Impact of Obstructive Sleep Apnea on Diabetic Retinopathy Progression and Systemic Complications

Abstract

Purpose:

Evaluate the risk of diabetic retinopathy progression and systemic vascular events, including death, in patients with non-proliferative diabetic retinopathy (NPDR) with obstructive sleep apnea (OSA).

Design:

Retrospective cohort study.

Methods:

Electronic chart query using TriNetX (Cambridge, MA, USA), an electronic health records network comprising data from over 124 million patients. Patients with NPDR with and without OSA were identified. Patients were excluded if they had history of proliferative disease (PDR), diabetic macular edema (DME), or prior ocular intervention (intravitreal injection, laser, or pars plana vitrectomy). Propensity score matching was performed to control for baseline demographics and comorbidities. Rate of progressing to vision threatening complications (VTCs), need for ocular intervention, and systemic events was measured at 1, 3, and 5 years.

Results:

11,931 patients in each group were analyzed after propensity score matching. There was elevated risk of PDR in the OSA cohort at 1 (RR: 1.34, P<0.001), 3 (RR: 1.31, P<0.001), and 5 years (RR: 1.28, P<0.001). There was elevated risk of DME in the OSA group at all time points: 1 (RR: 1.31, P<0.001), 3 (RR: 1.19, P<0.001), and 5 years (RR: 1.18, P<0.001). With respect to ocular interventions, there was an increased risk of intravitreal injection in OSA patients at 1 (RR: 1.59, P<0.001), 3 (RR: 1.58, P<0.001), and 5 years (RR: 1.54, P<0.001), and similar trends were noted with laser photocoagulation, but not vitrectomy. Regarding systemic events, NPDR patients with OSA had a greater risk of stroke (1 year RR: 1.80, P<0.001; 3 year RR: 1.56, P<0.001; 5 year RR: 1.49, P<0.001), myocardial infarction (1 year RR: 1.51, P<0.001; 3 year RR: 1.46, P<0.001; 5 year RR: 1.43, P<0.001), and death (1 year RR: 1.31, P<0.001; 3 year RR: 1.19, P<0.001; 5 year RR: 1.15, P<0.001).

Conclusions:

There is an increased rate of DR progression to VTCs, need for ocular intervention, and systemic complications, including death, for patients with OSA. We emphasize the need for improved screening measures of patients with NPDR and potential OSA.

Table of Contents Statement:

In this cohort study that included 9,367 patients with diabetic retinopathy and obstructive sleep apnea, there was an increased rate of vision threatening complications and need for ocular intervention through 5 years of follow-up when compared to a matched cohort of diabetic retinopathy patients without obstructive sleep apnea. There was additionally an increased risk of experiencing systemic cardiovascular complications, including death, in patients with diabetic retinopathy and obstructive sleep apnea during the study period.

Introduction

Diabetic retinopathy (DR) and its sequelae, retinal neovascularization leading to proliferative disease (PDR) and diabetic macular edema (DME), are a leading cause of preventable blindness in the working age population worldwide.^1,2 Development and progression of DR has been associated with numerous host risk factors including duration of diabetes, poor glycemic control, uncontrolled blood pressure, and dyslipidemia.³ In recent years, studies have emerged suggesting an association between DR and obstructive sleep apnea (OSA),⁴ a chronic sleep-related breathing disorder characterized by recurrent partial or complete cessation of airflow due to upper airway obstruction during sleep that results in intermittent nighttime hypoxia, hypercapnia, sleep fragmentation, and increased sympathetic output.⁵

OSA is estimated to affect an estimated 22% of men and 17% of women globally, with incidence figures increasing rapidly over time due to the concurrent rise in obesity.^4,6 Similarly, OSA tends to be more prevalent in patients with diabetes, affecting between 58-86% of afflicted individuals, and has been implicated as an independent factor to developing insulin resistance and consequent poor glucose control.^7,8 Within the retina, both conditions share pathophysiological overlap: inducing inflammation, oxidative stress, and tissue ischemia, ultimately resulting in upregulation of various pathways, cytokines, and growth factors (i.e. vascular endothelial growth factor (VEGF)).^9-12 Thus, it seems reasonable to deduce that OSA may potentiate the progression of DR to vision threatening complications (VTC).

Using a large aggregated electronic health record (EHR) dataset, our study aimed to ascertain the risk of DR progression and development of systemic vascular events, including death, in patients with non-proliferative diabetic retinopathy (NPDR) at baseline with and without known OSA.

Methods

A retrospective cohort study was conducted utilizing the TriNetX Health Research Network (Cambridge, MA, USA), which is a federated health research network comprising of numerous large health care organizations (HCO) within the United States and abroad. The current version of the network aggregates deidentified EHR data of more than 124 million patients within 80 HCOs spanning across nine countries, including 60 HCOs within the United States. TriNetX, LLC is compliant with the Health Insurance Portability and Accountability Act (HIPPA) rule and is certified to the ISO 27001:2013 standard. As this study only contained patient data displayed in an aggregate and deidentified form, it was determined to be exempted from Institutional Review Board approval.

The most updated data analyzed in this study was collected on October 29, 2023 from the TriNetX platform, which provided access to electronic medical records (diagnoses, procedures, medications, laboratory values, and genomic information). At the time of data querying, 124,043,499 patients were included in the TriNetX database. Our study used a retrospective cohort design by reviewing EHR from January 1, 2003 through October 29, 2023, and followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for reporting cohort studies.¹³

Two patient cohorts were identified for analyses. Each group included patients with pre-existing NPDR without DME or PDR, which were identified by their respective International Classification of Diseases, Tenth Revision (ICD-10) codes, or the previous version’s correlate code (ICD-9 prior to October 1, 20015 whenever not stated): Type 1 DM with mild (E10.329*), moderate (E10.339*), or severe (E10.349*) NPDR, and Type 2 DM with mild (E11.329*), moderate (E11.339*), or severe (E11.349*) NPDR. Patients were then subclassified based on whether they had a concurrent diagnosis of OSA (G47.33) combined with either confirmatory polysomnography testing with initiation of continuous positive airway pressure (CPAP) therapy or bilevel ventilation (Current Procedural Terminology (CPT) code: 95811) or documented CPAP management (CPT: 94660). Patients were required to have at minimum 1 year of history within the TriNetX system prior to the index date to meet eligibility for inclusion. Patients were excluded if they had any of the following at the index date: prior history of PDR (E10.35, E11.35, H43.1), prior diagnosis of DME (E10.321, E10.331, E10.341, E11.321, E11.331, E11.341) previous ocular intervention (intravitreal injection, laser photocoagulation, or surgical pars plana vitrectomy (PPV)). Patients were also removed from analyses if they developed retinovascular/exudative diseases that could confound the need for intravitreal injections: neovascular age-related macular degeneration (AMD; H35.32), branch retinal vein occlusion with cystoid macular edema (H34.83), or central retinal vein occlusion with cystoid macular edema (H34.81). Additionally, patients were excluded from the comparator control arm if they met the aforementioned criteria to be subsequently diagnosed with OSA during the study period. Finally, patients were excluded from consideration if they had any previous history of stroke (I63) or myocardial infarction (I21, I22) prior to the index date.

Baseline demographics such as age, sex, and race were recorded. Sex and race were determined based on the presence of these designations within the EHR. As the data is from a range of HCOs, there may be variation if self-reported gender and race versus institutionally classified sex and race was utilized. Other concomitant diagnoses were queried, including hypertension (I10-I1A), hyperlipidemia (E78), coronary artery disease (I25), chronic kidney disease (N18), overweight/obesity (E66), hypothyroidism (E03), polycystic ovary syndrome (E28.2), and nicotine dependence (F17) as a marker of smoking status.

Given that factors affecting the likelihood of a patient having OSA may also affect the likelihood of the same patient developing PDR, propensity score matching was included to address this potential bias. We included the following covariates in the propensity score model: age, sex, race, hypertension, hyperlipidemia, coronary artery disease, chronic kidney disease, overweight/obesity, hypothyroidism, polycystic ovary syndrome, nicotine dependence, hemoglobin a1c (HbA1c), body mass index (BMI), use of insulin, and use of novel hypoglycemics (glucagon-like peptide 1 (GLP-1) receptor agonists and sodium-glucose cotransporter-2 (SGLT-2) inhibitors). Propensity score matching was performed using the TriNetx built-in analysis platform (1:1 matching by nearest neighbor greedy matching algorithm with a caliper of 0.25 standard deviations).

The primary outcome measured was progression to vision-threatening complications (VTC: PDR or DME) in each group during the study period. Secondary outcome measures were need for ocular intervention to treat DR as reported by their respective CPT codes: intravitreal injections (CPT: 67028), panretinal laser photocoagulation (CPT: 67228), and/or surgical PPV (CPT: 67040, 67113). Additional secondary outcome measures were rates of systemic events, namely stroke, myocardial infarction (MI), or death during the study period. Outcomes were measured at the following follow-up periods: 1 year, 3 years, and 5 years after the index date. Patients were required to have at least 1 year of follow-up within the TriNetX system to be included for analyses. If patients experienced death prior to this interval, they were still included in the reporting. Univariate analyses were performed and risk ratios (RR) with 95% confidence intervals (CI) as well as risk differences (RD) were calculated to compare risk of ocular and systemic outcomes between the two cohorts. Significance tests were 2-sided and paired. P values were not adjusted for multiple analyses.

Results

Baseline Demographics

Among patients with treatment-naïve NPDR at baseline and with at least 1 year of follow-up recorded, 12,321 were identified with OSA at the index date, compared to 128,961 individuals with baseline NPDR without OSA. A total of 11,931 patients in each group were then further analyzed after propensity score matching (Table 1).

Table 1.

Baseline Characteristics of NPDR patients with and without OSA before and after Propensity Score Matching.

	Before PSM			After PSM
	NPDR with OSA (n=12,321)	NPDR without OSA (n=128,961)	P value	NPDR with OSA (n=11,931)	NPDR without OSA (n=11,931)	P value
Age, years (Mean ± SD)	67.9 ± 11.1	67.6 ± 13.8	0.11	67.8 ± 11.2	68.0 ± 12.6	0.20
Sex (%)
Female	48.0%	48.7%	<0.001	48.6%	48.1%	0.41
Male	50.5%	45.8%	<0.001	49.8%	50.5%	0.30
Race (%)
White	56.7%	51.5%	<0.001	56.6%	57.3%	0.31
Black	25.5%	20.2%	<0.001	25.3%	24.8%	0.38
Asian	3.2%	5.4%	<0.001	3.2%	3.2%	0.85
American Indian/Alaska Native	0.4%	0.5%	0.28	0.5%	0.3%	0.18
Other Race	2.4%	4.1%	<0.001	2.5%	2.4%	0.80
Unknown Race	10.3%	16.0%	<0.001	10.4%	10.5%	0.92
Ocular & Systemic Comorbidities (%)
Severe NPDR	4.5%	5.8%	<0.001	4.6%	4.1%	0.10
Hypertension	96.6%	73.5%	<0.001	96.4%	96.4%	0.73
Hyperlipidemia	87.2%	54.6%	<0.001	86.8%	86.9%	0.86
Coronary Artery Disease	48.2%	19.5%	<0.001	46.6%	47.2%	0.34
Chronic Kidney Disease	45.7%	21.3%	<0.001	44.1%	44.6%	0.46
Overweight/Obesity	84.1%	31.6%	<0.001	83.6%	82.9%	0.14
Hypothyroidism	23.1%	14.2%	<0.001	22.6%	22.8%	0.68
Polycystic Ovary Syndrome	1.4%	0.6%	<0.001	1.4%	1.5%	0.67
Nicotine Dependence	23.4%	11.4%	<0.001	22.4%	22.8%	0.40
Laboratory Values
HbA1c, % (Mean +/− SD)	7.96 ± 1.95 (n=11,082 (89.9%))	8.05 ± 1.96 (n=79,060 (61.3%))	<0.001	7.96 ± 1.96 (n=10,692 (89.6%))	8.07 ± 1.93 (n=10,695 (89.6%))	<0.001
BMI, kg/m2 (Mean ± SD)	37.9 ± 8.6 (n=8,754 (71.0%))	31.4 ± 7.2 (n=65,802 (51.0%))	<0.001	37.9 ± 8.6 (n=8,423 (70.6%))	34.4 ± 7.2 (n=8,505 (71.3%))	<0.001
Medications (%)
Insulin	81.1%	48.2%	<0.001	80.5%	80.0%	0.30
GLP-1 Agonists	27.5%	12.4%	<0.001	26.9%	26.7%	0.74
SGLT-2 Inhibitors	14.8%	9.1%	<0.001	14.9%	14.5%	0.40

Abbreviations: NPDR: non-proliferative diabetic retinopathy; OSA: obstructive sleep apnea; PSM: propensity score matching; SD: standard deviation; HbA1c: hemoglobin A1C; BMI: body mass index; GLP-1: glucagon-like peptide-1; SGLT-2: sodium-glucose cotransporter-2.

The average age of NPDR patients with and without OSA was 67.8 ± 11.2 years and 68.0 ± 12.6 years, respectively, with a roughly equal distribution of males and females. A substantial proportion of both groups had, in addition to their diabetes, concurrent hypertension (96.4% vs 96.4%), hyperlipidemia (86.8% vs 86.9%), or overweight/obesity (83.6% vs 82.9% in NPDR with OSA vs without OSA, respectively). Notably, of patients with documented values in the EHR, there were residual differences in baseline HbA1c between the cohorts (NPDR with OSA: 7.96 ± 1.96% vs NPDR without OSA: 8.07 ± 1.93%; P<0.001) and BMI (NPDR with OSA: 37.9 ± 8.6 vs NPDR without OSA: 34.4 ± 7.2; P<0.001) even after matching criteria had been applied.

Progression of Diabetic Retinopathy to Vision Threatening Complications

In the primary analysis, the rate of progression for patients with baseline NPDR to proliferative disease was higher in the OSA group at all time points: 1 year (RR: 1.34, 95% CI: 1.20-1.49, RD: 0.02, P<0.001), 3 years (RR: 1.31, 95% CI: 1.20-1.43, RD: 0.02, P<0.001), and 5 years (RR: 1.28, 95% CI: 1.18-1.39, RD: 0.02, P<0.001). By 5 years, 1,225 (10.4%) patients with OSA had converted to PDR, compared to 957 (8.1%) patients without OSA (Figure 1). A similar trend was observed for new-onset DME during the study period, with greater occurrence in the OSA group at all time points evaluated: 1 year (RR: 1.31, 95% CI: 1.19-1.45, RD: 0.02, P<0.001), 3 years (RR: 1.19, 95% CI: 1.10-1.29, RD: 0.02, P<0.001), and 5 years (RR: 1.18, 95% CI: 1.10-1.27, RD: 0.02, P<0.001). By 5 years, 1,479 (12.5%) NPDR patients with OSA had developed DME, compared to 1,253 (10.6%) individuals in the comparator group (Figure 1). The development of diabetic retinal detachment during the study period was also explored, and was not significantly different between both groups: 1 year (RR: 0.84, 95% CI: 0.50-1.41, RD: 0.0, P=0.51), 3 years (RR: 0.81, 95% CI: 0.55-1.19, RD: 0.0, P=0.28), and 5 years (RR: 0.85, 95% CI: 0.61-1.17, RD: 0.0, P=0.32). Overall, a low number of patients developed diabetic retinal detachments in both cohorts, occurring in 66 (0.6%) individuals with NPDR and OSA compared to 78 (0.7%) patients with NPDR without OSA at 5 years.

Figure 1.

Progression of Diabetic Retinopathy. Rates of conversion to proliferative diabetic retinopathy (PDR), new-onset diabetic macular edema (DME), and diabetic retinal detachment for patients with non-proliferative diabetic retinopathy (NPDR) with and without obstructive sleep apnea (OSA) at 1, 3, and 5 years of follow-up. NPDR

Need for Ocular Intervention

Patients with NPDR and OSA had significantly greater likelihood to require initiation of intravitreal pharmacotherapy at all time points throughout the study period: 1 year (RR: 1.59, 95% CI: 1.37-1.84, RD: 0.02, P<0.001), 3 years (RR: 1.58, 95% CI: 1.40-1.78, RD: 0.02, P<0.001), and 5 years (RR: 1.54, 95% CI: 1.38-1.72, RD: 0.02, P<0.001). By year 5, 763 (6.5%) individuals in the OSA group had initiated intravitreal injection therapy, compared to 496 (4.2%) without OSA (Figure 2). Similarly, we observed greater need for panretinal photocoagulation at all time points: 1 year (RR: 1.51, 95% CI: 1.21-1.90, RD: 0.01, P<0.001), 3 years (RR: 1.44, 95% CI: 1.21-1.70, RD: 0.01, P<0.001), and 5 years (RR: 1.52, 95% CI: 1.30-1.77, RD: 0.01, P<0.001). Laser photocoagulation was required in 397 (3.4%) patients with OSA versus 262 (2.2%) patients without OSA after 5 years (Figure 2). With respect to requiring surgical intervention, there was no significant differences observed in vitrectomy rates amongst NPDR patients with and without OSA: 1 year (RR: 1.02, 95% CI: 0.69-1.51, RD: 0.0, P=0.92), 3 years (RR: 1.11, 95% CI: 0.85-1.46, RD: 0.0, P=0.45), and 5 years (RR: 1.16, 95% CI: 0.91-1.46, RD: 0.0, P=0.23). An overall low instance of diabetic vitrectomy was performed through 5 years throughout both cohorts: 148 (1.3%) patients with OSA versus 128 (1.1%) without underwent the procedure (Figure 2), reflective of the lower starting DR severity at index.

Figure 2.

Need for Ocular Intervention. Progression to requiring intravitreal pharmacotherapy, panretinal laser photocoagulation, or surgical pars plana vitrectomy for patients with non-proliferative diabetic retinopathy (NPDR) with and without obstructive sleep apnea (OSA) at 1, 3, and 5 years of follow-up.

Risk of Systemic Complications

Patients with NPDR and OSA experienced significantly greater systemic events, including death, throughout the study period. The rate of stroke was higher in OSA patients at 1 year (RR: 1.80, 95% CI: 1.58-2.06, RD: 0.02, P<0.001), 3 years (RR: 1.56, 95% CI: 1.41-1.73, RD: 0.03, P<0.001), and 5 years (RR: 1.49, 95% CI: 1.37-1.63, RD: 0.03, P<0.001). After 5 years, the incidence of stroke was 9.5% (1,125 patients) versus 6.4% (753 patients) in NPDR patients with OSA compared to without OSA, respectively (Figure 3). Similarly, the incidence of myocardial infarction was greater in the OSA cohort at 1 year (RR: 1.51, 95% CI: 1.32-1.72, RD: 0.02, P<0.001), 3 years (RR: 1.46, 95% CI: 1.33-1.61, RD: 0.03, P<0.001), and 5 years (RR: 1.43, 95% CI: 1.32-1.55, RD: 0.03, P<0.001), occurring in 10.7% (1,261 patients) of the group compared to 7.5% (882 patients) of the NPDR without OSA cohort (Figure 3). Finally, patients with OSA had a greater mortality rate through 5 years, 14.0% (1,654 patients), compared to those without OSA, 12.2% (1,441 patients; Figure 3). Depending on the time frame, patients with OSA demonstrated an increased risk of passing than those without OSA (1 year: RR: 1.31, 95% CI: 1.15-1.49, RD: 0.01, P<0.001; 3 year: RR: 1.19, 95% CI: 1.10-1.29, RD: 0.02, P<0.001; 5 year: RR: 1.15, 95% CI: 1.08-1.23, RD: 0.02, P<0.001; Figure 4).

Figure 3.

Systemic Events. Incidence of developing stroke, myocardial infarction, and death for patients with non-proliferative diabetic retinopathy (NPDR) with and without obstructive sleep apnea (OSA) at 1, 3, and 5 years of follow-up.

Figure 4.

Risk Ratios. Risk ratios of ocular and systemic outcomes for patients with non-proliferative diabetic retinopathy (NPDR) with and without obstructive sleep apnea (OSA) at 1, 3, and 5 years of follow-up.

Discussion

A host of studies have implicated OSA with various ophthalmologic conditions, including nonarteritic anterior ischemic optic neuropathy (NAION),¹⁴ floppy eyelid syndrome,¹⁵ retinal vein occlusion (RVO),¹⁶ and central serous chorioretinopathy (CSCR).¹⁷ Within the retina, OSA has been postulated to contribute to impaired retinal microcirculation, hypertensive retinopathy, formation of retinal microaneurysms, tissue ischemia, and production of pro-angiogenic factors that may result in downstream neovascularization and/or macular edema.¹⁸

A direct relationship between OSA and DR is challenging to elucidate due to phenotypic overlap and potential confounding from comorbidities. Researchers over the years have explored this potential association with mixed results.^10,20-26 In one of the largest published studies, Chang et al found a positive association between severe OSA and DR in 317 patients.²⁰ Compared to patients with mild-to-moderate OSA, patients with severe OSA were found to be at a two- to threefold greater risk of having DR, PDR, and DME.²⁰ In a separate study of 230 patients, Altaf et al observed a dose-effect relationship showing that compared to no OSA, moderate-to-severe OSA was significantly associated with progression to advanced stages of DR.¹⁰ Further supporting the notion that OSA severity correlates to DR, Tan et al reported on 170 patients with diabetes and OSA, and found a higher prevalence of DR in patients with severe OSA compared to those with mild-to-moderate OSA.²¹ Conversely, Schober et al found no difference in the presence of DR between patients with mild OSA versus those with moderate-to-severe OSA when evaluating a larger cohort of 556 patients.²² In one of the largest cross-sectional studies on this topic evaluating 2,927 patients with sleep apnea and milder forms of sleep-disordered breathing, investigators from the Sleep Heart Health Study found no significant correlation with retinopathy (as measured by presence of retinal microvascular abnormalities on color fundus photographs) after adjusting for age, sex, parent study, BMI, hypertension, systolic blood pressure, smoking status, and diabetes.¹⁹ It is possible that that differences that result in progression may not be visible by fundus photography alone and/or that the effect of OSA on the progression of retinopathy is greater than its effect on the incidence of retinopathy.

The findings of our study are informative due to the substantial number of patients included over an extended period, and call into question an overreliance on traditional pharmacotherapy for managing diabetes without consideration for the presence of OSA. This is suggested by the significant differences observed in baseline HbA1c between both groups even after propensity matching, such that the group with OSA had lower HbA1c values yet experienced worse outcomes in most of the parameters measured in the study. Though we did not find a significant difference in progression to diabetic tractional retinal detachments or need for vitrectomy between the two groups, this could be partly explained by recent shifts in practice patterns to manage early conversion to PDR primarily with intravitreal anti-VEGF monotherapy or together with laser photocoagulation, which may help stave off progression to end-stage disease and resultant need for diabetic vitrectomy. Furthermore, there was a smaller proportion of patients that began the study period at a level of severe NPDR, which is congruous with the overall lower instances of progression to advanced disease we observed across both groups. That we observed a greater risk of converting to PDR, DME, need for ocular interventions (intravitreal injection and laser photocoagulation), as well as systemic vascular events, including death, over all time periods highlights a unique opportunity for eyecare providers to institute screening measures to capture patients with DR who may have concurrent undiagnosed OSA, and refer them for higher-level specialty care early on.

One time-efficient solution for larger scale screening of DR patients at risk of OSA would be questionnaire tools, such as the STOP-BANG survey, to detect mild, moderate, and severe OSA, after which patients can undergo formal diagnostic polysomnography testing.²⁷ The STOP-BANG questionnaire consists of eight yes-or-no questions based on the major risk factors for OSA. The name itself is an acronym for the first letter of each symptom or physical attribute often associated with OSA: Snoring, Tiredness, Observed apnea, Pressure (hypertension), BMI, Age, Neck circumference, and Gender.²⁷ Beyond a patient’s physical profile, there may be some cues the practitioner may consider to determine if a patient may be at risk for OSA. Blegen et al observed an association between OSA and the presence of cotton wool spots in NPDR. In their study, cotton wool spots were identified in 12 of 118 patients without OSA (10.2%), compared with 11 of 32 patients with OSA (34.4%).²⁸ Furthermore, for patients already receiving intravitreal injection therapy, several studies have demonstrated a poorer response to VEGF therapy for exudative retinal diseases, including DME and neovascular AMD, in patients with OSA.^29,30 Ultimately, diagnosing and treating OSA in DR patients earlier on in the disease process can aid in preventing progression of retinopathy, potentially enhance response to conventional intravitreal injection therapy, and may help prevent downstream serious cardiovascular consequences of the disease, including MI, stroke, and death, which we observed nearly a two-fold greater mortality risk over 5 years in the present study.

An additional consideration is the interplay between obesity with OSA and diabetes. While obesity is a known mediator between both conditions, our study design attempted to mitigate the potential confounding impact of adiposity on the observed findings through propensity score matching criteria. It is important to note that when adjusting for overweight/obesity by diagnosis (ICD-10: E66), we were able to achieve more balanced subgroupings with 83.6% (NPDR with OSA) vs. 82.9% (NPDR without OSA) of patients that underwent subsequent analysis, and were also able to account for conditions that can influence obesity and OSA alike, such as hypothyroidism and polycystic ovary syndrome. However, as an added measure, we propensity score matched for BMI values, when available in the EHR, which were registered in 70.6-71.3% of participants in each cohort after matching was completed. Even after propensity matching, there were still residual differences, though reduced, in the BMI observed between both cohorts: 37.9 ± 8.6 (NPDR with OSA) vs. 34.4 ± 7.2 (NPDR without OSA). Thus, it warrants consideration that our observations may not be fully attributable to OSA in isolation, and could have secondary overlapping impact from degree of obesity. Separately, the data on overweight/obesity and BMI give rise to the possibility that some percentage of patients in the NPDR without OSA group also have undiagnosed OSA. This thought is not unprecedented, as OSA is known to be severely underdiagnosed even in extremely at-risk populations, with some studies estimating that more than 85% of patients with clinically significant OSA as having never been diagnosed.^31-33 Assuming this is at least partly true in our patient cohorts, these results may be actually understating the true impact of OSA on DR progression, and the observed disparity between the two groups may be even greater than what we are reporting herein.

Strengths of this study include its substantial sample size involving multiple HCOs from large medical care systems representing a heterogenous patient population, both geographically and racially. TriNetX is a live real-world database with new patient data being added daily, allowing for observations and trends to be captured in real time without delays. The platform itself has been validated by numerous comprehensive study designs across many specialties in medicine, with more recent publications emerging from the field of ophthalmology.^34,35

This particular investigation represents the largest aggregate group of EHR records examining OSA’s impact on DR progression to date by a significant margin, as previous studies have been limited by the number of patients included. Additionally, the majority of previous studies exploring the relationship of OSA with DR have been cross-sectional in nature. Only the aforementioned study of 230 patients by Altaf et al was longitudinally designed over an average follow-up interval of 43 months.¹⁰ Due to the unique real time access to longitudinal clinical data afforded by the TriNetX platform, we were able to follow our patient cohort as far out as 5 years without major attrition. Our study maintained stringent inclusion criteria by excluding patients with any prior history of known PDR, DME, or ocular intervention (intravitreal injections, laser photocoagulation, and/or vitrectomy) and utilized a well-matched control population to account for various systemic differences in order to best isolate the impact of OSA. For example, groups were matched on use of insulin and other hypoglycemic agents, and more specifically the newer classes such as the sodium-glucose cotransporter-2 (SGLT-2) inhibitors and glucagon-like peptide 1 (GLP-1) agonists, which may differentially impact rates of DR progression based on emerging literature.^36-38 Lastly, this is the only study to examine longer-term outcomes, both locally and systemically, of OSA on DR patients with such a robust sample size.

Our study has several limitations, particularly challenges inherent to interrogating large data sets utilizing aggregate EHR data retrospectively. Namely, we are reliant upon accurate ICD-10 diagnosis and coding by practitioners, therein exists the possibility of misclassification of correct stage of DR and/or concomitant present of DME. Additionally, though strict inclusion/exclusion criteria were enforced to best isolate treatment-naïve NPDR patients, we cannot exclude that some patients may have received previous ocular treatment separated longitudinally by time or from receiving care at separate institutions which may have not been a member of the TriNetX research network. That only a minority of patients enrolled in this study were severe NPDR at the index visit (<5% in each cohort after propensity score matching) lends confidence to the notion that the groups being investigated were truly treatment-naïve, and that patients with any previous therapy were being adequately excluded from consideration. Finally, though the criteria used to identify patients with OSA based on combined ICD-10 and CPT codes is robust and dependable, we are unable to gain deeper insights into the severity of OSA or subsequent management of it due to not having access to polysomnography reports. Apnea-hypopnea index (AHI), a measurement of respiratory events by sleep hour, helps to determine OSA severity as mild (>5 events and <15 events), moderate (>15 events and <30 events), and severe (>30 events/hour). When a home sleep study is done, AHI can be represented by various quantitative metrics such as the respiratory event index (REI) or respiratory disturbance index (RDI).³⁹ One may surmise that having more severe OSA would correlate to worse ocular and systemic outcomes, as was corroborated in the previously mentioned publications,^10,20,21 but we are unable to measure for this in our study population due to a lack of current coding available for OSA severity. It would be prudent for future investigations in this space to more definitively explore the association between severity of OSA and DR outcomes.

Perhaps more importantly, is the inability of our study to measure patient compliance to recommended OSA therapy, which is typically mechanical ventilation with CPAP. Indeed, various studies have reported wide-ranging adherence to CPAP, between 46-83%,⁴⁰ due to discomfort from the bulky masks, facial rash, or claustrophobia.⁴¹ When compliance is attained, however, several studies have shown a direct association between increased CPAP use and more optimal DR outcomes.^42,43 To that end, our study correlates the presence of OSA and its impact on DR progression rates as well as systemic events, but we cannot comment on whether and how much does treatment adherence modify or potentially even revert that risk. That being stated, the criteria for inclusion herein combined ICD-10 coding of OSA together with CPT for use of CPAP; thus, the OSA cases included are presumably severe as they required CPAP management. That we still observed such differences in outcomes even with these patients receiving treatment (assuming compliance), argues in favor of these findings being even more significant.

In conclusion, this study demonstrates that NDPR patients with OSA carried an increased risk of developing VTCs, progression to proliferative disease or DME, and need for subsequent ocular intervention with intravitreal injection pharmacotherapy or laser photocoagulation over a 5-year period compared to patients without OSA. Furthermore, we observed significantly elevated risk of experiencing systemic events in individuals with OSA, including stroke, MI, and death during the same time interval. Further prospective studies are warranted to better characterize this complex relationship between DR and OSA; however, our findings emphasize the need for improved screening measures of patients with NPDR being monitored by eyecare providers to potentially uncover underlying OSA and generate referrals for appropriate management.

Highlights.

• In this cohort study that included 9,367 patients with diabetic retinopathy and obstructive sleep apnea, there was an increased rate of vision threatening complications and need for ocular intervention through 5 years of follow-up when compared to a matched cohort of diabetic retinopathy patients without obstructive sleep apnea.

• There was additionally an increased risk of experiencing systemic cardiovascular complications, including death, in the cohort of patients with diabetic retinopathy and obstructive sleep apnea during the 5-year period.

• The results emphasize the need for improved screening measures of patients with diabetic retinopathy by eyecare providers to potentially uncover underlying obstructive sleep apnea as a risk factor for progression of ocular as well as systemic disease.

Biography

Dr. Ehsan Rahimy is a vitreoretinal surgeon at the Palo Alto Medical Foundation and also an adjunct clinical assistant professor of ophthalmology at Stanford University. He completed his residency training at the Jules Stein Eye Institute at UCLA and fellowship at Wills Eye Hospital where he was a recipient of the Heed Fellowship and Ron Michels Fellowship Foundation awards. His research interests include aggregate healthcare data within retina, ophthalmology, and broader systemic interactions.