When using the criterion of a CRT ≥ 300 µm, the present study found a very high incidence of PV-DME of 40.1%, with the majority subtype being DRT alone (66.2%), followed by CME alone (27.0%). This high incidence indicated that 40% of PDR patients may need further treatment, such as intravitreal drug injection, to regain visual function even after PPV. Although a direct comparison was difficult, the current reported incidence was higher than in other studies. A retrospective cohort study showed that of the 124 eyes that underwent PPV for PDR, 10 eyes (8.06%) presented postoperative ME.8 Sun et al conducted a study focused on the effect of preoperative injection on vitreous inflammatory cytokines and chemokines and showed that 20 eyes (33.90%) presented ME.12 In contrast with previous studies, we excluded patients with an unreadable preoperative OCT image and postoperative silicone oil tamponade to focus on the incidence of PV-DME. Patients frequently required the removal of the silicone oil within the 3-month follow-up, and the silicone oil frequently led to decreased macular thickness and microstructural changes.13, 14 Moreover, the definition of ME was different among studies. For example, Sun et al determined the presence of ME only by the macular structure (such as DRT, SRD, and CME), but not CRT measurement.
In the present study, the preoperative CRT and intraoperative ILM peeling were independent risk factors for newly developed PV-DME, while intraoperative TA injection prevented PV-DME. Doncel-Fernández et al reported that patients without known risk factors for developing CME and major surgical complication presented 9.08 times more probability to develop pseudophakic ME when with a preoperative central macular thickness > 260.5 µm.15 In the present study, a thicker preoperative CRT was also found to be associated with PV-DME.
Whether ILM peeling improves anatomical and functional outcomes has been extensively studied. It has been reported that additional removal of the ILM at the time of PPV may be helpful by removing all tractional forces, inhibiting the reproliferation of fibrous astrocytes,16 and preventing postoperative ERM formation.17 Additionally, the mean thickness of the ILM was significantly higher in diabetic patients, which may disturb the outflow of fluids accumulating into the retina and limit the diffusion of oxygen from the vitreous.18, 19 ILM removal improves exchanges with the oxygen-rich vitreous and promotes the discharge of liquid from the retina. A recent randomized clinical trial showed that the patients who underwent ILM peeling had a better best-corrected visual acuity and a lower incidence of ERM at 6 months than patients without ILM peeling.20 There was also a trend towards lower central macular thickness on OCT and a lower incidence of DME treatment postoperatively.20 Hu et al indicated that compared with vitrectomy alone, vitrectomy with ILM peeling had higher rate of CRT reduction in patients with DME.21 However, the result of our study showed that the patients who underwent ILM peeling during PPV were more likely to present PV-DME. Given the close proximity of the ILM to the inner retina and its interdigitation with Müller cell footplates, ILM peeling would unavoidably cause damage to Müller cells.22 Müller cells release factors that induce tight junction formation in retinal vessels. Abnormalities in Müller cells caused by ILM peeling probably affect this barrier property in the retinal vessels.23 In addition, the role of Müller cells in controlling the movement of water and ions allows them to buffer intraretinal increases of potassium ions. Ischemia and inflammation can alter the potassium channels of Müller cells and cause them to accumulate intracellular fluid.24 Moreover, we considered that during the procedure of ILM peeling, additional traction was applied to the retina, causing PV-DME. Chromophore toxicity, inflammation, and ganglion cell damage caused by ILM peeling, and the vulnerable vascular bed, lower integrity of the endothelium, and the loss of pericytes of patients with diabetes also contribute to PV-DME development.
The current first-line treatment for DME is the intravitreal injection of anti-VEGF agent, because of its apparent effect in reducing the VEGF level, and a certain effect on the decrease in the levels of downstream inflammatory cytokines.25, 26 However, unexpectedly, neither preoperative nor intraoperative intravitreal anti-VEGF injection was found to prevent PV-DME in the present study. It has been reported that after preoperative intravitreal anti-VEGF injection, the VEGF level dramatically decreased, whereas the influence on the vitreous inflammatory cytokines and chemokines in PDR was limited.12 Moreover, the decreased half-lives and increased clearance of anti-VEGF drugs in vitrectomized eyes probably also contribute.27 Despite the preoperative intravitreal injection of anti-VEGF agent, a series of cytokines, including IL-1β, TNF-α, CXCL9, G-CSF, MCP-1, and RANTES, remained at a high level and may lead to PV-DME.12 Yoshida et al reported that elevated MCP-1 and IL-6 levels may indicate prolonged inflammation even after successful vitrectomy, which was significantly associated with PV-DME.7 Rather than with an anti-VEGF agent, we found an intraoperative intravitreal TA injection effectively prevented PV-DME development in this study. Takamura et al reported that intravitreal TA injection at the end of vitrectomy may inhibit postoperative inflammation and ME, which was consistent with our study.28 Intraoperative intravitreal TA injection can effectively inhibit inflammatory cytokine production, leukostasis, and the phosphorylation of cell-junction proteins.23
For patients with preoperative DME, 78.3% of them retained DME during the 3 months after PPV in the present study. Fortunately, the mean CRT displayed a stable trend for reduction after PPV. A greater reduction of CRT in these patients was associated with a thicker preoperative CRT, which may be due to a greater potential for CRT reduction in these patients after surgery. However, intraoperative ILM peeling and intravitreal anti-VEGF or TA injection were not found to be associated with CRT reduction. Patients with DME frequently display retinal structural abnormalities, including ellipsoid zone disruption, external limiting membrane disruption, and blood-retinal barrier breakdown, compared with patients without DME.26, 29, 30 Therefore, ILM peeling appears insignificant for exacerbating the macular microstructure. Patients with preoperative DME frequently require a long time and multiple anti-VEGF treatments. Therefore, a single operative injection may not be sufficient.
Our study has some strengths. First, we reported the incidence of PV-DME and DME progression after PPV in a large sample size of PDR patients. Second, preoperative and postoperative macular morphology and CRT were precisely evaluated by OCT. This study also has several limitations. First, due to the retrospective design, there were considerable discrepancies in surgical methods and the follow-up time. Second, the incidence of PV-DME may be overestimated. Patients excluded from this study tended to have poor follow-up. These patients probably had a good visual outcome and less severe DME.