Early-Onset Type 2 Diabetes and Mood, Anxiety, and Stress-Related Disorders: A Genetically Informative Register-Based Cohort Study

Type 2 diabetes has been increasing in prevalence for decades, and this trend has been particularly evident in younger adults (1). The earlier onset of type 2 diabetes poses long-lasting health challenges to patients. While it is clear that people with early-onset type 2 diabetes have worse physical health outcomes, including faster and more disruptive disease progression and increased risks of diabetes complications and mortality (2,3), less is known about their mental health.

The high occurrence of psychiatric disorders—especially mood disorders (including unipolar depression and bipolar disorder) (4,5), anxiety (6), and stress-related disorders (7)—in type 2 diabetes is well documented; however, studies examining psychiatric burden in early-onset type 2 diabetes remain scarce. In only two studies, both cross-sectional, have investigators explicitly focused on early-onset type 2 diabetes regarding psychiatric disorders; they reported higher prevalence of depressive symptoms among younger patients (age <45 years) than their older counterparts (8,9). Since these psychiatric comorbidities were associated with poorer diabetes management and life-course outcomes (10,11), a refined exploration of their association with early-onset type 2 diabetes in younger adults is valuable for planning preventive and early detection strategies.

Moreover, the underlying factors contributing to the comorbidity between early-onset type 2 diabetes and these psychiatric disorders are not yet clear. Beyond the potential bidirectional effects, where having one condition increases the risk of the other, the comorbidity seems to share pathogenesis mechanisms (10,12). In previous genome-wide association studies (GWAS), investigators observed shared genetic risk variants between type 2 diabetes and mood disorders (13). However, GWAS identify common genetic variants, which does not account for all disease heritability, and thus capture a fraction of genetic overlap between conditions. Familial coaggregation studies provide insights on whether conditions cluster together within families and are often the first step in genetic epidemiological research to explore whether genetics and/or shared environments explain an association (14). Quantitative genetic modeling quantifies the degree to which the phenotypic correlation between conditions can be attributed to genetic and environmental factors (15). Thus, a study with use of these two complementary genetically informative methods can further disentangle contributions from genetic and environmental factors to the co-occurrence between type 2 diabetes and these psychiatric disorders, which is important for identifying preventive and therapeutic targets.

In this population-based cohort study, we aimed to investigate the association and familial coaggregation between early-onset type 2 diabetes and clinically diagnosed mood, anxiety, and stress-related disorders and to quantify contributions of genetic and environmental factors to these associations.

In this population-based cohort study, we linked data from multiple Swedish nationwide registers. The Total Population Register was used to identify the study cohort and obtain information on migration and vital status (16). The Multi-Generation Register (MGR), a component of the Total Population Register, was used to identify biological kinships (16). Data on clinical diagnoses coded with ICD-8, -9, and -10 codes were obtained from the National Patient Register (NPR) (17). Information on medication dispensations was obtained from the Prescribed Drug Register (PDR), where filled prescriptions are documented classified with Anatomical Therapeutic Chemical (ATC) codes (18). Details regarding the registers used in this study can be found in Supplementary Table 1.

We included individuals born in Sweden between 1968 and 1998. Individuals were excluded if they 1) had no identifiable biological parents, 2) were diagnosed with a chromosomal abnormality, or 3) emigrated or died before age 5 years. The cohort selection process is shown in Supplementary Fig. 1. The cohort was followed up from birth until emigration, death, or 31 December 2013—whichever occurred first.

Four types of relative pairs were identified within the cohort: full siblings, maternal half-siblings, paternal half-siblings, and full cousins. These four types of relatives represent different levels of genetic and environmental relatedness.

The study was approved by the Regional Ethical Review Board in Stockholm, Sweden (Dnr2013/862-31/5). Register-based studies are exempt from informed consent in Sweden (19).

We defined early-onset type 2 diabetes as being diagnosed with type 2 diabetes before age 45 years, in line with previous studies (20,21). Diagnoses of type 2 diabetes were identified with use of data from the NPR and the PDR. For diagnoses firstly made with ICD-8/9 code 250.0, which does not differentiate between type 1 and type 2 diabetes, we required a subsequent diagnosis of ICD-10 code E11 and prescriptions of oral glucose-lowering medications (ATC code A10B) and no diagnosis of type 1 diabetes (ICD-10 code E10). Furthermore, we excluded individuals with diagnoses made before age 18 years because, during the period under consideration in this study, type 2 diabetes was rare in the Swedish pediatric population (22).

Psychiatric diagnoses were obtained from the NPR. Five categories were examined, including any diagnosis and specific diagnosis of unipolar depression, bipolar disorder, anxiety, and stress-related disorders. We excluded individuals who received any of the diagnoses before age 5 years because clinical diagnoses of these psychiatric disorders made at that young age are rare and can be phenotypically different from the diagnoses made at a later age (23). A complete list of ICD-8, -9, and -10 codes used can be found in Supplementary Table 2.

Using logistic regression models, we first estimated the association between early-onset type 2 diabetes and the examined psychiatric disorders by comparing individuals with and without type 2 diabetes. We then evaluated the familial coaggregation pattern across different types of relatives by estimating the risk of psychiatric disorders in individuals with a relative with and without type 2 diabetes.

The interpretation of familial coaggregation analyses consists of two aspects. First, a higher risk of psychiatric disorders in relatives of individuals with type 2 diabetes would suggest that shared familial liability contributes to the observed phenotypic associations between type 2 diabetes and these psychiatric disorders. Second, differences in the magnitudes of risk estimates across types of relatives would indicate the source of the familial liability, i.e., genetic or environmental. On average, full siblings share 50% of segregating genes, maternal and paternal half-siblings share 25%, and cousins share 12.5%. Full siblings and maternal half-siblings generally share a more similar environment than paternal half-siblings (e.g., prenatal intrauterine environment), while cousins share even fewer environmental factors compared with siblings (24). Therefore, a greater association in full siblings than maternal half-siblings would indicate the contribution of shared genetic factors, while a greater association in maternal than paternal half-siblings would suggest the contribution of shared environmental factors.

We repeated all analyses using Cox regression models to account for the temporal order of disorder occurrence. Age was used as the underlying timescale (i.e., we compared individuals of the same age), and the exposure of interest (i.e., type 2 diabetes or psychiatric disorder) was modeled as a time-varying variable. In the association analyses, individuals were regarded as unexposed before their first diagnosis and exposed thereafter. In the familial coaggregation analysis, individuals were considered unexposed before their relative’s first diagnosis and exposed after that. Follow-up time was calculated from birth until the first diagnosis of either outcome, emigration, death, or end of the study. We also compared models in which the exposure was type 2 diabetes with models in which the exposure was a psychiatric disorder to explore whether the temporal ordering of these diagnoses influences the associations.

We further examined whether the observed familial associations could be explained by a direct effect of type 2 diabetes on the studied psychiatric disorders, which we analyzed by adjusting for type 2 diabetes in the relatives. If the association remains statistically significant after adjustment, it further supports the presence of shared familial risk factors (14). Considering sex differences in the incidence of type 2 diabetes and psychiatric disorders, we also performed analyses for males and females separately.

Analyses were adjusted for sex and birth cohorts (six categories with 5-year intervals). A cluster robust sandwich estimator was calculated to avoid distribution assumptions and account for family data dependence. Data management was performed with SAS (version 9.4), and analyses were conducted with drgee and survival packages in R software (version 4.1.2).

Finally, we applied quantitative genetic modeling to estimate the contribution of genetic and environmental effects to the co-occurrence of early-onset type 2 diabetes and psychiatric disorders. These analyses were conducted for full sibling and maternal half-sibling pairs. To avoid dependencies between sibling pairs, we selected only one sibling pair (the oldest, to ensure a longer follow-up time) from each family. The liability threshold model was chosen, assuming an underlying normal distributed liability of disease. If an individual has a diagnosis, the risk liability is assumed to be above an estimated threshold. The threshold was allowed to be freely estimated across sibling types, since the prevalence of disorders often differs between full siblings and maternal half-siblings.

Three types of effects were considered in the modeling: additive genetic effects (A), representing the cumulative influences of multiple genes that are presumed to contribute to the phenotypes; shared environmental effects (C), indicating the environmental impacts that made the two individuals in a sibling pair similar to each other; and the unique environmental effects (E), representing the environmental effects that influence one individual but not the other in a sibling pair. In addition to the full ACE model (including A, C, and E), we also fitted a nested AE model (including A and E). The Akaike information criterion (AIC) was used to compare the model fit, where the model with the smallest AIC was selected. The so-called direct symmetric approach was used to decompose the variance and covariance between the phenotype correlations to avoid incorrect inference (25).

Sex and birth cohorts were included as covariates, and the models were fitted with use of the OpenMx package (version 2.19.8) (26) in R.

Descriptive statistics for the study cohort are presented in Table 1. A total of 3,061,192 individuals were included in the study cohort, among whom 7,896 (0.3%) were diagnosed with type 2 diabetes before age 45 years (median age at diagnosis 34.3 years [interquartile range 28.4, 39.1]). Individuals with early-onset type 2 diabetes had higher prevalence of all the examined psychiatric disorders than individuals without. Descriptive statistics for the different types of relatives are presented in Supplementary Table 3.

Associations and familial coaggregation patterns between early-onset type 2 diabetes and the examined psychiatric disorders are presented in Fig. 1. Individuals with type 2 diabetes were at statistically significantly higher risk of being diagnosed with any of the studied psychiatric disorders (adjusted odds ratio [OR] 3.62 [95% CI 3.44, 3.80]) as well as with the specific disorders: unipolar depression (3.97 [3.75, 4.22]), bipolar disorder (4.17 [3.68, 4.73]), anxiety disorders (3.76 [3.54, 3.99]), and stress-related disorders (3.35 [3.11, 3.61]). Relatives of individuals with early-onset type 2 diabetes also had higher overall risks of the examined psychiatric disorders (ORs ranging from 1.03 to 1.57) compared with relatives of individuals without early-onset type 2 diabetes. All risks were substantially attenuated but persisted and remained statistically significant among full siblings of individuals with type 2 diabetes, including any diagnosis (1.47 [1.37, 1.59]) and specific diagnosis of unipolar depression (1.50 [1.36, 1.64]), bipolar disorder (1.57 [1.27, 1.96]), anxiety disorders (1.47 [1.34, 1.62]), and stress-related disorders (1.57 [1.40, 1.75]). Associations were stronger in full siblings compared with maternal half-siblings, indicating genetic contributions to the co-occurrence of type 2 diabetes and the examined psychiatric disorders. Furthermore, the magnitudes of associations were comparable between maternal and paternal half-siblings, suggesting a lack of contribution from shared environmental factors. Notably, the statistically significant associations observed in half-siblings and cousins further suggested the existence of familial liability between type 2 diabetes and psychiatric disorders.

Associations and familial coaggregation patterns remained consistent in Cox regression analyses (Supplementary Table 4). The associations were generally stronger when type 2 diabetes was modeled as the outcome (hazard ratios ranging from 2.87 to 3.97) than when it was modeled as the exposure (hazard ratios ranging from 1.33 to 2.50).

After adjustment for type 2 diabetes in the relatives, the observed familial associations remained statistically significant with slightly attenuated ORs (Supplementary Table 5). Sex-specific analyses showed similar patterns of results in males and females, although the CIs were wider than the original estimates (Supplementary Table 6).

In the quantitative genetic modeling analyses, AICs, in general, favored the AE models (Supplementary Table 7). Heritability was estimated to be 60% in early-onset type 2 diabetes and 52% in any diagnosis and specifically 45% in unipolar depression, 61% in bipolar disorder, 47% in anxiety disorders, and 43% in stress-related disorders (Supplementary Table 8). Statistically significant phenotypic correlations were observed between type 2 diabetes and all examined psychiatric disorders, with tetrachoric correlation coefficients ranging from 0.19 to 0.24 (Fig. 2 and Supplementary Table 9). Genetic factors explained a large proportion of these phenotypic correlations: 78% for any diagnosis, 51% for unipolar depression, 62% for bipolar disorder, 64% for anxiety disorders, and 58% for stress-related disorders, with the rest attributed to unique environmental factors (Fig. 2 and Supplementary Table 10).

In this nationwide population-based study, we observed associations between early-onset type 2 diabetes and mood, anxiety, and stress-related disorders. By using complementary genetically informative designs—familial coaggregation analysis and quantitative genetic modeling—we showed that these associations could partly be attributed to familial liability, with shared genetics being the primary contributor.

In agreement with previous studies (8,9), we found that individuals with early-onset type 2 diabetes have higher odds of having the examined psychiatric disorders, with similar patterns of associations observed in men and women. These associations persisted in taking the temporal order of diagnosis into account, regardless of type 2 diabetes being modeled as the exposure or the outcome. However, it should be noted that we did not intend to infer causality using the temporal order of diagnoses because the recorded date of diagnosis is merely an approximation of the occurrence of health problems.

Familial coaggregation analyses showed that the relatives of individuals with early-onset type 2 diabetes were more likely to have been diagnosed with all examined psychiatric disorders compared with relatives of individuals without type 2 diabetes, suggesting that shared familial liability contributed to the co-occurrence of the conditions. This inference was further supported by the sensitivity analyses where statistically significant familial associations remained after adjustment for type 2 diabetes in the relatives, since if these psychiatric disorders were simply a consequence of having type 2 diabetes, the associations would be expected to disappear after the adjustment. However, as the associations remained significant, this result supports the idea that type 2 diabetes and these psychiatric disorders share familial liability.

Moreover, all associations were stronger in full siblings than maternal half-siblings, suggesting that genetics plays a role in the shared familial liability. This inference is further supported by results from quantitative genetic modeling, where additive genetics was found to explain much of the phenotypic correlation. These findings are consistent with evidence from previous GWAS, which have revealed that shared susceptibility loci (i.e., pleiotropic genes) increase risk for both type 2 diabetes and mood disorders (13), including unipolar depression (27) and bipolar disorder (10,28). To our knowledge, the shared genetic risk between type 2 diabetes and anxiety and stress-related disorders has not been investigated. A recent GWAS observed significant genetic correlations between anxiety and stress-related disorders with higher waist-to-hip ratio and BMI, as well as overweight and obesity (29), which are known risk factors and share susceptibility gene variants with type 2 diabetes (30). Future research with use of molecular genetic data could be valuable for disentangling the underlying genetic pathogenesis of the comorbidity between type 2 diabetes and these psychiatric disorders.

Notably, up to 49% of the phenotypic correlations between early-onset type 2 diabetes and these psychiatric disorders were attributed to unique environmental factors. One possible explanation is that causal environmental factors influence both type 2 diabetes and psychiatric disorders. For example, multiple behavioral and psychological factors such as physical inactivity, unhealthy diet, disturbed sleep patterns, and chronic stress may be some common environmental denominators and mediators. Another possible explanation is that one disorder increases the risk of the other disorder via environmental mechanisms. On the one hand, having early-onset type 2 diabetes from a younger age may elevate the psychological stress level and potentially directly increase the risk of onset or deterioration of psychiatric conditions (31). On the other hand, the phenomenology of psychiatric symptoms, e.g., depressive symptoms, may contribute to the development of an unfavorable lifestyle, e.g., increased sedentary behavior, unhealthy dietary habits, and smoking. In addition, some pharmacotherapies for psychiatric disorders, such as using atypical antipsychotics and some antidepressants, have been liked with adverse metabolic outcomes, which may subsequently accelerate development of type 2 diabetes (32). These findings highlight the potential of identifying modifiable factors that can alleviate these comorbidities.

Furthermore, our findings are consistent with evidence suggesting shared physiological mechanisms linking type 2 diabetes and these psychiatric disorders. For example, a well-recognized mechanism involves the dysregulation of the hypothalamic-pituitary-adrenal axis and its consequence of hypercortisolism, which has been linked to disrupted glucose homeostasis and pathogenesis of many psychiatric disorders (33). Another potential mechanism is chronic inflammation and changes in immune-inflammatory regulations, such as elevated levels of cytokines and upregulated inflammasomes (34). Other alternative mechanisms, such as the leptin-melanocortin pathway (36), a critical neuroendocrine regulator of energy homeostasis, and the microbiota-gut-brain axis (36), have also gained increasing attention. These biological pathways may act as common underlying mechanisms affecting the liability to type 2 diabetes and psychiatric disorders and/or mediating their relationships. Nevertheless, mechanistic studies remain warranted to advance our knowledge of the possible pathophysiological pathways and mediators behind the observed associations.

The main strength of the current study was the use of a representative population-based cohort with large sample size and extensive follow-up. The prospectively collected data in the nationwide registers minimized selection, recall, and report bias. However, some limitations should be noted. We relied on diagnosis records from registers. Thus, our results depend on diagnosed patients who actively sought medical help and/or with more severe symptoms, which could lead to underestimating the occurrence of type 2 diabetes and psychiatric disorders. This register-based nature of our study also restricted us from analyzing the causal relationship, as the precise time of disease onset is not well captured in using medical health records. Moreover, in an observational study, we could not examine the involvement of potential pathophysiological pathways or the role of biological factors (e.g., transcription factors, cortisol levels, and inflammatory parameters) in the observed associations. Furthermore, the statistical power in quantitative genetic modeling is limited, which compromises the precision of the estimation for contributions from genetic and environmental components.

Our findings upheld current guidelines that type 2 diabetes onset at a younger age should be given special attention regarding psychological burden (37,38). The integration of mental health support within a coordinated multidisciplinary diabetes care team should be encouraged (38). Given the shared liability of type 2 diabetes with mood, anxiety, and stress-related disorders, cross-diagnostic prevention and treatment paradigms should be supported. First, individuals with early-onset type 2 diabetes may be genetically predisposed to the studied psychiatric disorders. Early detection of mental health conditions followed by adequate psychological support may prevent further progression of psychiatric morbidity. Specifically, it may be advisable to initiate psychological screening and follow-up earlier in patients with personal or family history of mental health conditions. Second, mental health professionals should be aware of familial predisposition toward type 2 diabetes in patients with mood, anxiety, and stress-related disorders. For individuals with type 2 diabetes or a family history of type 2 diabetes, alternative evidence-based nonpharmacological intervention, such as cognitive behavioral therapy and interpersonal therapy (38), and medication with less diabetogenic effects for managing psychiatric conditions, should be considered. This is to minimize the potential adverse metabolic effects of some psychotropic medications. Moreover, promoting a healthy lifestyle, such as a balanced diet and physical activity, can be important and beneficial in preventing and managing both type 2 diabetes and examined psychiatric disorders. The Lancet Psychiatry Commission recommended that the Diabetes Prevention Program be used as a lifestyle intervention model in the management of psychiatric disorders (39), which further puts forward the possibility of incorporating programs for co-occurred type 2 diabetes and psychiatric disorders.

In conclusion, we showed that early-onset type 2 diabetes is associated with a higher risk of unipolar depression, bipolar disorder, anxiety, and stress-related disorders. These psychiatric disorders demonstrated familial liability with type 2 diabetes, mainly attributed to genetic factors. Our findings highlight the need for clinical vigilance for mental health burden in people with early-onset type 2 diabetes. We further shed light on the critical role of genetics and modifiable environmental risk factors in the shared liability, suggesting that future research to identify risk factors that contribute to the comorbidity is needed.

Funding. Financial support was provided through the Swedish Research Council (2017-00788) and Karolinska Institutet, Strategic Research Program in Neuroscience (StratNeuro). A.B. was also supported by the Stockholm County Council (clinical research appointment 20180718) while working on this project. M.Le. received funding from the European Union Horizon 2020 research and innovation program under the Marie Sklodowska-Curie grant agreement no. 721567.

Duality of Interest. H.L. has served as a speaker for Medici, Evolan Pharma, and Shire Sweden and has received research grants from Shire Sweden, outside of the submitted work. M.La. has received lecture honoraria from Lundbeck. M.Le. is employed at Janssen Pharmaceutical Companies of Johnson & Johnson. J.F.L. has coordinated an unrelated study on behalf of the Swedish Inflammatory Bowel Disease Quality Register (SWIBREG), which has received funding from Janssen Corporation. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.L. and A.B. conceived and designed the study. S.L. analyzed data and wrote the first draft of the manuscript. R.K.-H. and M.Le. assisted with the study design and data analysis. M.Le., J.F.L., S.G., P.L., M.La., S.E.B., M.J.T., H.L., R.K.-H., and A.B. assisted with the methods and contributed to interpreting the results and writing the final manuscript. All authors approved the final version of the manuscript to submit for publication. S.L. and A.B. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.