FormalPara Key Summary Points

Why carry out this study?

Sleep plays a prominent role in the pathogenesis of vision-threatening diabetic retinopathy.

The study aimed to assess the association of poor sleep quality with vision-threatening diabetic retinopathy in a population of people with diabetes.

What was learned from the study?

Sleep quality was poor in 71.90% of patients with vision-threatening diabetic retinopathy, significantly higher compared with those patients without sight-threatening diseases.

The study showed a novel association of vision-threatening diabetic retinopathy with sleep disruption in a diabetic cohort, indicating a possible connection between the two potential areas (the brain and the retina).

Ophthalmologists and eye care professionals may wish to consider therapy of both sleep disruption and vision-threatening diabetic retinopathy in a clinical setup.

Introduction

Vision-threatening diabetic retinopathy (VTDR), a long-term microvascular complication of diabetes in the eye, is a major cause of preventable blindness in working-age adults [1, 2]. Vision-threatening diabetic retinopathy can be manifested as severe background diabetic retinopathy (DR), proliferative DR, and diabetic macular edema, which can occur at any stage of DR [3].

The end stage of VTDR is retinal detachment, with no promising treatment currently available. Worldwide, about 3.7 million people with diabetes lost their sight due to VTDR [4]. Early assessment of the risk factors, such as sleeping quality, could be a potential alternative to prevent the diseases.

Sleep is a complex, highly organized biological process essential for the normal functioning of the human body [5]. It is achieved by the coordination of the retina and the thalamic area of the brain. Poor quality of sleep has been reported to have an effect on diabetes and its major microvascular complications such as diabetic retinopathy [6,7,8].

Two sides of the theory have been suggested, regarding the association of poor sleep quality with advanced diabetic retinopathy. The first side of the theory is that neuronal damage to the retinohypothalamic tract, due to diabetic retinopathy, could influence the circadian cycle and causes poor quality of sleep [9, 10]. Additionally, DR was also found to be linked with low melatonin production in the brain [11,12,13]. The other side of the theory is that sleep enhances the antioxidant capacity of the brain and retina, and plays a great role in the immune modulatory function. Thus, disturbed sleep behavior can cause oxidative stress and fuels uncontrolled inflammation in the retinal cells that leads to the development of advanced diabetic retinopathy [5, 14, 15].

Studies supporting the association of sleep with diabetic retinopathy is scarce. There have been a few studies attempted in various regions of the world. For instance, a cross-sectional survey by Tan et al. [16] investigated the association of sleep quality with diabetic retinopathy and reported that the overall quality of sleep was significantly related to VTDR. However, Dutta et al. [12], Morjaria et al. [17], and Fitrada et al. [18] observed a link between sleep quality and diabetic retinopathy regardless of severity.

Regarding the impact of poor sleep quality on diabetes and its major vascular complications, we are inquisitive about whether disturbed sleep quality has an impact on the advanced stage of diabetic retinopathy . We explored the association of poor sleep quality with VTDR in diabetic patients (with matched sleep durations) who followed a regular schedule at the retinal clinic of a comprehensive specialized center.

Methods

A hospital-based matched case–control study was conducted from 6 May 2022 to 6 September 2022. The study was conducted mainly during the summer, and the participants were recruited from the retina clinic at the University of Gondar Tertiary Eye Care and Training Center, the diagnostic and follow-up center in Ethiopia for patients with vitreoretinal disorders, and systemic co-morbidities, including diabetes and hypertension. The study excluded patients specifically with the following conditions: concomitant posterior uveitis syndrome, peripheral neuropathy, pruritic skin disorder, mental illness, thyroid diseases, chronic obstructive pulmonary diseases, severe kidney dysfunction, and cognitive impairment.

Sample Size Determination

The sample size was calculated with a pair-matched case–control study design sample size calculation in Epi Info version 7 [20]. Proportions were taken from a much similar study conducted in India [12]. The reference study consists of patients with non-DR and DR, with moderate and advanced diabetic retinopathy, that corresponds with VTDR for this intended study. Thus, the proportions taken from the previous study to calculate the sample size would not negatively represent the study population and affect the validity of our report, as the parameters represent the intended figures of the study . Accordingly, the proportion of pairs of cases and controls who were exposed was 17, the proportion of pairs of cases who were exposed and controls who were not exposed was 46, the proportion of pairs of cases who were not exposed and controls who were exposed was 17, and the proportion of pairs of cases and controls who were not exposed was 24. Accordingly, the computer-generated sample size was 63 discordant pairs.

Sampling Technique

For both cases and controls, recruitment of patients with diabetes (type 1 and type 2) was done consecutively after a complete examination in the retina clinic from 6 May 2022 to 6 September 2022. The cases and controls were matched by the duration of diabetes. The study was done in line with the relevant ethical proceedings pertaining to a human study, as stated in the Declaration of Helsinki.

Compliance to Ethics Guidelines

The study was reviewed by the Ethical Review Board of University of Gondar for ethical soundness, and finally a certificate of ethical approval (Reference. Vice President Research and Technology Transfer /05/839/2022) was obtained. Informed consent was received from participants for the clinical evaluation of VTDR and the administration of PSQ for the assessment of quality of sleep.

Ophthalmic Evaluation

We recruited cases, aged 18 years and older, diagnosed with VTDR and controls, aged 18 year and older, without VTDR. The diagnosis of VTDR was carried out by a senior ophthalmologist who is a vitreoretinal surgeon. The ophthalmologist examined patients under indirect slit-lamp biomicroscopy, and he applied the protocol of Early Treatment of Diabetic Retinopathy Study (ETDRS) [19] to grade severity of the retinopathy. Whenever there were confusing features of VTDR; exudates and or hemorrhages close to the macular area, the ophthalmologist utilized ocular coherence tomography (OCT) to define the cases of VTDR.

Definition of Risk Factors

Fasting blood sugar level was defined as poor when the current fasting blood glucose level was higher than 150 mg/dL and good when the level was 150 mg/dL and lower (Guideline of the American Diabetic Association). Hypertension in participants was marked “yes” if they were previously diagnosed and followed, and “no” if there was no previous hypertension diagnosis, as referred from the participants medical folder. The educational status of participants was categorized as “formal education” for participants who received formal education running from elementary school through to higher education, and “no formal education’” for participants who can read and write or do neither.

Assessment of Quality of Sleep

This was a case–control study of patients who were approached to report on their sleeping status over 1 month, preceding the study. A validated questionnaire was translated into the local language and was used to collect the data on the quality of sleep. Generally, the data collection tool was organized into three parts; part I involved a questionnaire on sociodemographic characteristics, part II involved a clinical data checklist form, and part III involved the validated Pittsburgh Sleep Quality Index (PSQI) questionnaire [21].

Quality of sleep was assessed by the PSQI questionnaire, which consists of 19 self-related questions combined to form seven component scores. Each component had a range of 0–3 points; a score of 0 indicating “no difficulty” and a score of 3 indicating “severe difficulty.” Seven component scores were then added to calculate the overall PSQI score that runs from 0 to 21; with 0 indicating “no difficulty” and 21 indicating “severe difficulty” in all sleep components. Poor quality of sleep was defined by an overall PSQI score of greater than five [22].

Data Analysis

The collected data were coded and entered into Epi Info 7, and then exported into SPSS version 20 for analysis. Pearson’s chi-squared and Fisher’s exact tests were done to detect significant differences in proportions of poor sleep quality between cases and controls. A two-tailed p value was reported and considered statistically significant when the value was < 0.05. An independent t-test was employed to compare the mean overall sleep quality among the two cohorts.

Duration of diabetes is a known and consistent factor for predicting VTDR [1]. Therefore, an attempt was made to make a strict (one-to-one) matching of cases and controls, with disease duration as a matching variable, to evaluate the effects of poor sleep quality on VTDR, while canceling the effects of confounders in the analysis. We inferred disease duration from the patients’ report on “how long they have had the disease,” and attempted to match a VTDR participant with diabetes, with a NVTDR patient with diabetes that had a similar disease duration.

The analysis was done with the conditional logistic regression model, the standard tool for the analysis of matched case–control studies. The modeling strategy was employed with the assumption of the Cox regression model. A matching number has been given to link each pair of participants with diabetes (cases and control), and the outcome status of the subject was defined and given a value of 0 for the control and 1 for the case.

Finally, a response variable (time) was created and given a value of 1 for the outcome of interest (poor quality of sleep) and 2 for good quality of sleep. By entering the time (response) variable in the time option, the outcome status (with a defined value of 1) in the status option, and matching number in the strata option, a Cox regression was performed.

A bivariable matched analysis was executed for each variable; those variables having p < 0.20 were included in the matched multivariable model. Adjusted matched odds ratios (mAORs) with 95% confidence intervals (CI) and p values were estimated for the multivariable matched analysis. Variables at p < 0.05 were considered statistically significant and independently associated with poor quality of sleep.

Results

Baseline Characteristics of Cases and Controls

A total of 126 (63 duration-matched pairs) with diabetes were included in this study. The mean age of diabetes was 52 ± 11.00 years in the case group and 58 ± 12.50 years in the control group.

The age range was 22–75 years and 30–88 years for cases and controls, respectively. The number of elderly patients (> 64 years) included was 10 in the case group versus 26 in the control group. The mean fasting blood sugar level was 152.20 ± 44 in the case group and 148 ± 47.40 in the control group. The baseline characteristics were comparable between the two groups, except for occupational status (Fisher’s exact test, p = 0.01) (Table 1).

Table 1 Baseline characteristics of cases and controls, 2022 (n = 126)
Full size table

Quality of Sleep for VTDR Compared with Controls

Generally, the overall quality of sleep was poor in 46 (80.70%) of patients with any degree of diabetic retinopathy versus 11 (19.30%) of patients without any retinopathy. The overall quality of sleep was poor among 41 (71.90%) of the cases, significantly higher compared with the 16 (28.10) of the controls (p < 0.0001) (Table 2).

Table 2 Sleep quality parameters and overall sleep quality among cases and controls, 2022 (n = 126)
Full size table

An independent-sample t-test was conducted to compare the mean global PSQ score for cases and controls. There was significant difference [t(degree of freedom = 105.80, p < 0.00] in the scores, with the mean score for cases (M = 7.10, SD = 4.30) higher than the control group (M = 3.60, SD = 2.70). The magnitude of the difference in the means (mean difference = 3.50, 95% CI: 2.25–4.75) was significant (Fig. 1).

Fig. 1
figure 1

Comparison of global means PSQ Scores in patients with DM with VTDR and NVTDR at retina clinic of Comprehensive Specialized Hospital at University of Gondar Northwest Ethiopia, 2022. VTDR vision-threatening diabetic retinopathy, NVTDR non-vision-threatening diabetic retinopathy, M mean, MD mean difference, asterisk: significant p value less than 0.001 and PSQ Pittsburgh Sleep Quality index

Full size image

Considering the details of sleep quality parameters, our study demonstrated significant differences between the two cohorts, except for the use of sleep medication. The subjective sleep latency was very good in only 16 (29.60%) of the cases versus 38 (70.40%) of the controls. Sleep latency was 15 min or less for 20 (40%) of cases versus 30 (60%) of controls. The duration of sleep was greater than 7 h for 40 (39.60%) of cases, versus 61 (60.40) of controls. Low sleep efficiency (< 65%) was noted for 15 (100%) of cases, against 51 (60.70%) of controls with high sleep efficiency (> 85%). High sleep disturbance was observed in 2 (100%) of cases. The daytime dysfunction was not problematic for 21 (35.60%) of cases versus 38 (64.40%) of controls. The use of sleep medication was comparable between the two cohorts; sleep without medication was noted among 59 (49.20%) of cases versus 61 (50.80%) of controls (p = 0.34) (Table 2).

Our bivariable analysis revealed that age greater than 55 years [crude odds ratio (COR) 0.38, 95% CI: 0.15–0.96], government employee (COR 9.03, 95% CI: 2.04–40.0), and poor sleep quality (COR 5.17, 95% CI: 2.16–12.38) were significantly associated with VTDR in diabetes. However, according to the multivariable regression, poor sleep quality alone significantly predicted the formation of VTDR in the given diabetic cohort [adjusted odds ratio (AOR) 8.48, 95% CI: 2.42–29.67] (Table 3).

Table 3 Bivariable and multivariable conditional logistic regression analysis for determinants of poor sleep quality in participants with VTDR 2022 (n 126)
Full size table

Discussion

Evidence has been emerging on the effect of sleep disruption on the development of diabetic retinopathy [6,7,8]. However, to the best of the authors’ knowledge, this is the first duration-matched case–control study to explore the association of VTDR with poor sleep quality among patients with diabetes at a retina clinic of a comprehensive specialized center.

In this study, it was observed that the majority of participants with vision-threatening diabetic retinopathy had poor sleep quality, compared with participants without VTDR [diabetes with VTDR, 41 (71.90%) versus diabetic controls, 16 (28.10)]. Poor quality of sleep also significantly predicted the presence of VTDR in a given diabetic cohort, if the other variables were kept constant . The present finding is consistent with the previous study conducted by Tan et al. [16], who reported that poor quality of sleep was significantly associated with presence and progression of VTDR.

Nevertheless, studies by Dutta et al. [12], Morjaria et al. [17], and Fitrada et al. [18] reported comparable quality of sleep among patients with severe DR and those without DR . The recruitment of normal individuals as a control, fewer patients with VTDR, strict exclusion criteria, and effects of potential confounders, such as duration of diabetes and comorbidities, in the previous studies could reflect the variation. In particular, Dutta et al. [12] and Fitrada et al. [18] investigated the poor quality of sleep, mostly on patients with type 2 diabetes, with restricted duration of the disease, age groups, fasting blood glucose level, and participants free of comorbidities, such as hypertension, that may increase severity of retinopathy influenced by the quality of sleep.

Additionally, a study by Dutta et al. [12] was mainly performed on moderate types of DR (participants are limited in terms of VTDR). Similarly, Morjaria et al. [17] included normal (nondiabetic) patients as a control group, and participants with mode-related complaints were not allowed in either cohort. On the other hand, our study was based on a diabetic group, and we did not exclude patients who had mode-related complaints.

Including nondiabetic patients in the cohort could yield a better quality of sleep, as normal individuals could have better quality of sleep than patients with diabetes [23]. Additionally, patients with mood-related complaints could have frequent instances of sleepiness, which could alter the quality of sleep score obtained from the previous studies.

In terms of specific sleep quality characteristics, compared with diabetic participants with non-VTDR, diabetic participants with VTDR claimed to have very bad sleep quality, 2 (100%) had slept less than 5 h a day [diabetes with non-VTDR 11 (91.70%), versus diabetes with VTDR 1 (8.30%)], had poor sleep efficiency [diabetes with non-VTDR, 15 (100%)], and frequent dysfunction during the daytime [diabetes with non VTDR, 17 (35.40%) versus diabetes with VTDR, 31(64.60%)].

In a population-based sample of 1231 Asian adults, Tan et al. [16] noted a significant association between a longer duration of sleep and VTDR. On the contrary, Jee et al. [24] reported a significant association between any duration of sleep with VTDR . Jee et al. [24] investigated the association between the occurrence of VTDR and sleep variability in the older diabetes population of over 40 years, when sleep variability could be limited to have a significant variation (most of the study subjects slept around the mean duration of 6–8 h).

The association between short duration of sleep and VTDR could be related to oxidative damage and resultant inflammation in the retinal cells [5], while the association between long duration of sleep with VTDR could not be explained via the disturbed circadian physiology or inflammation, thus it remains elusive.

Dutta et al. [12] recently explored the association of poor quality of sleep parameters with VTDR in patients with type 2 diabetes (70 cases and 70 controls), and reported a similar finding to the present study—daytime dysfunction was associated with the severity of diabetic retinopathy. However, Morjaria et al. [17] found a comparable quality of sleep parameters, with varying severity of DR. The involvement of nondiabetic patients in the former study could be responsible for the difference.

Limitations

The study employed a duration-matched case–control study design to better control confounders and demonstrate the effects of poor sleep quality on VTDR. Hence, the study retrieved information from the patients’ most recent 1-month exposure. The study was unable to establish causal relation between VTDR and poor quality of sleep. The direction of causality remains unclear and requires a better study design, with a large representative sample. In addition, hemoglobin A1C could have provided a better estimate of the glycemic state of participants, compared with the fasting blood sugar level used in the present study. Finally, sleep is typically measured with polysomnography; thus, an inability to employ the instrument could lead to misclassification error.

Conclusions

In this study, a significant association was observed between overall poor sleep quality and VTDR in a diabetic cohort. Subsequently, the features of poor sleep quality showed significant variation between VTDR and non-VTDR sufferers. The study emphasizes the need to consider management of both conditions to improve outcomes for patients with diabetic retinopathy. Further studies with a longitudinal design and objective sleep assessment are needed to establish a causal relationship between sleep disruption and vision-threatening diabetic retinopathy.