Background
Sleep apnoea is a common sleep-disordered breathing condition that affects an estimated 17% of women and 34% of men aged 30–70 in the general population [
1‐
3]. It is characterised by repetitive reduction or cessation of airflow in the upper airway during sleep [
4] which is believed to result in transient hypoxia hypoperfusion of the optic nerve and spikes in intracranial pressure, all of which elevate the risk of optic neuropathy [
5]. Indeed, higher rates of various forms of optic neuropathy, including non-arteritic anterior ischemic optic neuropathy [
6,
7] and papilloedema [
8], have been found in patients with sleep apnoea relative to controls.
Several previous studies have examined the link between obstructive sleep apnoea (OSA) and glaucoma, the most common cause of optic neuropathy, but findings have been inconsistent [
6,
9‐
18]. Glaucoma itself is a leading cause of irreversible blindness worldwide [
19], and the number of people with glaucoma will increase to 76 million in 2020 and to 112 million in 2040 [
19,
20]. In a longitudinal study of over 7000 adults over 40 years old, Lin et al. [
10] reported that those with OSA have a 5-year glaucoma hazard ratio of 1.7, adjusted for comorbidities. Meta-analyses [
13‐
15] have similarly estimated pooled odds ratios for glaucoma of up to 2.5 in those with OSA and 5.5 in those with severe OSA. Additionally, findings from several studies have generally agreed that individuals with OSA have, on average, thinner retinal nerve fibre layers (RNFL) [
2,
3,
21‐
23] and macular ganglion cell complex [
24] compared to age-matched controls. Recently, Wozniak et al. [
25] reported that the presence of OSA was independently associated with a faster rate of RNFL thinning in patients with primary open-angle glaucoma, after accounting for comorbidities. However, a number of large studies [
6,
16,
18] have failed to find a significantly increased rate of glaucoma in patients with OSA. For example, Keenan et al. [
18] reviewed over 3 million hospital records of patients aged 55 years and over and reported that the rate ratio for glaucoma in patients with sleep apnoea was not increased compared to the reference cohort.
Age-related macular degeneration (AMD) causes progressive visual impairment with its prevalence increasing within ageing populations [
26]. Interestingly, Keenan et al. [
18] found that patients with sleep apnoea had a 50% increased rate of subsequent AMD diagnosis, compared to a reference cohort. There have also been case reports [
27,
28] of poorer response to anti-vascular endothelial growth factor therapy for exudative AMD in patients with sleep apnoea. However, given that only one study has reported on the possible increased incidence rates of AMD in patients with sleep apnoea, more data are required to assess this link.
Findings on the association between sleep apnoea and glaucoma have been conflicting and very few studies have examined the link between the former condition and AMD. Large-scale cohort studies would provide evidence to determine the risks of glaucoma and AMD in patients with sleep apnoea relative to participants without the condition. In this study, we investigated the associations using two large prospective cohort studies.
Discussion
In two large-scale cohorts of middle-aged and older adults from the UK and Canada, we found a higher incidence of glaucoma and AMD in participants with sleep apnoea relative to those without sleep apnoea. The associations were independent of age, sex, socioeconomic status, ethnic group, smoking, diabetes, cardiovascular diseases, blood pressure, and cholesterol levels.
Despite plausible mechanisms linking sleep apnoea and glaucoma, the relationship has been debated among researchers. Previous studies reporting an association between sleep apnoea and glaucoma have focused on Asian populations [
10,
41]. For example, Lin et al.’s population-based matched-cohort study in Taiwan [
10] reported a 5-year glaucoma HR of 1.7 in those with sleep apnoea, relative to those without sleep apnoea. This figure is not too dissimilar to the current findings of 1.3 and 1.4 in the UK and Canada, respectively. However, our results do not agree with those of Keenan et al. [
18] whose study was also based in the UK but did not report a higher rate of glaucoma in those with sleep apnoea. Indeed, most studies reporting no association [
16,
17] or heterogeneous results [
13,
14] on the relationship between sleep apnoea and glaucoma risk had examined European or Caucasian populations. Thus, it could be possible that any association between the two conditions is more apparent in Asian populations. Unfortunately, we were unable to do further subgroup analysis for the Asian ethnic group due to the relatively small numbers of participants of Asian ancestry in the current study.
It remains curious as to why the current findings varied from those of Keenan et al. given that both studies were conducted in the UK population. A possible explanation for the difference may be the method of sampling and determining sleep apnoea, glaucoma, and AMD cases. Keenan et al. relied only on hospital records and thus may not have captured all (or even) most cases except for the most severe ones. An issue with only limiting analyses to severe cases is that it assumes that the hypothesis severity of sleep apnoea is linked with the development of glaucoma or AMD; however, it is still unclear if this is the case. The current study additionally utilised self-reported data and general practitioner record linkage, in addition to hospital-linkage records, which allowed us to include the mild and moderate cases which do not require any in-clinic procedure (e.g. sleep apnoea controlled by lifestyle modifications, medically controlled glaucoma, and non-exudative AMD). While we found a comparable glaucoma incidence to Lin et al.’s population-based matched-cohort study in Taiwan [
10], the latter study reported a high incidence rate of glaucoma (per 1000 person-years of 11.26 and 6.76 for subjects with and without OSA, respectively), nearly four times that in the UK Biobank. The higher incidence rate in Lin et al.’s study may reflect the higher prevalence of glaucoma in East Asian countries (3 to 4%) compared to Europe (2.4%) [
10,
19]. Moreover, Lin et al.’s cohort included a relatively large proportion of older participants (more than 70 years), which is likely to have a higher glaucoma incidence, relative to the UK Biobank (40 to 69 years) [
10]. From the CLSA study, on average, the participants were 5 years older than those from the UK Biobank; indeed, the incidence rate of glaucoma was also higher in the CLSA cohort than the UK Biobank. Moreover, the prevalence of sleep apnoea in the CLSA is much higher than that in the UK Biobank, which could be due to the higher prevalence of obesity and older population in the CLSA as well as different methods of ascertaining sleep apnoea cases. However, we observed similar association patterns for different baseline characteristics with sleep apnoea in the two large biobanks (Table
1 and Table
2), suggesting they are largely comparable. Furthermore, the associations between sleep apnoea and the risk of glaucoma or AMD are similar in the two cohorts.
In our sex subgroup analysis, the association in women seemed to be attenuated in the CLSA relative to the UK Biobank. However, the confidence intervals largely overlap. Furthermore, the confidence intervals were wider in the CLSA compared with the UK Biobank, and we found no evidence of interaction effects of sleep apnoea with sex in both the UK Biobank and the CLSA. These results indicate there might be no meaningful differences for the associations between men and women between the UK Biobank and CLSA.
Instead of using the risk of incidence for glaucoma as an outcome measure, some studies used pre-clinical measures of retinal nerve fibre layer to investigate the link between OSA and optic neuropathy, and consistently found thinner retinal nerve fibre layers in middle-aged or older adults with sleep apnoea but without glaucoma [
3,
23,
24,
42]. Recently, we reported that an association between thinner RNFL thickness and OSA could even be observed in young adults of European ancestry with healthy eyes [
2]. In a supplementary report, Wozniak et al. [
25] observed that the rate of global RNFL thinning in patients with glaucoma was on average twice as fast in those with co-existing OSA, compared to those with no OSA; hence the presence of OSA may not only be a risk factor of glaucoma, but it may also accelerate glaucoma progression.
Few studies have investigated the association between sleep apnoea and AMD. Our study revealed that sleep apnoea may be associated with an increased AMD risk, supporting Keenan et al.’s record-linkage study [
18] that showed a 44% increase in AMD incidence in those with OSA, relative to a reference cohort. There is evidence that hypoxia and oxidative stress associated with OSA activates inflammatory processes [
43,
44], which are believed to play a key role in the pathogenesis of AMD [
45‐
47]. Individuals with sleep apnoea have also been observed to have thinner choroids [
48], a manifestation of AMD [
49], relative to controls, suggesting poorer perfusion to the outer retinal layers. Our results suggest patients with sleep apnoea might benefit from eye health screening, not only for the early detection of glaucoma but also AMD.
A main strength of the current analysis is the large scale and prospective nature of the UK Biobank and CLSA, which have detailed information on socioeconomic characteristics, physical measurements, and biomarkers from blood assays that enabled us to adjust for potential confounders in our analysis. The prospective studies also allow us to evaluate the reverse associations between glaucoma and AMD with the risk of sleep apnoea, where glaucoma or AMD status at baseline was used as the exposure, and the outcome was the incident risk of sleep apnoea during follow-up. We found no evidence of a reverse association between glaucoma and sleep apnoea risk in both UK Biobank and CLSA cohorts (Additional file
1: Tables S9 and S10), indicating the main findings that sleep apnoea is associated with a higher risk of glaucoma are unlikely to be driven by confounding factors. We observed some evidence of the reverse association between AMD and sleep apnoea risk in the UK Biobank but not in the CLSA. It is possible that there are some shared risk factors for both sleep apnoea and AMD in the UK Biobank to lead to the reverse association results, although various potential confounders were adjusted for (e.g. model 5 in Table
3). These results suggest glaucoma is not associated with the risk of sleep apnoea. However, the clinical utility of screening for sleep apnoea in patients with AMD needs further study. In our UK Biobank analysis, we used ICD codes to identify cases, which only captures patients who needed a procedure or an inpatient stay for that disorder, potentially missing cases who only had their condition diagnosed as an outpatient. To address this, we used self-reported information and record-linkage data to capture these cases. However, a limitation of self-reported data is its susceptibility to recall error, especially in age at diagnosis or misunderstanding of disease (e.g. misinterpret glaucoma as cataract). This is unlikely to affect the data significantly as the majority cases were identified from health episode statistics and general practitioner record-linkage data (Fig.
1 and Additional file
1: Figure S1). When restricting the analysis to participants having any hospital inpatient records, the association results were essentially unchanged (Additional file
1: Table S11). We further removed self-reported sleep apnoea, glaucoma, or AMD cases, and the results are essentially the same. An inherent limitation when dealing with large population-based samples is the limited ability to ascertain a high-level detail of the phenotype or disease. For instance, the detailed information for glaucoma and AMD subtypes, and continuous variables, such as apnoea-hypopnoea index, are not available, which impede further analysis for different disease subtypes and linear correlation analyses for continuous variables. Since different disease subtypes, such as low-tension glaucoma and high-tension glaucoma, may have different pathological mechanisms, their association patterns with sleep apnoea could be different [
22,
50]. We did identify associations between sleep apnoea and broadly defined glaucoma and AMD, the true effect may be larger in particular subtypes. However, it would be very difficult to collect the detailed data in large-scale population-based studies such as the UK Biobank and the CLSA. A further limitation of this analysis is the lack of detailed information on treatment for sleep apnoea, such as with continuous positive airway pressure (CPAP). However, the use of CPAP is likely to ameliorate glaucoma and AMD, and it will generally bias toward the null. Moreover, a genetic study has found a high degree of homogeneity between cases from hospital records and self-reported questionnaires in the UK Biobank [
51]. There is evidence of a “healthy volunteer” selection bias in the UK Biobank [
52], where we observed lower incidence rates for glaucoma and AMD compared with previous reports. However, a recent study has shown that risk factor associations in the UK Biobank are broadly generalisable [
53].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.