Online electronic databases (PubMed, Web of Science, and the Cochrane Library) were searched with an end date of August 12, 2024. The following MeSH terms were used in [Title/Abstract]: “Macular Degeneration or Age-Related Macular Degeneration or AMD or ARMD or nAMD”, and “ranibizumab or Lucentis”, and “aflibercept or Eylea”. Additionally, the “related articles” function was used to complement the searches of the reference lists of all retrieved studies.

Only observational studies that reported results of comparison between aflibercept monotherapy and ranibizumab monotherapy for wAMD patients with no previous therapy, and that had at least one quantitative outcome of visual function or retinal morphology reported, were included. When multiple published articles described the same population, the most recent or complete report was used. Only studies published in peer-reviewed journals and in English were considered, irrespective of main outcomes, date, region or publication types. Conference papers/abstracts, editorials, letters to authors, reviews, commentaries and news were not included in the review. In addition, journal articles were excluded if they met the following criteria: 1) Follow-up time less than 3 months; 2) Case reports; 3) Pre-clinical studies; 4) Relatively small sample size (less than 20 eyes/group).

Figure 1 shows a flow chart of the selection process used to identify relevant studies. Data of included studies were extracted and summarized independently by two reviewers (X.W. and C.Y.). Any disagreement was resolved by a third reviewer (J.D.). The main outcomes were BCVA, CRT and sfCT. The other outcomes were incidence of “DM” and sfPEDH.

Studies were rated for the level of evidence provided according to criteria by the Centre for Evidence-Based Medicine (Oxford, UK). The methodological quality of all cohort studies was assessed by the Newcastle-Ottawa scale (NOS) [8] , which consists of three factors: patient selection, comparability of the study groups, and assessment of outcome (Supplementary Table S1 ). A 10-point scale was used and a score of 0 - 9 was allocated to each study. Two reviewers (X.W. and C.Y.) assessed the quality of the studies. Any discrepancies were resolved by a third reviewer (J.D.). Observational studies achieving a score of seven or more points were considered to be of high quality.

All analyses were performed using Review Manager 5.3 (Cochrane Collaboration,

Oxford, UK). The weighted mean difference (WMD) and odds ratio (OR) were used to analyze continuous and dichotomous variables, respectively. All results were reported with 95% confidence intervals (CIs). If continuous data were presented as means and range values, the standard deviation (SD) were calculated using the technique described by Hozo et al. [9] . An OR of less than 1 favored the Aflibercept group. Heterogeneity between studies was assessed by the χ² and I² statistic. The random-effects model was used if the p value was less than 0·1, otherwise, the fixed-effects model was reported. A two-tailed p < 0.05 was considered significant. Begg’s test and funnel plot analyses were used to assess the publication bias.

Subgroup analyses were performed to compare BCVA at baseline of less than or more than 0.6 logMAR (55 ETDRS letters) based on a recent study which reported that BCVA improvement was different between eyes with a baseline acuity at more than 0.6 logMAR vs. eyes with lower baseline BCVA [7] .

As a result of the literature search, 802, 1225 and 114 studies published in English in PubMed, Web of Science and Cochrane, respectively, were identified. After checking for duplications, 1035 studies were kept for stepwise review. Of these studies, 106 articles that were relevant to the study topic were retained for full text review. Finally, after full-text review of these 106 articles, 21 studies met the inclusion criteria [10] - [30] . Agreement between the two reviewers was 95% for study selection and 95% for quality assessment of trials after examination of references listed for studies. The literature search process was summarized in Figure 1 .

There was one prospective study [28] and the remaining studies were retrospective. In total, there were 6832 eyes in aflibercept and 6172 eyes in ranibizumab groups, respectively. Follow-up time varied from 1 to 24 months, the majority of studies had a 3-months (13, 61.9%) and 12-months (14, 66.7%) follow-up after the initial treatment. Only two studies have reported the change of BCVA and CRT at 24-months follow-up, however, they have both been analyzed in a previous meta-analysis and are not reported here [7] . Detailed information about follow-up and regimen was summarized in Supplementary Table S2 .

Quality assessment showed one trial with a score of 5, which was a relatively lower score compared to the other studies [11] . The quality of remaining studies was relatively high, with an average score of 7.9 (Supplementary Table S3 ). Two articles recorded the same population with different outcomes reported, therefore, both studies were included [16] [17] .

The measured BCVA values were converted to logarithm of the minimum angle of resolution (logMAR) units. Effects of aflibercept and ranibizumab on mean BCVA change were compared at two different time points of follow-up (3 months and 12 months). Although BCVA change has been compared in a previous meta-analysis [7] , more accurate results could be obtained in this work with the inclusion of five more studies.

The pooled BCVA change data from 13 studies [10] - [12] [15] [16] [18] - [21] [23] [26] [28] [29] showed no difference in BCVA at 3 months (WMD: −0.01; 95% CI: −0.04 to 0.01; p = 0.42), while the data from 14 studies [13] - [16] [19] [20] [22] - [27] [29] [30] showed borderline significance in favor of aflibercept (WMD: −0.04; 95% CI: −0.07 to 0.00; p = 0.04) at 12 months post initiation of treatment ( Figure 2 ). Due to high heterogeneity observed, a random-effects model was used at 3 months (χ² = 19.73, df = 12, p = 0.07; I² = 39%) and at 12 months (χ² = 33.91, df = 13, p = 0.001; I² = 62%). No publication bias was detected by Begg’s test for the comparison effects on BCVA at 3 months (p = 0.583; Figure 3(A) ) and at 12 months (p = 0.381; Figure 3(B) ). At one year follow-up visit, this outcome differs from the same measure reported in a previous meta-analysis which showed no significant difference [7] , most likely due to an increased number of included studies (fourteen in the current one vs. eleven in the previous one). However,

high heterogeneity among studies indicated a relatively weak advantage in favor of aflibercept at 12 months.

The CRT was defined as the distance between the inner surface of the neurosensory retina and the retinal pigment epithelium (RPE) in the central retina. Compared with the previous meta-analysis, two more studies [28] [29] reported the treatment effects of CRT thinning between aflibercept and ranibizumab in routine clinical practice. The two pooled analyses from nine [10] [12] [15] [16] [19] [20] [26] [28] [29] and eight [13] [15] [16] [19] [20] [24] [26] [29] studies showed no significant difference in CRT change between two groups at the 3 months(WMD: −2.93; 95% CI: −20.10 to 14.25; p = 0.74) and 12 months (WMD: 2.64; 95% CI: −18.55 to 23.83; p = 0·81) visit, respectively. No heterogeneity was observed among the studies at 3months (χ² = 8.39, df = 8, p = 0.4; I² = 5%) and 12 months (χ² = 4.40, df = 7, p = 0·73; I² = 0%) ( Figure 4 ). No publication bias was detected by Begg’s test for the comparison effects on CRT at 3 months (p = 0.466) and at 12 months (p = 0.174).

The sfCT was defined as the vertical distance between Bruch’s membrane and the chorioscleral interface. Six studies [10] [12] [17] [18] [21] [28] including 741 eyes assessed sfCT change at 3 months of follow-up. Despite some tendency for more pronounced thinning in sfCT of aflibercept group, the analysis showed no significant difference between the two groups (WMD: −8.98; 95% CI: −19.38 to 1.43; p = 0·09), with significant heterogeneity among studies (χ² = 30.68, df = 5, p < 0.0001; I² = 84%) ( Figure 5 ). No publication bias was detected by Begg’s test (p = 0.707).

Subgroup analysis

Subgroup analyses performed by Zhang et al. have been shown that lower baseline BCVA (<0.6 logMAR) might lead to a better visual outcome with the use of aflibercept at 12 months [7] . However, a subgroup analysis of 3-month post

treatment initiation could be also valuable. Therefore, we performed subgroup analyses based on both 3-month and 12-month treatment outcomes for different baseline BCVA levels.

In terms of BCVA improving, there was no significant difference either at 3 months (WMD: 0.01; 95% CI: −0.02 to 0.03; p = 0.71) or 12 months (WMD: −0.04; 95% CI: −0.09 to 0.01; p = 0.17) between the two groups for baseline BCVA better than 0.6 logMAR. For baseline BCVA worse than 0.6 logMAR, no significant difference was observed in 3 months (WMD: −0.02; 95% CI: −0.07 - 0.02; p = 0.35), but a borderline significant difference in favor of aflibercept was observed for BCVA change at one year post treatment (WMD: −0.05; 95% CI: −0.10 to 0.00; p = 0.05).

In terms of CRT thinning, no significant difference at 3 months (WMD: 7.41; 95% CI: −12.26 to 27.07; p = 0.46) and 12 months post treatment (WMD: 12.55;

95% CI: −11.52 to 36.63; p = 0.31) was observed in eyes with baseline BCVA better than 0.6 logMAR. When the eyes with a baseline BCVA worse than 0.6 logMAR was compared, no significant difference was observed at one year post treatment (WMD: −31.42; 95% CI: −76.05 to 13.21; p = 0.17), but a significant difference in favor of aflibercept in CRT change at 3 months follow-up (WMD: −36.19; 95% CI: −71.47 to −0.92; p = 0.04).

In terms of sfCT thinning, two studies showed that aflibercept was slightly superior in patients with baseline BCVA better than 0.6 logMAR (WMD: −12.67; 95% CI: −21.33 to −4.02; p = 0.004). Only one study reported change in sfCT in eyes with relatively worse baseline BCVA, which made it impossible to perform analysis for this morphological parameter.

The post-treatment incidence of complete resolution of the accumulation of intraretinal or subretinal fluid (referred to as “DM’’) was evaluated. Pooling the data from three studies [10] [16] [28] we assessed the incidence of “DM” in 343 eyes after a treatment course of 3 months. The eyes in aflibercept-treated group showed a higher incidence of “DM” compared to the ranibizumab-treated group (OR: 2.26; 95% CI: 1.33 to 3.82; p = 0.003). No significant heterogeneity was observed in the included studies (χ² = 2.64%, df = 2, p = 0.27; I² = 64%) ( Table 1 ). No publication bias was detected by Begg’s test (p = 1.00).

The sfPEDH was defined as the distance between the outer border of Bruch’s membrane and the inner border of the RPE at the fovea. Pooling the data assessed the mean change in absolute thickness of sfPEDH in 274 eyes at 3 months [10] [19] and in 212 eyes at 12 months [19] [24] . Compared to ranibizumab, aflibercept showed to be slightly more effective in reducing sfPEDH by average ~43.9 μm (95% CI: −73.88 to −13.87; p = 0.004) at 3 months, and by ~34.1 μm (95% CI: −76.30 to 8.07; p = 0.11) at 12 months. There was no significant heterogeneity at 3 months (χ² = 0.03, df = 1, p = 0·86; I² = 0%) and 12 months (χ² = 1.96, df = 1, p = 0.16; I² = 49%) ( Table 1 ). No publication bias was detected by Begg’s test for the comparison effects on sfPEDH at 3 months (p = 1.00) and at 12 months (p = 1.00).

BCVA: best-corrected visual acuity; logMAR: logarithm of minimum angle of resolution; CRT: central retinal thickness; sfPEDH: subfoveal pigment epithelial detachment height; sfCT: subfoveal choroidal thickness; DM: “dry macula”; WMD: weighted mean difference; OR: odds ratio; 95% CI: 95% credible interval.