Quantitative assessment of the bidirectional relationships between diabetes and depression

Diabetes and depression impose an enormous public health burden and the present study aimed to assess quantitatively the bidirectional relationships between the two disorders. We searched databases for eligible articles published until October 2016. A total of 51 studies were finally included in the present bidirectional meta-analysis, among which, 32 studies were about the direction of depression leading to diabetes, and 24 studies about the direction of diabetes leading to depression. Pooled results of the 32 eligible studies covering 1274337 subjects showed that depression patients were at higher risk for diabetes (odds ratio (OR) = 1.34, 95% confidence intervals (CI) = [1.23, 1.46]) than non-depressive subjects. Further gender-subgroup analysis found that the strength of this relationship was stronger in men (OR = 1.63, 95%CI = [1.48, 1.78]) than in women (OR = 1.29, 95%CI = [1.07, 1.51]). For the direction of diabetes leading to depression, pooled data of 24 articles containing 329658 subjects showed that patients with diabetes were at higher risk for diabetes (OR = 1.28, 95%CI = [1.15, 1.42]) than non-diabetic subjects. The available data supports that the relationships between diabetes and depression are bidirectional and the overall strengths are similar in both directions. More mechanistic studies are encouraged to explore the molecular mechanisms underlying the relationships between the two diseases.


INTRODUCTION
Diabetes is a chronic metabolic disease with increasing prevalence worldwide nowadays owing to the lifestyle modifications and increasing life expectancy [1]. According to the International Diabetes Federation, 415 million adults have diabetes and this figure is expected to rise to 642 million by 2040 [2]. As high blood glucose levels can damage blood vessels, nerves, eyes and kidneys, patients with diabetes may suffer from a number of serious complications. Depression is a common psychiatric mood disorder with increasing prevalence and is one of the largest single causes of disability around the world [3].
Diabetes and depression carry a large public health burden presently and also a heavy burden on health budgets in the future. In recent years, accumulating studies indicate the close associations between the two chronic diseases, diabetes and depression [4][5][6][7]. It is found that comorbid diabetes and depression is common, which results in significant detrimental impact on health outcomes, including a significantly increased risk of mortality and increased healthcare cost [8]. Although there is a high rate of comorbid depression in patients with diabetes, it is unrecognized and untreated largely [9,10]. Moreover, the prevalence of diabetes and depression differs in men and women. Women are about twice as likely as men to develop depression, while the prevalence of diabetes is higher in men than women [11,12]. Thus, the present study aimed to perform a bidirectional random effects meta-analysis to quantitatively assess the strength of the relationships between diabetes and depression, and to examine the difference of the strength in males and females, to provide implications for future interventional studies of the two diseases.

Risk of diabetes in patients with depression
Selection process of studies included in the present study was shown in Figure 1. There were 32 eligible studies covering 1274337 subjects included in the assessment of risk of diabetes in patients with depression. Table 1 showed the first author, year of publication, country, number of subjects, the odds ratio (OR) and the corresponding 95% confidence intervals (CI), extracted from each study. The metaanalytic results were shown in Figure 2, which indicated that depression patients were at higher risk of developing diabetes (OR = 1.34, 95%CI = [1.23, 1.46]). Statistically moderate heterogeneity was found among these studies (p = 0.000, I 2 = 68.4%).

Risk of diabetes in male patients with depression
Effects of gender on the strength of the bidirectional relationships between diabetes and depression were examined through further gendersubgroup analysis. A total of 10 studies were included in the male subgroup analysis. Table 2 listed the main characteristics about the number of male subjects, OR and the corresponding 95% CI of each study. According to the meta-analytic results shown in Figure  3, depression patients were at 63% increased risk for developing diabetes (OR = 1.63, 95%CI = [1.48, 1.78]) than non-depressive male subjects, and heterogeneity was not found among these studies (p = 0.446, I 2 = 0.0%).

Risk of diabetes in female patients with depression
There were 11 eligible studies included in the female subgroup analysis. Table 3 showed the extracted information of each eligible study. The pooled OR indicated that depression patients were at higher risk for diabetes (OR = 1.29, 95%CI = [1.07, 1.51], Figure 4) than non-depressive female subjects and statistically moderate heterogeneity was found among these studies (p = 0.000, I 2 = 75.3%).

Risk of depression in patients with diabetes
For the direction of diabetes leading to depression, there were a total of 24 eligible studies covering 329658 subjects. Table 4 collected the extracted main characteristics, including the first author, year of publication, country, and number of subjects, OR and the corresponding 95% CI of each included study. The pooled results were shown in Figure 5, which indicated that patients with diabetes were at higher risk for developing depression (OR = 1.28, 95%CI = [1.15, 1.42]) than nondiabetic subjects with statistically significant heterogeneity among these studies (p = 0.000, I 2 = 62.5%).

Publication bias
Funnel plot and Egger's test were employed to estimate the publication bias among the included studies. The funnel plot showed generally symmetrical distribution, suggesting no evidence of publication bias for the outcomes in both directions. Egger's test also found low possibility of publication bias (p = 0.262 for the direction of depression leading to diabetes and p = 0.326 for the direction of diabetes leading to depression).

DISCUSSION
Diabetes and depression are two highly prevalent chronic diseases around the world, and increasing evidence indicates the high rates of co-occurrence of the two diseases. The present study was designed to conduct a quantitative assessment on the bidirectional relationships between the two diseases to provide directions for their prevention and treatment. It was found that there was a 34% increased risk of developing diabetes in depression patients than non-depressive subjects, and a 28% increased risk of developing depression in patients with diabetes compared with non-diabetic subjects. Thus, the study provides evidence that the relationships between diabetes and depression are bidirectional with similar strengths in both directions.
The mechanisms accounting for the relationships between diabetes and depression should be multiple in terms of biological, psychological and socioeconomic determinants [64][65][66]. First, depression can be regarded as an additional risk factor for diabetes. Depression patients are less likely to be compliant with dietary and weight loss recommendations and are more prone to be lack of physical exercise, which can lead to worsening of obesity and insulin resistance, and thus may increase the risk of developing diabetes. Some drugs used to treat depression may lead to weight gain and obesity, which could make individuals susceptible to the development of diabetes [67,68]. Second, diabetes can also increase the risk of depression. For patients with diabetes, poor control of blood sugar, strict diet and physical exercise requirements, and treatment may increase the incidence of depression. It was found that chronic stress could cause hyper-activation of the hypothalamic-pituitary-adrenal axis and an increase in cortical, which has been proposed to be an important pathway to interpret the clinical relationships between diabetes and depression [69][70][71]. In addition, on the basis of the large-scale genome wide association studies about type 2 diabetes and major depressive disorder, we performed a bioinformatics analysis on the genetic overlap between the two diseases [72]. The overlapped single nucleotide polymorphisms and functional enrichment pathway of the annotated genes were identified, which provided clues for future treatment strategies for the two diseases [72].
Our analysis has several limitations. First, we analyzed the effect of gender on the risk of diabetes in patients with depression, while the current available data cannot permit us to perform the gender subgroup analysis for the direction of diabetes leading to depression. Second, as many studies did not specify the type of diabetes, we analyzed the overall relationships of diabetes with depression. Third, several case-control and cross-sectional studies were employed in the bidirectional analysis and thus we cannot exclude the recall and selection bias.

Search strategy and selection criteria
With the combination of the following terms "depression" or "depressive" and "diabetes", the Medline and PsycINFO databases were scanned for articles written in English and published in peer-reviewed journals up to October 2016. The conference proceedings and reference lists from retrieved articles were also screened to obtain additional relevant reports. Cohort, cross-sectional and case-control studies providing OR and 95%CI, or enough data to calculate the two parameters, of risk of developing diabetes in patients with depression or vice versa, were eligible for inclusion. Two investigators (QSZ and LS) independently screened the literature and extracted information from the eligible studies. Any disagreement was resolved by consensus. The flow chart in Figure 1 summarized the references selection process. In total, 51 studies were finally included in the bidirectional analysis , among which, 32 studies were about the risk of diabetes in patients with depression, and 24 studies were about the risk of depression in patients with diabetes.

Meta-analytic methods
The meta-analysis was performed for the extracted fully adjusted ORs employing the Stata statistical software version 12.0 (Stata Corp LP, College Station, Texas) with the random effects model. To assess the effect of gender on the relationships between diabetes and depression, the gender-subgroup analysis was further performed. Heterogeneity was evaluated employing Q-test and I 2 score. Funnel plot and Egger's test were used to examine the publication bias among the included studies.

CONCLUSIONS
In summary, the present quantitative assessmentprovides evidence that the relationships between diabetes and depression are bidirectional with similar strengths in both directions, which implies that physicians should be aware of the co-occurrence of depression and diabetes.