Risk of Depression and Suicidality among Diabetic Patients: A Systematic Review and Meta-Analysis

The purpose of this study is to conduct a systematic review and meta-analysis to evaluate the risk of depression and suicidality among diabetic patients. Methods: Medline, PubMed, EMBASE, Cochrane library, and Psych INFO were searched for studies published from 2008 onwards. Meta-analysis was conducted to estimate the pooled effect size. Sources of heterogeneity were investigated by subgroup analysis and meta-regression. Results: In total, 5750 articles were identified and of those, 17 studies on suicidality and 36 on depression were included in this study. Our analysis suggests a positive relationship between diabetes and depression (cohort studies odds ratio (OR) 1.49, 95% confidence interval (CI): 1.36–1.64 and cross-sectional studies OR 2.04, 95% CI, 1.73–2.42). Pooled OR values for suicidal ideation, attempted suicide, and completed suicide were 1.89 (95% CI: 1.36–2.63), 1.45 (95% CI: 1.07–1.96), and 1.85 (95% CI: 0.97–3.52), respectively. All findings were statistically significant except for completed suicide. Conclusions: The increased risk of depression and suicidality in diabetic patients highlights the importance of integrating the evaluation and treatment of depression with diabetes management in primary healthcare settings. Further research in this area is needed.


Introduction
Diabetes is considered one of the largest global epidemics and constitutes a public health emergency in many countries [1]. The global prevalence of diabetes among adults has nearly doubled in the last couple of decades, rising from 4.7% in 1980 to 8.5% in 2014, resulting in approximately 1.6 million deaths annually [2]. Diabetes is psychologically demanding given the chronic and large burden placed on diabetics for the self-management of their disease. Diabetic patients face several psychological challenges as a result of their illness, which may include: adherence to medical treatment and lifestyle modifications, need for continued monitoring for glycemic control, concerns for complications and disabilities, interference of symptoms with daily activities, and psychosocial difficulties at personal and interpersonal levels [3], all of which may eventually lead and interconnect to depression and in some cases suicide.
Depression is a common mental illness that negatively impacts productivity and quality of life [4]. It is reported that patients with depression die 5-10 years earlier than those without depression [5]. Evidence suggests that the co-morbidity of depression and diabetes is relatively common [6][7][8].
A previous meta-analysis estimated the prevalence of depression to be twice as high among diabetics compared to the general population [6]. Other systematic reviews corroborated these findings and

Data Sources and Search Strategies
A systematic literature search was undertaken using five relevant databases: PubMed, Medline, EMBASE, Psych INFO, and Cochrane Library. Search strategies focused on three major domains: (1) diabetes; (2) depression or suicidality; and (3) quantitative outcome measures.

Eligibility Criteria and Study Selection
In our study, eligible articles were required to: (1) be published in the English language in peer-reviewed journals since 2008, and be available in full text; (2) assess patients with type 1 and/or type 2 diabetes mellitus who either self-reported a physician's diagnosis of diabetes, were prescribed anti-diabetic medications, or were participants in studies using laboratory-based assessments; and (3) evaluate depressive disorders, or use of antidepressants or depressive symptoms based on validated standardized questionnaires. Studies were also selected by scanning the reference lists of previous systematic reviews on the topic. Articles had to include both exposed and non-exposed groups of participants and provide sufficient data to calculate odds ratio for depression among diabetics. Eligible articles were identified through title and abstract screening, followed by full text review. Two reviewers independently evaluated studies for relevance in a standardized manner. Non-agreement was resolved by discussion and adjudication.

Data Extraction
Data extraction strategies were developed and pilot-tested on 20 randomly selected included studies and then modified accordingly. Information extracted in duplicate from studies included: author, publication year, country, study design, follow-up time (for cohort studies), settings, total number, age and sex of study participants, method of diabetes evaluation, method of depression or suicidality assessment, reported effect measure with 95% confidence interval, and covariates for adjusted effect measures. Unadjusted odds ratios and 95% confidence intervals were calculated using the number of events in the exposed and non-exposed groups.

Risk of Bias
Two reviewers independently assessed the validity of eligible studies by using a modified version of the Newcastle Ottawa Scale (NOS) [26]. The NOS assesses representativeness of the study sample, comparability between respondents and non-respondents, ascertainment of depression or suicidality, and thoroughness of the reported descriptive statistics. Studies were rated as having a low, moderate, or high risk of bias.

Data Analysis
Statistical analysis was performed using the Comprehensive Meta-analysis software-version 3 (CMA-3) [27]. For depression, pooled odds ratios (both adjusted and unadjusted) were the effect measure of interest. While for suicidality, both odds ratios and pooled prevalence among diabetics were the focus of our analysis. As heterogeneity was likely to exist, a random effect model was used to calculate pooled estimates, which allows for estimating both within and between studies variation. We examined heterogeneity using the Cochran's Q heterogeneity test and I 2 as a measure for inconsistency [28]. Influential analysis was conducted for the effect of each study on the pooled estimate by reassessing estimates after removal of one study at the time. To visually assess for publication bias, the funnel plot method was used. The Egger's regression intercept method was also used to confirm and statistically test for the bias observed in the funnel plot [29]. When publication bias was detected, a Duval and Tweedie trim and fill method was used to calculate the adjusted effect size [30]. All analysis used a 5% level of significance (α = 0.05).

Study Selection
The PRISMA flow chart depicting the study selection is shown in Figure 1. A comprehensive search of the literature yielded a total of 5750 articles from which 427 articles qualified for full text screening and 355 were excluded due to one or more of the following conditions: (1) their full text was not available; (2) they were not written in English; (3) they focused on special populations; and/or (4) reported inadequate data or mixed outcomes. In total, 50 studies were analyzed, with 33 reporting data only on depression, 14 only on suicidality, and three studies reporting data on both depression and suicidality among diabetic patients [31][32][33].
Depression studies were stratified and examined on the bases of their study design. Tables 1 and 2 include summary of the characteristics of depression studies included in the review. (1) Cohort Studies There were 12 cohort studies included for depression. One study [34] reported two datasets for patients with prevalent and incident diabetes, leaving a total of 13 datasets for analysis. Three studies [35][36][37] used incident prescription of antidepressants as a proxy for depression diagnosis, five studies [34,[38][39][40][41] used clinical diagnostic criteria, and three [42][43][44] relied on questionnaires. One study used both questionnaires and prescription of antidepressants [42]. The follow-up period ranged from two [34] to 15 years [36]. Most studies were from Europe or North America, while only two studies were from Asia [38,39]. Ascertainment of diabetes was made by physician diagnosis (International Classification of Diseases (ICD) code) [34,36,[38][39][40]43], use of anti-diabetic medications [35,37,43], self-reported diagnosis of diabetes [44,45], and/or laboratory assessment [42,43].

Suicidality
Suicidality was stratified and examined on the bases of outcome. Tables 3 and 4 include summary of the characteristics of suicidality studies included in the review

Depression
Depression studies were stratified and examined on the bases of their study design. Tables 1 and 2 include summary of the characteristics of depression studies included in the review.

Suicidality
Suicidality was stratified and examined on the bases of outcome. Tables 3 and 4 include summary of the characteristics of suicidality studies included in the review.     (

1) Completed Suicide
Seven studies reported data on completed suicide (all cohort). Suicidal death was confirmed by either International Classification of Diseases (ICD) codes or examination of death certificate.
(2) Suicidal Attempts Five studies assessed suicidal attempts and self-harm (one cohort, one nested case-control and three cross-sectional). Attempted suicide was evaluated based on either ICD codes or self-reported information. (

3) Suicidal Ideation
Nine cross-sectional studies examined suicidal ideation. Three of them evaluated type 1 diabetics, one study evaluated type 2 diabetic patients, who were on insulin [32], while the remaining studies included both type 1 and type 2 patients. Two studies had minors as participants (adolescents) [25,73]. All studies used self-reported information to evaluate for the presence of suicidal ideation either as a response to a single question, or as a part of depression screening questionnaires, or during a standardized interview.

Depression
Depression results are reported as odds ratios (adjusted and unadjusted) by study design. The overall OR based on all depression studies (cohort and cross-sectional) was 1.79 (95% CI: 1.62-1.99), with significant heterogeneity (I 2 = 98.09%, Q = 9.66, and p-value < 0.001).

Suicidality
Suicidality results are reported as prevalence and odds ratios (adjusted and unadjusted) stratified by outcome.

Figure 2.
Forest plot for the association between depression and diabetes from cohort studies. Estimates are in the center of the box and lines represent 95% confidence intervals (CI). Diamond shows the pooled odds ratio size and its 95% CI.

Suicidality
Suicidality results are reported as prevalence and odds ratios (adjusted and unadjusted) stratified by outcome.

Suicidality
The results for suicidality subgroup analysis are included in Table 6, Figure 3. Due to the limited number of studies assessing attempted suicide, no subgroup analysis was conducted.

Influential Analysis and Publication Bias
Influential analysis revealed that no single study had a substantial influence on either the adjusted or unadjusted effect estimates. There was evidence of publication bias for depression among cohort studies based on the inspection of the funnel plot ( Figure 4) and Egger's test (p-value = 0.027). To account for this bias, a Duval and Tweedie test was used and reported an adjusted OR 1.25 (95% CI: 1.15-1.37). There was no evidence of publication bias for depression among cross-sectional studies (p-value = 0.721) and suicidality studies (p-value = 0.860).

Influential Analysis and Publication Bias
Influential analysis revealed that no single study had a substantial influence on either the adjusted or unadjusted effect estimates. There was evidence of publication bias for depression among cohort studies based on the inspection of the funnel plot ( Figure 4) and Egger's test (p-value = 0.027). To account for this bias, a Duval and Tweedie test was used and reported an adjusted OR 1.25 (95% CI: 1.15-1.37). There was no evidence of publication bias for depression among cross-sectional studies (p-value = 0.721) and suicidality studies (p-value = 0.860).

Discussion
To our knowledge, this study is one of the first meta-analyses that reports the prevalence of suicidality among diabetic patients and assesses the association between diabetes, depression, and suicidality using data from observational studies. Our results show that diabetic patients are more likely to have depression, experience suicidal ideations and attempt suicide compared to

Discussion
To our knowledge, this study is one of the first meta-analyses that reports the prevalence of suicidality among diabetic patients and assesses the association between diabetes, depression, and suicidality using data from observational studies. Our results show that diabetic patients are more likely to have depression, experience suicidal ideations and attempt suicide compared to non-diabetic patients. The findings of this study help highlight diabetics as a high-risk group for depression and suicidality.
In regards to depression, based on 13 cohort studies and 23 cross-sectional studies, our results suggest that there is a significant association between depression and diabetes. It is interesting to note that this association maintained its strength even after adjusting for potential confounders. Previous systematic reviews have shown that patients diagnosed with diabetes are at higher risk for depression [7,78]. Similarly, our analysis of cohort studies corroborates the existing evidence [78] and suggests a directionality, whereby diabetes may play a causal role in the development of depression. However, the psychological burden of diabetes may lead to but does not fully account for the increased risk of depression [3]. Other potential physiological contributors include: activation of the HPA axis and SNS [11,79], chronic inflammation [80], and cerebral vascular changes induced by diabetes [81]. Additionally, some common medications used for the treatment of diabetes have been linked to a higher risk of depression [82,83].
In our study, the calculated odds ratio for cross-sectional studies (assessing prevalence of depression) was higher than the one in cohort studies (assessing incidence of depression), corroborating the results reported in a previous meta-analysis [6]. This finding may be explained in part by: (1) the longer duration of depression among diabetic patients, and (2) the potential bidirectional association between diabetes and depression. The longer duration of depression among diabetic patients may be due to the increased likelihood to experience treatment resistant and recurrent forms of depression [20]. This leads to a buildup of chronic cases. The potential bidirectional association between diabetes and depression has also been examined in the literature [42,84]. Depression has been linked to the development of type 2 diabetes with the use of certain antidepressants, which are known to have clinical effects on glucose homeostasis and weight gain [85]. Additionally, depression may have a negative effect on a patient's lifestyle choices including physical activity, leading to an increased risk to develop diabetes [86]. Thus, it is unsurprising that the risk of prevalent depression was higher than that of incident depression among diabetic patients in our study.
Suicidality among diabetic patients has not been fully elucidated. Our study examined the prevalence and risk of suicidality among diabetic patients in order to address the existing gap in the literature. In the general population, a study involving 17 countries found the lifetime prevalence of suicidal ideation and attempts to be at 9.2% and 2.7%, respectively [87]. By comparison, our study found that the prevalence of suicidal ideation among patients with diabetes was much higher, at a reported 16.2% (95% CI: 8.5-28.5%), while the rate of attempted suicide was similar, at 2.7% (95% CI: 0.9-7.8%).
Depression is one of the leading risk factors for suicidality [22,88]. In our study, diabetic patients were found to be twice as likely to experience suicidal ideations compared to non-diabetics. When examining attempted suicide, diabetics were also significantly at higher risk compared to non-diabetics. These findings are concerning and help highlight the vulnerability of diabetic patients possibly progressing from suicidal ideation to attempted suicide. Despite the increased risk of suicidal ideation and attempts, diabetic patients did not experience a significant risk of suicidal death (OR = 1.85, 95% CI: 0.97-3.52, p-value = 0.061). This could be attributed to several reasons connected to suicide, including: (1) stigmatization; (2) misclassification; (3) low occurrence; (4) limited details and number of studies; and (5) lack of research into the distinct predictive factors for suicidal death.
Stigma related to suicide is a major barrier in accurately reporting and tracking of suicidal deaths [89]. The high level of stigmatization in many countries, where suicide is considered to be immoral and illegal, might lead to underreporting of suicide as a cause of death [90]. This in turn would negatively impact accuracy of suicide rates reported in large-scale epidemiological studies [90]. Misclassification of suicidal death due to insulin overdoses as accidental death or death due to natural causes is another possibility that need to be further evaluated. Studying suicide is also statistically challenging because of the relative low occurrence of the event and the need for large samples to obtain reliable estimates [91]. Additionally, in the present systematic review, the limited number of studies assessing suicidal attempts lacked some key details. For example, there was dearth of information assessing the seriousness of the attempt (i.e., extent of hospital care required afterward or whether the attempt resulted in permanent disabilities). Furthermore, it has been suggested in the literature that suicidal ideations and behaviors might have different predictors than completed suicide [92,93]. Therefore, we cannot fully rely on suicidal ideations to understand completed suicide. Factors that may play a role in moving from one condition to another remain unclear for diabetic patients and need to be carefully investigated.

Strengths and Limitations
This study has several strengths. It provides an up-to-date literature review, which includes both depression and suicidality as outcomes of interest among diabetic patients. It provides evidence that can be used as a reference point for future research focusing on specific diabetic sub-populations. Finally, it takes into consideration several important factors (i.e., method of depression evaluation, type of diabetes, whether diabetes was incident or prevalent, and geographical location) in the analysis of our data, which adds to the robustness of this study.
However, there are also several limitations we need to consider. First, there are few studies assessing suicidality, especially attempted suicide. Second, there was a marked heterogeneity among the included studies. Third, we included studies that used prescription of antidepressants as a proxy for depression. This method is quite controversial especially among diabetic patients since antidepressants are commonly used for symptomatic treatment of diabetic neuropathic pain, which may falsely increase reported estimates. However, our subgroup analysis showed that differences in depression evaluation methods did not significantly impact our results. Fourth, confounding bias could not be entirely eliminated. Fifth, association of diabetes and suicidal attempts and suicidal ideation do not imply causation. Finally, most studies included in our systematic review were from developed countries and therefore, the results of our analysis need to be interpreted with caution, as they may not be generalizable to the developing world.

Implications for Future Research and Clinical Practice
There are significant gaps and opportunities for research in this important public health topic. Further research is warranted to: (1) examine the factors, mechanisms and transitional triggers implicated in the association between diabetes, depression and suicidality; (2) assess the role and impact of different diabetes management strategies on the patient's risk of depression and suicidality; and (3) evaluate the cultural and ethnic differences as they relate to diabetes, depression and suicidality. Clinicians should be aware and receive cross-training in order to be better prepared to address the higher risk for depression and suicidality among diabetic patients. Additionally, several other initiatives can be considered, including: (1) early detection and treatment of depression, which would improve diabetes control and consequently, delay the development of diabetic and depression related complications; (2) the development of robust and standardized validated screening tools for diabetic patients at risk for depression and suicidality; and (3) timely design and implementation of comprehensive interventions that take into account the complex inter-relationships between depression and suicidality, so as to improve the quality of life of diabetic patients.

Conclusions
Our study found that diabetes is associated with an increased risk for depression, suicidal ideation and suicidal attempts but not completed suicide. Therefore, efforts for early detection and effective screening are urgently needed in primary care settings. Appropriate training for healthcare providers in the field of suicidality screening and depression management would help mitigate the negative impacts on a diabetic patient's quality of life and reduce the growing burden on healthcare systems. Given the limited number of studies on this important topic, further research is needed.
Author Contributions: J.M., R.E., Y.B., and L.T. were involved in the study conception and design. J.M. and R.E. were responsible for the data analysis. All authors contributed to the discussion, interpreted the findings, helped write, reviewed/edited the manuscript for intellectual content, and read and approved the final manuscript.
Funding: This research was supported in part by an internal grant from the School of Public of Public Health, University of Saskatchewan.