Death by Suicide in Graves' Disease and Graves' Orbitopathy: A Nationwide Danish Register Study

Abstract

Background:

Graves' disease (GD) is associated with excess morbidity and mortality, but little is known about unnatural manners of death and the potential relation with Graves' orbitopathy (GO). This study investigated the risk of unnatural death in Graves' patients with and without orbitopathy compared to matched control populations.

Methods:

This was a cohort study covering all adult Danes (≥18 years) diagnosed with GD or GO during 1995–2012. Median follow-up time was 7.9 years (range 0–17.5 years). Utilizing the Danish Register of Causes of Death and the Danish National Patient Registry, 28,461 subjects with GD and 3965 with GO were identified and matched for age and sex with four subjects from the background population. The manner of death was identified, and hazard ratios (HR) for mortality due to unnatural deaths (accident, suicide, violence/homicide, and unknown) were calculated using Cox regression analyses, adjusted for pre-existing somatic and psychiatric morbidity.

Results:

In Graves' disease overall (GD + GO), there was an increased risk of death from unknown unnatural manners (HR = 2.01 [confidence interval (CI) 1.17–3.45], p = 0.012) and of suicide, although the latter difference was not with certainty statistically significant (HR = 1.43 [CI 1.00–2.04], p = 0.053). There was no significant difference in risk of death from suicide in GD subjects compared to their controls (HR = 1.27 [CI 0.85–1.89], p = 0.253). However, GO patients had a significantly higher risk of death from suicide (HR = 2.71 [CI 1.16–6.32], p = 0.022).

Conclusions:

Mortality by suicide was increased in Graves' disease overall, most significantly in patients with GO, also after adjustment for pre-existing somatic and psychiatric disease. These findings indicate that GD and GO may have a significant role in the pathophysiological mechanisms of suicidal behavior. Beyond independent confirmation, reasons for this need to be explored in order to introduce preventive measures.

Introduction

Graves' disease (GD) is associated with an increased somatic (1) as well as psychiatric comorbidity (2), with possible profound effects on the mental health. In fact, the majority of GD patients experience some psychiatric symptoms, such as anxiety, nervousness, irritability, anger, sadness, poor concentration, and sometimes even changes in personality (3,4). Quality-of-life impairments, as determined by disease-specific and generic health-related quality-of-life instruments, also persist after therapy (5). In GD patients with involvement of the eyes (Graves' orbitopathy [GO]), which is clinically the most severe manifestation of hyperthyroidism (6), studies have demonstrated a significantly reduced quality of life (7), increased psychosocial morbidity (8), and have also reported psychotic symptoms (9). Moreover, GD patients, both with and without orbitopathy (10), have an increased risk of long-term sickness absence, work disability, unemployment, loss of labor market income, and receiving a disability pension (10 –12).

In a recent study of the entire Danish population (13), increased mortality was found in patients with a history of GD whether with (23%) or without orbitopathy (19%), compared to a matched control population. The negative impact on quality of life with an increased somatic and psychiatric morbidity raises the question of whether these patients could have an increased risk of unnatural death due to suicide, accidents, or violence/homicide. With regard to suicide, this has previously been shown in other chronic diseases with a major impact on quality of life, such as diabetes mellitus (14,15). However, little is known about unnatural deaths in GD and GO patients. To the authors' knowledge, there is only one prior study on this subject, which found no significantly increased risk of suicide in hyperthyroid patients (16).

Denmark's long tradition for storing information on its citizens in nation-wide registers (17) provides a unique opportunity for large scale, population-based investigations. This study aimed to address (i) the risk of unnatural death in GD overall compared to an adequately matched control population, and (ii) the risk of unnatural death in GD patients without and with GO separately compared to their respective control populations.

Materials and Methods

Data sources

In this register-based cohort study, data were retrieved from the Danish health registers listed below. By means of a unique personal identification number, the registers were linked, and information was recovered on an individual level.

The Danish Civil Registration System (DCRS) contains information on demographics, vital status, date of death, and residence of all persons living or having lived in Denmark (18). The Danish National Patient Registry (DNPR) holds information on all diagnoses given at all hospital in- and outpatient visits in Denmark (19). Registration of diagnoses is based on the Danish edition of the International Classification of Diseases, using the 10th revision (ICD-10) from 1994 up until now. The DNPR has high validity, with misclassification of thyroid diagnoses occurring in <2% (20,21). The quality of data on psychiatric diagnoses, based on register-based information, has also been found to be acceptable (22). The Danish National Prescription Registry (DNPrR) holds information on prescription drugs dispensed from Danish pharmacies since 1995 (20). Danish pharmacies are required by law to register all prescriptions dispensed at an individual level. The registry holds, among other things, information on medical products coded according to the Anatomic Therapeutical Chemical (ATC) classification system and the date of dispensing. Because antithyroid as well as antipsychotic, antidepressant, and anxiolytic drugs are prescribed drugs, the DNPrR is a highly valid source of information for the purpose of this study (1). The Danish Register of Causes of Death (DRCD) covers all deaths among citizens dying in Denmark (23). It holds information on date, manner, and cause of death.

Study population

Admissions to hospital wards have been registered in the DNPR since 1977. However, registration of outpatient visits did not start until 1995. The DNPrR, with information on prescription drugs, was also first established in 1995. Since GD patients, as well as the majority of patients with psychiatric disorders, are often solely treated as outpatients, the start date for inclusion in this study was set to the year 1995.

Subjects with GD were defined by an ICD-10 code E050 in the DNPR. GO subjects were retrieved from the subjects with GD and defined by the ICD-10 codes H052 and/or H062. Control subjects were defined by lack of thyroid- or orbitopathy-related ICD-10 codes in the DNPR and no evidence of dispensed prescriptions of thyroid hormone or antithyroid drugs in the DNPrR. The first six months of 1995 were used as a washout period in order to ensure inclusion of only incident cases. Therefore, all cases appearing with a thyroid diagnosis in this period were excluded from the study. A total of 32,426 individuals with GD were identified, of whom 28,461 had GD without orbitopathy and 3965 had GO. Each case was randomly matched for age and sex with four control subjects.

Mortality and manners of death

Information on date and manner of death was obtained from the DRCD (23). Manners of unnatural death were categorized into accidental, suicidal, violence/homicide, and unknown.

Pre-existing morbidity

The Charlson score (CS) was used as a measure of pre-existing somatic morbidity in order to adjust for confounding. It includes 19 disease categories (vascular disease, heart failure, myocardial infarction, cerebrovascular disease, dementia, hemiplegia, chronic lung disease, rheumatic disease, liver disease, liver failure, gastric ulcer, kidney disease, diabetes mellitus without complication, diabetes mellitus with complications, cancer without metastases, cancer with metastases, lymphoma, leukemia, and acquired immune deficiency syndrome). Each is assigned a score from 1 to 6, depending on severity. For each individual, the CS is the sum of scores for all their conditions. Each increment in the CS level has been associated with a 2.3-fold [confidence interval (CI) 1.9–2.8] increase in the 10-year mortality risk in a cohort of 685 breast cancer patients (24). The CS has been validated for different morbidities, including non-malignancies (25). The CS was based primarily on information from the DNPR. However, to ensure inclusion of patients with diabetes, cardiovascular diseases, and lung diseases, many of whom are solely seen in primary care, information was also used from the DNPrR for these disease categories. Thus, users of anti-diabetics (ATC code: A10), cardiovascular-related drugs (ATC codes: B01, C01, C03, C07, C08, C09, or N02), and drugs for obstructive airway diseases (ATC code: R03) were classified as having diabetes, cardiovascular diseases, or lung diseases, respectively.

In order to adjust for pre-existing psychiatric morbidity, all individuals with a prior psychiatric diagnosis in the DNPR, or with evidence of treatment with antipsychotic (ATC: N05A), antidepressant (ATC: N06A), or anxiolytic (ATC: N05BA) medication in the DNPrR, were classified as having had a pre-existing psychiatric morbidity.

Statistical analyses

Group frequencies were compared with the Pearson chi-square test or Fisher's exact test when one or more of the cells had an expected frequency of ≤5. Group means were compared by a t-test, and group medians by the Mann–Whitney test. Risk of unnatural death in GD overall and in GD and GO separately was evaluated by Cox's proportional hazard model, and time since diagnosis was the underlying time variable. Hence, person years of follow-up were accumulated from the date of diagnosis, and terminated at the date of death, emigration, or end of follow-up (December 31, 2012), whichever came first. The validity of the proportional hazards assumption was evaluated by inspection of Schoenfeld residuals versus follow-up time (no significant associations were found). Significant differences were defined as a p-value of <0.05, using two-tailed tests. All analyses were conducted using Stata v14.1 (Stata Corp., College Station, TX).

Ethical considerations

There was no patient involvement in the design or analysis of this study. The data used in the study were anonymized so that the identity of all patients remained unknown to the investigators, and the project was approved by the Danish Data Protection Agency (project number: 704047).

Results

The baseline characteristics of GD and GO cases as well as their control subjects are summarized in Table 1. Compared to their controls, both GD and GO subjects had a significantly higher pre-existing morbidity and shorter time of follow-up (p < 0.001). There was no significant difference in CS between GD and GO individuals (p = 0.248), but the median age at time of diagnosis was higher in the GD group than it was in the GO group (Table 1). The median time of follow-up was 7.9 years (range 0–17.5 years).

Table 1.

Baseline Characteristics for Graves' Disease Individuals Without (GD) and With Graves' Orbitopathy (GO) and Their Respective Controls

	GD			GO
	Cases	Controls	p	Cases	Controls	p	GO versus GD p
Number	28,461	113,844	—	3965	15,860	—	—
Median age at diagnosis (range), years	55 (18–102)	55 (18–102)	—	52 (18–98)	52 (18–98)	—	<0.001
Number of females	23,268 (82%)	93,072 (82%)	—	2911 (73%)	11,644 (73%)	—	<0.001
Number with CS = 0	7056 (25%)	50,116 (44%)	<0.001	900 (23%)	7502 (47%)	<0.001	0.004
Number with CS = 1	11,967 (42%)	39,140 (34%)	<0.001	1773 (45%)	5301 (33%)	<0.001	0.001
Number with CS = 2	5586 (20%)	15,372 (14%)	<0.001	752 (19%)	2039 (13%)	<0.001	0.326
Number with CS >2	3852 (14%)	9216 (8%)	<0.001	540 (13%)	1018 (7%)	<0.001	0.884
Median time of follow-up (range)	7.1 (0–17.5)	8.0 (0–17.5)	<0.001	7.9 (0–17.5)	8.5 (0–17.5)	<0.001	<0.001

CS, Charlson score, a measure of comorbidity.

Risk and manners of unnatural death in GD overall (GD and GO)

There was a significantly higher overall incidence of unnatural deaths, specifically of suicide, and unknown manners of death in individuals with GD overall compared to controls (Table 2). The hazard ratios (HR) for the different manners of death (Table 3) showed an increased risk of unknown manners of death in GD overall, also after adjustment for pre-existing somatic and psychiatric morbidity (HR = 2.01 [CI 1.17–3.45], p = 0.012). There was also an increased risk of suicide, although this effect attenuated after adjustment (HR = 1.43 [CI 1.00–2.04], p = 0.053].

Table 2.

Mortality in Relation to the Manner of Unnatural Deaths in Graves' Disease Overall and in Graves' Disease Without (GD) and With Orbitopathy (GO) Compared to Their Respective Controls

		Number dead
	Manners of unnatural death	Cases	(%)	Controls	(%)	p
GD + GO	Accident	156	(0.48)	532	(0.41)	0.085
	Suicide	43	(0.13)	103	(0.08)	0.007
	Violence/homicide	0	(0)	5	(0.00)	0.590
	Unknown	21	(0.06)	37	(0.03)	0.004
	All	220	(0.68)	677	(0.52)	0.001
GD	Accident	145	(0.51)	498	(0.44)	0.114
	Suicide	33	(0.12)	90	(0.08)	0.070
	Violence/homicide	0	(0)	5	(0.00)	0.590
	Unknown	17	(0.06)	33	(0.03)	0.020
	All	195	(0.69)	626	(0.55)	0.009
GO	Accident	11	(0.28)	34	(0.21)	0.456
	Suicide	10	(0.25)	13	(0.08)	0.015
	Violence/homicide	0	(0)	0	(0)	—
	Unknown	4	(0.10)	4	(0.03)	0.056
	All	25	(0.63)	51	(0.32)	0.009

Statistically significant values are shown in bold.

Table 3.

Hazard Ratios and Confidence Intervals for Different Manners of Unnatural Death in Graves' Disease Overall and in Graves' Disease Without (GD) and With Orbitopathy (GO) Compared to Their Respective Controls

		HR [CI]
	Manners of unnatural death	Unadjusted	p	Adjusted ^a	p
GD + GO versus controls	Accident	1.26 [1.05–1.51]	0.011	0.96 [0.80–1.15]	0.672
	Suicide	1.79 [1.25–2.55]	0.001	1.43 [1.00–2.04]	0.053
	Unknown	2.39 [1.40–4.01]	0.001	2.01 [1.17–3.45]	0.012
	All	1.39 [1.20–1.62]	<0.001	1.08 [0.93–1.26]	0.322
GD versus controls	Accident	1.27 [1.06–1.53]	0.011	0.98 [0.82–1.18]	0.849
	Suicide	1.59 [1.07–2.37]	0.022	1.27 [0.85–1.89]	0.253
	Unknown	2.20 [1.22–3.95]	0.008	1.84 [1.02–3.32]	0.044
	All	1.36 [1.15–1.59]	<0.001	1.06 [0.90–1.25]	0.477
GO versus controls	Accident	1.39 [0.70–2.74]	0.345	0.94 [0.47–1.87]	0.861
	Suicide	3.30 [1.45–7.52]	0.005	2.71 [1.16–6.32]	0.022
	Unknown	4.20 [1.05–16.79]	0.042	3.79 [0.90–15.85]	0.068
	All	2.10 [1.30–3.38]	0.002	1.54 [0.95–2.52]	0.081

Although data are provided on violence/homicide in Table 2, the fact that numbers are very low (n = 5) prohibit the calculation of hazard ratios in Table 3. Statistically significant values are shown in bold.

Adjusted for both somatic morbidity (using the CS) and for psychiatric morbidity prior to the thyroid diagnosis.

HR, hazard ratio; CI, confidence interval.

Risk and manners of unnatural death in GD

GD patients had a significantly higher overall incidence of unnatural deaths, and specifically of unknown manners of death, than their controls (Table 2). The adjusted HR for the different manners of death showed a significantly increased risk of unknown manners of death but not of suicide in GD (HR = 1.84 [CI 1.02–3.32], p = 0.044, and HR = 1.27 [CI 0.85–1.89], p = 0.253, respectively; Table 3).

Risk and manners of unnatural death in GO

GO individuals had a significantly higher overall incidence of unnatural deaths, specifically of suicide, compared to controls (Table 2). There was also a significantly higher risk of suicide in GO compared to controls, both before (HR = 3.30 [CI 1.45–7.52], p = 0.005] and after adjusting for pre-existing somatic and psychiatric morbidity (HR = 2.71 [CI 1.16–6.32], p = 0.022; Table 3).

Discussion

In this large-scale, population- and register-based, long-term follow-up study of more than 30,000 Danish patients with GD, the type and magnitude of unnatural deaths were explored. The most striking finding was that of a 101% increased risk of death in an unknown manner compared to a sex- and age-matched control population. There was also a 43% increased risk of death from suicide. However, despite the huge numbers investigated, this increased risk was not with certainty statistically significant (p = 0.053) after adjustment for pre-existing somatic and psychiatric morbidity. After subdivision according to presence of GO, there was an excess mortality of 84% by unknown unnatural causes in individuals with GD without orbitopathy, and an 171% increased risk of suicide in GO patients. Excess mortality from suicide in this range or higher has also been found in other autoimmune conditions such as type 1 diabetes (26), inflammatory bowel disease (27), and rheumatoid arthritis (28). Having access to data on deaths from accidents, violence/homicide, and a group labeled “unknown,” these were included, although the groups were small and the manner of death more uncertain. As in previous register-based studies from Denmark (29), death from accidents, violence/homicide, and unknown causes were not included in the suicide category, as these cannot be classified as certain suicides. It could be speculated that these three groups, especially the unknown group, also include some deaths from suicide. Clearly, if some of the deaths registered as unknown manner of death were due to suicides, the reported risk rates for suicides in this study would be even higher.

At variance with the present study, in the only other study on this topic with a similar sized group of hyperthyroid patients who were selected on the basis of having had radioiodine or thyroid surgery, Abraham-Nordling et al. (16) did not find an increased risk of death from suicide. Reasons for this discrepancy are not obvious. Their findings concerned both GD and toxic nodular goiter but without specific data on GO. The Swedish study (16) included individuals diagnosed between 1950 and 2005 at a time when diagnostic tools were less sophisticated and therapy more ablative, that is, surgery or radioactive iodine, than the current propensity of offering antithyroid drug therapy in patients diagnosed between 1995 and 2012 (6,30). All Graves' patients were included on a national level, and validated Danish registries were employed (18,20,23), while the Swedish study mainly included individuals from one region, Stockholm (16), and in contrast to the four matched controls in this study, it took advantage of one cohort hospitalized due to non-toxic goiter, and one undergoing cholecystectomy without control for pre-existing morbidity. Thus, it is suggested that the disease burden of the control groups in the two studies may well have differed and that the studies are de facto incomparable. Neither of the studies adequately accounted for a number of pertinent variables, such as smoking behavior (31), length of hyperthyroidism before diagnosis (32), and long-term thyroid function control (33), all of which significantly impact morbidity and mortality.

Are the present findings plausible? Pointing toward a biological mechanism, the development of mood disorders and suicide has been linked to inflammation caused by infection and/or autoimmunity (34 –36). On top of that, it is currently accepted that morbidity, both before and after a diagnosis of GD, is increased (1). Moreover, although discussed until recently (32), it is beyond doubt that mortality is also increased (37), especially if the disease is poorly controlled (33). Furthermore, in addition to the psychiatric manifestations of early and untreated GD (9), the condition is also associated with a diagnosis of psychiatric disorders (anxiety disorders, depression, and psychoses) in the long term, all of which lead to excess mortality (2). Based on the aforementioned, accepting that this study does not prove causality, the findings are considered plausible. Furthermore, the strong association of GO with deaths from suicide can intuitively be understood in view of the huge impact of GO on quality of life, appearance, work disability, and early retirement (8,10,11).

Classification of a thyroid diagnosis of GD may be inaccurate. However, as mentioned previously, it has been shown that misclassification occurs in <2% (20,21). Undoubtedly, not all GO cases were classified as such. However, it has been shown that the cases with moderate to severe GO are correctly classified (38). It is generally accepted that the quality and validity of register-based data on somatic and psychiatric diagnoses are reliable (19,22). Although this study is nationwide, the data are based on rather crude register-based material, and information about life-style, smoking, and drinking habits is lacking. Thus, it is not possible to disentangle the relative importance of these confounders further. It is also important to point out that the control population only included individuals who had at least one in- or outpatient contact during the study period. Accepting that these individuals may have an increased risk of unnatural death compared to the general background population, the results may in fact underestimate the risk for unnatural death in GD.

It is concluded that mortality by suicide seems to be increased in GD, most significantly in patients with GO. In addition to independent confirmation of the data, there are ways of extending the insight provided here. Understanding the potential role of thyroid dysfunction versus thyroid autoimmunity could be advanced by incorporating another case group suffering from nodular toxic goiter. As recently done for overall mortality in hyperthyroidism (33), analyzing the consequences for mortality of the length of thyroid dysfunction before and during therapy would offer insight into a potential dose–response relationship and thereby touch upon causality. The fact that divergent conclusions have been reached, as for the association between hyperthyroidism and mortality from suicide, suggests that the type of thyroid therapy may be crucial, and this needs further study. Equally important is the understanding of the link between psychosocial consequences of GO (8,10,11) and therapy with immunosuppressant drugs, including glucocorticoids, for the excess risk of suicide.

Footnotes

Acknowledgments

Grants were received from the Danish thyroid patient organization and Musikforlæggerne Agnes and Knut Mørk's Foundation. In addition, a pre-graduate grant from Odense University Hospital to C.F.S. and an unrestricted research grant from the Novo Nordisk Foundation to L.H. supported this study. This paper was accepted as an oral communication at the Annual Meeting of the European Thyroid Association, Belgrade, Serbia, September 9–12, 2017.

Author Disclosure Statement

No competing financial interests exist.

References

Brandt

, Thvilum

, Almind

, Christensen

, Green

, Hegedüs

, Brix

. 2013. Morbidity before and after the diagnosis of hyperthyroidism: a nationwide register-based study. PLoS One, 8:e66711.

Brandt

, Thvilum

, Almind

, Christensen

, Green

, Hegedüs

, Brix

. 2014. Hyperthyroidism and psychiatric morbidity: evidence from a Danish nationwide register study. Eur J Endocrinol, 170:341–348.

Cooper

. 2003. Hyperthyroidism. Lancet, 362:459–468.

Stern

, Robinson

, Thorner

, Arruda

, Prohaska

, Prange

Jr . 1996. A survey study of neuropsychiatric complaints in patients with Graves' disease. J Neuropsychiatry Clin Neurosci, 8:181–185.

Cramon

, Winther

, Watt

, Bonnema

, Bjorner

, Ekholm

, Groenvold

, Hegedüs

, Feldt-Rasmussen

, Rasmussen ÅK 2016. Quality-of-life impairments persist six months after treatment of Graves' hyperthyroidism and toxic nodular goiter: a prospective cohort study. Thyroid, 26:1010–1018.

Smith

, Hegedüs

. 2016. Graves' disease. N Engl J Med, 375:1552–1565.

Wiersinga

. 2012. Quality of life in Graves' ophthalmopathy. Best Pract Res Clin Endocrinol Metab, 26:359–370.

Kahaly

, Petrak

, Hardt

, Pitz

, Egle

. 2005. Psychosocial morbidity of Graves' orbitopathy. Clin Endocrinol (Oxf), 63:395–402.

Hazen

, Sherry

, Parangi

, Rabito

, Sadow

. 2015. Case records of the Massachusetts General Hospital. Case 10-2015. A 15-year-old girl with Graves' disease and psychotic symptoms. N Engl J Med, 372:1250–1258.

10.

Nexo

, Watt

, Pedersen

, Bonnema

, Hegedüs

, Rasmussen

, Feldt-Rasmussen

, Bjorner

. 2014. Increased risk of long-term sickness absence, lower rate of return to work, and higher risk of unemployment and disability pensioning for thyroid patients: a Danish register-based cohort study. J Clin Endocrinol Metab, 99:3184–3192.

11.

Brandt

, Thvilum

, Hegedüs

, Brix

. 2015. Hyperthyroidism is associated with work disability and loss of labour market income. A Danish register-based study in singletons and disease-discordant twin pairs. Eur J Endocrinol, 173:595–602.

12.

Ponto

, Merkesdal

, Hommel

, Pitz

, Pfeiffer

, Kahaly

. 2013. Public health relevance of Graves' orbitopathy. J Clin Endocrinol Metab, 98:145–152.

13.

Schwensen

, Brandt

, Hegedüs

, Brix

. 2017. Mortality in Graves' orbitopathy is increased and influenced by gender, age and pre-existing morbidity: a nationwide Danish register study. Eur J Endocrinol, 176:669–676.

14.

Handley

, Ventura

, Browne

, Rich

, Attia

, Reddy

, Pouwer

, Speight

. 2016. Suicidal ideation reported by adults with type 1 or type 2 diabetes: results from Diabetes MILES-Australia. Diabet Med, 33:1582–1589.

15.

Yamauchi

, Inagaki

, Yonemoto

, Iwasaki

, Akechi

, Sawada

, Iso

, Noda

, Tsugane

, Group

. 2016. History of diabetes and risk of suicide and accidental death in Japan: The Japan Public Health Centre-based Prospective Study, 1990–2012. Diabetes Metab, 42:184–191.

16.

Abraham-Nordling

, Lönn

, Wallin

, Yin

, Nyren

, Tullgren

, Hall

, Törring

. 2009. Hyperthyroidism and suicide: a retrospective cohort study in Sweden. Eur J Endocrinol, 160:437–441.

17.

Thygesen

, Daasnes

, Thaulow

, Brønnum-Hansen

. 2011. Introduction to Danish (nationwide) registers on health and social issues: structure, access, legislation, and archiving. Scand J Public Health, 39:12–16.

18.

Pedersen

. 2011. The Danish Civil Registration System. Scand J Public Health, 39:22–25.

19.

Lynge

, Sandegaard

, Rebolj

. 2011. The Danish National Patient Register. Scand J Public Health, 39:30–33.

20.

Kildemoes

, Sørensen

, Hallas

. 2011. The Danish National Prescription Registry. Scand J Public Health, 39:38–41.

21.

Vestergaard

, Mosekilde

. 2002. Fractures in patients with hyperthyroidism and hypothyroidism: a nationwide follow-up study in 16,249 patients. Thyroid, 12:411–419.

22.

Davis

KAS

, Sudlow

CLM

, Hotopf

. 2016. Can mental health diagnoses in administrative data be used for research? A systematic review of the accuracy of routinely collected diagnoses. BMC Psychiatry, 16:263.

23.

Helweg-Larsen

. 2011. The Danish Register of Causes of Death. Scand J Public Health, 39:26–29.

24.

Charlson

, Pompei

, Ales

, MacKenzie

. 1987. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis, 40:373–383.

25.

Tang

, Wan

, Bailey

. 2008. Performance of comorbidity measures to predict stroke and death in a community-dwelling, hypertensive Medicaid population. Stroke, 39:1938–1944.

26.

Wang

, An

, Shi

, Zhang

. 2017. Management of endocrine disease: suicide risk in patients with diabetes: a systematic review and meta-analysis. Eur J Endocrinol, 177:R169–R181.

27.

Gradus

, Qin

, Lincoln

, Miller

, Lawler

, Sørensen

, Lash

. 2010. Inflammatory bowel disease and completed suicide in Danish adults. Inflamm Bowel Dis, 16:2158–2161.

28.

Ogdie

, Maliha

, Shin

, Love

, Baker

, Jiang

, Choi

, Gelfand

. 2017. Cause-specific mortality in patients with psoriatic arthritis and rheumatoid arthritis. Rheumatology (Oxford), 56:907–911.

29.

Nordentoft

, Wahlbeck

, Hällgren

, Westman

, Osby

, Alinaghizadeh

, Gissler

, Laursen

. 2013. Excess mortality, causes of death and life expectancy in 270,770 patients with recent onset of mental disorders in Denmark, Finland and Sweden. PLoS One, 8:e55176.

30.

Hegedüs

. 2009. Treatment of Graves' hyperthyroidism: evidence-based and emerging modalities. Endocrinol Metab Clin North Am, 38:355–371, ix.

31.

Hegedüs

, Brix

, Vestergaard

. 2004. Relationship between cigarette smoking and Graves' ophthalmopathy. J Endocrinol Invest, 27:265–271.

32.

Brandt

, Green

, Hegedüs

, Brix

. 2011. A critical review and meta-analysis of the association between overt hyperthyroidism and mortality. Eur J Endocrinol, 165:491–497.

33.

Lillevang-Johansen

, Abrahamsen

, Jørgensen

, Brix

, Hegedüs

. 2017. Excess mortality in treated and untreated hyperthyroidism is related to cumulative periods of low serum TSH. J Clin Endocrinol Metab, 102:2301–2309.

34.

Dantzer

, O'Connor

, Freund

, Johnson

, Kelley

. 2008. From inflammation to sickness and depression: when the immune system subjugates the brain. Nat Rev Neurosci, 9:46–56.

35.

Lund-Sørensen

, Benros

, Madsen

, Sørensen

, Eaton

, Postolache

, Nordentoft

, Erlangsen

. 2016. A nationwide cohort study of the association between hospitalization with infection and risk of death by suicide. JAMA Psychiatry, 73:912–919.

36.

Benros

, Waltoft

, Nordentoft

, Ostergaard

, Eaton

, Krogh

, Mortensen

. 2013. Autoimmune diseases and severe infections as risk factors for mood disorders: a nationwide study. JAMA Psychiatry, 70:812–820.

37.

Brandt

, Thvilum

, Almind

, Christensen

, Green

, Hegedüs

, Brix

. 2013. Graves' disease and toxic nodular goiter are both associated with increased mortality but differ with respect to the cause of death: a Danish population-based register study. Thyroid, 23:408–413.

38.

Laurberg

, Berman

, Pedersen

, Andersen

, Carlé

. 2012. Incidence and clinical presentation of moderate to severe Graves' orbitopathy in a Danish population before and after iodine fortification of salt. J Clin Endocrinol Metab, 97:2325–2332.