Spectral-Domain Optical Coherence Tomography as an Indicator of Fluorescein Angiography Leakage from Choroidal Neovascularization

Abstract

Purpose.: To evaluate spectral-domain optical coherence tomography (SD-OCT) findings that predict angiographic leakage in choroidal neovascularization (CNV).

Methods.: SD-OCT and fluorescein angiography (FA) images of 93 eyes of 93 patients were retrospectively analyzed. All patients were previously treated with anti–vascular endothelial growth factor agents for CNV from age-related macular degeneration. FA images were analyzed to assess the presence of leakage. SD-OCT images were analyzed to identify the overall presence of fluid, as well as specific patterns of fluid presentation, including intraretinal cystic spaces (ICS), retinal pigment epithelium detachment (PED), and neurosensory detachment (NSD). The presence of ultrastructural features such as intraretinal hyperreflective flecks and the inherent reflectivity and boundary definition of the subretinal material were evaluated. Both the association and the sensitivity, specificity, and both positive and negative predictive values of SD-OCT findings compared with FA leakage were calculated.

Results.: A statistically significant association between SD-OCT findings and FA leakage was found for eyes that displayed fluid, NSD, intraretinal flecks, and low reflectivity or undefined boundaries from subretinal material, and not for PED or ICS. Sensitivity and specificity for SD-OCT findings were, respectively: 94% and 27% for fluid; 68% and 88% for NSD; 81% and 83% for intraretinal flecks; 63% and 92% for undefined boundaries of subretinal material; and 94% and 87% for low reflectivity from subretinal material.

Conclusions.: The evidence of fluid on SD-OCT is sensitive but nonspecific in identifying FA leaky CNV. The assessment of neurosensory detachment as well as other ultrastructural elements may increase the specificity of analysis.

Age-related macular degeneration (AMD) is the leading cause of blindness in people aged 50 years or older in developed countries. ^1,2 In the neovascular form, fluorescein angiography (FA) has been instrumental in the diagnostic and treatment decision-making guidelines arising from multiple clinical trials, ^{3 –6} and is particularly useful in characterizing neovascular activity through the dynamic visualization of dye leakage. 

More recently, optical coherence tomography (OCT) ⁷ has become a fundamental tool in the diagnosis and follow-up of this disease. ⁸ The pivotal minimally classic/occult trial of the anti–vascular endothelial growth factor (VEGF) antibody ranibizumab in the treatment of neovascular AMD and the anti-VEGF antibody for the treatment of predominantly classic choroidal neovascularization (CNV) in AMD trials established the role of monthly ranibizumab in the treatment of neovascular AMD. ^9,10 In clinical practice, though, a significant proportion of retina specialists use a variable dosing regimen that uses OCT to help determine which lesions may benefit from treatment with an anti-VEGF agent. In the prospective OCT imaging of patients with neovascular AMD treated with intraocular ranibizumab (PrONTO) study, evidence or persistence of fluid as seen by OCT was considered a sufficient element for retreatment. ¹¹ In this trial, as well as in the study of ranibizumab in patients with subfoveal choroidal neovascularization secondary to AMD and the safety assessment of intravitreal ranibizumab (Lucentis) for AMD studies, ¹² an increase of >100 μm in central retinal thickness as measured by OCT was also considered a sign of needed retreatment. In all these studies, the OCT parameter most often used to identify lesions that may benefit from anti-VEGF treatment was the presence of fluid visualized in OCT images or assessed using retinal thickness measurements. 

Despite the wide acceptance of this treatment strategy, the strength and reliability of OCT as an independent modality to accurately determine CNV lesion leakage is not well understood. This was acknowledged in the PrONTO study, where another, independent retreatment parameter was the FA evidence of new classic CNV. ¹¹  

In this study, to better assess the role of OCT in predicting angiographic leakage status, OCT images were compared in lesions that displayed leakage by FA with lesions that did not show leakage by FA. Furthermore, different patterns of fluid presentation as well as “ultrastructural” elements were assessed in OCT images: these features include the presence of intraretinal hyperreflective flecks, which were previously described in nonneovascular AMD, ¹³ and the intrinsic boundary definition and reflectivity of CNV material, which have been suggested to be indicators of CNV evolution toward a fibrotic stage. ¹⁴  

Methods

Sample Collection

A cross-sectional, retrospective study of FA and OCT images on 93 eyes of 93 consecutive patients with CNV from neovascular AMD was conducted at the Eye Clinic, Sacco Hospital, University of Milan. The study was approved by the local institutional review board and adhered to the tenets of the Declaration of Helsinki. 

Eligible eyes were identified retrospectively through a database search, and consent from enrolled patients was obtained. Inclusion criteria were: clinical history of AMD and FA diagnosis of subfoveal CNV, ¹⁵ FA and SD-OCT imaging performed using a commercial imaging device (Heidelberg HRA + OCT Spectralis; Heidelberg Engineering, Heidelberg, Germany); previous treatment with an anti-VEGF agent (ranibizumab or bevacizumab) for CNV; and all the analyzed FA and SD-OCT examinations had to be scheduled and performed following the regular, standard practice of the clinic (SD-OCT + FA are routinely acquired 1 month after any anti-VEGF agent injection and every 3 months thereafter). Exclusion criteria were: any previous laser treatment, photodynamic therapy, or vitreoretinal surgery on the study eye; significant macular hemorrhage that obscured the lesion; and a spherical refractive error >6 diopters. 

FA Evaluation

The dynamic video FA (first 30 seconds) and FA images from the early and late phases of the study were analyzed by two observers (GA, MC). A determination was made by each observer regarding the type of CNV lesion (CNV with a classic component or occult) present and the leakage status of each lesion. The FA data sets provided to the observers for examination were given a numerical identifier, and examiners were masked from any other patient data including SD-OCT images. In accordance with the modified Macular Photocoagulation Study grading retreatment criteria used in the Treatment of AMD with Photodynamic Therapy and Verteporfin in Photodynamic Therapy studies, ¹⁵ lesions were divided into two subgroups: CNV showing FA evidence of leakage or lesions that did not show leakage. Cases of disagreement between observers regarding lesion activity were defined as questionable. In these cases, input from a third observer was used to reach an open consensus. 

After FA grading was completed, the observers were then asked to select SD-OCT B-scans passing through the CNV lesion, as shown by FA images, and to specify which ones were localized on the classic component of the lesions (if present). 

SD-OCT Evaluation

All the analyzed SD-OCT images were acquired using the same volume protocol using the high-speed modality (768 A-scans/B-scan), and included a 30° scanned area and 19 B-scans per data set. All the B-scans were produced from a real-time image averaging process to reduce speckle noise, and the number of individual frames per B-scan was variable, ranging from 16 to 63. 

All the B-scan images from the SD-OCT volume data sets were analyzed by two different examiners (AG, CL) that were masked to all other patient data including FA grading and color photography. Each examiner evaluated every B-scan in each data set for the presence or absence of the following characteristics: intraretinal cystic spaces (ICS), without differentiation in retinal layer localization, content, or number/density; retinal pigment epithelium (RPE) detachment (PED), defined as a localized elevation of RPE due to fluid or fibrovascular tissue; and neurosensory retinal detachment (NSD), defined as a fluid detachment of retinal layers from the RPE. An eye was labeled as having “fluid” present by the finding of at least one of these parameters. In addition, observers were asked to analyze B-scans passing through the FA images of the CNV lesions that were previously selected by FA graders. For these B-scans, SD-OCT graders used standard images (Fig. 1F1) to assess for the presence or absence of intraretinal hyperreflective flecks, defined as the presence of multiple, hyperreflective, discrete spots of approximately 20 to 30 μm or less, often localized to the boundary of neurosensory detachment (Fig. 1A, black arrows). Lesions showing a classic component on FA were analyzed to identify the following ultrastructural SD-OCT features: poorly defined boundaries of subretinal lesion material, described as an ill-defined and hardly discernible interface of the subretinal material from the surrounding retinal tissue (Fig. 1B, white arrow); and low lesion optical reflectivity, defined as the mean white color intensity of the lesion (Fig. 1C, asterisk) compared with the reflectivity of the nearby, uninvolved RPE (Fig. 1C, black arrow). Once again, examiners were masked from all other patient data including FA images when evaluating for these SD-OCT features. 

Figure 1.

 

SD-OCT ultrastructural features in choroidal neovascularization: standard images used for the SD-OCT–based classification. (A) Intraretinal hyperreflective flecks, defined as multiple, hyperreflective, discrete spots, often localized to the boundary of neurosensory detachment (black arrows). (B) Poorly defined boundaries of the lesion material, defined as a hardly discernible interface of the subretinal material from the surrounding retinal tissue (white arrow). (C) Low lesion optical reflectivity, defined as the mean white color intensity of the lesion (*) compared with the reflectivity of the nearby, uninvolved RPE (black arrow).

Statistical Analysis

Statistical analysis was performed using a statistical software package (Statgraphics ver. 5.1; Statistical Graphics Corp., Herndon, VA) and R language for statistical computing (R Development Core Team; provided in the public domain by R Foundation for Statistical Computing, Vienna, Austria; available at: http://www.r-project.org/, accessed April 20, 2011). Analysis of FA and SD-OCT images was repeated a second time after 1 week, and both intra- and interobserver variability were assessed with Cohen's kappa (κ) test. The association between FA presence of leakage and the different OCT parameters in the study was assessed with Fisher's Exact Test. P < 0.05 was considered a statistically significant association. 

For SD-OCT parameters that were found to be associated with FA presence of leakage, the sensitivity (SE), specificity (SP), and both positive (PPVs) and negative predictive values (NPVs) were calculated by using FA leakage as the gold standard. This analysis was repeated after excluding FA questionable cases, defined as those in which the reviewers disagreed on the leakage status and a third observer was used to reach an open consensus. The studied population was also stratified based on the median number of anti-VEGF agent injections received and on the median number of months from diagnosis of CNV. SE, SP, and both PPV and NPV were recalculated after this stratification. 

The association between the different variables evaluated in this study was further examined by multifactorial regression analysis using a generalized linear model for binomial variables (logit). This analysis was computed by comparing the presence of FA leakage (independent variable) with the evidence of SD-OCT features, the number of anti-VEGF injections, and the duration of the pathology. The correlation of these different variables was tested using a phi (φ) correlation analysis. 

Results

A total of 93 eyes from 93 eligible and consecutive patients with CNV from AMD were evaluated. Sample characteristics are summarized in Table 1. There was good intra- and interobserver agreement in the evaluation of FA leakage as well as in SD-OCT parameter grading (Table 2). 

Table 1.

 

Sample Composition and Demographic Characteristics

Characteristic	Lesions with Classic Component			Occult Lesions			Entire Sample
	FA Leakage	FA No Leakage	Total	FA Leakage	FA No Leakage	Total	FA Leakage	FA No Leakage	Total
Number	33	24	57	19	17	36	52	41	93
Sex, M:F	16:17	9:15	25:32	7:12	9:8	16:20	23:29	18:23	41:52
Mean age, y	76.4 (10.7)	78.2 (10.5)	77.2 (10.5)	76.0 (12.1)	77.4 (13.9)	76.6 (12.8)	76.3 (11.1)	77.9 (11.1)	77.0 (11.4)
BCVA	0.37 (0.25)	0.46 (0.25)	0.41 (0.25	0.39 (0.28)	0.37 (0.25)	0.38 (0.26)	0.38 (0.26)	0.42 (0.25)	0.40 (0.25)
Number of anti-VEGF Treatment	7.2 (3.9)	6.5 (2.8)	6.9 (3.5)	6.8 (3.7)	6.4 (3.7)	6.5 (2.8)	7.1 (3.8)	6.3 (3.1)	6.7 (3.5)
Months from last anti-VEGF treatment	3.8 (3.1)	7.1 (4.9)	5.2 (4.3)	4.5 (2.6)	5.7 (3.2)	5.1 (2.9)	4.0 (2.9)	6.5 (4.3)	5.1 (3.8)
Months from diagnosis of CNV pathology	12.4 (7.6)	13.8 (9.8)	13.0 (8.5)	14.4 (8.9)	14.2 (8.5)	14.3 (8.6)	13.1 (8.0)	14.0 (9.1)	13.5 (8.5)

 Data are shown as mean ± SD. BCVA, best-corrected visual acuity.

Table 2.

 

Agreement in FA Leakage and Spectral-Domain Optical Coherence Tomography Parameters Grading (Cohen's kappa)

Parameter	Lesions with Classic Component		Occult Lesions		Entire Sample
	Intra-observer	Inter-observer	Intra-observer	Inter-observer	Intra-observer	Inter-observer
FA leakage	0.89	0.85	0.72	0.67	0.82	0.78
PED	0.84	0.78	0.79	0.92	0.92	0.93
NSD	0.85	0.96	0.83	0.94	0.84	0.95
ICS	0.75	0.92	0.72	0.94	0.74	0.94
Flecks	0.75	0.89	0.77	0.88	0.76	0.89
Undefined boundaries	0.78	0.88	n.a.	n.a.
Low reflectivity	0.78	0.92	n.a.	n.a.

 n.a., not available.

The presence of SD-OCT parameters in FA leaky and nonleaky lesions is shown in Table 3. Overall, SD-OCT evidence of fluid was present in 94% (49/52) of leaky lesions, compared with 73% (30/41) of nonleaky lesions (P = 0.007). In the occult lesion subgroup though, SD-OCT evidence of fluid was seen in both leaky lesions (100%, 19/19) and nonleaky lesions (88%, 15/17), leading to a nonsignificant association. SD-OCT evidence of PED was not associated with FA leakage status in the entire sample, as well as in the lesion type subgroups. It should be noted that in classic lesions PED was rare in both FA leaky and nonleaky lesions (15% [5/33] and 4% [1/24], respectively), whereas in occult CNV it was frequently found in both FA leaky and nonleaky lesions (79% [15/19] and 71% [12/17], respectively). NSD evidence in SD-OCT was strongly associated with FA leakage in the entire sample and in the CNV type subgroups. In particular, NSD was found in only 12% (5/41) of the FA nonleaky lesions, whereas 67% (35/52) of FA leaky CNV lesions showed this finding. ICS evidence in SD-OCT was not associated with the FA leakage status and was equally present in leaky and nonleaky lesions (52% [27/52] in FA leaky lesions vs. 56% [23/41] in nonleaky CNV). 

Table 3.

 

Association between Evidence of Fluorescein Angiography Activity and the Presence of Spectral-Domain Optical Coherence Tomography Parameters

Parameter	Lesions with Classic Component			Occult Lesions			Entire Sample
	FA Leakage	FA No Leakage	P *	FA Leakage	FA No Leakage	P *	FA Leakage	FA No Leakage	P *
Fluid	30/33	15/24	0.0187	19/19	15/17	0.2159	49/52	30/41	0.0074
PED	5/33	1/24	0.3845	15/19	12/17	0.7060	20/52	13/41	0.5216
NSD	20/33	2/24	<0.0001	15/19	3/17	0.0006	35/52	5/41	<0.0001
ICS	17/33	17/24	0.1777	10/19	6/17	0.3351	27/52	23/41	0.8343
Flecks	24/33	2/24	<0.0001	18/19	5/17	<0.0001	42/52	70/41	<0.0001
Undefined boundaries	21/33	2/24	<0.0001	n.a.	n.a.
Low reflectivity	31/33	3/24	<0.0001	n.a.	n.a.

 n.a., not available.

*  Fisher's Exact Test.

SD-OCT evidence of intraretinal hyperreflective flecks was strongly associated with FA leakage status (Fig. 2). Remarkably, 95% (18/19) of the occult FA leaky lesions showed this finding. The analysis of SD-OCT characteristics of subretinal material in classic lesions revealed that the presence of undefined boundaries and low reflectivity were both associated with the FA leakage status. In particular, 94% (31/33) of the FA leaky lesions showed low reflectance from subretinal material. The three cases of patients showing FA leaky lesions without evidence of fluid on SD-OCT are displayed in Figures 2C–E. In these cases, one showed all three ultrastructural findings (intraretinal flecks, subretinal material with a poorly defined boundary, and low optical reflectivity), one showed intraretinal flecks, and the last one showed subretinal material with an undefined boundary and low reflectivity, but with no evidence of intraretinal flecks. 

Figure 2.

 

SD-OCT features in angiographically leaky and nonleaky lesions. (A) A case of occult CNV that shows leakage at FA (left). SD-OCT (right) shows neurosensory detachment and intraretinal cystic spaces, as well as intraretinal hyperreflective flecks (black arrow). (B) A case of an FA nonleaky fibrotic CNV lesion that displays in SD-OCT highly reflective subretinal material (white arrow) with well-defined borders and intraretinal cystic spaces in the overlying retina. (C, D) Three cases of FA leaky lesions with no evidence of fluid in SD-OCT images. In the first case (C) the SD-OCT image shows intraretinal hyperreflective flecks (black arrows) and subretinal material with low reflectivity and undefined boundary (white arrow). In the second case (D) the SD-OCT image shows only undefined boundaries and low reflectivity from subretinal material (white arrow) and no intraretinal flecks. In the third case (E) the SD-OCT image shows some intraretinal flecks (black arrow).

The SE, SP, and both PPVs and NPVs were calculated for the presence of fluid, NSD, intraretinal flecks, and subretinal material characteristics using FA leakage as the gold standard (Table 4). Over the entire sample, SD-OCT evidence of fluid showed a high sensitivity (94%), but very low specificity (27%). This result was even more pronounced in the occult subgroup (100% vs. 12%, respectively). In contrast, NSD evidence had a high specificity (88%) and modest sensitivity (67%). The presence of intraretinal flecks had high sensitivity and specificity values (81% and 83%, respectively) over the entire sample. This was likely due to a high specificity (92%) in classic lesions and high sensitivity (95%) in occult CNV. Considering subretinal material characteristics, evidence of low reflectivity had high sensitivity and specificity values (94% and 87%). 

Table 4.

 

Sensitivity, Specificity, Positive and Negative Predictive Values for Spectral-Domain Optical Coherence Tomography Parameters Compared with Fluorescein Angiography Activity

Parameter	Lesions with Classic Component				Occult Lesions				Entire Sample
	SE (95% CI)	SP (95% CI)	PPV (95% CI)	NPV (95% CI)	SE (95% CI)	SP (95% CI)	PPV (95% CI)	NPV (95% CI)	SE (95% CI)	SP (95% CI)	PPV (95% CI)	NPV (95% CI)
Fluid	91% (76–98)	37% (19–59)	67% (51–80)	75% (43–95)	100% (79–100)	12% (2–38)	56% (38–72)	100% (20–100)	94% (83–98)	27% (15–43)	62% (50–73)	79% (49–94)
NSD	61% (42–77)	92% (72–99)	91% (70–98)	63% (45–78)	79% (54–93)	82% (56–96)	83% (58–96)	78% (52–93)	67% (53–79)	88% (73–95)	87% (72–96)	68% (53–80)
Flecks	73% (54–86)	92% (72–99)	92% (73–99)	71% (52–85)	95% (72–99)	71% (44–89)	78% (56–92)	92% (62–99)	81% (67–90)	83% (67–92)	86% (72–94)	77% (62–88)
Undefined boundaries	63% (45–79)	92% (75–99)	91% (70–98)	65% (46–80)	n.a.	n.a.	n.a.	n.a.
Low reflectivity	94% (78–99)	87% (67–97)	91% (75–98)	91% (70–98)	n.a.	n.a.	n.a.	n.a.

 Fluid was considered present if either neurosensory detachment, pigment epithelium detachment, or intraretinal cystic spaces were present.

 CI, confidence interval; n.a., not available.

When this analysis was repeated after excluding the cases with questionable FA leakage, there was a slight increase in specificity in lesions with a classic component (41% vs. 37%) and in the entire sample set (30 vs. 27%) (Table 4, Supplementary Table S1). 

In the population subset that had a number of anti-VEGF agent injections less than the median (seven treatments), the presence of fluid on SD-OCT showed higher sensitivity (100% vs. 89%) but less specificity (18% vs. 37%) in detecting FA leakage compared with lesions that had seven or more treatments (Supplementary Tables S2 and S3). The presence of NSD and intraretinal flecks on SD-OCT showed a higher specificity (91% vs. 84% and 86% vs. 79%, respectively) in detecting FA leakage in these patients (Supplementary Tables S2 and S3). 

In the analysis of the population subset that had a duration of CNV pathology less than the median time (10 months), there was a slight increase in the sensitivity of fluid on SD-OCT to detect FA leakage, particularly in occult lesions (17% vs. 9%), compared with lesions with a greater duration (≥10 months) (Supplementary Tables S4 and S5). The ultrastructural features seen with SD-OCT showed, in general, higher sensitivity and specificity in detecting FA leakage when the CNV diagnosis was more recent (Supplementary Tables S4 and S5). 

A multifactorial regression analysis was used to test for an association between the presence of FA leakage and the different SD-OCT variables when considered together. In lesions with a classic component (Table 5), statistically significant results for intraretinal flecks (P = 0.036) and low reflectivity of subretinal material (P = 0.010) were found. All the other SD-OCT features, including the presence of fluid, neurosensory detachment, and undefined boundaries of subretinal material, did not show a statistically significant result (P > 0.05). The number of anti-VEGF treatments and the duration of the pathology did not influence these results (P > 0.05). 

Table 5.

 

Multifactorial Regression Analysis with the Presence of Leakage by Fluorescein Angiography as Independent Variable

Parameter	Lesions with Classic Component		Occult Lesions
	z Value	P	z Value	P
Fluid	0.235	0.814	0.004	0.996
PED	−0.225	0.822	1.058	0.290
NSD	0.607	0.544	1.420	0.152
ICS	−0.808	0.418	1.580	0.114
Flecks	2.092	0.035	2.278	0.023
Undefined boundaries	−0.161	0.872	n.a.	n.a.
Low reflectivity	2.552	0.010	n.a.	n.a.
Number of anti-VEGF*	−0.566	0.571	1.049	0.294
Time from diagnosis†	0.097	0.923	0.134	0.893

 Fluid was considered present if either neurosensory detachment, pigment epithelium detachment, or intraretinal cystic spaces were present. n.a., not available.

*  Seven or more anti-VEGF injections.

†  Ten months or more from choroidal neovascularization diagnosis.

When the multifactorial regression analysis was repeated for the occult lesions (Table 5), we found a statistically significant result for intraretinal flecks (P = 0.022). All the other SD-OCT features studied, including the presence of neurosensory detachment, did not show a statistically significant result (P > 0.05). Again, the number of anti-VEGF treatments and the duration of the pathology did not influence these results (P > 0.05). 

The phi coefficients of correlation for lesions with a classic component (Table 6) showed a good correlation between FA leakage and intraretinal flecks (φ = 0.64) and low reflectivity of subretinal material (φ = 0.82). Evidence of neurosensory detachment showed a substantial correlation with the presence of FA leakage (φ = 0.53), intraretinal flecks (φ = 0.5), and low reflectivity of subretinal material (φ = 0.51). The presence of undefined boundaries of the subretinal material showed a substantial correlation with the presence of FA leakage (φ = 0.56) and low reflectivity of subretinal material (φ = 0.68), and a moderate correlation with the evidence of intraretinal flecks (φ = 0.47). 

Table 6.

 

Phi Coefficients of Correlation (φ) of the Different Study Parameters in Lesions with a Classic Component

Parameter	FA Leakage	Fluid	PED	NSD	ICS	Flecks	Undefined Boundaries	Low Reflectivity	Number of Anti-VEGF*	Time from Diagnosis†
FA leakage	1.00
Fluid	0.34	1.00
PED	0.18	0.04	1.00
NSD	0.53	0.32	0.20	1.00
ICS	−0.19	0.36	−0.30	−0.38	1.00
Flecks	0.64	0.30	0.26	0.50	−0.18	1.00
Undefined boundaries	0.56	0.25	0.30	0.23	−0.05	0.47	1.00
Low reflectivity	0.82	0.45	0.17	0.51	−0.09	0.54	0.68	1.00
Number of anti-VEGF*	0.04	−0.15	0.33	0.10	−0.14	0.16	−0.01	0.01	1.00
Time from diagnosis†	−0.06	0.01	−0.06	0.05	0.12	−0.14	−0.19	−0.02	0.03	1.00

 Fluid was considered present if either neurosensory detachment, pigment epithelium detachment, or intraretinal cystic spaces were present.

*  Seven or more anti-VEGF injections.

†  Ten months or more from choroidal neovascularization diagnosis.

The phi coefficients of correlation for occult lesions (Table 7) showed a good correlation between FA leakage and intraretinal flecks (φ = 0.68), and neurosensory detachment (φ = 0.61). Moreover, intraretinal flecks and neurosensory detachment were well correlated with each other (φ = 0.64). 

Table 7.

 

Phi Coefficients of Correlation (φ) of the Different Study Parameters in Occult Lesions

Parameter	FA Leakage	Fluid	PED	NSD	ICS	Flecks	Number of Anti-VEGF*	Time from Diagnosis†
FA leakage	1.00
Fluid	0.26	1.00
PED	0.10	0.42	1.00
NSD	0.61	0.24	0.06	1.00
ICS	0.17	0.22	0.01	−0.11	1.00
Flecks	0.68	0.07	−0.17	0.64	−0.03	1.00
Number of anti-VEGF*	0.11	−0.26	0.03	0.17	−0.06	0.02	1.00
Time from diagnosis†	0.01	−0.01	0.03	−0.06	−0.06	0.02	−0.23	1.00

 Fluid was considered present if either neurosensory detachment, pigment epithelium detachment, or intraretinal cystic spaces were present.

*  Seven or more anti-VEGF injections.

†  Ten months or more from choroidal neovascularization diagnosis.

Discussion

Choroidal neovascularization in AMD is one of the most common causes of central vision loss in Western countries. ¹⁶ Both evaluating treatment approaches with currently available drugs and determining the efficacy of new drugs in clinical trials rely heavily on the use of imaging modalities such as FA and OCT. In the OCT era, the strategy of “OCT presence of fluid = need for treatment” has rapidly gained popularity among the ophthalmologic community. ¹¹ Largely because SD-OCT is a fast, reliable, noninvasive tool that can easily detect the presence of fluid, this modality has gradually overtaken the FA assessment of leakage as a means to assess lesion activity. Since data are lacking regarding the long-term visual acuity impact of SD-OCT versus FA approaches, it would be of interest to study the interchangeability of these two diagnostic evaluations. 

The results from this study show that, in general, evidence of fluid by SD-OCT is sensitive in detecting FA leaky lesions (94%), but it has a very low specificity (27%). The specificity slightly increases when considering patients that have had a fewer number of anti-VEGF injections or that have had a more recent diagnosis of CNV, but the overall value still remains low. These results appear to be in accordance with a previous report by Khurana et al., ¹⁷ who reported a high sensitivity (90%) and a low specificity (47%) for a similar analysis. In our study, this outcome is evident for all CNV subtypes and is most dramatic for occult lesions (100% sensitivity, 12% specificity). 

Examining specific patterns of fluid accumulation can significantly affect the specificity of SD-OCT evaluation with regard to having an FA leaky lesion. The presence of a neurosensory detachment showed a very strong association and high specificity as indicators of an angiographically leaky lesion. In contrast, the presence of a PED or intraretinal cystic spaces was not associated with FA evidence of leakage. Specifically, in occult lesions we found a high frequency of PED in both leaky and nonleaky lesions. Similarly, when taking the entire sample into consideration, intraretinal cystic spaces were found equally in both leaky and nonleaky CNV. Although the former is not surprising and in line with previous reports, ¹⁸ the frequent evidence of intraretinal cystic spaces in nonleaky lesions seems to be in contrast with previous studies that considered the appearance of macular cysts to represent the earliest manifestations of recurrent CNV ¹¹ ; however, this observation is not without precedence because Liakopoulos et al. ¹⁹ reported that, in some cases, CNV angiographic activity was not associated with intraretinal cystic spaces on OCT. Furthermore, they hypothesized that retinal degeneration in old or chronic CNV lesions may be associated with the loss of neuronal retinal tissue and formation of cystic spaces that do not exhibit leakage on FA. Zweifel et al. ²⁰ described a possible arrangement of degenerative photoreceptors in a circular or ovoid conformation, common to several advanced pathologies affecting the outer retina and the RPE, and labeled this as outer retinal tubulation. The authors suggested that these findings may be misinterpreted as intraretinal or subretinal fluid, possibly prompting unnecessary interventions. In the present study, we did not differentiate between cystic spaces in the retina from outer retinal tabulation; however, we were careful to distinguish between the terms intraretinal cystic spaces and cystic edema or cystic “fluid,” and suggest that the presence of such spaces do not necessarily mark the presence of edema that may accompany angiographically leaky CNV. 

To further enhance the specificity and the sensitivity of SD-OCT to detect lesions that may show leakage with FA, we evaluated other “ultrastructural” features. Due to improvements in axial resolution and acquisition speed, SD-OCT instruments have minimized limitations seen in time-domain systems ²¹ that were used in previous anti-VEGF trials. ^{9 –12} This particular system allowed for image averaging that reduces speckle noise and increases image quality, as well as simultaneous acquisition of FA and SD-OCT images that permitted point-to-point OCT evaluation of a particular angiographic finding. Using this last feature, SD-OCT B-scan images that precisely corresponded to areas of CNV as identified by FA could be evaluated. 

Intraretinal hyperreflective flecks showed a very strong association with leakage detected by FA. Both sensitivity and specificity of this parameter were high. The histopathology underlying this correlation is unknown. Schuman et al. ¹³ studied the association of this feature with drusen, and hypothesized that the origin of this alteration could be due to deposits of pigment or from the migration of macrophages that have engulfed degenerative material from the synaptic terminals of damaged photoreceptors. The increased density of the flecks found in this study, compared with that observed by Schuman and colleagues, could be due to increased pigment or inflammatory cell migration that may be associated with CNV. Similar findings were found by Ahlers et al. ²² in central serous chorioretinopathy. The authors speculated that these deposits may represent inflammatory debris or residual rhodopsin and outer cone segments. Bolz et al. ²³ described hyperreflective foci in diabetic macular edema, and hypothesized that they may correspond to lipid or lipoprotein deposits, lipid-laden macrophages, or a combination of both. 

The appearance of subretinal material on SD-OCT could represent another element to predict whether a lesion that has a classic component shows leakage on FA. Classic CNV lesions, unlike occult ones, exist in the subretinal space and thus allow for more direct evaluation by SD-OCT because there is no signal attenuation from the RPE. In this study, the evaluation of the intrinsic reflectivity of subretinal material was qualitative. Moreover, any qualitative evaluation of reflectivity may be affected by media opacities lying above the lesion. Despite this limitation, lesions that displayed low intrinsic reflectivity of this material showed both high sensitivity and specificity (94% and 87%, respectively). Although an enhancement in subretinal material reflectivity by OCT has been described as a possible sign of fibrous tissue ingrowth from the CNV complex, ¹⁴ as suggested by Liakopoulos et al., ¹⁹ at this time it is not possible to distinguish if the origin of the subretinal material is from only the neovascular membrane because other components such as lipid, fibrin, and other unspecified material may be present. Furthermore, the biological basis of this ultrastructural finding is unknown, although it may be explained by an influx of inflammatory cells or fluid that can accompany active lesions. A possible role of inflammation in all stages of AMD, ^24,25 including the breakdown of Bruch's membrane ²⁶ and the influx of inflammatory cells such as macrophages, ²⁷ has been suggested. 

In clinical practice, the SD-OCT features evaluated in this study are often found together in the same case. The results of the multifactorial regression analysis suggest that, in such cases, the evidence of intraretinal flecks may be the most important feature in predicting the presence of FA leakage. Moreover, in lesions with a classic component, the low reflectivity of subretinal material is another important feature that may help predict angiographic leakage. In this analysis, other parameters such as a neurosensory detachment, which showed a good correlation with FA leakage by univariate analysis (Fisher's Exact Test), did not show an association with the presence of FA leakage. These results may be better explained by analyzing the phi coefficients of correlation. In this analysis, the presence of fluid, neurosensory detachment, and undefined boundaries of subretinal material showed a substantial or good correlation with the evidence of intraretinal flecks and low reflectivity of subretinal material, or both. This may explain why when the parameters are analyzed together in a multifactorial regression model, the features that have a lower association with the presence of FA leakage (evidence of fluid, neurosensory detachment, and undefined boundaries of subretinal material) apparently lose their predictive value compared with that of the other variables (evidence of intraretinal flecks and low reflectivity of subretinal material) that are more strongly associated with the presence of FA leakage. Interestingly, in the three cases of patients with FA leaky CNV that did not show evidence of fluid by SD-OCT, each lesion showed intraretinal flecks and/or low reflective subretinal material. This suggests that these two features may also be of particular value in predicting leaky CNV lesions that do not show evidence of traditional fluid on SD-OCT. 

This study has several limitations. First, the assessment of leakage by FA is not always easy and reliable. Although in this study the inter- and intraobserver agreement values were high for the entire sample, in the occult subtype group the interobserver kappa was acceptable, but not excellent. We suppose that these values may be even lower when comparing observers from different centers and with different experience, as described by Holz et al. ²⁸ To reduce the complexity of FA grading, in the group of lesions with no leakage we did not differentiate between cases of dye staining or complete absence of fluorescence. A second limitation is that the confidence intervals in the sensitivity and specificity analysis were quite broad. Although this study is a “first step” in the correlation of lesion activity as defined by FA with SD-OCT features, further studies are warranted in a larger pool of patients. Third, this study does not follow changes over time, both with SD-OCT and FA. Although our results provide information on the sensitivity and specificity of several SD-OCT findings in predicting FA leakage from lesions, further studies following CNV over time may help to better clarify the biological basis of these findings. Fourth, the retrospective nature of the study needs to be carefully considered when generalizing the results. For example, in this specific data set the lesions with FA leakage tended to be seen in eyes with more recent treatment by an anti-VEGF agent injection. A possible explanation for this is that these eyes had a less robust response to treatment and required more frequent dosing. This fact may represent a confounding factor when attempting to generalize the results. Moreover, the acquisition protocol varied in the number of frames used for the real-time averaging process. This may have reduced the ability to assess for the presence of certain features, such as intraretinal flecks, when the averaging was reduced. 

In an era where SD-OCT may often be the predominantly used instrument to guide treatment decisions, it appears useful to identify parameters that help in identifying lesions that would show leakage with FA. This study suggests that an evaluation of lesion activity in the clinical setting may be enhanced by differentiating the pattern of fluid presentation by SD-OCT. Moreover, the study of other ultrastructural features such as intraretinal flecks and low reflectivity of subretinal material may further increase the accuracy of SD-OCT analysis. Nevertheless, additional studies are needed to confirm the clinical and scientific impact of these findings. The influence of different anti-VEGF agents, as well as other AMD treatments such as photodynamic therapy, on ultrastructural SD-OCT features may be of interest. Moreover, histologic and immunohistochemical studies to investigate the nature of these SD-OCT findings may help to shed light on the pathophysiology of neovascular AMD. 

Supplementary Materials

Table st1, DOC - Table st1, DOC 

Table st2, DOC - Table st2, DOC 

Table st3, DOC - Table st3, DOC 

Table st4, DOC - Table st4, DOC 

Table st5, DOC - Table st5, DOC 

Footnotes

 Disclosure: A. Giani, None; C. Luiselli, None; D.D. Esmaili, None; P. Salvetti, None; M. Cigada, None; J.W. Miller, None; G. Staurenghi, Heidelberg Engineering (C)