Assessment of Retinal Morphology with Spectral and Time Domain OCT in the Phase III Trials of Enzymatic Vitreolysis

Abstract

Purpose.: To determine the relative ability of time domain (TD)-optical coherence tomography (OCT) compared with spectral domain (SD)-OCT to assess vitreoretinal interface abnormalities and pharmacologic treatment of symptomatic vitreomacular adhesion (VMA)/traction (VMT) with or without full-thickness macular hole (FTMH), and the reproducibility of trained readers' evaluation of these images in an interventional phase III program of ocriplasmin.

Methods.: Eyes from the MIVI-TRUST program with concurrent SD-OCT and TD-OCT at baseline and day 28 were included. Pairwise intermodality agreement frequency and interreader reproducibility were calculated for baseline OCT features and the study endpoints of VMA resolution and FTMH closure.

Results.: A total 186 eyes (186 patients) met the inclusion criteria for this study. There was excellent agreement between TD-OCT and SD-OCT for the reader-determined presence or absence of VMA (96.7%), FTMH (97.1%), and all other baseline parameters except epiretinal membrane (84.3%), which was detected at a significantly greater rate with SD-OCT than TD-OCT (44.6% vs. 35.3%, P < 0.001). There was excellent agreement for the study endpoints of VMA resolution (95.4%) and FTMH closure (100%) at day 28. Interreader reproducibility was similar but consistently greater with SD-OCT than TD-OCT to detect baseline VMA (kappa 0.6 vs. 0.52); FTMH (kappa 0.9 vs. 0.78); and epiretinal membrane (kappa 0.65 vs. 0.45).

Conclusions.: Readers using SD-OCT or TD-OCT have similar ability to assess vitreoretinal interface abnormalities and outcomes of enzymatic vitreolysis. SD-OCT may be superior for formal clinical trial grading due to greater interreader reproducibility and, therefore, decreased need for arbitration of discrepant values. (ClinicalTrials.gov numbers,NCT00781859, NCT00798317.)

Introduction

Optical coherence tomography (OCT) produces cross-sectional retinal images in vivo that are more sensitive than biomicroscopy to identify vitreomacular adhesion (VMA) and disorders of the vitreoretinal interface. ¹ VMA is the persistent attachment of the posterior hyaloid to the macula during the formation of a partial posterior vitreous detachment. It may produce tractional forces that lead to a spectrum of disorders, such as vitreomacular traction syndrome (VMT) and full-thickness macular hole (FTMH). ^2–5 It may exacerbate cystoid macular edema (CME) associated with epiretinal membrane (ERM), retinal vein occlusion, and diabetic retinopathy. ^6–8 In eyes with high myopia and staphyloma, VMA may be associated with progressive retinal schisis and posterior retinal detachment. ⁹  

Symptomatic VMA/VMT with or without FTMH was treated with enzymatic vitreolysis in prospective phase III trials that showed statistically significant and clinically relevant VMA resolution and FTMH closure at day 28 after treatment, when compared with placebo injection. ¹⁰ The primary and secondary trial endpoints were determined by teams of certified OCT readers that graded images from time domain OCT (TD-OCT) machines used at all clinical sites throughout follow-up. Spectral domain OCT (SD-OCT) machines achieve greater image resolution than TD-OCT machines, thanks to faster data acquisition and improved broadband light sources. ^11,12 However, it is unclear whether SD-OCT, when compared with TD-OCT, improves readers' capability to interpret vitreoretinal interface disorders or the results of pharmacologic vitreolysis. Significant differences may have implications for the ability of investigators and clinicians to compare the results of recent TD-OCT trials with the results of future SD-OCT trials for the pharmacologic or surgical treatment of symptomatic VMA/VMT and FTMH. 

The Microplasmin IntraVitreal Injection for Traction Release without Surgical Treatment (MIVI-TRUST) phase III program included an SD-OCT substudy in which both SD-OCT and TD-OCT imaging were obtained at clinical sites that had both OCT modalities. The purpose of this study was to determine the agreement between spectral and time domain OCT when used to evaluate vitreoretinal interface disorders and the pharmacologic treatment of symptomatic VMA/VMT and FTMH. Our secondary purpose was to determine the ability of trained readers to reproducibly grade VMA and associated retinal morphology on SD-OCT compared with TD-OCT. 

Methods

The MIVI-TRUST program comprised two large multicenter, prospective, randomized, placebo-controlled, phase III trials (ClinicalTrials.gov identifiers NCT00781859 and NCT00798317) that compared the efficacy of a single 125-μg ocriplasmin intravitreal injection (ThromboGenics NV, Leuven, Belgium) with a single placebo injection for obtaining VMA release without vitrectomy. The MIVI-TRUST program recruited 652 subjects from 90 centers across the United States and Europe. One eye with symptomatic VMA/VMT was enrolled per subject, according to inclusion and exclusion criteria that have been previously described. ¹⁰ All subjects received thorough counseling, signed a written informed consent form prior to enrollment in the program, and then were followed for 6 months with serial OCT scans obtained at each visit. The trial procedures were approved by the institutional review board (IRB) at each participating study site. The present study was approved by the Duke University IRB and followed the tenets of the Declaration of Helsinki. 

In the present study, we enrolled all eyes in the MIVI-TRUST program that had paired TD-OCT and SD-OCT images for the baseline visit and for the day 28 clinical endpoint visit. Eyes were excluded if they were missing either TD-OCT or SD-OCT scans at either visit. In the MIVI-TRUST program, all study centers were required to use the same OCT equipment (Stratus OCT, software version 6.0; Carl Zeiss Meditec, Dublin, CA) for TD-OCT imaging at each study visit. Technicians were certified for these trials and followed a scan protocol emphasizing appropriate focus, saturation, and scan line placement. 

The Stratus OCT protocol for each visit required a macular thickness map (MTM) and a fast macular thickness (FTM) map, each consisting of six equally spaced, 6-mm radial lines centered on the fovea and separated 30° apart in a rotational manner. Each radial line was comprised of 512 A-scans per line for MTM and 128 A-scans per line for FTM, with an axial resolution of 10 μm per pixel. The protocol also required a 10-mm vertical crosshair scan at 90° through the optic disc, a 10-mm horizontal crosshair scan at 180° through the optic disc, and a 10-mm offset high-resolution scan through the center of the fovea and the optic disc. This offset scan line was angled at 5° for right eyes and 355° for left eyes. 

Two SD-OCT devices with distinct imaging protocols, the Cirrus HD-OCT (software version 5.2; Carl Zeiss Meditec, Dublin, CA) and Spectralis OCT (software version 5.3; Heidelberg Engineering, Carlsbad, CA), were used in the SD-OCT substudy. For Cirrus, three volumetric scan patterns with an axial resolution of 5 μm per pixel were obtained. These patterns included a macular volume cube that covered a 6-mm × 6-mm area of the retina with 128 horizontal line scans and 512 A-scans per line, an optic nerve cube that covered a 6-mm × 6-mm area of the retina with 200 horizontal line scans and 200 A-scans per line, and a 5-line raster scan that consisted of five consecutive 6-mm horizontal high-resolution scan lines spaced 250 μm apart (covering a 6-mm × 1-mm area centered on the fovea) with 4096 A-scans per line. For Spectralis, two volumetric 30° × 15° raster scan patterns were obtained, covering an 8.6-mm × 4.3-mm area of the retina. Each volumetric pattern was composed of 19 consecutive horizontal high-resolution scan lines, spaced 240 μm apart, with 1536 A-scans per line and an axial resolution of 4 μm per pixel. One volume pattern was centered on the fovea. The second volume pattern was centered between the fovea and the optic disc, covering both regions. 

All OCT scans from the MIVI-TRUST phase III program were submitted to the Duke Reading Center (Duke University, Durham, NC), where the coded scans for a given subject and visit date were graded by two certified readers in a masked and independent manner. A data specialist entered all concordant values from the two readers into the trial database and flagged all discrepant values. A third certified senior reader arbitrated the discrepant values. The senior reader reconciled all reader disagreements according to his best judgment and expertise, recording his decision as the final arbitrated value that the data specialist entered into the trial database. All reader disagreements that remained controversial despite arbitration were presented to the director of grading (CAT) or the reading center director (GJJ) for a final decision. Each OCT variable was graded as present, absent, or unreadable (due to poor quality or centration of the scan), according to either the agreed decision between masked readers or the arbitrated decision when the readers disagreed. After finishing arbitration and data entry, another masked data specialist or project manager verified the accuracy of all values entered into the trial database. The verified data were used to compare TD- and SD-OCT agreement on each study variable. 

VMA was defined on OCT as attachment of the posterior vitreous hyaloid to the macula with vitreous detachment visible on opposite sides of the adhesion site, based on examination of one or more OCT B-scans within the central 6-mm field. VMA width was graded categorically based on whether the widest site of adhesion to the macula was broad (>1500 μm), focal (≤1500 μm), or could not be determined. For multifocal VMA sites with intervening vitreous detachment, the VMA width was determined by the greatest sum of attachment widths in a single radial or raster line scan. Subparameters of VMA included adhesions associated with FTMH, ERM, or central foveal deformation. Additional graded categorical variables included the following: FTMH, lamellar hole or pseudohole (LH/PH), macular hole operculum, ERM, inner retinal or foveal deformation by ERM, intraretinal cystoid spaces representing CME, subretinal fluid (SRF), and retinoschisis (RS) defined as retinal splitting beyond the central 1-mm diameter of the fovea. Additional subparameters included CME or SRF within 500 μm from the foveal center. 

Categorical data were reported as the percent frequency of occurrence based on final arbitrated values. If an OCT variable was graded as unreadable for any subject, then the subject's data point was excluded from analysis of the total paired agreement for this variable. The agreement of TD-OCT versus SD-OCT for each of these features was reported as the percent agreement and the Cohen kappa coefficient with 95% confidence intervals (CI). Pairwise comparisons of statistical significance and symmetry of agreement for each parameter were performed with the McNemar-Bowker χ² test, where P < 0.05 was considered statistically significant. The same tests were repeated to determine the paired agreement of Stratus TD-OCT with each specific SD-OCT device, Cirrus and Spectralis. 

Interreader reproducibility for TD-OCT and SD-OCT grading was defined as the initial agreement between two certified readers to detect VMA and associated OCT features prior to arbitration. Reader agreement for a variable detected with one OCT modality was determined by the difference of the total number of eligible eyes graded and the number of eyes requiring arbitration due to discordant values. Interreader reproducibility for TD-OCT and SD-OCT was compared by reporting the percent reader agreement and the Cohen kappa coefficient with 95% CI for each parameter. Statistical analyses were performed with statistical modeling software (SAS JMP Pro 9.0; SAS Institute, Inc., Cary, NC). 

Results

Paired Comparison of SD-OCT versus TD-OCT

A total of 186 eyes (186 subjects) from 30 centers participating in the MIVI-TRUST program met the enrollment criteria for this study. Among this cohort, there were 86 right eyes and 100 left eyes. A total of 744 OCT scans were evaluated, consisting of SD-OCT and TD-OCT for baseline and day 28 visits for all 186 enrolled eyes. 

Readers detected baseline VMA-associated pathology with similar frequency on SD-OCT and TD-OCT for nearly all parameters. Only the frequency of ERM detection differed significantly (SD-OCT 45% versus TD-OCT 35%). Among eyes with agreement on baseline VMA or baseline FTMH from paired SD-OCT and TD-OCT scans, the primary trial endpoint of VMA resolution and the secondary trial endpoint of FTMH closure at day 28 were detected with equal frequency by both machines (Table 1). 

Table 1. 

 

Frequency of Pathology and Study Endpoints Detected with SD- versus TD-OCT among Eyes with Dual Imaging

Group	% Eyes (no. of eyes/total eyes)		P Value*
	SD-OCT	TD-OCT
Baseline foveal VMA	93.6% (174/186)	94.6% (176/186)	0.89
Broad (>1500 μm)	21.7% (36/166)	22.3% (37/166)	0.76
Focal (≤1500 μm)	78.3% (130/166)	77.7% (129/166)	For broad vs. focal
Deformation by VMA	90.1% (164/182)	92.3% (168/182)	0.32
Resolution of VMA	20.6% (36/175)	21.7% (38/175)	0.48
Baseline FTMH	20.8% (36/173)	22.5% (39/173)	0.18
Closure of FTMH	40.0% (14/35)	40.0% (14/35)	1.0
Baseline LH/PH	11.1% (19/171)	10.5% (18/171)	0.81
Baseline ERM	44.6% (82/184)	35.3% (65/184)	<0.001
Baseline Retinoschisis	13.3% (23/173)	11.6% (20/173)	0.55
Baseline CME	83.9% (151/180)	81.7% (147/180)	0.32
Central 1-mm CME	99.3% (140/141)	96.5% (136/141)	0.10
Baseline SRF	42.0% (73/174)	42.0% (73/174)	1.0
Central 1-mm SRF	98.4% (60/61)	98.4% (60/61)	1.0

 FTMH <400 μm.

*  McNemar-Bowker χ² test for disagreement.

Concordance, defined as the percent agreement between OCT modalities when the pathology was present or absent, was excellent for baseline VMA (97%) and VMA-associated pathology. Concordance for the presence or absence of ERM (84%) was the lowest among all baseline parameters. There was excellent concordance (95%) for VMA resolution and perfect concordance (100%) for FTMH closure at day 28 (Table 2). 

Table 2. 

 

Concordance of Pathology and Study Endpoints Detected with SD- versus TD-OCT among Eyes with Dual Imaging

Group	% Eyes (no. of eyes/total eyes)				Kappa (95% CI)
	SD and TD Agree Yes	SD and TD Agree No	SD Yes TD No	TD Yes SD No
Baseline foveal VMA	92.4% (172/186)	4.3% (8/186)	1.1% (2/186)	2.2% (4/186)	0.57 (0.36–0.79)
Broad (>1500 μm)	18.7% (31/166)	74.7% (124/166)	3.0% (5/166)	3.6% (6/166)	0.81 (0.70–0.92)
Focal (≤1500 μm)	74.7% (124/166)	18.7% (31/166)	3.6% (6/166)	3.0% (5/166)	For broad vs. focal
Deformation by VMA	86.8% (158/182)	4.4% (8/182)	3.3% (6/182)	5.5% (10/182)	0.45 (0.23–0.68)
Resolution of VMA	18.9% (33/175)	76.5% (134/175)	1.7% (3/175)	2.9% (5/175)	0.86 (0.77–0.96)
Baseline FTMH	20.2% (35/173)	76.9% (133/173)	0.6% (1/173)	2.3% (4/173)	0.91 (0.84–0.99)
Closure of FTMH	40.0% (14/35)	60.0% (21/35)	0	0	–
Baseline LH/PH	5.8% (10/171)	84.2% (144/171)	5.3% (9/171)	4.7% (8/171)	0.48 (0.27–0.70)
Baseline ERM	32.1% (59/184)	52.2% (96/184)	12.5% (23/184)	3.2% (6/184)	0.67 (0.57–0.78)
Baseline retinoschisis	5.2% (9/173)	80.3% (139/173)	8.1% (14/173)	6.4% (11/173)	0.33 (0.13–0.54)
Baseline CME	78.3% (141/180)	12.8% (23/180)	5.6% (10/180)	3.3% (6/180)	0.69 (0.55–0.83)
Central 1-mm CME	95.7% (135/141)	0	3.6% (5/141)	0.7% (1/141)	–
Baseline SRF	35.1% (61/174)	51.1% (89/174)	6.9% (12/174)	6.9% (12/174)	0.72 (0.61–0.82)
Central 1-mm SRF	98.4% (60/61)	1.6% (1/61)	0	0	–

 FTMH <400 μm.

A post hoc review of spectral and time domain OCT imaging was performed in eyes with OCT modality disagreement for baseline VMA, FTMH, or day 28 VMA resolution. With SD-OCT raster scans, readers failed to detect baseline VMA in four eyes, baseline FTMH in four eyes, and persistent day 28 VMA in three eyes that were detected with TD-OCT radial scans (Fig. 1). However, with TD-OCT radial scans centered on the fovea, readers failed to detect baseline FTMH in one eye and persistent day 28 VMA in two eyes that had eccentric pathology detected with SD-OCT raster scans. With TD-OCT, readers also failed to detect baseline VMA in two eyes and day 28 VMA in three eyes that were detected with SD-OCT due to greater signal strength at the vitreoretinal interface and posterior vitreous cortex (Fig. 2). 

Figure 1. 

 

OCT nonconcordance of VMA between TD-OCT and SD-OCT modalities. (A) Macular hole, temporal epiretinal membrane, and nasal VMA (arrow) detected by TD-OCT readers. (B) Macular hole and epiretinal membrane detected by SD-OCT; however, persistent VMA was not sampled, possibly because adhesion was only present between horizontal raster scan lines, and therefore not detected by SD-OCT readers.

Figure 2. 

 

OCT nonconcordance of VMA between TD-OCT and SD-OCT modalities. (A) Poor reflectance signal, or possibly location of adhesion between radial scan lines, inhibits VMA (arrow) from being detected by TD-OCT readers. (B) Persistent VMA (arrow) detected by SD-OCT readers.

Paired Comparison of Specific SD-OCT Machines versus TD-OCT

In this study, SD-OCT scans were obtained with Cirrus from 119 eyes and with Spectralis from 67 eyes. Readers detected baseline VMA and FTMH with similar frequency when comparing Cirrus with Stratus and when comparing Spectralis with Stratus. ERM was detected significantly more frequently with Cirrus than with Stratus (Fig. 3). Similarly, ERM was detected more frequently with Spectralis than with Stratus, but this difference did not reach statistical significance. Each SD-OCT machine, when compared with Stratus, identified with similar frequency the study endpoints of VMA resolution and FTMH closure (Table 3). 

Figure 3. 

 

OCT nonconcordance of epiretinal membrane (ERM) between TD-OCT and SD-OCT modalities. (A) ERM was not detected by TD-OCT readers at the site of inner retinal hyperreflectivity adjacent to VMA. (B) Visible separation of ERM from inner retinal surface (arrow) allowed SD-OCT readers to detect the ERM adjacent to VMA.

Table 3. 

 

Frequency of Pathology and Study Endpoints Detected with Cirrus and Spectralis versus Stratus among Eyes with Dual Imaging

Group	% Eyes (no. of eyes/total eyes)		P Value*
	SD-OCT	TD-OCT
Cirrus (n = 119)
Baseline foveal VMA	94.1% (112/119)	95.0% (113/119)	0.98
Resolution of VMA	21.2% (24/112)	23.0% (26/112)	0.41
Baseline FTMH	26.6% (29/109)	26.6% (29/109)	1.0
Closure of FTMH	46.4% (13/28)	46.4% (13/28)	1.0
Baseline ERM	45.3% (53/117)	33.3% (39/117)	<0.001
Spectralis (n = 67)
Baseline foveal VMA	92.5% (62/67)	94.0% (63/67)	0.55
Resolution of VMA	19.4% (12/62)	19.4% (12/62)	1.0
Baseline FTMH	10.9% (7/64)	15.6% (10/64)	0.08
Closure of FTMH	14.3% (1/7)	14.3% (1/7)	1.0
Baseline ERM	43.3% (29/67)	38.8% (26/67)	0.37

 FTMH <400 μm.

*  McNemar-Bowker χ² test for disagreement.

Concordance of each specific SD-OCT machine with Stratus was excellent for baseline VMA and FTMH, but concordance was lower for ERM. Concordance for VMA resolution was excellent for Cirrus (95%) and Spectralis (97%) with respect to Stratus. Concordance for FTMH closure was perfect (100%) for both Cirrus and Spectralis with respect to Stratus (Table 4). 

Table 4. 

 

Concordance and Distribution of Grading for Cirrus and Spectralis versus Stratus among Eyes with Dual Imaging

Group	% Eyes (no. of eyes/total eyes)				Kappa (95% CI)
	SD and TD Agree Yes	SD and TD Agree No	SD Yes TD No	TD Yes SD No
Cirrus (n = 119)
Baseline foveal VMA	93.3% (111/119)	4.3% (5/119)	0.8% (1/119)	1.6% (2/119)	0.60 (0.33–0.87)
Resolution of VMA	19.5% (22/113)	75.2% (85/113)	1.8% (2/113)	3.5% (4/113)	0.85 (0.72–0.97)
Baseline FTMH	25.7% (28/109)	72.5% (79/109)	0.9% (1/109)	0.9% (1/109)	0.95 (0.89–1.02)
Closure of FTMH	46.4% (13/28)	53.6% (15/28)	0	0	1.0
Baseline ERM	31.6% (37/117)	53.0% (62/117)	13.7% (16/117)	1.7% (2/117)	0.68 (0.55–0.81)
Spectralis (n = 67)
Baseline foveal VMA	91.0% (61/67)	4.5% (3/67)	1.5% (1/67)	3.0% (2/67)	0.53 (0.18–0.88)
Resolution of VMA	17.8% (11/62)	79.0% (49/62)	1.6% (1/62)	1.6% (1/62)	0.90 (0.76–1.04)
Baseline FTMH	10.9% (7/64)	84.4% (54/64)	0	4.7% (3/64)	0.80 (0.58–1.02)
Closure of FTMH	14.3% (1/7)	85.7% (6/7)	0	0	1.0
Baseline ERM	32.8% (22/67)	50.7% (34/67)	10.5% (7/67)	6.0% (4/67)	0.66 (0.48–0.84)

 FTMH = <400 μm.

Interreader Reproducibility of SD-OCT versus TD-OCT Grading

Tables 5 and 6 present interreader reproducibility at baseline and day 28, respectively. For baseline visits, all 186 study eyes had initial reader values available for reproducibility analysis. For day 28 visits, 185 study eyes had initial reader values available. At baseline, SD-OCT had higher interreader agreement, and therefore lower rate of arbitration than TD-OCT for all parameters (Table 5). At day 28, SD-OCT had higher interreader reproducibility than TD-OCT for all parameters except OCT features associated with ERM (Table 6). At baseline and day 28, with both OCT modalities, interreader reproducibility was greatest for FTMH. At baseline, with both OCT modalities, interreader reproducibility was lowest for broad versus focal VMA. At day 28, reproducibility with TD-OCT was lowest for broad versus focal VMA, but reproducibility with SD-OCT was lowest for ERM with central foveal deformation. 

Table 5. 

 

Baseline Interreader Agreement with SD- versus TD-OCT

Group	% Eyes (no. of eyes/total eyes)		% Eyes (no. of eyes/total eyes)	Kappa (95% CI)
	SD Reader Agreement	Kappa (95% CI)	TD Reader Agreement
Foveal VMA	80.1% (149/186)	0.60 (0.49–0.72)	75.8% (141/186)	0.52 (0.39–0.64)
Broad vs. focal VMA*	73.1% (136/186)	0.46 (0.34–0.59)	60.8% (113/186)	0.22 (0.07–0.36)
FTMH	95.2% (177/186)	0.90 (0.84–0.96)	88.2% (164/186)	0.78 (0.70–0.87)
ERM	82.3% (153/186)	0.65 (0.54–0.75)	72.6% (135/186)	0.45 (0.32–0.58)
Any retinal deformation by ERM	86.6% (161/186)	0.73 (0.63–0.83)	78.0% (145/186)	0.56 (0.44–0.68)
Central 1-mm retinal deformation by ERM	80.1% (149/186)	0.60 (0.49–0.72)	77.4% (144/186)	0.55 (0.43–0.67)
ERM at the site of VMA	81.7% (152/186)	0.63 (0.52–0.75)	78.0% (145/186)	0.56 (0.44–0.68)
CME	88.7% (165/186)	0.77 (0.68–0.87)	83.3% (155/186)	0.67 (0.56–0.77)
SRF	85.5% (159/186)	0.71 (0.61–0.81)	78.5% (146/186)	0.57 (0.45–0.69)

 FTMH <400 μm.

*  Modified Cohen Kappa for grading as VMA width broad, focal, or unable to determine.

Table 6. 

 

Day 28 Interreader Agreement with SD- versus TD-OCT

Group	% Eyes (no. of eyes/total eyes)		% Eyes (no. of eyes/total eyes)
	SD Reader Agreement	Kappa (95% CI)	TD Reader Agreement	Kappa (95% CI)
Foveal VMA	86.5% (160/185)	0.73 (0.63–0.83)	77.8% (144/185)	0.56 (0.44–0.68)
Broad vs. focal VMA*	77.8% (144/185)	0.56 (0.44–0.68)	73.5% (136/185)	0.47 (0.34–0.60)
FTMH	96.8% (179/185)	0.94 (0.88–0.99)	90.8% (168/185)	0.82 (0.73–0.90)
ERM	81.6% (151/185)	0.63 (0.52–0.74)	80.0% (148/185)	0.60 (0.48–0.72)
Any retinal deformation by ERM	81.1% (150/185)	0.62 (0.51–0.73)	81.1% (150/185)	0.62 (0.51–0.73)
Central 1-mm retinal deformation by ERM	75.1% (139/185)	0.50 (0.38–0.63)	79.5% (147/185)	0.59 (0.47–0.71)
ERM at the site of VMA	76.8% (142/185)	0.54 (0.41–0.66)	78.9% (146/185)	0.58 (0.46–0.70)
CME	94.1% (174/185)	0.88 (0.81–0.95)	80.0% (148/185)	0.60 (0.48–0.72)
SRF	89.2% (165/185)	0.78 (0.69–0.87)	81.6% (151/185)	0.63 (0.52–0.74)

 FTMH <400 μm.

*  Modified Cohen Kappa for grading VMA width as broad, focal, or unable to determine.

Discussion

In this report, we found that certified readers detected baseline VMA, OCT features associated with VMA, and the trial endpoints of VMA resolution and FTMH closure, with very similar rates on SD-OCT and TD-OCT imaging. We also observed similar rates in a comparison of pathology grading between TD-OCT and two specific widely used SD-OCT instruments. The interreader grading reproducibility was generally high for both OCT modalities, although the agreement among primary readers for baseline pathology was greater on SD-OCT than TD-OCT. We included only a subset of eyes from the MIVI-TRUST phase III program that had dual OCT imaging; therefore, this study was not designed to verify the previously reported efficacy of ocriplasmin on the treatment of symptomatic VMA/VMT and macular hole. ¹⁰  

Studies to determine agreement of TD-OCT and SD-OCT quantitative and qualitative parameters have been reported previously for neovascular AMD, ^13–15 diabetic macular edema, ^13,16 and uveitis. ^13,17 These reports suggest that OCT machines should not be used interchangeably for quantitative retinal thickness measurements of healthy and pathologic eyes. ^13,18–20 To the best of our knowledge, literature is absent that evaluates the relative merits of TD-OCT and SD-OCT for evaluating vitreoretinal interface disorders in prospective clinical trials. With the transition from TD-OCT to SD-OCT imaging in randomized clinical trial protocols, there is now a greater need to understand the impact that spectral domain technology would have had on trial outcomes determined by TD-OCT. 

This study analyzed prospective data from the MIVI-TRUST program, and showed that readers can effectively assess baseline VMA and associated pathology with both time domain and spectral domain OCT. Discrepancies in the assessment of VMA most often were the result of differences in the area sampled by the radial versus raster line orientation of the scanning protocols, and not due to advantages in spatial resolution. This finding suggests that a combination of radial and parallel line scans with either OCT modality will improve the accuracy of OCT grading more than a single-orientation volumetric scan obtained with time domain or spectral domain OCT. 

The only anatomic feature that was detected at a significantly different rate by SD-OCT than with TD-OCT was ERM, although the detection discrepancy was still much lower than that reported previously. In a study of eyes with a clinical diagnosis of ERM, Falkner-Radler et al. ²¹ found that well-differentiated ERM was detected in 61% by SD-OCT grading, but only 32% by TD-OCT. In the present study, ERM was detected by SD-OCT but not TD-OCT in approximately 12% of evaluated eyes. We speculate that SD-OCT readers detected ERM more often than TD-OCT readers for two reasons. First, higher axial resolution leads to superior visualization of separations between ERM and the internal limiting membrane. Second, greater scanning density with the SD-OCT raster protocol permits observation of small sites of ERM separation. These results imply that TD-OCT trial results for ERM should be interpreted with caution. 

The relative ability of clinicians to use SD-OCT when compared with TD-OCT in daily clinical practice to detect VMA and associated pathology remains unclear. In the present report, certified OCT readers were trained to carefully review all radial scans obtained with Stratus, or all raster scans in volume cubes obtained with Cirrus and Spectralis. We hypothesize that if clinicians adopt this technique and review the entire available scan, both types of machines will enable nearly equal detection of VMA and most types of associated pathology, with the exception of ERM. We expect that clinicians will be able to identify ERM more frequently with SD-OCT, for reasons mentioned above. 

Interreader grading reproducibility was generally high with both spectral and time domain OCT. We and others have reported high quantitative and qualitative inter- and intrareader reproducibility with TD-OCT in eyes with neovascular AMD, ^22–24 and with SD-OCT in healthy eyes. ²⁵ We have recently reported excellent intrareading center team reproducibility for TD-OCT grading of baseline VMA, broad versus focal VMA width, baseline FTMH, and baseline ERM in the MIVI-TRUST program. ²⁶ In the present study, we found a similar high degree of interreader reproducibility with both spectral and time domain OCT for most parameters assessed. Of the parameters measured, readers agreed least when differentiating broad from focal VMA, on both TD-OCT and SD-OCT images. To determine broad versus focal VMA, readers measured the macular adhesion maximum transverse width. The measurement variability among readers was likely due to two factors: first, higher subjectivity among readers in measuring horizontal VMA width, relative to scoring other VMA-associated pathology; and second, different sectioning patterns—radial versus raster line orientation—for the two modalities. In previous reports, the maximum vitreomacular traction width has been highly variable, ranging from 3500 to 6000 μm in pilot studies of 7 to 19 eyes. ²⁷ Not only does VMA have highly variable horizontal diameters, but OCT reconstructions of VMA often reveal asymmetric conoid adhesions to retinal tissue that may create ambiguity among readers or clinicians when choosing the horizontal plane in which to take measurements. ²⁷ We could not find previous reproducibility studies of VMA width measurement, so we currently have no basis with which to compare this finding in the present study. 

Although grading reproducibility was generally high, it was modestly but consistently higher among SD-OCT readers on all baseline parameters and most day 28 parameters. According to our reading center protocols, discrepant values produced by a masked reader pairs were submitted to a senior reader who arbitrated each disagreement. Since fewer discrepant values required arbitration with SD-OCT, less time dedicated to the arbitration process would result in greater efficiency of data collection during SD-OCT grading. However, this increased efficiency is offset somewhat by the increased time required for readers to review each SD-OCT scan volume, due to the greater number of scan lines that must be examined in a spectral domain macular volume cube than in a TD-OCT macular thickness map. Further grading time studies would be needed to assess the relative overall efficiency of spectral and time domain OCT grading efficiency in a reading center setting. 

In conclusion, our results indicate that both TD-OCT and SD-OCT can be used effectively to assess the morphology of vitreoretinal interface disorders in multicenter clinical trials. Our results also show that trained readers can assess pharmacologic vitreolysis equally with TD-OCT and SD-OCT. The pairwise comparison of TD-OCT with SD-OCT validates the TD-OCT grading protocol for MIVI-TRUST and the data obtained from this clinical interventional program. However, for the purpose of formal OCT grading for multicenter clinical trials, SD-OCT may be slightly superior due to increased initial reader agreement and, therefore, increased efficiency in the grading process. New clinical trials will inevitably transition to SD-OCT imaging exclusively, but for the purpose of scientific discussion, the results reported in future trials based exclusively on SD-OCT can be confidently compared to the results obtained from this phase III program. 

Acknowledgments

The authors thank Russell Burns and Cindy Heydary of the Duke Reading Center for research assistance. 

Footnotes

 Presented in part at the annual meeting of the Macula Society, Boca Raton, Florida, March 2011.

Footnotes

 Disclosure: F.A. Folgar, None; C.A. Toth, Genentech (F), Bioptigen (F), Alcon (F), Physical Sciences, Inc. (F, C), P; F.C. DeCroos, None; A. Girach, ThromboGenics NV (E); S. Pakola, ThromboGenics NV (E), P; G.J. Jaffe, Heidelberg Engineering (C), Neurotech (C), SurModics (C)