To evaluate the utility of preoperative optical coherence tomography (OCT) and multifocal electroretinography (mfERG) in prediction of visual outcomes after idiopathic epiretinal membrane (ERM) surgery.
Retrospective, observational case series.
One hundred eyes of 100 patients with idiopathic unilateral ERM who underwent vitrectomy for ERM removal were retrospectively reviewed. Correlations between preoperative data (OCT and mfERG) and final best-corrected visual acuity (BCVA) were investigated using Pearson correlation analysis. One-way analysis of variance (ANOVA) was used to determine whether final BCVA and mfERG values differed among subgroups varying in photoreceptor integrity status. Receiver operating characteristic (ROC) curve analysis was performed to obtain a cutoff value of the P1 implicit time predicting visual recovery (final BCVA ≥20/25).
BCVA significantly improved, and 65 of 84 eyes (77%) achieved visual recovery of more than 2 Snellen lines after ERM surgery. Final BCVA was significantly correlated with preoperative photoreceptor integrity and P1 implicit time. The area under the ROC (AUROC) curve was statistically significant when P1 implicit time was examined, and the cutoff value for good visual prognosis was 40.81 msec (sensitivity: 72.7%; specificity: 81.3%).
Photoreceptor disruption detected by OCT and P1 implicit time delay on mfERG were significant predictors of poor visual recovery after ERM surgery.
Idiopathic epiretinal membrane (ERM) is a common disorder, occurring in about 7% of the population. A fibrocellular growth in the form of an ERM induces tangential tractional force on the retina, leading to macular constriction and thickening. When patients present with significant symptoms including metamorphopsia, reduced visual acuity (VA), and/or monocular diplopia, surgical removal of the membrane is performed as a standard treatment. It has been shown that visual improvement after surgery is attained by 70% to 80% of patients.
Many researchers have sought to evaluate macular structure and function before and after ERM removal using multifocal/focal electroretinography (mfERG/fERG) and optical coherence tomography (OCT). Recently, Suh and associates demonstrated that photoreceptor disruption detected by OCT is a predictor of poor visual outcome after ERM surgery and prompt surgical intervention is required to prevent irreversible photoreceptor damage. Although dysfunction of the innermost macular layers may be a primary impairment caused by ERM, the fact that photoreceptor integrity per se is a significant prognostic factor indicates that mfERG mapping of outer retinal function may be valuable in prediction of visual outcome after ERM surgery.
In the present study, we evaluated the prognostic values of preoperative OCT findings and mfERG responses with respect to visual outcomes after ERM surgery.
We retrospectively reviewed data for consecutive patients who were diagnosed with unilateral idiopathic ERM and operated on by a single surgeon (H.J.K.) for ERM removal between January 1, 2004 and March 31, 2008 at the vitreoretinal service clinic of the Yonsei University Eye, Ear, Nose, and Throat Hospital (Seoul, Korea). Patients who were followed up for at least 12 months after surgery were included in the analysis. The exclusion criteria were: 1) presence of a secondary ERM caused by diabetic retinopathy, venous occlusion, retinal detachment (RD), uveitis, or trauma; 2) refractive error ≥ ± 6.0 diopters; and 3) other ocular pathologies that might interfere with visual function, including severe cataract of more than grade 2 nuclear sclerosis and/or cortical opacity, glaucoma, or age-related macular degeneration.
Information collected from each patient included age, gender, duration of symptoms, presence of comorbid medical conditions (including diabetes mellitus and hypertension), lens status (phakic or pseudophakic), and concomitant surgical procedures (if any) during ERM removal. Snellen best-corrected visual acuity (BCVA) testing, slit-lamp biomicroscopy, and indirect ophthalmoscopy were performed at baseline and 6 and 12 months after surgery. BCVA was measured using a Snellen visual acuity chart and converted into the logarithm of the minimal angle of resolution (logMAR) prior to statistical analysis. VA at 12 months after surgery was used as final analysis because VA was stabilized at that time. OCT and mfERG examinations were performed before surgery.
OCT employed an OCT3 instrument (Stratus Zeiss Humphrey, San Leandro, California, USA). The macula was scanned in the horizontal and vertical meridians using the standard linear crosshair pattern, with a scan length of 6 mm centered through the fovea. The central macular thickness (CMT) was manually measured on horizontal OCT3 images of all eyes by calibrated calipers at the vitreous-retina and sensory retina–retinal pigment epithelium (RPE) interfaces.
Eyes were categorized into 3 groups depending on the appearance of the inner segment/outer segment (IS/OS) junction on OCT images of the fovea: intact, disrupted, and uncheckable groups. An intact photoreceptor line was defined as a regular continuation of the hyperreflective line corresponding to the IS/OS junction. A disrupted photoreceptor line was characterized by a hyporeflective discontinuity in the hyperreflective IS/OS junction. If the IS/OS junction was not clearly delineated because of severe macular edema or some other artifact, the eye was regarded as uncheckable. IS/OS line grading was determined by 2 authors (J.H.K. and H.J.K.) who were masked to VA data.
All mfERG data were recorded using the RETI scan multifocal system (Roland Consult, Brandenburg, Germany). Stimulation and recording of mfERG responses were performed using the m-sequence technique, according to International Society for Clinical Electrophysiology of Vision (ISCEV) guidelines. The typical waveform of the mfERG response (the first-order response) is a biphasic wave with an initial negative deflection followed by a positive peak. These peaks are called N1 and P1, respectively. The amplitude and implicit time from 3 different retinal eccentricities (0–5 degrees, 5–10 degrees, and 10–15 degrees) were analyzed. We defined the first and second ring as area 1 and the first, second, and third rings as area 2. Normal fellow eyes were also measured as control information. Averages of mfERG amplitude and implicit time were used in analysis.
All procedures were performed by the same surgeon (H.J.K.), using a standard 3-port pars plana vitrectomy with complete removal of the ERM with the aid of retinal forceps. Internal limiting membrane (ILM) peeling was performed on 16 eyes after 0.03% indocyanine green (ICG) dye injection. No silicone oil or intraocular gas was introduced after vitrectomy in any enrolled eye. If the eyes were phakic, combined cataract surgery was performed in older patients for preventive purposes. When cataract developed after vitrectomy during the follow-up period, additional cataract surgery was done. Concomitant cataract surgery was performed on 58 eyes and sequential cataract surgery on 6 eyes.
Continuous values are expressed as mean ± standard deviation. The paired t test was used to compare BCVA before and after surgery and mfERG values of affected and normal fellow eyes. Correlations between BCVA results and preoperative OCT and mfERG values were analyzed using the Pearson test. One-way analysis of variance (ANOVA) was used to determine whether VA parameters (final BCVA and perioperative BCVA change) and mfERG values were significantly dependent on photoreceptor status. To identify factors related to visual prognosis, multivariate regression analysis was performed using final BCVA value as a dependent variable. A receiver operating characteristic (ROC) curve was constructed to obtain a cutoff value allowing prediction of visual prognosis. We tried to obtain a P1 implicit time cutoff value in prediction of a good visual outcome after ERM surgery. It was defined arbitrarily as a final BCVA of 20/25 or better. All statistical analyses were performed using SPSS version 17.0 (SPSS Inc, Chicago, Illinois, USA) for Windows. The level of statistical significance was set at P < .05 in all instances.
Of the 100 eyes included in our study, 16 eyes were excluded for various reasons, such as loss before 12-month follow-up (10 eyes), recurrence of ERM requiring additional surgery (3 eyes), and the presence of visually significant cataracts at final analysis (3 eyes). Therefore, 84 eyes of 84 patients (23 men and 61 women) ultimately met the study criteria. The membrane was successfully removed from all treated eyes, and no serious intraoperative or postoperative complications were observed. The mean age of patients was 63.0 ± 9.2 years (range 18–78 years). The mean preoperative CMT was 408.1 ± 122.0 μm (range 174–697 μm). Based on preoperative IS/OS status, 59 eyes (70.2%) showed intact IS/OS line, whereas 12 (14.3%) and 13 eyes (15.5%) had disrupted and uncheckable IS/OS line, respectively. Eleven patients were preoperatively pseudophakic.
The preoperative and postoperative BCVA values are shown in Figure 1 . Average VA improved significantly at 12 months postoperatively compared with baseline (paired t test, P < .05). Sixty-five of 84 eyes (77%) achieved visual recovery of more than 2 Snellen lines. A significant visual improvement was evident between 6 and 12 months postoperatively.
The final BCVA was significantly poorer in the disrupted IS/OS junction group than in the intact group ( Figure 2 ) . Mean final BCVA was 0.64 ± 0.32 (logMAR) in the disrupted group and 0.07 ± 0.19 in the intact group ( P < .001; t test). Mean BCVA change was slightly higher in the intact group (0.26 ± 0.19) than in the disrupted group (0.21 ± 0.21), but the difference between these values did not reach statistical significance ( P > .05; t test).
P1 implicit time was significantly delayed in both areas 1 ( P = .008) and 2 ( P < .001) in the disrupted group compared with the intact group, and the N1 implicit time difference in area 1 was marginally significant in a between-group comparison ( P = .035) ( Table 1 ). No other mfERG parameter differed significantly among subgroups classified by photoreceptor status.
|Preoperative Photoreceptor Status|
|Intact (n = 59)||Disrupted (n = 12)||Uncheckable (n = 13)||P Value a|
|P1 amplitude (nV/deg 2 )||58.28 ± 16.59||59.01 ± 28.66||53.14 ± 12.66||.676|
|P1 implicit time (ms)||40.90 ± 2.27||43.64 ± 2.42||41.95 ± 3.55||.008 b|
|N1 amplitude (μV)||0.28 ± 0.10||0.31 ± 0.16||0.26 ± 0.07||.563|
|N1 implicit time (ms)||19.59 ± 3.55||22.63 ± 2.79||20.94 ± 3.00||.035 b|
|P1 amplitude (nV/deg 2 )||44.54 ± 12.99||43.09 ± 20.16||41.45 ± 9.84||.785|
|P1 implicit time (ms)||39.79 ± 1.72||42.75 ± 2.58||40.70 ± 2.85||.000 b|
|N1 amplitude (μV)||0.28 ± 0.09||0.29 ± 0.13||0.26 ± 0.08||.807|
|N1 implicit time (ms)||20.03 ± 3.42||22.25 ± 1.81||21.04 ± 1.64||.107|
Table 2 summarizes the mfERG responses of affected and fellow eyes in areas 1 and 2. Mean preoperative P1 amplitude and implicit time of affected eyes differed significantly from those of normal fellow eyes in both areas 1 and 2 ( P < .05). Neither the N1 amplitude nor the N1 implicit time of affected eyes was significantly different from that of fellow eyes.
|Area 1||Area 2|
|P1 amplitude (nV/deg 2 )|
|Affected eye||57.10 ± 19.83||43.45 ± 14.67|
|Fellow eye||68.91 ± 18.67||51.69 ± 14.36|
|P value a||<.001||<.001|
|P1 implicit time (ms)|
|Affected eye||41.51 ± 2.66||40.44 ± 2.41|
|Fellow eye||39.50 ± 1.94||38.81 ± 1.75|
|P value a||<.001||<.001|
|N1 amplitude (μV)|
|Affected eye||0.30 ± 0.11||0.29 ± 0.09|
|Fellow eye||0.32 ± 0.08||0.31 ± 0.07|
|P value a||.176||.272|
|N1 implicit time (ms)|
|Affected eye||20.73 ± 2.23||20.91 ± 1.68|
|Fellow eye||20.04 ± 2.01||20.21 ± 1.53|
|P value a||.086||.058|
Correlation analyses of preoperative mfERG responses vs preoperative and final BCVA are summarized in Tables 3 and 4 , respectively. Although no significant correlations between preoperative BCVA and P1 amplitudes in either area 1 or 2 were evident, P1 implicit time was significantly correlated with preoperative BCVA in both area 1 (r = 0.315, P = .008) and area 2 (r = 0.450, P < .001) ( Table 3 ). The correlation of final BCVA and mfERG responses showed similar tendency ( Table 4 ). The average preoperative CMT was 408.1 ± 122.0 μm, and CMT was significantly correlated with final BCVA (r = 0.475, P < .001; Figure 3 ) .
|P1 Amplitude||P1 Implicit Time|
|Area 1||ϒ = −0.032||ϒ = 0.315|
|( P = .790)||( P = .008)|
|Area 2||ϒ = −0.113||ϒ = 0.450|
|( P = .353)||( P < .001)|
|P1 Amplitude||P1 Implicit Time|
|Area 1||ϒ = −0.047||ϒ = 0.425|
|( P = .696)||( P < .001)|
|Area 2||ϒ = −0.147||ϒ = 0.493|
|( P = .223)||( P < .001)|