Diabetic retinopathy (DR) and diabetic macular edema (DME) are two of the leading causes of vision loss in adults [1]. An estimated one-third of people with diabetes worldwide have signs of DR, and approximately one in ten will develop vision-threatening diabetic retinopathy (VTDR), which includes severe non-proliferative DR (NPDR), proliferative DR (PDR) and/or clinically significant macular edema (CSME) [2].

DR and DME have a profoundly detrimental impact on vision-specific functioning [3] (i.e. reading and driving), mobility and independence and quality of life (QoL), [4] particularly in the VTDR stage [3, 5]. However, the association between DR and DME with psychosocial outcomes such as depression and anxiety is less well understood [6]. Qualitative work suggests that DR has a considerable emotional impact, resulting in feelings of distress, anger, anxiety and low mood [5, 7]. However, studies exploring the quantitative impact of DR on these psychosocial aspects are more limited [8, 9]. In this paper, we have used the term “psychosocial” as an umbrella term to capture a wide range of potential psychosocial factors including depression, depressive disorder, anxiety, vision-specific distress, diabetes-specific distress and emotional and social well-being.

A meta-analysis by de Groot and associates in 2001 showed a significant association between complications of diabetes and depression. In particular, ten out of 27 cross-sectional studies included in the analysis showed a significant correlation (combined p value of < 0.001) between DR and depressive symptoms [10]. Similarly, Fenwick and colleagues in 2011 reported that DR and associated visual loss were independently associated with poor emotional well-being [11]. Recently, several quantitative studies have demonstrated an association between DR and poor psychosocial outcomes [9, 12,13,14]. Others have shown that depression is associated with the presence and incidence of DR [15,16,17,18,19], whereby several pathophysiological mechanisms associated with depression (e.g. alteration in insulin and glucose resistance, dysregulation of the hypothalamic–pituitary–adrenal (HPA) axis and increase in circulating cytokines) have been proposed to play contributory roles in DR pathogenesis. This highlights a plausible bi-directional relationship between DR and psychosocial functioning.

With diabetes reaching epidemic proportions, the prevalence and incidence rates of DR/DME are increasing [2, 20]. A more comprehensive understanding of the relationship between DR/DME and psychosocial outcomes is therefore required so that effective interventions for patients can be developed and implemented. Therefore, this systematic review aims to determine the relationship between DR/DME and psychosocial functioning.

Methods

Literature search

A systematic review of current literature was conducted. Four databases, including PubMed, Medline, Embase and Cochrane library, were searched for articles evaluating the relationship between DR/DME and psychosocial functioning. No limitation was placed on the year of publication, with the earliest paper dating back to 1988, up until September 2017. The following search terms were used: (diabetic retinopathy OR diabetic macular [o]edema OR diabetic eye disease OR diabetic microvascular complications) AND (depression OR depressive disorder OR anxiety OR diabetes-related distress OR vision-specific distress OR diabetes-specific distress OR emotional well-being OR quality of life OR mental health OR psychological and social functioning). The reference lists of included articles were searched manually to identify and extract other potentially relevant articles.

Inclusion criteria

We based our eligibility criteria on the PICOS (population, intervention, comparison, outcomes, study design) framework recommended by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Supplementary Table 1) [21]:

  1. (1)

    Population Studies involving human participants with either type 1 or type 2 diabetes were included.

  2. (2)

    Intervention Both interventional (randomised control trials (RCT) and non-RCTs) and observational (cross sectional, case–control and prospective) studies were included.

  3. (3)

    Comparison Those without DR/DME and without psychological issues

  4. (4)

    Outcomes The outcomes were the level of psychosocial functioning measured by the use of validated questionnaires or via clinical diagnosis using the Diagnostic and Statistical Manual for Mental Disorders (DSM-IV). In several of the studies, the outcome was the prevalence, severity, incidence or progression of DR/DME

  5. (5)

    Study design Quantitative

  6. (6)

    Exposures The exposure was clinically diagnosed DR/DME. We accepted studies using different assessment methods, including but not limited to fundus photograph, fundoscopy, direct or indirect ophthalmoscopy, fluorescein angiography, clinicians’ diagnosis and hospital clinic notes. We also included studies using different scales to grade DR severity, such as the Early Treatment Diabetic Retinopathy Study (ETDRS) scale. In several of the studies, the exposure was psychosocial functioning (depressive symptoms or clinically diagnosed depression, measured by clinical diagnoses using the DSM-IV or by a validated questionnaire (e.g. Hospital Anxiety and Depression Scale, Beck Depression Inventory)).

Exclusion criteria

The following types of articles were excluded:

  1. (1)

    Review papers or editorials

  2. (2)

    Qualitative papers

  3. (3)

    Non-English papers

  4. (4)

    Studies with irrelevant exposure and/or outcome measures

  5. (5)

    Studies using utility instruments to assess psychosocial functioning

  6. (6)

    Studies that reported results for QoL as a whole without specific results for psychosocial functioning

  7. (7)

    Studies assessing psychosocial outcomes of diabetic complications without specific and separate results for DR/DME

Quality of evidence assessment

The quality of observational studies was assessed using a modified version of the Newcastle–Ottawa Scale (NOS) [22]. Originally designed to assess prospective and case–control studies, an adapted version of the NOS was used in the current study for the assessment of cross-sectional studies [23]. The NOS uses three main bias-reducing criteria at the study level to award up to a maximum of nine stars: (a) the selection and representativeness of the participants (maximum of four stars), (b) the comparability of groups (maximum of two stars) and (c) the ascertainment of exposure (for case–control) or outcome (for prospective and cross sectional) (maximum of three stars). Following previous reviews, studies assigned 0–4, 5–7 and ≥ 8 stars were considered as low, medium and high quality, respectively [24].

For the evaluation of interventional studies (non-RCT and RCT), the modified Downs and Black Checklist [25] was used, which measures the risk of bias at the study level via 27 criteria, giving a maximum score of 28 points. The domains covered included reporting, external validity, internal validity and assessment of statistical methods. The total points for each article were then divided by the total possible points—a score of 1 represents the highest possible quality article.

Sensitivity analysis

In order to ensure that study quality did not influence the outcomes of this systematic review, we conducted a sensitivity analysis. This involved analysing and synthesising the outcomes of the 20 studies rated as “high quality” by the NOS and the two studies rated as high quality by the modified Downs and Black Checklist.

Data extraction

The following relevant data were extracted from each article reviewed based on the “Strengthening the Reporting of Observational Studies in Epidemiology” (STROBE) statement [26]: year, author, study design, sample size (both diabetes and DR), exposure and outcome measures assessed and method of assessment, adjustments of confounders in analyses, patients’ diabetes type and a summary of the pertinent findings. (Tables 3, 4).

Results

Study characteristics

Of the 1827 screened titles, 179 were assessed for eligibility. After excluding 136 articles that did not meet the inclusion criteria, 42 were included in the systematic review (Supplementary Fig. 1). They comprised of four interventional (1 RCT and 3 non-RCTs) and 38 observational studies (33 cross sectional and five prospective). A summary of the data extracted from the 42 studies is presented in Tables 3 and 4.

Measurement of exposures and outcomes

For the measurement of psychosocial outcome/exposure, most studies (n = 20, 47.6%) utilised questionnaires that assessed generic health-related QoL with specific analyses on psychosocial outcomes (e.g. the Mental Health Composite Score (MCS) of the Short Form instruments [SF-12 (n = 4, 9.52%), SF-36 (n = 3, 7.14%)], VRQoL (e.g. the National Eye Institute Visual Functioning Questionnaire (NEI VFQ)-25 (n = 10, 23.8%)) or depression (e.g. Patient Health Questionnaire-9 (PHQ-9) (n = 3, 7.14%)) from the patient’s perspective. Two studies measured depression or depressive symptoms clinically using the Diagnostic and Statistical Manual for Mental Disordered (DSM).

DR was assessed using fundus photographs (number of fields unspecified; n = 3), seven-field fundus photography (n = 6), stereoscopic fundus photography (n = 1), four-field stereo retinal colour photography (n = 1), two-field colour fundus photography (n = 3), ophthalmologist examination (n = 11), record linkage (n = 5) and symptomatic eye problems (n = 1). Five studies did not specify the method of DR measurement.

Methodological quality

Of the 38 observational studies, the majority (92%) had moderate to high NOS scores, with 20 classified as “high quality” (≥ 8 stars) and 15 as “moderate quality” (5–7 stars). The remaining three studies were classified as “poor quality” (≤ 4 stars) (Table 1). Of the four interventional studies (both RCT and non-RCT) (Table 2), two studies were classified as “high quality” (≥ 0.8) and two as “moderate quality” (0.5–0.79).

Table 1 Ratings of articles reviewed
Table 2 Summary of data extracted from four studies looking at the impact of treatment therapies on psychological outcomes in patients with DR

Associations between DR (exposure) and psychosocial functioning

DR and poor psychosocial functioning was significantly associated in 20 of 28 observational studies (Table 3). Of these, eight cross-sectional studies found a significant association between the presence of DR and poor psychosocial functioning, [12, 27,28,29,30,31,32,33,34] including greater odds of depression, depressive symptoms and anxiety. For instance, Le Floch and colleagues showed that compared to no DR, any DR was associated with a higher risk of depressive symptoms, defined by a score of > 0 on the Mini Geriatric Depression Scale (p < 0.05) [12]. Of the eight studies, five adjusted for important demographic and clinical variables such as Haemoglobin A1c (HbA1c), duration of diabetes and lipid levels [12, 28,29,30].

Table 3 Summary of data extracted from the 38 observational studies included in the systematic review

Eight other cross-sectional studies found that severe DR, compared to early stage or no DR, was significantly associated with worse psychosocial outcomes. Mazhar and associates found that the decline in scores on the mental health domains of the NEI VFQ-25 and the SF-12 was modest until severity reached moderate NPDR in at least one eye, after which the decline in mental health became significantly steeper [35]. The remaining four studies evaluated the association between DR-associated vision loss and psychosocial outcomes [14, 32, 36, 37], with three studies reporting a significant link between reduced visual acuity (VA) and poor vision in the better eye with significantly impaired psychosocial functioning [32, 36, 37]. However, Fenwick and colleagues found, in path analyses of 514 patients with DR using DR-specific item banks, that the relationship between self-reported visual symptoms and emotional distress was mediated by mobility, inconvenience, activity limitation and social restriction (p < 0.05) [14].

Eight cross-sectional studies using mental health subscale scores from various QoL or depression questionnaires reported non-significant associations between DR and psychosocial functioning, [38,39,40,41,42,43,44] suggesting that the relationship between psychosocial functioning and DR may be driven by other medical or sociodemographic factors. Hirai and colleagues initially found, in univariate analysis, that compared to those with less severe DR and no visual impairment, a higher proportion of individuals with depression was observed among those with more severe DR and visual impairment. These associations attenuated in multivariable analyses, [38] with employment status (i.e. being unemployed vs employed) emerging as the main factor affecting MCS score.

Associations between depression (exposure) and DR

Current evidence for the bi-directional relationship between poor psychosocial outcomes and DR was limited to depression, with no studies reporting other aspects of psychosocial functioning (e.g. anxiety, emotional well-being). Of the ten relevant studies, [16,17,18,19, 45,46,47,48] six cross-sectional and three longitudinal studies found a significant, independent association between depression and DR. In the three longitudinal studies, antecedent clinical depression, [48] more severe depression and presence of depression at baseline [18] independently increased the risk of both DR incidence and progression, independent of glycemic control [18] and other health behaviours such as smoking, physical activity and diet [17,18,19,20, 46,47,48].

Roy and colleagues [18] found that type 1 DM patients with high Beck Depression Inventory (BDI) scores at both baseline and 6-year follow-up visits were more likely to show progression of DR (OR 2.44; 95% CI 1.01–5.88; p = 0.049) and progression to PDR (OR 3.19; 95% CI 1.30–7.87; p = 0.01) at follow-up than patients with low BDI scores. Similarly, Sieu and associates [19] showed that severe depression at baseline was associated with an independent increased risk of incident DR in patients with type 2 DM [OR 1.03; 95% CI 1.00–1.05] as well as shortened time to incident DR (hazard ratio = 1.03; 95% CI 1.01–1.04). The risk of incident DR was estimated to increase by up to 15% for every significant increase in depressive symptoms severity (5-point increase on the PHQ-9). Also, in patients with type 2 DM and depression, Yekta and co-workers [15] reported that individuals using antidepressants were less likely to have DR (OR 0.50, 95% CI: 0.31–0.82, p < 0.05) compared to those who were not. Only one study did not find any significant association between depressive symptoms and DR; [49] however, this was a small, unadjusted, cross-sectional study where details of the fundus examination were not well specified.

Medical interventions for DR on psychosocial functioning

Three studies explored the impact of medical interventions for DR on psychosocial functioning (Table 4). Loftus and colleagues [50] compared the effect on psychosocial outcomes of an intravitreal pegaptanib sodium injection to a sham injection in patients with DME and reported a clinically meaningful improvement in the mental health domain of the NEI VFQ-25 after week 102 of pegaptanib treatment (p < 0.05), with a concomitant statistically significant improvement in VA (6.1 letters with pegaptanib vs. 1.3 letters for sham, p < 0.01). Similar results were observed in three other non-RCT studies assessing different treatment options for DR/DME [51,52,53].

Table 4 Summary of data extracted from four studies looking at the impact of treatment therapies on psychological outcomes in patients with DR

Sensitivity analysis

After synthesising the results from 22 studies rated as “high quality” by either the NOS or the modified Downs and Black checklist, the findings remained similar to those of the full systematic review (n = 42 studies). In summary, DR was significantly associated with poor psychosocial functioning in 11 of 13 high-quality observational studies, [4, 9, 12,13,14, 28, 29, 32, 54,55,56] with more severe DR was independently associated with worse psychosocial outcomes. Three cross-sectional studies did not find significant associations between DR and psychosocial functioning. Furthermore, the link between depressive symptoms and risk of DR remained, with three cross-sectional and two longitudinal high-quality studies showing that presence of depressive symptoms was associated with increased presence, incidence and progression of DR [15,16,17,18, 58].

Discussion

In this systematic review of the relationship between DR and psychosocial functioning, we found that DR/DME and related visual impairment, especially in more severe stages of DR, were significantly cross sectionally and longitudinally associated with poorer psychosocial outcomes, including higher levels of depression, anxiety and worse scores on mental health domains of health- and vision-related QoL questionnaires. Importantly, the relationship between depression and DR appears to be bi-directional, as the presence of depression or depressive symptoms is linked with incident, progression and severe DR. Our findings support the need for interventions to improve psychosocial well-being in patients with DR and also highlight the importance of prevention, early detection and management of depression in those with diabetes to reduce the development and progression of DR.

Our findings show that not only was DR-induced visual loss associated with poor mental health, but also DR alone was independently linked with worse psychosocial outcomes. This result suggests that factors beyond vision loss, such as contrast sensitivity, visual field loss and loss of colour and contrast may be important for mental health, [59,60,61] although studies specifically exploring this topic are lacking. Indeed, while visual impairment is important clinically, it only explains 30–40% of the variance in QoL in people with DR [62]. In addition, it is possible that loss of daily living activities and social life resulting from DR-related vision loss are responsible for declines in psychosocial well-being. As shown by Fenwick and colleagues, [14] the association between visual symptoms and emotional distress in people with DR was mediated by factors such as mobility, activity limitation, inconvenience and social restriction. People with DR may benefit from early interventions to support continued participation in activities of daily living; support groups to minimise social isolation and psychological support from a multidisciplinary team as needed.

Our findings also suggest a bi-directional relationship between DR and depression. Several plausible pathophysiological mechanisms can explain this relationship. Roy and colleagues suggest that alteration in insulin and glucose resistance and dysregulation of the HPA axis are some pathophysiological mechanisms of depression. HPA axis dysregulation and resulting hypercortisolemia may be associated subsequently with changes in insulin resistance, leading to the pathogenesis of DR [18]. Chen and colleagues further add that increase in circulating cytokines (seen in both progression of diabetes and depression) and insulin deficiency leads to fluctuations in blood glucose level, abnormal neural development and neurocognitive defect [6, 63]. Future research to fully understand the temporal relationship, and associated underlying mechanisms, between depression and DR is warranted so that interventions for at-risk patients can be implemented. Furthermore, more research on the impact of anxiety and emotional well-being on DR is warranted, as our study found a paucity of evidence in this area.

Based on the results of our systematic review, we provide a series of recommendations for preventing and improving poor psychosocial functioning in patients with DM and DR.

Primary prevention

Preventing, early detection and optimal management of DR

Given that DR, particularly late-stage DR, is associated with increased likelihood of poor psychosocial functioning, preventing DR from developing or slowing its progression is paramount. Early screening and regular retinal examination is the cornerstone of effective diabetes management, aiming to detect DR before it causes visual loss so that effective treatment can be given [64, 65] and complications such as poor psychosocial functioning can be avoided. Hand in hand with screening and slowing the progression to late-stage DR are effective diabetes management via optimal glucose, blood pressure and lipids control. Landmark trials such as the UK Prospective Diabetes Study (UKPDS) have shown that for every 1% reduction in HbA1c there is a 25% reduction in microvascular complications [66]. Interventions to help patients make lifestyle modifications (e.g. diet and physical activity) [67, 68] and incorporating personalised care planning [69] have been shown to be effective in improving diabetes control indicators and capability to self-manage their conditions, and implementation of such strategies to the broader diabetes population is needed.

Improving the early detection and referral of poor psychosocial functioning in patients with DR

Despite clear diagnostic guidelines for depression and the availability of good screening tools such as the PHQ-9, screening for depression is not routinely done for patients with DR during their clinic visits, resulting in a lack of timely detection and management [70]. A patient-reported outcome measure (PROM) tailored specifically to patients with DR would assist in the detection of poor mental health and inform subsequent referral pathways for continued care. RetCAT [71, 72] is an item bank and computerised adaptive testing (CAT) system comprising domains of DR-specific QoL including emotional well-being, social well-being and concerns. With an average of 7 items required to achieve precise measurement (e.g. standard error of measurement 0.387 [equivalent to 0.85 reliability]) of each emotional trait, RetCAT enables brief and yet robust measurement [71]. Ideally, RetCAT would be implemented in tertiary eye clinics allowing patients to take the surveys prior to consultation, with a report generated and stored in the patient’s online medical records for discussion during the clinical consultation. Specific management strategies, including counselling, occupational therapy and vision rehabilitation, will be linked to the relevant RetCAT scores allowing patients to get the subsidiary care that they need. Moreover, with RetCAT performed at every visit, the patient’s QoL status can be monitored over time. Evidence suggests that integration of PROMs in healthcare organisations, along with practitioner buy-in and support through education and training about the usefulness of such enterprises, can substantially improve patient care and physician–patient relationships [73].

Secondary prevention

Improving the management of pre-existing depression in patients with diabetes

Optimal management of depression through both medical and psychological therapy in patients with diabetes is essential to reduce the risk of patients developing DR. Integrated care which coordinates ophthalmologist and psychiatric/psychological referrals, treatments and follow-up visits is likely to maximise efficiency and lead to effective patient-centred care [74].

  1. a.

    The use of pharmacotherapy suggests that antidepressants can reduce depressive symptoms and slow the development of DR, possibly through control of the immunoinflammatory response, improved medication adherence and health behaviours leading to better diabetes control [6].

  2. b.

    Evidence-based psychological strategies such as problem-solving therapy (PST) and Acceptance and Commitment Therapy (ACT) have been shown to be effective in reducing depressive symptoms in those with vision impairment [75] and DR and need to be considered [76].

A dedicated tool to determine the psychosocial impact of DR would be useful to inform management options. However, at present, most instruments rely on subscales in vision-related QoL tools like the IVI [77] and NEI VFQ [78] or even generic HRQOL tools such as the MCS of the SF instruments, which may not be sensitive to mental health outcomes specific to this condition. New instruments in development such as RetCAT [71] are needed to provide researchers, clinicians and rehabilitation workers with comprehensive and disease-specific tools to monitor and provide targeted interventions around mental health in people with DR.

Our systematic review has several strengths. First, most of the studies had sound methodological and study qualities, with more than half attaining high NOS scores. Second, our studies had wide geographic diversity, which aided in the generalisability of our results. We did not limit the timeframe, allowing a broad range of literature from 1988 to 2017. Last, we included studies with DR/psychosocial outcomes as both exposure and outcome, allowing to assess the potential for a bi-directional relationship.

However, we acknowledge certain limitations. The studies were largely cross sectional, with a handful of observational and interventional studies, and a lack of RCTs. Future longitudinal studies are warranted to more accurately assess the causality between DR and depression and to monitor disease progression. Given the potential for time-varying confounding, adjustments to properly estimate the relationship between DR and depression will be essential in subsequent longitudinal studies [79]. Another limitation was the lack of uniform measures to assess DR and its severity. While some studies used the ETDRS to categorise DR severity [80], others utilised ophthalmologist assessments or did not report the assessment method. This variation may have affected the comparability of studies. We included studies of all levels of quality in our review which could have reduced the robustness of our findings. However, when we conducted a sensitivity analysis including only those studies with a ‘high-quality’ rating, results were very similar. Finally, we were not able to conduct a meta-analysis due to the large number of outcomes and outcome measures considered in this review. Using a standardised, valid PROM for DR QoL such as RetCAT, subsequent pooling of data for comparison may be possible.

Conclusions

Our systematic review found that DR/DME negatively affects psychosocial outcomes, reinforcing the need for primary physicians to continue targeting primary prevention of DR/DME and the importance of tight control of existing DR/DME. Similarly, with depression as an independent risk factor for development and progression of DR/DME, this condition should be detected and treated early in patients with diabetes to reduce the incidence and progression of VTDR.