1 INTRODUCTION

Diabetes mellitus is one of the leading causes of years lived with disability worldwide,1 reflecting a global core health challenge of the 21st century. Furthermore, type 2 diabetes (T2D) represents the most common diabetes type, with a substantially increasing number of cases.2 Moreover, T2D is linked to depression. Compared to the general population, the prevalence of major depressive disorder (MDD) is remarkably higher among persons with T2D.3 Based on a meta-analysis, approximately one in four T2D patients shows concurrent depressive symptoms or disorders4—whereas only one in ten adults reports depressive symptoms in the general population in Germany.5

Given the emotional distress due to living with T2D,6 it seems not surprising, that people with T2D have a higher risk to develop a MDD compared to people without T2D.7

Besides, there is strong evidence for a bi-directional relationship since meta-analyses of longitudinal studies also indicate an increased risk for incident T2D associated with depression, presumably owing to shared common risk factors such as adverse socioeconomic or lifestyle conditions and alterations in metabolic networks.7-9 However, the evidence is heterogenous and various concepts and definitions of depression have been used, including self-reported depressive symptoms and psychological distress. Furthermore, the use of antidepressant medication was shown to be associated with an increased risk for incident diabetes, which may confound the association between depression and diabetes.8 Thus, the specific role of clinically relevant MDD regarding T2D risk remains unclear. Moreover, previous studies did not examine the prospective impact of depression on changes in blood glucose levels, also covering prodromal changes from normoglycaemia to prediabetes. In addition, a subgroup meta-analysis indicated that the association of depression and T2D risk may differ between men and women, but only a small number of studies have reported sex-specific estimates, so far.7

Therefore, the present study aimed to analyse the impact of clinically relevant MDD on 12-year changes in glycaemic status among adults without diabetes at baseline, based on a nationwide, population-based cohort from Germany. We specifically asked for men and women: 1) Is MDD associated with prospective changes in glycaemic status, including remission and progression? 2) Is MDD associated with the onset of T2D?

2 METHODS 2.1 Data basis

Figure 1 shows a participant flow diagram describing the data basis and definition of the study population.

Participant flow diagram describing the definition of the study population (n = 1,766) and glycaemic status changes between baseline (GNHIES98-MHS) and follow-up (DEGS1). Frequencies of itemized changes: n1 = 162 (29 men, 133 women), n2 = 28 (10 men, 18 women), n3 = 812 (340 men, 472 women), n4 = 153 (99 men, 54 women), n5 = 26 (10 men, 16 women), n6 = 184 (92 men, 92 women), n7 = 53 (13 men, 40 women), n8 = 6 (4 men, 2 women), n9 = 18 (6 men, 12 women), n10 = 212 (118 men, 94 women), n11 = 33 (18 men, 15 women), n12 = 79 (37 men, 42 women). GNHIES98-MHS: German National Health Interview and Examination Survey 1998, mental health supplement; DEGS1: German Health Interview and Examination Survey for Adults; MDD: major depressive disorder

Data basis was a cohort study of adults aged 18–65 years and without diagnosed or unknown diabetes at baseline who participated in the mental health supplement of the ‘German National Health Interview and Examination Survey 1998’ (GNHIES98-MHS) in 1997–1999 and within the examination part of the ‘German Health Interview and Examination Survey for Adults’ (DEGS1) in 2008–2011. Both national health surveys targeted non-institutionalized adults residing in Germany; design and methods have previously been described in detail.10-12 In brief, a two-stage stratified cluster sampling procedure was used in GNHIES98 to select participants aged 18–79 years (n = 7,124). Overall, 87.6% (n = 4,181) of GNHIES98 participants aged 18–65 years also took part in the mental health supplement GNHIES98-MHS, among them n = 3,783 participants without diagnosed or unknown diabetes. The subsequent DEGS1 (n = 8,151) combined a nationally representative survey (n = 4,192, 18–79 years of age) and a longitudinal follow-up of former GNHIES98 participants (n = 3,959, 18–91 years), including n = 2,452 GNHIES98-MHS participants without diagnosed or unknown diabetes. Among them, n = 1,891 also took part in the DEGS1 examination part.

For this study, we excluded persons with missing information for defining glycaemic status changes between baseline and follow-up (n = 21), with missing information for any covariables (n = 88), and individuals with incident type 1 or gestational diabetes at follow-up (n = 16). The final study sample was n = 1,766 (men: n = 776; women: n = 990).

GNHIES98 and DEGS1 were conducted according to the Federal and State Commissioners for Data Protection guidelines. DEGS1 was approved by the local ethics committee at Charité – Universitätsmedizin Berlin in October 2008 (No: EA2/047/08). The implementation of the surveys conformed to the principles of the Helsinki Declaration. Participants provided written informed consent prior to participation.

2.2 Glycaemic changes

Diagnosed diabetes was defined as self-reported physician-diagnosed diabetes in standardized computer-aided personal interview (CAPI) or as taking any antidiabetic medication (Anatomical Therapeutic Chemical (ATC) Code A10) within the 7 days preceding the interview. Using HbA1c measurements, participants without diagnosed diabetes were categorized into normoglycaemia [< 39 mmol/mol (<5.7%)], prediabetes [39 ≤ HbA1c < 48 mmol/mol (5.7–6.4%)] and unknown diabetes [≥48 mmol/mol (≥6.5%)] according to previous definitions using recommendations of the American Diabetes Association.13 For this study, diagnosed and unknown diabetes at follow-up were summarized as ‘diabetes’. HbA1c was assessed from venous blood samples drawn in both surveys, using a Diamant high-performance liquid chromatography (HPLC) analyzer (Bio-Rad Laboratories, Munich, Germany) in GNHIES98, and an immunoturbidimetric method (ARCHITECT ci8200; Abbott, Wiesbaden, Germany) in DEGS1. Both methods were traceable to the National Glycohemoglobin Standardization Program and the comparability of methods was carefully assessed in a metabolically healthy subset of study participants aged 18–39 years as previously shown.13, 14

Glycaemic changes over time were categorized into ‘remission’ (from prediabetes to normoglycaemia), ‘stability’ (unchanged normoglycaemic or prediabetic status) and ‘progression’ (from normoglycaemia to prediabetes or diabetes, or from prediabetes to diabetes).

2.3 MDD assessment

Lifetime MDD (‘yes’ vs. ‘no’) was assessed by clinically trained interviewers at baseline in the GNHIES98-MHS and according to diagnostic criteria of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV),15 using a modified German version of the WHO Composite International Diagnostic Interview (DIA-X/M-CIDI),16 a standardized and fully-structured clinical face-to-face interview. DSM-IV criteria for MDD diagnosis require at least five of nine symptoms persisting nearly every day for a minimum of 2 weeks, accompanied by clinically significant distress and impairment. Depressive mood and/or decreased interest/pleasure should be present. MDD exclusion criteria are lifetime (hypo)manic episodes and depressive symptoms due to direct physiological effects of a substance, a general medical condition or attributable to grief.

2.4 Covariables

All covariables described below were assessed at baseline.

Standardized self-administered questionnaires were used to collect information on sex, age, educational level (categorized as ‘low’, ‘medium’ vs. ‘high’ according to the Comparative Analysis of Social Mobility in Industrial Nations criteria, CASMIN) and living alone (being the only person living in the household, ‘yes’ vs. ‘no’).

Social support was categorized into ‘low’ (up to three persons) versus ‘high’ (at least four persons), based on a question of the Oslo-3 Social Support Scale (“How many people are so close to you that you can count on them if you have serious personal problems?”).17

The German Diabetes Risk Score (GDRS)18 was used to estimate the predicted 5-year risk of T2D, based on self-reported information on age, smoking, physical activity, prevalent hypertension, family history of diabetes, intake of coffee, wholegrain and red meat, as well as standardized measures of body height and waist circumference, as previously described.19

Comorbid conditions were assessed via CAPI. Calculating the number of self-reported physician-diagnosed chronic diseases other than diabetes (categorized into ‘no’, ‘one’ vs. ‘at least two’), we considered the following conditions: a lifetime history of cancer, myocardial infarction, stroke, chronic heart failure, osteoarthritis, osteoporosis, Parkinson's disease, and cirrhosis of the liver as well as asthma, rheumatoid arthritis, gout, hepatitis, gastric-duodenal ulcus, epilepsy, hypertension and dyslipidaemia within the past 12 months.

Information on antidepressant medication use (past 7 days) was recorded by trained health professionals via computer-assisted brown bag review of all medications—prescribed and Over-The-Counter (OTC) products. Universal product codes on the original medication containers were coded according to the ATC classification system. An ATC code N06A was considered as ‘antidepressants’ (including herbal antidepressants).

Self-reported utilization of health services (past 12 months) was assessed for several outpatient medical and therapeutic services via self-administered questionnaire. Mental health care use was defined as using a psychiatrist, neurologist or psychotherapist at least once during the past year (‘yes’ vs. ‘no’).

2.5 Statistical analysis

Percentages, means and 95% confidence intervals (95% CIs) were calculated to describe the study participants’ characteristics. All analyses were performed separately for men and women. Significance of differences was evaluated for each of the 10 variables based on Rao-Scott Chi-Square test (for categorial variables) or Adjusted Wald test (for continuous variables), using a significance level of 0.005 due to Bonferroni correction for multiple testing (calculated by dividing the conventional significance level by the number of variables: 0.05/10). The mean difference of HbA1c measurements (follow-up minus baseline) is only reported for participants without self-reported diagnosed diabetes at follow-up (740 men and 933 women), since accompanied antidiabetic treatment may affect blood glucose levels.

The impact of MDD on prospective glycaemic changes was evaluated based on multinomial logistic regression analyses (reported as risk rate ratio [RRR], which can be interpreted as odds ratio [OR], with 95% CI; reference category: ‘stability’), adjusted for potential confounders at baseline, derived from the literature and available data in the survey: sex, educational status, living alone, social support, chronic conditions, 5-year T2D risk (including age), antidepressant medication and mental health care use—as well as the interaction of sex with MDD. As a result, we conducted sex-stratified analyses and added the considered covariables consecutively: Model 1 adjusted for age. Model 2 adjusted for age, educational status, living alone, social support and chronic conditions. Model 3 additionally accounted for the 5-year T2D risk (including age) without considering age separately anymore. Model 4 additionally adjusted for antidepressant medication and mental health care use.

In addition, the association of baseline MDD with incident T2D was evaluated based on sex-stratified multivariable logistic regression analyses (reported as OR with 95% CI), using MDD and consecutively added covariables (see Model 1 – 4) as independent variables.

A conventional significance level of 5% was considered statistically significant (two-sided tests). All statistical tests were performed with Stata SE 17.0 and its survey design procedures, using cohort-specific weighting factors adjusting for the demographic-geographic population structure and re-participation probabilities. The re-participation probability of GNHIES98-MHS participants without diabetes in the examination part of DEGS1 was derived from a generalized linear mixed model including the independent variables age at the time of GNHIES98-MHS (six categories) and at the time of DEGS1 (eight categories) as well as education (three categories), income (three categories), smoking (yes or no), migration background (yes or no) and 12-month MDD (yes or no) at the time of GNHIES98-MHS.

3 RESULTS 3.1 Study population and descriptive results

Figure 1 shows the GNHIES98-MHS-cohort and itemized glycaemic status changes between baseline and follow-up, depending on initial MDD status. Among 1,766 study participants, a total number of n = 293 participants aged 18–65 years was diagnosed with lifetime MDD at baseline (72 men, 221 women). Thus, lifetime MDD prevalence was 15.1% (men: 8.9%, women: 21.4%; p-value for sex difference <0.001). After a mean follow-up period of 11.9 years (range: 9.6–14.1 years), 69.5% of the participants with baseline MDD showed stability of their initial glycaemic status (men: 59%, women: 73.9%), 23% progressed to prediabetes or diabetes (men: 27%, women: 21.3%) and 7.5% showed remission from prediabetes to normoglycaemia (men: 14%, women: 4.8%). Among participants without MDD, 67.1% showed stability (men: 66.3%, women: 68.1%), 22.4% progressed (men: 21.6%, women: 23.3%) and 10.5% remitted (men: 12.1%, women: 8.6%). The proportion of incident cases of diagnosed or unknown T2D after 12 years was 5.9% among persons with baseline MDD (men: 7.7%; women: 5.2%) and 7.1% among individuals without MDD (men: 6.4%; women: 7.9%). Excluding participants with diagnosed diabetes at follow-up (36 men and 57 women), the mean difference of HbA1c measurements was 0.19% for participants with MDD (men: 0.19%, women: 0.19%) and 0.14% for participants without MDD (men: 0.16%, women: 0.12%).

Baseline characteristics of the study population are displayed in Table 1 for men and women with and without MDD. After applying a Bonferroni correction, the following significant sex differences were observed: Among women, there was a lower proportion of prediabetes (p < 0.001), a higher mean age (p < 0.001), a lower HbA1c mean value (p < 0.001), a higher antidepressant application frequency (p < 0.001), and a higher mental health care utilization rate (p = 0.002) compared to men. No significant differences of baseline characteristics were observed with regard to MDD status, except that participants with MDD reported mental health care use more often than participants without MDD (men: p = 0.002, women: p < 0.001).

TABLE 1. Baseline characteristics of the study population Men Women With MDD (n = 72) Without MDD (n = 704) Total (n = 776) With MDD (n = 221) Without MDD (n = 769) Total (n = 990) Total, % (95% CI) 8.9 (6.9, 11.5) 91.1 (88.5, 93.1) 100 21.4 (18.4, 24.9) 78.6 (75.1, 81.6) 100 Glycaemic status, % (95% CI) Normoglycaemia 67.3 (52.4, 79.4) 69.8 (64.6, 74.4) 69.5 (64.6, 74.1) 80.8 (73.1, 86.8) 80.2 (76.5, 83.4) 80.3 (76.8, 83.4) Prediabetes 32.7 (20.6, 47.6) 30.2 (25.6, 35.4) 30.5 (25.9, 35.4) 19.2 (13.2, 26.9) 19.8 (16.6, 23.5) 19.7 (16.6, 23.2) HbA1c in %, mean (95% CI) 5.5 (5.3, 5.6) 5.4 (5.3, 5.4) 5.4 (5.4, 5.5) 5.3 (5.2, 5.3) 5.3 (5.3, 5.4) 5.3 (5.3, 5.3) Age (years), mean (95% CI) 39.6 (36.0, 43.2) 39.2 (37.9, 40.5) 39.2 (38.0, 40.4) 42.3 (40.4, 44.3) 41.9 (40.6, 43.1) 42.0 (40.9, 43.1) Educational level, % (95% CI) low 45.7 (30.9, 61.3) 43.7 (38.6, 49.0) 43.9 (38.8, 49.1) 40.4 (33.0, 48.2) 44.1 (39.0, 49.3) 43.3 (38.7, 48.0) medium 45.7 (32.2, 59.8) 41.3 (36.5, 46.2) 41.6 (36.9, 46.5) 47.3 (40.4, 54.3) 46.4 (41.5, 51.2) 46.5 (42.4, 50.7) high 8.7 (4.5, 16.2) 15.0 (12.2, 18.3) 14.5 (11.8, 17.5) 12.4 (8.1, 18.5) 9.6 (7.1, 12.8) 10.2 (7.9, 13.0) Living alone, % (95% CI) no 25.5 (13.9, 42.0) 28.2 (23.4, 33.6) 28.0 (23.3, 33.2) 24.3 (17.6, 32.6) 20.8 (16.7, 25.7) 21.6 (17.8, 25.9) yes 74.5 (58.0, 86.1) 71.8 (66.4, 76.6) 72.0 (66.8, 76.7) 75.7 (67.5, 82.4) 79.2 (74.3, 83.4) 78.4 (74.1, 82.2) Social support, % (95% CI) low 30.3 (19.0, 44.6) 35.1 (30.9, 39.8) 34.7 (30.6, 39.1) 40.5 (32.7, 48.9) 34.0 (29.6, 38.7) 35.4 (31.4, 39.6) high 69.7 (55.4, 81.0) 64.9 (60.2, 69.3) 65.3 (61.0, 69.5) 59.5 (51.1, 67.3) 66.0 (61.3, 70.4) 64.6 (60.4, 68.6) Chronic somatic conditions, % (95% CI) no 55.4 (40.0, 69.8) 57.3 (52.4, 62.0) 57.1 (52.5, 61.6) 47.0 (38.6, 55.5) 58.0 (53.4, 62.5) 55.6 (51.5, 59.7) 1 33.1 (20.7, 48.4) 29.4 (25.2, 33.9) 29.7 (25.7, 34.0) 32.0 (23.9, 41.3) 24.8 (21.2, 28.8) 26.4 (23.1, 29.9) ≥ 2 11.6 (6.2, 20.5) 13.4 (10.7, 16.5) 13.2 (10.9, 15.9) 21.1 (14.6, 29.4) 17.2 (14.1, 20.7) 18.0 (15.0, 21.5) 5-year T2D risk in %, mean (95% CI) 2.3 (1.6, 2.9) 2.7 (2.4, 3.1) 2.7 (2.4, 3.0) 1.5 (1.2, 1.8) 2.2 (1.6, 2.7) 2.0 (1.6, 2.4) Antidepressant medication use, % (95% CI) no 97.6 (93.0, 99.2) 98.4 (96.7, 99.2) 98.3 (96.8, 99.1) 93.4 (89.4, 96.0) 94.0 (91.7, 95.7) 93.9 (92.0, 95.4) yes 2.4 (0.8, 6.9) 1.6 (0.8, 3.3) 1.7 (0.9, 3.2) 6.6 (4.0, 10.6) 6.0 (4.3, 8.3) 6.1 (4.6, 8.0) Mental health care use, % (95% CI) no 87.1 (76.7, 93.2) 96.3 (94.4, 97.6) 95.5 (93.7, 96.8) 84.5 (78.3, 89.2) 93.0 (90.5, 95.0) 91.2 (88.6, 93.3) yes 12.9 (6.8, 23.3) 3.7 (2.4, 5.6) 4.5 (3.2, 6.3) 15.5 (10.8, 21.7) 7.0 (5.0, 9.6) 8.8 (6.7, 11.4) Note The final study sample comprised n = 1,766 adults aged 18–65 years without diabetes at baseline who participated in the mental health supplement of the ‘German National Health Interview and Examination Survey 1998’ (GNHIES98-MHS) in 1997–1999 and re-participated an average of 12 years later in the examination part of the ‘German Health Interview and Examination Survey for Adults’ (DEGS1) in 2008–2011, after excluding persons with incident type1 diabetes or gestational diabetes at follow-up (n = 16), with missing information for defining glycaemic status at baseline or follow-up (n = 21), and with missing information for any covariables (n = 88). MDD: major depressive disorder; 95% CI: 95% confidence interval.

Baseline characteristics depending on participants’ glycaemic status are provided as supporting information (see Table S1). After applying a Bonferroni correction, participants with prediabetes had a higher HbA1c mean value (men: p < 0.001, women: p < 0.001), a higher mean age (men: p < 0.001, women: p < 0.001), a higher occurrence of chronic somatic conditions (men: p = 0.005, women: p < 0.001), and a higher 5-year T2D risk (men: p < 0.001, women: p < 0.001) compared to participants with normoglycaemia. Furthermore, the proportion of living alone differed depending on glycaemic status (men: p = 0.001, women: p = 0.005), as well as the educational level among women (p = 0.001).

3.2 Association between MDD and glycaemic changes

In the overall study population, MDD had no effect on remission (RRR = 0.80; 95% CI: 0.43, 1.49; p = 0.482) or progression (RRR = 1.08; 95% CI: 0.70, 1.66; p = 0.732) of glycaemic status over time (fully adjusted model, ref.: stability), but there was a significant interaction effect of sex and MDD on remission (p-value for interaction = 0.049) and a tendency for an interaction effect of sex and MDD on progression (p-value for interaction = 0.082). Table 2 shows the results from multinomial logistic regression analyses for men and women.

TABLE 2. Impact of lifetime major depressive disorder (MDD) at baseline on glycaemic changes at follow-up Men Women Remission (n = 102) Stability (n = 478) Progression (n = 196) Remission (n = 108) Stability (n = 677) Progression (n = 205) RRR (95% CI) p-value RRR RRR (95% CI) p-value RRR (95% CI) p-value RRR RRR (95% CI) p-value Model 1 1.30 (0.51, 3.30) 0.574 1.0 (ref) 1.42 (0.73, 2.79) 0.300 0.50 (0.27, 0.90) 0.022 1.0 (ref) 0.82 (0.49, 1.37) 0.444 Model 2 1.28 (0.51, 3.20) 0.602 1.0 (ref) 1.51 (0.82, 2.80) 0.184 0.44 (0.24, 0.82) 0.011 1.0 (ref) 0.80 (0.47, 1.37) 0.415 Model 3 1.31 (0.53, 3.23) 0.554 1.0 (ref) 1.62 (0.89, 2.97) 0.115 0.45 (0.24, 0.84) 0.012 1.0 (ref) 0.88 (0.52, 1.49) 0.630 Model 4 1.48 (0.60, 3.70) 0.394 1.0 (ref) 1.80 (0.97, 3.33) 0.063 0.43 (0.23, 0.82) 0.011 1.0 (ref) 0.90 (0.53, 1.55) 0.708 Note Relative Risk Ratios (RRR) with 95% confidence intervals (95% CI) and p-values from multinomial logistic regression. Bold type indicates significant differences (local significance level α = 0.05). Model 1: adjusted for age; Model 2: adjusted for age, educational status, living alone, social support, chronic somatic conditions; Model 3: adjusted for educational status, living alone, social support, chronic somatic conditions and 5-year T2D risk; Model 4: see Model 3, additionally adjusted for antidepressant medication and mental health care use; all covariables at baseline.

Among women, baseline MDD was associated with a lower chance for remission, irrespective of covariates (model 4: RRR = 0.43; 95% CI: 0.23, 0.82; p = 0.011). There were no significant associations between MDD and progression (model 4: RRR = 0.90; 95% CI 0.53, 1.55; p = 0.708).

Among men, there were no significant associations of MDD with remission (model 4: RRR = 1.48; 95% CI 0.60, 3.70; p = 0.394) or progression (model 4: RRR = 1.80; 95% CI: 0.97, 3.33; p = 0.063).

3.3 Association between MDD and T2D incidence

Table 3 shows the impact of MDD on incident T2D for men and women. For both sexes, MDD had no significant effect on incident T2D (men: OR = 1.58; 95% CI: 0.55, 4.52; p = 0.389; women: OR = 0.76; 95% CI: 0.37, 1.58; p = 0.457; see model 4).

TABLE 3. Impact of lifetime major depressive disorder (MDD) at baseline on incident type 2 diabetes (T2D) at follow-up Men (n = 776) Women (n = 990) OR (95% CI) p-value OR (95% CI) p-value Model 1 1.37 (0.52, 3.61) 0.517 0.66 (0.33, 1.32) 0.241 Model 2 1.25 (0.45, 3.47) 0.668 0.60 (0.29, 1.26) 0.174 Model 3 1.47 (0.53, 4.08) 0.456 0.74 (0.35, 1.56) 0.431 Model 4 1.58 (0.55, 4.52) 0.389 0.76 (0.37, 1.58) 0.457 Note At follow-up, n = 65 men and n = 71 women showed incident T2D. Odds Ratios (OR) with 95% confidence intervals (95% CI) and p-values from logistic regression. Model 1: adjusted for age; Model 2: adjusted for age, educational status, living alone, social support, chronic somatic conditions; Model 3: adjusted for educational status, living alone, social support, chronic somatic conditions and 5-year T2D risk; Model 4: see Model 3, additionally adjusted for antidepressant medication and mental health care use; all covariables at baseline. 4 DISCUSSION 4.1 Key findings

Based on a prospective population-based study, we examined the association of MDD with glycaemic changes and incident T2D after 12 years among adults without diabetes in Germany. Overall, we found no significant effect of MDD on remission or progression over time, but there was a sex-specific impact on glycaemic changes: Women with versus without MDD showed a remarkably lower chance for remission from prediabetes to normoglycaemia, irrespective whether covariables were considered or not. Among men, baseline MDD was not significantly related to glycaemic changes. For both sexes, MDD showed no significant association with incident T2D.

4.2 Surrounding evidence

Overall, previous meta-analyses of prospective data revealed relative risks for incident T2D associated with depression ranging from 1.26 to 1.60, but findings largely vary between studies.7-9 When considering only clinically relevant interview-defined MDD as within our study, one longitudinal analysis indicated a pronounced T2D risk (RR = 2.23).20 Furthermore, studies controlling for unknown diabetes at baseline showed a slightly higher T2D risk associated with depression (RR = 1.54),9 whereas studies using physiological examinations or clinical records for T2D assessment at follow-up showed smaller effects compared to studies focusing on self-reported diabetes, solely.7 Moreover, also the age range of studies may be crucial, since a subgroup meta-analyses indicated a higher T2D risk associated with depression for persons aged less than 50 years (RR = 1.96) compared to persons ≥50 years (RR = 1.50).7

Based on a subgroup meta-analysis it was already suggested, that the association of depression and incident T2D varies by sex, with a presumably more pronounced T2D risk among men (RR = 1.57) compared to women (RR = 1.26).7 Thus, our results may partly correspond to previous findings, since effect sizes and courses of associations among men also indicate an aggravation of glycaemic status and an increased T2D risk associated with MDD over time—and low statistical power may account for the lacking significance of findings within this specific subgroup (see strengths and limitations). However, MDD had no impact on glycaemic status progression or on incident T2D among women within our study—albeit the outlined analysis indicated that the association may be weaker compared to men, but still apparent.7 Nevertheless, three of the included studies revealed an association between depression and incident T2D among men, solely, and further research on sex differences is still needed.

Finally, by focusing on changes in blood glucose levels over time, our findings reveal an entirely different role of depression between men and women. Thus, depression among women may not provoke an aggravation of gly