DIABETICMedicine

DOI: 10.1111/dme.14054

Systematic Review or Meta-analysis

Longitudinal associations between depression and
diabetes complications: a systematic review and meta-
analysis

A. Nouwen1 , M. C. Adriaanse2 , K.vanDam1, M. M. Iversen3 , W. Viechtbauer4 , M.

Peyrot3,5, I. Caramlau6, A. Kokoszka7 , K. Kanc8, M.de Groot9 , G. Nefs10,11,12 and F.
Pouwer13,14,15 for the European Depression in Diabetes (EDID) Research Consortium
1
Middlesex University, London, UK , 2Vrije Universiteit, Amsterdam, The Netherlands , 3Western Norway University of Applied Sciences, Bergen, Norway ,
4
Maastricht University, Maastricht, The Netherlands , 5Loyola University Maryland, Baltimore, USA , 6Beaumont Hospital, Dublin, Ireland , 7Medical University of
Warsaw, Poland , 8Jazindiabetes (Diabetes & Me), Private Diabetes Centre, Ljubljana, Slovenia , 9Indiana University School of Medicine, Indianapolis, USA ,
10
Tilburg University, The Netherlands , 11Radboud University Medical Center, Nijmegen, The Netherlands , 12Diabeter, Rotterdam, The Netherlands , 13University
of Southern Denmark, Odense, Denmark , 14STENO Diabetes Center Odense, Odense, Denmark and 15School of Psychology, Deakin University, Geelong,
Australia

Accepted 17 June 2019

Abstract
To conduct a systematic review and meta-analysis of longitudinal studies assessing the bi-directional association between
depression and diabetes macrovascular and microvascular complications. Embase, Medline and PsycINFO databases
were searched from inception through 27 November 2017. A total of 4592 abstracts were screened for eligibility. Meta-
analyses used multilevel random/mixed-effects models. Quality was assessed using the Newcastle-Ottawa scale. Twenty-
two studies were included in the systematic review. Sixteen studies examined the relationship between baseline
depression and incident diabetes complications, of which nine studies involving over one million participants were
suitable for meta-analysis. Depression was associated with an increased risk of incident macrovascular (HR = 1.38; 95%
CI: 1.30–1.47) and microvascular disease (HR = 1.33; 95% CI: 1.25–1.41). Six studies examined the association
between baseline diabetes complications and subsequent depression, of which two studies involving over 230 000
participants were suitable for meta-analysis. The results showed that diabetes complications increased the risk of
incident depressive disorder (HR = 1.14; 95% CI: 1.07–1.21). The quality analysis showed increased risk of bias notably
in the representativeness of selected cohorts and ascertainment of exposure and outcome. Depression in people with
diabetes is associated with an increased risk of incident macrovascular and microvascular complications. The
relationship between depression and diabetes complications appears bi-directional. However, the risk of developing
diabetes complications in depressed people is higher than the risk of developing depression in people with diabetes
complications. The underlying mechanisms warrant further research.
Diabet. Med. 00: 1–11 (2019)

prevalence of diabetes complications [5,6]. More specifically,

Introduction
a meta-analysis of 27 studies focusing on the associations of
Diabetes is associated with long-term complications includ- depression with macrovascular complications (such as coro-
ing microvascular and macrovascular disease. Prospective nary artery disease), microvascular complications (diabetic
studies have also shown a significant bi-directional relation- retinopathy, neuropathy, nephropathy or end stage renal
ship between Type 2 diabetes and depression [1,2]. Meta- disease) and sexual dysfunction found a significant and
analytical evidence of longitudinal studies indicates that consistently positive relationship [7]. Effect sizes (correla-
diabetes increases the risk of developing depression by tions) were small to moderate, ranging from 0.17 to 0.32
approximately 25% [3,4] and that depression increases the (overall effect size of 0.25) and were similar across Type 1
risk of incident Type 2 diabetes by 40–60% [2,3]. and Type 2 diabetes study samples. However, all of the
Depression not only increases emotional suffering in included studies had a cross-sectional design, precluding the
people with diabetes, it is also associated with elevated identification of directions and pathways. For example,
similar to the longitudinal relationship between depression
Correspondence to: Arie Nouwen. E-mail: [email protected]

ª 2019 Diabetes UK 1
DIABETICMedicine Depression and diabetes complications  A. Nouwen et al.

titles and abstracts. Selected papers were retrieved and two

What’s new? other authors independently scrutinized the full text papers
• Numerous studies have examined the longitudinal for inclusion. Disagreements were resolved by discussion
relationship between depression and diabetes compli- between the authors and the first author (AN).
cations but comprehensive evidence about the magni- Articles meeting the following study criteria were included:
tude and direction is unavailable. (i) involving adults (>18 years old) at the time of follow-up
with Type 1 or Type 2 diabetes; (ii) published in peer-
• The current study shows that the relationship between reviewed, English language journals; and (iii) examining the
depression and diabetes complications appears bi- longitudinal relationship between depression and at least one
directional with depression being associated with an long-term complication of Type 1 or Type 2 diabetes.
increased risk of developing incident macrovascular and Diabetes complications included microvascular diseases
microvascular complications, and diabetes complica- (retinopathy, neuropathy, nephropathy, sexual dysfunction,
tions increasing the risk of subsequent depression. diabetic foot) and macrovascular diseases (angina, stroke,
• The increase in risk of developing diabetes complica- coronary artery disease, myocardial infarction, peripheral
tions in depressed people (by 38 and 33% for vascular disease). Studies solely focusing on amputations
macrovascular and microvascular complications, related to diabetes complications (n = 7) and studies
respectively) is higher than the increase in risk of examining the relationship between depression, diabetes
developing depression (by 9 and 24% for macrovascu- and cardiovascular mortality were not included as these
lar and microvascular complications, respectively) in have been the topic of recent systematic review/meta-
people with diabetes complications. analyses [9,10]. Moreover, amputations reflect treatment
decisions, which may confound their relationship with
depression. Studies were also excluded if they focused on
and diabetes, depression and diabetes complications may be gestational diabetes, impaired glucose tolerance or pre-
related either bi-directionally (depression may increase the diabetes. Studies examining the relationship of diabetes
risk of incident diabetes complications and vice versa) or complications with either lifetime or current depression were
only uni-directionally. included.
Since the publication of that meta-analysis, a number of
prospective population-based studies have examined the lon-
gitudinal relationships between depression and diabetes com- Data extraction and quality assessment
plications. However, comprehensive analysis of the magnitude One author (AN) performed data extraction, which was
and direction of these relationships is unavailable. Therefore, verified by two other authors (MCA, MMI). Data of interest
the aim was to examine the directional relationships between included: (i) name of first author; (ii) publication year; (iii)
depression and diabetes macrovascular and microvascular country; (iv) study design; (v) number of participants; (vi)
complications by conducting a systematic review of longitudi- mean age/age range; (vii) gender; (viii) diabetes type; (ix)
nal epidemiological studies. Where possible, meta-analyses diabetes duration; (x) follow-up length; (xi) depression
were carried out to quantify the sizes of the associations. assessment method; (xii) diabetes assessment method; (xiii)
diabetes complications assessment method; and (xiv) results
regarding associations between depression and diabetes
Methods
complication(s).
Method of depression assessment could be either a
Data sources and searches
diagnosis of depression assessed by a diagnostic psychiatric
This study is reported according to the Preferred Reporting interview or assessment of depressive symptoms using a self-
Items for Systematic Reviews and Meta-Analyses (PRISMA) report questionnaire. Type of diabetes and diabetes compli-
[8]. Literature searches were conducted using Embase, cations could be assessed either via self-reporting or
Medline and PsycInfo from inception through 27 November extracted from medical records. Diabetes complications
2017, including references of eligible articles. Databases were could be assessed via self-reporting, with the use of diagnos-
searched using Medical Subject Headings (MeSH) and tic tests, or extracted from medical records.
Boolean operators. The search terms (for Embase see the Quality assessment of included studies was conducted by
supporting information) were adapted to meet the require- two authors (AN, KvD) based on the Newcastle-Ottawa
ments of each database. scale [11] for cohort studies. Three domains were evaluated
for each study: selection of the study groups, comparability
of the groups and ascertainment of outcome of interest
Study selection
(Table 1). There were three levels of evidence: low risk of
The results of the searches were divided into batches of 500, bias (a score of 7–9), moderate risk of bias (5–6) and a high
and for each batch two authors independently screened the risk of bias (score < 5) [12].

2 ª 2019 Diabetes UK
 Table 1 Results of quality assessment of studies identified in the systematic review, based on the Newcastle-Ottawa scale

ª 2019 Diabetes UK
Selection Comparability Outcome

Controlled for
Ascertainment Demonstration minimum:
Selection of exposure that outcome of Complications/ Controlled Was follow-up
Representativeness of the (predictor - interest was not depression for long enough Adequacy of
of the exposed non-exposed depression/ present at start at baseline or anything Assessment for outcome follow up of Total
Systematic Review or Meta-analysis

cohort cohort complication) of study excluded else of outcome to occur? cohorts score

Acceptable a, b a, b a, b a a a a, b a a, b 9
Depression to diabetes complications
Black et al., 2003 [16] b a c b a a c a b 6
Clouse et al., 2003 [38] c a b b b a b a b 6
Orchard 2003 [23]† c a c a a a d a b 6
Roy et al, 2007a [29]* b a c a a a b a b 8
Lin et al., 2010 [17]‡ b a c b a a b a b 7
Scherrer et al., 2011 [18] c a a a a a b a – 7
Sullivan et al., 2012 [31] d a c b a a d b d 3
Ting et al., 2013 [32] c a a a a a b a b 8
Nefs et al., 2015 [19] b a c b a a b a d 6
Novak et al, 2016 [36] c a a a a a b a b 8
Roy et al., 2007b [30]* c a c a a a c a a 6
Gonzalez et al., 2010 [41]§ c a c a a a c a a 6
Iversen et al., 2015 [43] a a c a a a c a b 7
Jani et al., 2016 [45] b a c b b a b b – 4
Ismail et al., 2017 [35] b a c b b a b b b 5
Trento et al., 2017 [34] b a c b a a a a b 7
Diabetes complications to depression
Katon et al., 2009 [40]‡ b a a b a a c a b 7
Vileikyte et al., 2009 [39]§ c a c b a a c a b 5
Pan et al., 2012[42] b a a a a a a a d 8
Jacob and Kostev, 2016 [33] a a a a a a b a – 8
Bell et al., 2017 [44] b a c a a a c a b 7
Desch^enes et al., 2017 [37] a a c a a a c a c 6

*Hispanic Established Population for Epidemiologic Study of Elderly (EPESE) cohort

†
Epidemiological Diabetes Cohort (EDC) study
‡
Pathways cohort
§
Baltimore, State College and Manchester Hospital cohorts
–
Retrospective analyses
For a summary and scoring of the Newcastle-Ottawa scales, see the supporting information.

3
DIABETICMedicine
DIABETICMedicine Depression and diabetes complications  A. Nouwen et al.

The extracted data of the 30 studies are presented in

Data synthesis and analysis
Table S1A-C.
The meta-analysis used multilevel random/mixed-effects Where more than one study relied on the same cohort (six
models with random effects for studies and estimates within papers from the Epidemiological Diabetes Cohort study [20–
studies, which accounts for the dependency in the underlying 25] and four from the Pathways Epidemiological study
true effects for studies supplying multiple estimates (in some [17,26–28]), only the most recent or the most complete study
studies, the same participants were assessed for multiple [17,23] was selected for the review, resulting in 22 papers
macrovascular and microvascular outcomes). Although cor- being included in the systematic review. Of those, 16 papers
relation in the sampling errors is not automatically consid- (including 1025 563 people with diabetes) examined the
ered in this model, theoretical considerations and simulation relationship of depression to incident diabetes complications
studies [13] have shown that the multilevel model automat- (Table S1A,B) while 6 papers (including 239 519 people with
ically subsumes this source of dependency into the correla- diabetes) examined the relationship of diabetes complica-
tion between the true effects and therefore provides valid tions to incident depression (Table S1C). Data came
inferences about the fixed effects. As a sensitivity analysis, we predominantly from the USA but also from Canada, China,
also used cluster-robust inference methods, [13–15] which Italy, Norway, Taiwan, Germany, the UK and The Nether-
yielded identical conclusions. lands.
Criteria for meta-analysis were outcomes reported as Three of the 22 studies examined people with Type 1
proportional HRs and 95% CIs. For some studies, to obtain diabetes, [23,29,30] nine examined people with Type 2
the estimate of interest (e.g. reference group people with diabetes, [16–19,31–35] and 6 examined people with either
diabetes and no depression vs. people with diabetes and Type 1 or Type 2 diabetes [36–41]. Four studies did not
depression) we recalculated HRs using the procedures specify the type of diabetes [42–45].
described in Van Dooren et al [9]. For Black et al. [16] we Twelve studies quantified the results in terms of Cox
combined the HRs for the group of participants with diabetes proportional HRs, [16–19,31–33,36,38,41,42,45] while
without any depressive symptoms (Center for Epidemiologic others used ORs, [29,30,35,40,43,44] RR [37], or provided
Studies Depression Scale = 0) and those with diabetes with regression coefficients [39]. Two papers briefly explained
minimal depression (Center for Epidemiologic Studies that depression did not increase incident diabetes complica-
Depression Scale = 1–15) into the reference group. For Lin tions but neither provided numerical data [23,34]. Of the 12
et al. [17] we combined the groups with probable minor studies reporting Cox proportional HRs, 10 concerned
depression (Patient Health Questionnaire-9 algorithm) and depression as a risk factor for the onset of diabetes
no depression into the reference group and compared them complications and 2 concerned the relationship from dia-
with the group with probable major depression betes complications to incident depression.
(Patient Health Questionnaire-9 algorithm). For Scherrer
et al. [18] we used the information from the four-group
Depression as a risk factor for diabetes complications
scenario (neither diabetes nor major depressive disorder,
major depressive disorder alone, Type 2 diabetes alone, and Nine of the 10 studies reporting HRs and 95% CI were
co-morbid major depressive disorder and Type 2 diabetes) to suitable for meta-analysis; one study was not suitable for
obtain the HRs for the comparison of the two diabetes inclusion because a continuous depression measure was used
groups. For the meta-analyses, all outcome variables were rather than dichotomized depression [41]. Nine studies
dichotomized. allowed for a meta-analysis testing depression as a risk
factor for macrovascular disease. Four of the nine studies
provided data on depression as a risk factor for microvas-
Results
cular complications, which could be used in an additional
meta-analysis. Of the nine studies [16–19,31,32,36,38,45],
Search results
most used a composite macrovascular outcome including
Figure 1 shows the results of the selection process. The conditions such as myocardial infarction, coronary artery
searches yielded 7457 potential articles. After removing disease, congestive heart failure and stroke, but some also
conference papers, books, dissertation abstracts and dupli- included cardiovascular procedures (e.g. percutaneous coro-
cates within or between the databases, 4591 abstracts nary artery intervention, coronary artery bypass grafting,
remained. Twenty-nine eligible papers were selected, 24 abdominal aortic aneurysm repair and revascularization of
from Embase and 5 additional papers from Medline (no the lower extremity [16]) or unstable angina [31] to obtain a
additional papers from PsycInfo). One of the authors single estimate of macrovascular disease. However, one
identified an additional paper [19] not included in the search study [36] reported separate outcomes for stroke and coro-
results, bringing the total to 30 papers. Searches of the nary heart disease for the same sample, which were included
reference lists did not identify additional potential papers. individually.

4 ª 2019 Diabetes UK
 Systematic Review or Meta-analysis DIABETICMedicine

Records identified Records identified Records identified

through Embase through Medline through PsycInfo
(n = 4885) (n = 1629) (n = 943)

Identification
Conference papers,
Records identified through database
books, dissertations,
searching
duplicates removed
(n = 7457)
(n = 2866)

Additional records identified

through other sources
(n = 1)
Screening

Records screened Records excluded

(n = 4591) (n = 4493)

Full-text articles excluded with reasons

(n = 69)
Not focused on diabetes population (n = 23)
Not focused on diabetes complications (n =
10)
Eligibility

Full-text articles assessed for Outcome = recurrent foot ulcers (n = 2)

eligibility Outcome = cognitive decline/dementia (n = 4)
(n = 99) Outcome = mortality (n = 5)
Other outcomes (n = 7)
No comparison of complications vs. no
complications (n = 3)
Cross-sectional (n = 12)
Participants were adolescents (n = 1)
Complications to persistent depression (n = 2)

Studies identified
(n = 30) Studies from the same cohort
with duplicate information
(n = 8)
Studies included Epidemiology of Diabetes
(n = 22) Complication (EDC) (n = 5)
depression -> complications (n = 16) Pathways (n = 3)
Inclusion

complications -> depression (n = 6)

Studies included in the meta- Studies not included in the meta-

analyses (n = 11) analyses (n = 11)
Depression -> complications (n = 9) Depression -> complications (n = 7)
Complications -> depression (n = 2) Complications -> depression (n = 4)

FIGURE 1 Flowchart of study selection process.

Of the four studies also analysing microvascular compli- with the estimated overall HR based on all estimates
cations [16,17,31,36], three included a composite microvas- combined, is shown in Figure 2.
cular outcome [16,17,31], while Novak et al. [36] only used However, there was considerable heterogeneity between
incident chronic kidney disease as an outcome. studies [Cochran’s Q (13) = 92.30, P < 0.001; I2 = 86%].
To examine whether the HRs of developing complications in Standardized residuals and Cook’s distances showed that
people with diabetes and depression differed according to the Black et al. [16] was an influential study and the source of the
type of complication (macro/micro), we first carried out a heterogeneity in these data (primarily due to a much larger
moderator analysis. The results showed that the relationship effect size than other studies for microvascular disorders). This
between depression and complication development did not study was also the only one in the initial meta-analysis where
differ according to type of complication [v2(1) = 0.67, P = 0.41], complications were based on self-reporting; therefore, we
allowing us to analyse an overall measure of complications. repeated the above analyses without its inclusion, and regard
In this initial analysis, those with depression were on those secondary results as our study’s primary findings.
average 51% more likely to develop a complication than Moderator analysis of complication type was significant
those without depression (HR = 1.51; 95% CI: 1.23–1.86). A for these 11 studies [v2 (1) = 18.82, P < 0.001]; therefore,
forest plot of the HRs and 95% CI of each study, together results are reported separately for macrovascular and

ª 2019 Diabetes UK 5
DIABETICMedicine Depression and diabetes complications  A. Nouwen et al.

FIGURE 2 Forest plot of HRs of longitudinal studies included in the meta-analysis examining associations from baseline depression to incident
diabetes complications, grouped by macrovascular complications and microvascular complications. Abbreviations: CAD, coronary artery disease;
CHD, coronary heart disease; CVD, cerebrovascular disease; MI, myocardial infarction; PVD, peripheral vascular disease.

microvascular complications. Depression was associated was a linear relationship between increased risk and the
with an increased risk of macrovascular (HR = 1.38; 95% severity of depressive symptoms [43].
CI: 1.30–1.47) and microvascular (HR = 1.33; 95% CI:
1.25–1.41) complications (Figure 2). Sensitivity analyses
Diabetes complications as a risk factor for depression
using the cluster-robust inference approach yielded identical
conclusions in all of the analyses reported above. Six studies examined the hypothesis that diabetes complica-
A number of studies (n = 7) identified in the systematic tions increase the risk of incident depression
review reported incompatible statistics and could not be [33,37,39,40,42,44]. For the two largest of these studies
included in the meta-analysis (e.g. ORs instead of HRs; for (combined N = 234 628), nine HRs of diagnosed depression
full details, see Table S1). In terms of macrovascular were suitable for meta-analysis [33,42]. Therefore, we tested
complications, the results of two studies supported the whether any diabetes complication increased the risk of
findings of the meta-analysis [29,35]. Another study[23] incident depressive disorder. Results showed that diabetes
found a significantly increased risk only for angina (HR = complications increased the risk of incident depressive
1.40; 95% CI: 1.06–1.84), but not for the objective, disorder (HR = 1.14; 95% CI: 1.07–1.21). Analysis using
coronary artery disease (no statistics were provided for the the cluster-robust approach confirmed these results. Because
latter). In terms of microvascular complications, results were moderator analysis of complication type was significant [χ2
mixed, with three studies supporting [30,41,43] and two (1)=3.91, p = .048], we report the results of the meta‐analysis
failing to support the findings of the meta-analysis [34,35]. separately for macrovascular and microvascular complica-
Regarding incident foot ulcers, one study found that there tions. Both macrovascular complications (HR = 1.09; 95%

6 ª 2019 Diabetes UK
 Systematic Review or Meta-analysis DIABETICMedicine

CI: 1.02–1.17) and microvascular complications (HR=1.24;

Quality of evidence
95% CI: 1.12–1.37) were associated with an increased risk of
depressive disorder. For a Forest plot of these results, see Of the 22 included studies, 2 showed a high risk of bias, 8
Figure 3. showed a moderate risk of bias and 12 showed a low risk of
The remaining four studies [37,39,40,44] reported statis- bias (Table 1). Risk of bias was mainly due to the use of
tics that were incompatible with their inclusion in the meta- selected rather than representative cohorts (8 of 22 studies),
analysis (e.g. ORs or RRs instead of HRs; for full details, see use of self-reporting or failure to report the use of secure
Table S1C). One study[44] found that diabetes complications records or structured interviews to ascertain exposure (15 of
(not specified) increased the odds of developing depressive 22 studies) or outcome (9 of 22 studies), and failure to
symptoms (OR = 1.46; 95% CI 1.14–1.86). The results of the demonstrate that the outcome of interest (diabetes compli-
three studies that did examine specific diabetes complications cations/depression) was not present at baseline (10 of 22
mirrored those of the two studies in the meta-analyses. studies, although all but 3 studies [35,38,45] statistically
Regarding macrovascular complications, one study[37] controlled for this).
found an increased risk of depressive symptoms with
coronary artery disease, cerebral vascular disease and
Discussion
peripheral vascular disease, while another study[40] found
this to be the case for macrovascular events or procedures, This systematic review identified nine studies suitable for
but not for stroke. Regarding microvascular complications, meta-analysis which showed that in people with diabetes,
one study[40] found no significantly increased risk of those with co-morbid depression have a 38 and 33%
depressive symptoms with either retinopathy or nephropa- increased risk of developing macrovascular and microvascu-
thy. Two other studies [37,39] found an increased risk of lar complications, respectively, compared with controls
depressive symptoms with neuropathy, but not with without depression. Eight studies were identified as unsuit-
retinopathy or foot problems [37]. (For further details, see able for inclusion in the meta-analysis. For macrovascular
Table S1C). complications, the results of these studies generally

FIGURE 3 Forest plot of HRs of longitudinal studies included in the meta-analysis examining associations from baseline diabetes complications,
grouped by macrovascular and microvascular complications, to incident depression. Abbreviations: CHD, coronary heart disease; CVD,
cerebrovascular disease; MI, myocardial infarction; PVD, peripheral vascular disease.

ª 2019 Diabetes UK 7
DIABETICMedicine Depression and diabetes complications  A. Nouwen et al.

concurred with those obtained in the meta-analysis, except glycaemic control, which, in turn, is associated with
for one [23]. For microvascular complications, results were increased risk of complications[48]. Moreover, the UK
mixed, with three supporting [30,41,43] and two failing to Prospective Diabetes Study has shown that a 1% in reduction
support[34,35] the findings of the meta-analysis. Of the in HbA1c may delay the onset of diabetes complications by
latter, one examined participants during their first two years 21% [49]. However, research found no evidence that
of newly diagnosed diabetes [35], and the other had cognitive depression delays insulin initiation [50,51].
function as the main predictor variable [34]. More recently, biological processes associated with obe-
Finally, six identified studies examined the relationship sity, insulin resistance and persistently poor glucose control,
from diabetes complications to incident depression have been implicated in the development of diabetes com-
[33,37,39,40,42,44]. In the meta-analysis of the two largest plications in the context of depression. More specifically,
studies [33,42] (total N = 234 628), diabetes complications inflammatory processes associated with both diabetes and
increased the risk of developing depressive disorder by 14%. depression [52] have been identified as possible mechanisms
Although this increase in risk was found to be higher for [53]. Other studies have shown that depression and diabetes
microvascular than for macrovascular complications (by 24 are associated with endothelial function, increasing the risk
vs. 9%, respectively), because of the small number of studies of macrovascular disease [54], although the underlying
these results should be interpreted with caution. The results of mechanisms have yet to be elucidated.
the six studies not included in the meta-analyses (Table S1C) The mechanisms underlying the increased risk of incident
mimic those of the meta-analysis and suggest that having depression as a result of diabetes complications are likely to
macrovascular and microvascular diabetes complications may be similar to those hypothesized for diabetes as a risk for
increase the risk of developing depression. However, across depression onset. Although inflammation and other biolog-
the six studies examining the relationship of diabetes compli- ical factors may play a role [52,54], the epidemiological
cation as a risk factor for depression, we did not find evidence evidence points to the role of diabetes burden and distress as
of a systematic effect of specific complications. contributing factors [55,56]. The results of our study,
Taken together, our results provide support for a bi- showing no systematic effect of specific diabetes complica-
directional relationship between diabetes complications and tions, is consistent with the latter. It should be noted that
depression. In other words, depression (both depressive only one study[39] examining the relationship of complica-
symptoms and major depressive disorder) increases the risk tions with incident depression controlled for pain as a
of incident macrovascular and microvascular diabetes com- possible contributor.
plications, and the presence of diabetes complications Overall, the results of the current study extend the findings
increases the risk of significant depressive symptoms and/or of a previous systematic review/meta-analysis of cross-
possible depressive episode. However, the increase in risk of sectional studies [7] and shows that depression is associated
developing diabetes complications (by 38 and 33% for with an increased risk of diabetes complications. Impor-
macrovascular and microvascular complications, respec- tantly, given that depression is a treatable condition, the
tively) in depressed people is higher than the increase in risk clinical implication of this study is that depression could be a
of developing depression (by 9 and 24% for macrovascular preventable risk factor for diabetes complications. Future
and microvascular complications, respectively) in people studies are needed to examine whether intensive and
with diabetes complications. These findings parallel those successful treatment of depression in diabetes does indeed
found for the bi-directional longitudinal relationships reduce the risk of future complications [57].
between diabetes and depression with a modest increased The current study also adds to an increasing number of
relative risk (15%) of developing depression in people with longitudinal studies showing that depression in diabetes,
diabetes and a higher relative risk (60%) for the development even when using self-report questionnaires, can have serious
of diabetes in depressed people [2]. consequences, including reduced self-care [39], increased
These findings may reflect different underlying mecha- cognitive decline [58], dementia [59] and mortality [9].
nisms for the relationship from depression to diabetes or A number of limitations should be taken into consider-
diabetes complications than for the relationship from ation when interpreting the results of this study. First,
diabetes or diabetes complications to depression. While the despite the large number of participants upon which the
exact nature of these underlying mechanisms remains meta-analyses were based, the relatively small number of
unknown, there are a number of common processes that studies available (n = 9 for the relationship from depression
may result in these negative outcomes. First, depression is to diabetes complications and n = 2 for the relationship
often accompanied by behavioural changes, including from diabetes complications to depression) precluded exam-
reduced self-care and medication adherence [46], increased ination of the contribution of factors such as specific
smoking, reduced physical activity and increased sedentary diabetes complications, type of depression measurement,
behaviours, and increased intake of high-calorie food [47]. and type of diabetes. Moreover, although some studies
These behaviours may be particularly problematic in the included participants with Type 1 diabetes with paediatric
context of diabetes and are likely to adversely affect onset, teasing out the effects of depression on diabetes

8 ª 2019 Diabetes UK
 Systematic Review or Meta-analysis DIABETICMedicine

complications in Type 1 vs. Type 2 diabetes remains a topic

Competing interests
for future research.
Second, the quality assessment of the included studies No potential or existing conflicts of interest relevant to this
found a high risk of bias in two studies due to, among other article were reported.
factors, the use of self-report measures to ascertain exposure
or outcome. Self-reporting should be avoided when objective
REFERENCES
diagnostic tests are available, such as in the case of diabetes
complications. However, as there are no known biomarkers 1 Golden SH. Examining a bidirectional association between depres-
for depression, its assessment remains based on symptoma- sive symptoms and diabetes. JAMA 2008; 299: 2751–2759.
2 Mezuk B, Eaton WW, Albrecht S, Golden SH. Depression and type
tology; although structured diagnostic interviews are con- 2 diabetes over the lifespan: A meta-analysis. Diabetes Care 2008;
sidered the gold standard for identifying depression, self- 31: 2383–2390.
report symptom questionnaires are commonly used. While 3 Rotella F, Mannucci E. Depression as a risk factor for diabetes: a
the use of such questionnaires may be regarded as introduc- meta-analysis of longitudinal studies. J Clin Psychiatry 2013; 74:
ing bias, research suggests that symptom questionnaires and 31–37.
4 Nouwen A, Winkley K, Twisk J, Lloyd CE. Type 2 diabetes
diagnostic interviews assess different constructs, with ele- mellitus as a risk factor for the onset of depression: a
vated depressive symptoms more reflective of general and systematic review and meta-analysis. Diabetologia 2010; 53:
diabetes-specific emotional distress than of clinical depres- 2480–2486.
sion [60]. Furthermore, although most studies provided 5 Lustman PJ, Clouse RE. Depression in diabetic patients: the
follow-up data, only a few studies provided information as relationship between mood and glycemic control. J Diabetes
Complications 2005; 19: 113–122.
to how many people had died from diabetes complications 6 van Steenbergen-Weijenburg KM, van Puffelen AL, Horn EK,
during the follow-up period. Because mortality is higher in Nuyen J, van Dam PS, Beekman AT et al. More co-morbid
people with depression, it is likely that the results from the depression in patients with Type 2 diabetes with multiple compli-
meta-analysis are conservative. Third, the included studies cations. An observational study at a specialized outpatient clinic.
did not provide information regarding the specific set of Diabet Med 2011; 28: 86–89.
7 de Groot M, Anderson R, Freedland KE, Clouse RE, Lustman PJ.
complications individuals developed during the study. It is Association of depression and diabetes complications: a meta-
therefore impossible to ascertain whether depression leads to analysis. Psychosom Med 2001; 63: 619–630.
the development of only single complications or a cluster of 8 Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting
complications. Furthermore, some of the studies in the meta- items for systematic reviews and meta-analyses: the PRISMA
analysis[17,31,32,36] included not only diabetes complica- statement. PLoS Med 2009; 6: e1000097.
9 Van Dooren FEP, Nefs G, Schram MT, Verhey FRJ, Denollet J,
tions but also surgical interventions, which may or may not Pouwer F. Depression and risk of mortality in people with diabetes
have resulted from the onset or worsening of diabetes mellitus: a systematic review and meta-analysis. PLoS One 2013; 8:
complications and may have confounded the observed e57058.
relationships. Fourth, it is unclear how clinical management, 10 O’Neill K, McNamara B. Comorbid depression and risk of lower
access to medical care, disease management supplies and extremity amputation in people with diabetes: systematic review
and meta-analysis. BMJ Open Diabetes Res Care 2017; 5:
medications, financial burden of diabetes, and other socio- e000366.
economic factors play into the development of depression or 11 Wells GA, Shea B, O’Connell D, Peterson J, Welch V, Losos M
diabetes complications and the relationship between the et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality
two. of nonrandomized studies in meta-analyses. Ottawa Hosp Res Inst
In summary, depression in people with diabetes is associ- 2013; 3: 1–4.
12 McPheeters ML, Kripalani S, Peterson NB, Idowu RT, Jerome RN,
ated with an increased risk of incident macrovascular and Potter SA et al. Closing the quality gap: revisiting the state of the
microvascular complications. This relationship is likely to be science (vol. 3: quality improvement interventions to address health
bi-directional as diabetes complications were also found to disparities). Evid Rep Technol Assess (Full Rep) 2012; 208(3): 1–
increase the risk of incident depression. It remains to be 475.
determined whether effective treatment of depression reduces 13 Moeyaert M, Ugille M, Beretvas SN, Ferron J, Bunuan R, Van den
Noortgate W. Methods for dealing with multiple outcomes in
the risk of developing diabetes complications. meta-analysis: a comparison between averaging effect sizes, robust
variance estimation and multilevel meta-analysis. International J
Acknowledgements Soc Res Methodol 2017; 20: 559–572.
14 Van den Noortgate W, Lopez-Lopez JA, Marin-Martinez F,
The authors thank Vicky Wallace of the Barnes Library, Sanchez-Meca J. Meta-analysis of multiple outcomes: a multilevel
University of Birmingham, UK, for help with the literature approach. Behav Res Methods 2015; 47: 1274–1294.
15 Hedges LV, Tipton E, Johnson MC. Robust variance estimation in
searches.
meta-regression with dependent effect size estimates. Res Synth
Methods 2010; 1: 39–65.
16 Black S, Markides K, Ray L. Depression predicts increased
Funding sources
incidence of adverse health outcomes in older Mexican Americans
There was no funding source for this study. with Type 2 diabetes. Diabetes Care 2003; 26: 2822–2828.

ª 2019 Diabetes UK 9
DIABETICMedicine Depression and diabetes complications  A. Nouwen et al.

17 Lin EH, Rutter CM, Katon W, Heckbert SR, Ciechanowski P, progression in patients with type 2 diabetes. Eur J Ophthalmol
Oliver MM et al. Depression and advanced complications of 2017; 27: 278–280.
diabetes: a prospective cohort study. Diabetes Care 2010; 33: 264– 35 Ismail K, Moulton CD, Winkley K, Pickup JC, Thomas SM,
269. Sherwood RA et al. The association of depressive symptoms and
18 Scherrer JF, Garfield LD, Chrusciel T, Hauptman PJ, Carney RM, diabetes distress with glycaemic control and diabetes complications
Freedland KE et al. Increased risk of myocardial infarction in over 2 years in newly diagnosed type 2 diabetes: a prospective
depressed patients with type 2 diabetes. Diabetes Care 2011; 34: cohort study. Diabetologia 2017; 60: 2092–2102.
1729–1734. 36 Novak M, Mucsi I, Rhee CM, Streja E, Lu JL, Kalantar-Zadeh K
19 Nefs G, Pop VJM, Denollet J. Depressive symptom clusters et al. Increased risk of incident chronic kidney disease, cardiovas-
differentially predict cardiovascular hospitalization in people with cular disease, and mortality in patients with diabetes with
type 2 diabetes. Psychosomatics 2015; 56: 662–673. comorbid depression. Diabetes Care 2016; 39: 1940–1947.
20 Lloyd CE, Kuller LH, Ellis D, Becker DJ. Wing RR, Orchard TJ. 37 Desch^enes SS, Burns RJ, Pouwer F, Schmitz N. Diabetes compli-
Coronary artery disease in IDDM. Gender differences in risk cations and depressive symptoms. Psychosom Med 2017; 79: 603–
factors but not risk. Arterioscler Thromb Vasc Biol 1996; 16: 720– 612.
726. 38 Clouse R, Lustman P, Freedland K, Griffith L, McGill J, Carney R.
21 Fickley CE, Lloyd CE, Costacou T, Miller RG, Orchard TJ. Type A Depression and coronary heart disease in women with diabetes.
behavior and risk of all-cause mortality, CAD, and CAD-related Psychosom Med 2003; 65: 376–383.
mortality in a type 1 diabetic population: 22 years of follow-up in 39 Vileikyte L, Peyrot M, Gonzalez JS, Rubin RR, Garrow AP,
the Pittsburgh epidemiology of diabetes complications study. Stickings D et al. Predictors of depressive symptoms in persons
Diabetes Care 2013; 36: 2974–2980. with diabetic peripheral neuropathy: a longitudinal study. Dia-
22 Orchard TJ, Virella G, Forrest KY-Z, Evans RW, Becker DJ. betologia 2009; 52: 1265–1273.
Lopes-Virella MF. Antibodies to oxidized LDL predict coronary 40 Katon WM, Russo JP, Lin EH, Heckbert SR, Ciechanowski P,
artery disease in type 1 diabetes a nested case-control study from Ludman EJ et al. Depression and diabetes: factors associated with
the Pittsburgh Epidemiology of Diabetes Complications Study. major depression at five-year follow-up. Psychosomatics 2009; 50:
Diabetes 1999; 48: 1454–1458. 570–579.
23 Orchard TJ, Olson JC, Erbey JR, Williams K, Forrest KY, 41 Gonzalez JS, Vileikyte L, Ulbrecht JS, Rubin RR, Garrow AP,
Smithline Kinder L et al. Insulin resistance-related factors, but Delgado C et al. Depression predicts first but not recurrent diabetic
not glycemia, predict coronary artery disease in type 1 diabetes: 10- foot ulcers. Diabetologia 2010; 53: 2241–2248.
year follow-up data from the Pittsburgh epidemiology of diabetes 42 Pan HH, Li CY, Chen PC, Lee MD, Liang CY, Hou WH et al.
complications study. Diabetes Care 2003; 26: 1374–1379. Contributions of diabetic macro-vascular complications and hip
24 Forrest KY, Becker DJ, Kuller LH, Wolfson SK, Orchard TJ. Are fracture to depression onset in elderly patients with diabetes: an 8-
predictors of coronary heart disease and lower-extremity arterial year population-based follow-up study. J Psychosom Res 2012; 73:
disease in type 1 diabetes the same? A prospective study. 180–184.
Atherosclerosis 2000; 148: 159–169. 43 Iversen MM, Tell GS, Espehaug B, Midthjell K, Graue M, Rokne B
25 Kinder LS, Kamarck TW, Baum A, Orchard TJ. Depressive et al. Is depression a risk factor for diabetic foot ulcers? 11-year
symptomatology and coronary heart disease in Type I diabetes follow-up of the Nord-Trøndelag Health Study (HUNT). J
mellitus: a study of possible mechanisms. Health Psychol 2002; 21: Diabetes Complications 2015; 29: 20–25.
542–552. 44 Bell T, Davila AL, Clay O, Markides KS, Andel R, Crowe M. The
26 Sieu N, Katon W, Lin EHB, Russo J, Ludman E, Ciechanowski P. association between cognitive decline and incident depressive
Depression and incident diabetic retinopathy: a prospective cohort symptoms in a sample of older Puerto Rican adults with diabetes.
study. Gen Hosp Psychiatry 2011; 33: 429–435. Int Psychogeriatrics 2017; 29: 1317–1325.
27 Williams LH, Rutter CM, Katon WJ, Reiber GE, Ciechanowski P, 45 Jani BD, Cavanagh J, Barry SJE, Der G, Sattar N, Mair FS.
Heckbert SR et al. Depression and incident diabetic foot ulcers: a Relationship between blood pressure values, depressive symptoms,
prospective cohort study. Am J Med 2010; 123: 748–754. and cardiovascular outcomes in patients with cardiometabolic
28 Yu MK, Weiss NS, Ding X, Katon WJ, Zhou XH, Young BA. disease. J Clin Hypertens 2016; 18: 1027–1035.
Associations between depressive symptoms and incident ESRD in a 46 Gonzalez JS, Peyrot M, McCarl LA, Collins EM, Serpa L, Mimiaga
diabetic cohort. Clin J Am Soc Nephrol. 2014; 9: 920–928. MJ et al. Depression and diabetes treatment nonadherence: a meta-
29 Roy MS, Peng B, Roy A. Risk factors for coronary disease and analysis. Diabetes Care 2008; 31: 2398–2403.
stroke in previously hospitalized African-Americans with Type 1 47 Penninx BWJH. Depression and cardiovascular disease: epidemi-
diabetes: A 6-year follow-up. Diabet Med 2007; 24: 1361–1368. ological evidence on their linking mechanisms. Neurosci Biobehav
30 Roy MS, Roy A, Affouf M. Depression is a risk factor for poor Rev 2017; 74: 277–286.
glycemic control and retinopathy in African-Americans with type 1 48 Zoungas S, Chalmers J, Ninomiya T, Li Q, Cooper ME, Colagiuri
diabetes. Psychosom Med 2007; 69: 537–542. S et al. Association of HbA1c levels with vascular complications
31 Sullivan M, O’Connor P, Feeney P, Hire D, Simmons DL, Raisch and death in patients with type 2 diabetes: evidence of glycaemic
DW et al. Depression predicts all-cause mortality. Diabetes Care thresholds. Diabetologia 2012; 55: 636–643.
2012; 35: 1708–1715. 49 Stratton IM. Association of glycaemia with macrovascular and
32 Ting RZW, Lau ESH, Ozaki R, Lau WW, Kong AP, Luk AO et al. microvascular complications of type 2 diabetes (UKPDS 35):
High risk for cardiovascular disease in Chinese type 2 diabetic prospective observational study. BMJ 2000; 321: 405–412.
patients with major depression - A 7-year prospective analysis of the 50 Nefs G. The longitudinal association between depressive symptoms
Hong Kong Diabetes Registry. J Affect Disord 2013; 149: 129–135. and initiation of insulin therapy in people with type 2 diabetes in
33 Jacob L, Kostev K. Prevalence of depression in type 2 diabetes primary care. PLoS One 2013; 8: e78865.
patients in German primary care practices. J Diabetes Complica- 51 Iversen MM, Nefs G, Tell GS, Espehaug B, Midthjell K, Graue M
tions 2016; 30: 432–437. et al. Anxiety, depression and timing of insulin treatment among
34 Trento M, Charrier L, Salassa M, Merlo S, Passera P, Baltatescu A people with type 2 diabetes: nine-year follow-up of the Nord-
et al. Cognitive function may be a predictor of retinopathy Trøndelag Health. J Psychosom Res 2015; 79: 309–315.

10 ª 2019 Diabetes UK
 Systematic Review or Meta-analysis DIABETICMedicine

52 Laake JP, Stahl D, Amiel SA, Petrak F, Sherwood RA, Pickup JC 59 Katon W, Pedersen HS, Ribe AR, Fenger-Grøn M, Davydow D,
et al. The association between depressive symptoms and systemic Waldorff FB et al. Effect of depression and diabetes mellitus on the
inflammation in people with type 2 diabetes: findings from the risk for dementia. JAMA Psychiatry 2015; 72: 612–619.
South London Diabetes Study. Diabetes Care 2014; 37: 2186– 60 Fisher L, Skaff MM, Mullan JT, Arean P, Mohr D, Masharani U
2192. et al. Clinical depression versus distress among patients with type 2
53 Fossati P, Radtchenko A, Boyer P. Neuroplasticity: from MRI to diabetes: not just a question of semantics. Diabetes Care 2007; 30:
depressive symptoms. Eur Neuropsychopharmacol 2004; 14: 542–548.
S503–S510.
54 van Agtmaal MJM, Houben AJHM, Pouwer F, Stehouwer CDA,
Schram MT. Association of microvascular dysfunction with late- Supporting Information
life depression. JAMA Psychiatry 2017; 74: 729–739.
Additional supporting information may be found online in
55 Demakakos P, Zaninotto P, Nouwen A. Is the association between
depressive symptoms and glucose metabolism bidirectional? Evi- the Supporting Information section at the end of the article.
dence from the English Longitudinal Study of Ageing. Psychosom
Text S1. Search criteria Embase.
Med 2014; 76: 555–561.
56 Nouwen A, Nefs G, Caramlau I, Connock M, Winkley K, Lloyd Text S2. Summary and scoring of the Newcastle-Ottawa
CE et al. Prevalence of depression in individuals with impaired scale.
glucose metabolism or undiagnosed diabetes: a systematic review
and meta-analysis of the European Depression in Diabetes (EDID) Table S1A. Depression to macrovascular complications
research consortium. Diabetes Care 2011; 34: 752–762.
(overview of all included studies sorted by publication date
57 Baumeister H, Hutter N, Bengel J. Psychological and pharmaco-
logical interventions for depression in patients with diabetes in ascending order).
mellitus: an abridged Cochrane review. Diabet Med 2014; 31: Table S1B. Depression to microvascular complications
773–786. (overview of all included studies sorted by publication date
58 Demakakos P, Muniz-Terrera G, Nouwen A. Type 2 diabetes, in ascending order).
depressive symptoms and trajectories of cognitive decline in a
Table S1C. Complications to depression (overview of all
national sample of community-dwellers: a prospective cohort
study. PLoS One 2017; 12: e0175827. included studies sorted by publication date in ascending order).

ª 2019 Diabetes UK 11