Abstract
Purpose:
Obstructive sleep apnea (OSA) may increase the risk of age-related macular degeneration (AMD) due to repetitive oxygen deprivation or other mechanisms, though whether OSA increases the risk of AMD progression is unknown. We analyzed associations between OSA and AMD risk in the Taiwanese population.
Methods:
We identified patients diagnosed with OSA between 2000 and 2018 in the Taiwan National Health Insurance Research Database and used 1:1 propensity score matching on demographics and co-morbidities to create a non-OSA cohort. We used Cox proportional hazard modeling to investigate the risk of AMD and the risk of progression from nonexudative to exudative AMD in OSA versus non-OSA patients.
Results:
A total of 66,869 OSA patients were matched with 66,869 non-OSA patients. The hazard ratio (HR) of AMD in the OSA cohort was 1.36 (95% confidence interval[CI]: 1.31–1.43, p<.0001). The HR for progression from nonexudative to exudative AMD for the OSA cohort was 0.94 (95%CI: 0.77–1.14, p=0.5073).
Conclusion:
OSA is associated with a higher risk of developing AMD, and further research is warranted to determine whether OSA patients may benefit from regular ophthalmic examinations. Patients with OSA and existing non-exudative AMD did not show an increased risk of AMD progression.
Keywords: Obstructive sleep apnea, age-related macular degeneration, exudative age-related macular degeneration
Brief summary statement for index:
Using the Taiwan National Health Insurance Research Database, we found that patients with obstructive sleep apnea (OSA) had 1.36 fold increased risk of developing age-related macular degeneration (AMD) compared to matched peers without OSA, but no increased risk of progression to exudative AMD.
Introduction
Age-related macular degeneration (AMD) is the primary cause of vision loss in developed countries, with the expected prevalence estimated to be 288 million by 2040.1 AMD is characterized by drusen formation, abnormalities and geographic atrophy in the retinal pigment epithelium (RPE) and typically affects individuals 55 years and older.2 AMD is generally divided into nonexudative (non-neovascular or dry) and exudative (neovascular or wet). In late AMD, patients develop vision-threatening symptoms, and some patients progress to exudative AMD. Although there is no accepted treatment for late non-neovascular AMD, exudative AMD is managed with intravitreal anti-vascular endothelial growth factor injections, an effective but expensive option to prevent blindness and restore vision.3
Obstructive sleep apnea (OSA) has been associated with AMD and other ocular conditions such as glaucoma, ischemic optic neuropathy, central serous chorioretinopathy, and retinal vein occlusion.4 OSA is the most common form of sleep-disordered breathing, characterized by intermittent upper airway collapse, impaired ventilation, hypoxia, hypercapnia, and disturbed sleep.5 A potential mechanism in OSA that may be associated with AMD is the occurrence of repetitive oxygen deprivation, leading to angiogenesis as a response to compensate for insufficient oxygen.6,7 Although OSA is associated with a higher risk of AMD in population studies,8,9 it is not yet known whether OSA is associated with progression from nonexudative AMD to exudative AMD.
Asia is projected to see the largest increase of AMD cases globally1 as the aging population increases. The Taiwan National Health Insurance Research Database (NHIRD) contains data from Taiwan’s universal single-payer national health insurance program, encompassing 99% of the population. It includes comprehensive longitudinal de-identified beneficiary data, including demographic information, medical appointments, and diagnosis records. Utilizing this unique dataset, we investigated the risk of AMD and compared progression from nonexudative to exudative AMD in patients with and without OSA.
Methods
Data Source
We performed a retrospective cohort study using the NHIRD and the Taiwan Death Record (TDR) between 2000 and 2021. Taiwan’s National Health Insurance (NHI) system has been run by the government since 1995. By law, every Taiwanese citizen with an official residency or foreigner living in Taiwan with an Alien Resident Certificate, regardless of age, gender, or employment status, must enroll in the NHI program. Up to 2018, 23,948,000 beneficiaries, over 99% of the total population in Taiwan, were enrolled in the NHI system10 The NHIRD contains registration information and data on original claims for reimbursement, including comprehensive information on demographics, dates of clinical visits, medical diagnoses, medical expenditures, prescriptions, examinations, and procedures. In Taiwan, the Household Registration Act requires reporting of personal status changes with any death or presumption of death. All original death certificates in Taiwan are centralized in the Provincial Department of Health, Office of Statistics. The accuracy of cause-of-death coding in Taiwan has been validated previously.11 Since 2015, researchers must perform onsite analysis at the Health and Welfare Data Center (HWDC), which centralizes the NHIRD and TDR through a remote connection to the Ministry of Health and Welfare (MOHW) servers. All datasheets have been linked with individual personal identification numbers (PINs), but PINs are encrypted for research. No electronic devices may be brought into a center, and individual-level data cannot be released. All analysis results are reviewed by data custodians to prevent patient identity disclosure. The Institutional Review Board of Chang Gung Medical Foundation approved this study and waived participants’ consent (202301359B0).
Cohort selection
For our analytic cohort, we included incident patients diagnosed with OSA in the NHIRD between 2000 and 2018. A diagnosis of OSA was determined by International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes: 780.51, 780.53, 780.57 and Tenth Revision (ICD-10-CM) code: G47.33 during at least two outpatient visits or one inpatient visit. The OSA index date was defined as the first date of OSA diagnosis. Exclusion criteria included: (1) unreasonable death records before the index date (n=47), (2) missing sex or age (n=3013), (3) age < 50 years old (n=184,547), (4) AMD history before the index date (n=7,321), and (5) missing polysomnography (n=62,991). Because of the large number of patients without OSA in the database, the non-OSA control cohort was first matched 10:1 with the OSA cohort on birth year. Then 1:1 propensity score matching (PSM) was performed between the OSA and non-OSA cohorts using comparable covariates (birth year, sex, year of diagnosed OSA, comorbidities, and co-medications) at the index date to reduce confounding (Figure 1A). For matching on the OSA index date, non-OSA patients who sought treatment other than OSA within 1 month of the OSA diagnosis in the OSA group were selected.
Figure 1.

Flow chart of participant selection. For propensity score matching (PSM) on the OSA index date, non-OSA patients who sought treatment other than OSA within 1 month of the OSA diagnosis in the OSA group were selected.
For AMD incidence, duration at risk was defined from the index date until the occurrence of AMD, loss to follow-up, death, or December 31, 2021, whichever occurred first. For the progression to exudative AMD analysis, only those who had developed AMD were included, and those who had identical dates of exudative and nonexudative AMD diagnoses were excluded. Duration at risk was defined from the AMD index date (newly diagnosed nonexudative AMD after the OSA index date) until exudative AMD, loss to follow-up, death, or December 31, 2021, whichever occurred first (Figure 1).
Ascertainment of study outcome
Nonexudative AMD was defined by ICD-9-CM codes 362.50 (unspecified AMD), 362.51 (nonexudative AMD), ICD-10-CM: H35.30 (unspecified AMD), or H35.31 (nonexudative AMD) occurring after the OSA index date for two or more outpatient visits or at least one inpatient visit. Exudative AMD was defined by ICD-9-CM 362.52 or ICD-10-CM H35.32 occurring after the nonexudative AMD index date for two or more outpatient visits or at least one inpatient visit. We performed a sensitivity analysis where the definition of exudative AMD required both ICD code for AMD and the use of either ranibizumab (covered by NHI since 2011) or aflibercept (covered by NHI since 2014).
Ascertainment of covariates
Demographic variables included the calendar year of OSA diagnosis, age, sex, income level, occupation, and residence region. Income level was categorized by quartile. The occupation was divided into five categories: (1) dependent, financially supported by an insured family member, (2) civil servants with full-time or regular jobs in the government or public affiliation, (3) non-manual workers working in privately owned institutions, (4) manual workers, including blue-collar workers, farmers, or fishers; and (5) others, including veterans, members of low-income families, and substitute service draftees. Residence region was categorized into north, middle, south, and outlying islands, or not specified.
Comorbidities of interest included hypertension (ICD-9-CM: 401, 402; ICD-10-CM: I10-I16), hyperlipidemia (ICD-9-CM: 272; ICD-10-CM: E78), and the conditions included in the Charlson Comorbidity Index (CCI).12 The CCI score is calculated based on the presence/absence of variously weighted comorbidities, including myocardial infarction (1 point), congestive heart failure (1 point), peripheral vascular disease (1 point), cerebrovascular disease (1 point), dementia (1 point), chronic pulmonary disease (1 point), rheumatic disease (1 point), peptic ulcer disease (1 point), mild liver disease (1 point), diabetes (1), diabetes with end organ damage (2 points), hemiplegia or paraplegia (2 points), any tumor, leukemia, lymphoma (2 points), moderate or severe liver disease (3 points), metastatic solid tumor (6 points), and AIDS/HIV+ (6 points). Comorbidity diagnosis codes were required at two or more outpatient visits or at least one inpatient visit before the index date to reduce misclassification (see Table S1, Supplemental Digital Content 1). Baseline medications included antiplatelets (aspirin, clopidogrel, ticlopidine, ticagrelor) and antihypertensive drugs (angiotensin-converting-enzyme inhibitors, angiotensin receptor blockers, beta-blockers, diltiazem, verapamil, dihydropyridine calcium channel blockers). Baseline medications were restricted to those prescribed at least once within three months before the index date. Please refer to the National Health Insurance Administration website for detailed medication codes.13
Statistical analysis
The incidence rate of AMD was the number of patients newly diagnosed with AMD after the OSA index date divided by the person-years involved. The progression rate from nonexudative to exudative AMD was calculated in person-years based on the participants with nonexudative AMD who went on to develop exudative AMD later. Hence, we excluded patients whose first nonexudative and exudative AMD diagnosis dates were the same. The 95% confidence interval (CI) of the rate was based on the relationship between the F and binomial distribution.14
For the PSM, the OSA cohort was matched with those in the non-OSA cohort in a 1:1 ratio within a caliper width equal to 0.2 of the standard deviation of the logit of the estimated propensity score. We included all demographics, comorbidities, and co-medications (Table 1) in the generalized boosted model (GBM) to obtain the propensity score. The generalized boosted model involves an iterative process with multiple regression trees to capture complex and nonlinear relationships between the two cohorts and the covariates without overfitting the data.15 Thus, the GBM model often achieves an optimal balance between two cohorts, less affected by extreme weights and model misspecification.16 We computed the absolute standardized mean difference (ASMD) to measure the balance of covariates between the two groups.17 An ASMD greater than 0.1 was considered statistically significant.
Table 1.
Baseline characteristics of obstructive sleep apnea (OSA) cohort and non-OSA cohort after propensity score matching.
| OSA cohort (n=66869) | Non-OSA (n=66869) | ASMD | |
|---|---|---|---|
| Age (years) | |||
| Mean±SD | 60.0±8.1 | 60.3±8.2 | 0.0291 |
| <55 | 20024 (30.0%) | 20024 (30.0%) | 0 |
| 55–64 | 29974 (44.8%) | 29974 (44.8%) | |
| 65–74 | 12386 (18.5%) | 12386 (18.5%) | |
| 75–84 | 3938 (5.9%) | 3938 (5.9%) | |
| ≥85 | 547 (0.8%) | 547 (0.8%) | |
| Male | 46901 (70.1%) | 46901 (70.1%) | 0 |
| Income | 0.0239 | ||
| 0–24% | 27738 (41.5%) | 28313 (42.3%) | |
| 25–49% | 5079 (7.6%) | 5054 (7.6%) | |
| 50–74% | 13319 (19.9%) | 13376 (20.0%) | |
| ≥75% | 20733 (31.0%) | 20126 (30.1%) | |
| Occupation | |||
| Dependent | 16611 (24.8%) | 17053 (25.5%) | 0.0577 |
| Civil servants | 3071 (4.6%) | 2876 (4.3%) | |
| Non-manual worker | 17984 (26.9%) | 17300 (25.9%) | |
| Manual worker | 19647 (29.4%) | 19939 (29.8%) | |
| Other | 9556 (14.3%) | 9701 (14.5%) | |
| Region | 0 | ||
| North | 36756 (55.0%) | 36756 (55.0%) | |
| Middle | 22348 (33.4%) | 22348 (33.4%) | |
| South | 7503 (11.2%) | 7503 (11.2%) | |
| Outlying Islands | 239 (0.4%) | 239 (0.4%) | |
| Not specified | 23 (0.0%) | 23 (0.0%) | |
| Index year | 0 | ||
| 2000–2004 | 4558 (6.8%) | 4558 (6.8%) | |
| 2005–2009 | 13579 (20.3%) | 13579 (20.3%) | |
| 2010–2014 | 26113 (39.1%) | 26113 (39.1%) | |
| 2015–2018 | 22619 (33.8%) | 22619 (33.8%) | |
| Comorbidities | |||
| Hypertension | 25768 (38.5%) | 25186 (37.7%) | 0.0179 |
| Hyperlipidemia | 28440 (42.5%) | 27901 (41.7%) | 0.0163 |
| Coronary artery disease | 8848 (13.2%) | 7536 (11.3%) | 0.0599 |
| Myocardial infarction (1 point) | 1601 (2.4%) | 1433 (2.1%) | 0.0169 |
| Congestive heart failure (1 point) | 683 (1.0%) | 678 (1.0%) | 0.0007 |
| Peripheral vascular disease (1 point) | 2166 (3.2%) | 2069 (3.1%) | 0.0083 |
| Cerebrovascular disease (1 point) | 13285 (19.9%) | 12550 (18.8%) | 0.0278 |
| Dementia (1 point) | 1316 (2.0%) | 1210 (1.8%) | 0.0116 |
| Chronic obstructive pulmonary disease (1 point) | 28123 (42.1%) | 27996 (41.9%) | 0.0038 |
| Rheumatic disease (1 point) | 411 (0.6%) | 332 (0.5%) | 0.0159 |
| Peptic ulcer disease (1 point) | 33723 (50.4%) | 34401 (51.5%) | 0.0203 |
| Mild liver disease (1 point) | 22150 (33.1%) | 22268 (33.3%) | 0.0037 |
| Diabetes (1 point) | 17432 (26.1%) | 16917 (25.3%) | 0.0176 |
| Diabetes with end organ damage (1 point) | 4875 (7.3%) | 4604 (6.9%) | 0.0158 |
| Hemiplegia or paraplegia (2 points) | 704 (1.1%) | 640 (1.0%) | 0.0096 |
| Moderate or severe renal disease (2 points) | 7374 (11.0%) | 6912 (10.3%) | 0.0224 |
| Any tumor, leukemia, lymphoma (2 points) | 6233 (9.3%) | 6469 (9.7%) | 0.012 |
| Moderate or severe liver disease (3 points) | 149 (0.2%) | 180 (0.3%) | 0.0094 |
| Metastatic solid tumor (6 points) | 367 (0.6%) | 508 (0.8%) | 0.0262 |
| AIDS/HIV+ (2 points) | 26 (0.0%) | 29 (0.0%) | 0.0022 |
| Charlson Comorbidity Index | |||
| Mean±SD | 2.4±2.1 | 2.4±2.2 | 0.0046 |
| 0 | 11252 (16.8%) | 11492 (17.2%) | 0.0084 |
| 1 | 15186 (22.7%) | 15469 (23.1%) | |
| 2 | 14097 (21.1%) | 14107 (21.1%) | |
| 3 | 9904 (14.8%) | 9871 (14.8%) | |
| 4+ | 16430 (24.6%) | 15930 (23.8%) | |
| Medications | |||
| Antiplatelet* | 15620 (23.4%) | 14629 (21.9%) | 0.0354 |
| Antihypertensive drugs† | 35633 (53.3%) | 36518 (54.6%) | 0.0266 |
| Anticoagulants‡ | 1236 (1.9%) | 998 (1.5%) | 0.0278 |
ASMD, absolute standardized mean difference. AIDS/HIV+, acquired immune deficiency syndrome/human immunodeficiency virus-positive. SD, standard deviation.
(points): the score involved for Charlson Comorbidity Index
Antiplatelet: aspirin100, clopidogrel, ticlopidine, ticagrelor.
Antihypertensive drugs: ACEI/ARB, betablocker, diltiazem/verapamil, dihydropyridine calcium-channel blockers, loop diuretics (bumetanide furosemide).
Anticoagulants: warfarin, dabigatran, rivaroxaban, apixaban, edoxaban.
We used the Cox proportional hazard model to obtain the hazard ratio (HR) of AMD or progression of exudative AMD between the OSA vs. non-OSA groups. The robust sandwich method in the Cox model was used to handle the dependency of the matched data. We also applied the Fine-Gray model, which accounts for deaths as competing risk events, to obtain the subdistribution hazard ratio (SHR), given the high mortality rates in elderly populations18 Note that the cohort variable (OSA vs. non-OSA) was the only independent variable included in the Cox and Fine-Gray models because these cohorts were well-balanced in demographics, comorbidities, and co-medications.19 Statistical analyses were performed using SAS, version 9.4 (SAS Institute, Cary, NC, USA). P < 0.05 was considered statistically significant.
Results
A total of 66869 patients were enrolled in the OSA cohort and were matched with 66869 patients in the non-OSA cohort using 1:1 PSM. (Figure 1) The OSA and non-OSA groups were therefore similar in terms of demographics, comorbidities, and co-medications (Table 1). The mean age was 60, 70% were male, about 42% were in the lowest income quartile, a quarter were dependent, 55% lived in North Taiwan, and only 6.8% were first diagnosed with OSA during 2000–2004. Peptic ulcer disease was the most common comorbidity (51%), followed by hyperlipidemia (42%), chronic obstructive pulmonary disease (42%), hypertension (38%), mild liver disease (33%), diabetes (26%), and cerebrovascular disease (19%). The mean Charlson Comorbidity Index was 2.4 (standard deviation ±2). About half of the patients were prescribed antihypertensive drugs, and 22% were prescribed antiplatelets.
A total of 3,505 (6.11 per 1000 person-years) and 2,524 (13.76 per 1000 person-years) patients developed AMD in the OSA and non-OSA cohorts, respectively. The HR of AMD in the OSA cohort was 1.36 (95% CI: 1.31–1.43) (Figure 2). The significance between the two cohorts remained after accounting for competing risk of death (SHR 1.41, CI: 1.34–1.48). The elevated risk became apparent two years after OSA diagnosis for the OSA cohort (Figure 2).
Figure 2.

The cumulative incidence rate of AMD for the OSA cohort vs. non-OSA cohort.
For those who had AMD, 244 (6.9%) in the OSA cohort and 222 (8.8%) in the non-OSA cohort, had either identical first diagnosis dates for exudative and nonexudative AMD or had a diagnosis of exudative AMD without prior nonexudative AMD diagnosis and were excluded from the progression to exudative AMD analysis. Among the remaining OSA cohort with incident nonexudative AMD, 232 developed exudative AMD, equivalent to a progression rate of 13.76 per 1000 person-years. Among the non-OSA cohort with incident nonexudative AMD, 171 developed exudative AMD, equivalent to the progression rate 14.85 per 1000 person-years, although the difference between groups was not statistically significant. The HR and SHR for progression from nonexudative AMD to exudative AMD for the OSA cohort were 0.94 (p=0.5073) and 0.95 (p=0.5921), respectively (Figure 3A). Because it takes time to progress, those diagnosed with exudative AMD within one month of the nonexudative AMD index date (47 in the OSA group and 54 in the non-OSA group) were further excluded. The risk of progression to exudative AMD in those with OSA remained insignificant (HR 1.12, p=0.3337) and SHR 1.14, p=0.2680) (Figure 3B). Results from the sensitivity analysis were similar when exudative AMD required both ICD codes for AMD and the use of either ranibizumab or aflibercept (see Figure S1, Supplemental Digital Content 2; Figure S2, Supplemental Digital Content 3).
Figure 3.

The cumulative incidence rate for progression to exudative AMD in patients with nonexudative AMD for the OSA and non-OSA cohorts. (A) followed from AMD index date (i.e. newly diagnosed nonexudative AMD after the OSA index date), (B) followed from 1 month after AMD index date (excluding patients with exudative AMD diagnosis within one month of nonexudative AMD diagnosis).
Discussion
In this nationwide cohort study of 66,869 patients with OSA and 66,869 PSM non-OSA patients in the Taiwan NHIRD followed for at least 3 years and up to 22 years, we found that OSA was associated with a 1.36-fold increased risk of developing AMD. We found no significant increase in risk for progression to exudative AMD in association with OSA. Our results confirm previous findings that OSA is associated with an increased risk of AMD. The primary hypothesized mechanism for this association is OSA-induced hypoxia, however, the lack of increased risk of exudative AMD in our study may suggest that conversion from nonexudative to exudative AMD is associated with other factors or disease processes.
Previous studies have found positive associations between OSA and risk of AMD incidence, with similar hazard ratios (HR). One study using ICD code data from 502,505 patients from the UK Biobank found that individuals with OSA had a higher incidence of AMD (HR 1.39, 95% CI 1.15, 1.68) compared to those without OSA over a follow-up period of 8 years.8 The authors also analyzed patient-reported disease data from the Canadian Longitudinal Study on Aging (n=24,073) and found a risk ratio of 1.39 (95% CI 1.08–1.77) for AMD in those with OSA versus no OSA over 3 years of follow-up. Similarly, another UK record linkage study with data collected from 67,786 patients between 1999 and 2011, found a rate ratio of 1.44 for developing AMD (95% CI 1.32, 1.57) in those with OSA compared to a reference cohort.9
Multiple mechanisms have been proposed to explain the association between AMD and OSA. OSA-related hypoxia and hypercapnia can affect choroidal layers by altering vascular resistance, permeability, and blood flow, which leads to the progressive dysfunction of the RPE.7,20 The RPE is a key player in maintaining the function of photoreceptors and the immune defense of the macula.21,22 Reduced perfusion and oxygenation due to OSA-associated hypoxia may lead to the apoptosis of retinal ganglion cells and loss of retinal nerve fiber layer thickness that occurs gradually or intermittently over longer periods.23 Various studies have also suggested that recurrent hypoxia and oxidative stress linked to OSA initiate inflammation,24 which has been implicated to have a significant role in the development and progression of AMD.6,20,25 AMD pathogenesis has also been shown to be associated with the dysregulation of mitochondrial function in RPE cells.22,26 It is possible that oxidative stress from OSA could disrupt mitochondrial function and the metabolic ecosystem in the retina, which reduces glucose availability to photoreceptors leading to the impairment or loss of photoreceptors.22 Lastly, the RPE is vulnerable to oxidative stress due to its high metabolic activity and abundant oxygen consumption– a factor that together with deteriorating function, further stimulates the vicious cycle that promotes the development of chronic inflammation in the retina leading to the progression of AMD.6,21
To our knowledge, no study has ever investigated whether patients with OSA have an elevated progression rate from nonexudative AMD to exudative AMD. In this study, we saw the progressive rate from nonexudative AMD to exudative AMD was similar between the two study cohorts. Although we hypothesized that patients with AMD and OSA might be at higher risk for progression to exudative AMD, we did not find an increased risk for exudative AMD in patients with OSA compared to those without OSA. The systemic OSA–related mechanisms that contribute to increased risk of nonexudative AMD (hypoxia, inflammation, mitochondrial dysfunction, etc.) may differ from nonsystemic RPE/choroidal interactions involved in the progression to exudative AMD such as localized perfusion defects.27 One study in 38 patients with OSA and exudative AMD found that those who were undergoing continuous positive airway pressure therapy (CPAP) for OSA achieved a better response to intravitreal bevacizumab (better visual acuity and less central retinal thinning) compared to those with untreated OSA.28 The pathogenesis of exudative AMD is complex and involves immune responses leading to the secretion of inflammatory cytokines including VEGF, which stimulates epithelial cell proliferation and choroidal neovascularization.29 The newly formed blood vessels leak fluid, which disrupts and harms the layer of photoreceptors, ultimately causing vision impairment.30 Progression to exudative AMD is associated with multiple risk factors in addition to hypoxia and oxidative stress, such as genetics, inflammation, and environmental factors. Further investigation into the risk of exudative AMD conversion in OSA patients and the mechanisms underlying these associations is warranted.
There are several limitations to this study. First, we used data from primarily Asian patients in Taiwan, and these results may not be generalizable to other populations; however, it is one of the few studies that focuses on an Asian population. Second, extensive database studies such as this one depend on national healthcare claim records, in which the precision of ICD coding may be imprecise. Coding for exudative versus non-exudative AMD may not be specified. Nevertheless, the risk of AMD in patients with OSA in our study was comparable to previous population studies. Third, 7 to 9% of patients with AMD who had a first diagnosis of both nonexudative and exudative AMD on the same date or were diagnosed with exudative AMD without a prior nonexudative AMD diagnosis were not eligible for the progression to exudative AMD analysis. Fourth, bevacizumab is not covered by Taiwan’s NHI due to its off-label use in the eye. If patients received intravitreal bevacizumab for exudative AMD, we would have missed these cases. However, this would have underestimated the true number of AMD conversions and biased our results towards the null further. Fifth, some studies suggest an association between OSA and polypoidal choroidal vasculopathy (PCV),31 which is prevalent in Taiwan.32 Whether PCV is a subtype of exudative AMD33 or a different clinical entity remains controversial.34 To our knowledge, no ICD code is specific enough for PCV. In Taiwan, PCV is coded as exudative AMD because the treatment of PCV is similar to exudative AMD. We used the same ICD code to define exudative AMD for the OSA and non-OSA cohorts, so nondifferential misclassification bias towards the null35 may exist in this study. As we already saw a significant association between OSA and overall AMD, the unbiased estimate of the association would have been further from the null than what was observed. Lastly, we were unable to obtain data related to body mass index (BMI), smoking, intake of age-related eye disease (including vitamin C, vitamin E, beta-carotene, copper (cupric oxide), zinc, lutein, and zeaxanthin),36 and genetic markers of OSA. Investigating the relationship between genetic risk factors for OSA and AMD would be interesting for future studies. Although we controlled for an extensive list of comorbidities and demographic variables using PSM, residual confounders may have impacted the association between OSA and the risk of AMD in our study. Furthermore, as ICD coding for obesity is under-reported in NHIRD, we did not include obesity in this study. Notably, Keenan et al. highlighted that the association between obesity and AMD is weaker compared to that between OSA and AMD, suggesting a separate significant impact of OSA.9
Data from the Taiwan NHIRD suggests that individuals with OSA are associated with a higher risk of developing AMD. Patients with OSA and nonexudative AMD did not show an increased risk for progression to exudative AMD compared to non-OSA patients, and these results warrant further study. This relationship between OSA and AMD has essential implications for understanding the risk factors for both conditions. In particular, further research is warranted to determine whether OSA patients may benefit from undergoing routine ophthalmic examinations, and physicians caring for patients with AMD should be aware of sleep-related symptoms and potential concurrent OSA.
Supplementary Material
Supplemental Digital Content 1.docx: Table S1
Supplemental Digital Content 2.docx: Figure S1
Supplemental Digital Content 3.docx: Figure S2
Funding:
Chang Gung Memorial Hospital (CFRPG1N0011) funded this research. Dr. C. Lee has grant support from the National Institutes of Health grant R01AG060942, the Latham Vision Research Innovation Award (Seattle, WA), the Klorfine Family Endowed Chair, the Karalis Johnson Retina Center, and by an unrestricted grant from Research to Prevent Blindness. The sponsors or funding organizations had no role in the design or conduct of this research.
Footnotes
Financial disclosures: None of the authors has any financial/conflicting interests to disclose.
References
- 1.Wong WL, Su X, Li X, et al. Global prevalence of age-related macular degeneration and disease burden projection for 2020 and 2040: a systematic review and meta-analysis. Lancet Glob Health. 2014;2(2):e106–16. [DOI] [PubMed] [Google Scholar]
- 2.Deng Y, Qiao L, Du M, et al. Age-related macular degeneration: Epidemiology, genetics, pathophysiology, diagnosis, and targeted therapy. Genes Dis. 2022;9(1):62–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Wong TY, Liew G, Mitchell P. Clinical update: new treatments for age-related macular degeneration. Lancet. 2007;370(9583):204–206. [DOI] [PubMed] [Google Scholar]
- 4.Ahn J, Gorin MB. The Associations of Obstructive Sleep Apnea and Eye Disorders: Potential Insights into Pathogenesis and Treatment. Current Sleep Medicine Reports. 2021;7(3):65–79. [Google Scholar]
- 5.Eckert DJ, Malhotra A. Pathophysiology of adult obstructive sleep apnea. Proc Am Thorac Soc. 2008;5(2):144–153. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Chen NN, Chen CY, Wang JJ, et al. Functional and Anatomical Outcomes of Anti-Vascular Endothelial Growth Factor Treatment for Exudative Age-Related Macular Degeneration with or without Obstructive Sleep Apnea. Int J Mol Sci. 2023;24(8). doi: 10.3390/ijms24087285 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Nakayama LF, Tempaku PF, Bergamo VC, et al. Obstructive sleep apnea and the retina: a review. J Clin Sleep Med. 2021;17(9):1947–1952. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Han X, Lee SSY, Ingold N, et al. Associations of sleep apnoea with glaucoma and age-related macular degeneration: an analysis in the United Kingdom Biobank and the Canadian Longitudinal Study on Aging. BMC Med. 2021;19(1):104. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Keenan TDL, Goldacre R, Goldacre MJ. Associations between obstructive sleep apnoea, primary open angle glaucoma and age-related macular degeneration: record linkage study. Br J Ophthalmol. 2017;101(2):155–159. [DOI] [PubMed] [Google Scholar]
- 10.Hsieh CY, Su CC, Shao SC, et al. Taiwan’s National Health Insurance Research Database: past and future. Clinical Epidemiology. 2019;11:349–358. doi: 10.2147/clep.s196293 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Lu TH, Lee MC, Chou MC. Accuracy of cause-of-death coding in Taiwan: types of miscoding and effects on mortality statistics. Int J Epidemiol. 2000;29(2):336–343. [DOI] [PubMed] [Google Scholar]
- 12.Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992;45(6):613–619. [DOI] [PubMed] [Google Scholar]
- 13.Health Insurance Drug Item Inquiry. National Health Insurance Administration Ministry of Health and Welfare; 2022. Accessed April 2, 2024. https://www.nhi.gov.tw/en/mp-2.html [Google Scholar]
- 14.Zar JH. Biostatistical Analysis. Prentice Hall, Inc.; 1999. [Google Scholar]
- 15.McCaffrey DF, Griffin BA, Almirall D, Slaughter ME, Ramchand R, Burgette LF. A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Stat Med. 2013;32(19):3388–3414. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Lee BK, Lessler J, Stuart EA. Weight trimming and propensity score weighting. PLoS One. 2011;6(3):e18174. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Yang Dalton. A unified approach to measuring the effect size between two groups using SAS. SAS Glob Forum. http://www.lerner.ccf.org/qhs/software/lib/stddiff.pdf [Google Scholar]
- 18.Lau B, Cole SR, Gange SJ. Competing risk regression models for epidemiologic data. Am J Epidemiol. 2009;170(2):244–256. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Austin PC. The use of propensity score methods with survival or time-to-event outcomes: reporting measures of effect similar to those used in randomized experiments. Stat Med. 2014;33(7):1242–1258. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Sia JT, Lee EPX, Cheung CMG, et al. Associations between age-related macular degeneration and sleep dysfunction: A systematic review. Clin Experiment Ophthalmol. 2022;50(9):1025–1037. [DOI] [PubMed] [Google Scholar]
- 21.Kauppinen A, Paterno JJ, Blasiak J, Salminen A, Kaarniranta K. Inflammation and its role in age-related macular degeneration. Cell Mol Life Sci. 2016;73(9):1765–1786. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Brown EE, DeWeerd AJ, Ildefonso CJ, Lewin AS, Ash JD. Mitochondrial oxidative stress in the retinal pigment epithelium (RPE) led to metabolic dysfunction in both the RPE and retinal photoreceptors. Redox Biol. 2019;24:101201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Zhao XJ, Yang CC, Zhang JC, Zheng H, Liu PP, Li Q. Obstructive Sleep Apnea and Retinal Nerve Fiber Layer Thickness: A Meta-analysis. J Glaucoma. 2016;25(4):e413–8. [DOI] [PubMed] [Google Scholar]
- 24.Maniaci A, Iannella G, Cocuzza S, et al. Oxidative Stress and Inflammation Biomarker Expression in Obstructive Sleep Apnea Patients. J Clin Med Res. 2021;10(2). doi: 10.3390/jcm10020277 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Tan W, Zou J, Yoshida S, Jiang B, Zhou Y. The Role of Inflammation in Age-Related Macular Degeneration. Int J Biol Sci. 2020;16(15):2989–3001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Terluk MR, Kapphahn RJ, Soukup LM, et al. Investigating mitochondria as a target for treating age-related macular degeneration. J Neurosci. 2015;35(18):7304–7311. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Spaide RF. Choriocapillaris Flow Features Follow a Power Law Distribution: Implications for Characterization and Mechanisms of Disease Progression. Am J Ophthalmol. 2016;170:58–67. [DOI] [PubMed] [Google Scholar]
- 28.Schaal S, Sherman MP, Nesmith B, Barak Y. UNTREATED OBSTRUCTIVE SLEEP APNEA HINDERS RESPONSE TO BEVACIZUMAB IN AGE-RELATED MACULAR DEGENERATION. Retina. 2016;36(4):791–797. [DOI] [PubMed] [Google Scholar]
- 29.Penn JS, Madan A, Caldwell RB, Bartoli M, Caldwell RW, Hartnett ME. Vascular endothelial growth factor in eye disease. Prog Retin Eye Res. 2008;27(4):331–371. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Flores R, Carneiro Â, Vieira M, Tenreiro S, Seabra MC. Age-Related Macular Degeneration: Pathophysiology, Management, and Future Perspectives. Ophthalmologica. 2021;244(6):495–511. [DOI] [PubMed] [Google Scholar]
- 31.Kumar A, Kumawat D, Sundar MD, et al. Polypoidal choroidal vasculopathy: a comprehensive clinical update. Ther Adv Ophthalmol. 2019;11:2515841419831152. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Kawasaki R, Yasuda M, Song SJ, et al. The prevalence of age-related macular degeneration in Asians: a systematic review and meta-analysis. Ophthalmology. 2010;117(5):921–927. [DOI] [PubMed] [Google Scholar]
- 33.Yannuzzi LA, Sorenson J, Spaide RF, Lipson B. Idiopathic polypoidal choroidal vasculopathy (IPCV). 1990. Retina. 2012;32 Suppl 1:1–8. [PubMed] [Google Scholar]
- 34.Laude A, Cackett PD, Vithana EN, et al. Polypoidal choroidal vasculopathy and neovascular age-related macular degeneration: same or different disease? Prog Retin Eye Res. 2010;29(1):19–29. [DOI] [PubMed] [Google Scholar]
- 35.Yland JJ, Wesselink AK, Lash TL, Fox MP. Misconceptions About the Direction of Bias From Nondifferential Misclassification. Am J Epidemiol. 2022;191(8):1485–1495. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Agrón E, Mares J, Clemons TE, et al. Dietary Nutrient Intake and Progression to Late Age-Related Macular Degeneration in the Age-Related Eye Disease Studies 1 and 2. Ophthalmology. 2021;128(3):425–442. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
