Macular Morphology and Visual Acuity in Year Five of the Comparison of Age-related Macular Degeneration Treatments Trials (CATT)

Abstract

Objective:

To evaluate associations of morphologic features with 5-year visual acuity (VA) in the Comparison of Age-related Macular Degeneration (AMD) Treatments Trials (CATT).

Design:

Cohort study within a randomized clinical trial.

Participants:

Participants in CATT.

Methods:

Eyes with AMD-associated choroidal neovascularization (CNV) and VA between 20/25 and 20/320 were eligible. Treatment was assigned randomly to ranibizumab or bevacizumab and to 3 dosing regimens for 2 years and was at the ophthalmologists’ discretion thereafter.

Main Outcome Measures:

VA; thickness and morphological features on optical coherence tomography; lesion size and foveal composition on fundus photography and fluorescein angiography.

Results:

VA and image gradings were available for 523 of 914 (57%) participants alive at 5 years. At 5 years, 60% of eyes had intraretinal fluid (IRF), 38% had subretinal fluid (SRF), 36% had sub-retinal pigment epithelium (RPE) fluid, and 66% had subretinal hyper-reflective material (SHRM). Mean (SD) foveal center thickness (μm) was 148 (99) for retina, 5 (21) for SRF, 125 (107) for subretinal tissue complex, 11 (33) for SHRM, and 103 (95) for RPE+RPE elevation. SHRM, thinner retina, greater CNV lesion area and foveal center pathology (all p<0.001) and IRF (p<0.05), were independently associated with worse VA. Adjusted mean VA letters was 62 for no pathology in the foveal center, 61 for CNV, fluid, or hemorrhage, 65 for non-geographic atrophy (GA), 64 for non-fibrotic scar, 53 for GA, and 56 for fibrotic scar. Incidence or worsening of eight pathological features (foveal GA, foveal scar, foveal CNV, SHRM, foveal IRF, retinal thinning, CNV lesion area, and GA area) between years 2 and 5 were independently associated with greater loss of VA from year 2 to 5, and VA loss from baseline to year 5.

Conclusions:

Associations between VA and morphologic features previously identified through year 1 were maintained or strengthened at year 5. New foveal scar, CNV, intraretinal fluid, SHRM and retinal thinning, development or worsening of foveal GA, and increased lesion size, are important contributers to the VA decline from year 2 to 5. A significant need to develop therapies to address these adverse pathological features remains.

During years 1 and 2 of the Comparison of Age-related Macular Degeneration (AMD) Treatment Trials (CATT), anti VEGF therapy with either ranibizumab (Lucentis®; Genentech, South San Francisco, CA) and bevacizumab (Avastin®; Genentech) resulted in rapid and sustained reduction in all types of retinal fluid and thickness, stabilization of lesion growth, and reduction in vascular leakage, and an associated improvement in visual acuity (VA)^1–4. Intraretinal fluid, but not subretinal fluid or sub-RPE fluid was independently associated with worse VA. Very thin or very thick retinas, thick subretinal tissue, atrophy and fibrotic scar were associated with worse VA^3,4.

Participants continued to receive anti-VEGF therapy during the ensuing three years of CATT in a real-world treatment scenario whereby the participant’s ophthalmologist chose the anti-VEGF agent and dosing frequency according to her/his best judgement. As we have recently reported, the average VA at the five-year follow-up time point had declined from year 2, to a level that was below the baseline value at trial entry. Among eyes followed to 5 years, mean VA had improved from 62 to 70 letters by year 2, but had declined to 59 letters by year 5, representing an 11 letter loss⁵. Furthermore, other studies, including those from electronic medical record databases and those from interventional trials have demonstrated a decline in visual acuity with long-term anti-VEGF therapy^6–8. However, there are very little data that address directly the morphological correlates to this visual acuity decline, and how this information could be applied in the clinic or in clinical trials. In the present report, we assessed the associations between macular morphology and VA during 5 years of anti-VEGF treatment, and we explored the retinal anatomic factors that contributed to the year 5 VA results in a real-world treatment setting.

Methods

Study Population

Details of the design and methods for CATT have been published previously^1,2. A total of 1185 subjects were enrolled from 43 US clinical centers between February 2008 and December 2009. Only one eye per subject, the study eye, was treated as a part of the clinical trial. Inclusion criteria included subject age >50 years, presence of previously untreated active choroidal neovascularization (CNV) secondary to AMD in the study eye, and VA between 20/25 and 20/320. Choroidal neovascularization was considered active when leakage or increased stippling was seen on FA and intraretinal, subretinal, and/or sub-RPE fluid was documented on OCT. Choroidal neovascularization or its sequelae (i.e., pigment epithelium detachment, hemorrhage, blocked fluorescence, macular edema, or fluid) needed to involve the center of the fovea. For the CNV to be considered secondary to AMD, at least 1 druse >63 μm needed to be present in the study eye or fellow eye, or the fellow eye needed to have CNV or geographic atrophy. Participants were initially assigned randomly with equal probability to 1 of 4 treatment groups in year 1: (1) ranibizumab monthly, (2) bevacizumab monthly, (3) ranibizumab as needed (pro re nata [PRN]), or (4) bevacizumab PRN. During the second year participants in the monthly arms were re-randomized to either continue on monthly treatment or switch to PRN therapy². Participants were released from the study protocol after year 2, and were treated with anti-VEGF therapy (aflibercept, ranibizumab, or bevacizumab) at dosing intervals as determined by the treating physician’s best judgement. They were recalled for an eye examination, and ancillary image assessment at approximately 5 years (5.5 ± 0.6 years). Those participants who were assessed at this follow-up visit, and their associated clinical data comprised the CATT follow-up study (CATT FS). For ease of description, in the following text, we refer to data collected in CATT FS as the year 5 time point.

The institutional review boards associated with each center approved the study. All participants provided written informed consent. The study was performed in compliance with the Health Insurance Portability and Accountability Act.

Study Procedures

Follow-up methods

A detailed description of the methods used to enroll CATT FS subjects, and the procedures performed at the year 5 visit have been described previously⁵. Briefly, all patients who enrolled in the clinical trial, except for those known to be deceased at 2 years, were targeted for participation in CATT FS. Returning patients had a dilated eye examination, refraction and best-corrected VA measurement, optical coherence tomography (OCT), fundus color stereophotography (FP) and fluorescein angiography (FA). All examinations were performed by study-certified personnel following the same protocols used during the clinical trial. OCT, FP, and FA were obtained at baseline, and at pre-specified intervals through year 2, at variable frequencies during years 2-5 and on all CATT FS participants, when possible, at year 5. Procedures used for the acquisition, analysis and grading of OCT, FA and color fundus photography images have been published previously^1,3.Time domain OCT images were obtained throughout year one, while 22.6% of scans were obtained on spectral domain (SD) OCT in year 2 as has been reported previously². All CATT FS OCT images were obtained with SDOCT.

Data and Statistical Analysis

Only patients with a VA measurement between 53 months (4.4 years) and 83 months (6.9 years) after the date of treatment assignment in the clinical trial were included in the data analyses. To evaluate the association of OCT thickness measurements with VA, thickness measurements were also divided into categories in the same way as in previous year 1 and year 2 reports^1,2 . To evaluate the association of each type of OCT fluid with VA, OCT fluid was categorized as one of the following: no fluid, extrafoveal fluid, and foveal-center fluid. The associations of OCT thickness and OCT fluid with VA were assessed using analysis of variance for any difference among the categories and linear trend p-value for the ordered measurements. A post-hoc analysis of all possible pairwise comparisons with adjustment for multiple comparisons was performed using the Hochberg procedure⁹.

The association of retinal morphology findings from FP, FA, or OCT with VA at year 5, and the association of morphological change with VA change from year 2 to year 5, were analyzed with multiple regression models. Backward variable selection processes were used by retaining only variables with p<0.05 in the final multivariate model. All statistical analyses were performed in SAS version 9.2 (SAS Inc., Cary, NC), and 2-sided P values <0.05 were considered to be statistically significant.

Results

Association of Morphological Features on Optical Coherence Tomography and Visual Acuity by Univariate Analysis

Morphological and VA data were available on 523 patients in CATT FS and comprised the analysis population (Figure 1, Table 1; available at www.aaojournal.org). At 5 years, on OCT, 60% had intraretinal fluid (IRF), 38% had subretinal fluid (SRF), 36% had subretinal pigment epithelium fluid (sub-RPEF), 66% had subretinal hyper-reflective material (SHRM), and 21% had outer retinal tubulations (ORTs). The mean (SD) thickness in um was 279 (160) for foveal center total thickness, 148 (99) for retinal thickness, 5 (21) for SRF, 125 (107) for subretinal tissue complex, 11 (33) for SHRM, and 103 (95) for RPE+RPE elevation. Overall, the VA was slightly worse in eyes with macular fluid, compared to those without, 63 vs 59 and 61 letters for eyes without fluid, or those with non-foveal or foveal fluid, respectively (Figure 2A). However, mean VA differed, depending on the specific type of fluid. As was seen at years 1 and 2, relative the mean VA in eyes with no IRF (68 letters) , mean VA was worse for eyes with extrafoveal IRF (57 letters; p<0.001) and worse still for those with foveal IRF (44 letters ;p<0.001, Figure 2B). In contrast, relative to mean VA in eyes with extrafoveal SRF (57 letters), mean VA was better for eyes with foveal SRF (68 letters, p=0.02), and similar to those without SRF (61 letters; Figure 2C). A trend towards better VA in Eyes with foveal sub-RPEF had better mean VA (73 letters) than eyes without sub-RPEF (60 letters; p=0.006) or those with extrafoveal sub-RPEF (60 letters; p=0.01); Figure 2D). As observed previously through year 2 ¹⁰, eyes with SHRM had worse mean VA, particularly if it involved the foveal center (41 for foveal SHRM, 63 for extrafoveal SHRM, and 72 letters for no SHRM; p<0.001 for each pairwise comparison). Eyes with ORT also had worse mean VA when compared to those without ORT (52 vs. 63 letters; p<0.001).

Figure 1.

Flow chart for analysis cohort

Table 1:

Univariate analysis for the association of optical coherence tomography morphological features with visual acuity at Year 5 (N=523)

OCT Feature at Year 5	n (%)	Unadjusted Mean (SE) of VA Score in Letters at Year 5	P-value*
Any fluid			0.40
No Fluid	89 (17.0%)	63.2 (2.8)
Fluid not in foveal center	323 (61.8%)	59.4 (1.3)
Fluid in foveal center	108 (20.7%)	60.9 (2.1)
Unknown	3 (0.6%)	30.3 (14.7)
Intraretinal fluid present			<0.001
No Fluid	206 (39.4%)	67.8 (1.4)
Fluid not in foveal center	269 (51.4%)	57.3 (1.5)
Fluid in foveal center	43 (8.2%)	44.3 (3.6)
Unknown	5 (1.0%)	37.2 (9.8)
Subretinal fluid present			0.02
No Fluid	315 (60.2%)	61.1 (1.4)
Fluid not in foveal center	154 (29.4%)	57.4 (2.0)
Fluid in foveal center	45 (8.6%)	68.4 (1.9)
Unknown	9 (1.7%)	34.2 (7.3)
Sub-RPE fluid present			0.02
No Fluid	327 (62.5%)	60.1 (1.4)
Fluid not in foveal center	161 (30.8%)	59.5 (1.8)
Fluid in foveal center	28 (5.4%)	73.1 (1.9)
Unknown	7 (1.3%)	25.3 (8.7)
Subretinal hyper reflective material present			<0.001
No	177 (33.8%)	71.9 (1.2)
Yes, not at foveal center	211 (40.3%)	62.7 (1.4)
Yes, at foveal center	132 (25.2%)	41.0 (2.3)
Unknown	3 (0.6%)	42.3 (11.6)
Retinal thickness (microns)			<0.001 (<0.001)
<120	188 (35.9%)	50.2 (1.8)
120-212	273 (52.2%)	68.7 (1.2)
>212	60 (11.5%)	54.2 (3.1)
Unknown	2 (0.4%)	12.0 (12.0)
Subretinal tissue complex thickness (microns)			0.004 (0.002)
<50	122 (23.3%)	63.5 (2.0)
>=50, <90	138 (26.4%)	64.4 (1.8)
>=90, <160	125 (23.9%)	55.8 (2.3)
>=160	138 (26.4%)	56.9 (2.1)
Total retinal thickness by quartile (microns)			<0.001 (0.009)
<185	139 (26.6%)	51.2 (2.1)
>=185, <235	121 (23.1%)	65.6 (2.1)
>=235, <330	131 (25.0%)	66.5 (1.8)
>=330	132 (25.2%)	58.4 (2.1)
Total retinal thickness (microns)			<0.001 (0.01)
0-325	382 (73.0%)	60.8 (1.2)
>325 - 425	70 (13.4%)	64.9 (2.0)
>425 - 550	30 (5.7%)	62.4 (4.6)
>550	38 (7.3%)	45.6 (4.5)
Subretinal fluid thickness			0.07 (0.02)
0	469 (89.7%)	59.9 (1.1)
>0, <25	13 (2.5%)	66.4 (4.2)
>= 25	32 (6.1%)	69.2 (2.2)
Sub-retinal tissure complex thickness			0.01 (0.002)
0-75	208 (39.8%)	64.4 (1.5)
>75 - 160	175 (33.5%)	58.2 (1.9)
>160 - 275	97 (18.5%)	58.1 (2.4)
>275	41 (7.8%)	53.9 (4.5)
Outer retinal tubulation present			<0.001
No	403 (77.1%)	63.2 (1.1)
Yes	112 (21.4%)	51.5 (2.5)
Unknown	8 (1.5%)	27.4 (8.3)
Macular atrophy present			<0.001
No	284 (54.3%)	65.4 (1.3)
Yes	232 (44.4%)	55.1 (1.6)
Unknown	7 (1.3%)	19.6 (7.7)

CNV = choroidal neovascularization; RPE = retinal pigment epithelium; SE = standard error; VA = visual acuity.

^*P-value in the parenthesis was for test of trend, while P-value not in the parenthesis was for the test of overall difference from one way analysis of variance, and Unknown category is not included for P-value calculation.

Figure 2. Correlation of retinal fluid on OCT with visual acuity.

A) Visual acuity vs. any fluid. B) Visual acuity vs. intraretinal fluid. C) Visual acuity vs sub-retinal fluid. D) Visual acuity vs. sub-RPE fluid.

Correlation of Optical Coherence Tomography-Determined Thickness Measurements with Visual Acuity by Univariate Analysis

At year 5, eyes with total thickness >550 μm had markedly worse mean VA (46 letters) than that eyes with < 550 μm (mean of 61-65 letters) (all p<0.05 when compared to the 3 other thickness groups; Figure 3A and Table 1, available at www.aaojournal.org). The relationship between VA and retinal, subretinal fluid, subretinal tissue complex, and SHRM thickness was determined. As shown previously for CATT 1- and 2-year data, and prominently at year 5, eyes with very thin (<120 μm; 50 letters) or thick retinas (>212 μm; 54 letters) had worse mean VA than eyes with normal retinal thickness (120-212 μm; 69 letters; all P<0.001; Figure 3B). When foveal SRF data were stratified by subretinal fluid thickness categories ( 0 μm, 1-25 μm, and >25 μm), eyes with foveal SRF thickness > 0 μm, had better mean VA (69 letters), than eyes with SRF thickness of 0 μm (60 letters, Figure 3C), and eyes with foveal SRF had better mean VA (68 letters) than eyes with no SRF, 61 letters, or extrafoveal SRF 57, Table 1, available at www.aaojournal.org). Increasing thickness of the subretinal tissue complex was associated with increasingly worse mean VA (linear trend P=0.002), Figure 3D).

Figure 3. Visual acuity with thickness measurements on OCT at baseline, and years 1, 2, and 5.

A) Visual acuity vs. total thickness. B) Visual acuity vs. retinal thickness. C) Visual acuity vs sub-retinal fluid thickness. D) Visual acuity vs. sub-retinal tissue complex thickness.

Correlation of Fundus Features Determined on Fluorescein Angiograms and Color Fundus Photographs with Visual Acuity on Univariate Analysis

At year 5, larger neovascular lesion area was associated with worse VA (P<0.0001; Table 2; available at www.aaojournal.org). Eyes with lesion area 5 mm² or less had a mean VA of 72 letters vs. 49 letters among eyes with lesion area great than 20 mm². The presence and type of pathology in the foveal center as determined by FP and FA was associated with worse VA (P<0.0001; Table 2: available at www.aaojournal.org). Eyes with no pathology in the foveal center had the best mean VA of 70 letters, whereas the mean for eyes with scar or with geographic atrophy was 46 letters.

Table 2:

Univariate analysis for mean visual acuity at Year 5 by neovascular lesion area and foveal center pathology at Year 5 (N=523)

Fundus Feature at Year 5	N	Unadjusted Mean (SE) of VA Score in Letters at Year 5	P-value*
Total area of CNV lesion (mm2)			<0.001 (<0.001)
>=0 to <=5	113	71.6 (1.5)
>5 to <=10	109	64.7 (1.9)
>10 to <=20	116	59.9 (2.1)
>20	91	48.5 (2.9)
Unknown	94	52.9 (2.8)
Foveal center pathology			<0.001
None	85	70.3 (2.1)
Fibrotic Scar	91	46.4 (3.0)
Geographic atrophy	84	45.9 (2.7)
Non-geographic atrophy	121	69.9 (1.4)
CNV or fluid, hemorrhage	53	64.5 (2.9)
Non-fibrotic scar	26	71.1 (2.6)
Other, CG/CD**	63	58.6 (2.8)

CNV = choroidal neovascularization; RPE = retinal pigment epithelium; SE = standard error; VA = visual acuity.

^*P-value in the parenthesis was for test of trend, while P-value not in the parenthesis was for the test of overall difference from one way analysis of variance, and Unknown category is not included for P-value calculation.

^**This category includes RPE Tear(n=4), other(n=6) and can not grade/can not determine (n=53).

Multivariate Analysis of the Association Between Visual Acuity and Optical Coherence Tomography and Fundus Features

The presence and foveal involvement of each of the 3 types of fluid on OCT, the thickness of each of the 3 retinal layers, the lesion size and the foveal pathology were considered simultaneously in a multivariate linear regression model of VA and a reduced, final model was determinated by backward variable selection. (Table 3). We found that the presence and foveal involvement of IRF and SHRM, retinal thickness, particularly <120 μm, larger total CNV lesion area and the type of foveal pathology were all independently associated with worse VA at year 5. Of note, eyes with foveal GA or fibrotic scar, had worse VA when compared to those without foveal pathology, or those with non-geographic atrophy, non-fibrotic scar or CNV/fluid/hemorrhage (Table 3). This finding is notable given the large proportion of eyes at year 5 with foveal GA and fibrotic scar; Figures 4A-D; available at www.aaojournal.org). Foveal SRF was independently associated with better VA at CATT year 2². In contrast, at year 5, while eyes with foveal SRF had better mean VA than those without foveal SRF on univariate analysis, this relationship was not significant on multivariate analysis (p=0.14).

Table 3:

Multivariate analysis for the association of optical coherence tomography and fundus photography morphological features with visual acuity at Year 5 (N=513)

Optical Coherence Tomography and Fundus Features at Year 5	N	Adjusted Mean (SE) VA Score in Letters at Year 5	P-value*
Total area of CNV lesion (mm2)			<0.001
>=0 to <=5	112	67.8 (2.0)
>5 to <=10	109	62.5 (1.9)
>10 to <=20	114	61.5 (1.8)
>20	91	55.6 (2.1)
Unknown	87	53.5 (2.5)
Foveal center pathology			<0.001
None	84	61.9 (2.3)
Fibrotic Scar	89	56.4 (2.2)
Geographic atrophy	83	52.6 (2.2)
Non-geographic atrophy	120	65.1 (1.8)
CNV or fluid, hemorrhage	53	61.4 (2.7)
Non-fibrotic scar	26	63.6 (3.8)
Other, CG/CD	58	66.0 (3.0)
Intraretinal fluid present			0.045
No Fluid	204	62.6 (1.4)
Fluid not in foveal center	268	60.6 (1.2)
Fluid in foveal center	41	51.6 (4.0)
Subretinal hyper reflective material present			<0.001
No	176	66.0 (1.6)
Yes, not at foveal center	207	63.7 (1.3)
Yes, at foveal center	130	48.7 (1.9)
Retinal thickness (microns)			<0.001
<120	184	54.2 (1.5)
120-212	270	64.8 (1.2)
>212	59	62.1 (3.2)

CNV = choroidal neovascularization; RPE = retinal pigment epithelium; SE = standard error; VA = visual acuity.

^*P-values were from the multivariate regression models with all these morphological variables in the same model.

Figure 4.

Pathology in the foveal center at A) Baseline, B) Year 1, C) Year 2, D) Year 5

Morphological Associations with Visual Acuity Decline From Year 2 to Year 5

The mean VA declined approximately 2 ETDRS lines between year 2 and year 5 to a level below baseline.⁵ Accordingly, we explored whether adverse pathological features that developed, or worsened from year 2 to year 5 accounted for the decline. We defined adverse features as those morphological characteristics that developed and/or increased in size from year 2 to year 5 and that on multivariate analysis were statistically significantly associated with a 3-line (15 letters) VA worsening of VA. These adverse pathological features included: area of CNV lesion, area of GA, new foveal GA, new foveal scar, new foveal CNV, new SHRM within the center 1 mm, new foveal intraretinal fluid, new retinal thinning, (Table 4, 5 available at www.aaojournal.org). Even among eyes that developed no adverse features, mean VA declined 3 letters (from 73 to 70 letters) from year 2 to year 5. However, when two or more adverse morphological features developed, the mean decrease was approximately 3 lines, from 69 letters to 56 letters (Figure 5). In addition, the mean VA declined more between year 2 to year 5 when there were abnormal features, but the average VA at each time point was worse from year 1 to year 5, as the number of adverse features increased (Figure 6). Eyes that developed GA, scar, or CNV in the foveal center after year 2 had a mean loss greater than three lines between years 2 and 5 while those with foveal pathology already present at 2 years had a mean loss of 1 to 2 lines. (Table 5; available at www.aaojournal.org). Eyes without foveal pathology at 5 years had a mean loss of approximately 1.5 lines.

Table 4:

Multivariate analysis for the association of morphological change with 3 line VA loss or worse between Year 2 and Year 5

		Without ≥3 lines loss from Year 2 to Year 5 (N=362)	With ≥3 lines loss from Year 2 to Year 5 (N=130)	P-value
Foveal center pathology				0.09
	New Fov GA at Year 5	29 (8.0%)	22 (16.9%)
	New Fov Scar at Year 5	17 (4.7%)	15 (11.5%)
	New Fov CNV at Year 5	10 (2.8%)	9 (6.9%)
	Existing Fov GA at Year 2	15 (4.1%)	8 (6.2%)
	Existing Fov Scar at Year 2	34 (9.4%)	21 (16.2%)
	Existing Fov CNV at Year 2	18 (5.0%)	2 (1.5%)
	Other foveal pathology features at year 5	200 (55.2%)	45 (34.6%)
	No foveal pathology at Year 5	39 (10.8%)	8 (6.2%)
Foveal intraretinal fluid				0.03
	No	343 (94.8%)	111 (85.4%)
	Existing at Year 2	10 (2.8%)	6 (4.6%)
	New at Year 5	9 (2.5%)	13 (10.0%)
SHRM at center 1mm				0.01
	No	213 (58.8%)	37 (28.5%)
	Existing at Year 2	85 (23.5%)	60 (46.2%)
	New at Year 5	64 (17.7%)	33 (25.4%)
Retinal thinning				0.009
	No	255 (70.4%)	67 (51.5%)
	Existing at Year 2	48 (13.3%)	36 (27.7%)
	New at Year 5	59 (16.3%)	27 (20.8%)
Change in area of GA between Year 2 and Year 5 (mm2)				0.03
	Size change >2	38 (10.5%)	25 (19.2%)
	Size change ≤2	33 (9.1%)	13 (10.0%)
	No GA at year 2 or 5 or both	288 (79.6%)	88 (67.7%)
	Unknown GA status in Year 2 or 5	3 (0.8%)	4 (3.1%)
Area of GA at Year 2 (mm2)				0.03
	Size >2	32 (8.8%)	23 (17.7%)
	Size ≤2	39 (10.8%)	18 (13.8%)
	No GA at year 2	291 (80.4%)	89 (68.5%)

Table 5:

Multivariate analysis for the association of morphological change with VA change between Year 2 and Year 5 (N=508)

		N	VA change between Year 2 and Year 5 Mean (SE)	P-value
Change in area of CNV lesion between Year 2 and Year 5 (categorical, mm2)				0.04
	≤0	86	−6.7 (1.9)
	>0, ≤2	103	−7.3 (1.8)
	>2, ≤5	87	−10.3 (1.9)
	>5	127	−11.1 (1.6)
	Unknown	105	−13.9 (1.8)
Foveal center pathology				<0.001
	New Fov GA at Year 5	52	−17.3 (2.5)
	New Fov Scar at Year 5	34	−20.2 (3.1)
	New Fov CNV at Year 5	19	−17.7 (4.1)
	Existing Fov GA at Year 2	25	−9.8 (3.6)
	Existing Fov Scar at Year 2	56	−9.5 (2.6)
	Existing Fov CNV at Year 2	21	−6.1 (4.0)
	Other foveal pathology features at year 5	253	−7.7 (1.2)
	No foveal pathology at Year 5	48	−6.8 (2.7)
SHRM at center 1mm				<0.001
	No	255	−6.4 (1.2)
	Existing at Year 2	155	−13.2 (1.6)
	New at Year 5	98	−14.2 (1.8)

Figure 5. Mean visual acuity over time among eyes without morphological features vs. eyes with at least one new adverse morphological features developed between year 2 to year 5.

The adverse features included the following: foveal GA; foveal scar, foveal CNV, SHRM at 1 mm center, foveal intraretinal fluid, and retinal thinning, each that developed after year 2; area of CNV lesion increased by >2 mm² between year 2 and Year 5; change of GA area >2mm² between year 2 and Year 5.

Figure 6. Mean visual acuity over time by groups of eyes defined based on number of new adverse morphological features developed between year 2 to year 5.

The adverse features were the same as those described in Figure 5, and included foveal GA; foveal scar, foveal CNV, SHRM at 1 mm center, foveal intraretinal fluid, and retinal thinning, each that developed after year 2; area of CNV lesion increased by >2 mm² between year 2 and Year 5; change of GA area >2mm² between year 2 and Year 5.

To better understand the effect of foveal GA and fibrosis on this VA decline, we calculated mean VA over time (baseline, years 1, 2 and 5) for eyes with and without GA or fibrosis by year 5, adjusting for all baseline predictors of scar and GA ^11,12 in the multivariate analysis. As is seen in Figure 7, although eyes with GA and/or fibrosis at year 5 had worse VA at baseline, years 1 and 2 compared to eyes without GA or scar by year 5, the VA difference at year 5 widened markedly, with adjusted mean VA 49 letters in eyes with GA or fibrosis, and 66 letters in eyes without this pathology.

Figure 7.

Visual acuity over time with and without foveal GA and fibrosis

Discussion

During year 5 of this study, the strength of the years 1 and 2 associations between VA and morphologic features and quantitative measurements determined on OCT, FA and FP were maintained or strengthened. In particular IRF, SHRM, foveal GA and fibrotic scar, an abnormally thin or thick retina, larger CNV area, and increasing sub-RPE tissue complex thickness were associated with significantly worse VA, whereas eyes with SRF and sub RPE fluid had better VA. Unlike years 1 and 2, when VA was improved or stabilized relative to baseline, VA tended to worsen to below baseline by year 5, coincident with an increased proportion of eyes with abnormally thin retinas, increased lesion size, GA and subretinal fibrotic scar. Furthermore, the number of new adverse pathological features from year 2 to year 5 was associated with worse VA throughout the sudy, and a greater drop in VA between year 2 and year 5.

A key 1 -and 2-year study finding was that IRF, as determined by OCT, had a negative impact on VA at all time points examined. The strength of this association was even greater by year 5. When other potential confounding variables were controlled, foveal IRF was independently associated with worse VA over the entire study duration. The presence of other pathologic features such as GA and fibrotic scar did not worsen the negative impact of IRF on VA. The proportion of eyes with IRF, seen as round hyporeflective spaces on OCT, steadily increased from 45% at year 1 to 50 % at year 2, to 61% at year 5⁵. Despite the increased proportion of eyes with IRF, there was a higher proportion of eyes with retinal thinning (thickness < 120 μm) at year 5 compared to those at years 1 and 2, 36% compared to 21% and 22% at years 1 and 2, respectively. We have previously speculated that some of the hyporeflective cystoid structures seen on OCT that persisted on anti-VEGF therapy was not fluid that leaked from CNV, but, rather, may have represented tissue loss mediated by non-VEGF-driven mechanisms, such as cell death. Our year-5 observations, that there were a higher proportion of eyes with hyporeflective cystic spaces, an increased percentage with abnormally thin retinas and an even stronger negative correlation between IRF fluid and VA, when compared to preceding years, are consistent with this hypothesis.

Eyes with foveal SRF had better VA at year 5 than eyes that did not, an effect that was even more pronounced than it was at year 2. The reason for this association remains unclear, although it has been hypothesized that SRF could serve to protect the photoreceptors from potential toxicity related to direct contact with underlying diseased RPE. Furthermore, the SRF could be a biomarker for CNV that provides trophic support to the overlying retina. Alternatively or perhaps in addition, the SRF might protect photoreceptors from direct infiltrative damage by serving as a fluid buffer between the outer segments and the CNV below it, or the SRF itself may contain neuroprotective substances. It is also conceivable that some of the eyes had associated central serous retinopathy, as part of a pachychoroid syndrome, which may have a better VA outcome¹³. To this point, although it is beyond the scope of this manuscript, we are currently evaluating the relationship between choroidal thickness and morphology in eyes of CATT participants. The beneficial effect on VA disappeared when we adjusted for IRF, SHRM, and total CNV lesion size. These data indicate that SRF is associated with at least one of these variables (IRF absence, SHRM absence or small CNV size), so that once these factors are taken into account, there is no additional association between VA and SRF.

At year 5, in contrast to earlier time points, foveal sub-RPE fluid was associated with better VA on univariate analysis. The reason for better VA in these eyes, many whose sub-RPE fluid reflects a serous pigment epithelial detachment, is unclear. It is possible that in some eyes, sub-RPE fluid reflects type 1 CNV that provides trophic support to the retina. Regardless, one possible conclusion is that in the absence of other signs of active CNV, one may withhold anti-VEGF treatment if sub-RPE fluid does not change from one exam to the next. However, the study was not designed to determine the effect of withholding treatment in eyes with a particular fluid type, since the protocol mandated that all eyes were to be treated when IRF, SRF, or sub-RPE fluid was present. A randomized study to compare the effects of withholding anti-VEGF treatment when there is unchanged subretinal or sub-RPE fluid would be required to determine whether or not these types of persistent fluid should always be treated.

Foveal SHRM was independently associated with worse VA. In fact, this pathology was associated with the worst VA of all single parameters that were studied. Furthermore, when SHRM resolved by year 2, the VA was much better than when it persisted. We have previously shown that SHRM is associated with poor VA, probably because of overlying photoreceptor damage¹⁰. Subretinal hyper-reflective material components typically include CNV, fibrin, fibrosis, blood, and fibrotic scar^10,11,14,15. It is likely that fibrotic scar is the main SHRM component in these eyes with late-stage anti-VEGF-treated CNV lesions¹⁴. However, the study was not designed to correlate the specific location of fibrotic scar as determined by FA and FP, with SHRM, as assessed on SDOCT. To better address this point, our group has undertaken a study to register images obtained on SDOCT, with FP and FA images, so that we can correlate one-to one, the pathology observed on these modalities.

The VA worsened significantly from year 2 to year 5 when new adverse pathological features developed, that included foveal GA, fibrotic scar, CNV, SHRM within the central 1 mm, foveal intraretinal fluid, retinal thinning, an increased CNV lesion size more than 5 mm², or increased GA area >2mm². Furthermore, a greater number of these new adverse pathological features, was associated with worse VA at at all time points from baseline, and a greater VA drop between year 2 and year 5, respectively. The presence of these adverse features goes a long way to explain why VA declined from year to year 5 of the study. Unfortunately, recent phase 3 combination therapy interventional studies designed to try to improve VA by targeting PDGF-related pathology such as fibrosis and SHRM failed to meet their therapeutic endpoints¹⁶. Accordingly, there remains a significant unmet need to develop treatments that can limit scar formation and that can prevent GA development. Furthermore, our data to show progressive retinal thinning, outer retinal tubulation indicative of photoreceptor degeneration, and intraretinal hyporeflective cavitary spaces, all point to an unmet to develop neuroprotective strategies to accompany anti-VEGF treatment. Finally, the independent association of increased lesion size with worse VA highlights the need to develop treatments that limit lesion size, and not just specific lesion components.

Although we have highlighted pathological features that were associated with the VA decline from year 2 to year 5, these associations do not tell the whole story. Even among eyes without any adverse pathology at year 5, the VA still declined by 3 letters from year 2 to year 5. Clearly, there are factors that we not yet been identified to account for this observation. For example synaptic re-organization associated with the underlying disease, and lost neuroprotective effects induced by anti-VEGF therapy might play a role in the VA decline, but may not have been detected by the imaging modalties used in this study¹⁷. Further investigations that focus on reasons for the VA decline despite a lack of obvious pathology seen on OCT, FP, or FA are warranted.

In the present study, we conducted several novel analyses to determine anatomical correlates to visual acuity decline with long-term anti-VEGF therapy. Expert readers evaluated images in a standardized manner to determine the impact of a variety of anatomic features on final VA, and to assess the effect of change in morphological features with change in visual acuity over time, to help explain the VA decline from year 2 to year 5. Furthermore, we determined how the number of adverse anatomic features affected VA. Previous studies have examined anatomic factors that correlate with the final VA in an effort to explain the VA decline over time with long-term anti-VEGF therapy. For example, Gillies and colleagues reported that fibrosis and atrophy, as reported by the treating ophthalmologist, might account for decreased VA after 7 years of anti-VEGF therapy⁶. However, in CATT we were able to delve more deeply into the causes of visual acuity loss over time. Color photographs, fluorescein angiograms, and OCTs were each analyzed at several time points, which allowed us to evaluate important morphological characteristics such as lesion growth and lesion size, intraretinal and intraretinal and subretinal fluid and the relationship between changes in morphology with changes in visual acuity and the relative contribution of the different morphological features to visual acuity changes. In contrast, the Gillies study design, in which morphology -visual acuity correlations depended on the treating ophthalmologists’ reports, and which were obtained only at a single time point, precluded these types of analyses. In the Seven-Up study, anatomic features at final follow-up were also correlated with VA⁷. The area of macular atrophy as assessed on FAF images was associated with VA, but not subretinal fibrosis, as determined by fundus photography. However, in that study, only a small number of subjects were studied (65 of 155 eligible subjects), and the study may have been underpowered to detect anatomic correlates such as subretinal fibrosis. Furthermore, the study only assessed VA-morphological correlates at the study endpoint but did not analyze changes in morphology to explain the drop in visual acuity that occurred after the initial VA improvement.

There has been much debate about whether retina specialists under-treat with anti-VEGF therapy in a real world setting, and whether undertreatment could account for the observed VA decline after 2 years of treatment in CATT and other studies^7,18,19. There are undoubtedly some patients who do not receive sufficient anti-VEGF treatment, as evidenced by persistent VEGF-driven pathology such as intraretinal fluid that we observed at years 2 and year 5, and the slightly worse average VA (2.3 letters) seen at year 2 among eyes treated PRN when compared to those treated monthly. Alternative approaches to maximize physician practice efficiency and patient compliance, such as anti-VEGF sustained drug delivery systems would likely help in this regard. However, there are many patients who do receive aggressive treatment and it is clear that undertreatment cannot and does not account for all of the VA decline that we observed between years 2 and 5. First, during the first 2 years of CATT, a significant proportion of eyes developed visually adverse pathology including foveal GA, fibrotic scar, retinal thinning, and lesion growth despite monthly anti-VEGF treatment for 2 years, and the rate of GA was higher among eyes that received a greater number of injections^2,3. Accordingly, more injections do not necessarily translate to prevention of VA loss in some cases. Second, there is no evidence that additional anti-VEGF therapy will prevent GA, and the development of GA or expansion from existing GA was a major contributor to VA decline between years 2 and 5. Finally, there was a large number of eyes that retained excellent vision between years 2 and 5 despite receiving no additional injections during that time. As such, the ideal number of anti-VEGF injections given over many years that will yield an optimal VA result is unknown and likely varies greatly among patients.

We identified morphological features that developed or worsened after year 2 of CATT that help to explain the pronounced visual acuity decline from years 2-5, and which highlight an unmet need for specific new treatments.