Janus Kinase Inhibitor Therapy and Risk of Age-Related Macular Degeneration in Autoimmune Disease

Key Points

Question

Are Janus kinase inhibitors (JAKis) in autoimmune diseases associated with a reduction in age-related macular degeneration (AMD)?

Findings

Using MarketScan and Optum administrative claims data (2010-2022), this cohort study compared AMD incidence among 29 586 propensity score–matched patients who received JAKi-based immunotherapy vs non–JAKi-based immunotherapy for autoimmune diseases. Those who received JAKi therapy had a lower AMD incidence in both MarketScan and Optum than those who received non-JAKi therapy.

Meaning

Adults with autoimmune diseases who receive JAKi therapy may have a reduced risk of AMD onset.

This cohort study evaluates the association between Janus kinase inhibitor therapy and age-related macular degeneration in autoimmune disease.

Abstract

Importance

The involvement of chronic inflammation in the pathogenesis of age-related macular degeneration (AMD) opens therapeutic possibilities to AMD management.

Objective

To determine whether Janus kinase inhibitors (JAKis) are associated with a reduced risk of AMD development in patients with autoimmune diseases.

Design, Setting, and Participants

This retrospective observational cohort study used administrative claims data from Merative MarketScan research databases (Commercial and Medicare Supplemental) and Optum Clinformatics Data Mart databases between January 1, 2010, and January 31, 2022. Patients with autoimmune diseases satisfying study eligibility criteria and who received JAKi treatment (9126 in MarketScan and 5667 in Optum) were propensity score matched (1:1) to identical numbers of study-eligible patients who received non–JAKi-based immunotherapy.

Exposure

Treatment duration of 6 months or longer.

Main Outcomes and Measures

Incidence rates of AMD (exudative and nonexudative) over the first 6 to 18 months of treatment were determined, and bayesian Poisson regression models were used to estimate incidence rate ratios, 95% CIs, and posterior probabilities of AMD.

Results

After matching, female sex represented the majority of the patient population in both MarketScan and Optum (14 019/18 252 [76.6%] and 8563/3364 [75.2%], respectively in the JAKi patient population). More than 60% of the patient population was older than 55 years of age in both cohorts. Over the specified treatment period, a 49% relative reduction in incidence of AMD was observed among patients who received JAKi therapy (10/9126 events; adjusted incidence rate ratio [AIRR], 0.51; 95% CI, 0.19-0.90) vs those who received non-JAKi therapy (43/9126 events; AIRR, 1 [reference]) in MarketScan, and a 73% relative reduction in incidence of AMD was observed among patients who received JAKi therapy (3/5667 events; AIRR, 0.27; 95% CI, 0.03-0.74) vs those who received non-JAKi therapy (21/5667 events; AIRR, 1 [reference]) in Optum. The absolute percentage reductions were 0.36% (MarketScan) and 0.32% (Optum), favoring patients who received JAKi therapy. Posterior probabilities of the adjusted risk being less than unity were 97.6% (MarketScan) and 98.9% (Optum) for those who received JAKi therapy vs those who received non-JAKi therapy in MarketScan and Optum, respectively.

Conclusions and Relevance

JAKi use may be associated with a reduced risk of incident AMD in US adults with major autoimmune diseases. The absolute percentage reduction is consistent with a potential role for JAKi in this population. Future studies with long-term follow-up are recommended to investigate the association between JAKi use and incident AMD in other disease indications. Investigation into the role of systemic inflammation and JAK–signal transducers and activators of transcription signaling in AMD may improve understanding of the pathophysiology of AMD and lead to new treatment options.

Introduction

Age-related macular degeneration (AMD), a progressive degeneration of the outer retina and retinal pigment epithelium, is the leading cause of irreversible moderate to severe vision loss among individuals 50 years and older worldwide.¹ AMD progresses from early and intermediate stages with drusen deposition between the retinal pigment epithelium and the Bruch membrane, pigmentary changes in the retinal pigment epithelium, and subtle visual changes, to an advanced stage with choroidal neovascularization or geographic atrophy and irreversible central vision loss.² In 2020, approximately 19.8 million individuals had AMD in the US³ and 196 million worldwide.⁴

Although the etiology is incompletely understood, evidence suggests a role for inflammation in AMD pathogenesis and progression.^5,6,7 Chronic low-level local inflammation is indicated by the presence of inflammatory mediators in drusen⁸ and inflammatory cell infiltrates in affected ocular tissues.⁹ Polymorphism has been identified in several immune-related genes associated with the risk of AMD.¹⁰ A recent transcriptomewide association study reported marked upregulation of genes associated with complement system activation in advanced AMD (ocular and nonocular tissues).¹¹ Chronic inflammation caused by complement system overactivation is implicated in advanced AMD with geographic atrophy (retinal pigment epithelium/photoreceptor loss).¹²

Systemic inflammation may also play a role in AMD pathogenesis.^13,14,15,16 Myeloproliferative neoplasms are associated with chronic inflammation^17,18 and an increased risk of AMD.^13,14 Other studies have shown an association between AMD and gut microbiota alterations that may promote inflammation.¹⁵ Moreover, use of oral metformin (which may reduce chronic low-grade inflammation) is associated with a decreased AMD risk in patients with and without diabetes.^16,19

Among key inflammation mediators, Janus kinase (JAK) tyrosine kinases are critical in enabling cellular signal transduction at cytokine and growth factor receptors.²⁰ The JAK signal transducers and activators of transcription (STAT) signaling pathway has a central role in immune system activation, inflammation, and hematopoiesis, and its dysregulation has been implicated in various autoimmune diseases and malignancies.²⁰ Expression or activation of JAK-STAT signaling components has been demonstrated in human retinal pigment epithelium cells exposed to inflammatory cytokines,^21,22 murine models of light-induced and inherited retinal degeneration,^23,24,25 and choroidal neovascular membranes from patients with AMD.²⁶ Given the potential involvement of JAK-STAT–mediated inflammation in AMD pathogenesis, we postulated that use of JAK inhibitor (JAKi) therapy in at-risk individuals may decrease AMD incidence. This study of real-world data compared the AMD incidence between patients who received JAKi- and non–JAKi-based immunotherapy for autoimmune diseases.

Methods

Study Design and Data Source

This retrospective, observational cohort study used US administrative claims data (January 1, 2010, through January 31, 2022) from the Merative MarketScan research databases (MarketScan Commercial and MarketScan Medicare Supplemental) and Optum Clinformatics Data Mart databases. The MarketScan databases provide integrated, longitudinal patient-level claims data from commercial and public (including Medicare and Medicaid) insurance programs and include records for more than 273 million individuals from 1995 to date. Optum’s Clinformatics Data Mart comprises commercial and Medicare Advantage longitudinal medical and pharmacy claims data from approximately 17 to 19 million patients across the continental US. All data used herein were deidentified, per the Health Insurance Portability and Accountability Act. Institutional review board approval was waived for this study. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Patient Identification

Both families of databases were screened to identify, using International Classification of Diseases, Ninth and Tenth Revisions (ICD-9 and ICD-10), Clinical Modification (CM) codes (eTable 1 in Supplement 1), patients who (1) had 1 or more diagnoses of Crohn disease, ulcerative colitis, rheumatoid arthritis, psoriatic arthritis, ankylosing spondylitis, atopic dermatitis, psoriasis, hidradenitis suppurativa, or polyarticular juvenile idiopathic arthritis between January 1, 2010, and June 30, 2021; (2) had received a first JAKi prescription (upadacitinib, baricitinib, tofacitinib, or fedratinib), other immunotherapy, namely antitumor necrosis factor agents (adalimumab, certolizumab, etanercept, golimumab, or infliximab), interleukin (IL)–23 (risankizumab, guselkumab, tildrakizumab, or ustekinumab), IL-6 (tocilizumab or sarilumab), IL-17 (secukinumab, ixekizumab, or brodalumab), immunoglobulin E (omalizumab), or other disease-modifying antirheumatic drug (hydroxychloroquine, sulfasalazine, methotrexate, leflunomide, rituximab, abatacept, azathioprine, cyclosporine, apremilast, or mycophenolate mofetil), as identified by the relevant National Drug Code, Current Procedural Terminology/Healthcare Common Procedure Coding System, or ICD, Tenth Revision (ICD-10), Procedure Coding System code (eTable 2 in Supplement 1), between July 1, 2010, and June 30, 2021 (index period); (3) were 40 years or older on the date of first immunotherapy prescription (index date); and (4) had been continuously enrolled in the database for 6 months or longer immediately before and after the index date. Patients who had received prescriptions for both JAKi- and non–JAKi-based immunotherapy were assigned to the JAKi treatment group (prior treatment was not a reason for patient exclusion).

Patients with 1 or more diagnosis codes for AMD (nonexudative, drusen [degenerative], or exudative; ICD-9-CM codes 362.50, 362.51, 362.52, and 362.57; ICD-10-CM codes H35.31, H35.32, and H35.36) during the preindex period (ie, between first diagnosis and initiation of treatment) were excluded from the study. Patients who switched from non-JAKi therapy to JAKi therapy (or vice versa) or discontinued treatment during the postindex period (after index treatment initiation) were censored at the time of the postindex switch or discontinuation.

Propensity Score Matching

To minimize selection bias, propensity score matching (1:1) between patients who received JAKi therapy and those who received non-JAKi therapy was performed.²⁷ Propensity scores were estimated using a logistic regression model: the dependent variable was JAKi exposure and the independent variables (covariates) included age at index date, sex, geographic region, race and ethnicity (Optum database only; reported here as recorded in Optum and included because it is considered an important covariate), first autoimmune disease diagnosis, time from first autoimmune disease diagnosis to index date, Charlson Comorbidity Index (CCI) category, and nonocular baseline comorbidity (dyslipidemia, hypertension). A look-back period of 6 months was applied for identification of events preceding the index date. Each patient exposed to JAKi therapy was matched with the closest patient exposed to non-JAKi therapy with regard to age, sex, region, race and ethnicity (available in the Optum database only), first autoimmune disease diagnosis, time from first autoimmune disease diagnosis to index date, CCI category, dyslipidemia, and hypertension using a 1:1 nearest neighbor matching technique, with a caliper of 0.2 × standard deviation of the estimated logit propensity score.²⁸ For several of these covariates (age category, CCI category, and first autoimmune disease diagnosis), an exact match between patient and partner (ie, standardized mean difference [SMD] of zero) was required.

The success of propensity score matching was assessed by comparing the prematch and postmatch balance of identified covariates. A between-cohort SMD (mean difference of the average standard deviation of the variable’s distribution across the 2 cohorts) of 0.1 or less was considered indicative of good balance. Key baseline demographic and clinical covariates with a between-cohort SMD greater than 0.1 were subjected to backward or piecewise selection based on their deemed association with exposure (JAKi treatment) and outcome in the logistic regression model.

Outcome Measures

The primary study end point was AMD development after immunotherapy initiation, indicated by the presence of an AMD diagnosis of the nonexudative (ICD-9-CM codes 362.50, 362.51; ICD-10-CM code H35.31), drusen (degenerative) (ICD-9-CM code 362.57; ICD-10-CM code H35.36), or exudative (ICD-9-CM code 362.52; ICD-10-CM code H35.32) type. Patients were followed up from treatment initiation (index date) until an AMD diagnosis, end of insurance plan enrollment, or the study end date of January 31, 2022 (whichever occurred earliest). The postindex follow-up period was limited to 6 to 18 months in part because the number of patients who received JAKi therapy had already decreased by up to 62% in our study populations at day 500.

Statistical Analyses

The AMD incidence rate ratio (IRR) in patients who received JAKi therapy vs those who received non-JAKi therapy was estimated using bayesian Poisson regression (eTable 3 in Supplement 1), adjusting for age group and sex, and offset by the logarithm of time at risk. Noninformative, normally distributed priors were selected for the regression coefficients. The regression coefficient of JAKi (βJAKi) in the Poisson regression model represented the logarithm of IRR; thus, the IRR is estimated by the posterior mean of exponential(βJAKi). Additionally, the exact posterior distribution of IRR was estimated from posterior sampling draws and used to calculate highest posterior density intervals, which are free from any distributional assumptions. IRR was considered to be statistically significant if the 95% highest posterior density interval of exp(βJAKi) did not include 1. An observed estimate of reduction in AMD risk was deemed the most likely value in our analyses comparing patients who received JAKi therapy and those who received non-JAKI therapy (reference group) if the null value (IRR = 1.0)was not contained within the 95% CI. Additionally, the posterior probability of IRR less than 1 was reported; a probability value close to 1 (0.97-1.00) was considered as evidence of a lower risk of AMD for the JAKi-exposed group vs the non–JAKi-exposed group.

Propensity score matching and statistical analyses were conducted using SAS version 9.04, release 3.82 (SAS Institute). A 2-sided P value ≤.05 was considered statistically significant.

Results

Study Population

Initial screening identified 6 673 213 individuals in MarketScan and 3 097 904 in Optum with diagnoses of the autoimmune diseases of interest between January 1, 2010, and June 30, 2021. Of the total included, 390 091 patients in MarketScan (11 109 patients who received JAKi therapy and 378 982 who received non-JAKi therapy) and 221 152 patients in Optum (6567 who received JAKi therapy and 214 585 who received non-JAKi therapy) satisfied the study eligibility criteria (Figure 1).

Figure 1. Cohort Selection.

AMD indicates age-related macular degeneration; AS, ankylosing spondylitis; CD, Crohn disease; DMARD, disease-modifying antirheumatic drug; HS, hidradenitis suppurativa; IBD, inflammatory bowel disease; Ig, immunoglobulin; IL, interleukin; JAKi, Janus kinase inhibitor; PsA, psoriatic arthritis; RA, rheumatoid arthritis; TNF, tumor necrosis factor; UC, ulcerative colitis.

Comparison of baseline characteristics of patients in the JAKi and non-JAKi groups indicated that both groups were predominantly female (MarketScan: 8653/11 109 [77.9%] vs 249 621/378 982 [65.9%], respectively; Optum: 5000/6567 [76.1%] vs 139 852/214 585 [65.2%], respectively) and generally older than 55 years (eTable 4 in Supplement 1). Information on race and ethnicity (limited to the Optum database) indicated that 6852/221 152 patients (3.1%) were Asian, 21 333/221 152 (9.7%) were Black, 21 167/221 152 (10.5%) were Hispanic, and 159 746/221 152 (72.2%) were White. Of the autoimmune diseases of interest, JAKi therapy appeared to be directed preferentially toward rheumatoid arthritis, which accounted for 68.0% of all first autoimmune disease diagnoses among patients exposed to JAKis (MarketScan, 8024/11 109 [72.2%]; Optum, 3995/6567 [60.8%]) compared with 41.0% among patients who were not exposed (MarketScan, 164 947/378 982 [43.5%]; Optum, 78 227/214 585 [36.5%]) (eTable 4 in Supplement 1). The mean (SD) time from first autoimmune disease diagnosis to treatment initiation was considerably longer for JAKi users (MarketScan, 1549 [973] days; Optum, 1545 [1058] days) than for nonusers (MarketScan, 285 [567] days; Optum, 400 [688] days), suggesting that JAKi use is mainly reserved for second-line therapy in these indications. In both the MarketScan and Optum populations, cataract (28 015/390 091 [7.2%] and 25 786/221 152 [11.7%], respectively) and ocular surface disease (31 480/390 091 [8.1%] and 25 370/221 152 [11.5%], respectively) were the most common ocular comorbidities, while hypertension (137 417/390 091 [35.2%] and 109 174/221 152 [49.4%], respectively) and dyslipidemia (112 252/390 091 [28.8%] and 94 276/221 152 [42.6%], respectively) were the most common nonocular comorbidities (eTable 4 in Supplement 1).

Propensity Score–Matched Cohorts

Among the eligible study population, a total of 9126 patients who received JAKi therapy from the MarketScan dataset were matched with 9126 who did not receive JAKi therapy with respect to age, sex, region, first autoimmune disease diagnosis, time from first autoimmune disease diagnosis to index date, CCI category, dyslipidemia, and hypertension. A total of 5667 patients who received JAKi therapy from the Optum dataset were successfully matched to an identical number of patients who did not receive JAKi therapy. Some patients could not be matched, hence the difference between patients who were analyzed (9126 in MarketScan and 5667 in Optum) and those who were exposed to JAKis (11 109 in MarketScan and 6567 in Optum). Postmatch SMDs across JAKi users and nonusers were less than 0.1 for all matched variables, indicating good balance of covariates (Table 1).

Table 1. Standardized Mean Differences (SMDs) in Demographic and Clinical Covariates Before and After Propensity Score Matching.

Variable	MarketScan		Optuma,b
	Prematch SMD	Postmatch SMD	Prematch SMD	Postmatch SMD
Time from AID to index date	1.59	0.02	1.28	0.01
Categorical ageb	0.17	0.0	0.24	0
Sex	0.27	−0.01	0.24	−0.02
Race and ethnicityc	NA	NA	0.05	0
Region	0.20	0.05	0.12	0.05
Ankylosing spondylitisb	−0.06	0.0	−0.02	0
Atopic dermatitisb	−0.24	0.0	−0.21	0
Crohn diseaseb	−0.30	0.0	−0.28	0
Hidradenitis suppurativab	−0.07	0.0	−0.05	0
Polyarticular juvenile idiopathic arthritisb	0.05	0.0	0.07	0
Psoriasisb	−0.42	0.0	−0.33	0
Psoriatic arthritisb	0.03	0.0	0.12	0
Rheumatoid arthritisb	0.61	0.0	0.52	0
Ulcerative colitisb	−0.09	0.0	−0.14	0
Categorical CCIb	0.64	0.0	0.45	0
Baseline dyslipidemia	0.07	−0.04	−0.04	−0.01
Baseline hypertension	0.07	−0.04	−0.03	−0.02

Abbreviations: AID, autoimmune disease; CCI, Charlson Comorbidity Index; NA, not available.

^a Patients of unknown sex and unknown region were excluded from matching.

^b Exact match required.

^c Race and ethnicity data were available in the Optum database only; reported here as recorded in Optum and included because it is considered an important covariate.

Demographic and baseline clinical characteristics of the propensity score–matched cohorts from each database are summarized in Table 2. Postindex follow-up times were similar for patients who received JAKis and those who did not in the matched MarketScan cohorts (mean [SD], 2.6 [1.9] years and 2.8 [2.0] years, respectively) and Optum cohorts (mean [SD], 2.9 [2.1] years and 3.2 [2.4] years, respectively). The proportion of patients aged 65 years and older was higher in Optum than Marketscan (Optum, 34.1%; MarketScan, 16.7%), as well as the mean CCI (Optum, 1.9-2.0; MarketScan, 1.5-1.6) and the prevalence of atopic dermatitis (Optum, 6.3%; MarketScan, 3.2%), psoriasis (Optum, 12.8%; MarketScan, 8.5%), and rheumatoid arthritis (Optum, 57.9%; MarketScan, 67.2%). Ocular comorbidities were well balanced (SMDs all <0.1) between the matched cohorts of individuals who received and did not receive JAKis, with cataract (MarketScan, 691/9126 [7.6%] vs 805/9126 [8.8%]; Optum, 608/5667 [10.7%] vs 651/5667 [11.5%], respectively) and ocular surface disease (MarketScan, 824/9126 [9.0%] vs 1049/9126 [11.5%]; Optum, 662/5667 [11.7%] vs 769/5667 [13.6%], respectively) being the most frequent.

Table 2. Baseline Demographic and Clinical Characteristics of Patients Exposed to and Not Exposed to Janus Kinase Inhibitor (JAKi) Therapy After Propensity Score Matching.

Variable	Patients in MarketScan, No. (%)		Patients in Optum, No. (%)
	Exposed to JAKis (n = 9126)	Not exposed to JAKis (n = 9126)	Exposed to JAKis (n = 5667)	Not exposed to JAKis (n = 5667)
Age, y
41-54	3553 (38.9)	3553 (38.9)	1682 (29.7)	1682 (29.7)
55-64	4049 (44.4)	4049 (44.4)	1989 (35.1)	1989 (35.1)
65-74	1052 (11.5)	1052 (11.5)	1331 (23.5)	1331 (23.5)
≥75	472 (5.2)	472 (5.2)	493 (10.6)	493 (10.6)
Sex
Female	6988 (76.6)	7031 (77.0)	4259 (75.2)	4304 (75.9)
Male	2138 (23.4)	2095 (23.0)	1408 (24.9)	1363 (24.1)
Race and ethnicitya
Asian	NA	NA	125 (2.2)	125 (2.2)
Black	NA	NA	620 (10.9)	620 (10.9)
Hispanic	NA	NA	686 (12.1)	686 (12.1)
White	NA	NA	4236 (74.8)	4236 (74.8)
Region
Northeast	1436 (16.7)	1488 (16.3)	529 (9.3)	579 (10.2)
Midwest	1345 (14.7)	1272 (13.9)	1209 (21.3)	1139 (20.1)
South	3646 (39.9)	3739 (41.0)	2826 (49.9)	2868 (50.6)
West	1129 (12.4)	1026 (11.2)	1103 (19.5)	1081 (19.1)
Unknown	1570 (17.2)	1601 (17.5)	0 (0)	0 (0)
Metropolitan
Rural	2208 (24.2)	2088 (22.9)	NA	NA
Urban	6918 (75.8)	7038 (77.1)	NA	NA
First autoimmune diagnosis
Ankylosing spondylitis	256 (2.8)	256 (2.8)	203 (3.6)	203 (3.6)
Atopic dermatitis	291 (3.2)	291 (3.2)	354 (6.3)	354 (6.3)
Crohn disease	158 (1.7)	158 (1.7)	103 (1.8)	103 (1.8)
Hidradenitis suppurativa	46 (0.5)	46 (0.5)	49 (0.9)	49 (0.9)
Polyarticular juvenile idiopathic arthritis	79 (0.9)	79 (0.9)	63 (1.1)	63 (1.1)
Psoriasis	776 (8.5)	776 (8.5)	725 (12.8)	725 (12.8)
Psoriatic arthritis	755 (98.3)	755 (98.3)	528 (9.3)	528 (9.3)
Rheumatoid arthritis	6130 (67.2)	6130 (67.2)	3280 (57.9)	3280 (57.9)
Ulcerative colitis	635 (7.0)	635 (7.0)	362 (6.4)	362 (6.4)
Time from first AID diagnosis to index date, d
Mean (SD)	1348 (894)	1329 (890)	1430 (1012)	1421 (1016)
Median (IQR) [range]	1166 (655-1878) [0-4088]	1154 (652-1845) [0-4094]	1184 (696-2110) [0-4358]	1194 (607-2119) [0-4333]
CCI
Mean (SD)	1.5 (1.3)	1.6 (1.5)	1.9 (1.8)	2.0 (2.0)
Median (IQR) [range]	1 (1-2) [0-12]	1 (1-2) [0-14]	1 (1-2) [0-18]	1 (1-2) [0-19]
CCI class
0	1216 (12.3)	1216 (12.3)	789 (13.9)	789 (13.9)
1	4778 (52.4)	4778 (52.4)	2415 (42.6)	2415 (42.6)
2	1716 (18.8)	1716 (18.8)	1048 (18.5)	1048 (18.5)
≥3	1416 (15.5)	1416 (15.5)	1415 (25.0)	1415 (25.0)

Abbreviations: AID, autoimmune disease; CCI, Charlson Comorbidity Index; NA, not available.

^a Race and ethnicity data were available in the Optum database only; reported here as recorded in Optum and included because it is considered an important covariate.

Incidence of AMD

For the MarketScan cohorts, the AMD incidence rate over the first 6 to 18 months postindex was lower among patients who received JAKis (1.96 per 1000 patient-years) than in those who did not (3.58 per 1000 patient-years), with a relative reduction of 49% among patients who received JAKis (10/9126 events; adjusted IRR [AIRR], 0.51; 95% CI, 0.19-0.90) vs those who did not (43/9126 events; AIRR, 1 [reference]). Similarly, among the Optum cohorts, JAKi-exposed patients had a lower AMD incidence rate (0.86 per 1000 patient-years) than patients exposed to non-JAKis (2.75 per 1000 patient-years), with a 73% relative risk reduction among patients who received JAKi therapy (n = 3/5667 events; AIRR, 0.27; 95% CI, 0.03-0.74) vs (21/5667 events; AIRR, 1 [reference]) in those who did not receive JAKi therapy. Findings were similar when index year was added as covariate in the adjusted model (sensitivity analysis). The absolute percentage reductions were 0.36% (MarketScan) and 0.32% (Optum), favoring patients who received JAKis.

Cumulative incidences of AMD over time among patients who received JAKis and those who did not are shown in Figure 2 (MarketScan) and Figure 3 (Optum). Bayesian modeling indicated that the probability of avoiding AMD onset (relative risk <1) in patients who received JAKi therapy was 97.59% (MarketScan) and 98.94% (Optum).

Figure 2. Cumulative Incidences of Age-Related Macular Degeneration Over the First 6 to 18 Months Postindex Among Patients in MarketScan Receiving Either Janus Kinase Inhibitor (JAKi) Therapy or Other Immunotherapy for Autoimmune Disease.

Figure 3. Cumulative Incidences of Age-Related Macular Degeneration Over the First 6 to 18 Months Postindex Among Patients in Optum Receiving Either Janus Kinase Inhibitor (JAKi) Therapy or Other Immunotherapy for Autoimmune Disease.

Discussion

This retrospective cohort study, based on 2 administrative claims databases representing approximately 9.8 million US patients with autoimmune disease, provides comprehensive, up-to-date information on the relative risk of incident AMD associated with JAKi use vs other immunotherapies for this indication. Findings from the MarketScan and Optum populations indicated that JAKi use (over other immunotherapies) in individuals older than 40 years with autoimmune diseases was associated with a markedly reduced risk of incident AMD over the first 6 to 18 months of treatment. Substantially lower incidence rates of AMD were observed among individuals who received JAKi therapy vs those who did not in the MarketScan and Optum populations. Moreover, the absolute percentage reduction is consistent with a potential role for JAKi in this population of patients with autoimmune disease. A sensitivity analysis conducted in the patient subpopulation aged 50 years and older (the commonly accepted minimum age threshold for AMD development) yielded similar results. The validity of these findings is reinforced by the study’s use of propensity-matching techniques to adjust for residual (ie, nonautoimmune) disease differences and by replication in 2 separate database populations.

The reason for the difference in the extent of risk reduction between the MarketScan (49%) and Optum (73%) study populations is unclear. Within each database population, patients exposed to and not exposed to JAKis were closely matched for demographic and clinical characteristics and follow-up duration. Nevertheless, the 2 database populations, drawn from different health care organizations and insurance programs, differed in several respects after propensity score matching, including the proportion of patients 65 years and older , mean CCI, and prevalence of atopic dermatitis, psoriasis, and rheumatoid arthritis. The more pronounced reduction in AMD risk seen in the Optum database population may potentially be related to this being an older age group with greater comorbidity.

Involvement of local inflammation in AMD pathogenesis is widely accepted^6,29,30; however, emerging evidence suggests that systemic inflammation may also contribute to this process. Although the retina is an immune-privileged tissue protected by the blood-retina barrier and the local immune regulatory system, the blood-retina barrier does not completely prevent entry of circulating cells.^31,32 Features suggestive of systemic inflammatory involvement in the development of AMD include hyperactivation of the JAK2/STAT3 and NF-kβ signaling pathways^29,33 and elevation of serum levels of C-reactive protein and various inflammatory cytokines, including IL-1β, IL-4, IL-8, IL-17, and tumor necrosis factor α.^34,35,36 Investigation of the association between JAKi use and AMD incidence may provide insight into the complex immunological and inflammatory processes at play in its pathophysiology and lead to improved treatment options for AMD. However, due to systemic drug administration, the present study is unable to differentiate between the potential systemic vs local (retinal) anti-inflammatory effects of JAKi therapy in reducing the risk of AMD.

Strengths and Limitations

Strengths of the present study lie in its large sample size of more than 14 000 patients, its analysis of 2 distinct population-based data sources, and its use of propensity score–matching techniques to adjust for differences in demographic and clinical features between the study cohorts. Additional study attributes include the selection of a geographically and clinically disparate patient population receiving treatment under different reimbursement scenarios. Although the proportion of individuals who received JAKis represented approximately 3% of the entire patient cohort, it was larger than the required sample size for a given success probability in a logistic regression model (as previously defined).^37,38 However, as a retrospective, nonrandomized, observational study, the analysis has several limitations. The databases indicated whether patients received an AMD diagnosis-related bill, but their electronic health records and clinical information were not available. Therefore, the clinical setting in which AMD was diagnosed could not be established, representing a potential confounder, although access to ophthalmology services and the likelihood of receiving an AMD diagnosis should not have differed between treatment groups. Clinical characteristics were determined using ICD diagnosis codes on nondiagnostic claims, which introduces the potential for coding errors. Despite good balance in baseline characteristics between the matched patient cohorts, the possibility of bias due to imbalance in nonidentified covariates cannot be excluded. For example, the underlying autoimmune disease could potentially contribute to the pathogenesis of the disease (eg, in the prematched population, individuals who received JAKi therapy were more likely to have rheumatoid arthritis as first diagnosis) or could affect the chances of having access to an ophthalmologist and receiving a diagnosis of AMD (ie, surveillance bias, wherein incidence and risk trends are affected by differences in screening frequency between groups). The closely matched comorbidity rates suggested by ICD-9-CM and ICD-10-CM diagnostic coding may nevertheless conceal clinically important differences in disease severity between the cohorts. Limitations concerning the recording and identification of AMD data should also be considered. Furthermore, the study data constitute a sample of convenience drawn from patients with autoimmune diseases and continuous enrollment in the MarketScan and Optum databases, and the patients and practice patterns in the constituent health care organizations may not be fully representative at the national level. It is also noteworthy that we were unable to match all patients exposed to JAKis, which contributed to attrition; use of JAKi therapy was considered a binary parameter (not time dependent); unknown confounders may have affected the results; subgroup analyses comparing outcomes in patients who never received JAKi therapy to various subgroups of patients who received non–JAKi-based treatment were not conducted; and the postindex follow-up was limited to 18 months. Moreover, it is not possible to generalize the study findings to populations who do not have inflammatory or autoimmune diseases.

Conclusions

In conclusion, our study provides evidence of an association between JAKi therapy and a reduced risk of developing AMD in patients with autoimmune diseases. Further research is required to determine the robustness and duration of this effect in larger and/or more representative patient populations and its applicability to other diseases or indications. Importantly, the mechanisms underlying the prophylactic action of JAKi agents against AMD, and how they differ from those of non-JAKi immunotherapies, remain to be determined. However, our findings suggest that reducing systemic inflammation by inhibiting JAK-STAT signaling represents an additional therapeutic option for AMD management in clinical practice.

Supplement 1.

eTable 1. International Classification of Diseases, Clinical Modification, Diagnosis Codes

eTable 2. NDC Billing Codes, CPT/HCPCS Procedure and Supply Codes, and ICD-10-PCS Procedure Codes

eTable 3. Demographic and Baseline Clinical Characteristics Prior to Propensity Score Matching

eTable 4. Outcome model

Supplement 2.

Data sharing statement