Abstract
Objective:
To predict 2-year visual acuity (VA) responses to anti-vascular endothelial growth factor (anti-VEGF) therapy using early morphological and functional responses in patients with neovascular age-related macular degeneration (nAMD).
Design:
Cohort within a randomized clinical trial.
Participants:
1185 participants with untreated active nAMD and best corrected VA (BCVA) 20/25 to 20/320 at baseline.
Methods:
Secondary analysis of data from participants randomized to either ranibizumab or bevacizumab and to one of three dosing regimens. Associations of 2-year BCVA responses with baseline morphological and functional characteristics and their change from baseline at 3 months were assessed using univariable and multivariable linear regression models for BCVA change and logistic regression models for ≥3-line BCVA gain from baseline. The performance of predictions for 2-year BCVA outcomes using these characteristics were assessed using R2 for BCVA change and area under the Receiver Operating Characteristic (ROC) curve (AUC) for ≥3-line BCVA gain.
Main Outcome Measures:
BCVA change and ≥3-line gain from baseline at year 2.
Results:
In multivariable analyses that included previously reported significant baseline predictors (baseline BCVA, baseline macular atrophy, baseline retinal pigment epithelium elevation (RPEE) maximum width and early BCVA change from baseline at 3 months, new RPEE occurrence at 3 months was significantly associated with more BCVA gain at 2-year (10.2 letters vs. 3.5 letters for RPEE resolved, p<0.001), and none of the other morphological responses at 3 months were significantly associated with BCVA responses at 2 year. These significant predictors moderately predicted 2-year BCVA gain with a R2=0.36. Baseline BCVA and ≥3-line BCVA gain at 3 months predicted 2-year ≥3-line gain with AUC 0.83 (95% CI: 0.81–0.86).
Conclusion:
Most structural responses on OCT at 3 months were not independently predictive of the 2-year BCVA responses, which was associated with baseline factors and the 3 month BCVA response to anti-VEGF therapy. A combination of baseline predictors, early BCVA and morphological responses at 3 months only moderately predicted the long-term BCVA responses. Future research is needed to better understand the factors contributing to the variation of long-term vision outcome with anti-VEGF therapy.
INTRODUCTION
Anti-vascular endothelial growth factor (anti-VEGF) therapy is the primary treatment for neovascular age-related macular degeneration (nAMD), the major cause of vision loss in individuals 50 years or older in developed countries.1,2 However, patients’ responses to anti-VEGF vary substantially, 3–9 and several baseline factors including age, baseline visual acuity (VA), area of choroidal neovascularization (CNV) and OCT features are predictive of the VA responses in eyes enrolled in CATT after 1 and 2 years of anti-VEGF treatment.10,11 There is evidence that for eyes with nAMD, an early response to anti-VEGF may predict the long-term VA response from several years of anti-VEGF therapy.10,12–14 In a previous secondary analysis of the CATT data, we found that the VA response at week 12 was the strongest predictor of the visual outcomes after one- or two-year treatment of nAMD with ranibizumab or bevacizumab, but it only predicted less than 50% of the variation in VA response.10 Menghini et al. reported similar results that an initial gain in VA from the loading phase treatment for nAMD is predictive of good long-term response, but an initial poor response is not predictive of poor long-term response.12
On the other hand, early morphological responses may also predict long-term VA response. For example, eyes with intraretinal fluid (IRF) at 12 weeks have a poor VA prognosis, but subretinal fluid at 12 weeks does not predict final visual acuity at 1 year.2 Another study also found that regardless of whether subretinal fluid persisted or not, VA outcomes were similar after adjusting for other variables.13 However, there is a weak association between a reduction of central subfield retinal thickness of 20% or more at three weeks and long-term VA at 1 and 3 years.13
To evaluate whether the early optical coherence tomography (OCT) morphological responses predict long-term vision response and whether the combination of early visual acuity responses and morphological responses can improve the prediction of the long-term vision outcome from anti-VEGF treatment for nAMD, we performed a secondary data analysis from the Comparison of AMD Treatment Trials (CATT). Predicting the long-term vision outcomes using the early treatment responses will allow us to forecast a patient’s clinical course and adjust the expectations of ophthalmologists and patients; such prediction also provides useful information for considering alternative treatments (e.g., combination therapy) if the improvement in long-term vision outcome is unlikely with the current drug. This paper reports our findings on whether early morphological responses and VA response at 3 months predict the VA response at 2 years.
METHODS
This is a secondary analysis of data collected during the 2-year CATT Study. The CATT was conducted in accordance to the tenets of the Declaration of Helsinki and the Health Insurance Portability and Accountability Act. The Institutional Review Boards of clinical centers participating in the CATT approved the study protocol and written informed consent was obtained from all patients. The details of the study design and methods have been reported in previous publications 3,4 and in ClinicalTrials.gov (NCT00593450). Here we only describe the major features related to this secondary analysis.
Study Participants:
Each clinical center’s institutional review board approved the study protocol, and each patient provided informed consent. The patients were enrolled from 43 clinical centers in the United States and randomized to one of four treatment groups: (1) Lucentis monthly; (2) Avastin monthly; (3) Lucentis as needed (pro re nata, PRN); and (4) Avastin PRN.
The trial eligibility criteria included age at least 50 years, untreated active CNV due to AMD in the study eye (one eye per patient), and best corrected visual acuity (BCVA) between 20/25 and 20/320 on electronic VA testing. An eye was considered to have active CNV if there was both leakage of dye on fluorescein angiography and fluid, located either within or below the retina or below the RPE, on time-domain OCT.
Study Procedures:
Patients provided information about their demographic and medical history at the baseline visit. Study-certified photographers acquired stereoscopic, color fundus photographs and fluorescein angiograms at baseline, year 1, and year 2. Study-certified OCT imagers acquired OCT images from all participants at baseline, week 4, 8, 12, 24, 52, 76 and 104. Trained, masked readers at the Duke Reading Center evaluated OCT images of the study eye for presence and location (intraretinal, subretinal, sub-RPE) of fluid, subretinal hyper-reflective material (SHRM) and retinal pigment epithelium (RPE) elevation. Readers measured the retinal, subretinal fluid (SRF), and subretinal tissue complex thickness following a standard grading protocol.15
At baseline and weeks 4, 12, 24, 36, 52 (Year 1), 64, 76, 88, and 104 (Year 2) of follow-up, study-certified VA examiners, who were masked to the treatment assignments, used the Electronic Visual Acuity (EVA) Tester to measure VA after refraction (i.e., BCVA) in both eyes following the Diabetic Retinopathy Clinical Research Network’s protocol.16 The VA scores from EVA ranged from 0 to 100, corresponding with the Snellen equivalents of worse than 20/800 to 20/10.
Statistical Analysis:
We evaluated associations between 3-month responses (morphological responses in OCT and VA response) and 2-year BCVA change from baseline using univariable and multivariable linear regression models. We also evaluated associations of 3-month responses with ≥3-line gain in BCVA from baseline at year 2 using univariable and multivariable logistic regression models. The 3-month OCT responses evaluated as predictors in these analyses include presence and thickness of intraretinal fluid, presence and thickness of subretinal fluid, presence and thickness of sub-RPE fluid, presence of subretinal hyper reflective material (SHRM), presence of RPE elevation (RPEE), RPEE maximum height, and RPEE maximum width, presence of epiretinal membrane, and vitreomacular attachment within central 3 mm. Univariable analyses were first performed to evaluate the association between the VA responses at 2 years (e.g., BCVA change from baseline and BCVA gain of ≥3 lines from baseline at year 2) with each of the OCT morphological features/measures at 3 months, the changes of these morphological features/measures from baseline at 3 months, and the variability in OCT thickness within the first three months (e.g., standard deviation of OCT thickness calculated using measures at baseline, months 1, 2, 3 for each study eye). The multivariable analyses were then used to determine the significant predictors for the BCVA change from baseline at year 2 and BCVA gain of ≥3 lines from baseline at year 2. The initial multivariable models included OCT features or their changes from baseline at 3 months with p<0.10 from univariable analyses and the previously reported statistically significant baseline predictors for VA responses at 2 years.10 The multivariable models went through backward variable selection by only keeping the statistically significant predictors.17 The performance of the predictors for predicting 2-year BCVA change from baseline in these regression models was evaluated by R2, which was calculated as the percent of variance of Year 2 VA response that was explained by the predictors. R2 values range between 0 and 1; 0 means there is only random variation whereas 1 means there is perfect prediction. The performance of the predictors for predicting ≥3-line gain from baseline at year 2 was evaluated using the area under the Receiver Operating Characteristic (ROC) curve (AUC) from logistic regression models.
For the secondary analysis, the study eyes in both the ranibizumab and bevacizumab treatment groups were combined because their effects on visual acuity outcomes were similar.1 The categorizations of OCT and fundus features/measures followed the same approach as that in previous CATT publications.10,11,18 The multiple comparisons from evaluations of many factors for their associations with visual outcomes were not corrected with two-sided p<0.05 considered statistically significant. All these analyses were performed using SAS v9.4 (SAS Institute Inc., Cary, NC).
RESULTS
Univariable Analysis for Association between OCT Features at Month 3 and 2-Year VA Responses
In univariable analyses (Table 1) for OCT features at 3 months, none of these OCT features at 3 months were significantly associated with the BCVA change from baseline at 2 years (all p≥0.09).
Table 1.
Univariable analysis for associations between OCT features at 3 months and visual acuity outcomes at 2 years
| Change in visual acuity from baseline at 2 years | BCVA gain ≥3 lines from baseline at 2 years | |||||
|---|---|---|---|---|---|---|
| OCT features at 3 months | # of eyes | Mean (SE) in letters | P-value | n (%) | Odds Ratio (95% CI) | P-value |
| Intraretinal fluid | 0.14 | 0.40 | ||||
| No fluid | 443 | 6.93 (0.78) | 128 (28.9%) | 1.05 (0.75–1.48) | ||
| Fluid not under foveal center | 248 | 7.34 (1.04) | 82 (33.1%) | 1.27 (0.87–1.87) | ||
| Fluid under foveal center | 254 | 4.70 (1.03) | 71 (28.0%) | 1.0 | ||
| Subretinal fluid | 0.42 | 0.47 | ||||
| No fluid | 625 | 6.68 (0.66) | 194 (31.0%) | 1.20 (0.83–1.76) | ||
| Fluid not under foveal center | 148 | 4.85 (1.35) | 40 (27.0%) | 0.99 (0.60–1.61) | ||
| Fluid under foveal center | 172 | 7.04 (1.25) | 47 (27.3%) | 1.0 | ||
| Sub-RPE fluid | 0.23 | 0.03 | ||||
| No fluid | 599 | 7.12 (0.66) | 193 (32.2%) | 1.64 (1.10–2.49) | ||
| Fluid not under foveal center | 138 | 6.14 (1.37) | 36 (26.1%) | 1.22 (0.71–2.07) | ||
| Fluid under foveal center | 160 | 4.70 (1.28) | 36 (22.5%) | 1.0 | ||
| Any of intraretinal fluid, subretinal fluid, or sub-RPE fluid | 0.09 | 0.11 | ||||
| No fluid | 263 | 7.71 (1.01) | 86 (32.7%) | 1.35 (0.97–1.88) | ||
| Fluid not under foveal center | 231 | 7.40 (1.08) | 76 (32.9%) | 1.36 (0.96–1.92) | ||
| Fluid under foveal center | 449 | 5.21 (0.77) | 119 (26.5%) | 1.0 | ||
| Subretinal hyper reflective material (SHRM) | 0.43 | 0.01 | ||||
| No | 431 | 5.96 (0.80) | 110 (25.5%) | 1.00 | ||
| Yes | 523 | 6.81 (0.73) | 175 (33.5%) | 1.47 (1.11–1.96) | ||
| RPE elevation | 0.73 | 0.33 | ||||
| No | 129 | 5.91 (1.46) | 43 (33.3%) | 1.22 (0.81–1.80) | ||
| Yes | 831 | 6.45 (0.58) | 242 (29.1%) | 1.0 | ||
| Epiretinal membrane | 0.29 | 0.36 | ||||
| No | 799 | 6.71 (0.58) | 240 (30.0%) | 1.21 (0.81–1.83) | ||
| Yes | 141 | 4.72 (1.38) | 37 (26.2%) | 1.0 | ||
| Vitreomacular attachment within central 3 mm | 0.91 | 0.71 | ||||
| No | 835 | 6.47 (0.57) | 247 (29.6%) | 0.92 (0.59–1.45) | ||
| Yes | 102 | 6.28 (1.62) | 32 (31.4%) | 1.0 | ||
| Intraretinal fluid thickness (um) | 0.10 | 0.28 | ||||
| ≤120 | 170 | 6.68 (1.27) | 53 (31.2%) | 1.48 (0.87–2.56) | ||
| >120 to ≤212 | 676 | 6.81 (0.64) | 205 (30.3%) | 1.42 (0.91–2.29) | ||
| >212 | 115 | 3.29 (1.54) | 27 (23.5%) | 1.0 | ||
| Subretinal fluid thickness (um) | 0.55 | 0.58 | ||||
| 0 | 795 | 6.16 (0.59) | 239 (30.1%) | 1.26 (0.80–2.03) | ||
| >0 to ≤25 | 60 | 8.53 (2.14) | 19 (31.7%) | 1.36 (0.67–2.72) | ||
| >25 | 106 | 6.67 (1.61) | 27 (25.5%) | 1.0 | ||
| Sub-RPE fluid thickness (um) | 0.20 | 0.16 | ||||
| ≤100 | 553 | 6.72 (0.70) | 156 (28.2%) | 1.08 (0.71–1.67) | ||
| >100 to ≤240 | 276 | 6.80 (1.00) | 94 (34.1%) | 1.42 (0.90–2.26) | ||
| >240 | 131 | 3.95 (1.45) | 35 (26.7%) | 1.0 | ||
| RPEE maximum height (um) | 0.29 | 0.52 | ||||
| ≤3 | 349 | 6.88 (0.87) | 109 (31.2%) | 1.27 (0.81–2.01) | ||
| >3 to ≤5 | 206 | 5.59 (1.14) | 56 (27.2%) | 1.04 (0.64–1.73) | ||
| >5 to ≤11.5 | 268 | 7.59 (1.00) | 85 (31.7%) | 1.30 (0.82–2.09) | ||
| >11.5 | 129 | 4.61 (1.44) | 34 (26.4%) | 1.0 | ||
| RPEE maximum width (um) | 0.40 | 0.89 | ||||
| ≤12 | 282 | 5.91 (0.97) | 84 (29.8%) | 1.07 (0.71–1.62) | ||
| >12 to ≤31.5 | 258 | 7.57 (1.01) | 76 (29.5%) | 1.05 (0.69–1.60) | ||
| >31.5 to ≤55 | 194 | 7.65 (1.16) | 62 (32.0%) | 1.18 (0.76–1.84) | ||
| >55 | 183 | 5.60 (1.20) | 52 (28.4%) | 1.0 | ||
In the univariable analysis for associations between these OCT features and a BCVA gain of ≥3 lines from baseline at 2 years (Table 1), absence of sub-RPE fluid (OR=1.64, p=0.03) and presence of SHRM (OR=1.47, p=0.01) at 3 months were significantly associated with higher likelihood of gaining ≥ 3 lines from baseline at 2 years (Table 1).
Univariable Analysis for Associations between Change in OCT Features from Baseline at Month 3 and 2-Year VA Responses
In univariable analyses, change of several OCT features from baseline at 3 months were significantly associated with greater BCVA gain at 2 years (Table 2), including resolution of foveal center intraretinal fluid (p=0.003), resolution of SHRM (p=0.02), new development of RPEE (p<0.0001), greater decrease in intraretinal fluid thickness (p<0.001), greater decrease in sub-RPE fluid thickness (p=0.02), and change of retinal thickness from >212 um to ≤212 um (p=0.004).
Table 2.
Univariable analysis for associations between change from baseline in OCT features at 3 months and visual acuity outcomes at 2 years
| Change in visual acuity from baseline at 2 years | BCVA gain ≥3 lines from baseline at 2 years | |||||
|---|---|---|---|---|---|---|
| Change in OCT features from baseline at 3 months | # of Eyes | Mean (SE) in letters | p-value | n (%) | Odds Ratio (95% CI) | p-value |
| Change in intraretinal fluid | 0.003 | 0.002 | ||||
| Foveal center fluid resolved | 296 | 9.03 (0.95) | 110 (37.2%) | 1.79 (1.28–2.49) | ||
| Foveal center fluid remained or new foveal center fluid developed | 251 | 4.71 (1.03) | 71 (28.3%) | 1.19 (0.83–1.71) | ||
| Other | 382 | 5.51 (0.84) | 95 (24.9%) | Ref. | ||
| Change in subretinal fluid | 0.82 | 0.46 | ||||
| Foveal center fluid resolved | 241 | 6.53 (1.06) | 67 (27.8%) | 0.84 (0.60–1.18) | ||
| Foveal center fluid remained or new foveal center fluid developed | 171 | 7.12 (1.26) | 47 (27.5%) | 0.83 (0.56–1.21) | ||
| Other | 529 | 6.23 (0.71) | 166 (31.4%) | 1.0 | ||
| Change in sub-RPE fluid | 0.06 | 0.03 | ||||
| Foveal center fluid resolved | 150 | 8.52 (1.32) | 51 (34.0%) | 1.20 (0.81–1.75) | ||
| Foveal center fluid remained or new foveal center fluid developed | 152 | 4.16 (1.31) | 32 (21.1%) | 0.62 (0.40–0.94) | ||
| Other | 538 | 6.53 (0.70) | 162 (30.1%) | 1.0 | ||
| Change in subretinal hyper reflective material | 0.02 | <0.0001 | ||||
| New occurrence | 45 | 1.36 (2.46) | 9 (20.0%) | 0.47 (0.21–0.95) | ||
| Resolved | 251 | 7.53 (1.04) | 81 (32.3%) | 0.89 (0.64–1.23) | ||
| Remained negative | 173 | 4.01 (1.26) | 28 (16.2%) | 0.36 (0.23–0.56) | ||
| Remained positive | 474 | 7.33 (0.76) | 165 (34.8%) | 1.0 | ||
| Change in RPEE | <0.0001 | 0.005 | ||||
| New occurrence | 73 | 13.01 (1.92) | 31 (42.5%) | 1.92 (1.17–3.13) | ||
| Resolved | 71 | 2.06 (1.95) | 17 (23.9%) | 0.82 (0.45–1.42) | ||
| Remained negative | 52 | 10.85 (2.27) | 23 (44.2%) | 2.07 (1.16–3.65) | ||
| Remained positive | 750 | 5.80 (0.60) | 208 (27.7%) | 1.0 | ||
| Change in epiretinal membrane | 0.23 | 0.14 | ||||
| New occurrence | 46 | 7.11 (2.41) | 16 (34.8%) | 2.23 (0.99–5.01) | ||
| Resolved | 47 | 5.83 (2.38) | 15 (31.9%) | 1.96 (0.87–4.42) | ||
| Remained negative | 721 | 6.77 (0.61) | 216 (30.0%) | 1.79 (1.05–3.20) | ||
| Remained positive | 88 | 2.47 (1.74) | 17 (19.3%) | 1.0 | ||
| Change in VMT-vitreous attached within central 3 mm | 0.42 | 0.92 | ||||
| New occurrence | 21 | 10.14 (3.56) | 7 (33.3%) | 1.06 (0.36–2.89) | ||
| Resolved | 41 | 9.63 (2.55) | 13 (31.7%) | 0.98 (0.43–2.20) | ||
| Remained negative | 749 | 6.26 (0.60) | 219 (29.2%) | 0.88 (0.54–1.47) | ||
| Remained positive | 78 | 5.90 (1.85) | 25 (32.1%) | 1.0 | ||
| Change in intraretinal fluid thickness from baseline (um) | 0.02 | <0.001 | ||||
| ≤−60 | 308 | 8.06 (0.94) | 116 (37.7%) | 2.09 (1.49–2.95) | ||
| >−60 to ≤−10 | 301 | 6.76 (0.95) | 90 (29.9%) | 1.48 (1.04–2.10) | ||
| >−10 | 348 | 4.53 (0.89) | 78 (22.4%) | 1.0 | ||
| As continuous | −0.02 (0.01) | <0.001 | 1.00 (0.99–1.00) | 0.04 | ||
| Change in subretinal fluid thickness from baseline (um) | 0.54 | 0.44 | ||||
| <0 | 299 | 6.19 (0.96) | 81 (27.1%) | 0.92 (0.56–1.52) | ||
| 0 | 554 | 6.15 (0.71) | 173 (31.2%) | 1.12 (0.71–1.80) | ||
| >0 | 104 | 8.07 (1.63) | 30 (28.8%) | 1.0 | ||
| As continuous | 0.01 (0.01) | 0.09 | 1.00 (1.00–1.00) | 0.69 | ||
| Change in sub-RPE fluid thickness from baseline | 0.004 | 0.003 | ||||
| ≤−100 | 302 | 8.62 (0.95) | 110 (36.4%) | 1.81 (1.28–2.56) | ||
| >−100 to ≤−16 | 326 | 6.47 (0.92) | 95 (29.1%) | 1.30 (0.92–1.84) | ||
| >−16 | 328 | 4.21 (0.91) | 79 (24.1%) | 1.0 | ||
| As continuous | −0.01 (0.00) | 0.02 | 1.00 (1.00–1.00) | 0.98 | ||
| Change in status of retinal thickness <120 pm | 0.36 | 0.62 | ||||
| New occurrence | 127 | 8.00 (1.47) | 42 (33.1%) | 1.44 (0.67–3.24) | ||
| Resolved | 57 | 6.28 (2.20) | 14 (24.6%) | 0.95 (0.38–2.40) | ||
| Remained negative | 730 | 6.31 (0.61) | 217 (29.7%) | 1.23 (0.63–2.60) | ||
| Remained positive | 43 | 2.79 (2.53) | 11 (25.6%) | 1.0 | ||
| Change in status of retinal thickness >212 pm | 0.004 | <0.0001 | ||||
| New occurrence | 35 | 4.60 (2.79) | 9 (25.7%) | 1.17 (0.45–2.91) | ||
| Resolved | 261 | 9.26 (1.02) | 108 (41.4%) | 2.39 (1.36–4.38) | ||
| Remained negative | 582 | 5.67 (0.68) | 149 (25.6%) | 1.17 (0.68–2.09) | ||
| Remained positive | 79 | 2.77 (1.86) | 18 (22.8%) | 1.0 | ||
| Change in RPEE maximum height | 0.30 | 0.15 | ||||
| ≤−4 | 240 | 6.08 (1.05) | 68 (28.3%) | 0.68 (0.45–1.02) | ||
| >−4 to ≤−1 | 276 | 5.93 (0.98) | 80 (29.0%) | 0.70 (0.47–1.04) | ||
| >−1 to ≤1 | 222 | 6.39 (1.09) | 60 (27.0%) | 0.64 (0.42–0.97) | ||
| >1 | 182 | 8.66 (1.21) | 67 (36.8%) | 1.0 | ||
| As continuous | 0.13 (0.09) | 0.14 | 1.02 (0.99–1.04) | 0.18 | ||
| Change in RPEE maximum width | 0.77 | 0.59 | ||||
| ≤−13 | 217 | 6.26 (1.10) | 59 (27.2%) | 0.76 (0.50–1.15) | ||
| >−13 to ≤−2 | 233 | 6.06 (1.06) | 66 (28.3%) | 0.80 (0.53–1.21) | ||
| >−2 to ≤6 | 213 | 7.26 (1.11) | 62 (29.1%) | 0.83 (0.55–1.26) | ||
| >6 | 203 | 7.40 (1.14) | 67 (33.0%) | 1.0 | ||
| As continuous | 0.03 (0.03) | 0.25 | 1.01 (1.00–1.01) | 0.07 | ||
Similarly, change of several OCT features from baseline at 3 months were significantly associated with gain of ≥3 lines from baseline at 2 years (Table 2), including resolution of foveal center intraretinal fluid (p=0.002), resolution of foveal center sub-RPE fluid (p=0.03), resolution of SHRM (p<0.0001), new development of RPEE (p=0.005), greater decrease in intraretinal fluid thickness (p<0.001), greater decrease in sub-RPE fluid thickness (p=0.003), and change of retinal thickness from >212 um to ≤212 um, (p<0.0001).
Univariable Association Between Variability in OCT Thickness within the First 3 Months of Treatment and 2-Year VA Responses
To evaluate whether the fluctuations of OCT thickness in the first 3 months of treatment were associated with 2-year VA responses, the standard deviation (SD) for intraretinal fluid thickness, subretinal fluid thickness, and sub-RPE fluid thickness in each eye was calculated using their measured values from baseline and weeks 4, 8 and 12, and associations of SDs of these fluid thickness with 2-year VA responses were evaluated using univariable regression models (Table 3). Only variation of sub-RPE fluid thickness was significantly associated with 2-year BCVA change from baseline; the eyes with small SD (≤15 um) had a BCVA gain of 2.8 letters, which was significantly smaller than eyes with SD >15 um (p<0.001). The variation for intraretinal fluid thickness (p=0.28), subretinal fluid thickness (p=0.13) and total OCT thickness (p=0.09) were all not significantly associated with 2-year BCVA change from baseline (Table 3).
Table 3.
Univariable analysis on association between OCT thickness within first three months (time points included: baseline, months 1, 2 and 3) and visual acuity outcome at 2 years
| Change in visual acuity from baseline at 2 years | BCVA gain ≥3-line from baseline at 2 years | |||||
|---|---|---|---|---|---|---|
| Variability in OCT thickness in the first 3 months | # of eyes | Mean (SE) Letters | P-value | n (%) | Odds Ratio (95% CI) | P-value |
| Standard deviation of intraretinal fluid thickness in the first 3 months (um) | 0.28 | 0.001 | ||||
| ≤15 | 246 | 5.01 (1.05) | 55 (22.4%) | 0.50 (0.34–0.74) | ||
| >15 to ≤20 | 124 | 4.91 (1.49) | 27 (21.8%) | 0.48 (0.29–0.79) | ||
| >20 to ≤50 | 398 | 6.92 (0.83) | 128 (32.2%) | 0.82 (0.59–1.15) | ||
| >50 | 263 | 7.27 (1.02) | 96 (36.5%) | 1.0 | ||
| Standard deviation of subretinal fluid thickness in first 3 months | 0.13 | 0.79 | ||||
| 0 | 527 | 6.16 (0.72) | 161 (30.6%) | 1.19 (0.83–1.72) | ||
| ≤14 | 149 | 6.88 (1.35) | 46 (30.9%) | 1.21 (0.76–1.93) | ||
| >14 to ≤40 | 151 | 8.67 (1.34) | 44 (29.1%) | 1.11 (0.70–1.78) | ||
| >40 | 204 | 4.54 (1.16) | 55 (27.0%) | 1.0 | ||
| Standard deviation of sub-RPE fluid thickness in first 3 months | <0.001 | <0.0001 | ||||
| ≤15 | 269 | 2.80 (1.00) | 52 (19.3%) | 0.45 (0.30–0.67) | ||
| >15 to ≤35 | 239 | 7.14 (1.06) | 69 (28.9%) | 0.76 (0.52–1.11) | ||
| >35 to ≤70 | 258 | 9.07 (1.02) | 93 (36.0%) | 1.06 (0.74–1.52) | ||
| >70 | 265 | 6.45 (1.01) | 92 (34.7%) | 1.0 | ||
| Standard deviation of total OCT thickness in first 3 months (um) | 0.09 | <0.0001 | ||||
| ≤40 | 261 | 4.17 (1.02) | 49 (18.8%) | 0.39 (0.26–0.58) | ||
| >40 to ≤70 | 249 | 6.48 (1.05) | 73 (29.3%) | 0.69 (0.48–1.01) | ||
| >70 to ≤120 | 267 | 7.63 (1.01) | 89 (33.3%) | 0.84 (0.58–1.20) | ||
| >120 | 254 | 6.96 (1.04) | 95 (37.4%) | 1.0 | ||
When associations between variability in thickness and gain of ≥ 3 lines were evaluated (Table 3), smaller variations of intraretinal fluid thickness (p=0.001), sub-RPE fluid thickness (p<0.0001), and total OCT thickness (p<0.0001) were significantly associated with lower likelihood of a gaining ≥3 lines in BCVA.
Multivariable Associations Between All Factors (at Baseline and 3 Months) and 2-Year VA Responses
The multivariable analysis initially included all the significant baseline predictors previously reported from CATT,10,11 BCVA change from baseline at 3 months and significant OCT features and their changes from baseline at 3 months. After backward variable selection, only worse baseline BCVA (p<0.0001), absent baseline macular atrophy in the study eye (p=0.01), RPEE maximum width at baseline (p=0.02), new occurrence of RPEE at 3 months (p=0.04), and greater BCVA gain at 3 months (p<0.0001) were significantly associated with more BCVA gain from baseline at 2 years (Table 4). These significant factors predicted 2-year BCVA gain with a R2 of 0.36 (Figure 1).
Table 4.
Multivariable analysis for associations between baseline factors, OCT responses and visual acuity response at 3 months and visual acuity outcomes at 2 years*
| Change in visual acuity from baseline at 2 years | BCVA gain ≥3-line from baseline at 2 years | ||||||
|---|---|---|---|---|---|---|---|
| Predictors | # of eyes | Mean (SE) Letters | p-value | # of eyes | N (%) | Odds Ratio (95% CI) | p-value |
| Baseline BCVA in study eye | <0.0001 | <0.0001 | |||||
| 20/100 to 20/320 | 222 | 11.54 (0.90) | 250 | 134 (53.6%) | 14.3 (8.19, 25.0) | ||
| 20/50 to 20/80 | 328 | 6.22 (0.74) | 352 | 135 (38.4%) | 8.25 (4.80, 14.2) | ||
| 20/25 to 20/40 | 338 | 3.77 (0.73) | 361 | 18 (5.0%) | 1.0 | ||
| Baseline macular atrophy in study eye | 0.01 | ||||||
| No | 822 | 6.98 (0.46) | |||||
| Yes | 66 | 2.62 (1.63) | |||||
| Baseline RPEE maximum width | 0.02 | ||||||
| ≤12 | 226 | 6.87 (1.19) | |||||
| >12 to ≤31.5 | 220 | 4.87 (0.93) | |||||
| >31.5 to ≤55 | 215 | 8.84 (0.94) | |||||
| >55 | 227 | 6.11 (0.94) | |||||
| Change in RPEE from baseline at 3 months | 0.04 | ||||||
| New occurrence | 73 | 10.19 (1.92) | |||||
| Resolved | 67 | 3.54 (1.66) | |||||
| Remained negative | 52 | 9.34 (2.15) | |||||
| Remained positive | 696 | 6.43 (0.55) | |||||
| BCVA change from baseline at 3 months (per letter increase) | 0.77 (0.04) | <0.0001 | |||||
| Gain of >3 lines from baseline at 3 months | <0.0001 | ||||||
| No | 787 | 150 (19.1%) | Ref | ||||
| Yes | 176 | 137 (77.8%) | 10.0 (6.25, 14.3) | ||||
Only factors with statistically significant association (p<0.05) with the specific BCVA outcome in the final multivariable model were reported in this table.
The R2 =0.36 for predicting BCVA change from baseline at 2 years from the multivariable model all these statistically significant predictors in multivariable model.
The area under ROC curve=0.83 (95% CI: 0.81–0.86) from the multivariable model for predicting BCVA gain ≥3 lines from baseline at 2 years with all these statistically significant predictors in multivariable model.
Figure 1: Scatterplot for observed BCVA change at 2-years vs. Predicted BCVA change at 2-year based on multivariable model presented in Table 4.
The scatterplot with LOESS line shows the agreement between observed BCVA change and predicted BCVA change from baseline at 2 years. These data points are scattered around the LOESS line with R2=0.36 (R2 ranges from 0 to 1, with 1 indicating perfect prediction).
In the multivariable analysis for ≥3 lines BCVA gain from baseline at 2 years, only worse baseline BCVA (p<0.0001) and presence of BCVA gain ≥3 lines at 3 months (p<0.0001) were significantly associated with BCVA gain of ≥3 lines at 2 years (Table 4). These two factors predicted a BCVA gain of ≥3 lines at 2-years moderately well with an AUC of 0.83 (95% CI: 0.81–0.86) (Figure 2).
Figure 2. ROC curve for the multivariable regression analysis predicting BCVA gain ≥3-line from baseline at year 2.
The ROC curve shows the prediction for BCVA gain ≥3-line from baseline at year 2 using baseline BCVA and BCVA gain of ≥3-line from baseline at 3 months after treatment. The area under ROC curve is 0.83 (95% CI: 0.81–0.86) indicating moderate prediction. The area under ROC curve ranges from 0 to 1 with 1 indicating perfect prediction.
DISCUSSION
In this secondary analysis of CATT data, we performed a detailed evaluation for associations between early morphological responses in OCT features at 3 months and the long-term VA responses at 2 years. Although changes of five OCT features at 3 months were found to be significantly associated with better VA response at 2-years in the univariable analysis; including resolution of intraretinal fluid, decrease of sub-RPE fluid thickness, retinal thickness >212 um, resolution of SHRM, and development of RPEE; their associations were not significant after they were adjusted by baseline predictors and early VA response at 3 months, with the exception of the development of RPEE which remained statistically significant in multivariable analyses. The combination of baseline predictors and the early morphological and vision responses only moderately predicted the VA responses at 2-years (R2=0.36 for BCVA change from baseline, AUC=0.83 for BCVA gain of ≥3 lines).
Previous analysis of data from IVAN and CATT found that greater variation in retinal thickness across the entire duration of the trial in eyes treated with anti-VEGF for neovascular AMD was associated with worse VA response, higher risk of developing fibrosis and macular atrophy.19 In contrast from the previous analysis, this analysis evaluated how the variation of retinal thickness during the first 3 months of anti-VEGF treatment was associated with 2-year VA responses. We found that variation of the sub-RPE fluid within the first three months was significantly associated with VA responses (p<0.001) in a non-monotone manner. Eyes in the third quartile of variation (standard deviation 35 to 70) had the greatest BCVA gain (gain of 9 letters compared to 2.8 letters in the 1st quartile and 6.5 letters in the 4th quartile) and the highest percentage of eyes with a BCVA gain of ≥3 lines (36.0% as compared with 19% in the 1st quartile, and 34.7% in the 4th quartile). Greater variation of the intraretinal fluid thickness within the first three months was also associated with a higher percentage of eyes with a BCVA gain ≥3 lines (p=0.001). However, these significant association did not remain after they were adjusted by baseline predictors and early VA response at 3 months.
In multivariable analyses, worse baseline BCVA score, absence of baseline macular atrophy and baseline maximum RPEE width were significantly associated with more BCVA gain from baseline at 2-years. When considering the early vision response and morphological responses with adjustment of significant baseline predictors, we found that only more BCVA gain from baseline at 3 months and new occurrence of RPEE at 3 months were significantly associated with greater BCVA gain from baseline at 2 years. This finding is consistent with previous research that found early VA response was a good predictor of late VA response.2,10,12 However, these baseline predictors and early VA and morphological responses at 3 months only moderately predicted long-term VA responses at 2-years with a R2 of 0.36 (i.e., these predictors only predicted 36% of variation of in BCVA change from baseline at 2-year). This moderate prediction performance could be due to some eyes undertaking an unpredictable course after 3 months of anti-VEGF treatment, including developing macular atrophy or scars or reactivating nAMD, which could all lead to poor VA outcomes.
When evaluating factors associated specifically with a BCVA gain of ≥3 lines at 2 years, only baseline BCVA and gaining ≥3 lines in BCVA at 3 months from baseline were significantly associated with BCVA gain of ≥3 lines at 2 years in multivariable analysis. These two factors predicted BCVA gain of ≥3 lines at 2 years with AUC of 0.83. These results align with past research that has concluded early VA response is a strong predictor of late VA response.2,10,12
A prior study used machine learning of SD-OCT images to identify the image biomarkers for predicting the VA responses in eyes treated with anti-VEGF for neovascular AMD.20 The machine learning technique found that the horizontal extension of intraretinal cystoid fluid in the foveal region was the most relevant biomarker for visual acuity response at 12 months, whereas subretinal fluid and pigment epithelial detachments (PED) were not predictive. Consistent with our analysis, their study found that the strongest predictive factor for VA outcomes at 1 year was the visual acuity response during the initiation loading phase.
Previous analysis of CATT data found that RPEE anywhere in the macula was common (85%) at baseline and was independently associated with worse BCVA, lower mean increase in VA score, and a lower proportion with ≥3 lines gained in BCVA at one-year after anti-VEGF treatment. Based on this finding, it was hypothesized that eyes with RPE elevation at baseline due to sub-RPE hemorrhage, neovascularization or fibrosis, or drusen alone, could signal abnormal RPE functional activity and may have consequently adversely affected BCVA.11 However, this analysis of RPE evaluation during follow-up found that new occurrence of RPEE at 3 months after anti-VEGF treatment was independently associated with more BCVA gain from baseline at 2 years. This finding is unexpected, and the reason is unknown on why baseline RPEE was associated with worse VA outcomes, but new occurrence of RPEE at follow-up was associated more BCVA gain from baseline at 2 years. It is possible that the RPEE during follow-up after anti-VEGF treatment is different from baseline RPEE before anti-VEGF treatment (in terms of its cause and composition), leading to a different effect on VA. Future study is needed to better understand the new occurrence of RPEE during anti-VEGF treatment and to confirm or refute our finding.
In summary, this secondary analysis was performed to find early morphological responses that could be used to predict the long-term VA responses at 2 years. Although we found a few OCT morphological responses that predicted VA response at 2 years on univariable analysis, these morphological responses did not independently predict the long-term VA response; thus, they did not add much beyond what were already predicted by the baseline predictors and early VA response at 3 months. A combination of baseline predictors and early vision and morphological responses at 3 months only moderately predicted the long-term VA responses in terms of VA change from baseline and gain of ≥3 lines from baseline at 2 years. Consequently, early response in OCT features are not strong predictors for long-term vision outcomes in response to anti-VEGF therapy for patients with nAMD. Future research is needed to better understand the factors contributing to the variation of long-term vision response.
Supplementary Material
Precis:
In the secondary analysis of the data from the Comparison of AMD Treatments Trials, we found a combination of baseline predictors, early vision and morphological responses at 3 months moderately predicted 2-year vision responses.
Acknowledgments
Supported by cooperative agreements U10 EY017823, U10 EY017825, U10 EY017826, U10 EY017828 and R21EY028998 from the National Eye Institute, National Institutes of Health, Department of Health and Human Services.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References:
- 1.Smith W, Assink J, Klein R, et al. Risk factors for age-related macular degeneration: Pooled findings from three continents. Ophthalmology. 2001;108(4):697–704. [DOI] [PubMed] [Google Scholar]
- 2.Ashraf M, Souka A, Adelman RA. Age-related macular degeneration: using morphological predictors to modify current treatment protocols. Acta Ophthalmol. 2018;96(2):120–133. [DOI] [PubMed] [Google Scholar]
- 3.Martin DF, Maguire MG, Ying GS, Grunwald JE, Fine SL, Jaffe GJ. Ranibizumab and bevacizumab for neovascular age-related macular degeneration. N Engl J Med. 2011;364(20):1897–1908. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Martin DF, Maguire MG, Fine SL, et al. Ranibizumab and bevacizumab for treatment of neovascular age-related macular degeneration: two-year results. Ophthalmology. 2012;119(7):1388–1398. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Krebs I, Schmetterer L, Boltz A, et al. A randomised double-masked trial comparing the visual outcome after treatment with ranibizumab or bevacizumab in patients with neovascular age-related macular degeneration. Br J Ophthalmol. 2013;97(3):266–271. [DOI] [PubMed] [Google Scholar]
- 6.Kodjikian L, Souied EH, Mimoun G, et al. Ranibizumab versus Bevacizumab for Neovascular Age-related Macular Degeneration: Results from the GEFAL Noninferiority Randomized Trial. Ophthalmology. 2013;120(11):2300–2309. [DOI] [PubMed] [Google Scholar]
- 7.Heier JS, Brown DM, Chong V, et al. Intravitreal aflibercept (VEGF trap-eye) in wet age-related macular degeneration. Ophthalmology. 2012;119(12):2537–2548. [DOI] [PubMed] [Google Scholar]
- 8.Chakravarthy U, Harding SP, Rogers CA, et al. Ranibizumab versus bevacizumab to treat neovascular age-related macular degeneration: one-year findings from the IVAN randomized trial. Ophthalmology. 2012;119(7):1399–1411. [DOI] [PubMed] [Google Scholar]
- 9.Chakravarthy U, Harding SP, Rogers CA, et al. Alternative treatments to inhibit VEGF in age-related choroidal neovascularisation: 2-year findings of the IVAN randomised controlled trial. Lancet. 2013;382(9900):1258–1267. [DOI] [PubMed] [Google Scholar]
- 10.Ying GS, Maguire MG, Daniel E, et al. Association of Baseline Characteristics and Early Vision Response with 2-Year Vision Outcomes in the Comparison of AMD Treatments Trials (CATT). Ophthalmology. 2015;122(12):2523–2531. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ying GS, Huang J, Maguire MG, et al. Baseline predictors for one-year visual outcomes with ranibizumab or bevacizumab for neovascular age-related macular degeneration. Ophthalmology. 2013;120(1):122–129. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Menghini M, Kloeckener-Gruissem B, Fleischhauer J, et al. Impact of loading phase, initial response and CFH genotype on the long-term outcome of treatment for neovascular age-related macular degeneration. PLoS One. 2012;7(7):e42014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Gonzalez VH, Campbell J, Holekamp NM, et al. Early and Long-Term Responses to Anti-Vascular Endothelial Growth Factor Therapy in Diabetic Macular Edema: Analysis of Protocol I Data. Am J Ophthalmol. 2016;172:72–79. [DOI] [PubMed] [Google Scholar]
- 14.Bressler NM, Beaulieu WT, Maguire MG, et al. Early Response to Anti-Vascular Endothelial Growth Factor and Two-Year Outcomes Among Eyes With Diabetic Macular Edema in Protocol T. Am J Ophthalmol. 2018;195:93–100. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Decroos FC, Toth CA, Stinnett SS, Heydary CS, Burns R, Jaffe GJ. Optical coherence tomography grading reproducibility during the Comparison of Age-related Macular Degeneration Treatments Trials. Ophthalmology. 2012;119(12):2549–2557. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Beck RW, Moke PS, Turpin AH, et al. A computerized method of visual acuity testing: adaptation of the early treatment of diabetic retinopathy study testing protocol. Am J Ophthalmol. 2003;135(2):194–205. [DOI] [PubMed] [Google Scholar]
- 17.Chowdhury MZI, Turin TC. Variable selection strategies and its importance in clinical prediction modelling. Fam Med Community Health. 2020;8(1):e000262. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Jaffe GJ, Martin DF, Toth CA, et al. Macular morphology and visual acuity in the comparison of age-related macular degeneration treatments trials. Ophthalmology. 2013;120(9):1860–1870. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Evans RN, Reeves BC, Maguire MG, et al. Associations of Variation in Retinal Thickness With Visual Acuity and Anatomic Outcomes in Eyes With Neovascular Age-Related Macular Degeneration Lesions Treated With Anti-Vascular Endothelial Growth Factor Agents. JAMA Ophthalmol. 2020;138(10):1043–1051. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Schmidt-Erfurth U, Bogunovic H, Sadeghipour A, et al. Machine Learning to Analyze the Prognostic Value of Current Imaging Biomarkers in Neovascular Age-Related Macular Degeneration. Ophthalmol Retina. 2018;2(1):24–30. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.