Baseline visual acuity gaps in age-related macular degeneration: AREDS smokers and non-smokers Authors: Madeline Pan, Neha Sunkara, Biai Dominique Elmir Digbeu, Praveena K. Gupta Published: 2026 ABSTRACT BackgroundAge-related macular degeneration (AMD) is a leading cause of blindness in developed countries. Smoking is the only modifiable risk factor and is associated with faster AMD progression. While Age-Related Eye Disease Study (AREDS) supplements are recommended for intermediate AMD, their impact on visual acuity over time remains unclear.ObjectiveTo describe baseline disparities and visual acuity trajectories in patients with intermediate dry AMD stratified by AREDS use and smoking status.DesignRetrospective observational study of patients with intermediate dry AMD conducted under IRB protocol 22–0267. Data collected from January 1, 2014 to January 1, 2024.SettingSingle academic medical center.ParticipantsA total of 450 patients (759 eligible eyes) with intermediate dry AMD (ICD code H35.3192) were included. Patients were stratified by smoking status and AREDS use. Participants with fewer than three ophthalmology visits during the 10-year study period were excluded.ExposureAREDS supplementation (AREDS1 or AREDS2) and smoking status (current/former vs. never).Main outcomes and measuresVisual acuity measured in Logarithm of the Minimum Angle of Resolution (LogMAR) at first and final clinic visits. ANOVA, Chi-square tests, and mixed effect regression models were used for analysis. Changes in LogMAR over time were calculated descriptively.ResultsAmong 450 patients (mean age, 82.32 years), 255 were smokers (423 eyes) and 195 were non-smokers (336 eyes). Baseline vision was better in non-smokers (mean LogMAR 0.28; SD 0.28) and smokers (mean LogMAR 0.30; SD 0.33) taking AREDS compared to non-smokers (mean LogMAR 0.48; SD 0.44) and smokers (mean LogMAR 0.45; SD 0.47) not taking AREDS (P < 0.0001). All groups experienced visual acuity decline from baseline with no significant differences in final visual acuity among the groups in a mean follow up period of 3.62 years. In a sub-analysis of greater than 5 years follow up mean years of 6.93 ± 1.55 years), a statistically significant difference in final LogMAR values across groups was observed (p = 0.033). The Smokers & AREDS group demonstrated the best final visual acuity (mean 0.39, SD 0.40), followed by Smokers & No-AREDS (mean 0.44, SD 0.43), Non-smokers & AREDS (mean 0.46, SD 0.47), and Non-smokers & No-AREDS (mean 0.72, SD 0.60).Conclusions and relevanceThese results suggest that less than 5 years AREDS users had larger decline in vision from baseline, protective associations of AREDS supplementation may become more discernible over longer follow-up horizons. Real-world observational findings should be interpreted cautiously and do not contradict AREDS efficacy shown in randomized trials.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12886-026-05035-w. Introduction Age-related macular degeneration (AMD) is the leading cause of blindness in developed countries, particularly among individuals over 60 years old. With approximately 200 million people affected globally and prevalence rising alongside aging populations, AMD represents a growing public health burden [1, 2]. AMD progresses along a continuum from early to intermediate to late stages, with intermediate dry AMD representing a critical intervention window before irreversible vision loss occurs [3]. While a subset of patients advance to wet AMD treatable with anti-VEGF therapy, photodynamic therapy, or emerging gene therapy approaches [4] the majority remain on the dry AMD trajectory, where roughly 10–15% ultimately develop advanced disease [5]. It is at this stage that antioxidant supplementation has demonstrated meaningful benefit: the Age-Related Eye Disease Study (AREDS) established that a formulation of vitamins C, E, beta-carotene, and zinc reduces the risk of progression to advanced AMD by 25% [6]. This benefit is structural; a reduction in progression to choroidal neovascularization or geographic atrophy rather than a direct improvement in visual acuity, which remains a secondary and imperfectly correlated surrogate outcome. The application of AREDS supplementation, however, is complicated by smoking status. Smoking is the most significant modifiable risk factor for AMD among several that have been identified, including older age, hypertension, cardiovascular disease, and family history [7, 8], with current smokers carrying a greater than 6-fold increased risk of neovascular AMD compared to nonsmokers and former smokers retaining more than 3 times the risk [9, 10]. At the same time, beta-carotene a core AREDS component increases lung cancer risk in smokers by approximately 18% [11]. This led to the development of AREDS2, which substitutes lutein and zeaxanthin for beta-carotene, and is now the formulation recommended by the National Eye Institute for patients who smoke [12]. Smoking therefore shapes both AMD risk and the choice of supplementation regimen. What remains poorly understood is whether smoking modifies the effectiveness of AREDS supplementation on visual outcomes. Observational studies are limited by baseline vision differences and selection bias, which may confound real-world assessments [13]. Crucially, smoking cessation alone may not immediately reverse accumulated retinal damage, meaning that smokers and non-smokers on supplementation may follow divergent visual trajectories even under equivalent treatment. The primary objective of this study is descriptive: to characterize baseline visual acuity disparities and longitudinal acuity trajectories among AREDS users stratified by smoking status in intermediate dry AMD. A secondary aim is to explore whether any differential patterns by smoking status emerge over time, with the important caveat that this observational design is not powered to assess causal AREDS effectiveness. This gap leaves clinicians without evidence-based guidance on whether AREDS provides equivalent benefit across smoking strata, and this study aims to contribute to that evidence base while emphasizing characterization of differences rather than causal inference. Methods A retrospective observational study using chart review in EPIC electronic medical record was conducted in compliance with the University of Texas Medical Branch Institutional Review Board protocol #22–0267. Study personnel collected relevant demographic and clinical patient information for patients diagnosed with intermediate dry AMD, identified using ICD-9 and ICD-10 billing codes, including H35.3192. The review period spanned from January 1, 2014, to January 1, 2024. AREDS use was documented in the patient’s chart. For those who did not have AREDS listed in their chart during the study period, those patients were part of the non-AREDS group. It was assumed that for patients taking AREDS, the time interval that they took them was the time between their first and last appointments. However, whether patients had been taking AREDS before 2014 was unable to be confirmed, given that EPIC did not have records predating 2014. Given the retrospective nature, patient consent was not required. Patients were eligible for inclusion if they were between the ages of 19–100, had a primary diagnosis of intermediate dry AMD, had at least 3 documented ophthalmology clinic visits during the study period, had a documented smoking history (categorized as current, former, or never), and had documented use or non-use of AREDS supplements. Specific drusen size and pigmentary changes were not documented given the study was a retrospective chart review. Of the initial 1,880 patients screened, 1,430 were excluded based on the following criteria: any history of intraocular or subtenon steroid injections, focal/grid laser photocoagulation, panretinal photocoagulation, or intraocular/refractive surgery in the study eye(s); any previous intravitreal injections; a diagnosis of neovascular AMD, diabetic retinopathy, central retinal vein or artery occlusion, retinal detachment or tear, vitelliform disease, Stargardt disease, or retinal deposits not attributable to AMD; incomplete documentation of smoking history or AREDS supplement use; and fewer than 3 office visits within the study period. Patients with a LogMAR value ≥ 2.4 at baseline or final assessment were also excluded. Demographic and clinical information were categorized by age, sex, race, comorbidities (hypertension, hyperlipidemia, obesity, diabetes, and chronic kidney disease), smoking status, and AREDS use. Patients were stratified into 4 groups based on smoking status (current/former vs. never) and AREDS supplement use. This study was unable to distinguish between AREDS1 and AREDS2 formulations due to limitations in electronic health record documentation. This distinction is clinically relevant, particularly for smokers given the beta-carotene risk, and is discussed further in the limitations. AREDS2 became the predominant formulation at our institution approximately in 2014–2015 following the publication of the AREDS2 trial results; however, we cannot confirm the exact formulation received by individual patients in this retrospective dataset. Visual acuity at the initial and final visit was assessed using LogMAR values. The primary outcomes included changes in LogMAR and associations with smoking history and AREDS use. The ANOVA test and Chi-square test were used to compare baseline characteristics and group differences. Mixed effect regression models were applied to account for individual or group differences to assess differences in baseline and final visual acuity within groups over the follow-up period. Fixed effects were group, age, race, sex, comorbidities, and the individual ID was used as a random effect to account for nested eyes within each individual. In the mixed-effects models, individual patient ID was included as a random intercept to account for the correlation between fellow eyes contributed by the same patient. The eye therefore served as the unit of analysis, nested within a patient-level random effect, with each patient contributing up to two eyes. Results Overview and baseline disparities (Table 1) Table 1Summary statistics and demographicsVariablesNon-smokers & No-AREDS N = 34 individuals (60 eyes)Non-smokers & AREDS N = 161 individuals (276 eyes)Smokers & No-AREDS N = 57 individuals (91 eyes)Smokers & AREDS N = 198 individuals (332 eyes)P-valueAge, mean (SD)83.38 (9.48)83.09 (8.66)82.86 (8.98)81.35 (8.99)0.2380*Sex, n (%) Female Male21 (61.76)13 (38.24)117 (72.67)44 (27.33)31 (54.39)26 (45.61)100 (50.51)98 (49.49)0.0003**Race Caucasian Non-Caucasian18 (52.94)16 (47.06)130 (80.75)31 (19.25)40 (70.18)17 (29.82)167 (84.34)31 (15.66)0.0002**HTNHLDObesityDiabetesCKD32 (94.12)22 (64.71)11 (32.35)18 (52.94)8 (23.53)131 (81.37)106 (65.84)50 (31.06)44 (22.73)41 (25.47)42 (73.68)33 (57.89)15 (26.32)20 (35.09)12 (21.05)171 (86.36)145 (73.23)58 (29.29)56 (28.28)44 (22.22)0.0367**0.12820.90040.0201**0.8703Follow-up (yrs), mean (SD)4.37 (2.95)3.76 (2.46)3.33 (2.67)3.46 (2.47)0.1759*Baseline LogMAR, mean (SD)0.48 (0.44)0.28 (0.28)0.45 (0.47)0.30 (0.33)< 0.0001*Final LogMAR, mean (SD)0.50 (0.47)0.42 (0.45)0.52 (0.51)0.40 (0.37)0.0673*P < 0.05 indicates statistical significance | *ANOVA test | **Chi-square test | ***Eye-level values Table 1Summary statistics and demographicsVariablesNon-smokers & No-AREDS N = 34 individuals (60 eyes)Non-smokers & AREDS N = 161 individuals (276 eyes)Smokers & No-AREDS N = 57 individuals (91 eyes)Smokers & AREDS N = 198 individuals (332 eyes)P-valueAge, mean (SD)83.38 (9.48)83.09 (8.66)82.86 (8.98)81.35 (8.99)0.2380*Sex, n (%) Female Male21 (61.76)13 (38.24)117 (72.67)44 (27.33)31 (54.39)26 (45.61)100 (50.51)98 (49.49)0.0003**Race Caucasian Non-Caucasian18 (52.94)16 (47.06)130 (80.75)31 (19.25)40 (70.18)17 (29.82)167 (84.34)31 (15.66)0.0002**HTNHLDObesityDiabetesCKD32 (94.12)22 (64.71)11 (32.35)18 (52.94)8 (23.53)131 (81.37)106 (65.84)50 (31.06)44 (22.73)41 (25.47)42 (73.68)33 (57.89)15 (26.32)20 (35.09)12 (21.05)171 (86.36)145 (73.23)58 (29.29)56 (28.28)44 (22.22)0.0367**0.12820.90040.0201**0.8703Follow-up (yrs), mean (SD)4.37 (2.95)3.76 (2.46)3.33 (2.67)3.46 (2.47)0.1759*Baseline LogMAR, mean (SD)0.48 (0.44)0.28 (0.28)0.45 (0.47)0.30 (0.33)< 0.0001*Final LogMAR, mean (SD)0.50 (0.47)0.42 (0.45)0.52 (0.51)0.40 (0.37)0.0673*P < 0.05 indicates statistical significance | *ANOVA test | **Chi-square test | ***Eye-level values Summary statistics and demographics 21 (61.76) 13 (38.24) 117 (72.67) 44 (27.33) 31 (54.39) 26 (45.61) 100 (50.51) 98 (49.49) 18 (52.94) 16 (47.06) 130 (80.75) 31 (19.25) 40 (70.18) 17 (29.82) 167 (84.34) 31 (15.66) HTN HLD Obesity Diabetes CKD 32 (94.12) 22 (64.71) 11 (32.35) 18 (52.94) 8 (23.53) 131 (81.37) 106 (65.84) 50 (31.06) 44 (22.73) 41 (25.47) 42 (73.68) 33 (57.89) 15 (26.32) 20 (35.09) 12 (21.05) 171 (86.36) 145 (73.23) 58 (29.29) 56 (28.28) 44 (22.22) 0.0367** 0.1282 0.9004 0.0201** 0.8703 P < 0.05 indicates statistical significance | *ANOVA test | **Chi-square test | ***Eye-level values 450 patients (759 eyes) met inclusion criteria. At the patient level, there were no statistically significant differences in age or follow-up duration between the four groups (P = 0.24 and P = 0.17, respectively). Sex and race distributions differed significantly, with a higher proportion of females in the Non-smokers & AREDS group (72.67%; P = 0.0003) and a higher proportion of Caucasians in the Smokers & AREDS group (84.34%; P = 0.0002). Hypertension was more prevalent in the Non-smokers & No-AREDS group (94.12%; P = 0.037), as was diabetes (52.94%; P = 0.02). Mean follow-up duration was 3.62 years (SD 2.54). Notably, the Non-smokers & No-AREDS and Smokers & No-AREDS groups were considerably smaller (n = 34 and n = 57 individuals, respectively) than the AREDS groups; findings from these subgroups should be interpreted with appropriate caution given reduced statistical power. Visual acuity trajectories (Table 2) Table 2Paired T-test comparing Baseline and Final LogMAR Value within each exposure group, N = 759 eyesGroupBaseline LogMAR ValueFinal LogMAR Value ValueDifference between Baseline and Final LogMAR ValueN eyesMean (SD)N eyesMean (SD)N eyesMean (SD)P-valueNon-smokers & No-AREDS600.48 (0.44)600.50 (0.47)60-0.02 (0.37)0.738Non-smokers & AREDS2760.28 (0.28)2760.42 (0.45)276-0.14(0.37)< 0.0001Smokers & No-AREDS910.45 (0.47)910.52 (0.51)91-0.07 (0.32)0.037Smokers & AREDS3320.30 (0.33)3320.40 (0.37)332-0.10 (0.33)< 0.0001Total7590.33 (0.35)7590.43 (0.43)759-0.10 (0.35)< 0.0001 Table 2Paired T-test comparing Baseline and Final LogMAR Value within each exposure group, N = 759 eyesGroupBaseline LogMAR ValueFinal LogMAR Value ValueDifference between Baseline and Final LogMAR ValueN eyesMean (SD)N eyesMean (SD)N eyesMean (SD)P-valueNon-smokers & No-AREDS600.48 (0.44)600.50 (0.47)60-0.02 (0.37)0.738Non-smokers & AREDS2760.28 (0.28)2760.42 (0.45)276-0.14(0.37)< 0.0001Smokers & No-AREDS910.45 (0.47)910.52 (0.51)91-0.07 (0.32)0.037Smokers & AREDS3320.30 (0.33)3320.40 (0.37)332-0.10 (0.33)< 0.0001Total7590.33 (0.35)7590.43 (0.43)759-0.10 (0.35)< 0.0001 Paired T-test comparing Baseline and Final LogMAR Value within each exposure group, N = 759 eyes -0.14 (0.37) Overall, mean visual acuity declined from 0.33 LogMAR (~ 68 ETDRS letters) at baseline to 0.43 LogMAR (~ 63 ETDRS letters) at the final visit, representing a mean loss of approximately 5 letters (p < 0.0001). AREDS users in both smoking strata entered the study with better baseline vision (~ 70–71 letters) compared to non-users (~ 60–62 letters), a difference that largely accounts for the observed patterns in longitudinal decline. Non-smokers & No-AREDS showed no statistically significant change over follow-up (mean difference − 0.02 ± 0.37, p = 0.7380), while all other groups demonstrated statistically significant worsening. The greatest absolute letter loss occurred in the Non-smokers & AREDS group (~ 7 letters), followed by Smokers & AREDS (~ 5 letters) and Smokers & No-AREDS (~ 3.5 letters). These differences are better explained by the ceiling effect of superior baseline acuity, regression to the mean, and selection bias toward earlier disease identification in AREDS users, rather than a differential treatment response. Regression models depicting mean change in LogMAR value (Table 3) In the baseline model, Non-smokers & No-AREDS and Smokers & No-AREDS had significantly worse baseline acuity than the Smokers & AREDS reference group (β = 0.19, p = 0.0007 and β = 0.14, p = 0.0017, respectively). In the final vision model, none of these group-level differences remained statistically significant, and age was the sole independent predictor of final acuity (β = 0.01, p < 0.0001). The attenuation of group effects between baseline and final models reinforces the interpretation that observed unadjusted differences reflect baseline disparities rather than differential acuity trajectories. In addition, adjusted longitudinal analysis evaluating the rate of visual acuity decline over time after incorporating mixed-effects model predicting the LogMar value difference between final and baseline measurements (ΔVA=VAfinal−VAbaseline) was also done (supplementary Table 1). We adjusted this model for relevant covariates and for the interaction between follow-up time (duration in years) and AREDS groups. The interaction term was added to assess whether AREDS users decline faster or slower over time after adjustment. This additional analysis provided a more clinically interpretable assessment of disease progression and helped mitigate the potential bias introduced by differing baseline acuity levels. Results from the analysis indicated no significant differences in the rates of decline between AREDS groups. The interaction between follow-up time and AREDS groups was also not significant. Table 3Mixed effect regression model depicting the change in mean in LogMAR valueVariableBaseline LogMARFinal LogMARβ95% CIP-valueβ95% CIP-value Group (Reference = Smokers & AREDS) Non-smokers & No-AREDS0.190.08, 0.29 0.001 0.06-0.08, 0.200.380 Non-smokers & AREDS-0.01-0.08, 0.050.6560.00-0.08, 0.080.931 Smokers & No-AREDS0.140.05, 0.23 0.002 0.08-0.03, 0.200.153 Age0.010.00, 0.01 0.015 0.010.01, 0.01 < 0.0001 Non-Caucasians vs. Caucasians (Reference)-0.03-0.10, 0.040.3760.03-0.06, 0.120.558 Male vs. Female (Reference)0.02-0.03, 0.080.3950.03-0.05, 0.100.473 Hypertension0.02-0.06, 0.090.655-0.03-0.13, 0.070.579 Hyperlipidemia-0.05-0.11, 0.010.1080.01-0.07, 0.090.777 Obesity0.02-0.05, 0.080.5790.01-0.07, 0.100.803 Diabetes0.01-0.05, 0.080.6530.04-0.04, 0.130.301 Chronic Kidney Disease0.04-0.03, 0.110.2330.01-0.08, 0.100.794 Duration (years)-0.01-0.02, 0.00 0.016 0.01-0.01, 0.020.240 Table 3Mixed effect regression model depicting the change in mean in LogMAR valueVariableBaseline LogMARFinal LogMARβ95% CIP-valueβ95% CIP-value Group (Reference = Smokers & AREDS) Non-smokers & No-AREDS0.190.08, 0.29 0.001 0.06-0.08, 0.200.380 Non-smokers & AREDS-0.01-0.08, 0.050.6560.00-0.08, 0.080.931 Smokers & No-AREDS0.140.05, 0.23 0.002 0.08-0.03, 0.200.153 Age0.010.00, 0.01 0.015 0.010.01, 0.01 < 0.0001 Non-Caucasians vs. Caucasians (Reference)-0.03-0.10, 0.040.3760.03-0.06, 0.120.558 Male vs. Female (Reference)0.02-0.03, 0.080.3950.03-0.05, 0.100.473 Hypertension0.02-0.06, 0.090.655-0.03-0.13, 0.070.579 Hyperlipidemia-0.05-0.11, 0.010.1080.01-0.07, 0.090.777 Obesity0.02-0.05, 0.080.5790.01-0.07, 0.100.803 Diabetes0.01-0.05, 0.080.6530.04-0.04, 0.130.301 Chronic Kidney Disease0.04-0.03, 0.110.2330.01-0.08, 0.100.794 Duration (years)-0.01-0.02, 0.00 0.016 0.01-0.01, 0.020.240 Mixed effect regression model depicting the change in mean in LogMAR value Sub-analysis: greater than 5 years follow-up Among the 124 individuals (216 eyes) with more than 5 years of follow-up (mean 6.93 ± 1.55 years), a statistically significant difference in final visual acuity across groups was observed (p = 0.0334). The Smokers & AREDS group retained the best final acuity (~ 65.5 ETDRS letters), while Non-smokers & No-AREDS demonstrated the worst (~ 49.0 letters), a gap exceeding the 15-letter threshold for clinically significant change. These findings are hypothesis-generating and should be interpreted with caution: this subgroup is smaller and more selective, and the observed pattern likely reflects the combined influence of baseline disparities, longer disease duration, and survivor bias. Full sub-analysis results are presented in Table 4. Table 4Visual acuity comparison in sub-analysis ≥ 5-year follow-up amongst groupsVariableNonsmokers & No-AREDS N = 13 (21 eyes)Nonsmokers & AREDS N = 44 (82 eyes)Smokers & No-AREDS N = 14 (21 eyes)Smokers & AREDS N = 53 (92 eyes)Total N = 124 (216 eyes)P-value*Follow-up (yrs), mean (SD)7.23 (2.04)6.99 (1.45)7.34 (1.61)6.70 (1.49)6.93 (1.55)0.227Baseline LogMAR, mean (SD)†0.57 (0.56)0.24 (0.26)0.37 (0.39)0.28 (0.33)0.30 (0.35)0.001Final LogMAR, mean (SD)†0.72 (0.60)0.46 (0.47)0.44 (0.43)0.39 (0.40)0.45 (0.46)0.033* ANOVA test † Eye-level values NS = Non-smokers; S=Smokers P < 0.05 indicates statistical significance* ANOVA test. † Eye-level values. P < 0.05 indicates statistical significance. Table 4Visual acuity comparison in sub-analysis ≥ 5-year follow-up amongst groupsVariableNonsmokers & No-AREDS N = 13 (21 eyes)Nonsmokers & AREDS N = 44 (82 eyes)Smokers & No-AREDS N = 14 (21 eyes)Smokers & AREDS N = 53 (92 eyes)Total N = 124 (216 eyes)P-value*Follow-up (yrs), mean (SD)7.23 (2.04)6.99 (1.45)7.34 (1.61)6.70 (1.49)6.93 (1.55)0.227Baseline LogMAR, mean (SD)†0.57 (0.56)0.24 (0.26)0.37 (0.39)0.28 (0.33)0.30 (0.35)0.001Final LogMAR, mean (SD)†0.72 (0.60)0.46 (0.47)0.44 (0.43)0.39 (0.40)0.45 (0.46)0.033* ANOVA test † Eye-level values NS = Non-smokers; S=Smokers P < 0.05 indicates statistical significance* ANOVA test. † Eye-level values. P < 0.05 indicates statistical significance. Visual acuity comparison in sub-analysis ≥ 5-year follow-up amongst groups * ANOVA test † Eye-level values NS = Non-smokers; S=Smokers P < 0.05 indicates statistical significance * ANOVA test. † Eye-level values. P < 0.05 indicates statistical significance. Supplementary Fig. 1 illustrates mean baseline and final LogMAR values plotted by year of follow-up for each of the four stratified groups. The visualization shows better baseline acuity among AREDS users with substantial convergence of final visual acuity across groups, reinforcing the interpretation that observed differences are largely driven by baseline disparities rather than differential treatment response. Discussion This retrospective observational study evaluated visual acuity outcomes among patients with intermediate dry AMD stratified by AREDS supplementation status and smoking history over a mean follow-up of 3.62 years (SD 2.54). Across all 759 eyes, mean visual acuity declined from 0.33 LogMAR (~ 68 ETDRS letters) at baseline to 0.43 LogMAR (~ 63 ETDRS letters) at the final visit, representing a mean loss of approximately 5 letters overall (p < 0.0001). While statistically significant worsening was observed in three of the four exposure groups, the magnitude of decline varied meaningfully across groups and warrants careful interpretation in the context of study design limitations. AREDS users demonstrated slightly better baseline visual acuity than non-users, yet showed larger unadjusted increases in LogMAR over the follow-up period. Non-smokers using AREDS experienced the greatest mean letter loss (~ 7 letters), followed by smokers using AREDS (~ 5 letters) and smokers not using AREDS (~ 3.5 letters), while non-smokers not using AREDS showed no statistically significant change. Importantly, after adjustment for baseline visual acuity and relevant covariates in the mixed effects regression model, final visual acuity did not differ significantly across exposure groups (Table 3). This pattern of better initial vision, greater observed decline, but statistically similar adjusted final outcomes highlights the complexity of interpreting longitudinal visual acuity changes in real-world. In addition, the attenuation of group effects between the baseline and final models reinforces the interpretation that observed unadjusted differences are driven primarily by baseline disparities rather than differential acuity trajectories. These unadjusted differences are more parsimoniously explained by three interrelated phenomena. First, a ceiling effect: patients entering with better baseline acuity (~ 71 letters in Non-smokers & AREDS vs. ~60 letters in Non-smokers & No-AREDS) had greater measurable vision to lose over time. Second, regression to the mean: patients with strong baseline acuity are statistically more likely to show larger absolute declines at follow-up independent of any treatment effect. Third, selection bias: patients who elect to take AREDS supplementation may be those identified earlier in the disease course, more engaged in ophthalmic care, and more adherent to follow-up all of which favor better initial acuity but do not necessarily confer protection against long-term decline. A pre-specified sub-analysis restricted to 124 individuals (216 eyes) with more than 5 years of follow-up demonstrated a statistically significant difference in final visual acuity across groups (p = 0.03). The approximately 16.5-letter gap between the best and worst performing groups exceeds the 15-letter threshold for clinically significant change. However, this sub-analysis is hypothesis-generating rather than confirmatory. The subgroup is smaller and more selective, and differences in final acuity across groups likely reflect the compounding influence of baseline disparities and survivor effects over longer observation periods. Language suggesting a treatment effect would not be appropriate; these findings instead motivate prospective trials with longer follow-up and structured adherence monitoring. Another important finding is that smoking status did not emerge as a meaningful independent predictor of visual acuity in either regression model after covariate adjustment. This null finding warrants careful interpretation. The grouping of current and former smokers into a single category likely dilutes the true effect of cumulative tobacco exposure, as retinal damage may persist differently after cessation and lifetime pack-year burden varies considerably within this combined group. The absence of pack-year data precluded any dose-response analysis, which is an important limitation. It is plausible that smoking heterogeneity obscured associations in the regression analyses, particularly given the relatively small size of the No-AREDS subgroups. Future studies should capture pack-year history and analyze current and former smokers separately to better characterize these relationships. An important clinical consideration is the distinction between AREDS1 and AREDS2 formulations, which this study was unable to address due to limitations in electronic health record documentation. This is particularly relevant for smokers: AREDS1 contains beta-carotene, which increases lung cancer risk in smokers, while AREDS2 substitutes lutein and zeaxanthin and is now the recommended formulation for this population [14–16]. At our institution, AREDS2 became the predominant formulation approximately in 2014–2015; however, individual-level formulation data could not be confirmed retrospectively. This ambiguity may have influenced findings in the smoker subgroups and should be addressed in future research. Beyond visual outcomes, the public health and economic implications of improved AMD management are substantial. AMD is a leading cause of irreversible blindness in older adults and is projected to affect nearly 300 million people globally by 2040 [17]. Vision loss significantly impacts independence, increases fall risk, and is associated with depression and cognitive decline [18]. From a financial standpoint, late-stage AMD imposes over $49 billion in economic burden annually in the U.S. alone [19]. This study also revealed a notable prevalence of diabetes in the AREDS smoker cohort, a known confounder in visual decline [20]. Future prospective studies should control for comorbidities and incorporate imaging-based phenotyping. Limitations Several limitations temper the interpretation of these findings. The retrospective design introduces selection bias and limits causal inference. AREDS use was ascertained from patient-reported documentation and actual adherence between visits could not be confirmed. Baseline disease severity including drusen area, volume, and pigmentary abnormality grading was not captured. Smoking status was classified as current or former without pack-year quantification, limiting dose-response analysis and potentially diluting smoking-related effects; the grouping of current and former smokers likely obscures heterogeneity in cumulative tobacco exposure and residual retinal damage after cessation. Group sizes were unequal, with the Non-smokers & No-AREDS and Smokers & No-AREDS groups being notably smaller, which limits statistical power and robustness for those comparisons; findings from these groups should be considered exploratory. The inability to distinguish AREDS1 from AREDS2 is an additional limitation, particularly relevant for smokers given beta-carotene exposure risk. Finally, the patient-level random intercept structure assumes symmetric correlation between fellow eyes and may not fully capture bilateral disease at differing stages. Conclusions In this observational study of intermediate dry AMD, AREDS users began with better baseline visual acuity and demonstrated greater unadjusted declines over time. After adjustment for baseline acuity and relevant covariates over a mean 3.62-year follow-up, final visual acuity did not differ significantly across exposure groups, and age was the sole independent predictor of final acuity. The observed pattern of decline among AREDS users is attributable to baseline disparities, regression to the mean, and unmeasured confounding rather than a supplementation effect. The sub-analysis restricted to patients with greater than 5 years of follow-up revealed a statistically significant difference in final acuity across groups, but these findings are hypothesis-generating and should not be over-interpreted given the smaller, more selective subgroup. These findings should not be interpreted as contradicting the established efficacy of AREDS therapy demonstrated in randomized controlled trials. Future prospective studies incorporating imaging-based phenotyping, structured adherence assessments, pack-year documentation, separate analysis of current and former smokers, and distinction between AREDS1 and AREDS2 formulations are needed to more definitively characterize the impact of antioxidant supplementation on visual acuity trajectories in smokers and non-smokers with intermediate AMD. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1