Theories on autoimmune aspects of schizophrenia invoke the notion of early infection by microorganisms possessing antigens that are so similar to tissue in the CNS that resulting antibodies act against the brain (1–5). Comparisons of schizophrenia patients and healthy subjects have revealed differences in immunologic parameters (6), but there have been failures to replicate. There has been repeated evidence of a genetic locus for schizophrenia in the area of the human leukocyte antigens (HLA), also with failures to replicate (7–9). Obstetric complications have been implicated in schizophrenia, and some have speculated that infection in the mother produces antibodies that are transmitted to the fetus, producing autoantibodies that disrupt neural development and raise risk for schizophrenia (10, 11).
Schizophrenia patients or their relatives have been reported to have either higher or lower than expected prevalences of some autoimmune disorders, including rheumatoid arthritis (12), type 1 diabetes (13), thyroid disorders (13, 14), and celiac disease (15). This article presents a systematic comparison of the prevalence of 29 autoimmune disorders for all patients in the nation of Denmark diagnosed with schizophrenia between 1981 and 1998 and their parents along with a group of healthy subjects and their parents.
Data from Danish registers were linked using the unique personal identification number that has been allocated to all residents in Denmark since 1968 (16). Socioeconomic information was obtained from the Integrated Database for Longitudinal Labor Market Research, which was created by linking a number of employment and education databases (17).
The Danish Psychiatric Register, which records contacts with psychiatric facilities throughout Denmark, was the source of patients with schizophrenia (18, 19). There are no private psychiatric facilities in Denmark, and all treatment is free of charge. Information on autoimmune diseases originated from the National Patient Register, which has collected data on all admissions to Danish hospitals since 1977 (20). Since 1995 it has included all contacts in emergency rooms and outpatient clinics. Diagnoses in the psychiatric and patient registers were according to ICD-8 until the end of 1993 and according to ICD-10 from the beginning of 1994. The disease categories used here were designed to separate, as much as possible, syndromes with an autoimmune basis from similar syndromes with other causes.
The study procedures were approved by the Danish Data Protection Board and the Johns Hopkins Bloomberg School of Public Health Committee on Human Research.
The 7,704 schizophrenia patients comprised all persons over age 15 admitted to a Danish psychiatric facility for the first time between 1981 and 1998 with a diagnosis of schizophrenia and known maternal identity. This group of patients, 66% of whom were male, has been described elsewhere (21). For each patient, 25 comparison subjects matched by year of birth and sex were selected randomly from a 5% sample of the Integrated Database for Longitudinal Labor Market Research. Comparison subjects were excluded if they had been admitted to a psychiatric facility before the first admission of the patient. In 92% of the patients and 96% of the comparison subjects, the father was known. Socioeconomic and disease information on the patients, comparison subjects, and their parents relates to the status just before the first contact with the patient. Wealth of the parents was organized into the highest quartile of the two parents for each individual, based on the distribution of wealth in the entire Integrated Database for Longitudinal Labor Market Research.
Statistical methods included estimation of prevalence proportions in patients and comparison subjects, and conditional logistic regression analyses that yielded an incidence rate ratio. The strata for the logistic regression are formed by the matching variables of year of birth and sex. Multivariate models adjust for known risk factors for schizophrenia: urbanicity of birth, socioeconomic status, and family history of schizophrenia. An initial logistic regression model predicts the occurrence of schizophrenia from prior occurrence of any of 29 autoimmune diseases. Later models are developed that predict schizophrenia from the occurrence of each of the specific autoimmune diseases. Separate prediction models are developed from the autoimmune status of the patients and from the parents of the patients to suggest effects of genetics as distinct from environment.
There were 29 autoimmune diseases with which either the patient or a parent was diagnosed before the patient had been diagnosed with schizophrenia. t1 shows prevalence data. There were 175, 16, and two patients with one, two, and three autoimmune diseases, respectively. Schizophrenia was associated with nearly 50% higher lifetime prevalence of one or more autoimmune disorders (t2). The analysis in t2 adjusts for known risk factors for schizophrenia (as well as controlling for sex and age in the matching process), revealing relationships that mirror the scientific literature (22).
Nine autoimmune diseases had higher lifetime prevalence among schizophrenia patients than among comparison subjects at a 95% level of statistical significance: thyrotoxicosis, intestinal malabsorption, acquired hemolytic anemia, chronic active hepatitis, interstitial cystitis, alopecia areata, myositis, polymyalgia rheumatica, and Sjögren’s syndrome (t3). Two disorders had sizable incidence rate ratios but did not meet traditional levels of significance: thyroiditis (incidence rate ratio=3.3) and ankylosing spondylitis (incidence rate ratio=2.7). Even so-called significant findings were based on small numbers: a single case of Sjögren’s disorder produced the incidence rate ratio of 12.5; likewise three cases of acquired hemolytic anemia produced the large incidence rate ratio of 12.5.
Twelve autoimmune diseases had higher prevalence among parents of schizophrenia patients than among parents of comparison subjects: thyrotoxicosis, thyroiditis, type 1 diabetes, intestinal malabsorption, pernicious anemia, acquired hemolytic anemia, interstitial cystitis, psoriasis, seropositive rheumatoid arthritis, other rheumatoid arthritis, dermatomyositis, and Sjögren’s syndrome (t4). Chronic active hepatitis, alopecia areata, myositis, and polymyalgia rheumatica were autoimmune disorders with higher prevalences than expected in the schizophrenia patients but not in their parents.
Five autoimmune disorders appeared more frequently in patients with schizophrenia prior to schizophrenia onset as well as in the patients’ parents: thyrotoxicosis, intestinal malabsorption, acquired hemolytic anemia, interstitital cystitis, and Sjögren’s syndrome.
The relationship between intestinal malabsorption, or celiac disease, and schizophrenia was noticed as early as 1961 (23). Celiac disease is an immune-mediated reaction to gluten that presents with diarrhea, weight loss, and abdominal complaints as well as a range of less common signs and symptoms, including some psychiatric and neurological symptoms (24, 25). The psychological interpretation for the first case series was reinterpreted by Dohan (26) as an inherited defect in which the environmental trigger of gluten precipitated schizophrenia in some individuals. Dohan presented two series of ecological data supporting the idea: one temporal series from countries in Europe during World War II (26) and a number of comparisons in the western Pacific based on anthropological data (27). The literature includes case studies, biological explanations for the association, and clinical trials of gluten withdrawal (e.g., references 28–34). Data from the Oxford Record Linkage Study (35) revealed an odds ratio of about three for cross-sectional comorbidity of schizophrenia and celiac disease, and we have published a short report on celiac disease and schizophrenia that used a nearly identical dataset to this one (15).
Autoimmune thyroiditis is characterized by hypothyroidism with clinical manifestations of goiter and lymphocyte infiltration of the gland (36). Thyrotoxicosis (Graves disease) causes sustained hyperthyroidism with clinical complications in thyroid-associated ophthalmology and dermopathy (37). Kraepelin reported clinical observations of enlarged thyroid gland in dementia praecox (38). Subsequent clinical evidence has shown an excess of thyroid hormone dysfunction in schizophrenia, and some have attributed these observations to dysfunction of the hypothalamus-pituitary-thyroid axis and neuroleptic medications (39–42). Our findings on thyroid disorders confirm prior research (13, 14, 43, 44).
Acquired hemolytic anemia is the clinical manifestation of the production of antibodies against red blood cells (45, 46). Our findings on acquired hemolytic anemia are consistent with a prior study (13) that showed an excess of hemolytic anemia in (just two) relatives of schizophrenia patients compared with the single instance in healthy subjects (odds ratio=2.02, 95% CI=0.11–121.6).
Interstitial cystitis is a bladder condition characterized by increased urinary frequency or pelvic pain (47, 48). Although interstitial cystitis has been recently genetically linked to panic disorder (49), we are unaware of prior research linking it to schizophrenia. The causes and pathogenesis for interstitial cystitis are not known.
Sjögren’s syndrome is characterized by progressive destruction of the exocrine glands, manifesting in a decrease in the production of saliva and tears (50). Sjögren’s syndrome generally affects women, and the estimated prevalence varies from approximately 0.1%–6% (51–53). It sometimes occurs with other rheumatic disorders such as systemic lupus erythematosus, but we are unaware of any prior research linking it to schizophrenia.
This analysis does not support certain results in the epidemiologic literature on schizophrenia. For example, two studies have reported an excess of type I diabetes in the relatives of individuals with schizophrenia (13, 44), consistent with the data reported here in t4; other studies, however, have reported a negative association between the two disorders (54).
The most consistent finding in the area of schizophrenia and autoimmune diseases is the negative relationship with rheumatoid arthritis (12, 55, 56). The incidence rate ratio for the schizophrenia patients was very close to 1.0 in this analysis, whereas in most other studies rheumatoid arthritis is much less common in individuals with schizophrenia. Our analysis required that the rheumatoid arthritis appear before the patient was diagnosed with schizophrenia, which may have influenced this result, since many cases of rheumatoid arthritis have onset much later than the age at onset for schizophrenia. The incidence rate ratio for parents of schizophrenia patients was greater than 1.0, contrary to the expected direction. There are two small studies of rheumatoid arthritis in the mothers of schizophrenia patients (13, 57), which suggest an inverse relationship.
Parallel Genetic Studies of Schizophrenia and Autoimmune Diseases
One of the possible hypotheses for our observed results is that schizophrenia shares a genetic diathesis with the family of autoimmune diseases, yielding the nearly 50% increase in risk for other autoimmune disorders as shown in t2. For example, Becker and colleagues (58) hypothesized that common complex diseases may be a result of the collective effects of disease-specific loci, common nondisease-specific loci, and specific environmental triggers, the so-called common variants/multiple diseases hypothesis. This contrasts slightly with the notion that a single or limited number of genes specific to each autoimmune disorder might be associated with schizophrenia—straightforward pleiotropy (59). The data relevant to these hypotheses, for the most part, come from separate, parallel genetic research studies of specific disorders. For autoimmune diseases, a source of general vulnerability may be the HLA system. Association studies have highlighted the role of HLA genes for certain autoimmune diseases (60, 61). Case/control and family studies suggest that genes in HLA class II regions (e.g., HLA-DR3, DQA1, DRB1) are related to thyrotoxicosis and thyroiditis (60), but the evidence is limited (62). Several studies have suggested the HLA-related susceptibility for celiac disease lies in the DQ alleles (63, 64), but there is some evidence for linkage with other HLA regions (65, 66). There is a line of research on HLA class II (e.g., DQA, DRB1, and DQB1) in relation to the primary Sjögren’s syndrome (67–69). Little is known about associations linking HLA and interstitial cystitis. The epidemiologic association between all these disorders could be a result of 1) direct involvement of HLA antigens or 2) physical closeness between loci for the autoimmune disorders and schizophrenia loci in HLA regions.
Outside the HLA region, the search for variants for common autoimmune diseases has not, as yet, suggested many clusters related to schizophrenia (70). An exception is that linkage studies have suggested that schizophrenia and celiac disease may have genes that are close to each other or identical (71, 72). Another exception is that the HOPA (human opposite paired) gene on chromosome Xq13, which codes for a coactivator for the T4 receptor and in which mutations have been found associated with hypothyroidism (73), has recently been linked with the elevated risk of schizophrenia (74, 75). Because the HOPA gene is expressed throughout the CNS and other tissues, especially in the period of fetal development, the abnormality in the HOPA gene is hypothesized to raise risk for schizophrenia. Separate association studies have connected the 5,10-methylenetetrahydrofolate reductase (MTHFR) gene to schizophrenia (76) and to rheumatoid arthritis (77), and a similar pattern exists for the IL1B gene (78, 79). The CTL-4 gene has been associated with schizophrenia in at least one study (80), and with type 1 diabetes, autoimmune thyroid diseases, and rheumatoid arthritis in numerous studies (58). There have also been suggestive association study findings for the IL10 gene (81) and type 1 diabetes, rheumatoid arthritis, and Sjögren’s syndrome (58) as well as for the TNF gene (82) and psoriasis, rheumatoid arthritis, and type 1 diabetes (58).
These data suffer from important limitations. The rarity of the autoimmune disorders is problematic, even for this study that involved an entire nation. Ascertainment was based on diagnosis received in normal medical specialty settings. There is likely to be underascertainment, since individuals with many of these disorders will not always be in attendance at a specialty clinic or inpatient setting. The argument could be made that schizophrenia, and many autoimmune diseases, are so serious that they inevitably end up in specialty treatment and the register system. But some diseases, such as hypothyroid disorder and type 1 diabetes, may be treated by primary care practitioners and never enter the registers. However, for any given disease, these biases would exist equally for patients and comparison subjects, and for parents of both groups; as a consequence the net effect is to lower the degree of association, making the findings conservative.
Another explanation for the low prevalence of autoimmune diseases is that many of them have onset later than schizophrenia, so that the prevalence in patients is much lower than what might be expected. These findings—even findings on parents of patients—may be limited to a subset of autoimmune diseases that have early onset. This possibility cannot be addressed with these data. Another limitation of the present study is the fact that the analyses were carried out in subjects matched by gender, making it impossible to examine gender-related differences, even though the etiology of autoimmune diseases and of schizophrenia probably differ by sex (83, 84). Finally, we cannot be certain that treatment for autoimmune disease is not a risk factor for schizophrenia.
Conclusions from these analyses, especially when the focus is on individual disorders, must necessarily be circumspect because of the opportunistic nature of the statistical analysis. Results reviewed from genetic association and linkage studies likewise are suggestive at best. On the other hand, findings concerning celiac disease and autoimmune thyroid diseases are consistent with the scientific literature, and this analysis is a confirmation based on a stronger dataset than has existed before. In future clinical studies it may be interesting to search for a family history of autoimmune diseases, and specific autoantibodies, in patients with schizophrenia. Eventually, individual or family disease comorbidity may help to elucidate shared etiologic pathways.
Received Nov. 29, 2004; revision received Feb. 9, 2005; accepted March 10, 2005. From the Department of Mental Health, Bloomberg School of Public Health, The Johns Hopkins University; the National Centre for Register-Based Research, University of Aarhus, Aarhus, Denmark; and the Institute for Basic Psychiatric Research, Psychiatric Hospital, University Hospital, Aarhus, Denmark. Address correspondence and reprint requests to Dr. Eaton, Department of Mental Health, Bloomberg School of Public Health, The Johns Hopkins University, HH 852, 624 N. Broadway St., Baltimore, MD 21205; firstname.lastname@example.org (e-mail).Supported by the Danish National Research Foundation, the Stanley Medical Research Institute, and NIMH grant 53188.Dr. Ewald died in January 2004.The authors thank Noel Rose for advice provided for this study.