Mothers with missing smoking registration were excluded from the study as were mothers who reported smoking cessation during the pregnancy (Fig. 1). Mothers were classified as smokers or nonsmokers according to the smoking status reported in the first pregnancy in the study period. For mothers with two or more pregnancies during the study period, we also ascertained the smoking status reported during the second pregnancy.
From Statistic Denmark, we obtained information on maternal cohabitation, income, origin and geographical residence at the time of first child's birth in the study period. For maternal cohabitation and origin, we replaced missing values by available information in the preceding or following five (origin) or three (cohabitation) years, whichever came first. Mothers with missing information on covariates were excluded from the study (Fig. 1).
We used information on maternal residence as an indicator variable for iodine intake. Iodine is essential for thyroid hormone synthesis, and in Denmark, regional differences in iodine intake exist due to different levels of iodine in drinking water: higher levels in East Denmark than in West Denmark (divided by the Great Belt).
The mothers were studied in relation to their first pregnancy during the study period leading to birth of a live-born child. In the main analyses, mothers with no diagnosis of thyroid dysfunction prior to birth of the first child were followed from birth of child to onset of thyroid dysfunction, death, emigration or 31 December 2008, whichever came first. We used Cox proportional hazards model to estimate adjusted hazard ratio (HR) with 95% confidence interval (95% CI) for the risk of hyper- and hypothyroidism in mothers who reported that they were smokers during the pregnancy in comparison with nonsmoking mothers. Analyses were stratified by maternal age at the time of child's birth, smoking intensity (≤ or >10 cigarettes/day) and geographical residence (East/West Denmark) at the time of child's birth.
To evaluate whether the association between smoking and thyroid dysfunction also applied to women who had never given birth to a child, we subsequently estimated adjusted odds ratio (OR) with 95% CI for maternal hyper- or hypothyroidism diagnosed before the pregnancy in mothers assumed to be smokers before the pregnancy (smoking reported during the pregnancy) in comparison with nonsmoking mothers. Finally, we estimated adjusted OR with 95% CI for onset of disease during the pregnancy in smoking mothers.
In sensitivity analyses, we repeated analyses according to different criteria for maternal thyroid dysfunction and included information on maternal smoking in subsequent pregnancies. Statistical analyses were performed using STATA version 11 (Stata Corp., College Station, Texas, USA), and a 5% level of significance was chosen. The study was approved by the Danish Data Protection Agency.
Maternal smoking during pregnancy
During their first pregnancy in the study period, altogether 89 022 mothers (19·7%) reported that they were smokers with the number of cigarettes per day distributed as follows: 0–5 (n = 26 623), 6–10 (n = 33 160), 11–20 (n = 23 804), >20 (n = 3120) and ‘unknown number’ (n = 2315). Smoking mothers were younger at the time of child's birth, were less often married and less often foreign born (Table 1). The frequency of maternal smoking in the study population changed over the years (Fig. 2) with a notable reduction during the study period (25·4% in 1996 to 11·8% in 2008).
Table 1. Characteristics of the mothers at birth of first child, 1996–2008, according to maternal smoking status reported during the pregnancy
Maternal smoking and the risk of hyper- and hypothyroidism
We identified 13181 (2·9%) of the mothers studied to have developed hyper- or hypothyroidism. For maternal onset of thyroid dysfunction before birth of the child (n = 4276), the median time from onset of disease to birth of the child was 3·1 years (range 0·01–30·1 years). For maternal onset of disease after birth of the child (n = 8905), the median follow-up time was 4·2 years (range 0·1–12·9 years). Table 2 focuses on the mother's first pregnancy in the study period and presents the number of mothers developing thyroid dysfunction according to the onset of disease in relation to birth of child, type of maternal thyroid dysfunction and maternal smoking status during the pregnancy. Overall, hyperthyroidism occurred more often and hypothyroidism less often among smoking mothers.
Table 2. Frequency of maternal hyper- and hypothyroidism according to maternal smoking status reported during the first pregnancy, 1996–2008
Maternal smoking status during the pregnancy registered in the Danish National Hospital Register (DNHR) at the time of child's birth.
All percentages correspond to the percentage of ‘all’ within the column (no smoking and smoking, respectively).
Result from χ2 test: smoking/nonsmoking vs thyroid dysfunction (yes/no).
No diagnosis of hyper- or hypothyroidism in DNHR and no dispensed prescription of thyroid medication before 1 January 2009.
Diagnosis of hyper- or hypothyroidism registered in DNHR before 1 January 2009 and/or dispensed prescription of thyroid medication before 1 January 2009: prescription only (n = 5731), diagnosis only (n = 231) and diagnosis and prescription (n = 7219).
Figure 3 focuses on the risk of having thyroid dysfunction diagnosed after the first pregnancy in the study period according to maternal smoking status reported during the pregnancy. In a multivariate model, smoking decreased the risk of hypothyroidism and increased the risk of hyperthyroidism. Notably, the protective role of smoking on hypothyroidism was most pronounced within 2 years after birth of the child. In the group of mothers with onset of hypothyroidism within 2 years after birth of the child (n = 1421), 22·2% had only dispensed prescriptions of thyroid medication within the period 2 years after birth of the child. The protective role of smoking was similar in this group (adjusted OR 0·79 (95% CI 0·73–0·86) and when excluding this group of mothers from analyses (0·75 (0·70–0·81)).
To evaluate whether smoking intensity during the pregnancy had an impact on the subsequent risk of maternal thyroid dysfunction within 2 years after birth of the child, analyses were stratified by number of cigarettes per day ≤10 (n = 59 783 mothers) or >10 (n = 26 924). Maternal smoking intensity during the pregnancy might be associated with a more pronounced effect, but this did not reach statistical significance (hypothyroidism: ≤10 cigarettes adjusted HR 0·64 (95% CI 0·52–0·77), >10 cigarettes 0·48 (0·35–0·66), hyperthyroidism: ≤10 cigarettes: 1·31 (1·12–1·53), >10 cigarette/day: 1·57 (1·28–1·92)).
Maternal age and iodine intake at the time of child's birth
Median maternal age at the time of child's birth was 29 years (25th–75th percentile: 26–32 years), and the protective role of smoking in relation to hypothyroidism was most pronounced in the youngest mothers (<30 years: adjusted HR 0·54 (95% CI 0·43–0·68), ≥30 years: 0·78 (0·70–0·86)). On the other hand, the risk of hyperthyroidism tended to be highest in the oldest mothers (<30 years: 1·26 (1·13–1·40), ≥30 years: 1·51 (1·35–1·69)). In this respect, it should be noted that among mothers with a first-time hospital diagnosis of hyperthyroidism, both an increased risk of Graves' disease (adjusted HR 1·44 (1·28–1·62)) and toxic nodular goitre (1·60 (1·16–2·21)) were observed.
In relation to maternal iodine intake, results did not indicate a different association between maternal smoking and the risk of hyper- or hypothyroidism according to geographical residence at the time of child's birth (hyperthyroidism West Denmark: adjusted HR 1·39 (1·26–1·55), East Denmark: 1·36 (1·20–1·53), hypothyroidism West Denmark: 0·69 (0·62–0·77), East Denmark 0·81 (0·73–0·89)).
Smoking and onset of thyroid dysfunction before and during pregnancy
Mothers smoking during the first pregnancy in the study period were assumed also to be smoking before the pregnancy. Figure 4 focuses on the risk of maternal thyroid dysfunction diagnosed before the first pregnancy in the study period according to maternal smoking status reported during the pregnancy. For maternal onset of disease before the pregnancy, the association between smoking and the risk of hyper- and hypothyroidism showed a divergent effect of smoking similar to the main analyses. The association was unaltered when restricting analyses to mothers who were giving birth to their first child in the study period and hence were nulliparous at onset of disease, but for hyperthyroidism, results were then at the borderline of statistical significance (Fig. 4).
For maternal onset of disease during the pregnancy, the number of exposed cases was relatively small (Table 2). Smoking decreased the risk of hypothyroidism (adjusted OR 0·67 (95% CI 0·46–0·98)), but did not significantly influence on the risk of hyperthyroidism (0·86 (0·60–1·22)).
Our findings did not substantially change when restricting analyses to mothers with both a diagnosis and a prescription of thyroid medication or mothers with a prescription only. Also, including mothers initially excluded from the main analyses due to inconsistent information on maternal thyroid dysfunction did not change results (data not shown).
The consistency of maternal smoking habits after the pregnancy was ascertained in the group of mothers with two or more births during the study period (n = 230 391). Information on smoking was obtained for the first and the second pregnancy in the study period, and 176 505 mothers (76·6%) reported that they were nonsmokers in both the first and the second pregnancy, whereas 29 182 (12·7%) mothers were smoking in both pregnancies. A group of mothers (n = 24 704; 10·7%) had inconsistent information on maternal smoking in the first and second pregnancy and this group was excluded from these analyses. After such exclusion, the association between maternal smoking and onset of thyroid dysfunction after birth of the first child was the same (hyperthyroidism adjusted HR 1·48 (1·30–1·68), hypothyroidism 0·60 (0·52–0·68)). Similarly, when further restricting analyses to mothers (n = 173 913) who were giving birth to their first-born child in the study period (hyperthyroidism adjusted HR 1·43 (1·23–1·66), hypothyroidism 0·60 (0·51–0·70) or looking at all mothers (n = 440 962) with a singleton pregnancy (hyperthyroidism adjusted HR 1·38 (1·28–1·50), hypothyroidism 0·75 (0·70–0·81)), associations were unaltered.
Based on Danish nationwide registration of maternal smoking during pregnancy, smoking decreased the risk of hypothyroidism and increased the risk of hyperthyroidism. Thus, our results corroborate an association between smoking and hyperthyroidism and in addition provide new evidence of an association between smoking and hypothyroidism. The protective role of smoking in relation to hypothyroidism was most pronounced within 2 years after birth of the child and in the youngest mothers.
Smoking and hyperthyroidism
Previous studies suggest an increased risk of Graves' disease and especially Graves' orbitopathy in smokers.[5-9] In Graves' disease, smoking may affect underlying immunological mechanisms and the ways smoking affects Graves' orbitopathy are not completely clear.
The association between smoking and toxic nodular goitre is less studied; however, a meta-analysis including three studies did not detect an association, but another study did reveal an increased risk in women. In our study population, the increased risk of hyperthyroidism did not differ according to a diagnosis of either Graves' disease or toxic nodular goitre; however, this distinction should be made with caution as it is based on the first-time hospital diagnosis and neither results of thyroidal scintigraphy nor measurements of TSH receptor antibodies may have been available. In addition, the number of mothers with a diagnosis of toxic nodular goitre was rather low compatible with the majority of women in reproductive age suffering from Graves' disease.
Toxic multinodular goitre is considered a late complication of nontoxic multinodular goitre, which is mainly caused by iodine deficiency.[23, 24] Thiocyanate from smoking is a competitive inhibitor of the sodium iodide transporter (NIS), which mediates iodide transport into the thyroid gland. Thus, smoking may aggravate iodine deficiency in the thyroid gland and increase the risk of multinodular goitre in smokers. We did not detect a difference in the impact of smoking on thyroid function according to maternal iodine intake defined by geographical residence. Previous studies[23, 26] have shown an interaction between smoking and iodine intake in relation to thyroid volume, but results have been inconsistent in relation to thyroid function.[26-28]
Smoking and hypothyroidism
In our study population, maternal smoking reduced the risk of hypothyroidism. Previous studies have led to inconsistent results[5, 6]; however, evidence from a number of studies suggests a negative association between smoking and the occurrence of thyroid autoantibodies[11-13] and a lower risk of TSH elevation[9, 10, 12] in smokers. Thus, smoking may protect against autoimmune hypothyroidism by reducing thyroid autoimmune activity and the occurrence of thyroid autoantibodies. Notably, in a mouse model of experimental autoimmune thyroiditis, the tobacco alkaloid anatabine ameliorated the disease. On the other hand, in a recent case–control study by Carlé et al., no significant association between current smoking and the subsequent development of overt autoimmune hypothyroidism was observed. However, differences in study design and study population exist, and although nonsignificant, the lower limit of the confidence interval was at the protective level we report in the present study.
For maternal smoking during the pregnancy, studies have led to different results regarding the smoking-induced alterations in maternal thyroid function tests. One study reported lower TSH, higher fT3 and unaltered fT4 in smokers throughout the pregnancy, whereas in a study of TPO-antibody-negative pregnant women in first trimester, TSH levels were not associated with maternal smoking and fT4 was lower in smokers.
Smoking has been causally linked to a number of autoimmune diseases, and smoking affects the immune system, but the precise mechanisms by which smoking may increase the risk of hyperthyroidism and decrease the risk of hypothyroidism are unknown. Smoking often increases the risk of autoimmune disease, but in inflammatory bowel disease, there is also a known divergent effect of smoking which has been shown to increase the risk of Crohn's disease and protect against ulcerative colitis.
Pregnancy and autoimmunity
Pregnancy induces a number of physiological changes in thyroid function and modulates the immune system. In summary, the pregnant state increases iodine requirements and suppresses the immune system with subsequent immune rebound in the postpartum period. Thus, both maternal smoking and pregnancy may influence thyroid autoimmunity. The protective role of smoking on the development of hypothyroidism was most pronounced within 2 years after birth of the child. Thus, it might be speculated that maternal smoking attenuates the aggravation of thyroid autoimmunity occurring in the postpartum period. One previous study specifically addressed the impact of maternal smoking on the risk of subsequently hypothyroidism in a cohort of pregnant women and found no association. However, the study population was smaller than that in the present study, and the authors used a broader definition of hypothyroidism including diagnoses of thyroiditis.
In relation to thyroid volume, a synergistic effect of pregnancy and smoking has been shown. The association between smoking and the subsequent risk of thyroid dysfunction after the pregnancy in our study population may not apply to nulliparous women. However, when restricting analyses to mothers giving birth to their first child, we found a similar association between maternal smoking and onset of disease before the pregnancy at a time when these women had never given birth to a child.
Strength and limitations
The strength of our study is the large study population and population-based design. We excluded mothers with no information on maternal smoking. However, the occurrence of thyroid disease did not differ in this group of mothers compared with those included in the study. In addition, mothers who reported smoking cessation during the first pregnancy in the study period were excluded as smoking habits before and after the pregnancy were considered more unpredictable in this subgroup.
The information obtained from DNHR on maternal smoking is registered by midwives at delivery. When the pregnant woman consults her general practitioner in early pregnancy, the current smoking status is registered and continually reassessed by midwives. However, registration procedures might differ between hospitals and it might happen that the smoking registration at delivery is not confirmed by the mother, but recorded from previous registration during the pregnancy. This could potentially underestimate the number of mothers with smoking cessation during the pregnancy. Furthermore, women might quit smoking prior to becoming pregnant.
We assumed that the mother had been smoking before the pregnancy if current smoking was reported during the pregnancy. However, we did not have information on the duration of maternal smoking. In a recent national survey of smoking habits in the Danish female population, the median age at regular smoking initiation was 15–16 years, suggesting that women who reported smoking during the pregnancy had been regular smokers for a considerable period of time. Concerning maternal smoking after the pregnancy, we found high agreement between the smoking status reported in the first and the second pregnancy for mothers with 2 or more births during the study period, and our results were consistent when limiting analyses to this group of mothers.
We were able to identify maternal thyroid dysfunction by both hospital diagnosis and prescription of thyroid medication. Validation of a diagnosis of thyroid disease registered in DNHR in the general Danish population showed only misclassification in 2% of the cases. The distinction between Graves' disease and toxic nodular goitre should be interpreted with caution as it was based on the first-time hospital diagnosis at a time when results of thyroid autoantibodies and scintigraphy might not yet have been available. Only part of the women identified with thyroid dysfunction had a hospital diagnosis registered, and we were not able to further evaluate the impact of smoking on specific subtypes of hyper- and hypothyroidism.
Smoking has numerous detrimental effects on human health, and maternal smoking during pregnancy may adversely affect foetal development and increase the risk of pregnancy complications. Thus, the decline over time in the number of mothers smoking during pregnancy in Denmark is favourable, even if this may lead to a small increase in the risk of hypothyroidism. Pregnant as well as nonpregnant women should always be encouraged not to start smoking or to quit smoking.
Our findings, that different autoimmune thyroid diseases are affected diversely by smoking, illuminate the complexity of thyroid autoimmunity. One or more of the many chemical compounds in tobacco may protect against autoimmune hypothyroidism, and further studies are needed to more precisely delineate these factors, which could open the possibility of developing new therapies, as already proposed based on studies in experimental animals.
In conclusion, Danish nationwide registration of maternal smoking during pregnancy corroborates an association between smoking and hyperthyroidism and adds further evidence to an association between smoking and hypothyroidism: smoking increased the risk of hyperthyroidism and decreased the risk of hypothyroidism.
Chun Sen Wu is supported by the individual postdoctoral grants from the Danish Medical Research Council (FSS: 12-32232).
Conflict of interest
The authors declare that they have no conflict of interest.