Associations Between Autoimmune Disease and the Development of Age-Related Macular Degeneration

Abstract

Purpose: The pathogenesis of age-related macular degeneration (AMD) likely implicates the dysregulation of immune response pathways. Several studies demonstrate that the pathogenic elements of AMD resemble those of autoimmune diseases, yet the association between AMD development and most autoimmune diseases remain unexplored.

Methods: We conducted a case-control analysis of patients ages 55 and older with new-onset International Classification of Diseases (ICD) coding of dry, wet, or unspecified AMD between 2005 and 2019 in the Merative MarketScan Commercial and Medicare Databases. The diagnosis of an autoimmune disease was defined by an outpatient or inpatient claim with a relevant ICD code in the 12 months before the index visit. Conditional multivariable logistic regression, adjusted for AMD risk factors, was used to calculate odd ratios and 95% confidence intervals.

Results: We identified 415,027 cases with new-onset ICD coding for AMD matched with propensity scores to 414,853 controls. In total, 16.1% of cases and 15.9% of controls were diagnosed with any autoimmune disease. The diagnosis of any autoimmune disease did not affect the odds of new-onset ICD coding for AMD in multivariable regression (OR = 1.01; 95% CI, 0.999–1.02). Discoid lupus erythematosus (OR = 1.29; 95% CI, 1.12–1.48), systemic lupus erythematosus (SLE) (OR = 1.21; 95% CI, 1.15–1.27), giant cell arteritis (OR = 1.19; 95% CI, 1.09–1.30), Sjogren's syndrome (OR = 1.17; 95% CI, 1.09–1.26), and Crohn's disease (OR = 1.13; 95% CI, 1.06–1.22) increased the odds of a new-onset ICD coding for AMD.

Conclusions: Most autoimmune diseases do not affect the odds of developing AMD but several common autoimmune disorders such as SLE and Crohn's disease were associated with modestly increased odds of AMD. Further studies are needed to validate and investigate the underlying mechanisms of these associations.

Age-related macular degeneration (AMD) is a leading cause of vision loss in adults over the age of 50, contributing to an estimated 1.8 million cases of blindness worldwide.¹ In early and intermediate stages of AMD, drusen and pigmentary changes occur, whereas in advanced stages of AMD, geographic atrophy and neovascularization occur.² Our understanding of the mechanisms driving these features has improved in recent years. Dysregulation of inflammatory, complement, oxidative stress, lipid, and angiogenic pathways have been implicated in AMD pathogenesis.^3–6 Additionally, advanced age, smoking, and obesity have been identified as risk factors for AMD.^7–9 The precise manner through which these pathways and risk factors interact, however, has yet to be fully characterized. 

There is also evidence to suggest that AMD may have elements resembling those seen in autoimmune diseases. Multiple studies have noted an increased presence of autoantibodies against retinal antigens, including glial fibrillary acidic protein, a marker of astrocytes, which maintain the blood-retinal barrier.^10–13 Drusen and serum from patients with AMD have been found to have higher levels of autoantibodies against carboxyethylpyrrole compared to healthy controls, and immunization with carboxyethylpyrrole adducts in mice induces an AMD-like pathology through the development of autoantibodies.^14–16 However, it has not been established whether these autoantibodies are intrinsic to the pathogenesis of AMD or if they arise nonspecifically in response to retinal inflammation and damage. 

A few studies have turned to large-scale patient databases to further understand possible epidemiological associations between autoimmune diseases and AMD development. However, these studies have been limited to associations between AMD and psoriasis, rheumatoid arthritis (RA), and systemic lupus erythematosus (SLE).^17–20 To circumvent shortcomings of the current literature, which has focused on selected systemic diseases, we used International Classification of Disease (ICD) codes to capture a wide range of diseases. Because the associations between AMD development and most autoimmune diseases remain unexplored, we conducted a case-control study in a nationwide commercial health insurance claims database to determine them. 

Methods

We conducted a case-control study in the Merative MarketScan Commercial and Medicare Databases between January 2003 and December 2019. These annual databases include the health services of employees, dependents, and retirees in the United States with primary or Medicare supplemental coverage through privately insured health plans. Approximately 19 million patients were in the database in 2003, 57 million patients in 2012, and between 19 million and 57 million patients in all other years. The University of Chicago Institutional Review Board exempted this study because personal identifiable information was not available in these data. 

Sample Identification

We identified cases as patients ages 55 and older with new-onset ICD code of dry, wet, or unspecified AMD (Supplementary Table S1). The date of the AMD diagnosis was defined as the cases’ index visits, and every case was required to be continuously enrolled in a health insurance plan with prescription drug coverage throughout the two years before that visit. Thus cases were identified between January 2005 and December 2019. 

Cases were also required to have had an outpatient or inpatient eye examination, as defined by Current Procedural Terminology (Supplementary Table S2), on the date of their index visit and in the year before their index visit. These eye examinations confirmed that AMD was diagnosed in an ophthalmic setting and was newly-onset. 

The age of cases at their index visit was grouped as 55–64, 65–74, 75–84, and ≥85 years, and their U.S. Census Bureau region was provided as Northeast, South, North Central, West, and Unknown. We exclude cases in Unknown regions because region was a matching variable (0.24% of cases). Comorbidities were selected to compute a modified Charlson Comorbidity Index (CCI).²¹ We excluded from the CCI calculation: (1) diabetes with and without complications because type I diabetes is an autoimmune disease, (2) peripheral vascular disease to stay consistent with similar AMD-related case-control studies,^22,23 and (3) rheumatologic disease as the association between autoimmune diseases and AMD development was the focus of this study. Thus the CCI range was 0 to 25, which we grouped as 0, 1, 2, and ≥3. We also identified risk factors for AMD, including female sex and diagnoses of diabetes, hyperlipidemia, obesity, smoking, nonproliferative diabetic retinopathy (NPDR), and proliferative diabetic retinopathy (PDR). Comorbidities and risk factors required an outpatient or inpatient claim with a relevant ICD code in the 12 months before and including the index visit (Supplementary Table S1). 

Every case was matched with propensity scores estimated by age, region, CCI group, and hypertension to a single control of the same year (Supplementary Note 1 for technical details of match).²⁴ These matching variables were selected based on the design of similar case-control studies of AMD.^22,23 We did not match on sex, diabetes, NPDR, PDR, hyperlipidemia, obesity, and smoking because we wanted to report their independent effects on AMD development and to avoid overmatching bias. Controls were selected from annual control pools comprised of patients meeting the same inclusion criteria as for the cases, except without an AMD diagnosis. Controls’ index visits were randomly selected eye examinations. Controls could be in control pools in multiple years if they had not yet been matched to a case or diagnosed with AMD. 

The exposures in this study were autoimmune diseases, defined by an outpatient or inpatient claim with a relevant ICD code in the 12 months before and including the index visit (Supplementary Table S3). Autoimmune diseases were grouped into one of seven disease classes, including (1) connective tissue, systemic, rheumatic, or multi-organ: systemic lupus erythematosus, rheumatoid arthritis, psoriatic arthritis, dermatomyositis, polymyositis, polymyalgia rheumatica, systemic scleroderma, Sjogren's syndrome, ankylosing spondylitis; (2) vasculitides: giant cell arteritis, granulomatosis with polyangiitis, IgA vasculitis, polyarteritis nodosa, Behcet's syndrome; (3) gastrointestinal: inflammatory bowel disease (Crohn's disease and ulcerative colitis), celiac disease, pernicious anemia, primary biliary cholangitis, autoimmune hepatitis; (4) neurological: Guillain-Barre syndrome, myasthenia gravis, multiple sclerosis; (5) endocrine: Graves’ disease, Hashimoto's thyroiditis, Addison's disease; (6) dermatological: dermatological psoriasis, localized scleroderma, discoid lupus erythematosus (DLE); and (7) other autoimmune diseases: autoimmune hemolytic anemia, immune thrombocytopenic purpura, acute rheumatic fever, and chronic rheumatic heart disease. 

Statistical Analysis

We performed descriptive statistics of the cases, controls, and control pools. For categorical variables, this included frequencies and percentages, and for continuous variables, this included means and standard deviations. We also calculated the exposure rates of cases and controls to any autoimmune disease, the seven autoimmune diseases classes, and every autoimmune disease. 

We tested the association between new-onset ICD coding of AMD and any autoimmune disease, the seven autoimmune disease classes, and every autoimmune disease in conditional univariate and multivariable logistic regressions. We implemented conditional logistic regression to account for the matched pairs (i.e., case-control) study design. This technique ensured that cases and controls of the same year, age group (not age), CCI group, hypertension status, and region were compared,²⁵ and that nonmatched cases were still able to have explanatory power. The dependent variable in each regression was an autoimmune disease or class. The multivariable analyses adjusted for the risk factors of AMD, including female sex, type I diabetes, type II diabetes, hyperlipidemia, obesity, smoking, NPDR, and PDR. The selection of these covariates was based on the design of similar case-control studies of AMD.^22,23 

Data analysis was performed in SAS software, version 9.4. Many regression analyses were performed in this study, so the type I error rate may have been inflated. Therefore statistical significance was set a priori at α = 0.01 (i.e., P < 0.01), a more stringent threshold than α = 0.05. 

Results

We identified 415,027 cases with new-onset ICD coding of AMD. Compared to the 27,870,104 patients in control pools without AMD and whom the controls were selected from, cases were older, more frequently resided in the Northeast and North Central regions, had higher average CCIs, and were more often diagnosed with hypertension (Table 1). 

Table 1.

 

Sample Characteristics Before Matching

 

Matching identified 414,853 controls. The 174 nonmatched cases were all 85 years and older (Table 2). Compared to controls, cases were more often female and diagnosed with NPDR and PDR and less often diagnosed with type 2 diabetes, hyperlipidemia, and obesity (Table 2). Cases and controls were balanced on the matching variables of age, region, year, CCI group, and hypertension, as defined by the absolute values of their standardized differences and distributions of their estimated propensity scores (Supplementary Tables S4 and S5, respectively).²⁶ Of the 415,027 cases, 284,133 (68.5%) had dry AMD, 37,505 (9%) had wet AMD, and 93,389 (22.5%) had unspecified AMD (Table 2). 

Table 2.

 

Sample Characteristics

 

Overall, 16.1% of cases and 15.9% of controls were diagnosed with any autoimmune disease (Table 3). The case and control exposure rates to the seven autoimmune disease classes were 5.4% and 5.2% for connective tissue, systemic, rheumatic, or multiorgan diseases; 1.6% and 1.6% for vasculitides; 3.4% and 3.2% for gastrointestinal; 1.6% and 1.5% for neurological; 0.7% and 0.7% for endocrine; 3.7% and 3.7% for dermatological; and 2.8% and 2.8% for other autoimmune diseases. The five most common autoimmune diseases among cases and controls were localized scleroderma, chronic rheumatic heart disease, rheumatoid arthritis, pernicious anemia, and IgA vasculitis. 

Table 3.

 

Autoimmune Class and Disease Exposures

 

Any autoimmune disease did not affect the odds of new-onset ICD coding for AMD in multivariable regression (OR = 1.01; 95% CI, 0.999–1.02; P = 0.08; Table 4). Among the autoimmune disease classes, connective tissue, systemic, rheumatic, or multiorgan and gastrointestinal diseases increased the odds of new-onset ICD coding for AMD in multivariable regression (connective tissue, systemic rheumatic, or multiorgan: OR = 1.03; 95% CI, 1.01–1.05; P = 0.0074; gastrointestinal: OR = 1.04; 95% CI, 1.02–1.07; P = 0.0009); vasculitides, neurological, endocrine, dermatological, and other autoimmune diseases did not affect the odds. Among the autoimmune diseases, systemic lupus erythematosus (OR = 1.21; 95% CI, 1.15–1.27; P < 0.0001), Sjogren's syndrome (OR = 1.17; 95% CI, 1.09–1.26; P < 0.0001), giant cell arteritis (OR = 1.19; 95% CI, 1.09–1.30; P = 0.0002), inflammatory bowel disease (OR = 1.09; 95% CI, 1.04–1.14; P = 0.0002), Crohn's disease (OR = 1.13; 95% CI, 1.06–1.22; P = 0.0005), and discoid lupus erythematosus (OR = 1.29; 95% CI, 1.12–1.48; P = 0.0005) increased the odds of new-onset ICD coding for AMD in multivariable regressions; the remaining autoimmune diseases did not affect the odds (Table 5). 

Table 4.

 

Univariate and Multivariable Conditional Logistic Regression Models for Autoimmune Classes (N = 829,880)

 

Table 5.

 

Odds Ratios of Autoimmune Diseases from Univariate and Multivariable Conditional Logistic Regression (N = 829880)

 

Discussion

In this nationwide case-control study of 415,027 cases of AMD and 414,853 matched controls, the presence of any autoimmune disease did not affect the odds of new-onset ICD coding of AMD, adjusting for sex, diabetes, NPDR, PDR, hyperlipidemia, obesity, and smoking. This is a reassuring finding and suggests that increased surveillance for AMD beyond what is recommended by age-based guidelines²⁷ is likely unwarranted for the majority of autoimmune diseases. Several autoimmune diseases, specifically SLE, Sjogren's syndrome, giant cell arteritis, Crohn's disease, and DLE, modestly increased the odds of new-onset ICD coding of AMD. 

Overall, 16.1% of AMD cases and 15.9% of controls carried an ICD code diagnosis of any autoimmune diseases. Within the developed world, the estimated prevalence of autoimmune diseases varies from 4.5% to 10%,^28,29 although it is widely suggested that the prevalence of autoimmune disease is rising.³⁰ Regardless, the prevalence of autoimmune disease in this study is much greater than what has been reported previously. This disparity may be due to enrolling an older patient population, with inclusion criteria requiring that cases and controls be at least 55 years in age. Such criteria likely capture patients diagnosed with an autoimmune disease regardless of its typical age of onset, given that such conditions are typically chronic and thus carried forward in the medical record. 

Our finding of an association between SLE and increased odds of new-onset ICD coding for AMD aligns with a previous case-control study of the United Kingdom General Practice Research Database. That study enrolled 18,007 cases with AMD and matched 86,169 controls on age, sex, and practice location, and it found an association between SLE and AMD (OR = 1.69; 95% CI, 1.05–2.72) after adjustment for confounders, including smoking, alcohol intake, body mass index, atherosclerotic disease, diabetes, heart failure, hyperlipidemia, hypertension, exposure to aspirin, exposure to hormone replacement therapy, exposure to cardiovascular disease drugs, and observation time in the database.²⁰ The smaller OR we observed herein may reflect the greater precision that comes with a larger population of cases and controls. Additionally, the authors of that study reported that AMD cases were likely advanced in nature with visual complaints at presentation.²⁰ We captured all stages of AMD, which may further contribute to differences in ORs between the studies. 

A small observational study of 65 patients with SLE matched 1:1 to controls on age and sex found patients with SLE had significantly greater levels of drusen, including medium or large drusen,³¹ the latter of which has been shown to increase the risk of progression to advanced AMD.³² The pathogenesis of SLE is hallmarked by complement system activation by circulating immune complexes and an accentuated inflammatory response.³³ Paradoxically, deficiencies of early complement proteins, particularly C1q and C4, are strongly linked with SLE development.³³ Polymorphisms of complement factor H have been identified as key genetic risk factors for AMD.^34,35 It has been proposed that these polymorphisms lead to sustained complement activation and inflammation that precipitates drusen formation and retinal pigment epithelium damage, both of which are hallmarks of AMD.^36,37 Hence, shared dysregulation of the complement system may underlie the association between SLE and AMD development in this study. However, AMD and SLE pathogenesis implicate different components of the complement pathway. Hence, further studies are required to clarify the mechanisms that drive the potential association we identified. 

Discoid lupus erythematosus was associated with the highest OR for new-onset ICD coding for AMD among all autoimmune diseases in this study. Approximately 20% of patients with SLE develop manifestations of DLE.³⁸ Patients with DLE have been shown to have lower levels of autoantibodies against nuclear antigens compared to patients with SLE, an indication that DLE has lower levels of systemic disease activity.^39,40 Hence, the association between DLE and AMD development may be driven by patients who carry an underlying diagnosis of SLE, where there is sustained inflammation that can cause disease involvement at distant organs. 

We also identified an association between inflammatory bowel disease (IBD), specifically Crohn's disease, and an increased odds ratio of new-onset ICD coding for AMD that has not previously described. In recent years, the concept of a gut-retina axis has been proposed that seeks to explain how dysbiosis, an imbalance of the gut microbiota, can propagate to the distant retina and precipitate disease activity.^41,42 Dysbiosis is linked with greater permeability of the intestinal wall, thereby facilitating the translocation of microbial endotoxins, which increase levels of systemic inflammation.⁴³ Gut epithelium hyperpermeability has been well documented in patients with both Crohn's disease and ulcerative colitis, allowing for bacteria and their byproducts to propagate systemically throughout the bloodstream.⁴⁴ Andriessen et al.⁴⁵ triggered gut dysbiosis by feeding mice a high fat diet and found that this state exacerbated choroidal neovascularization, likely through elevated cytokine production associated with intestinal hyperpermeability. Zhang et al.⁴⁶ have also shown in a germ-free mice model that the gut microbiome regulates chorioretinal gene expressions and the formation of choroidal neovascularization. Hence, the emerging gut-retina axis may help to explain the association between IBD and AMD development that we have identified. Sustained, low-grade inflammation may eventually propagate to the retina and precipitate the pathogenic events required for the inception of AMD. 

We additionally found that both Sjogren's syndrome and giant cell arteritis increased the odds of a new-onset ICD coding for AMD. Sjogren's syndrome is known to cause symptoms of dry eye due to lymphocytic infiltration and autoimmune-mediated destruction of the lacrimal glands,⁴⁷ whereas giant cell arteritis increases the risk of vision loss because of ischemia of the optic nerve from systemic inflammation.⁴⁸ However, links between these diseases and AMD have not been previously described. Sjogren's syndrome has previously been reported to occur in up to 14% to 17.8% of patients with SLE, as estimated by recent meta-analyses.^49,50 Hence, the association we report may be due to confounding by a co-occurrence of an SLE diagnosis in patients with Sjogren's syndrome. 

For now, it is unclear whether patients with SLE, IBD, Sjogren's syndrome, and giant cell arteritis should undergo increased surveillance for AMD. Preventative measures, such as smoking cessation and adherence to a Mediterranean diet, have been shown to decrease the risk of AMD development, so their potential benefits may warrant investigation or use among patients with these autoimmune diseases.^7,51,52 Patients with Sjogren's syndrome and giant cell arteritis likely follow up with an ophthalmologist for ocular complications related to those respective conditions. However, the associations between either SLE or IBD and AMD development need to be further validated before screening guidelines can be considered. 

A diagnosis of RA did not significantly affect the odds of new-onset ICD coding for AMD in this study. A prospective study of patients with RA estimated that the prevalence of AMD was tenfold lower in patients with RA compared with those in the general population.⁵³ The authors hypothesized that this risk reduction was mediated by anti-inflammatory agents used to treat RA. However, these findings have been disputed by two subsequent observational studies.^18,19 Keenan et al.¹⁸ constructed a cohort of 261,232 patients with RA and a reference cohort of 2,134,771 patients from an English dataset. Patients with RA had a modestly elevated risk ratio of AMD at 1.15 (95% CI, 1.12-1.19) that was highest after the first documentation of RA. Schnabolk et al.¹⁹ performed a retrospective cohort study of the MarketScan Databases and identified 37,252 patients with a diagnosis of RA and 37,094 control patients without an RA diagnosis. An RA diagnosis increased the risk of dry AMD by an odds ratio of 2.076 (95% CI, 1.98–2.18) but did not affect the risk of wet AMD. Additionally, the risk of AMD was not altered by the presence of monoclonal antibody therapy against tumor necrosis factor. Our findings conflict with these previous studies, which have suggested that RA is either protective against or increases the risk of AMD. It is possible that medications used to treat RA are biasing our findings towards the null. For example, in a clinical trial of 110 patients randomized to receive either 12 months of daily 400 mg hydroxychloroquine, an immunosuppressive agent used to treat or RA, or placebo,⁵⁴ patients receiving hydroxychloroquine showed slower visual deterioration and decreased formation and progression of drusen.⁵⁴ Thus future studies should control for the potential effects of autoimmune disease treatments. 

A previous case-control study of the Taiwan Longitudinal Health Insurance Database analyzed 2187 cases of wet AMD who were matched to 10,935 controls.¹⁷ Wet AMD was associated with a prior diagnosis of psoriasis (OR = 1.52; 95% CI, 1.07-2.15). We did not find an association between either psoriasis or psoriatic arthritis and new-onset ICD coding for AMD in this study. This conflicting data may be due to our combined analysis of dry, wet, and unspecified AMD. As expected, only 37,505 (9%) cases had wet AMD, so subgroup analysis would likely be required to corroborate the association uncovered in the Taiwan Database. Otherwise, there are also underlying differences in the enrolled study populations, because patients from the Taiwanese study were reported to be of majority Chinese ethnicity. 

Strengths of the study are as follows. We used a higher threshold to determine statistical significance (α = 0.01 vs. α = 0.05) because many regression analyses were performed. This methodology increases the type I error rate (i.e., rate of false-positive results). However, a higher threshold also increases the type II error rate (i.e., rate of false-negative results). Ulcerative colitis (OR = 1.06; 95% CI, 1.001-1.12) and primary biliary cholangitis (OR = 1.23; 95% CI, 1.03-1.46) would have been associated with increased odds of new-onset ICD coding of AMD had α = 0.05. Thus these diseases should be explored in future studies to assess for potential associations with AMD. 

A further strength of this study is that each multivariable regression adjusted for the risk factors of AMD. However, these regressions did not account for the fact that among the cases and controls with any autoimmune diseases, 21.8% of cases and 20.7% of controls had two or more autoimmune diseases. This finding is not necessarily surprising. Autoimmune diseases have been shown to co-occur in multiple observational studies.^29,55 Therefore the associations between AMD development and each autoimmune disease could be confounded by other autoimmune diseases. Although we did not investigate multicollinearity and interaction effects of autoimmune diseases in our study, future studies should do so, in addition to selecting patients with no co-occurrence of disease so that the association between that disease and AMD can be estimated precisely. 

Further limitations of the study include the possibility of confounding by anti-inflammatory and immunomodulatory medications used to treat autoimmune diseases. The agents used to treat autoimmune disease may also confound their associations with AMD development. We did not specify AMD type, including dry and wet forms or even by dry AMD subtypes, such as geographic atrophy. It is plausible to suspect that the associations we reported may change based on disease staging. Additionally, the use of an insurance claims database did not allow us to capture clinical characteristics that would stratify patients based on AMD severity, such as drusen size. Furthermore, coding practices are likely to be inconsistent across providers and may be prone to inaccuracies or missing diagnoses. Last, we did not undergo a formal variable selection process and instead selected covariates based on the design of similar case-control studies of AMD.^22,23 Therefore we may have included covariates that did not enhance model fit. 

It is unclear whether these statistically significant results also carry clinical significance given that the ORs in this study are modest. However, the prevalence of autoimmune diseases is rising and affects as many as 10% of individuals as of 2000 to 2019.^29,30 Therefore a sizeable and growing proportion of the general population may face increased odds of developing AMD. To further understand the degree to which the associations we describe are clinically meaningful, several follow-up studies may be useful. Studies should explore whether patients with autoimmune diseases are diagnosed with AMD at earlier ages and investigate for which subtypes of AMD these associations persist. For example, neovascular AMD accounts for only 10% of AMD diagnoses but causes up to 90% of its severe vision loss.⁵⁶ Associations with this subtype would thus be particularly meaningful. Additionally, it will be valuable to understand whether more severe manifestations of autoimmune diseases confer increased risk of AMD. With respect to SLE, for example, a severity index could be used to assess for a dose-response relationship between worsening SLE and increased odds of AMD development.⁵⁷ Alternatively, consideration could be given toward laboratory markers, which have been shown to correlate with disease activity in patients with Crohn's disease.⁵⁸ These quantitative markers would more precisely characterize the relationship between Crohn's disease and AMD development. 

In conclusion, our results are reassuring, demonstrating no association between the majority of autoimmune diseases and AMD development. Diseases such as SLE and IBD, however, may increase the odds of developing AMD. Further observational studies should confirm these results among different AMD subtypes while also considering the potential impact of immunosuppressive medications on AMD development. 

Acknowledgments

Supported by the UChicago Institute of Translational Medicine and the Bucksbaum Institute for Clinical Excellence. 

Disclosure: J. Moir, None; M.J. Hyman, None; J. Wang, None; A. Flores, None; D. Skondra, Biogen (C), Allergan (C), Focuscope (C), LaGrippe Research (C), Trinity Life Sciences (C)