Review
Depression and type 2 diabetes: a causal association? Adam G Tabák, Tasnime N Akbaraly, G David Batty, Mika Kivimäki Lancet Diabetes Endocrinol 2014; 2: 236–45 Published Online November 14, 2013 http://dx.doi.org/10.1016/ S2213-8587(13)70139-6 Department of Epidemiology and Public Health, University College London, London, UK (A G Tabák MD, T N Akbaraly PhD, G D Batty PhD, Prof M Kivimäki PhD); Semmelweis University Faculty of Medicine, 1st Department of Medicine, Budapest, Hungary (A G Tabák); Inserm U 1061, Montpellier, France (T N Akbaraly); University Montpellier I, Montpellier, France (T N Akbaraly); and Centre for Cognitive Ageing and Cognitive Epidemiology, University of Edinburgh, Edinburgh, UK (G D Batty) Correspondence to: Dr Adam G Tabák, Department of Epidemiology and Public Health, University College London, London WC1E 6BT, UK [email protected]
There is a controversy regarding whether depression and type 2 diabetes are causally linked. To assess this issue, we review key findings for the association between depression and diabetes, and its underlying mechanisms. Findings from meta-analyses of cohort studies show a modestly sized bidirectional association between depression and type 2 diabetes (ie, depression predicts diabetes onset and diabetes predicts future depression). However, depression-related biological alterations in the hypothalamic-pituitary-adrenal cortex axis and the sympathetic nervous system, and subclinical inflammation, are not consistently linked with increased diabetes risk. The evidence for an association between depression and glycaemic traits (eg, glucose, insulin, insulin sensitivity, and insulin secretion) is also mixed. Diabetes increases the risk of depression to the same extent as do other chronic disorders (eg, cardiac diseases, osteoarthritis, lung disease, and poor hearing). At present, the available evidence suggests that pathophysiological changes preceding the onset of type 2 diabetes might not cause depression, nor might depression directly increase the risk of developing type 2 diabetes. Despite insufficient robust causal evidence, treating physicians should be aware of the co-occurrence of depression and type 2 diabetes.
Introduction As a result of their increasing prevalence, type 2 diabetes and depression are major public health issues. Worldwide, more than 365 million people are estimated to have type 2 diabetes,1 and almost 300 million people have major depression.2 Both these disorders are projected to be among the five leading causes of disease burden by 2030.3 The co-occurrence of type 2 diabetes and depression has attracted much research interest. If this association was causal, it would have profound implications for prevention and treatment of these disorders. However, the association between diabetes and depression seems to be complex, and does not follow a simple cause-and-effect pattern. In this Review we describe the conceptual models central to this association, provide an up-to-date review and appraisal of the evidence, and assess the relevance of preventive and clinical guidelines for type 2 diabetes and depression in view of research advances. Figure 1 shows some of the physiological, biological, and behavioural factors that are potentially connected with both type 2 diabetes and depression and might be implicated in the relation between these diseases.
Conceptual models for the association between depression and type 2 diabetes Findings from a series of prospective cohort studies and meta-analyses suggest that depression is a risk factor for type 2 diabetes.4 However, the evidence is also compatible with two alternative explanations. One notion is that depression and diabetes share common antecedents, resulting in a non-causal association between depression and type 2 diabetes. Studies have attempted to rule out this explanation (the so-called common-soil hypothesis) by statistical analyses in which adjustment is made for potential common causes or confounding factors, such as physical inactivity and obesity. However, the association between depression and diabetes might still be accounted for by residual confounding as a result of unmeasured or poorly measured common-cause risk 236
factors. The second notion posits that preclinical and clinical type 2 diabetes might increase the likelihood of participants reporting depressive symptoms, consistent with a reverse causation explanation. The three conceptual models—causality, the common-soil hypothesis, and reverse causality—might simultaneously contribute to the association between depression and type 2 diabetes (panel). Some investigators refer to a bidirectional association, combining the first and third models (ie, that depression could increase the risk of diabetes and that type 2 diabetes, in turn, could increase the risk of depression). A similar bidirectional association is suggested to exist between type 1 diabetes and depression.5 In this Review we organise the key evidence according to these three notions.
Depression as a risk factor for the development of type 2 diabetes Possible pathophysiological links In the multistage model of diabetes development,6 pathological changes preceding type 2 diabetes include longstanding insulin resistance that is accompanied by a compensatory increase in insulin secretion.7 This stage is typically followed by a stable adaptation period in which increased insulin resistance is not fully compensated by insulin secretion, leading to increased glucose concentrations. In the third stage of diabetes development— the unstable early decompensation period—β-cell function further worsens and compensation for insulin resistance becomes inadequate, leading to rapidly increasing glucose concentrations and manifest diabetes.6 Some aspects of the physiological changes associated with depression overlap with those characterising diabetes causation. Depression is associated with increased activity of the hypothalamic-pituitary-adrenal axis and the sympathetic nervous system,8,9 resulting in increased release of cortisol and other glucocorticoids, catecholamines (eg, adrenaline and noradrenaline), growth hormone, and glucagon. Cortisol stimulates glucose production and increases lipolysis, resulting in www.thelancet.com/diabetes-endocrinology Vol 2 March 2014
Review
decreased insulin sensitivity and hampering insulin secretion from β cells. Indeed, it is postulated that chronically high cortisol concentrations (as reported in about 50% of patients with depression) might result in obesity, insulin resistance, and type 2 diabetes.8 Furthermore, the inflammatory hypothesis of depression proposes a bidirectional association between activation of various aspects of the immune system and depression.10 The role of inflammatory markers and mechanisms are also well known in the pathophysiology of type 2 diabetes, providing a potential additional link between these diseases.11
Health behaviours Depression
Diabetes diagnosis
Vascular complications
www.thelancet.com/diabetes-endocrinology Vol 2 March 2014
Physical activity
Biological mechanisms Obesity Sympathetic nervous system↑ Hypothalamicpituitary-adrenal axis↑
Effects on glycaemic traits If depression is a causal factor for the development of type 2 diabetes, people with depression would be at an increased risk of prediabetes (defined with use of fasting glucose, HbA1c concentration, or an oral glucosetolerance test).12 However, most evidence from crosssectional studies suggests that prediabetes (defined as impaired fasting glucose according to the American Diabetes Association,13,14 or impaired fasting glucose and impaired glucose tolerance according to WHO) or undiagnosed diabetes is not associated with an increased risk of depression.15,16 Findings from the Hoorn study17 suggested a sex interaction, with an increased risk of impaired glucose tolerance in women with depression but not in men. Additionally, a dose-response relation was reported in US army veterans and south-Asian people at risk of depression, increasing stepwise from normoglycaemia to impaired fasting glucose to hyperglycaemia.18,19 However, longitudinal evidence from a Swedish population-based study20 suggested a different sex interaction, with a doubling of the risk of prediabetes in men with depression but no effect in women during 8–10 years of follow-up.20 A further suggestion of a link between diabetes and depression is from reports of a positive association between continuous measures of depression (eg, severity scores) and glycaemia (eg, fasting glucose, 2 h post-load glucose, or HbA1c concentration). However, findings from at least one study21 showed increased scores for depressive symptoms in people with very low and high glycaemic measures after adjustment for several potential confounders, contradicting the notion of a direct link between depression and dysglycaemia. Several studies have assessed depressive symptoms and validated indicators of insulin resistance and insulin secretion in the general population,22,23 but these results are mixed. Some study findings show no association between depression and insulin resistance,24,25 whereas others suggest increased insulin resistance in people with depression as a result of obesity and abnormal concentrations of adipokines.26,27 Furthermore, findings from at least two studies using the gold-standard measure of glycaemic traits (frequently sampled intravenous
Disease-related factors
CNS↑ Subclinical inflammation
Drug therapy Diet
Self-management
Type 2 diabetes
Figure 1: Potential pathophysiological and behavioural mechanisms linking depression and type 2 diabetes CNS=central nervous system.
Panel: Plausible mechanisms underlying the association between depression and type 2 diabetes Depression as a cause of type 2 diabetes • Depression might induce biological changes that increase the risk of type 2 diabetes • Depression might affect diabetes risk indirectly by increasing health-risk behaviours • Antidepressant drugs might have diabetogenic side-effects • Depression might worsen adherence to dietary and exercise regimens Type 2 diabetes as a cause of depression • Preclinical changes (eg, increased blood glucose or altered insulin concentrations) might directly affect depression risk through biochemical mechanisms • Diagnosis of chronic disease, such as type 2 diabetes, is a negative life event increasing the risk of depression • The burden of dealing with diabetes complications might lead to symptoms of depression Non-causal association • The so-called common-soil hypothesis posits that the association between depression and diabetes results from factors affecting both disorders—eg, chronic inflammation, obesity, physical inactivity, and vascular risk
glucose-tolerance test) showed increased insulin resistance irrespective of other factors.28,29 Nonetheless, analyses from the large-scale US National Health and Nutrition Examination Survey30 did not reproduce these 237
Review
findings in young adults. The analyses showed a statistically significant sex interaction, with no association between depression and insulin resistance in women, but in men a paradoxical inverse association between depression and homoeostasis model assessment (HOMA) insulin resistance. The investigators suggested that, in older people, unmeasured comorbidities rather than depression per se could bring about increased insulin resistance. This finding has not been confirmed by other studies. Findings from a small study31 of children at risk of obesity showed that depressive symptoms predicted insulin resistance 6 years later, even after the effect of baseline risk factors and changes in BMI were taken into account. By contrast, findings from a large multiethnic study (the Study of Women’s Health Across the Nation32) showed that depression was related to higher HOMA insulin resistance, although this association was eliminated after adjustment for baseline central obesity. Furthermore, depression was unrelated to changes in insulin resistance, suggesting that central obesity could be a common underlying factor linking depression with insulin resistance. Insulin secretion was investigated in at least three studies but with opposing findings: acute insulin response was increased in two studies28,29 but decreased in a third,25 probably as a compensatory response to decreased insulin sensitivity.
Off-target effects of antidepressant treatment A potential confounding factor that can distort the association between depression and type 2 diabetes is antidepressant medication. This confounding would happen if effective treatment for depression also had a favourable effect on glycaemic traits. Findings from an observational study33 comparing glycaemic traits before and after drug therapy in people with depression, and at least one randomised placebo-controlled trial of an antidepressant,34 showed no effect of these drugs on insulin resistance despite improvements in depressive symptoms. Findings from two further studies, however, showed improved insulin sensitivity in patients with depression receiving drug therapy without any changes in obesity measures.28,35 Insulin secretion has been less frequently investigated than has insulin resistance, and the results shown by these studies might be specific to the drug used36 or compensatory to changes in insulin secretion;25,28 however, these findings do not indicate that a major bias by drug therapy is a plausible explanation for the association between depression and diabetes. Findings from several studies of patients with recorded diagnosed type 2 diabetes as the outcome suggest that antidepressant drugs could actually increase, rather than decrease, risk of type 2 diabetes. Long-term use of at least moderate doses of antidepressants (eg, tricyclic antidepressants and selective serotonin reuptake inhibitors) was associated with an increased risk of type 2 diabetes.37–39 This finding should be interpreted cautiously, because 238
the scarcity of evidence for specificity by drug molecule or drug group effect raises the possibility that the observed association could be an artifact rather than the result of a causal relation. This alternative explanation is supported by analysis of data from repeated screenings in the Whitehall II study,40 which showed that although incidence of recorded diagnosed diabetes was higher in antidepressant users than in non-users, antidepressant use was not associated with undiagnosed diabetes at follow-up, or with increased fasting, increased 2 h postload plasma glucose concentrations, or increasing glucose concentrations over time. Thus, antidepressant use seems not to increase glucose concentrations, but is associated with an increased likelihood of being diagnosed with diabetes. Because many people with diabetes and depression are not diagnosed, detection bias is one potential explanation for such a non-causal association. Patients treated for one disorder might also be likely to be diagnosed with another disorder because of their more frequent contact with medical care.40,41
Effects on disease prognosis Depression has been postulated to have a role at later stages of the disease process, affecting the prognosis of type 2 diabetes. In support of this notion, patients with diabetes and persistent or worsening depressive symptoms have worse adherence to dietary and exercise regimens than do non-depressed patients with diabetes.42 Depression could be associated with a worsening of metabolic control, poor adherence to drugs, and increased health-care expenditures.43 A longitudinal modelling analysis44 lent support to a conceptual model in which the effect of depression on glucose dysregulation in diabetic patients with comorbid depression was mediated through lower adherence to self-care.
Diabetes as a risk factor for depression Standard risk factors for depression include having biological relatives with depression, traumatic experiences as a child (eg, sexual abuse), stressful life events (eg, the death of a loved one), alcohol and drug misuse, and having a serious somatic illness. Findings from meta-analyses suggest that the association between depression and type 2 diabetes might be bidirectional— that is, in addition to depression predicting the development of type 2 diabetes, type 2 diabetes might predict the development of depression, with the strength of each association being modest.45,46 This finding, and the fact that indicators of prediabetes (such as metabolic syndrome) predict the onset of depression,47 highlight the importance of considering diabetes and prediabetes as risk factors for depression.
Depressogenic effects of prediabetes Findings from a meta-analysis48 of cross-sectional population-based studies published up to 2011 suggest that there might be a weak positive association between www.thelancet.com/diabetes-endocrinology Vol 2 March 2014
Review
insulin resistance and depressive symptoms in adults. However, the investigators reported large heterogeneity between studies, including slightly stronger associations in studies that used more precise measures of depression (interview-based or treated depression) and insulin resistance (non-fasting estimates).48 Yet, because some studies in this meta-analysis included people with diabetes (who are insulin resistant by diagnosis), the weak positive association might in fact be driven by the relation between manifest diabetes and depressive symptoms. Additionally, in one study the association between insulin resistance and depression was not reported within the normal range of glucose tolerance, despite large variation of both variables.49 Finally, a report from the British Women’s Heart and Health study50 showed opposing results, with a lower risk of treated depression and lower depressive scores in the most insulin-resistant non-diabetic people than in normoglycaemic participants. A non-linear association, with lower risk in nondiabetic non-insulin-resistant people and higher risk in non-diabetic insulin-resistant and diabetic insulinresistant people, is consistent with a potential role of insulin secretion as an underlying factor, because absolute insulin secretion is increased in insulinresistant people without diabetes and then decreases after diagnosis.51 However, insulin secretion has rarely been estimated in population-based samples, and the few reported studies provide highly inconsistent findings. Investigators of the RISC study52 reported decreased insulin secretion in participants with depression, findings from the PPP-Botnia53 study showed no association, and findings from a further study54 of adolescents showed reductions in the oral disposition index (decreased compensatory insulin secretion) in participants with depression. In a study55 comparing children with and without a family history of depression, participants at high risk of depression had increased blood pressure, increased
arterial stiffness, and decreased insulin sensitivity compared with the control group. This finding suggests again that unmeasured vascular factors could underlie the link between insulin resistance and the development of depression. In agreement with this possibility, no prospective association between baseline insulin resistance and follow-up depressive symptoms was reported in the Caerphilly56 or the Whitehall II57 prospective cohort studies, although in the Whitehall study women with reduced incident depressive symptoms had decreased insulin secretion.57 Taken together, the evidence at present is mixed and suggests that the pathological changes associated with diabetes development might not induce depressive symptoms. This notion is further supported by the fact that, although some study findings showed the expected dose-response from normoglycaemia to impaired glucose regulation to undiagnosed diabetes,18,19 in most studies only those with previously known diabetes had an increased risk of depression.4,13–16 Finally, findings from a recent meta-analysis58 showed that, although diagnosed diabetes was associated with increased risk of depressive symptoms, the risk of depression in individuals with undiagnosed diabetes or impaired glucose tolerance was similar to that of non-diabetic controls (figure 2). Findings from longitudinal studies also lend support to an association between prevalent diabetes and incident depressive symptoms, but they have contradictory results regarding prediabetes, with an increased risk of depression in some studies and a decreased risk in others.59,60
Effects of diabetes diagnosis on depression risk It is important to distinguish the effects of prediabetes on depression from those related to manifest diabetes. The diagnosis of diabetes itself could lead to increased depressive symptoms as a result of stress caused by the awareness of having a pernicious chronic condition.45 At least two year-by-year analyses of antidepressant
2·5
Odds ratio
2·0
1·5
1·0
0·5
Normal glucose metabolism
Impaired glucose metabolism
Undetected type 2 diabetes
Impaired glucose metabolism
Undetected type 2 diabetes
Diagnosed diabetes
Undiagnosed diabetes
Diagnosed diabetes
Figure 2: Effect of impaired glucose metabolism, undetected diabetes, and diagnosed type 2 diabetes on risk of depression Point estimates and 95% CIs of odds ratios for depression, based on a meta-analysis58 by the European Depression in Diabetes Research Consortium.
www.thelancet.com/diabetes-endocrinology Vol 2 March 2014
239
Review
prescription data suggest an increase in antidepressant use in the first year after diagnosis of type 2 diabetes.41,61 This finding is consistent with a short-term temporary adverse effect of type 2 diabetes on mental health. An alternative explanation is that (as previously described) regular contact with a physician after diagnosis of type 2 diabetes makes detection of unrecognised depression more likely; in this situation, one would expect prescription rates to have remained high after diagnosis over the observation period. However, in these studies,41,61 soon after the diagnosis of diabetes, prescription rates returned to the rate reported before diagnosis. This finding suggests that the temporary increase in antidepressant use around the time of diabetes onset might represent a temporary increase in depression risk as a result of the diagnosis.
target organ damage, especially of the eyes, kidneys, nerves, heart, and blood vessels. Long-term complications of diabetes include retinopathy potentially leading to blindness; nephropathy (a frequent cause of renal failure); peripheral neuropathy with a risk for foot ulcers, amputations, and Charcot joints; and autonomic neuropathy that can lead to gastrointestinal, genitourinary, and cardiovascular symptoms, and erectile dysfunction. Patients with diabetes also have an increased incidence of atherosclerotic cardiovascular, peripheral arterial, and cerebrovascular disease. The burden of dealing with diabetes complications and comorbid disorders (especially cardiovascular disease) might therefore lead to symptoms of depression.65,66
The common-soil hypothesis Effects of antidiabetic drugs Evidence from studies suggests that a group of antidiabetic drugs, thiazolidinediones, could reduce both depressive symptoms and insulin resistance in diabetic people with comorbid major depressive disorder.62 In a study comparing the effect of pioglitazone with metformin (another insulin sensitising drug), similar improvements in insulin resistance were reported, although patients in the metformin group had no reduction in depressive symptoms.62
Effects of diabetes complications and comorbidities At more advanced stages of diabetes, self-care behaviours, complications, and resulting physical limitations might cause an increased risk of depression or depressive symptoms. Complex self-management procedures, anxiety about treatment, and difficulties coping could increase depressive symptoms in people with diagnosed diabetes.63 Hypoglycaemia (an acute complication of diabetes) is associated with an increased risk of depression,64 although the direction of this association is not clear. The chronic increase of blood glucose in people with diabetes confers risk for long-term complications and
Healthy (reference) Diabetes Cardiac diseases Stroke Lung diseases Osteoarthritis Poor hearing Poor vision 1
2
3
4
5
Odds ratio (95% CI)
Figure 3: Co-occurrence of chronic disorders and depression in old age (>60 years) Point estimates and 95% CIs of odds ratios for depression, based on a meta-analysis67 of cross-sectional data.
240
6
A 2010 meta-analysis67 assessed the association between various chronic diseases and depression at older ages on the basis of cross-sectional data. Summary odds ratios for depression by diabetes status, cardiac diseases, stroke, lung diseases, arthritis, and functional impairment were similar (range 1·5–2·5; figure 3).67 Thus, there seems to be no specific association between depression and diabetes that would be distinct from those between depression and other chronic disorders. According to the Bradford Hill criteria for evaluation of causal associations on the basis of observational data, poor specificity could suggest a confounded association, attributable to a factor that affects both the exposure and the outcome—a common-soil factor.68 Figure 1 illustrates various common-soil factors that might contribute to both depression and type 2 diabetes. In prospective studies investigating depression as a risk factor for diabetes, adjustment for BMI and lifestyle factors (particularly physical activity) substantially decreased the excess risk of diabetes in depressed people, suggesting that obesity and physical activity might contribute to the association between depression and diabetes.69 Furthermore, findings from a meta-analysis70 of prospective cohort studies suggested that obesity could increase the likelihood of depressive symptoms, although people with chronic or repeated episodes of common mental disorders might also be at increased risk of weight gain.71 The role of a common underlying factor has been suggested by findings from a Swedish study72 showing that abdominal obesity and physical inactivity, but not insulin resistance, were associated with depressive symptoms. Similar findings were reported by investigators of the Hertfordshire cohort study,73 in which triglycerides were associated with depressive symptoms in both sexes, whereas depressive symptoms were associated with HDL cholesterol only in women and with insulin resistance only in men. In an Australian study,74 approximately 40% of the association between insulin resistance and depression was attributable to waist circumference. These potential common-soil factors increase the risk of developing type 2 diabetes and metabolic syndrome.75 In a www.thelancet.com/diabetes-endocrinology Vol 2 March 2014
Review
meta-analysis48 investigating the association between insulin resistance and depression, the adjusted association (which took into account age, sex, obesity, lifestyle factors, occupation, and family history of diabetes) was significant only in one study and non-significant in the others, even though most of the studies included in the meta-analysis were based on large samples of population-based cohorts. Socioeconomic adversity and chronic stress leading to so-called wear and tear on the body (ie, allostatic load) could be linked to an increased risk of both type 2 diabetes76,77 and depression.78,79 Several other clinical characteristics are also associated with both of these disorders. First, chronic inflammation is associated with an increased risk of depression,80,81 and several genetic variants associated with a better immune response to infection have been found to be related to an increased risk of depression in candidate gene studies and genomewide association studies.82 Additionally, inflammation has an important role in the pathological process leading to type 2 diabetes and is a result of this disease.11 As a result, inflammation might increase the risk of both depression and type 2 diabetes, contributing to a noncausal association between the two disorders. Second, obesity is associated with an increased risk of depression, although the evidence is not totally consistent and the association has been found to be bidirectional, with obesity increasing the risk of depression and depression increasing the risk of obesity. Supporting a pathway from obesity to depression, findings from a meta-analysis83 of interventional evidence showed a reduction in symptoms of depression in obese individuals participating in a trial of weight loss. Obesity is also one of the most important modifiable risk factors for development of type 2 diabetes, the cornerstone of lifestyle intervention to prevent or delay development of type 2 diabetes and its complications.84 Thus, obesity might represent a common cause for both depression and diabetes, inflating the association between depression and diabetes in studies not controlling for adiposity. Third, physical inactivity is a major risk factor for the development of type 2 diabetes.75 Furthermore, findings from a Cochrane review85 suggested that exercise improves depressive symptoms in people with a diagnosis of depression compared with no treatment or control intervention, although this effect was modest in the most methodologically robust trials. Observational study findings showed a similar association.86 A commonsoil explanation of the association between depression and diabetes is supported if physical activity causally affects both type 2 diabetes and depression risk. Fourth, high blood glucose (already in the prediabetic range) is associated with many vascular risk factors, including general and central obesity, high blood pressure, and high concentrations of triglyceride and lipoprotein.51 The vascular depression hypothesis suggests a link between vascular pathology and depression in later life.87 In agreement with this hypothesis, findings from a www.thelancet.com/diabetes-endocrinology Vol 2 March 2014
meta-analysis88 showed evidence of an association between composite vascular risk (combining different subsets of risk factors), Framingham risk score profile for stroke, or individual factors, and subsequent depression, although no association was reported between some major vascular risk factors (eg, hypertension, smoking, or dyslipidaemia) and depression. The association between manifest cardiovascular disease and depression was even stronger.67 Thus, vascular pathology, both clinical and subclinical, might partly underlie the association between depression and diabetes.
Present evidence in relation to clinical guidelines In the following section, we briefly review present treatment guidelines for diabetes and depression in view of the evidence described in this review. However, key questions, such as whether prevention of diabetes is of value in preventing depression and whether depression treatment in diabetes helps to prevent diabetes complications and lower mortality, are, as yet, unanswered.89 Indeed, we are not aware of efficacy trials directly addressing these questions. Thus, we assess the consistency of present treatment guidelines with existing observational evidence.
Diabetes The mention of depression in diabetes position statements varies. Guidance for diabetes prevention from the UK National Institute for Health and Care Excellence acknowledges the possibility that depression could be a barrier to intensive lifestyle intervention for diabetes prevention, although no recommendations about screening or therapy are provided.90 The Canadian Diabetes Association, by contrast, emphasises depression as a potential risk factor for the development of type 2 diabetes and recommends that screening for diabetes should be started at a younger age (
Depression and type 2 diabetes: a causal association? Adam G Tabák, Tasnime N Akbaraly, G David Batty, Mika Kivimäki Lancet Diabetes Endocrinol 2014; 2: 236–45 Published Online November 14, 2013 http://dx.doi.org/10.1016/ S2213-8587(13)70139-6 Department of Epidemiology and Public Health, University College London, London, UK (A G Tabák MD, T N Akbaraly PhD, G D Batty PhD, Prof M Kivimäki PhD); Semmelweis University Faculty of Medicine, 1st Department of Medicine, Budapest, Hungary (A G Tabák); Inserm U 1061, Montpellier, France (T N Akbaraly); University Montpellier I, Montpellier, France (T N Akbaraly); and Centre for Cognitive Ageing and Cognitive Epidemiology, University of Edinburgh, Edinburgh, UK (G D Batty) Correspondence to: Dr Adam G Tabák, Department of Epidemiology and Public Health, University College London, London WC1E 6BT, UK [email protected]
There is a controversy regarding whether depression and type 2 diabetes are causally linked. To assess this issue, we review key findings for the association between depression and diabetes, and its underlying mechanisms. Findings from meta-analyses of cohort studies show a modestly sized bidirectional association between depression and type 2 diabetes (ie, depression predicts diabetes onset and diabetes predicts future depression). However, depression-related biological alterations in the hypothalamic-pituitary-adrenal cortex axis and the sympathetic nervous system, and subclinical inflammation, are not consistently linked with increased diabetes risk. The evidence for an association between depression and glycaemic traits (eg, glucose, insulin, insulin sensitivity, and insulin secretion) is also mixed. Diabetes increases the risk of depression to the same extent as do other chronic disorders (eg, cardiac diseases, osteoarthritis, lung disease, and poor hearing). At present, the available evidence suggests that pathophysiological changes preceding the onset of type 2 diabetes might not cause depression, nor might depression directly increase the risk of developing type 2 diabetes. Despite insufficient robust causal evidence, treating physicians should be aware of the co-occurrence of depression and type 2 diabetes.
Introduction As a result of their increasing prevalence, type 2 diabetes and depression are major public health issues. Worldwide, more than 365 million people are estimated to have type 2 diabetes,1 and almost 300 million people have major depression.2 Both these disorders are projected to be among the five leading causes of disease burden by 2030.3 The co-occurrence of type 2 diabetes and depression has attracted much research interest. If this association was causal, it would have profound implications for prevention and treatment of these disorders. However, the association between diabetes and depression seems to be complex, and does not follow a simple cause-and-effect pattern. In this Review we describe the conceptual models central to this association, provide an up-to-date review and appraisal of the evidence, and assess the relevance of preventive and clinical guidelines for type 2 diabetes and depression in view of research advances. Figure 1 shows some of the physiological, biological, and behavioural factors that are potentially connected with both type 2 diabetes and depression and might be implicated in the relation between these diseases.
Conceptual models for the association between depression and type 2 diabetes Findings from a series of prospective cohort studies and meta-analyses suggest that depression is a risk factor for type 2 diabetes.4 However, the evidence is also compatible with two alternative explanations. One notion is that depression and diabetes share common antecedents, resulting in a non-causal association between depression and type 2 diabetes. Studies have attempted to rule out this explanation (the so-called common-soil hypothesis) by statistical analyses in which adjustment is made for potential common causes or confounding factors, such as physical inactivity and obesity. However, the association between depression and diabetes might still be accounted for by residual confounding as a result of unmeasured or poorly measured common-cause risk 236
factors. The second notion posits that preclinical and clinical type 2 diabetes might increase the likelihood of participants reporting depressive symptoms, consistent with a reverse causation explanation. The three conceptual models—causality, the common-soil hypothesis, and reverse causality—might simultaneously contribute to the association between depression and type 2 diabetes (panel). Some investigators refer to a bidirectional association, combining the first and third models (ie, that depression could increase the risk of diabetes and that type 2 diabetes, in turn, could increase the risk of depression). A similar bidirectional association is suggested to exist between type 1 diabetes and depression.5 In this Review we organise the key evidence according to these three notions.
Depression as a risk factor for the development of type 2 diabetes Possible pathophysiological links In the multistage model of diabetes development,6 pathological changes preceding type 2 diabetes include longstanding insulin resistance that is accompanied by a compensatory increase in insulin secretion.7 This stage is typically followed by a stable adaptation period in which increased insulin resistance is not fully compensated by insulin secretion, leading to increased glucose concentrations. In the third stage of diabetes development— the unstable early decompensation period—β-cell function further worsens and compensation for insulin resistance becomes inadequate, leading to rapidly increasing glucose concentrations and manifest diabetes.6 Some aspects of the physiological changes associated with depression overlap with those characterising diabetes causation. Depression is associated with increased activity of the hypothalamic-pituitary-adrenal axis and the sympathetic nervous system,8,9 resulting in increased release of cortisol and other glucocorticoids, catecholamines (eg, adrenaline and noradrenaline), growth hormone, and glucagon. Cortisol stimulates glucose production and increases lipolysis, resulting in www.thelancet.com/diabetes-endocrinology Vol 2 March 2014
Review
decreased insulin sensitivity and hampering insulin secretion from β cells. Indeed, it is postulated that chronically high cortisol concentrations (as reported in about 50% of patients with depression) might result in obesity, insulin resistance, and type 2 diabetes.8 Furthermore, the inflammatory hypothesis of depression proposes a bidirectional association between activation of various aspects of the immune system and depression.10 The role of inflammatory markers and mechanisms are also well known in the pathophysiology of type 2 diabetes, providing a potential additional link between these diseases.11
Health behaviours Depression
Diabetes diagnosis
Vascular complications
www.thelancet.com/diabetes-endocrinology Vol 2 March 2014
Physical activity
Biological mechanisms Obesity Sympathetic nervous system↑ Hypothalamicpituitary-adrenal axis↑
Effects on glycaemic traits If depression is a causal factor for the development of type 2 diabetes, people with depression would be at an increased risk of prediabetes (defined with use of fasting glucose, HbA1c concentration, or an oral glucosetolerance test).12 However, most evidence from crosssectional studies suggests that prediabetes (defined as impaired fasting glucose according to the American Diabetes Association,13,14 or impaired fasting glucose and impaired glucose tolerance according to WHO) or undiagnosed diabetes is not associated with an increased risk of depression.15,16 Findings from the Hoorn study17 suggested a sex interaction, with an increased risk of impaired glucose tolerance in women with depression but not in men. Additionally, a dose-response relation was reported in US army veterans and south-Asian people at risk of depression, increasing stepwise from normoglycaemia to impaired fasting glucose to hyperglycaemia.18,19 However, longitudinal evidence from a Swedish population-based study20 suggested a different sex interaction, with a doubling of the risk of prediabetes in men with depression but no effect in women during 8–10 years of follow-up.20 A further suggestion of a link between diabetes and depression is from reports of a positive association between continuous measures of depression (eg, severity scores) and glycaemia (eg, fasting glucose, 2 h post-load glucose, or HbA1c concentration). However, findings from at least one study21 showed increased scores for depressive symptoms in people with very low and high glycaemic measures after adjustment for several potential confounders, contradicting the notion of a direct link between depression and dysglycaemia. Several studies have assessed depressive symptoms and validated indicators of insulin resistance and insulin secretion in the general population,22,23 but these results are mixed. Some study findings show no association between depression and insulin resistance,24,25 whereas others suggest increased insulin resistance in people with depression as a result of obesity and abnormal concentrations of adipokines.26,27 Furthermore, findings from at least two studies using the gold-standard measure of glycaemic traits (frequently sampled intravenous
Disease-related factors
CNS↑ Subclinical inflammation
Drug therapy Diet
Self-management
Type 2 diabetes
Figure 1: Potential pathophysiological and behavioural mechanisms linking depression and type 2 diabetes CNS=central nervous system.
Panel: Plausible mechanisms underlying the association between depression and type 2 diabetes Depression as a cause of type 2 diabetes • Depression might induce biological changes that increase the risk of type 2 diabetes • Depression might affect diabetes risk indirectly by increasing health-risk behaviours • Antidepressant drugs might have diabetogenic side-effects • Depression might worsen adherence to dietary and exercise regimens Type 2 diabetes as a cause of depression • Preclinical changes (eg, increased blood glucose or altered insulin concentrations) might directly affect depression risk through biochemical mechanisms • Diagnosis of chronic disease, such as type 2 diabetes, is a negative life event increasing the risk of depression • The burden of dealing with diabetes complications might lead to symptoms of depression Non-causal association • The so-called common-soil hypothesis posits that the association between depression and diabetes results from factors affecting both disorders—eg, chronic inflammation, obesity, physical inactivity, and vascular risk
glucose-tolerance test) showed increased insulin resistance irrespective of other factors.28,29 Nonetheless, analyses from the large-scale US National Health and Nutrition Examination Survey30 did not reproduce these 237
Review
findings in young adults. The analyses showed a statistically significant sex interaction, with no association between depression and insulin resistance in women, but in men a paradoxical inverse association between depression and homoeostasis model assessment (HOMA) insulin resistance. The investigators suggested that, in older people, unmeasured comorbidities rather than depression per se could bring about increased insulin resistance. This finding has not been confirmed by other studies. Findings from a small study31 of children at risk of obesity showed that depressive symptoms predicted insulin resistance 6 years later, even after the effect of baseline risk factors and changes in BMI were taken into account. By contrast, findings from a large multiethnic study (the Study of Women’s Health Across the Nation32) showed that depression was related to higher HOMA insulin resistance, although this association was eliminated after adjustment for baseline central obesity. Furthermore, depression was unrelated to changes in insulin resistance, suggesting that central obesity could be a common underlying factor linking depression with insulin resistance. Insulin secretion was investigated in at least three studies but with opposing findings: acute insulin response was increased in two studies28,29 but decreased in a third,25 probably as a compensatory response to decreased insulin sensitivity.
Off-target effects of antidepressant treatment A potential confounding factor that can distort the association between depression and type 2 diabetes is antidepressant medication. This confounding would happen if effective treatment for depression also had a favourable effect on glycaemic traits. Findings from an observational study33 comparing glycaemic traits before and after drug therapy in people with depression, and at least one randomised placebo-controlled trial of an antidepressant,34 showed no effect of these drugs on insulin resistance despite improvements in depressive symptoms. Findings from two further studies, however, showed improved insulin sensitivity in patients with depression receiving drug therapy without any changes in obesity measures.28,35 Insulin secretion has been less frequently investigated than has insulin resistance, and the results shown by these studies might be specific to the drug used36 or compensatory to changes in insulin secretion;25,28 however, these findings do not indicate that a major bias by drug therapy is a plausible explanation for the association between depression and diabetes. Findings from several studies of patients with recorded diagnosed type 2 diabetes as the outcome suggest that antidepressant drugs could actually increase, rather than decrease, risk of type 2 diabetes. Long-term use of at least moderate doses of antidepressants (eg, tricyclic antidepressants and selective serotonin reuptake inhibitors) was associated with an increased risk of type 2 diabetes.37–39 This finding should be interpreted cautiously, because 238
the scarcity of evidence for specificity by drug molecule or drug group effect raises the possibility that the observed association could be an artifact rather than the result of a causal relation. This alternative explanation is supported by analysis of data from repeated screenings in the Whitehall II study,40 which showed that although incidence of recorded diagnosed diabetes was higher in antidepressant users than in non-users, antidepressant use was not associated with undiagnosed diabetes at follow-up, or with increased fasting, increased 2 h postload plasma glucose concentrations, or increasing glucose concentrations over time. Thus, antidepressant use seems not to increase glucose concentrations, but is associated with an increased likelihood of being diagnosed with diabetes. Because many people with diabetes and depression are not diagnosed, detection bias is one potential explanation for such a non-causal association. Patients treated for one disorder might also be likely to be diagnosed with another disorder because of their more frequent contact with medical care.40,41
Effects on disease prognosis Depression has been postulated to have a role at later stages of the disease process, affecting the prognosis of type 2 diabetes. In support of this notion, patients with diabetes and persistent or worsening depressive symptoms have worse adherence to dietary and exercise regimens than do non-depressed patients with diabetes.42 Depression could be associated with a worsening of metabolic control, poor adherence to drugs, and increased health-care expenditures.43 A longitudinal modelling analysis44 lent support to a conceptual model in which the effect of depression on glucose dysregulation in diabetic patients with comorbid depression was mediated through lower adherence to self-care.
Diabetes as a risk factor for depression Standard risk factors for depression include having biological relatives with depression, traumatic experiences as a child (eg, sexual abuse), stressful life events (eg, the death of a loved one), alcohol and drug misuse, and having a serious somatic illness. Findings from meta-analyses suggest that the association between depression and type 2 diabetes might be bidirectional— that is, in addition to depression predicting the development of type 2 diabetes, type 2 diabetes might predict the development of depression, with the strength of each association being modest.45,46 This finding, and the fact that indicators of prediabetes (such as metabolic syndrome) predict the onset of depression,47 highlight the importance of considering diabetes and prediabetes as risk factors for depression.
Depressogenic effects of prediabetes Findings from a meta-analysis48 of cross-sectional population-based studies published up to 2011 suggest that there might be a weak positive association between www.thelancet.com/diabetes-endocrinology Vol 2 March 2014
Review
insulin resistance and depressive symptoms in adults. However, the investigators reported large heterogeneity between studies, including slightly stronger associations in studies that used more precise measures of depression (interview-based or treated depression) and insulin resistance (non-fasting estimates).48 Yet, because some studies in this meta-analysis included people with diabetes (who are insulin resistant by diagnosis), the weak positive association might in fact be driven by the relation between manifest diabetes and depressive symptoms. Additionally, in one study the association between insulin resistance and depression was not reported within the normal range of glucose tolerance, despite large variation of both variables.49 Finally, a report from the British Women’s Heart and Health study50 showed opposing results, with a lower risk of treated depression and lower depressive scores in the most insulin-resistant non-diabetic people than in normoglycaemic participants. A non-linear association, with lower risk in nondiabetic non-insulin-resistant people and higher risk in non-diabetic insulin-resistant and diabetic insulinresistant people, is consistent with a potential role of insulin secretion as an underlying factor, because absolute insulin secretion is increased in insulinresistant people without diabetes and then decreases after diagnosis.51 However, insulin secretion has rarely been estimated in population-based samples, and the few reported studies provide highly inconsistent findings. Investigators of the RISC study52 reported decreased insulin secretion in participants with depression, findings from the PPP-Botnia53 study showed no association, and findings from a further study54 of adolescents showed reductions in the oral disposition index (decreased compensatory insulin secretion) in participants with depression. In a study55 comparing children with and without a family history of depression, participants at high risk of depression had increased blood pressure, increased
arterial stiffness, and decreased insulin sensitivity compared with the control group. This finding suggests again that unmeasured vascular factors could underlie the link between insulin resistance and the development of depression. In agreement with this possibility, no prospective association between baseline insulin resistance and follow-up depressive symptoms was reported in the Caerphilly56 or the Whitehall II57 prospective cohort studies, although in the Whitehall study women with reduced incident depressive symptoms had decreased insulin secretion.57 Taken together, the evidence at present is mixed and suggests that the pathological changes associated with diabetes development might not induce depressive symptoms. This notion is further supported by the fact that, although some study findings showed the expected dose-response from normoglycaemia to impaired glucose regulation to undiagnosed diabetes,18,19 in most studies only those with previously known diabetes had an increased risk of depression.4,13–16 Finally, findings from a recent meta-analysis58 showed that, although diagnosed diabetes was associated with increased risk of depressive symptoms, the risk of depression in individuals with undiagnosed diabetes or impaired glucose tolerance was similar to that of non-diabetic controls (figure 2). Findings from longitudinal studies also lend support to an association between prevalent diabetes and incident depressive symptoms, but they have contradictory results regarding prediabetes, with an increased risk of depression in some studies and a decreased risk in others.59,60
Effects of diabetes diagnosis on depression risk It is important to distinguish the effects of prediabetes on depression from those related to manifest diabetes. The diagnosis of diabetes itself could lead to increased depressive symptoms as a result of stress caused by the awareness of having a pernicious chronic condition.45 At least two year-by-year analyses of antidepressant
2·5
Odds ratio
2·0
1·5
1·0
0·5
Normal glucose metabolism
Impaired glucose metabolism
Undetected type 2 diabetes
Impaired glucose metabolism
Undetected type 2 diabetes
Diagnosed diabetes
Undiagnosed diabetes
Diagnosed diabetes
Figure 2: Effect of impaired glucose metabolism, undetected diabetes, and diagnosed type 2 diabetes on risk of depression Point estimates and 95% CIs of odds ratios for depression, based on a meta-analysis58 by the European Depression in Diabetes Research Consortium.
www.thelancet.com/diabetes-endocrinology Vol 2 March 2014
239
Review
prescription data suggest an increase in antidepressant use in the first year after diagnosis of type 2 diabetes.41,61 This finding is consistent with a short-term temporary adverse effect of type 2 diabetes on mental health. An alternative explanation is that (as previously described) regular contact with a physician after diagnosis of type 2 diabetes makes detection of unrecognised depression more likely; in this situation, one would expect prescription rates to have remained high after diagnosis over the observation period. However, in these studies,41,61 soon after the diagnosis of diabetes, prescription rates returned to the rate reported before diagnosis. This finding suggests that the temporary increase in antidepressant use around the time of diabetes onset might represent a temporary increase in depression risk as a result of the diagnosis.
target organ damage, especially of the eyes, kidneys, nerves, heart, and blood vessels. Long-term complications of diabetes include retinopathy potentially leading to blindness; nephropathy (a frequent cause of renal failure); peripheral neuropathy with a risk for foot ulcers, amputations, and Charcot joints; and autonomic neuropathy that can lead to gastrointestinal, genitourinary, and cardiovascular symptoms, and erectile dysfunction. Patients with diabetes also have an increased incidence of atherosclerotic cardiovascular, peripheral arterial, and cerebrovascular disease. The burden of dealing with diabetes complications and comorbid disorders (especially cardiovascular disease) might therefore lead to symptoms of depression.65,66
The common-soil hypothesis Effects of antidiabetic drugs Evidence from studies suggests that a group of antidiabetic drugs, thiazolidinediones, could reduce both depressive symptoms and insulin resistance in diabetic people with comorbid major depressive disorder.62 In a study comparing the effect of pioglitazone with metformin (another insulin sensitising drug), similar improvements in insulin resistance were reported, although patients in the metformin group had no reduction in depressive symptoms.62
Effects of diabetes complications and comorbidities At more advanced stages of diabetes, self-care behaviours, complications, and resulting physical limitations might cause an increased risk of depression or depressive symptoms. Complex self-management procedures, anxiety about treatment, and difficulties coping could increase depressive symptoms in people with diagnosed diabetes.63 Hypoglycaemia (an acute complication of diabetes) is associated with an increased risk of depression,64 although the direction of this association is not clear. The chronic increase of blood glucose in people with diabetes confers risk for long-term complications and
Healthy (reference) Diabetes Cardiac diseases Stroke Lung diseases Osteoarthritis Poor hearing Poor vision 1
2
3
4
5
Odds ratio (95% CI)
Figure 3: Co-occurrence of chronic disorders and depression in old age (>60 years) Point estimates and 95% CIs of odds ratios for depression, based on a meta-analysis67 of cross-sectional data.
240
6
A 2010 meta-analysis67 assessed the association between various chronic diseases and depression at older ages on the basis of cross-sectional data. Summary odds ratios for depression by diabetes status, cardiac diseases, stroke, lung diseases, arthritis, and functional impairment were similar (range 1·5–2·5; figure 3).67 Thus, there seems to be no specific association between depression and diabetes that would be distinct from those between depression and other chronic disorders. According to the Bradford Hill criteria for evaluation of causal associations on the basis of observational data, poor specificity could suggest a confounded association, attributable to a factor that affects both the exposure and the outcome—a common-soil factor.68 Figure 1 illustrates various common-soil factors that might contribute to both depression and type 2 diabetes. In prospective studies investigating depression as a risk factor for diabetes, adjustment for BMI and lifestyle factors (particularly physical activity) substantially decreased the excess risk of diabetes in depressed people, suggesting that obesity and physical activity might contribute to the association between depression and diabetes.69 Furthermore, findings from a meta-analysis70 of prospective cohort studies suggested that obesity could increase the likelihood of depressive symptoms, although people with chronic or repeated episodes of common mental disorders might also be at increased risk of weight gain.71 The role of a common underlying factor has been suggested by findings from a Swedish study72 showing that abdominal obesity and physical inactivity, but not insulin resistance, were associated with depressive symptoms. Similar findings were reported by investigators of the Hertfordshire cohort study,73 in which triglycerides were associated with depressive symptoms in both sexes, whereas depressive symptoms were associated with HDL cholesterol only in women and with insulin resistance only in men. In an Australian study,74 approximately 40% of the association between insulin resistance and depression was attributable to waist circumference. These potential common-soil factors increase the risk of developing type 2 diabetes and metabolic syndrome.75 In a www.thelancet.com/diabetes-endocrinology Vol 2 March 2014
Review
meta-analysis48 investigating the association between insulin resistance and depression, the adjusted association (which took into account age, sex, obesity, lifestyle factors, occupation, and family history of diabetes) was significant only in one study and non-significant in the others, even though most of the studies included in the meta-analysis were based on large samples of population-based cohorts. Socioeconomic adversity and chronic stress leading to so-called wear and tear on the body (ie, allostatic load) could be linked to an increased risk of both type 2 diabetes76,77 and depression.78,79 Several other clinical characteristics are also associated with both of these disorders. First, chronic inflammation is associated with an increased risk of depression,80,81 and several genetic variants associated with a better immune response to infection have been found to be related to an increased risk of depression in candidate gene studies and genomewide association studies.82 Additionally, inflammation has an important role in the pathological process leading to type 2 diabetes and is a result of this disease.11 As a result, inflammation might increase the risk of both depression and type 2 diabetes, contributing to a noncausal association between the two disorders. Second, obesity is associated with an increased risk of depression, although the evidence is not totally consistent and the association has been found to be bidirectional, with obesity increasing the risk of depression and depression increasing the risk of obesity. Supporting a pathway from obesity to depression, findings from a meta-analysis83 of interventional evidence showed a reduction in symptoms of depression in obese individuals participating in a trial of weight loss. Obesity is also one of the most important modifiable risk factors for development of type 2 diabetes, the cornerstone of lifestyle intervention to prevent or delay development of type 2 diabetes and its complications.84 Thus, obesity might represent a common cause for both depression and diabetes, inflating the association between depression and diabetes in studies not controlling for adiposity. Third, physical inactivity is a major risk factor for the development of type 2 diabetes.75 Furthermore, findings from a Cochrane review85 suggested that exercise improves depressive symptoms in people with a diagnosis of depression compared with no treatment or control intervention, although this effect was modest in the most methodologically robust trials. Observational study findings showed a similar association.86 A commonsoil explanation of the association between depression and diabetes is supported if physical activity causally affects both type 2 diabetes and depression risk. Fourth, high blood glucose (already in the prediabetic range) is associated with many vascular risk factors, including general and central obesity, high blood pressure, and high concentrations of triglyceride and lipoprotein.51 The vascular depression hypothesis suggests a link between vascular pathology and depression in later life.87 In agreement with this hypothesis, findings from a www.thelancet.com/diabetes-endocrinology Vol 2 March 2014
meta-analysis88 showed evidence of an association between composite vascular risk (combining different subsets of risk factors), Framingham risk score profile for stroke, or individual factors, and subsequent depression, although no association was reported between some major vascular risk factors (eg, hypertension, smoking, or dyslipidaemia) and depression. The association between manifest cardiovascular disease and depression was even stronger.67 Thus, vascular pathology, both clinical and subclinical, might partly underlie the association between depression and diabetes.
Present evidence in relation to clinical guidelines In the following section, we briefly review present treatment guidelines for diabetes and depression in view of the evidence described in this review. However, key questions, such as whether prevention of diabetes is of value in preventing depression and whether depression treatment in diabetes helps to prevent diabetes complications and lower mortality, are, as yet, unanswered.89 Indeed, we are not aware of efficacy trials directly addressing these questions. Thus, we assess the consistency of present treatment guidelines with existing observational evidence.
Diabetes The mention of depression in diabetes position statements varies. Guidance for diabetes prevention from the UK National Institute for Health and Care Excellence acknowledges the possibility that depression could be a barrier to intensive lifestyle intervention for diabetes prevention, although no recommendations about screening or therapy are provided.90 The Canadian Diabetes Association, by contrast, emphasises depression as a potential risk factor for the development of type 2 diabetes and recommends that screening for diabetes should be started at a younger age (
