All MHs closed after the first surgery, as confirmed by SD-OCT. At the 1-month follow-up visit, three patients (cases 7, 15, and 16) had a small peripheral superior retinal detachment that was successfully treated with revision of vitrectomy, cryocoagulation of the retinal break, and gas tamponade (20% sulfur hexafluoride) without affecting the final anatomic and functional outcomes. The VA improved significantly from 0.2 ± 0.1 (mean ± SD) preoperatively to 0.7 ± 0.2 at T12 (
Table 1). The mean change in VA was 5.5 lines. Mean final VA did not differ significantly between cases with ICG (6 eyes) and Brilliant Blue G (13 eyes) -assisted peeling of ILM: 0.73 and 0.64, respectively (Welch two-sample
t-test;
P = 0.41). During follow-up, the improvement in VA was significant at T1, T6, and T9 compared with the preceding follow-up visits (
Fig. 2). The “time-complete” statistical model showed that approximately 80% of VA variance was explained by the elapse of time (
R 2 = 0.7991). The mean preoperative VA was approximately 0.25 after removing the patient effect. At T1 it increased to 0.40, at T6 it increased by another 0.1, and at T9 it reached 0.66. When only the effect of T1 was tested, it was still significant, but the time-reduced model accounted for a smaller amount of the variance (
R 2 = 0.6168) than did the time-complete model. The two models were significantly different (
F test,
P < 0.00001); therefore, the hypothesis that VA recovery is only a matter of mechanical realignment of photoreceptors after surgery was rejected.
With regard to the postoperative features of the outer retina, the ELM was restored more rapidly than the other anatomic features (
Fig. 3). It was continuous in 10 out of 19 eyes at the T1 follow-up visit (
Fig. 4) and in 15 eyes at T3. It is noteworthy that the ELM appeared already continuous in 15 of the 19 eyes with persistent outer foveal hyporeflective defects such as disruption/interruption of the IS/OS junction and the outer foveal cystic space during follow-up (
Figs. 4 5 6–
7).
Foveal hyporeflective defects changed over time (
Fig. 3). Disruption/interruption at the IS/OS junction gradually decreased although restoration of the integrity took longer than reestablishment of the ELM. Thus no eyes revealed a continuous IS/OS line at T1, only one eye out of 19 at T3, five eyes at T6, eight eyes at T9, and 10 eyes at T12. It is noteworthy that there were no eyes with a continuous IS/OS line and an interrupted ELM anytime during follow-up. Foveal cysts (
Fig. 4) were visible in three eyes at T1 and in eight eyes (
Figs. 5 6–
7) during later follow-up periods as a subsequent stage to the disruption/interruption of the IS/OS line. The cystic spaces gradually filled, resulting in a continuous IS/OS line in five of these 11 eyes, leaving six eyes with persistent cysts. The cyst in one eye was resolved by T6 (
Fig. 6F), in two eyes by T9, and in two eyes by T12 (
Fig. 7H). The ELM was always continuous and rectilinear overlying the cysts (
Figs. 4 5 6–
7). Again, the final integrity of the IS/OS line did not differ between cases with ICG (6 eyes) and Brilliant Blue G (13 eyes) -assisted peeling of ILM (Fisher's exact test;
P = 1).
Persistent foveal detachment occurred in only two eyes at T1 (cases 11 and 14;
Fig. 6). Both had a small preoperative MH (<400 μm), and both showed a continuous, curvilinear ELM at T1 and a continuous IS/OS junction by T6 and T3, respectively.
Moderately reflective foveal lesions that partially or completely replaced the normal foveal architecture were present in 10 out of 19 eyes during follow-up. In one of these, they were no longer visible by T6 (
Fig. 7), and in two they were no longer visible by T9 (
Fig. 8). All three of these eyes showed a normal ONL and a continuous IS/OS line at the last follow-up visit. We did not find any relationship between the basal hole diameter and the development of moderately reflective foveal lesions. Alterations of the outer nuclear layer were always present in the 10 eyes with moderately reflective foveal lesions. All 10 eyes with a recovered IS/OS line at T12 had a normal-appearing ONL (
Figs. 6 7–
8). In contrast, only two of the nine eyes with a final persistent outer foveal hyporeflective defect revealed the presence of an intact ONL (
Fig. 4).
Statistical analysis of the complete model of changes in the ELM, IS/OS, and ONL revealed that the OCT findings explained most of the VA variations during the 1-year follow-up (R 2 = 0.7963). When the ELM became continuous, the VA increased by approximately 0.12, and this was significant (P = 0.025). When the ONL was present, VA increased by 0.11, and this was also significant (P = 0.038). Finally, a continuous IS/OS junction increased VA by >0.3 (P < 0.01). Among the three reduced models, all of them were significantly different from the complete model: R 2 = 0.6602 for ELM (P < 0.01), R 2 = 0.7863 for ELM-IS/OS (P = 0.04), and R 2 = 0.6894 for ELM/ONL (P < 0.01). This suggests that the combined recovery of ELM, IS/OS, and ONL determined most of VA improvement.