Introduction
Major depressive disorder (MDD) and type 2 diabetes are both important public health issues globally. It is estimated that around 322 million people had MDD and more than 425 million adults were living with diabetes worldwide in 2015 [
1,
2]. Epidemiological data showed that the prevalence of MDD among individuals with type 2 diabetes was twice that in those without diabetes [
3], which indicated a potential link between these two prevailing diseases.
Observational studies have reported that MDD is associated with an increased risk of type 2 diabetes through several biological alterations and unhealthy behaviours [
4,
5]. It has further been suggested that type 2 diabetes may play a role in the development of MDD, possibly by causing disability and comorbidity [
6]. However, whether the mutual association between MDD and type 2 diabetes is causal remains unclear due to potential residual confounding and reverse causation bias in observational studies [
7].
Mendelian randomisation (MR) is a method for assessing causal inference of an exposure on an outcome by using genetic variants as instrumental variables for the exposure [
8]. This technique diminishes residual confounding because genetic variants are randomly assorted at conception, thereby having no connection to self-selected lifestyle factors, behaviours and environmental factors. In addition, it overcomes reverse causality because genetic variants are fixed regardless of the development or progression of the disease (except for certain cancers).
MDD has been proposed as a risk factor also for other cardiometabolic diseases, such as coronary artery disease (CAD) and heart failure, in observational studies [
9‐
12]. Therefore, we conducted a bidirectional two-sample MR study to explore the causal associations of MDD with major cardiometabolic diseases as well as the causal role of cardiometabolic diseases for MDD.
Discussion
In the present bidirectional two-sample MR study, we observed a significant positive association of genetic liability to MDD with type 2 diabetes and CAD, and a suggestive association with heart failure. The reverse MR analysis provided no evidence that liability to type 2 diabetes, CAD or heart failure was related to MDD.
Adverse effects of MDD on type 2 diabetes development have been demonstrated consistently in observational studies. A meta-analysis of nine prospective studies proposed that depression was a risk factor for the onset of type 2 diabetes and reported a pooled RR of 1.26 (95% CI 1.13, 1.39), which is in agreement with our findings [
20]. Results of another updated meta-analysis, which included 13 prospective studies with 6916 incident cases of type 2 diabetes, showed that having depression increased the risk of type 2 diabetes by 60% [
21]. Additionally, a positive association between MDD and type 2 diabetes was supported by three large-scale cohort studies with medium or long follow-up periods (individual studies included 65,381 women followed up for 10 years, 11,694 adults for 6 years, and 5201 adults for 3.2 years) [
22‐
24]. In contrast, epidemiological data on the effect of type 2 diabetes on MDD risk are inconclusive [
25]. A systematic review including 83 studies found that type 2 diabetes could not independently predict MDD risk [
26], but some recently published studies reported a significant positive association in both directions [
21‐
24]. The present MR study found no evidence supporting a causal detrimental effect of type 2 diabetes on MDD. A cohort study that followed 65,381 women for 10 years found that the risk of developing clinical depression was higher in diabetic women, particularly in those who received insulin therapy, compared with non-diabetic women [
22]. Another cohort study reported that treated but not untreated type 2 diabetes was associated with an increased risk of depression [
23]. These studies suggest that MDD risk in type 2 diabetes may differ depending on type 2 diabetes treatment [
22,
23].
We found evidence that MDD may be causally associated with risk of CAD and possibly heart failure, thereby confirming the results of observational studies [
9‐
12,
27]. Notably, diabetes and CAD have been proposed as risk factors for heart failure, and CAD can explain more than 60% of heart failure cases [
28]. Thus, the suggestive association between MDD and heart failure might be partly mediated via type 2 diabetes and CAD.
There are several possible mechanisms that can explain the causal effect of MDD on type 2 diabetes risk. A series of biological abnormalities related to depression, including increased counter-regulatory hormone release and activity, alterations in glucose transport function and increased immunoinflammatory activation, may influence the risk of type 2 diabetes [
5]. In addition, lifestyle factors, such as smoking and alcohol consumption, may play a mediating role in the pathway from depression to type 2 diabetes [
29].
In traditional observational studies of the effect of depression on type 2 diabetes, the results are prone to be biased by reverse causality, as impaired glucose tolerance may lead to depression before diabetes symptoms manifest. Antidepressant medication, which has been demonstrated to be associated with hyperglycaemia, may introduce residual confounding [
25]. We used the MR study design, thereby avoiding reverse causality and minimising residual confounding. Another strength is that we extracted summary-level data from the hitherto largest GWAS for MDD and type 2 diabetes and, therefore, we had high power to detect even weak associations. A limitation of the present study is the potential of pleiotropy. However, we detected no directional pleiotropy in the MR-Egger regression analysis and the estimates were consistent when using the weighted median and MR-PRESSO analyses, which indicated a negligible distortion by potential pleiotropy. A previous MR study revealed that BMI was a risk factor for both MDD [
30] and type 2 diabetes [
31], indicating that BMI might be a pleiotropic factor or confounder in the pathway from MDD to type 2 diabetes. However, in a sensitivity analysis using BMI-adjusted estimates for the genetic associations with type 2 diabetes, the effect of MDD on type 2 diabetes remained at the conventional level of significance (
p < 0.05). This implies that the causal association between MDD and type 2 diabetes was not completely mediated or biased by BMI. Another limitation is that population stratification may have affected the result for the association between CAD and MDD because the GWAS of CAD included participants of different ancestry. However, as the majority (77%) of individuals were of European ancestry, any population stratification bias was expected to be small.
The present MR study strengthens the evidence that MDD is a potential risk factor for type 2 diabetes and CAD. However, there was no genetic support for a causal effect of type 2 diabetes or CAD on MDD. Whether MDD is causally related to heart failure needs further investigation. Given the high disease burden related to the causal link, it is recommended that MDD prevention, management and treatment should be enhanced for type 2 diabetes prevention.
Acknowledgements
Summary-level data for genetic associations with MDD, type 2 diabetes, CAD and heart failure were from the MDD GWAS dataset (the UK Biobank study, 23andMe and Psychiatric Genomics consortium), the DIAbetes Genetics Replication And Meta-analysis (DIAGRAM) consortium, the Coronary ARtery DIsease Genome wide Replication and Meta-analysis plus The Coronary Artery Disease Genetics (CARDIoGRAMplusC4D) consortium and the Heart Failure Molecular Epidemiology for Therapeutic Targets (HERMES) consortium. The authors thank all investigators for sharing these data.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.