The Broader Economic Value of Treatment for Diabetic Macular Edema

Diabetic macular edema (DME) is a complication of diabetic retinopathy (DR) in which retina vascular hyperpermeability results in the pathologic accumulation of fluid in the macula (1). Patients with DME will experience loss of central visual acuity (VA) or metamorphopsia, particularly if the DME involves the center of the macula. The natural history can be variable, but the Early Treatment of Diabetic Retinopathy Study (ETDRS) identified a 32% risk of moderate vision loss over 3 years if untreated (2). DME is the most common cause of vision loss in people with DR, and approximately half of people with DR may develop DME (3). More than 1.1 million people, or ∼3.8% of adults with diabetes aged ≥45 years, have DME (4,5). Risk factors for DME include diabetes duration, poor glycemic control, and hypertension (4,6).

The retinal vascular changes responsible for DME are not well understood. Fortunately, effective treatments have been developed, including anti–vascular endothelial growth factor (VEGF) injections, corticosteroid injections, and/or laser therapy. The clinical benefits of DME treatments are well established in large, pivotal clinical trials (7), and they are all cost effective relative to no treatment (8–11). Although laser therapy had long been the primary treatment option for DME, the advent of intravitreal therapy in the mid-2000s changed DME management paradigms. Intravitreal anti-VEGF therapies, which include ranibizumab, aflibercept, and off-label bevacizumab, have become the first-line treatment for DME, with intravitreal corticosteroids and laser therapy usually reserved as adjunct options if needed (12). The share of patients with DME who received anti-VEGFs increased from 5% in 2009 to 27.1% in 2014 (13). More recent studies estimated that 35–48% of Medicare beneficiaries with DME received anti-VEGFs in 2018 and 68% of patients with DME received anti-VEGFs in 2020 (12,14).

Understanding the economic value of DME treatment to patients and society is important, particularly in light of increasing diabetes prevalence in the U.S. (15). Previous economic studies have focused on utilization and costs of DME. For example, people with DME have 11.7 more health care visit days and 3.6 additional eyecare visits compared with those with diabetes alone (16). The estimated annual direct medical cost of DME (incremental to diabetes, analysis samples included people with and without complications, values shown in 2022 US dollars) ranges from $4,763 for Medicare beneficiaries and $3,597–$17,967 for commercially insured individuals (17–20). Of note, incremental annual medical costs increase to $36,163 among a visually impaired DME population (20). Finally, people with DME also have higher rates of absenteeism and workers’ compensation and disability claims (21).

While existing research has begun to explore the costs associated with DME and the cost effectiveness of specific DME therapies, no study to our knowledge has estimated the broader economic benefit associated with spending on DME treatment. To quantify the economic net value (benefit − cost) of DME treatment, we first simulated the direct and indirect benefits associated with treatment for individuals. We then used the individual-level estimates to calculate the aggregate value of improved vision and economic outcomes based on current U.S. treatment patterns.

Our study paired individual and cohort analyses to demonstrate the value of treatment for DME. First, we modeled the effect of DME treatment on individual outcomes using the Future Elderly Model (FEM). The FEM uses baseline SRV along with other demographic and health variables as inputs and predicts outcomes such as SRV, life expectancy, quality-adjusted life-years (QALYs), employment, and disability. The individual simulation does not model treatment patterns or treatment costs. To quantify the relative magnitude of costs and benefits of DME treatment, we used a cohort-level analysis. To calculate cohort-level benefits (direct and indirect), we scaled up the individual FEM estimates based on a real-world treatment mix. Treatment costs associated with this mix were derived using published real-world data on treatment frequency (12). The aggregate net value of treatment for DME in the U.S. is measured as total benefit minus treatment costs.

Our study used data from four sources. The individual simulation relied on data from the Health and Retirement Study (HRS). The treatment scenarios implemented in the simulation were derived using the National Health and Nutrition Examination Survey (NHANES) and clinical trial data from the Diabetic Retinopathy Clinical Research Network (DRCR.net). Finally, our cohort analysis incorporated current treatment patterns from the literature, which were derived from the Vestrum Health Retina database (Vestrum Health, Naperville, IL) (12).

The FEM relies primarily on the HRS, a nationally representative longitudinal panel of ∼20,000 Americans aged ≥51 years (22). The FEM uses demographic variables from the HRS, including age, sex, race/ethnicity, education, employment status, earnings, and disability claims, as well as health indicators, such as smoking status and diagnoses for heart disease, stroke, cancer, hypertension, diabetes, and lung disease. Our FEM simulation was based on HRS data from 2008 to 2014 (HRS waves 9–12).

The DME treatment scenarios modeled in the FEM were derived from publicly available DRCR.net clinical trial data (23). The DRCR.net was formed in 2002 through a National Eye Institute and National Institute of Diabetes and Digestive and Kidney Diseases–sponsored cooperative agreement. The objective was to develop a collaborative network to facilitate multicenter clinical trial research on DR, DME, and associated conditions. We used clinical trial data from Protocol A (A Pilot Study of Laser Photocoagulation for Diabetic Macular Edema), Protocol B (A Randomized Trial Comparing Intravitreal Triamcinolone Acetonide and Laser Photocoagulation for Diabetic Macular Edema), and Protocol T (A Comparative Effectiveness Study of Intravitreal Aflibercept, Bevacizumab, and Ranibizumab for Diabetic Macular Edema) to derive the laser, steroid, and anti-VEGF/untreated scenarios, respectively.

Because clinical trial data in ophthalmology report outcomes in terms of VA, we required a mapping from VA to SRV. While several nationally representative data sets have information on SRV, NHANES is the only one with information on both VA and SRV, making it ideal to create the VA-to-SRV mapping. NHANES is a nationally representative survey that combines interviews and physical examinations and is conducted every 2 years (24). Our mapping relied on data from the two most recent waves of NHANES with VA and SRV data (2005–2006 and 2007–2008).

Treatment patterns for our cohort analysis were derived from an analysis of individuals in the Vestrum Health Retina database who were newly diagnosed with DME between 2015 and 2020 (12). Vestrum is an aggregated longitudinal database of electronic medical records from a demographically and geographically diverse U.S. patient sample. It contains visit data from ∼1.5 million unique patients with >11 million encounters obtained from a panel of 350 retina specialists.

The FEM is a dynamic economic microsimulation model that has been used to study health care innovation in numerous contexts, including cancer and diabetes (25–27). The FEM uses data from the HRS to estimate transitions between health states and assess outcomes as a function of individual characteristics. These estimated transition probabilities are entered into the model to simulate how health and economic outcomes evolve over time. Changing FEM starting conditions and inputs allowed us to simulate how outcomes would differ under hypothetical scenarios, such as the development of a new medical treatment, and assess the value for each scenario.

For this study, we expanded the FEM to incorporate measures of vision from the HRS. Specifically, HRS respondents’ SRV was assessed during each study wave using the following question: “Rate your eyesight while wearing glasses or corrective lenses as usual.” Respondents reported vision on a five-point Likert scale as excellent, very good, good, fair, and poor; a small subset of respondents (<1%) provided a voluntary response of blind. We standardized SRV across the HRS and NHANES, which was used to create our VA-to-SRV mapping, by comparing the overall response distributions and collapsing SRV into four categories: excellent, good, fair, or poor (Supplementary Table 1).

We used an ordered probit regression to model SRV as a function of SRV from 2 years prior, sex, age, education, race/ethnicity, years since diabetes diagnosis, and indicators for select comorbidities, such as hypertension. We allowed for differences in the evolution of SRV over time associated with cataracts and cataract surgery, which are recorded for people aged >64 years in the HRS, by stratifying the sample and estimating separate models for people aged <65 and people ≥65 years with and without surgery (see Supplementary Table 2 for estimated model coefficients). To measure the impact of vision changes on nonvision outcomes, we included SRV from 2 years prior as a covariate in the transition models for mortality, quality of life (measured using the Health Utilities Index Mark 3), work, disability, and disability benefits (Supplementary Table 3). Additional FEM documentation and model validation results are presented in the Supplementary Appendix, Supplementary Tables 4–6, and Supplementary Figs. 2 and 3.

We relied on data from DRCR Protocols A, B, and T to derive the vision changes associated with laser monotherapy, steroid monotherapy, and anti-VEGF therapy, respectively. All three trials enrolled individuals aged ≥18 years with diagnosed diabetes, had at least one eye that met the study criteria, and had not been previously treated for DME with the intervention under consideration. All three studies reported 12- and 24-month outcomes.

DRCR Protocol A was the first study initiated by the DRCR.net and was a randomized clinical trial conducted between 2003 and 2008 (23). All participants (N = 263) were treated with laser photocoagulation by either the modified ETDRS technique or the mild macular grid technique. DRCR Protocol B was a randomized clinical trial conducted between 2004 and 2008. Participants (N = 693) were randomly assigned to laser therapy or steroid therapy (intravitreal triamcinolone [1 mg or 4 mg]) (23). DRCR Protocol T was a head-to-head comparison of anti-VEGF therapies that took place between 2012 and 2018 (23). Participants (N = 660) were randomly assigned to one of three treatment arms: aflibercept (2 mg), bevacizumab (1.25 mg), or ranibizumab (0.3 mg).

To derive the treatment effects for our model scenarios, we analyzed the baseline and 2-year VA data for individuals aged ≥51 years, which corresponds to the age range modeled in the FEM. The laser scenario pooled both trial arms from Protocol A, the steroid treatment scenario pooled both steroid arms from Protocol B, and the anti-VEGF scenario pooled all arms from Protocol T. For each scenario, we pooled data for people who received treatment for DME in either one or both eyes. In the sample used to construct the laser, steroid, and anti-VEGF scenarios, 23%, 29%, and 32% of people were being treated in both eyes, respectively.

Because the FEM requires vision measured in terms of SRV as an input, we converted the clinical trial VA data to SRV using a mapping derived from NHANES data. We limited the NHANES analysis sample to people with diabetes aged >50 years (n = 882) to match the FEM model age. People with excellent or good SRV were less likely to be visually impaired (defined as having worse than 20/40 vision in the better-seeing eye). Among those with excellent or good SRV, 4% and 10% were visually impaired compared with 22% and 44% of people with fair or poor SRV.

The VA-to-SRV mapping was created by estimating an ordered probit with SRV as the dependent variable. The model included eight categories for VA (20/25, 20/30, 20/40, 20/50, 20/60, 20/80, 20/200, and worse than 20/200) in the better- and worse-seeing eye, where 20/20 was the reference category. The model also controlled for demographic variables that were available in the DRCR clinical trials, NHANES, and HRS data sets: sex, 5-year age categories, race/ethnicity, and time since diabetes diagnosis. The estimated coefficients from the model are provided in the Supplementary Appendix and Supplementary Tables 14–16, along with additional validation results (Supplementary Tables 7–12). For the NHANES estimation sample with visual impairment (n = 131), the VA-to-SRV mapping predicted the following SRV distribution: 9% excellent, 31% good, 25% fair, and 35% poor.

We limited our FEM simulation sample to people with diabetes and matched their baseline vision and demographic variables to DRCR Protocol T. Individuals were simulated under four scenarios: untreated, anti-VEGF, laser, and steroid. The derivation of the three treatment scenarios was described in the previous section. The untreated scenario was derived by applying a model based on vision changes in an untreated DME population (28) to the Protocol T baseline population (see Supplementary Appendix for additional detail and Supplementary Tables 17 and 18) (23).

We did not run separate simulations for people with DME in one eye versus two eyes because the posttreatment SRV distribution for these groups is similar within each treatment scenario. Because we lacked data on long-run persistence of treatment effects (durability), we assumed that the vision benefits from treatment are permanent but allowed for SRV to decline naturally with age. Table 1 summarizes the treatment effects.

We simulated the following outcomes for each scenario: SRV, quality of life, employment, earnings, whether an individual claimed disability benefits, disability benefits (U.S. dollars), activities of daily living, independent activities of daily living, and mortality. We calculated disability-free life-years as the total years in which a person had no difficulty with any activities of daily living or independent activities of daily living and did not live in a nursing home. The value of a QALY was calculated as the total QALYs multiplied by $150,000 (29,30). Future dollar values were discounted at 3% per year. Individual outcomes were calculated for a lifetime horizon.

We derived our starting cohort size (N = 1,108,500) by applying DME prevalence to the most recent estimate of people with diabetes in the U.S. aged >45 years (the nearest age cutoff for which data were available that matched our model age of >50) (4,5). A published analysis of 11,042 eyes in the 2020 Vestrum database informed treatment patterns (both treatment mix and anti-VEGF injection frequency) for the cohort model (12). We compared benefits and costs in a cohort assuming 2020 real-world treatment patterns (68% anti-VEGF, 3% laser, 1% steroid, and 28% untreated) (12) with benefits in a cohort where 100% remained untreated.

The cost of treatment included drug and clinical treatment costs. The costs of office visits ($120), laser treatment ($532), and intravitreal injection ($107.72) were derived from the 2022 Physician Fee Schedule (31). We assumed a per-injection anti-VEGF drug cost of $865, which represents the weighted average of ranibizumab (19%; $1,841.50), aflibercept (36%; $921.60), and bevacizumab (45%; $67.86). The weighted steroid drug cost was $1,133, which included triamcinolone acetonide (33%; $17.63), dexamethasone intravitreal implant (64%; $1,400), and fluocinolone acetonide intravitreal implant (3%; $7,500). Treatment weights for anti-VEGFs and steroids were derived from the published 2020 Vestrum analysis (12).

To model estimated costs based on anti-VEGF utilization patterns, we created real-world and Protocol T frequency scenarios. The real-world frequency scenario was derived from the 2020 Vestrum data (12) and assumed that patients received four anti-VEGF injections in year 1 and three injections in subsequent years. The Protocol T frequency scenario corresponded to clinical trial treatment frequencies in years 1 and 2 (10 and 6 anti-VEGF injections, respectively) followed by 1 injection in all subsequent years per the Protocol T 5-year extension analysis (23,32). In both scenarios, laser monotherapy was modeled to be given three times in the 1st year, with no treatment in subsequent years, and steroid injection frequency varied by drug type, consistent with DRCR clinical trials. We assumed that untreated patients had four office visits annually.

Individual results from the microsimulation were aggregated according to the treatment weights (i.e., 2020 real-world treatment patterns or 100% untreated) to calculate benefits for a nationally representative DME cohort. Direct benefits were calculated as QALYs multiplied by $150,000 (29,30). Indirect benefits included earnings and government savings on disability payments.

We estimated several outcomes for the real-world and Protocol T scenarios: total treated, total direct benefit, total indirect benefit, and total treatment costs. Net value estimates were calculated as the difference between total benefits and total treatment costs. All future dollar values were discounted at a rate of 3% per year. We present outcomes for 2-, 5-, 10-, and 20-year time horizons.

Because we lacked long-run data on durability, we ran simulations that assumed that treatment benefits declined at 2, 6, and 12 years following treatment initiation. After that point, patients with DME experienced the same SRV trajectory as an untreated patient. We also conducted sensitivity analysis for the assumed value for QALYs ($50,000, $100,000, and $300,000) and the cost of anti-VEGF drugs ($432.49 and $1,297.47, corresponding to ±50% of the main parameter value).

All data sets used for this study and our model validation are publicly available (DRCR clinical trials, NHANES, HRS, National Health Interview Survey). The source of the DRCR clinical trial data is the DRCR Retina Network (unique Federal Award Identification Number UG1EY014231), but the analyses, content, and conclusions presented herein are solely the responsibility of the authors and have not been reviewed or approved by the DRCR Retina Network.

Table 2 presents the simulated lifetime outcomes for a hypothetical 51-year-old. A 51-year-old with DME who does not receive treatment has a life expectancy of 79 years (i.e., can expect to live an additional 28 years). An untreated individual will spend 15.5 years with excellent or good SRV and 12.5 years with fair or poor SRV. However, a person with DME who receives treatment will experience an increase in time with excellent or good SRV by ∼35% for the anti-VEGFs and ∼30% for laser or steroid scenarios. Treatment with anti-VEGFs increases QALY and disability-free life-years by 11% and 13%, respectively. Both laser and steroid treatment for DME increase QALY and disability-free life-years by 8% and 9%, respectively.

Even though our simulation cohort is relatively close to retirement age, treatment for DME improves labor market outcomes. A person who receives DME treatment works an additional 6–9% years compared with an untreated individual, as well as increases in their total earnings by 12–19%. Moreover, they will generate 10–17% in savings on government disability payments compared with an untreated individual. Anti-VEGFs are associated with the greatest benefit.

Figure 1 shows the cumulative direct and indirect patient benefits associated with current treatment patterns for a U.S. DME cohort. The total benefit increases from $5.6 billion 2 years after treatment initiation to $67.8 billion 20 years after treatment initiation. Indirect benefits, which include earnings and government savings from disability payments, are $1.0 billion and $4.8 billion in years 2 and 20 after treatment initiation, respectively.

We calculated net value (Table 3) assuming permanent durability and two injection frequency scenarios. The net value is −$5.8 billion and $1.0 billion in year 2 by Protocol T and real-world injection frequencies, respectively. In year 20, net value ranges from $52.8 billion (Protocol T) to $47.8 billion (real world).

Even with the most conservative assumption about durability (i.e., treatment benefits last 2 years), net value estimates are positive for all time horizons assuming real-world injection frequency and positive in years 5, 10, and 20 assuming Protocol T injection frequency. If we increase the weighted drug cost of anti-VEGFs by 50%, net value ranges from $38.8 billion to $45.7 billion over 20 years. Conversely, if we reduce the weighted cost of anti-VEGFs by 50%, net value ranges from $56.9 billion to $59.9 billion over 20 years. Full results for sensitivity analyses are presented in Supplementary Tables 22–27.

We simulated individual vision outcomes and patient benefits associated with treatment for DME. These benefits were aggregated to generate the total benefits for a DME cohort in the U.S., for which we also estimated costs and net value associated with current treatment patterns. Our findings indicate that treatment for DME generates substantial direct and indirect benefits for individuals. For a representative 51-year-old being treated for DME, anti-VEGFs provide the largest benefit, which mirrors findings from clinical trials (7). The difference in the posttreatment shift in the SRV distribution between the anti-VEGF and steroid/laser scenarios is the primary mechanism for the difference in economic outcomes. In particular, anti-VEGF treatment shifts 8.2% of the people with fair/poor baseline vision to excellent/good posttreatment vision compared with only 4.1% and 2.7% for laser and steroid treatments, respectively.

Because they were based on clinical trial data, the vision gains from our anti-VEGF model scenario are associated with a higher injection frequency (i.e., monthly) than typically used in real-world practice. Uninsured patients in particular may not be able to afford more frequent treatment visits or the associated anti-VEGF drug costs. Furthermore, while the rate of uninsured patients has fallen since the passage of the Affordable Care Act, an estimated 1.65 million people with diabetes were uninsured in 2016 (33). Our simulation results suggest that patients who opt for alternative treatment strategies that require fewer visits or may be less costly (e.g., laser) would still incur meaningful vision and economic benefits compared with remaining untreated.

Our cohort analysis shows that DME treatment generates substantial net value based on real-world treatment patterns. Under Protocol T–based assumptions for treatment frequencies, the net value is negative in year 2, reflecting the high cost of treatment from higher injection frequency associated with clinical trial design (10 and 6 injections in years 1 and 2, respectively) versus real-world utilization (4 and 3 injections in years 1 and 2, respectively), in which positive net value is seen at year 2. Short-run costs are primarily driven by anti-VEGF drug cost (in particular ranibizumab and aflibercept). While bevacizumab provides a less expensive option among the three anti-VEGFs, aflibercept was associated with slightly better vision gains in clinical trials (34). Nevertheless, under both scenarios, the net value becomes positive over longer time horizons as the extended benefit from treatment exceeds the upfront cost.

A comparison of individual-level benefits to microsimulation estimates in other disease areas provides a sense of the relative magnitude of our estimated benefits. A hypothetical 51-year-old treated with anti-VEGFs will have an additional 1.4 years of life expectancy, 2.1 QALYs, and 2.0 disability-free life-years. In contrast, the elimination of high blood pressure would yield an additional 0.72 years of life, 0.95 QALYs, and 0.47 disability-free life-years (26). Similarly, the elimination of diabetes would yield an additional 1.96 years of life and QALYs and 1.45 disability-free life-years. By comparison, the simulated QALY gains from immunotherapy treatments for non–small cell lung cancer range from 0.3 to 1.5 (35).

A review article indicates that people with diabetes miss between 1.6 and 6.5 more days of work per year compared with people without diabetes (36). Another study found that people with DME miss 0.75–2.71 more days of work than their healthy counterparts (21). In our simulation, someone treated with anti-VEGFs will earn ∼$1,000 more 2 years after treatment initiation. For someone earning the median wage in the U.S., this translates to ∼4.6 fewer days worked annually in the absence of treatment, which aligns with previous research (21, 36). Our cohort-level results are similar to findings from a recent study that modeled the patient benefit and social value from anti-VEGF treatment for wet age-related macular degeneration (21,37). While these models are not directly comparable, their results are similar in magnitude, with societal value from treatment of wet age-related macular degeneration ranging from $0.9 to $3.0 billion over 3 years.

While both direct and indirect benefits from DME treatment are large, the direct benefits dominate. This finding is not surprising for two reasons. First, because younger people work more, we expect that the indirect benefit from DME treatment will be relatively large in the short run and decrease as people retire. Indeed, simulated indirect benefits account for 19% and 7% of total benefit in years 2 and 20, respectively. Modeling a younger cohort would result in greater contribution from indirect benefits. Second, independent of labor market dynamics, vision benefits from treatment accrue over time, which means direct benefits increase at a greater rate than indirect benefits. For example, while cumulative indirect benefit increases by a factor of 4.5 between years 2 and 20, cumulative direct benefit increases by a factor of nearly 14 over the same time frame.

Although we demonstrate that DME treatments provide substantial economic benefits, these benefits could go largely unrealized if people with DME remain undiagnosed. Despite recommendations that people with diabetes should receive regular eye examinations (38), a recent study found that nearly 50% of people with diabetes did not receive an eye examination over the 5-year study period (39). Among people diagnosed with DME during the examination component of NHANES, only 44.7% reported being told by a physician that diabetes had affected their eyes or that they had DR (40). In contrast, our cohort model only includes eyes with a diagnosis of DME and likely underestimates the net value due to systemic underdiagnosis of disease.

This study has several limitations, many of which are inherent to all microsimulation models. Because our simulated vision changes in the treatment scenarios used 2-year clinical trial outcomes, we relied on assumptions to model how vision changed beyond 2 years. Nevertheless, total benefit is robust across a range of durability assumptions. A related issue is the lack of long-run data for treatment frequency and corresponding vision outcomes. While this prevented us from explicitly modeling treatment frequency (and corresponding costs) in the microsimulation, we estimated cohort-level costs and net value for a range of treatment frequencies.

Second, the clinical trial data used for our simulation scenarios might not be reflective of real-world DME cases, although this is a common limitation of modeling studies in general. While the demographic characteristics (age, sex, and baseline vision) were similar across our clinical trial data and the published real-world data, clinical trial treatment frequency is substantially higher (i.e., monthly injections) compared with real-world treatment. Because the FEM does not explicitly model treatment frequency, we had to consider its potential impact indirectly. The sensitivity analyses for treatment durability (2, 6, and 12 years) can be conceptually interpreted as the potential effect of lower treatment frequency on outcomes. For the cohort model, pairing the various durability assumptions and treatment frequency scenarios through sensitivity analyses provided an upper- and lower-bound estimate on the potential net value of treatment.

Third, the microsimulation model relied on HRS data, which only include people aged >50 years. In contrast, ∼14% of DRCR Protocol T participants and 10% of NHANES participants with DME were aged <50 years. Because our simulation cohort was closer to retirement age than the entire DME population, the impact of DME treatment on indirect benefits is likely understated. While the microsimulation had the capability to model people aged 18–50 years, the underlying data set for that group did not contain vision data (self-reported or otherwise). If a nationally representative panel data set with vision outcomes for a younger cohort becomes available, it would be worthwhile to incorporate it into the microsimulation.

Finally, NHANES is the only nationally representative data set with VA information collected in an examination setting. Although NHANES is ongoing, it stopped collecting vision data after 2007–2008. However, we confirmed that the SRV distribution has not changed much over the past decade using nationally representative data from the National Health Interview Survey (see Supplementary Table 12), suggesting that the lack of more recent data should not significantly impact our findings.

In conclusion, we found that treatment for DME provides substantial benefits to patients and society through improved vision, life expectancy, and quality-of-life gains. Furthermore, treatment provides indirect benefit in the form of increased employment, earnings, and reduced disability.

Funding. Partial funding for this project was provided by an unrestricted gift from the American Society of Retina Specialists (ASRS) to the authors affiliated with the University of Southern California.

G.E., P.J.F., J.E.K., and P.H. serve as leadership for ASRS, and J.B. is an employee of ASRS. Other than as indicated, ASRS had no role in the design and conduct of the study; collection, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.

Duality of Interest. K.M., J.K., B.T., and S.S. report unrestricted gifts from ASRS during the conduct of the study. G.E. reports holding stock in Regeneron and receiving research funding from Eyepoint and Opthea. P.J.F. reports grants or contracts from Genentech, Apellis, RegenxBio, Gyroscope, Regeneron, Opthea, Alexion, Alkeus, and Eyepoint; consulting fees from Genentech and Apellis; honoraria from Apellis, participation on the Genentech Faricimab advisory board; stock in Arctic, and slide preparations from Genentech. J.E.K. serves on the board of trustees for the American Society of Ophthalmology and reports consulting fees from Alimera Sciences, Allergan, Clearside Biomedical, Genetech, Outlook Therapeutics, Regeneron, Novartis, and Bausch and Lomb. P.H. reports consulting fees from Allergan, Apellis, Dutch Ophthalmic Research Center (DORC), Eyepoint, and Genentech; honoraria from Alcon, Apellis, DORC, and Genentech; and participation on advisory boards for Alimera, Alcon, Allergan, Apellis, DORC, Eyepoint, and Genentech. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. K.M., J.K., and B.T. performed the statistical analysis. K.M., J.K., B.T., and P.H. contributed to the analysis and interpretation of data. K.M., J.K., and P.H. contributed to the acquisition of data. K.M. and B.T. drafted the manuscript. K.M., B.T., G.E., P.J.F., J.E.K., S.S., and P.H. contributed to the concept and design. J.B. provided administrative, technical, and logistic support. S.S. and P.H. provided supervision. All authors contributed to the critical revision of manuscript for important intellectual content. K.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and accuracy of the data analysis.

Prior Presentation. Parts of this study were presented at the Annual Meeting of the American Society of Retina Specialists, New York, NY, 13–16 July 2022.