Examinations
The 76 eyes of 66 patients underwent a general ophthalmic evaluation and preoperative examinations, including BCVA assessment, biomicroscopy slit-lamp examination, fundus examination by a panfundoscopic contact lens, and indirect ophthalmoscopy. Cross-sectional images of the macular region were acquired along the horizontal plane through the foveal center using the Spectralis OCT (Heidelberg Engineering, Heidelberg, Germany) or in some cases by using SD-OCT (RTVue-XR platform SD-OCT, Optovue Inc., Fremont, CA, USA), the axial lengths were measured using partial coherence laser interferometry (Zeiss IOL Master 700; Carl Zeiss Meditec AG, Oberkochen, Germany). The presence of RT-associated RRD was confirmed by indirect ophthalmoscopy and B-scan ultrasonography (A and B Ultrasound Unit, Quantel Medical, Du Bois Loli, Auvergne, France). A postoperative microstructural evaluation was performed using SD-OCT Spectralis OCT), SD-OCT RTVue-XR platform SD-OCT), and a swept-source (SS)-OCT device (Topcon Medical Systems, Inc., Oakland, NJ, USA). All OCT images were analyzed by three experienced retina specialists (co-authors) from the three participating institutions.
Surgical technique
A standard 23- or 25-gauge 3-port pars plana vitrectomy (Alcon Constellation Vision System, Alcon Labs, Fort Worth, TX, USA) was performed in all eyes under local anesthesia and sedation by one of the authors (MAQR). The vitrectomy was performed using a contact wide-angle viewing precorneal lens system (ROLS reinverted system Volk Medilex, Miami, FL, USA), the Wide Angle Viewing System (WAVS) with the resight non-contact lens (Carl Zeiss Meditec AG, Jena Germany), or recently in the last four cases the Zeiss ARTEVO 800 digital ophthalmic 3-D head-up microscope with the resight non-contact lens system, which was implemented as a hybrid mode (coaxial and 3-d HD 4K monitor), and integrated transoperative OCT allowed retinal structural intraoperative imaging analysis and real-time detection of ERM proliferation, enabling a more precise membrane dissection and stripping. In addition to central vitrectomy, our standard technique includes the use of diluted triamcinolone acetonide adjuvant (Kenalog 40 mg/mL; Bristol-Myers Squibb, New York, NY, USA) to identify and better visualize the vitreous face, vitreous base, and its posterior border, and safely perform an integral removal of the cortical face from the surface of the retina using a silicone-tipped cannula with active suction prior to perfluorocarbon liquid (PFCL) infusion and reattachment of the retina. The vitreous was carefully and precisely removed mainly from the anterior border of the GRT. The absolute release of the posterior retinal flap was checked, and showed that it tended to bend on itself, as was in many surgical cases. Then, the anterior retinal flap was trimmed in most cases to avoid anterior PVR and the release of vascular endothelial growth factor from this ischemic retina. The retina was reattached by a PFCL-assisted technique to effectively perform hydropneumatic retinal manipulation and assisted subretinal fluid (SRF) endodrainage; in all the cases, we performed meticulous drying of the subretinal space and RPE along the edges of the GRT, and in some cases, a direct perfluorocarbon-silicon oil exchange was performed to avoid posterior retinal slippage.
Phacoemulsification with in-the-bag intraocular lens implantation techniques was uneventfully performed in all phakic eyes. Subsequently, the vitreous base was shaved 360°, assisted with scleral depression or recently by doing self-scleral depression assisted with a single 27-gauge (disposable Eckard TwinLight Chandelier; Dutch Ophthalmic Research Center International, DORC International, Zuidland Netherlands), that was inserted into the trocar, and the non-contact WAVS with the resight lens was carefully applied in regions where the retina was detached and not broken or attached. This assisted scleral depression allowed the complete removal of the vitreous traction from the GRT and careful shaving and debulking of the vitreous base using mostly closed port duty cycle with high speed and low vacuum levels to perform a safer shaving of peripheral vitreous mainly over areas of the detached retina without producing iatrogenic retinal tears. Our young patients generally showed vitreous that was attached or only partially detached, and removing the core vitreous was relatively straightforward. However, separation of the posterior hyaloid and other areas of adherent vitreous in the periphery with a very mobile retina was technically intricate, especially when concurrent lattice degeneration was present. Injection of a PFCL is used to flatten and unfold to manipulate and immobilize the posterior retina. Once the retina was reattached without an SB, performing meticulous peripheral vitrectomy and ensuring a complete vitreous release with trimming of the anterior giant retinal flap, after all other retinal tears were identified and laser-treated along with the retinal lateral posterior extension tears (horns tears), were crucial. Additional benefits of the vitrectomy technique in these eyes were the removal of all vitreous opacities, attending to opacified lens capsules, and addressing the cases where significant macular ERM proliferation pre- or transoperatively was confirmed. Once the retina was completely attached and before the SB placement, for the ERM proliferation/removal, a transoperative surgical macular staining was performed using 0.15 mL of a 0.25 mg/mL (0.025%) diluted isomolar solution (pH 7.4) of Brilliant Blue G dye (BBG), to selectively stain and remove the internal limiting membrane (ILM) en-bloc with the ERM. For the ILM–ERM en bloc removal technique, a 23- or 25-gauge diamond-dusted membrane scraper and 25-gauge 0.44 ILM forceps (Grieshaber Revolution DSP ILM forceps; Alcon Labs, Fort Worth, TX, USA) and a 23- or 25-gauge Finesse ILM flex loop microinstrument (Grieshaber; Alcon Labs) to facilitate the removal of ERM and ILM from arcade to arcade. In cases where the removal was performed in two steps (double staining technique), trypan blue 0.15% ophthalmic solution (Membrane Blue; Dutch Ophthalmic, Exeter, NH, USA) was instilled under air to remove the ERM proliferations after washing the dye; in the second step, the ILM was stained with the aforementioned BBG dye and removed (two-step or double staining technique removal).
We performed SRF endodrainage very slowly by implementing a first-step fluid-to-fluid exchange over the edge of the GRT to avoid posterior retinal slippery and to remove viscous proteinaceous SRF, to reduce the extent of SRF and minimize the chance of trapped SRF before proceeding to an air-fluid exchange and continuing with SRF drainage. Once the retina was completely free of vitreous traction and completely reattached, 360° continuous argon laser endophotocoagulation in three to four rows, mainly at the peripheral edge of the circumferential retinal giant tear and lateral posterior radial extensions (horns tears) of the giant lesion, was thoroughly performed. To completely dry out the subretinal space, a second air-fluid exchange was performed, and as the last surgical step, a non-expandable bubble containing 15% perfluoropropane (C3F8) gas mixture or lighter than water silicon oil was used as a long-acting tamponade at the end of the procedure in all cases. Figure 2 shows surgical approaches (images a–l).
As part of the standardized selected technique of the author, and once the retina is fully reattached only with vitrectomy techniques and laser retinopexy to avoid posterior retinal slippage or radial folds in all eyes, a methodical, standardized, classical complementary low-lying SB surgical procedure was performed in the eyes with GRTs < 180° (by one of the authors, MAQR) consistent with traditional 504, 503, 360° round Lincoff episcleral sponges (Storz model E-5395-4) and oval foam silicon sponges (506 style S 1981-5 or 501 style S 1981-4) with the new designed profile (Labtician Ophthalmics, Inc., Oakville, Ontario, Canada) around the equator according with the axial length of the eye, or standard 240 circling silicon band (style 240/S-2987 by DORC) and 41 circling silicon band (style 41/S-2970 by DORC) in some cases, the SB was fixed with polyester 5-0 MERSILENE® Polyester Sutures, double-armed 3/8 circle spatulated needle suture (ETHICON, Johnson & Johnson, Brunswick, NJ, USA).
Statistical methodology
A two-tailed Student’s t-test was used to assess the normality of the distribution of continuous variables, and eyes were divided into three groups according to tear magnitude: group 1, tear <180°; group 2, tear 180–270°; and group 3, tear >270°. Variables were expressed as frequencies for discrete factors and mean values for continuous factors. Statistical comparisons were performed using the two-tailed chi-square test for categorical variables, and a two-tailed Student’s t-test for continuous variables was appropriate. Significance was set at p-value <0.05. A two-tailed paired samples t-test was conducted to examine whether the mean difference in preoperative BCVA in logMAR units and final postoperative BCVA in logMAR units was significantly different from zero. Repeated measures analysis of variance (ANOVA) was applied to see the before-and-after differences among the tear magnitude groups associated with other study factors. A linear regression analysis was conducted to assess whether CSFT in microns, DONFL defects, ELM line, tear magnitude, and ellipsoid band status (inner segment/outer segment band-IS/OS zone) significantly predicted final postoperative BCVA in logMAR units. A multivariate binary logistic regression analysis was performed to evaluate possible factors important for lower final postoperative BCVA in logMAR units. The Kaplan–Meier method was used to evaluate the general survival for final postoperative BCVA logMAR units between the divided eye groups. All statistical analyses were performed using SPSS 27 (IBM Corp., Armonk, NY, USA). The size of the effect observed in Student’s t-test was also determined. Cohen's d was calculated as 1.87, resulting in a large effect size according to the 1988 Cohen conventions.17 Cohen’s d gives a biased estimate of the population effect size, especially when the samples are <20.18 Therefore, Cohen’s d is referred to as the uncorrected effect size. The corrected value is Hedges’ g, which for our study was 1.65, indicating a relatively large effect size.
We determined that the power of the Mann–Whitney U tests and the Kruskal–Wallis tests that we used were very good (power = 86.5% and power = 99.0%) for the given sample size (n = 76) and for a medium effect size (Cohen's d = 0.5).
The Shapiro–Wilk normality test results showed that most numerical data followed a normal distribution (p<0.05); hence, the nonparametric Mann–Whitney U test was used to investigate the associations of the preoperative BCVA and postoperative BCVA in terms of the differences in medians of these variables.
Statistical results
The patients’ general and demographic data are presented in Table 1. Among the 76 eyes analyzed, the general prevalence of preoperative primary ERM proliferation was 4.78% (11 eyes), but only eight eyes (10.52%) with attached retina underwent ERM-ILM surgery; however, this prevalence was not statistically considered due to the heterogeneity of criteria used to define preoperative primary or postoperative secondary ERM proliferation and because the GRT associated RRD eyes are in high risk of developing surgical complication with additional surgical procedures.
Throughout the postoperative follow-up, additional surgical procedures were performed in nine (11.84%) eyes in which significant and symptomatic macular folds were detected due to posterior retinal slippage, vitrectomy review by retained PFCLs in five eyes (6.57%) and reoperation for a macular hole that was not detected in the first surgical time in one eye (1.31%) as aforementioned (Figure 3). According to our statistical results there was no significant difference in postoperative ERM proliferation incidence by tear magnitude.
Structural results (SD-OCT patterns)
To describe the structural postoperative SD-OCT findings (Table 1), we used the terminology proposed by the International Nomenclature for Optical Coherence Tomography Panel report,20 which were correlated with the functional findings. The statistical program yielded the following SD-OCT abnormalities in the macula-off GRT-associated RRD group: ellipsoid band disruption was observed in 57.9%, CSFT abnormalities in 74.7%, and ELM line alterations in 42.1% of the eyes. In the macula-on GRT-associated RRD group, ellipsoid band disruption was observed in 41.3%, CSFT abnormalities in 62.3%, and ELM line alterations in 51.9% of eyes (Figure 4a-j, Figure 5a-l). The differences between these categorical variables were not considered significant (p>0.05%).
Table 1
Comparison between tear magnitude groups to measure the associations from the other study variables
|
Sample
|
Tear <180°
|
Tear 180-270°
|
Tear>270°
|
p-value
|
N
|
|
N=76
|
N=44
|
N=21
|
N=11
|
|
|
Age
|
43.0 (13.0)
|
44.1 (12.8)
|
40.0 (11.3)
|
44.7 (16.8)
|
0.453
|
76
|
Sex:
|
|
|
|
|
0.239
|
76
|
Female
|
36 (47.4%)
|
24 (54.5%)
|
9 (42.9%)
|
3 (27.3%)
|
|
|
Male
|
40 (52.6%)
|
20 (45.5%)
|
12 (57.1%)
|
8 (72.7%)
|
|
|
Eye:
|
|
|
|
|
0.986
|
76
|
Right
|
41 (53.9%)
|
24 (54.5%)
|
11 (52.4%)
|
6 (54.5%)
|
|
|
Left
|
35 (46.1%)
|
20 (45.5%)
|
10 (47.6%)
|
5 (45.5%)
|
|
|
Preop lens status:
|
|
|
|
|
0.426
|
76
|
Pseudophakic
|
13 (17.1%)
|
8 (18.2%)
|
2 (9.52%)
|
3 (27.3%)
|
|
|
Phakic
|
63 (82.9%)
|
36 (81.8%)
|
19 (90.5%)
|
8 (72.7%)
|
|
|
Etiology:
|
|
|
|
|
<0.001
|
76
|
Idiopathic
|
42 (55.3%)
|
30 (68.2%)
|
11 (52.4%)
|
1 (9.09%)
|
|
|
Myopia
|
15 (19.7%)
|
7 (15.9%)
|
7 (33.3%)
|
1 (9.09%)
|
|
|
Blunt trauma
|
5 (6.58%)
|
3 (6.82%)
|
2 (9.52%)
|
0 (0.00%)
|
|
|
Marfan
|
5 (6.58%)
|
0 (0.00%)
|
0 (0.00%)
|
5 (45.5%)
|
|
|
Stickler
|
5 (6.58%)
|
1 (2.27%)
|
1 (4.76%)
|
3 (27.3%)
|
|
|
Marchesani
|
4 (5.26%)
|
3 (6.82%)
|
0 (0.00%)
|
1 (9.09%)
|
|
|
Macula at surgery:
|
|
|
|
|
0.219
|
76
|
Macula-on
|
15 (19.7%)
|
7 (15.9%)
|
7 (33.3%)
|
1 (9.09%)
|
|
|
Macula-off
|
61 (80.3%)
|
37 (84.1%)
|
14 (66.7%)
|
10 (90.9%)
|
|
|
preop logMAR
|
1.07 (0.61)
|
1.13 (0.57)
|
0.82 (0.53)
|
1.30 (0.77)
|
0.063
|
76
|
Months of follow-up
|
41.1 (29.1)
|
38.0 (27.0)
|
51.2 (35.3)
|
33.9 (20.2)
|
0.157
|
76
|
Final postop logMAR
|
0.56 (0.26)
|
0.55 (0.27)
|
0.68 (0.24)
|
0.35 (0.15)
|
0.002
|
76
|
Recurrent RRD
|
18 (23.7%)
|
8 (18.2%)
|
8 (38.1%)
|
2 (18.2%)
|
0.253
|
76
|
Recurrence cause:
|
|
|
|
|
0.363
|
76
|
No
|
58 (76.3%)
|
36 (81.8%)
|
13 (61.9%)
|
9 (81.8%)
|
|
|
PVR
|
12 (15.8%)
|
4 (9.09%)
|
6 (28.6%)
|
2 (18.2%)
|
|
|
PVR+choroidals
|
5 (6.58%)
|
3 (6.82%)
|
2 (9.52%)
|
0 (0.00%)
|
|
|
Undetected macular hole
|
1 (1.32%)
|
1 (2.27%)
|
0 (0.00%)
|
0 (0.00%)
|
|
|
Tamponade type:
|
|
|
|
|
0.020
|
76
|
C3F8
|
53 (69.7%)
|
36 (81.8%)
|
12 (33.3%)
|
5 (9.09%)
|
|
|
|
|
|
|
|
|
|
Silicon
|
23 (30.3%)
|
8 (18.2%)
|
9 (42.9%)
|
6 (54.5%)
|
|
|
Additional surgery:
|
|
|
|
|
0.300
|
76
|
No
|
56 (73.7%)
|
35 (79.5%)
|
12 (57.1%)
|
9 (81.8%)
|
|
|
Silicon removal
|
2 (2.63%)
|
1 (2.27%)
|
1 (4.76%)
|
0 (0.00%)
|
|
|
Vitrectomy revision
|
18 (23.7%)
|
8 (18.2%)
|
8 (38.1%)
|
2 (18.2%)
|
|
|
Postop ERMs
|
18 (23.7%)
|
12 (27.3%)
|
4 (19.0%)
|
2 (18.2%)
|
0.745
|
76
|
ERM surgery:
|
|
|
|
|
0.899
|
76
|
ERM removal
|
8 (10.5%)
|
6 (13.6%)
|
1 (4.76%)
|
1 (9.09%)
|
|
|
No
|
58 (76.3%)
|
32 (72.7%)
|
17 (81.0%)
|
9 (81.8%)
|
|
|
Vitrectomy revision + ERM removal
|
10 (13.2%)
|
6 (13.6%)
|
3 (14.3%)
|
1 (9.09%)
|
|
|
Other complications:
|
|
|
|
|
0.451
|
76
|
Macular fold
|
9 (11.8%)
|
5 (11.4%)
|
4 (19.0%)
|
0 (0.00%)
|
|
|
No
|
61 (80.3%)
|
36 (81.8%)
|
16 (76.2%)
|
9 (81.8%)
|
|
|
Retained PFCL
|
5 (6.58%)
|
2 (4.55%)
|
1 (4.76%)
|
2 (18.2%)
|
|
|
Undetected macular hole
|
1 (1.32%)
|
1 (2.27%)
|
0 (0.00%)
|
0 (0.00%)
|
|
|
CSFT microns
|
209 (28.2)
|
210 (27.6)
|
199 (24.3)
|
228 (29.4)
|
0.019
|
76
|
Foveal contour:
|
|
|
|
|
0.067
|
76
|
Normal
|
23 (30.3%)
|
14 (31.8%)
|
3 (14.3%)
|
6 (54.5%)
|
|
|
Abnormal
|
53 (69.7%)
|
30 (68.2%)
|
18 (85.7%)
|
5 (45.5%)
|
|
|
Ellipsoid band:(IS/OS)
|
|
|
|
|
0.845
|
76
|
Normal
|
48 (63.2%)
|
28 (63.6%)
|
14 (66.7%)
|
6 (54.5%)
|
|
|
Disrupted
|
28 (36.8%)
|
16 (36.4%)
|
7 (33.3%)
|
5 (45.5%)
|
|
|
DONFL defects:
|
|
|
|
|
0.296
|
76
|
Absent
|
53 (69.7%)
|
30 (68.2%)
|
17 (81.0%)
|
6 (54.5%)
|
|
|
Present
|
23 (30.3%)
|
14 (31.8%)
|
4 (19.0%)
|
5 (45.5%)
|
|
|
ELM line:
|
|
|
|
|
0.635
|
76
|
Normal
|
47 (61.8%)
|
25 (56.8%)
|
14 (66.7%)
|
8 (72.7%)
|
|
|
Disrupted
|
29 (38.2%)
|
19 (43.2%)
|
7 (33.3%)
|
3 (27.3%)
|
|
|
C3F8, perfluoropropane; CSFT, central subfoveal thickness; DONFL, dissociated optic nerve fiber layer; ELM, external limiting membrane; ERM, epiretinal membrane; IS/OS, internal segment/external segment; PFCL, perfluorocarbon liquid; Postop, postoperative; Preop, preoperative; PVR, proliferative vitreoretinopathy; RRD, rhegmatogenous retinal detachment |
Functional results
The mean preoperative BCVA was 1.87±0.15 logMAR vs the mean postoperative BCVA of 0.35±0.21 in logMAR units with a p value < 0.05 (p = 0.01) (Table 1).
The Spearman's rank correlation coefficient test showed a moderate negative correlation (rho = −0.53, p < 0.01) of the postoperative BCVA (logMAR) with the CSFT in microns (Figure 6).
A chi-square test of independence was conducted to examine whether tear magnitude and etiology were independent. The results of the chi-square were significant based on an alpha value and p < .001, suggesting a relationship between tear magnitude and etiology. An ANOVA was conducted to determine whether there were significant differences in the final postoperative BCVA logMAR units by tear magnitude.
There were significant differences in final postoperative BCVA in logMAR units among the levels of tear magnitude (Table 1). The eta squared was 0.16, indicating tear magnitude explained approximately 16% of the variance in final postoperative BCVA. Regarding the main effect of tear magnitude, the mean final postoperative BCVA for tear magnitude was significantly larger than for tear >270º (p =.048). For the main effect of tear magnitude, the mean of final postoperative BCVA for tear 180-270º was significantly larger than for tear>270º ( p = .001). No other significant effects were found. However, paired t-tests were calculated between each pair of measurements to further examine the differences among the variables. Tukey pairwise comparisons (post hoc) were conducted for all significant effects, based on an alpha of 0.05.
The chi-square test results showed a significant relationship between tear magnitude and tamponade type based on an alpha value of 0.05, χ2(2) =7.69, p = 0.21.
A linear regression analysis was conducted to assess whether CSFT, DONFL defects, ELM line, tear magnitude, and ellipsoid band status (IS/OS) significantly predicted final postoperative BCVA in logMAR units. The Shapiro–Wilk test determined that results were not significant based on an alpha value of 0.05, W = 0.97, p = .09120. This result suggests the possibility that the residuals of the model were produced by a normal distribution, indicating that the assumption of normality was met.
Multicollinearity
High Variance inflation factors (VIFs) indicate increased effects of multicollinearity in the model. VIFs > 5 are cause for concern, whereas VIFs of 10 should be considered the maximum upper limit.22 VIFs for all predictors in the regression model were < 10. Table 2 presents the VIFs for each predictor in the model.
Table 2
Variance inflation factors for CSFT in microns, DONFL defects, ELM line, TEAR MAGNITUDE, and ELLIPSOID band
Variable
|
VIF
|
CSFT in microns
|
1.23
|
DONFL defects
|
1.29
|
ELM line
|
1.51
|
Tear magnitude
|
1.23
|
Ellipsoid band (IS/OS zone)
|
1.47
|
CSFT, central subfoveal thickness; DONFL, dissociated optic nerve fiber layer; ELM, external limiting membrane; IS/OS, internal segment/external segment; VIF, variance inflation factors |
The results of the linear regression model were significant (p < .001) indicating that approximately 37% of the variance in final postoperative BCVA in logMAR units were explained by CSFT in microns, DONFL defects, ELM line, tear magnitude, and ellipsoid band status (IS/OS zone). CSFT significantly predicted the final postoperative BCVA (p < .001). This indicated that, on average, a one-unit increase of CSFT in microns decreased the value of final postoperative BCVA in logMAR by 0.00 units. The category of DONFL defects did not significantly predict the final postoperative BCVA (p = .898). Based on this sample, this suggests that moving from the absent to presence of DONFL defects did not have a significant effect on the mean final postoperative BCVA. The disrupted category of the ELM line did not significantly predict final postoperative BCVA (p = .476). Moreover, this suggests that moving from the normal to disrupted category of ELM did not have a significant effect on the mean final postoperative BCVA. The tear 180–270° category of tear magnitude did not significantly predict final postoperative BCVA (p = .120), suggesting that moving from the tear 270° category of tear magnitude did not significantly predict final postoperative BCVA (p = .157) or have a significant effect on the mean final postoperative BCVA (Figure 7). The disrupted category of ellipsoid band did not significantly predict the final postoperative BCVA (p = .658), suggesting that moving from the normal to disrupted category of ellipsoid band status (EZ) did not have a significant effect on the mean final postoperative BCVA. Table 3 summarizes the results of regression model.
Table 3
Regression model parameter estimates to measure the association of the OCT biomarkers.
Variable
|
B
|
SE
|
90% CI
|
β
|
t
|
p
|
(Intercept)
|
1.46
|
0.21
|
[1.11, 1.81]
|
0.00
|
6.96
|
< .001
|
CSFT in microns
|
−0.00
|
0.00
|
[−0.01, −0.00]
|
−0.48
|
−4.55
|
< .001
|
DONFL (present)
|
0.01
|
0.06
|
[−0.10, 0.11]
|
0.01
|
0.13
|
.898
|
ELM line (disrupted)
|
0.05
|
0.06
|
[−0.06, 0.15]
|
0.08
|
0.72
|
.476
|
Tear magnitude (tear 180–270°)
|
0.09
|
0.06
|
[−0.01, 0.19]
|
0.16
|
1.57
|
.120
|
Tear magnitude (tear>270°)
|
−0.11
|
0.08
|
[−0.24, 0.02]
|
−0.15
|
−1.43
|
.157
|
Ellipsoid band (disrupted)
|
0.03
|
0.06
|
[−0.08, 0.13]
|
0.05
|
0.44
|
.658
|
Note: Results: F (6,69) = 6.77, p < .001, R2 = 0.37
CSFT, central subfoveal thickness; DONFL, dissociated optic nerve fiber layer; CI, confidence interval; ELM, external limiting membrane; OCT, optical coherence tomography; SE, standard error
|
The results of the two-tailed paired samples t-test was significant based on an alpha value of 0.05, (p < .001), indicating that the null hypothesis could be rejected. Thus, the difference in the mean preoperative BCVA in logMAR units and the mean of final postoperative BCVA in logMAR units was significantly different from zero. The mean preoperative logMAR was significantly higher than the mean final postoperative BCVA in logMAR units. The results are presented in Table 4. A bar plot of the means is presented in Figure 8.
Table 4
Two-tailed paired samples t-test for the difference between preoperative logMAR and final postoperative logMAR
PREOP logMAR
|
FINAL POSTOP logMAR
|
|
|
|
M
|
SD
|
M
|
SD
|
t
|
p
|
d
|
1.07
|
0.61
|
0.56
|
0.26
|
6.88
|
<0.001
|
0.79
|
Note: N = 76. Degrees of freedom for t-statistic = 75. d represents Cohen's d.
M, mean; POSTOP, postoperative; PREOP, preoperative; SD, standard deviation
|
To further understand the significant differences between preoperative BCVA and final postoperative BCVA, a repeated measures analysis of covariance (ANCOVA) with one within-subjects factor was conducted to determine whether significant differences existed between preoperative BCVA and final postoperative BCVA after controlling for tear magnitude and tamponade type. Table 5 shows the most significant factors that controlled the differences between preoperative BCVA and final postoperative BCVA.
Table 5
Repeated measures ANCOVA results for the difference between preoperative logMAR and final postoperative logMAR
Source
|
df
|
SS
|
MS
|
F
|
p
|
ηp2
|
Between-subjects
|
|
|
|
|
|
|
Tear magnitude
|
2
|
0.50
|
0.25
|
1.17
|
.317
|
0.03
|
Tamponade type
|
2
|
1.73
|
0.86
|
4.01
|
.022
|
0.10
|
Residuals
|
71
|
15.29
|
0.22
|
|
|
|
Within-subjects
|
|
|
|
|
|
|
Within factor
|
1
|
6.88
|
6.88
|
39.19
|
<.001
|
0.36
|
Tear magnitude: within.factor
|
2
|
3.04
|
1.52
|
8.66
|
<.001
|
0.20
|
Tamponade type: within.factor
|
2
|
0.60
|
0.30
|
1.71
|
.188
|
0.05
|
Residuals
|
71
|
12.46
|
0.18
|
|
|
|
ANCOVA, analysis of covariance; postop, postoperative; preop, preoperative; SS, sum of squares; MS, mean square |
The results were examined based on an alpha of 0.05. Tear magnitude was not significantly related to preoperative BCVA and final postoperative BCVA (p = .317). Tamponade type was significantly related to preoperative BCVA and final postoperative BCVA (p = .022). The main effect for the within-subjects factor was significant (p < .001), indicating significant differences between the values of preoperative BCVA and final postoperative BCVA after controlling for tear magnitude and tamponade type. The interaction effect between the within-subjects factor and tear magnitude was significant, (p < .001), indicating that the relationship between preoperative BCVA and final postoperative BCVA differed significantly between the levels of tear magnitude (Figure 9). The interaction effect between the within-subjects factor and tamponade type was not significant (p = .188), indicating that the relationship between preoperative BCVA and final postoperative BCVA was similar between the levels of tamponade type. Table 5 presents the ANCOVA results.
A binary logistic regression analysis was conducted to examine whether months of follow-up, tear magnitude, CSFT in microns, status of the macula at surgery, recurrent RRD, foveal contour appearance, ellipsoid zone (EZ), DONFL defects, and ELM line had a significant effect on the odds of observing the 20/300 or worse category of postoperative BCVA. The reference category for final BCVA was between 20/50 to 20/200.
The model was evaluated based on an alpha of 0.05 (Table 6). The overall model was significant (p < .001), suggesting that months of follow-up, tear magnitude, CSFT in microns, status of the macula at surgery, recurrent RRD, postoperative foveal contour appearance, EZ status, DONFL defects, and ELM line appearance had a significant effect on the odds of observing the 20/300 or worse category of postoperative BCVA. McFadden's R-squared was calculated to examine the model fit, where values > .2 are indicative of models with excellent fit.22 The McFadden R-squared value calculated for this model was 0.49. The regression coefficient for months of follow-up was significant (p = .020), indicating that for a one-unit increase in months of follow-up, the odds of observing the 20/300 or worse category of postoperative BCVA increased by approximately 4%. Additionally, the regression coefficient for macula-off at surgery was significant (p = .04), indicating that for a one-unit increase in macula-off at surgery, the odds of observing the 20/300 or worse category of postoperative BCVA increased by approximately 638%. However, the regression coefficient for foveal contour appearance was significant (p < .001), indicating that for a one-unit increase in foveal contour abnormality, the odds of observing the 20/300 or worst category of postoperative BCVA also increased.
Table 6
Multivariate logistic regression results
Model coefficients – logMAR post
|
|
95% Confidence interval
|
Predictor
|
Estimate
|
SE
|
Z
|
p
|
Odds ratio
|
Lower
|
Upper
|
Intercept
|
−1.5319
|
3.9450
|
−0.388
|
0.698
|
0.216
|
9.48e–5
|
492.81
|
Months of follow-up
|
0.0421
|
0.0184
|
2.294
|
0.022
|
1.043
|
1.0062
|
1.08
|
Tear magnitude:
|
|
|
|
|
|
|
|
Tear 180–270°–tear <180°
|
0.7277
|
0.8821
|
0.825
|
0.409
|
2.070
|
0.3675
|
11.66
|
Tear>270°–tear <180°
|
−1.0039
|
1.2716
|
−0.789
|
0.430
|
0.366
|
0.0303
|
4.43
|
CSFT in microns
|
−0.0232
|
0.0157
|
−1.476
|
0.140
|
0.977
|
0.9474
|
1.01
|
Macula at surgery:
|
|
|
|
|
|
|
|
Off–macula-on
|
1.9408
|
0.9464
|
2.051
|
0.040
|
6.964
|
1.0896
|
44.51
|
Foveal contour:
|
|
|
|
|
|
|
|
Abnormal–normal
|
4.3954
|
1.2975
|
3.387
|
< .001
|
81.074
|
6.3741
|
1031.19
|
Ellipsoid band:
|
|
|
|
|
|
|
|
Disrupted–normal
|
0.4868
|
0.9474
|
0.514
|
0.607
|
1.627
|
0.2541
|
10.42
|
DONFL defects:
|
|
|
|
|
|
|
|
Present–absent
|
−0.3840
|
1.0001
|
−0.384
|
0.701
|
0.681
|
0.0959
|
4.84
|
ELM line:
|
|
|
|
|
|
|
|
Disrupted–normal
|
−0.4387
|
0.9702
|
−0.452
|
0.651
|
0.645
|
0.0963
|
4.32
|
Note. Estimates represent the log odds of "logMAR post = 20/300 or worse" vs. "logMAR post = between 20/50 and 20/200"
CSFT, central subfoveal thickness; DONFL, dissociated optic nerve fiber layer; ELM, external limiting membrane; SE,standard error
|
A Cox proportional hazards model was used to determine whether tear magnitude had a significant effect on the hazard of final postoperative BCVA categories. The 20/300 or worse category of final postoperative BCVA was used to indicate survival, while the 20/50 to 20/200 group category was used to represent a hazard event. The model’s results were significant (p = .006), indicating that tear magnitude adequately predicted the hazard of the final postoperative BCVA (Table 7).
The coefficient for the tear 180–270° group category of tear magnitude was significant (p = .027), indicating that at any particular time, an observation in the tear 180–270° category had a hazard that is 0.33 times as large as the tear 270° was not significant ( p = .151), indicating that being in the tear >270° category of tear magnitude did not have a significant effect on the hazard of final postoperative BCVA (Table 7).
Table 7
Cox proportional hazards regression coefficients for tear magnitude
Variable
|
B
|
SE
|
95% CI
|
z
|
p
|
HR
|
Tear magnitude (tear 180–270°)
|
−1.10
|
0.50
|
[−2.07, −0.12]
|
−2.21
|
.027
|
0.33
|
Tear magnitude (tear >270°)
|
0.57
|
0.40
|
[−0.21, 1.35]
|
1.43
|
.151
|
1.77
|
CI, confidence interval; HR, hazard ratio; SE, standard error |
A calculation of survival probability according to the tear magnitude was performed and the results are shown in risk tables 7 and 8.
Kaplan–Meier survival plots
A Kaplan–Meier survival probability plot was included for tear magnitude. Each plot represents the survival probabilities of the different groups over time (Figure 10).
Table 8
Risk table for tear magnitude = tear 180–270°
Quantile
|
Time
|
No. at risk
|
Survival probability
|
Std. error
|
0%
|
6.000
|
21
|
1.000
|
0.000
|
25%
|
24.000
|
17
|
0.895
|
0.079
|
50%
|
38.000
|
11
|
0.763
|
0.139
|
75%
|
77.000
|
6
|
0.763
|
0.139
|
100%
|
118.000
|
1
|
0.610
|
0.264
|
No. number; Std., standard |
Table 9
Risk table for tear magnitude = tear >270°
Quantile
|
Time
|
No. at risk
|
Survival probability
|
Std. error
|
0%
|
11.000
|
11
|
0.909
|
0.095
|
25%
|
15.000
|
9
|
0.727
|
0.185
|
50%
|
28.000
|
7
|
0.519
|
0.302
|
75%
|
39.000
|
4
|
0.312
|
0.474
|
100%
|
77.000
|
1
|
0.000
|
Inf
|
No. number; Std., standard |
Additionally, we used generalized linear models (GLM) to further investigate potential associations of the postoperative BCVA with the other variables. The selected by “step” generalized additional model for the postoperative BCVA revealed that the postoperative BCVA was significantly associated with the number of follow up months (p < 0.01), the CSFT (p < 0.01), and the “Normal foveal contour” (p < 0.01), when adjusting for cofounders with multivariable analyses (Table 10).
Table 10
The “best” generalized linear model results of the postoperative BCVA in logMAR units.
|
Estimate
|
Std. Error
|
t value
|
p value
|
significance
|
(Intercept)
|
1.35736
|
0.1802913
|
7.529
|
1.45E−10
|
***
|
Male sex
|
−0.065232
|
0.0440148
|
−1.482
|
0.14288
|
|
Macula-on at surgery
|
−0.0822018
|
0.0561708
|
−1.463
|
0.14789
|
|
Follow-up in months
|
0.0021417
|
0.0007667
|
2.793
|
0.00675
|
**
|
Recurrent RRD
|
0.0846707
|
0.0526205
|
1.609
|
0.11216
|
|
CSFT in microns
|
−0.0037141
|
0.000813
|
−4.569
|
2.09E−05
|
***
|
Normal foveal contour
|
−0.2649488
|
0.049419
|
−5.361
|
1.04E−06
|
***
|
The significant variables are in bold writing and marked with *.
BCVA, best-corrected visual acuity; CSFT, central subfoveal thickness; E, error; RRD, rhegmatogenous retinal detachment; Std., standard
|