We calculated risk as hazard ratios (HRs) adjusted for age, sex, calendar year, and marital status. The broken horizontal line indicates HR of 1.00 (no difference); error bars, 95% CIs.
eAppendix 1. Information on Depression Obtained From the Danish Psychiatric Central Research Register and the Danish National Prescription Registry
eAppendix 2. Information on Severe Mental Illness Obtained From the Danish Psychiatric Central Research Register
eAppendix 3. Information on Diabetes Mellitus Obtained From the Danish National Diabetes Register
eAppendix 4. Information on Dementia Obtained From the Danish Psychiatric Central Research Register, the Danish National Patient Register, and the Danish National Prescription Registry
eAppendix 5. Information on Marital Status Obtained From the Danish Civil Registration System
eAppendix 6. Information on Chronic Diseases and Diabetic Complications Obtained From the Danish National Patient Register
Katon W, Pedersen HS, Ribe AR, Fenger-Grøn M, Davydow D, Waldorff FB, Vestergaard M. Effect of Depression and Diabetes Mellitus on the Risk for DementiaA National Population-Based Cohort Study. JAMA Psychiatry. 2015;72(6):612-619. doi:10.1001/jamapsychiatry.2015.0082
Copyright 2015 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
Although depression and type 2 diabetes mellitus (DM) may independently increase the risk for dementia, no studies have examined whether the risk for dementia among people with comorbid depression and DM is higher than the sum of each exposure individually.
To examine the risk for all-cause dementia among persons with depression, DM, or both compared with persons with neither exposure.
Design, Setting, and Participants
We performed a national population-based cohort study of 2 454 532 adults, including 477 133 (19.4%) with depression, 223 174 (9.1%) with DM, and 95 691 (3.9%) with both. We included all living Danish citizens 50 years or older who were free of dementia from January 1, 2007, through December 31, 2013 (followed up through December 31, 2013). Dementia was ascertained by physician diagnosis from the Danish National Patient Register or the Danish Psychiatric Central Register and/or by prescription of a cholinesterase inhibitor or memantine hydrochloride from the Danish National Prescription Registry. Depression was ascertained by psychiatrist diagnosis from the Danish Psychiatric Central Research Register or by prescription of an antidepressant from the Danish National Prescription Registry. Diabetes mellitus was identified using the National Diabetes Register.
Main Outcomes and Measures
We estimated the risk for all-cause dementia associated with DM, depression, or both using Cox proportional hazards regression models that adjusted for potential confounding factors (eg, demographics) and potential intermediates (eg, medical comorbidities).
During 13 834 645 person-years of follow-up, 59 663 participants (2.4%) developed dementia; of these, 6466 (10.8%) had DM, 15 729 (26.4%) had depression, and 4022 (6.7%) had both. The adjusted hazard ratio for developing all-cause dementia was 1.83 (95% CI, 1.80-1.87) for persons with depression, 1.20 (95% CI, 1.17-1.23) for persons with DM, and 2.17 (95% CI, 2.10-2.24) for those with both compared with persons who had neither exposure. The excess risk for all-cause dementia observed for individuals with comorbid depression and DM surpassed the summed risk associated with each exposure individually, especially for persons younger than 65 years (hazard ratio, 4.84 [95% CI, 4.21-5.55]). The corresponding attributable proportion due to the interaction of comorbid depression and DM was 0.25 (95% CI, 0.13-0.36; P < .001) for those younger than 65 years and 0.06 (95% CI, 0.02-0.10; P = .001) for those 65 years or older.
Conclusions and Relevance
Depression and DM were independently associated with a greater risk for dementia, and the combined association of both exposures with the risk for all-cause dementia was stronger than the additive association.
Diabetes mellitus (DM) and major depression are very common in Western populations. Type 2 DM occurs in approximately 8% to 14% of most Western populations,1 whereas approximately 25% of women and 16% of men will have a major depressive episode during their lifetime.2Quiz Ref ID As many as 20% of persons with type 2 DM have comorbid depression.3 Furthermore, a recent meta-analysis has established a bidirectional link between depression and DM.4 Patients with comorbid depression and DM have poorer adherence to diet, smoking cessation, exercise, and medication regimens to control DM.5 Depression is also associated with increased cortisol levels,6 autonomic nervous system dysregulation,7 and increased inflammation,8 all of which worsen glycemic control. Therefore, patients with comorbid depression and DM have an increased risk for microvascular and macrovascular complications and mortality.9
An extensive literature has also identified DM and depression as independent risk factors for dementia. A meta-analysis10 found that persons with DM have a 47% increased risk for all-cause dementia. Two recent meta-analyses11,12 found that depression doubled the subsequent risk for all-cause dementia. Quiz Ref IDTwo additional studies13,14 have shown that comorbid depression and type 2 DM were associated with a 2-fold greater risk for developing all-cause dementia compared with DM alone. A study of more than 29 000 patients with type 2 DM15 developed a 10-year risk prediction model for dementia that identified depression as an important risk factor. Another recent study using data from a large randomized clinical trial aimed at optimum control of glycemic and cardiovascular risk factors in patients with DM16 found that those with comorbid depression and type 2 DM compared with DM alone had a greater risk for cognitive decline during a 40-month period.
However, these prior studies have all been limited to cohorts of patients with DM13- 16 and so could not ascertain whether people with depression and DM have an elevated risk for dementia owing to an additive (or more than additive) interaction between the two. Given the increasing incidence of dementia in aging modern societies,17 understanding the risk associated with potentially modifiable depression and DM disease trajectories is essential.
In a population-based cohort of 2.4 million adults, we aimed to study the risk for all-cause dementia among persons with DM, depression, or both compared with persons who had neither illness. Given the rapidly increasing incidence of DM in younger groups18 and the demographic, clinical, and prognostic differences in patients who develop DM in middle vs older age,19,20 we also examined whether age (<65 vs ≥65 years) modified the risk for all-cause dementia in this population.
We conducted a population-based cohort study using data from the Danish Civil Registration System.21 This register includes information on sex and month of birth and continuously updated information on vital status and migration since 1968. In this register, Danish citizens are each assigned a unique personal identification number, providing accurate linkage to person-level data.21 Diagnoses in the registers are classified according to the Danish version of the International Classification of Diseases, Eighth Revision,22 before January 1, 1994. Hereafter, diagnoses were classified according to the International Statistical Classification of Diseases, 10th Revision.23 Our cohort included all individuals who were 50 years or older from January 1, 2007, through December 31, 2013, born in Denmark, free of dementia, and alive as of January 1, 2007. We followed up our sample through December 31, 2013, to ensure maximum validity of the dementia diagnoses24 and homogeneous calendar period.
The study protocol was approved by the Danish Data Protection Agency and the Danish Health and Medicines Authority. The need for informed consent was waived, and all data were deidentified.
Our primary independent variables of interest were the presence of depression, DM, or comorbid depression and DM. We identified individuals with depression by a diagnosis of depression made by a psychiatrist or by redemption of at least 1 prescription for an antidepressant using data from the Danish Psychiatric Central Research Register25 or the Danish National Prescription Registry, respectively (eAppendix 1 in the Supplement).26 The Danish Psychiatric Central Research Register contains diagnostic information on all psychiatric admissions since January 1, 1969, and outpatient specialty mental health visits since January 1, 1995.25 The Danish National Prescription Registry contains information on all prescriptions dispensed at Danish pharmacies since January 1, 1995, including the day of purchase and classification of drugs according to the anatomic-therapeutic-chemical classification.27 Individuals with schizophrenia, schizoaffective disorders, or bipolar disorder were censored at the date of diagnosis (eAppendix 2 in the Supplement). We supplemented our depression definition by identifying all antidepressant prescriptions (ie, selective serotonin reuptake inhibitors, monoamine oxidase inhibitors, and other nontricyclic antidepressants) redeemed from January 1, 1995 through December 31, 2013 (eAppendix 1 in the Supplement). Our primary definition of depression excluded redemption of tricyclic antidepressant prescriptions because of their frequent use for insomnia and/or pain and excluded bupropion hydrochloride and trazodone hydrochloride because neither was approved for the treatment of depression in Denmark during the study period.
Individuals diagnosed as having DM from January 1, 1990, through December 31, 2013, were identified in the National Diabetes Register using a validated algorithm28 (eAppendix 3 in the Supplement). Registration of DM is considered complete from 1995 onward, with a sensitivity of 86% and a positive predictive value of 89%.28
We identified incident all-cause dementia using data from the Danish National Patient Register,29 the Danish Psychiatric Central Research Register, and the Danish National Prescription Registry (eAppendix 4 in the Supplement). The Danish National Patient Register contains information on all Danish medical hospitalizations since January 1, 1977, and all outpatient contacts since January 1, 1995.29 Approximately two-thirds of all dementia cases in Denmark are diagnosed within the secondary health care system.24 Although the diagnosis of all-cause dementia in the Danish National Patient Register or the Danish Psychiatric Central Research Register has a positive predictive value of 86%,30 the validity is lower among individuals younger than 65 years for dementia subtypes.31 We identified all inpatient or outpatient contacts with a diagnosis of dementia made from January 1, 1969, through December 31, 2013, based on a validated algorithm.24 In addition, we supplemented our definition of all-cause dementia with redemption of at least 1 prescription for a cholinesterase inhibitor or memantine hydrochloride from January 31, 1995, through December 31, 2013. We excluded all cases of dementia prevalent before January 1, 2007, to identify all incident cases of dementia.
Covariates were chosen a priori based on their availability and prior research identifying their potential associations with depression, DM, and the risk for dementia.32 We obtained marital status information (defined as married or living in a registered partnership or single) from the Danish Civil Registration System (eAppendix 5 in the Supplement). We used the Danish National Patient Register to obtain data on all hospital contacts from January 1, 1977, through December 31, 2013, for 1 or more of the following chronic diseases: ischemic heart disease, congestive heart failure, peripheral vascular disease, atrial fibrillation/flutter, cerebrovascular disease, traumatic brain injury, chronic pulmonary disease, renal disease, retinopathy, and neuropathy (eAppendix 6 in the Supplement).
We used Cox proportional hazards regression models to estimate hazard ratios (HRs) and 95% CIs for the associations between depression and DM and the risk for all-cause dementia. Age was chosen as the underlying time scale, for which we corrected intrinsically. Individuals contributed at-risk time from January 1, 2007, or from their 50th birthday, whichever came last (delayed entry). Censoring occurred at the day of the dementia diagnosis, diagnosis of schizophrenia or bipolar disorder, death, immigration from Denmark, their 100th birthday, or January 1, 2014, whichever came first.
Apart from using age as the time scale, our primary regression model was adjusted for sex, marital status, and calendar period. Next, we adjusted for potential intermediates on the pathway from depression and DM to dementia, including medical comorbidities (ie, ischemic heart disease, congestive heart failure, peripheral vascular disease, atrial fibrillation/flutter, cerebrovascular disease, traumatic brain injury, and chronic pulmonary disease) and complications of DM (ie, renal disease, retinopathy, and neuropathy).10- 12 To minimize the possibility that any associations between depression and all-cause dementia risk could be confounded by similarities between late-life depressive symptoms and prodromal dementia,33 we added 2 years to the date of the initial diagnosis of depression or the initial prescription of an antidepressant. Furthermore, because guidelines recommend that individuals with suspected dementia have fasting blood glucose or hemoglobin A1c levels measured as part of the medical workup34 and are therefore likely to be diagnosed with DM soon after the dementia diagnosis, 1 year was added to the date of the initial diagnosis of DM. To validate this approach, we performed a sensitivity analysis in which we repeated our regression models stratified by time since the depression diagnosis and time since the DM diagnosis without postponement of these exposures; these models were adjusted for age, sex, and marital status.
We examined whether an additive interaction existed by testing the hypothesis of no excess hazard due to the interaction.35 We performed interaction analyses between DM and depression using the entire sample and stratifying by age (ie, ≥65 and <65 years) and calculated the attributable proportion due to interaction as a measure of the excess HR for individuals with both conditions not explained by the independent effects of either. In this setting, the attributable proportion (represented by AP) is given by the following formula35: APInteraction = (HRDepression + DM − HRDepression − HRDM + 1)÷ HRDepression + DM.All interaction analyses were adjusted for age, sex, marital status, and calendar period.
We conducted 2 secondary analyses. First, we created a categorical variable denoting early- vs late-onset DM using the median age of onset, 63 years, as the cut point. To facilitate this categorization based on the National Diabetes Register, this analysis was restricted to individuals born after 1932. Next, we ascertained the associations of our independent variables of interest with the risk for diagnosis of Alzheimer disease or diagnosis of vascular dementia individually in regression models adjusted for age, sex, calendar period, and marital status. In a sensitivity analysis, we examined whether our findings were affected by expanding our definition of depression to include prescription of tricyclic antidepressants.
We used 2-sided significance tests for all analyses, with statistical significance set at P < .05. The proportional hazards assumption was assessed graphically for all variables using the log-minus-log plots, and we found no violations. All statistical analyses were performed using commercially available software (Stata, version 13; StataCorp).
We followed up a cohort of 2 454 532 individuals for a total of 13 834 645 person-years, including 477 133 (19.4%) with a diagnosis of depression, 223 174 (9.1%) with a diagnosis of DM, and 95 691 (3.9%) with comorbid depression and DM. The mean age at the initial diagnosis of DM was 63.1 (SD, 12.0) years; at the initial diagnosis of depression, the mean age was 58.5 (SD, 13.5) years.
During the study period, 59 663 persons (2.4%) developed dementia. The mean age at the first diagnosis of dementia was 80.9 (SD, 8.7) years. Of those participants who developed dementia, 15 729 persons (26.4%) had depression alone, 6466 (10.8%) had DM alone, and 4022 (6.7%) had comorbid depression and DM (Table 1).
Compared with persons without depression or DM, DM alone was associated with a 20% greater risk for all-cause dementia (HR, 1.20 [95% CI, 1.17-1.23]); depression alone, with an 83% greater risk (HR, 1.83 [95% CI, 1.80-1.87]); and comorbid depression and DM, with a 117% greater risk (HR, 2.17 [95% CI, 2.10-2.24]) after adjustment for age, sex, calendar period, and marital status. The estimates decreased slightly after adjustment for chronic diseases (Table 2).
Quiz Ref IDAs shown in Figure 1, during the first year after the diagnosis of depression, the associated hazard for all-cause dementia was elevated nearly 7-fold (HR, 6.75 [95% CI, 6.55-6.95]), but thereafter it decreased consistently to approximately 2.00 (compared with individuals without depression). As shown in Figure 2, during the first year after the diagnosis of DM, the associated hazard for all-cause dementia was elevated 31% (HR, 1.31 [95% CI, 1.22-1.40]), with a decrease in subsequent years. The long-term HR rose 42% at 10 years after the diagnosis of DM (HR, 1.42 [95% CI, 1.38-1.47]).
Quiz Ref IDAmong participants younger than 65 years, the HRs for all-cause dementia were 2.93 (95% CI, 2.71-3.16) with depression alone, 1.71 (95% CI, 1.49-1.97) with DM alone, and 4.84 (95% CI, 4.21-5.55) with comorbid depression and DM (Table 3). The combined effect of the 2 illness exposures on all-cause dementia risk was larger than the sum of the 2 individual diseases; that is, the attributable proportion owing to the interaction was 0.25 (95% CI, 0.13-0.36; P < .001) for persons younger than 65 years and 0.06 (95% CI, 0.02-0.10; P = .001) for those 65 years or older. When we examined the impact of age at onset of DM, the HR for the association between early-onset DM and the risk for all-cause dementia was significantly higher (HR, 1.82 [95% CI, 1.73-1.91]) than that for late-onset DM (HR, 1.30 [95% CI, 1.24-1.36) (P < .001).
Quiz Ref IDDepression, DM, and their comorbid combination were all associated with an increased risk for Alzheimer disease (HR for DM alone, 1.06 [95% CI, 1.01-1.11]; HR for depression alone, 1.39 [95% CI, 1.35-1.44]; HR for comorbid depression and DM, 1.46 [95% CI, 1.37-1.55]). However, the magnitude of the associations of DM, depression, and their comorbid combination with the risk for vascular dementia was more pronounced (HR for DM alone, 1.55 [95% CI, 1.44-1.66]; HR for depression alone, 2.42 [95% CI, 2.29-2.55]; HR for comorbid depression and DM, 3.56 [95% CI, 3.28-3.86]). Finally, our results were unaffected by expanding our definition of depression to include redemption of a tricyclic antidepressant prescription (HR for DM alone, 1.20 [95% CI, 1.17-1.24]; HR for depression alone, 1.79 [95% CI, 1.75-1.82]; HR for comorbid depression and DM, 2.07 [95% CI, 2.01-2.14]).
In a nationwide, population-based cohort study of more than 2.4 million persons 50 years or older, DM and depression were associated with an increased risk for all-cause dementia, and the combined effect of both disorders appeared more than additive, especially among younger persons. Among those with comorbid depression and DM in our cohort, 6% of incident dementia may be accounted for by the interaction between depression and DM overall, and 25% may be accounted for among those younger than 65 years. Although the underlying risk for dementia is low in this younger group, the marked increase in the incidence of DM in younger groups18 makes this finding quite worrisome.
Our study extends beyond prior studies by identifying that, compared with a population without depression or DM, depression alone is associated with the highest relative risk for all-cause dementia. Further, we found similar results when examining associations with the risk for Alzheimer disease or vascular dementia, although the magnitude of the associations of depression and of comorbid depression and DM with the risk for vascular dementia was more pronounced, in line with the results of a recent meta-analysis.36 In addition, we found that having comorbid depression and DM is associated with a level of risk greater than that of the sum of the 2 illnesses. Although underlying causal mechanisms are unclear, one explanation could be that depression and DM have many shared risk factors for dementia, including increased inflammation, decreased insulin sensitivity, autonomic nervous system dysregulation, obesity, and vascular disease.37
Prior studies38 have found that patients with comorbid depression and DM are younger than those with DM alone and were diagnosed as having DM approximately 5 years earlier. Also, depression earlier in life may be a risk factor for developing type 2 DM.4 Given that depression in patients with DM is associated with poor self-care, nonadherence to treatment regimens, and adverse psychobiological changes,5- 8 this younger group with comorbid depression and DM may be vulnerable to developing dementia later in life.
From a public health perspective, developing screening and interventions to improve the quality of treatment of depression and DM in this subgroup of patients could be important in reducing the risk for dementia. A recent prospective cohort study of 1433 older adults with DM39 found that effective treatment of DM and depression and an improved diet could lead to as much as a 20% decrease in incident dementia. Primary care–based collaborative care models have been developed and shown to reduce depressive symptoms in patients with comorbid depression and chronic medical illnesses, such as DM and heart disease.40- 42 Although adequately powered trials of these interventions to test their effects on preventing dementia may require very large sample sizes (eg, >50 000 participants) and long durations (eg, 15-20 years), they are warranted given the societal costs of dementia.
Our study has several strengths and limitations. We followed up a nationwide cohort virtually without any losses, making nonresponse bias an unlikely explanation for our findings. Also, information on DM, depression, and dementia was collected prospectively and did not rely on patient or proxy recall. An important limitation of our study was that the population was from a single country with a well-developed national health care system and a relatively homogenous population, therefore limiting generalizability. However, this factor should improve internal validity, because the role of socioeconomic factors in health care–seeking behavior is expected to be minimal. Further, our definition of depression was based on a combination of psychiatric diagnoses and antidepressant prescription records, thereby introducing selection bias, because patients with more severe depression are more likely to be prescribed antidepressants and/or referred to psychiatrists.43,44 This issue is further complicated by our inability to capture individuals with depression who have not sought treatment.45 Similarly, the use of codes from the 8th and 10th revisions of the International Classification of Diseases to identify dementia and DM cases could miss patients with initial symptoms of these illnesses until symptoms or functional impairments become more prominent and limit our ability to differentiate dementia subtypes accurately.46,47 A further limitation was the lack of data on possible confounders such as health-risk behaviors, including smoking, obesity, and sedentary lifestyle. However, these lifestyle factors may be mediators of the associations presented herein, and previous studies did not find attenuation of the association of comorbid depression and DM with the risk for dementia after adjusting for health-risk behaviors.13,14 Finally, residual confounding remains a possibility, as in any observational study.
We found that depression and DM were both associated with a greater risk for all-cause dementia, Alzheimer disease, and vascular dementia. These associations appeared to be stronger among those individuals with depression alone compared with those with DM alone. Persons with comorbid DM and depression appeared to have the highest relative risk for dementia, and this association tended to be stronger than additive. The interaction between DM and depression tended to be particularly strong for individuals younger than 65 years. In light of the increasing societal burden of chronic diseases, further research is needed to elucidate the pathophysiologic mechanisms linking depression, DM, and adverse outcomes such as dementia and to develop interventions aimed at preventing these dreaded complications.
†Died March 1, 2015.
Submitted for Publication: November 12, 2014; final revision received January 20, 2015; accepted January 22, 2015.
Corresponding Author: Dimitry Davydow, MD, MPH, Division of Health Services Research and Psychiatric Epidemiology, Department of Psychiatry and Behavioral Sciences, University of Washington School of Medicine, 1959 NE Pacific St, Room BB1661, Campus Box 356560, Seattle, WA 98195 (firstname.lastname@example.org).
Published Online: April 15, 2015. doi:10.1001/jamapsychiatry.2015.0082.
Author Contributions: Dr Vestergaard had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Katon, Pedersen, Ribe, Fenger-Grøn, Vestergaard.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Katon, Ribe, Davydow.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Pedersen, Fenger-Grøn.
Obtained funding: Vestergaard.
Administrative, technical, or material support: Katon, Pedersen, Fenger-Grøn, Vestergaard.
Study supervision: Katon, Ribe, Davydow, Vestergaard.
Conflict of Interest Disclosures: None reported.
Funding/Support: The study was supported by an unrestricted grant from the Lundbeck Foundation and by grant KL2 TR000421 from the National Institutes of Health (Dr Davydow).
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.