Association between depression, anxiety, and treatment adherence in patients with diabetic macular edema

Abstract

BACKGROUND

Diabetic macular edema (DME) is a major cause of vision impairment among working-age adults with diabetes. Treatment adherence remains a considerable challenge, with psychological factors potentially playing a crucial role.

AIM

To investigate the relationship between depression, anxiety, and treatment adherence among patients with DME.

METHODS

A retrospective cohort study was conducted at our hospital January 2021 and August 2025, including 130 patients with DME. Depression and anxiety were assessed using the Patient Health Questionnaire-9 (PHQ-9) and Generalized Anxiety Disorder-7 (GAD-7) scale, respectively. Treatment adherence was evaluated using the Medication Adherence Report Scale-5 and appointment attendance rates. Visual function and quality of life were measured using the National Eye Institute Visual Function Questionnaire-25.

RESULTS

Among 130 patients, 44.6% (n = 58) demonstrated depression symptoms (PHQ-9 ≥ 5), and 37.7% (n = 49) showed anxiety symptoms (GAD-7 ≥ 5). Poor medication adherence (Medication Adherence Report Scale-5 < 17) was observed in 43.1% (n = 56) of patients. Depression severity was significantly associated with low medication adherence (r = -0.42, P < 0.001) and reduced appointment attendance (r = -0.38, P < 0.001). Similarly, anxiety was inversely correlated with medication adherence (r = -0.35, P < 0.001). In multivariate analysis, moderate-to-severe depression (PHQ-9 ≥ 10) (odds ratio = 3.42, 95% confidence interval: 1.68-6.95, P = 0.001) and moderate-to-severe anxiety (GAD-7 ≥ 10) (odds ratio = 2.86, 95% confidence interval: 1.35-6.04, P = 0.006) were independent predictors of poor medication adherence. Patients with depression had significantly lower National Eye Institute Visual Function Questionnaire-25 composite scores than those without depression (58.3 ± 14.2 vs 72.6 ± 12.8, P < 0.001). Treatment adherence partially mediated the relationship between depression and visual outcomes, accounting for 28.6% of the total effect.

CONCLUSION

Depression and anxiety significantly impact treatment adherence in DME patients. Integrated care approaches addressing both psychological and ophthalmological aspects are essential for optimizing treatment outcomes.

Key Words: Diabetic macular edema; Depression; Anxiety; Treatment adherence; Visual function; Quality of life

Core Tip: Depression and anxiety are frequent psychiatric comorbidities in patients with diabetic macular edema and play a crucial role in treatment adherence. This retrospective study found that both disorders were strongly associated with poor medication compliance and reduced appointment attendance, ultimately leading to impaired visual outcomes and lower quality of life. Treatment adherence partially mediated the link between depression and vision loss. These findings underscore the need for integrated management strategies that combine psychiatric care with ophthalmological treatment to optimize adherence and improve clinical prognosis in diabetic macular edema.

INTRODUCTION

Diabetic macular edema (DME), the leading cause of vision loss among individuals with diabetes mellitus, affects approximately 7% of the diabetic population worldwide[1]. As the global prevalence of diabetes increases, with an estimated 537 million adults living with the condition in 2021 and projections reaching 783 million by 2045, the DME burden is expected to expand substantially[2]. DME pathophysiology involves complex mechanisms, including chronic hyperglycemia-induced vascular damage, blood-retinal barrier breakdown, and inflammatory processes, leading to fluid accumulation in the macula[3]. DME management has been revolutionized by anti-vascular endothelial growth factor (anti-VEGF) therapy, which has become the first-line treatment for center-involving DME[4]. However, in real-world effectiveness often lags behind outcomes achieved in controlled clinical trials, with treatment adherence emerging as a critical determinant of therapeutic success[5]. Studies have demonstrated that patients in routine clinical practice achieve lower visual acuity gains than clinical trial participants, with poor adherence to treatment regimens being a major contributor[6].

Depression and anxiety are prevalent comorbidities among individuals with diabetes, as meta-analyses indicate that depression prevalence is approximately doubled in patients with diabetes compared with the general population[7]. The bidirectional relationship between diabetes and depression has been well-established, with depression increasing the risk of developing type 2 diabetes by 40%-60%, and diabetes predicting future depressive episodes[8]. This comorbidity results in substantial disease burden, including higher symptom severity, decreased treatment adherence, additive functional impairment, reduced quality of life, and increased mortality[9]. DME-associated psychological burden extends beyond the general challenges of diabetes management. The threat of vision loss creates significant emotional distress, with patients experiencing fear and anxiety regarding their future independence and quality of life[10]. The intensive treatment regimen required for DME, typically involving monthly intravitreal injections and frequent clinic visits, can be overwhelming for patients already managing multiple diabetes-related complications[11]. Reportedly, fear and anxiety related to injections represent major barriers to treatment adherence, affecting > 50% of patients[12].

Despite the recognized importance of psychological factors in chronic disease management, a paucity of research specifically examining the impact of depression and anxiety on treatment adherence in patients with DME remains. Studies have focused on either macrovascular complications or composite microvascular outcomes, with limited attention to the unique challenges experienced by individuals with vision-threatening diabetic eye disease[13]. Furthermore, the mechanisms through which psychological distress influences adherence behaviors in this population remain unclear. The assessment of treatment adherence in DME presents unique challenges, as it encompasses multiple dimensions, including medication adherence for systemic diabetes control, attendance at ophthalmology appointments for intravitreal injections, and adherence to lifestyle modifications[14]. Traditional adherence measures may not fully capture the complexity of DME management, necessitating comprehensive evaluation approaches that consider both objective and subjective indicators[15].

Recent evidence suggests that the relationship between psychological distress and treatment adherence may be mediated by several factors, including self-efficacy, social support, and illness perceptions[16]. Understanding these mediating pathways is crucial for developing targeted interventions to improve adherence and ultimately visual outcomes in patients with DME. Furthermore, the impact of psychological interventions on adherence and glycemic control in diabetic populations is promising, though specific studies in patients with DME remain limited[17]. The economic implications of poor treatment adherence in DME are substantial, resulting in increased healthcare utilization, emergency interventions, and indirect costs related to vision loss and disability[18]. Identifying modifiable psychological factors influencing adherence could inform cost-effective strategies for optimizing DME management and reducing the overall burden of diabetic eye disease. This study aimed to address these knowledge gaps by systematically examining the relationship between depression, anxiety, and treatment adherence in a cohort of patients with DME receiving anti-VEGF therapy.

MATERIALS AND METHODS

Study design and setting

This retrospective cohort study was conducted at Shanghai Jiao Tong University Sixth People’s Hospital between January 2021 and August 2025. The research protocol was approved by the Review Committee of Shanghai Sixth People’s Hospital, No. 2025-KY-305 (K), and all procedures adhered to the tenets of the Declaration of Helsinki. Given the retrospective nature of the study, the requirement for individual informed consent was waived by the Institutional Review Board. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines for reporting observational studies. The study was conducted in the Department of Ophthalmology, which served as a tertiary referral center for diabetic eye disease in the region. The department provided comprehensive care for diabetic retinopathy and DME, including medical management, laser photocoagulation, and intravitreal anti-VEGF therapy. The multidisciplinary team included retina specialists, diabetes educators, social workers, and mental health professionals, facilitating integrated care for patients with complex needs.

Study population

Inclusion criteria: (1) Age ≥ 18 years at the time of DME diagnosis; (2) Confirmed diagnosis of type 1 or type 2 diabetes mellitus based on American Diabetes Association criteria; (3) Diagnosis of center-involving DME, confirmed by optical coherence tomography (OCT), with central subfield thickness (CST) ≥ 250 μm in at least one eye; (4) Anti-VEGF therapy (ranibizumab, aflibercept, or conbercept) initiation during the study period; (5) Minimum follow-up duration of 12 months from treatment initiation; (6) Completion of psychological assessment tools [Patient Health Questionnaire-9 (PHQ-9) and Generalized Anxiety Disorder-7 (GAD-7)] within 3 months of treatment initiation; and (7) Availability of complete medical records including treatment history and adherence data.

Exclusion criteria: (1) Concurrent retinal diseases that could affect visual acuity or macular thickness (such as age-related macular degeneration, retinal vein occlusion, and epiretinal membrane); (2) History of vitrectomy or other intraocular surgery within 6 months before enrollment; (3) Diagnosed psychiatric disorders other than depression or anxiety (such as schizophrenia or bipolar disorder); (4) Cognitive impairment or dementia that could affect the ability to complete questionnaires or adhere to treatment; (5) Active substance-abuse disorder; (6) Terminal illness with life expectancy ≤ 12 months; or (7) Participation in other clinical trials that could influence treatment adherence or psychological status.

Data collection procedures

Data collection was performed through a systematic review of electronic medical records and paper-based clinical charts. A standardized data extraction form was developed and pilot-tested on 10 randomly selected records to ensure consistency and completeness. Two trained research assistants independently extracted data, with discrepancies resolved through discussion with a senior investigator. The inter-rater reliability for key variables was assessed using Cohen’s kappa coefficient, with values > 0.85 for all primary outcome measures. Demographic information collected included age, sex, ethnicity, education level, employment status, marital status, and health insurance type. Clinical data encompassed diabetes type and duration, most recent glycated hemoglobin (HbA1c) level, presence of diabetic complications (neuropathy, nephropathy, or cardiovascular disease), systemic medications, best-corrected visual acuity (BCVA) in both eyes, DME severity based on OCT findings, and treatment history, including number and type of intravitreal injections received.

Assessment of depression and anxiety

PHQ-9: Depression was assessed using the PHQ-9, a validated 9-item self-report instrument that measures the frequency of depressive symptoms over the past 2 weeks[19]. Each item was scored from 0 (not at all) to 3 (nearly every day), yielding a total score of 0-27. The PHQ-9 items corresponded directly to the nine diagnostic criteria for major depressive disorder in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition. Cut-off scores were applied as follows: 0-4, 5-9, 10-14, 15-19, and 20-27 indicate minimal, mild, moderate, moderately severe, and severe depression, respectively. A score of ≥ 10 has been shown to have an 88% sensitivity and specificity for detecting major depression. The PHQ-9 has been extensively validated in diabetic populations, demonstrating good internal consistency (Cronbach’s α = 0.83-0.89) and test-retest reliability [intraclass correlation coefficient (ICC) = 0.84].

GAD-7 scale: Anxiety symptoms were evaluated using the GAD-7, a 7-item questionnaire that assesses the severity of generalized anxiety disorder symptoms[20]. Respondents rate how often they have been bothered by each symptom over the past 2 weeks on a scale from 0 (not at all) to 3 (nearly every day), with total scores ranging from 0 to 21. Score interpretation was as follows: 0-4, 5-9, 10-14, and 15-21 for minimal, mild, moderate, and severe anxiety. A cut-off score of ≥ 10 has demonstrated optimal sensitivity (89%) and specificity (82%) for identifying generalized anxiety disorder. The GAD-7 has shown excellent internal consistency (Cronbach’s α = 0.92) and good test-retest reliability (ICC = 0.83) in various clinical populations, including patients with diabetes.

Assessment of treatment adherence

Medication adherence: Medication adherence was assessed using the Medication Adherence Report Scale-5 (MARS-5), a widely validated self-report measure[21] (Copyright information was shown in the Supplementary material). The MARS-5 comprises five items assessing common non-adherent behaviors (altering the dose, forgetting to take medication, stopping medication, skipping doses, and taking less than prescribed). Each item is rated on a 5-point scale (1 = always to 5 = never), yielding total scores of 5-25, with higher scores indicating better adherence. Adherence levels were categorized as: Good (score ≥ 23), moderate (score 17-22), and poor (score < 17) adherence. The MARS-5 has demonstrated good reliability and validity in diabetic populations (Cronbach’s α = 0.67-0.83) and significant associations with glycemic control. In addition, pharmacy refill data were collected to calculate the medication possession ratio for oral antidiabetic and antihypertensive medications over the 12-month follow-up period. A medication possession ratio ≥ 80% was considered adequate adherence.

Appointment adherence: Appointment adherence was evaluated through a comprehensive review of scheduled and completed ophthalmology visits. The appointment adherence rate was calculated as the number of completed visits divided by the total number of scheduled visits, expressed as a percentage. Appointments cancelled with at least a 24-hour notice and subsequently rescheduled within 7 days were not counted as missed appointments. For anti-VEGF injection appointments, we calculated the injection interval consistency, defined as the coefficient of variation of intervals between consecutive injections. Additional data collected included reasons for missed appointments (when documented), time for missed appointment rescheduling, and travel distance to the clinic.

Visual function and quality of life assessment

National Eye Institute Visual Function Questionnaire-25: Vision-related quality of life was assessed using the National Eye Institute Visual Function Questionnaire-25 (NEI VFQ-25), a validated instrument specifically designed for individuals with chronic eye diseases[22]. The questionnaire comprises 25 items across 12 subscales: General health, general vision, ocular pain, near activities, distance activities, social functioning, mental health, role difficulties, dependency, driving, color vision, and peripheral vision. Each subscale score is converted to a 0-100 scale, with higher scores indicating better vision-related functioning. A composite score is calculated as the unweighted average of the 11 vision-targeted subscale scores (excluding general health). The NEI VFQ-25 has demonstrated good psychometric properties in DME populations, with internal consistency of 0.71-0.85 for different subscales and significant correlations with visual acuity and retinopathy severity.

Clinical assessments

Visual acuity measurement: BCVA was measured using Early Treatment Diabetic Retinopathy Study (ETDRS) charts at a 4-meter testing distance under standardized illumination conditions. Visual acuity was recorded as the number of letters correctly identified and converted to logarithm of the minimum angle of resolution units for analysis. When ETDRS charts were not available, Snellen visual acuity measurements were converted to approximate ETDRS letter scores, using published conversion algorithms. The better-seeing eye, worse-seeing eye, and binocular visual acuity were all recorded, with the better-seeing eye visual acuity used as the primary measure for correlation analyses.

OCT: Spectral domain OCT was performed using Heidelberg Spectralis OCT (Heidelberg Engineering, Germany) to assess macular thickness and morphology. CST was measured using the ETDRS grid centered on the fovea. Additional parameters recorded included the presence of intraretinal fluid, subretinal fluid, and retinal layer disruption. OCT images were reviewed by two masked graders, with discrepancies resolved by a senior retina specialist. Changes in CST from baseline to 12 months were calculated as both absolute values and percentage changes.

Statistical analysis

Statistical analyses were performed using SPSS version 28.0 (IBM Corporation, Armonk, NY, United States) and R version 4.2.0 (R Foundation for Statistical Computing, Vienna, Austria). Descriptive statistics were calculated for all variables, with continuous variables expressed as means ± SD or medians with interquartile ranges depending on distribution normality, and categorical variables as frequencies and percentages. The Shapiro-Wilk test was used to assess the normality of continuous variables. Bivariate analyses examined associations between psychological variables (PHQ-9 and GAD-7 scores) and adherence outcomes. Pearson’s correlation coefficients were calculated for normally distributed continuous variables, and Spearman’s rank correlation for non-normally distributed data. The χ² tests or Fisher’s exact tests were employed for categorical variables. Independent samples t-tests or Mann-Whitney U tests compared adherence measures between depressed and non-depressed groups, and between anxious and non-anxious groups. Multiple linear regression models were constructed to examine the independent associations of depression and anxiety with medication adherence scores, controlling for potential confounders including age, sex, diabetes duration, HbA1c level, visual acuity, number of systemic medications, and socioeconomic factors. Logistic regression analyses evaluated predictors of poor adherence (MARS-5 score < 17 or appointment adherence < 80%). Variables with P < 0.20 in bivariate analyses were included in initial multivariate models, with final models determined using backward stepwise selection with a retention threshold of P < 0.05. Mediation analyses were conducted to examine whether specific factors (such as self-efficacy, social support, and illness perceptions) mediated the relationship between psychological symptoms and adherence outcomes. The PROCESS macro for SPSS was used to estimate direct and indirect effects with bootstrapped confidence intervals (5000 samples). Moderation analyses tested whether the association between psychological symptoms and adherence varied by patient characteristics such as age, disease duration, or baseline visual acuity. To account for potential clustering effects within treating physicians, multilevel models were constructed with patients nested within providers. ICCs were calculated to quantify the proportion of variance attributable to provider-level factors. Sensitivity analyses were performed excluding patients with severe depression or anxiety (scores in the highest quartile) to assess the robustness of findings. All statistical tests were two-tailed, with P < 0.05 considered statistically significant.

RESULTS

Participant characteristics

A total of 130 patients with DME, who met the inclusion criteria, were enrolled. The mean age was 62.4 ± 11.3 years, with a slight male predominance (53.8%). Many participants had type 2 diabetes (90.8%) with a mean disease duration of 14.6 ± 7.2 years. The mean baseline HbA1c was 8.2 ± 1.6%, indicating suboptimal glycemic control in the study population (Table 1).

Table 1 Baseline demographic and clinical characteristics of study participants, mean ± SD, n (%).

Characteristic	Total (n = 130)	Depression (n = 58)	No depression (n = 72)	P value
Demographics
Age, years	62.4 ± 11.3	60.8 ± 10.9	63.7 ± 11.5	0.142
Men	70 (53.8)	28 (48.3)	42 (58.3)	0.251
Education ≥ 12 years	78 (60.0)	30 (51.7)	48 (66.7)	0.083
Employed	52 (40.0)	18 (31.0)	34 (47.2)	0.059
Married	86 (66.2)	34 (58.6)	52 (72.2)	0.102
Diabetes characteristics
Type 2 diabetes	118 (90.8)	53 (91.4)	65 (90.3)	0.829
Duration, years	14.6 ± 7.2	15.8 ± 7.5	13.6 ± 6.8	0.086
HbA1c	8.2 ± 1.6	8.6 ± 1.7	7.9 ± 1.5	0.013
Diabetic neuropathy	48 (36.9)	28 (48.3)	20 (27.8)	0.016
Diabetic nephropathy	32 (24.6)	19 (32.8)	13 (18.1)	0.049
Ophthalmologic characteristics
BCVA, logMAR	0.48 ± 0.32	0.56 ± 0.34	0.42 ± 0.29	0.012
CST, μm	412 ± 98	428 ± 102	399 ± 93	0.093
Bilateral DME	72 (55.4)	38 (65.5)	34 (47.2)	0.037
Prior anti-VEGF injections	3.2 ± 4.1	3.8 ± 4.5	2.7 ± 3.7	0.134
Psychological measures
PHQ-9 score	7.3 ± 5.8	12.8 ± 4.2	2.9 ± 1.8	< 0.001
GAD-7 score	6.1 ± 5.2	9.4 ± 5.1	3.4 ± 3.2	< 0.001
Adherence measures
MARS-5 score	18.7 ± 3.2	17.2 ± 2.8	21.6 ± 2.9	< 0.001
Appointment adherence	76.4 ± 18.2	68.2 ± 19.3	83.0 ± 14.6	< 0.001

HbA1c: Glycated hemoglobin; BCVA: Best-corrected visual acuity; LogMAR: Logarithm of the minimum angle of resolution; CST: Central subfield thickness; DME: Diabetic macular edema; Anti-VEGF: Anti-vascular endothelial growth factor; PHQ-9: Patient Health Questionnaire-9; GAD-7: Generalized Anxiety Disorder-7; MARS-5: Medication Adherence Report Scale-5.

Prevalence of depression and anxiety

Depression symptoms (PHQ-9 score ≥ 5) were present in 58 patients (44.6%), with 22 (16.9%) meeting criteria for moderate-severe depression (PHQ-9 ≥ 10). Anxiety symptoms (GAD-7 score ≥ 5) were identified in 49 patients (37.7%), with 18 (13.8%) having moderate-severe anxiety (GAD-7 ≥ 10). Comorbid depression and anxiety symptoms were present in 36 patients (27.7%) (Table 2).

Table 2 Distribution of depression and anxiety severity among patients with diabetic macular edema, n (%).

Severity category	Depression (PHQ-9)	Anxiety (GAD-7)
Minimal (0-4)	72 (55.4)	81 (62.3)
Mild (5-9)	36 (27.7)	31 (23.8)
Moderate (10-14)	14 (10.8)	12 (9.2)
Moderately severe/severe (≥ 15)	8 (6.2)	6 (4.6)
Clinical threshold (≥ 10)	22 (16.9)	18 (13.8)
Any symptoms (≥ 5)	58 (44.6)	49 (37.7)
Comorbid depression and anxiety	36 (27.7)

PHQ-9: Patient Health Questionnaire-9; GAD-7: Generalized Anxiety Disorder-7.

Treatment adherence patterns

The mean MARS-5 score was 18.7 ± 3.2, with only 31 patients (23.8%) demonstrating good adherence (score ≥ 23). Poor adherence (score < 17) was observed in 56 patients (43.1%), while 43 patients (33.1%) showed moderate adherence (score 17-22). The mean appointment adherence rate was 76.4 ± 18.2%, with 42 patients (32.3%) having adherence rates < 70% (Table 3).

Table 3 Medication and appointment adherence by psychological status, n (%).

Adherence measure	Depression status			Anxiety status
	Present (n = 58)	Absent (n = 72)	P value	Present (n = 49)	Absent (n = 81)	P value
MARS-5 score
mean ± SD	17.2 ± 2.8	21.6 ± 2.9	< 0.001	17.2 ± 3.1	21.1 ± 2.8	< 0.001
Poor (< 17),	34 (58.6)	22 (30.6)	0.001	28 (57.1)	28 (34.6)	0.011
Moderate (17-22)	18 (31.0)	25 (34.7)		15 (30.6)	28 (34.6)
Good (≥ 23)	6 (10.3)	25 (34.7)		6 (12.2)	25 (30.9)
Appointment adherence
mean ± SD	68.2 ± 19.3	83.0 ± 14.6	< 0.001	69.8 ± 19.8	80.5 ± 16.1	0.001
< 70%	26 (44.8)	16 (22.2)	0.006	24 (49.0)	18 (22.2)	0.001
70%-89%	22 (37.9)	28 (38.9)		16 (32.7)	34 (42.0)
≥ 90%	10 (17.2)	28 (38.9)		9 (18.4)	29 (35.8)
Injection interval consistency
CV of intervals	38.4 ± 12.6	28.2 ± 10.3	< 0.001	36.8 ± 13.2	30.1 ± 11.4	0.003
Reasons for non-adherence
Forgetfulness	42 (72.4)	28 (38.9)	< 0.001	36 (73.5)	34 (42.0)	< 0.001
Cost concerns	22 (37.9)	18 (25.0)	0.111	19 (38.8)	21 (25.9)	0.120
Transportation issues	18 (31.0)	12 (16.7)	0.049	16 (32.7)	14 (17.3)	0.042
Fear of injections	28 (48.3)	14 (19.4)	< 0.001	31 (63.3)	11 (13.6)	< 0.001

CV: Coefficient of variation; MARS-5: Medication Adherence Report Scale-5.

Association between psychological factors and adherence

Significant negative correlations were observed between PHQ-9 scores and both MARS-5 scores (r = -0.42, P < 0.001) and appointment adherence (r = -0.38, P < 0.001). Similarly, GAD-7 scores were inversely correlated with MARS-5 scores (r = -0.35, P < 0.001) and appointment adherence (r = -0.33, P < 0.001) (Table 4). Further correlation analysis demonstrated robust relationships across multiple domains. The correlation between PHQ-9 and GAD-7 scores was strong (r = 0.68, P < 0.001), indicating substantial comorbidity. Both MARS-5 scores and appointment adherence showed significant positive correlation (r = 0.45, P < 0.001), suggesting these measures capture related but distinct aspects of treatment adherence. Visual outcomes (BCVA and CST change) were significantly correlated with both psychological measures and adherence indicators, supporting the hypothesized pathway from psychological distress, through reduced adherence to worse clinical outcomes. The NEI VFQ-25 composite score showed the strongest correlations with PHQ-9 (r = -0.48, P < 0.001) and GAD-7 (r = -0.41, P < 0.001), underscoring the profound impact of psychological comorbidities on vision-related quality of life beyond their effects on clinical parameters.

Table 4 Correlation matrix of psychological symptoms, adherence measures, and clinical outcomes.

Variable	1	2	3	4	5	6	7
1 PHQ-9 score	1.00
2 GAD-7 score	0.68c	1.00
3 MARS-5 score	-0.42c	-0.35c	1.00
4 Appointment adherence	-0.38c	-0.33c	0.45c	1.00
5 BCVA change (logMAR)	0.28b	0.22a	-0.31c	-0.36c	1.00
6 CST change (μm)	0.24b	0.19a	-0.26b	-0.29c	0.42c	1.00
7 NEI VFQ-25 composite	-0.48c	-0.41c	0.39c	0.43c	-0.52c	-0.38c	1.00

^aP < 0.05.

^bP < 0.01.

^cP < 0.001.

PHQ-9: Patient Health Questionnaire-9; GAD-7: Generalized Anxiety Disorder-7; MARS-5: Medication Adherence Report Scale-5; BCVA: Best-corrected visual acuity; CST: Central subfield thickness; NEI VFQ-25: National Eye Institute Visual Function Questionnaire-25; LogMAR: Logarithm of the minimum angle of resolution.

Predictors of poor treatment adherence

In multivariate logistic regression analysis, PHQ-9 score ≥ 10 [odds ratio = 3.42, 95% confidence interval (CI): 1.68-6.95, P = 0.001] and GAD-7 score ≥ 10 (odds ratio = 2.86, 95%CI: 1.35-6.04, P = 0.006) were significant predictors of poor medication adherence, after adjusting for demographic and clinical factors (Table 5).

Table 5 Multivariable logistic regression analysis of factors associated with poor medication adherence (Medication Adherence Report Scale-5 < 17).

Variable	Unadjusted OR (95%CI)	P value	Adjusted OR (95%CI)	P value
PHQ-9 ≥ 10	4.28 (2.18-8.42)	< 0.001	3.42 (1.68-6.95)	0.001
GAD-7 ≥ 10	3.65 (1.78-7.48)	< 0.001	2.86 (1.35-6.04)	0.006
Age (per 10 years)	0.82 (0.68-0.98)	0.032	0.88 (0.72-1.08)	0.214
Men	0.76 (0.42-1.38)	0.365	0.82 (0.43-1.56)	0.548
Diabetes duration > 15 years	1.68 (0.92-3.06)	0.092	1.42 (0.74-2.72)	0.292
HbA1c > 8%	2.14 (1.18-3.88)	0.012	1.86 (0.98-3.52)	0.058
Bilateral DME	1.92 (1.06-3.48)	0.032	1.74 (0.92-3.29)	0.089
Low education (< 12 years)	1.78 (0.98-3.24)	0.059	1.65 (0.87-3.14)	0.126
Living alone	2.35 (1.22-4.52)	0.011	2.12 (1.05-4.28)	0.036
> 5 medications daily	1.94 (1.07-3.52)	0.029	1.68 (0.88-3.21)	0.116

OR: Odds ratio; CI: Confidence interval; PHQ-9: Patient Health Questionnaire-9; GAD-7: Generalized Anxiety Disorder-7; DME: Diabetic macular edema; HbA1c: Glycated hemoglobin.

Visual function and quality of life outcomes

Patients with depression had significantly lower NEI VFQ-25 composite scores than those without depression (58.3 ± 14.2 vs 72.6 ± 12.8, P < 0.001). All subscales (except ocular pain) showed significant between-group differences (Table 6).

Table 6 National Eye Institute Visual Function Questionnaire-25 scores by depression and anxiety status, mean ± SD.

NEI VFQ-25 subscale	Depression status			Anxiety status
	Present (n = 58)	Absent (n =72)	P value	Present (n = 49)	Absent (n = 81)	P value
Composite score	58.3 ± 14.2	72.6 ± 12.8	< 0.001	60.2 ± 14.8	70.8 ± 13.6	< 0.001
General health	42.2 ± 18.6	56.3 ± 19.2	< 0.001	44.4 ± 19.8	54.6 ± 19.3	0.004
General vision	48.6 ± 16.4	62.8 ± 15.2	< 0.001	50.6 ± 17.2	60.8 ± 16.0	0.001
Ocular pain	72.4 ± 20.3	76.0 ± 18.6	0.292	71.9 ± 20.8	75.9 ± 18.6	0.258
Near activities	52.8 ± 18.9	68.4 ± 16.2	< 0.001	54.7 ± 19.2	66.5 ± 17.3	< 0.001
Distance activities	54.3 ± 19.7	70.2 ± 17.3	< 0.001	56.1 ± 20.3	68.4 ± 18.2	< 0.001
Social functioning	56.9 ± 21.4	76.4 ± 18.2	< 0.001	58.7 ± 22.1	74.2 ± 19.6	< 0.001
Mental health	45.3 ± 19.8	68.2 ± 17.4	< 0.001	47.8 ± 20.6	65.4 ± 19.2	< 0.001
Role difficulties	48.7 ± 22.6	69.8 ± 20.3	< 0.001	51.2 ± 23.4	67.3 ± 21.8	< 0.001
Dependency	52.6 ± 23.8	74.3 ± 21.2	< 0.001	54.8 ± 24.6	71.9 ± 22.4	< 0.001
Driving1	58.2 ± 24.3	72.6 ± 21.8	0.002	60.4 ± 24.8	70.8 ± 22.6	0.026
Color vision	68.5 ± 22.1	82.3 ± 18.4	< 0.001	70.4 ± 22.6	80.9 ± 19.2	0.006
Peripheral vision	62.9 ± 21.7	78.5 ± 19.3	< 0.001	64.8 ± 22.3	76.9 ± 20.1	0.002

¹Driving subscale calculated only for patients who reported driving (n = 86).

NEI VFQ-25: National Eye Institute Visual Function Questionnaire-25.

Mediation and moderation analyses

Mediation analysis revealed that treatment adherence partially mediated the relationship between depression and visual outcomes. The indirect effect of depression on BCVA change through medication adherence was significant (β = 0.082, 95%CI: 0.034-0.138, P = 0.002), accounting for 28.6% of the total effect (Table 7). Social support moderated the relationship between depression and medication adherence (interaction term: Β = -0.18, P = 0.012), with stronger negative effects of depression observed in patients with low social support.

Table 7 Mediation analysis of the relationship between depression/anxiety and visual outcomes.

Pathway	Effect	SE	95%CI	P value	Mediated
Depression - BCVA change
Total effect	0.287	0.062	0.165-0.409	< 0.001	-
Direct effect	0.205	0.058	0.091-0.319	< 0.001	71.4%
Indirect effect via MARS-5	0.082	0.026	0.034-0.138	0.002	28.6%
Depression - NEI VFQ-25
Total effect	-14.32	2.48	-19.24 to -9.40	< 0.001	-
Direct effect	-10.68	2.36	-15.36 to -6.00	< 0.001	74.6%
Indirect effect via MARS-5	-3.64	1.12	-5.92 to -1.48	0.001	25.4%
Anxiety - BCVA change
Total effect	0.224	0.058	0.110-0.338	< 0.001	-
Direct effect	0.162	0.054	0.056-0.268	0.003	72.3%
Indirect effect via appointment adherence	0.062	0.022	0.021-0.108	0.004	27.7%

SE: Standard error; CI: Confidence interval; BCVA: Best-corrected visual acuity; MARS-5: Medication Adherence Report Scale-5; NEI VFQ-25: National Eye Institute Visual Function Questionnaire-25.

DISCUSSION

This study provides comprehensive evidence regarding the substantial impact of depression and anxiety on treatment adherence among patients with DME. Our findings demonstrate that psychological comorbidities are highly prevalent in this population and are associated with substantial reductions in both medication adherence and appointment attendance, ultimately affecting visual outcomes and quality of life. The prevalence of depression and anxiety in our DME cohort aligns with previous studies in diabetic populations; however, rates were notably higher than those reported in general diabetes samples without vision-threatening complications[23]. This elevated prevalence likely reflects the additional psychological burden imposed by the threat of vision loss and the demanding treatment regimen required for DME management. The bidirectional relationship between diabetes and depression has been well-established with shared biological pathways, including hypothalamic-pituitary-adrenal axis dysregulation, inflammatory processes, and neurotransmitter alterations, contributing to this association[24]. The added stress of potential blindness and the need for frequent invasive procedures may further exacerbate psychological distress in patients with DME.

Our finding that depression severity was inversely correlated with medication adherence is consistent with extensive literature in chronic disease management[25]. Depressive symptoms can impair cognitive function, motivation, and self-care behaviors, creating barriers to optimal diabetes management. The mechanisms underlying this relationship are multifaceted, involving reduced self-efficacy, hopelessness regarding treatment outcomes, and impaired executive function that affects planning and organization of complex medication regimens[26]. Notably, we observed that mild depressive symptoms were associated with measurable decrements in adherence, suggesting that subclinical depression warrants attention in clinical practice. The association between anxiety and treatment adherence in our study was pronounced for appointment attendance and acceptance of intravitreal injections. This finding corroborates qualitative research, indicating that injection-related anxiety is a major barrier to DME treatment[27]. The anticipation of ocular discomfort, fear of complications, and anxiety regarding the injection procedure itself can lead to appointment avoidance and treatment discontinuation. Our data suggest that addressing injection-related anxiety through patient education, psychological support, and procedural modifications could potentially improve treatment adherence and outcomes.

The impact of psychological factors on visual function and quality of life extended beyond their effects on adherence. Patients with comorbid depression and anxiety reported significantly low scores across multiple domains of the NEI VFQ-25, including mental health, role difficulties, and social functioning. These findings align with previous research demonstrating that psychological distress amplifies the perceived impact of visual impairment on daily activities and social participation[28]. The relationship between psychological symptoms and vision-related quality of life appeared to be partially mediated by adherence behaviors, suggesting that interventions targeting both psychological well-being and adherence could have synergistic effects on patient-reported outcomes.

Our analysis identified several modifiable factors that influenced the relationship between psychological symptoms and adherence. Social support emerged as a significant protective factor, with patients reporting strong family support demonstrating better adherence than those with weak or no family support, despite psychological distress. This finding underscores the importance of involving family members and caregivers in DME management, particularly for patients with psychological comorbidities[29]. In addition, healthcare provider communication style and the quality of the patient-provider relationship moderated the impact of depression on adherence, suggesting that enhanced communication training for eye care professionals could improve outcomes in psychologically vulnerable patients.

The economic implications of our findings are substantial. Poor adherence due to psychological factors leads to suboptimal treatment outcomes, necessitating more intensive treatment, increased healthcare utilization, and higher rates of vision loss with associated disability costs. Cost-effectiveness analyses in other chronic diseases have demonstrated that integrated care models addressing both medical and psychological needs can be cost-saving in the long term[30]. Similar DME management-specific economic evaluations are warranted to inform healthcare policy and resource-allocation decisions.

Some limitations warrant consideration. First, the retrospective design precludes causal inference regarding the directionality of relationships between psychological symptoms and adherence behaviors. Second, self-report adherence measures such as MARS-5 may overestimate actual behaviors due to social desirability bias, although we attempted to mitigate this by also examining objective appointment attendance data. Third, the single-center setting at a tertiary hospital in Shanghai limits generalizability to other healthcare settings and geographic regions. Our patient population may represent more severe cases with complex comorbidities compared with community-based samples; moreover, region-specific cultural factors may influence psychological responses and help-seeking behaviors. Fourth, we did not systematically evaluate the potential differential effects of specific anti-VEGF agents (ranibizumab, aflibercept, or conbercept) on treatment adherence, nor did we assess injection-related fear and other psychological barriers beyond depression and anxiety. Fifth, regarding missing data, our study included only patients with complete psychological assessments and adherence data over 12 months, potentially introducing selection bias. The proportion of excluded patients due to incomplete data was approximately 18% (28 of 158 screened patients), primarily owing to missing follow-up psychological assessments. We did not employ imputation methods, which may limit the representativeness of our findings. Sixth, we did not assess longitudinal changes in psychological symptoms or evaluate the impact of psychological interventions on adherence outcomes. Future prospective studies with larger, more diverse samples and inclusion of psychological intervention components are needed to address these limitations.

CONCLUSION

This study demonstrates that depression and anxiety are highly prevalent among patients with DME and considerably impact treatment adherence, visual outcomes, and quality of life. The inverse relationship between psychological symptom severity and adherence behaviors highlights the need for integrated care approaches that address both ophthalmological and psychological aspects of DME management. Routine psychological screening, enhanced provider-patient communication, and targeted psychological interventions represent critical components of comprehensive DME care. With the growing global burden of diabetic eye disease, addressing psychological barriers to treatment adherence will be essential for optimizing visual outcomes and preserving quality of life in this vulnerable population. Based on our findings, we propose several practical recommendations for optimizing DME management. First, routine psychological screening using validated instruments (PHQ-9 and GAD-7) should be integrated into standard ophthalmology practice for patients with DME, with referral pathways established for those meeting clinical thresholds. Second, patient education programs should specifically address injection-related anxiety through demonstration videos, peer support groups, and procedural modifications to minimize discomfort. Third, collaborative care models involving ophthalmologists, endocrinologists, and mental health professionals should be implemented to provide integrated treatment, addressing both physical and psychological aspects of disease management. Fourth, reminder systems via text messaging or mobile applications may help address forgetfulness, which was the most common reason for non-adherence in our cohort. Fifth, for patients with transportation barriers or cost concerns, telemedicine follow-ups and financial assistance programs should be explored to improve accessibility. Future research should focus on longitudinal assessment of psychological symptom trajectories and their impact on long-term visual outcomes, evaluation of specific psychological interventions (cognitive-behavioral therapy and mindfulness-based approaches) tailored for patients with DME, investigation of biological mechanisms linking depression to diabetic complications, including neuroinflammatory pathways, and cost-effectiveness analysis of integrated care models compared with standard ophthalmological care alone. Such research will inform evidence-based policies and clinical guidelines for comprehensive DME management in the era of precision medicine.