Psychiatric disease in late adolescence and young adulthood. Foetal programming by maternal hypothyroidism?

Authors


Summary

Objective

Lack of maternal thyroid hormones during foetal brain development may lead to structural abnormalities in the brain. We hypothesized that maternal hypothyroidism during the pregnancy could programme the foetus to development of psychiatric disease later in life.

Design

Danish nationwide register study.

Participants

Singletons live-born 1980–1990.

Measurements

Cox proportional hazards model was used to estimate adjusted hazard ratio (aHR) with 95% confidence interval for offspring redemption of ≥2 prescriptions of a psychiatric drug from age 15 to 31 years.

Results

Among 542 100 adolescents and young adults included, altogether 3979 (0·7%) were born to mothers with hypothyroidism registered before 1996. In crude analyses, the use of a psychiatric drug was more frequent in late adolescence and young adulthood when the mother had hypothyroidism (P < 0·001); however, several possible confounders had to be taken into account. For example, mothers with hypothyroidism often also had a psychiatric registration (38·5% vs 27·7%, P < 0·001) and the use of psychiatric drugs changed over time. After adjustment for confounders including birth year, maternal age and maternal psychiatric history, maternal hypothyroidism was associated with an increased risk of having redeemed prescriptions of anxiolytics [aHR 1·23 (1·03–1·48)] and antipsychotics [aHR 1·22 (1·03–1·44)] in late adolescence and young adulthood. For antidepressants, aHR was 1·07 (0·98–1·17).

Conclusions

The association between maternal hypothyroidism and the use of a psychiatric drug in late adolescence and young adulthood was partly confounded by maternal psychiatric history, but foetal programming by maternal hypothyroidism may be part of the mechanisms leading to the use of anxiolytics and antipsychotics.

Introduction

Psychiatric disease is the leading cause of disease in adolescents and young adults, which is mainly attributable to depression, bipolar disease and schizophrenia.[1] The prevalence of depression is low in childhood, but rises sharply after puberty, and in late adolescence, the cumulative prevalence has been reported up to 20%.[2] Anxiety disorders are also common either alone or comorbid to depression.[3] Bipolar disease, which is characterized by mania, and schizophrenia both affect about one percentage of the population.[4, 5] The symptoms often start in late adolescence or early adulthood and may continue throughout life.[5]

The pathophysiological changes in psychiatric diseases have been extensively investigated by use of neuroimaging techniques. Structural brain changes observed in schizophrenia include enlarged ventricles and diminished volume of the hippocampus and amygdala.[6] Hippocampal volume might also be reduced in depression,[7] and amygdala has a central role in anxiety disorders.[8] The aetiology of psychiatric diseases is only partly understood, but the causation may start early in life. Foetal programming is the concept that predicts an in utero exposure to increase susceptibility to diseases later in life.[9] Hypothyroidism affects 2–3% of pregnant women either as overt or subclinical disease,[10] and thyroid hormones are important regulators of foetal brain development.[11] Lack of maternal thyroid hormones in early pregnancy may lead to structural and functional changes in the foetal brain including abnormal hippocampal development.[12, 13]

We speculated whether untreated or insufficiently treated maternal hypothyroidism during the pregnancy via foetal programming could increase the risk of psychiatric disease in late adolescence and young adulthood. More specifically, we hypothesized that lack of maternal thyroid hormones in early pregnancy could lead to structural or functional changes during foetal brain development and programme the foetus to develop psychiatric disease later in life. To examine our hypothesis, we used Danish nationwide registers. We identified a cohort of adolescents and obtained information on their use of anxiolytics, antipsychotics and antidepressants from the age of 15 years. In addition, we identified the subgroup of adolescents born to mothers who had been treated for hypothyroidism before, during or in the first years after the pregnancy.

As ‘control’ exposures, we studied children born to mothers first-time diagnosed and treated for hypothyroidism several years after the child's birth and children whose father had hypothyroidism.

Methods and materials

Study design and study population

We conducted a population-based cohort study using Danish nationwide registers. All Danish citizens are assigned a unique 10-digit personal identification number, which enables accurate linkage between the nationwide registers. In the Danish Civil Registration System,[14] we identified all live-born singletons in Denmark between 1 January 1980 and 31 December 1990. All data were linked in Statistic Denmark and were available only in encrypted form. The study was approved by the Danish Data Protection Agency.

Registers

The Danish National Hospital Register (DNHR)[15] 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 holds 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 and onwards. The Danish Psychiatric Central Register (DPCR)[16] holds data on all psychiatric hospital admissions in Denmark since 1969, and in addition, all psychiatric hospital outpatient visits since 1995. The Danish National Prescription Registry (DNPR)[17] holds data on all prescription drugs dispensed from Danish pharmacies since 1995 including the type of drug prescribed according to the Anatomical Therapeutic Chemical (ATC) classification system and the date of sale.

Maternal hypothyroidism

Hypothyroidism was defined by (1) a first-time hospital diagnosis of hypothyroidism plus at least one prescription of thyroid hormones (ATC H03A) and no prescription of antithyroid medication (ATC H03B) or (2) if no hospital diagnosis was registered, at least two prescriptions of thyroid hormones and no prescription of antithyroid medication. If only one prescription was registered in the last year of the follow-up period, it was classified as hypothyroidism as subsequent prescription could have been redeemed shortly after the end of the period. Hypothyroidism in women of reproductive age is most often of autoimmune origin.[18] According to ICD-classification, 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)].

The DNPR[17] was initiated in 1995 and, thus, after the time period when the adolescents and young adults were born. However, the first year of the prescription registration will include prevalent cases, and many of the mothers with redeemed prescriptions of thyroid hormone registered in 1995 would have been in therapy for hypothyroidism already around the time of the pregnancy. Time of maternal diagnosis of hypothyroidism in relation to the pregnancy period could only be determined for the group of singletons born to mothers with a hospital diagnosis of hypothyroidism before 1995 (n = 868).

Use of psychiatric drugs

We obtained information on the use of psychiatric drugs from the DNPR.[17] Prescriptions of the psychiatric drugs antipsychotics (N05A), antidepressants (N06A), anxiolytics (N05B) and lithium (N05AN) were identified. We predefined that a psychiatric drug had been used if minimum two redeemed prescriptions were registered within the drug group. For maternal and paternal psychiatric history, we also obtained information on hospital diagnosis of a psychiatric disease to include psychiatric diseases diagnosed before the time period when registration of prescriptions was initiated. The psychiatric diagnoses were ICD-8: 295.0-301.9 and ICD-10: F20.0-F48.9 including schizophrenia and psychosis, affective disorders and anxiety disorders. Following this, maternal and paternal psychiatric history was defined as a minimum of two redeemed prescriptions of a psychiatric drug from 1995 to 2005 or a psychiatric diagnosis from 1977 to 1994.

Covariates

From the Danish Medical Birth Registry,[19] we obtained information on maternal age and parity at time of child's birth, gender and birth weight of the child, 5-min Apgar score and gestational age at birth. From Statistic Denmark, we obtained information on maternal cohabitation, income, origin and geographical residence at time of child's birth. For maternal marriage and origin, we replaced missing values by available information in the preceding or following five (origin) or three (marriage) years, whichever came first. 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.[20] Information on maternal diabetes and preeclampsia was obtained from the DNHR and the DNPR.

Statistical analyses

The follow-up period was from the age of 15 years to the time of the first registration of redeemed prescription of psychiatric drug, emigration, death or 31 December 2010, whichever came first. The Cox proportional hazards model was used to estimate hazard ratio (HR) with 95% confidence interval (95% CI) for redemption of psychiatric drug in adolescents and young adults born to mothers registered with hypothyroidism before 1996 in comparison with the reference group (no maternal or paternal hyper- or hypothyroidism). The Cox proportional hazards assumption was justified in plots of log cumulative hazards and by Schoenfeld residuals. Robust standard errors were used to adjust for dependence between maternal multiple pregnancies. Crude and adjusted analyses were performed, and the adjusted model included birth year and gender of the child, maternal psychiatric history and the following maternal covariates obtained at the time of the child's birth: age, income, cohabitation, residence, parity and origin. Parallel analyses were performed considering paternal hypothyroidism as a ‘control’ exposure and considering the association with maternal diagnosis and treatment for hypothyroidism from 1996 to 2005 (>5 years after the birth of the child). Sensitivity analyses were performed including the evaluation of possible intermediates (birth weight and gestational age at birth, 5-min Apgar score and maternal history of diabetes and preeclampsia). Statistical analyses were performed using stata version 11 (Stata Corp., College Station, TX, USA) with a 5% level of significance.

Results

Study population

Altogether 542 100 adolescents born between 1 January 1980 and 31 December 1990 were included in the study (Fig. 1), and 0·7% (group 1a and 1b) were born to mothers registered with hypothyroidism before 1996. Table 1 gives characteristics of the adolescents at the time of their birth. When the mother had hypothyroidism registered before 1996, preterm birth occurred more often, and the children were more often large-for-gestational-age (LGA).

Table 1. Characteristics of the adolescents at time of their birth. Results are depicted for the group (group 4 in Fig. 1) with no maternal or paternal hyper- or hypothyroidism diagnosed before 2006 (nonexposed) and the group (group 1a and 1b in Fig. 1) with maternal hypothyroidism diagnosed before 1996 (maternal hypothyroidism)
Singletons (n)NonexposedaMaternal hypothyroidismb P c
526 2283979
n % n %
  1. a

    No maternal or paternal hyper- or hypothyroidism registered from 1977 to 2005.

  2. b

    Maternal hypothyroidism registered before 1996.

  3. c

    Chi-square test for categorical variables, Student's t-test for continuous variables (nonexposed vs maternal hypothyroidism).

  4. d

    Singletons with missing value on 5-min Apgar score not included (n = 4769).

  5. e

    Singletons with missing value on gestational age or gestational age <20 or >45 weeks (n = 17 104) not included.

  6. f

    Singletons with missing value on birth weight or birth weight <500 or >6000 g (n = 881) not included.

  7. g

    SGA small for gestational age (birth weight ≤10th percentile for gestational week and gender), AGA appropriate for gestational age, LGA large-for-gestational-age (birth weight ≥90th percentile for gestational week and gender).

Gender
Boy270 55751·4199450·10·102
Girl255 67148·6198549·9
5-min Apgar scored
0–628470·6290·70·106
7–10518 65499·4390899·3
Gestational agee
<37 weeks21 1534·21894·90·019
≥37 weeks488 10095·8366195·1
Birth weight (g)f
Mean (SD)3439(563)3480(602)<0·001
Birth weight categoriesg
SGA52 52810·33729·8<0·001
AGA402 28379·1291776·5
LGA54 02910·652413·7
Figure 1.

Flowchart illustrating the selection of the adolescents included in the study.

Use of a psychiatric drug

Overall, 67 826 adolescents (12·5%) had ≥2 redeemed prescriptions of a psychiatric drug (anxiolytics, antipsychotics or antidepressants) registered from the age of 15 years. All adolescents (n = 542 100) were followed from the age of 15 years to a median age of 24·1 years (range 15·0–31·0 years), and during follow-up, 34 275 were censored because of death (n = 2116) or emigration (n = 32 159).

First, we looked at the associations with maternal hypothyroidism diagnosed before 1996 (Table 2, part a). In the crude analysis, a significant increased risk of having redeemed prescriptions of anxiolytics and antipsychotics from the age of 15 years was observed, and the risk of having redeemed prescriptions of antidepressants was also marginally increased.

Table 2. Prescriptions of a psychiatric drug from the age of 15 years in adolescents and young adults exposed to maternal hypothyroidism diagnosed and treated before 1996 (a) or from 1996 (b); and in adolescents and young adults born from 1980 to 1990 and exposed to paternal hypothyroidism (c). Crude and adjusted hazard ratio (HR) with 95% confidence interval (95% CI) for minimum two redeemed prescriptions of a psychiatric drug. The reference (nonexposed) is the group of adolescents and young adults with no maternal or paternal hyper- or hypothyroidism diagnosed before 2006 (group 4 in Fig. 1)
 NonexposedExposedCrude modelAdjusted modela
n % n %HR95% CIHR95% CI
  1. Significant associations in the adjusted model are marked in bold.

  2. a

    Model included birth year (1980–1981, 1982–1983, 1984–1985, 1986–1987, 1988–1989, 1990), gender of the child (boy/girl) and the following maternal variables obtained at the time of the child's birth: income (1st, 2nd, 3rd, 4th quartile), residence (East/West Denmark divided by the great Belt), age (<20, 20–24, 25–29, 30–34, 35–39, ≥40 years), parity including index pregnancy (1, 2, 3, ≥4), cohabitation (married/not married), origin (born in Denmark/not born in Denmark) and maternal (a and b) or paternal (c) psychiatric history (yes/no).

(a) Maternal hypothyroidism 1977–1995 (Group 1a and 1b in Fig. 1)
Offspring prescriptions from age 15 years
Anxiolytics10 5432·01162·91·371·14–1·65 1·23 1·03–1·48
Antipsychotics14 3312·71413·51·221·03–1·44 1·22 1·03–1·44
Antidepressants58 49211·152013·11·091·00–1·191·070·98–1·17
(b) Maternal hypothyroidism 1996–2005 (Group 2a and 2b in Fig. 1)
Offspring prescriptions from age 15 years
Anxiolytics10 5432·02112·21·060·92–1·210·960·84–1·10
Antipsychotics14 3312·72933·11·080·96–1·211·040·92–1·17
Antidepressants58 49211·1114411·91·030·98–1·101·000·95–1·06
(c) Paternal hypothyroidism 1977–2005 (Group 1b, 2b and 3 in Fig. 1)
Offspring prescriptions from age 15 years
Anxiolytics10 5432·0502·10·980·74–1·300·920·70–1·21
Antipsychotics14 3312·7652·70·940·74–1·200·930·73–1·19
Antidepressants58 49211·130512·51·080·96–1·201·070·96–1·20

Possible confounders

End of follow-up was 31 December 2010; thus, the follow-up period was different for each birth year cohort. We therefore considered the possible influence of changes in the use of psychiatric drugs over time. According to the study design, all adolescents were followed up to at least 20 years of age. Looking at redeemed prescriptions before the age of 20 years only (Fig. 2), an increasing trend in the use of antidepressants (1·4% in 1980 to 5·9% in 1990, P < 0·001) and antipsychotics (0·5% in 1980 to 1·7% in 1990, P < 0·001) by birth year was observed. Thus, adolescents born in the last years of the study period were more likely to have redeemed prescriptions of these drugs before 20 years of age, and especially, the use of antidepressants had increased. On the other hand, the use of anxiolytics had declined (highest frequency 1·0% in 1983 declining to 0·7% in 1990, P < 0·001).

Figure 2.

Changes in the use of psychiatric drugs over time in the study population. The figure illustrates the percentages of adolescents and young adults in the final study population with minimum two redeemed prescriptions of anxiolytics, antidepressants and antipsychotics registered from age 15 to 20 years in each birth year cohort.

We considered the possible association with maternal and paternal history of psychiatric disease. Figure 3 illustrates the percentage of mothers and fathers with a psychiatric registration stratified by coexisting registration of hypothyroidism. In both groups, the frequency of a psychiatric registration was higher when the parent had hypothyroidism.

Figure 3.

Associations between parental hypothyroidism and psychiatric disease. Frequency of minimum two prescriptions of a psychiatric drug (anxiolytics, antipsychotics, antidepressants or lithium) from 1995 to 2005 or a diagnosis of psychiatric disease (schizophrenia and psychosis, anxiety disorder or affective disorder) from 1977 to 1994 in mothers and fathers in the absence and the presence of a concomitant registration of hypothyroidism, P-values are results of the chi-square test: no hypothyroidism vs hypothyroidism.

Adjusted analyses

In the adjusted model (last column, Table 2, part a), we included birth year of the child and maternal age (to adjust for differences in follow-up and trend in prescriptions over time), maternal history of psychiatric disease and other maternal characteristic a priori considered possible confounders. After such adjustment, the association between maternal hypothyroidism and the use of anxiolytics in late adolescence and young adulthood attenuated, but still revealed a significant increased risk. For the association between maternal hypothyroidism and the use of antipsychotics in the offspring, the adjusted estimate was similar to the crude estimate and still revealed an increased risk. Thus, in this case, the adjustment balanced the effect of the different confounders. For antidepressants, the association with maternal hypothyroidism showed a nonsignificant trend.

Control exposures

If other factors than an in utero exposure to low maternal thyroid hormone levels were to be involved in the associations, for example genetics, we would also expect an association with maternal hypothyroidism diagnosed and treated several years after the pregnancy or with paternal hypothyroidism. We identified 9607 singletons whose mother was first-time diagnosed and treated for hypothyroidism in the period 1996–2005 and, thus, more than 5 years after the birth of the child (median 16·4 years, range 5·3–25·9 years after). No association was observed between a later diagnosis of maternal hypothyroidism and the use of a psychiatric drug in the offspring (Table 2, part b). In our study population, 2433 adolescents had a father with hypothyroidism. Also, no association was observed between paternal hypothyroidism and the use of a psychiatric drug in the offspring (Table 2, part c).

We identified 2606 adolescents and young adults with hypothyroidism registered during the study period. The frequency of having ≥2 redeemed prescriptions of a psychiatric drug during follow-up was higher in this group (25·9% vs 12·5% when no hypothyroidism, P < 0·001). Associations were similar when this group was excluded from analyses (data not shown). A small group (n = 205) had a diagnosis of congenital hypothyroidism (all registered before the age of 3 months), and these adolescents and young adults did not have a higher frequency of psychiatric registration during follow-up (14·6% vs 12·5% when no hypothyroidism, P = 0·36).

Sensitivity analyses (data not shown)

Results did not change when additionally adjusting for maternal diabetes and preeclampsia or when excluding singletons with a diagnosis of epilepsy (n = 14 239). Also, results were similar when adjusting for birth weight categories (small (SGA) vs appropriate (AGA) vs large (LGA) for gestational age), gestational age at birth (<37 vs 37–41 vs ≥42 weeks) and 5-min Apgar score (0–6 vs 7–10). The exclusion of adolescents (n = 75) whose mother had a diagnosis of drug-induced hypothyroidism did not change results. Among adolescents and young adults identified with a psychiatric registration during follow-up, 2238 (3·3%) also had a registration before turning 15 years old. The exclusion of this group and also the exclusion of singletons with a registration of a psychiatric diagnosis and/or prescription of a psychiatric drug before age 15 years only (n = 8701) did not change results.

Discussion

Main findings

In a Danish nationwide cohort study, maternal hypothyroidism was associated with an increased risk of having redeemed two or more prescriptions of anxiolytics and antipsychotics in late adolescence and young adulthood. An increased risk was observed, even after adjustment for maternal history of psychiatric disease, and no association was observed with maternal hypothyroidism diagnosed more than 5 years after the pregnancy or with paternal hypothyroidism. The findings suggest that foetal programming by maternal hypothyroidism may be part of the mechanisms leading to some types of psychiatric disease in the offspring.

Maternal hypothyroidism in pregnancy

Hypothyroidism in pregnant women can be overt or subclinical. It can be detected and treated prior to the pregnancy start or undetected and untreated in the pregnancy. Considering maternal diagnosis and treatment before pregnancy, studies have indicated a high frequency of insufficiently treated hypothyroidism when tested in a subsequent pregnancy.[21, 22] In Denmark, no systematic screening for maternal thyroid dysfunction in pregnancy is implemented. Considering mothers first-time diagnosed and treated for hypothyroidism after a pregnancy, some of these women are likely to have had low thyroid hormone levels already in the pregnancy. Thyroid hormone levels may be abnormal for a long period of time[23], and symptoms may be absent or noncharacteristic even in the presence of overt disease.[24]

In the present study, we included adolescents born from 1980 to 1990 to obtain a long follow-up period. Registration of prescription of drugs in Denmark was initiated in 1995.[17] We previously reported an increased risk of preterm birth and LGA children, when the mother had hypothyroidism diagnosed before, during or after the pregnancy,[25] and in the present study, we similarly found a higher frequency of preterm birth and LGA when the mother had a registration of hypothyroidism before 1996. Thus, even though registration of prescriptions was initiated after the decade when the adolescents were born, some of these mothers are likely to have had hypothyroidism already at the time of the pregnancy either diagnosed before birth of the child (and potentially insufficiently treated in the pregnancy) or after birth of the child (and undetected in the pregnancy).

Foetal programming by maternal hypothyroidism

Maternal thyroid hormones play a crucial role in foetal brain development.[11] In early pregnancy before the onset of foetal thyroid hormone production, the foetus is dependent on maternal thyroid hormones. Thus, lack of maternal thyroid hormones in early pregnancy may affect foetal brain development. An area in the brain extensively studied is the hippocampus, which is involved in memory and learning. Studies in rats exposed to maternal hypothyroidism in early pregnancy[26] and in children with congenital hypothyroidism[27] have shown abnormal development of the hippocampus, and a recent study in humans evaluated the impact of maternal hypothyroidism in pregnancy on the development of the hippocampus in the offspring.[12] In this study, MRI of 24 children age 9–12 years born to mothers with hypothyroidism in the pregnancy had smaller volume of hippocampus on both right and left side and lower scores in memory tests than controls.

We found an increased use of antipsychotics in adolescents and young adults born to mothers with hypothyroidism even after the adjustment for maternal history of psychiatric disease.[28] The use of antipsychotics increased during the study period in accordance with a previous Danish study in children and adolescents up to age 17 years.[29] The main indication for this type of drug is schizophrenia and psychosis. Smaller total brain volume and enlarged ventricles have been reported in patients with schizophrenia,[6] and in a recent meta-analysis of 44 MRI studies, smaller volume of the hippocampus was found bilateral both in first-episode and chronic patients.[30] The fact that the changes were present already at the onset of the disease may indicate that schizophrenia is a neurodevelopmental disorder.[30]

We hypothesized that maternal hypothyroidism via foetal programming could increase the risk of psychiatric disease in late adolescence and young adulthood. Our results and existing literature on the structural changes in the brain after in utero exposure to maternal hypothyroidism, indeed suggest that maternal hypothyroidism via structural changes in the hippocampus may predispose to later development of schizophrenia. We did not find an association with paternal hypothyroidism or with maternal hypothyroidism diagnosed long time after birth of the child. These negative controls contradict a genetic link between hypothyroidism and psychiatric disease.

We also found an increased use of anxiolytics in adolescents and young adults born to mothers with hypothyroidism even after the adjustment for maternal history of psychiatric disease. The principle brain region involved in anxiety is the amygdala with reciprocal connections to the prefrontal cortex,[8] and development of these structures may have been affected. Considering functional abnormalities in anxiety disorders, the GABA system is involved,[31] and a recent experimental study reported that the development of GABAergic pre- and postsynaptic components in the rat hippocampus may be thyroid hormone dependent.[13] However, there is a considerable overlap between the use of anxiolytics in various types of psychiatric diseases, and our data do not allow a precise identification of the type of underlying disease.

It should be stressed that the present study by nature is a generation of hypothesis. Even if a number of factors fit the hypothesis of foetal programming as discussed above, other possibilities exist. The lack of an association with congenital hypothyroidism, which is associated with some degree of foetal hypothyroid in late pregnancy, may suggest that the important factor is lack of maternal thyroid hormone in early pregnancy. On the other hand, we cannot exclude that thyroid autoimmunity per se may be involved in our findings.

Methodological considerations

This large, nationwide study was designed with a long follow-up period to ascertain the risk of psychiatric disease in late adolescence and young adulthood, which is a common onset period. We were able to include information on redeemed prescription of drugs; thus, patients treated in general practice alone were also included, but we had no information on the dose of drug prescribed. In general, the validity of the Danish prescription registration is high.[32] For the registration of hospital diagnoses, thyroid diseases revealed misclassification in <2% of the cases.[33] We cannot exclude that the inclusion of outpatients from 1995 may have influenced on the information obtained about maternal thyroid and psychiatric disease before and after this point in time.

We are aware of the risk of immortal time bias when conditioning on maternal hypothyroidism registered after birth of the child. The mother has to survive from birth of child to registration of hypothyroidism; however, we estimate this bias to be small due to the low mortality in this age group in Denmark. It should be considered that only indicators of exposure were available and some of the singletons classified as exposed would have been unexposed. This nondifferential misclassification is expected to bias the associations towards null, and associations based on data without this misclassification (e.g. based on actual measurement of maternal thyroid hormone levels in early pregnancy) could potentially reveal a much higher risk.

The use of lithium was too limited to study associations with the use of this drug (0·1%). We were not able to evaluate the risk of specific types of underlying psychiatric disease because a hospital diagnosis of psychiatric disease was less frequent than prescription of a psychiatric drug. To increase the likelihood of an underlying psychiatric disease, we defined the outcomes under study as a minimum of two redeemed prescriptions. We were able to adjust for a number of confounders, but we did not have information on maternal smoking,[34] and unmeasured or residual confounding may still exist.

Conclusion

Lack of maternal thyroid hormones during foetal brain development may programme the foetus to develop psychiatric disease later in life. The hypothesis is biologically plausible, and associations between maternal hypothyroidism and the use of both antipsychotics and anxiolytics in late adolescence and young adulthood were observed even though only indicators of exposure were available. The concept of foetal programming by maternal hypothyroidism to later development of psychiatric disease is in line with our previous finding of autism spectrum disorder in such children.[35] The findings should be further investigated in studies with actual measurement of maternal thyroid hormone status in early pregnancy.

Acknowledgement

Chun Sen Wu is supported by the individual postdoctoral grants from the Danish Medical Research Council (FSS: 12-132232).

Ancillary