Post vitrectomy alterations of retinal microcirculation in complicated diabetic retinopathy with and without silicone oil

Abstract

BACKGROUND

Vitrectomy combined with silicone oil (SO) tamponade is a common treatment for complicated diabetic retinopathy (DR). However, changes of the fragile retinal microcirculation in patients with DR after SO tamponade remain unclear. In recent years, the imaging technique of optical coherence tomography angiography (OCTA) has revolutionized the evaluation of retinal microcirculation by enabling in vivo visualization and quantitative assessment of the retinal vascular networks noninvasively.

AIM

To explore changes of retinal microcirculation and visual function in patients with DR after SO tamponade using OCTA.

METHODS

We retrospectively analyzed data from patients with DR who underwent vitrectomy at our institution. Two sequential comparative analyses were performed: A crosssectional comparison between the SO and Air groups, followed by a longitudinal comparison between the SO and SO removal (SOR) groups. OCTA was used to measure retinal thickness (RT), vascular density (VD) in the superficial and deep capillary plexuses (DCP), and the foveal avascular zone (FAZ) area. Best-corrected visual acuity (BCVA) was also recorded.

RESULTS

Compared with the Air group, the SO group showed greater parafoveal and perifoveal RT at both 1 month and 3 months, lower parafoveal and perifoveal VD-DCP at 3 months, and larger FAZ areas at both time points. Correspondingly, the SO group also demonstrated worse BCVA. At 3 months after SOR, parafoveal and perifoveal VD in the DCP increased, accompanied by an improvement in BCVA compared with that during SO tamponade. Correlation analysis revealed that parafoveal and perifoveal RT and FAZ area were positively correlated with logMAR BCVA, whereas parafoveal and perifoveal VD in the DCP were negatively correlated.

CONCLUSION

SO was associated with altered retinal microcirculation and worse visual outcomes in patients with complicated DR after vitrectomy. Only partial improvement in some perfusion parameters and vision was observed after SOR.

Key Words: Optical coherence tomography angiography; Silicone oil; Vitrectomy; Diabetic retinopathy; Retinal microcirculation

Core Tip: This study explored changes of retinal microcirculation in patients with complicated diabetic retinopathy after silicone oil (SO) tamponade using optical coherence tomography angiography (OCTA). SO tamponade was associated with decreased deep retinal vascular density (VD), increased retinal thickness, and enlargement of the foveal avascular zone (FAZ). After SO removal, deep retinal VD partially increased, whereas FAZ enlargement persisted. OCTA may serve as a valuable tool for postoperative follow-up and clinical management of patients after vitrectomy.

INTRODUCTION

Diabetes mellitus (DM) has become a major global public health challenge, and its prevalence has steadily increased in recent decades. Among its numerous complications, diabetic retinopathy (DR) is one of the most common and vision-threatening conditions[1]. It is widely recognized as the leading cause of irreversible blindness among working-age adults worldwide[2,3]. The core pathological mechanism underlying DR involves progressive retinal microvascular damage, resulting in retinal ischemia and hypoxia[4,5].

For patients with advanced DR, pars plana vitrectomy (PPV) remains the first-line surgical approach to prevent further visual decline or blindness[6,7]. Although long-acting gas tamponades are available, silicone oil (SO) remains indispensable in complex cases such as tractional retinal detachment and severe preretinal proliferation[8,9]. However, complicated DR is characterized by pre-existing severe retinal microvascular damage and ischemia. Changes of the fragile retinal microcirculation after SO tamponade, as well as after SO removal (SOR), has not been fully elucidated. Dynamic and comprehensive investigations on this topic are still scarce.

In recent years, optical coherence tomography angiography (OCTA) has revolutionized the evaluation of retinal microcirculation. OCTA enables noninvasive, in vivo visualization and quantitative assessment of the retinal vascular networks, including the superficial capillary plexus (SCP) and deep capillary plexuses (DCP). It also provides precise measurements of key parameters such as foveal avascular zone (FAZ) area, vascular density (VD), and retinal thickness (RT), offering detailed insights into both the structural and functional status of the retinal microvasculature[10-12].

The present study used OCTA to investigate changes in retinal microcirculatory parameters in patients with DR who underwent PPV with or without SO tamponade, as well as after SOR. Through this analysis, we aimed to explore the correlation between SO and retinal perfusion as well as visual function recovery.

MATERIALS AND METHODS

Research participants

This study was approved by the Institutional Review Board of the First Affiliated Hospital of Soochow University and was conducted in accordance with the tenets of the Declaration of Helsinki. Written informed consent was obtained from all participants after a detailed explanation of the study procedures, potential risks, and the retrospective use of their clinical and imaging data for research purposes.

After ethical approval, we retrospectively reviewed the medical records of patients who underwent 23-gauge PPV at our institution between January 2024 and December 2024.

The inclusion criteria were as follows: (1) Diagnosis of proliferative DR requiring surgical intervention; (2) Age between 30 years and 75 years; and (3) Preoperative best-corrected visual acuity (BCVA) ranging from light perception to 0.2 (logMAR). The exclusion criteria were as follows: (1) Preexisting ocular diseases, including glaucoma, keratitis, uveitis, retinal vascular occlusion, ischemic optic neuropathy, or prior ocular trauma; (2) High myopia with refractive error > -10.0 diopters (D); (3) History of previous vitrectomy; (4) Preoperative significant tractional retinal detachment; (5) Uncontrolled systemic disease, defined as systolic blood pressure > 160 mmHg, fasting blood glucose > 10.0 mmol/L, or a recent history (within 3 months) of stroke, myocardial infarction, or coronary artery stenting; (6) Severe rheumatic or connective tissue disease; (7) Failure to complete the 3-month postoperative follow-up or inability to undergo SOR at 3 months postoperatively; and (8) Severe postoperative complications, including persistent intraocular pressure (IOP) elevation lasting > 1 month, recurrent vitreous hemorrhage, retinal redetachment, postoperative epiretinal membrane formation, or endophthalmitis.

Patients were categorized into three groups based on the intraocular tamponade material and follow-up status: (1) Air group, comprising patients who underwent PPV with intravitreal air tamponade; (2) SO group, comprising patients who underwent PPV with intravitreal SO tamponade; and (3) SOR group, consisting of a subset of patients from the SO group who underwent SOR 3 months postoperatively and were included in the longitudinal comparison. The participant flow diagram was presented in Figure 1.

Figure 1 Participant flow diagram. This flowchart depicts the screening, eligibility assessment, and exclusion of eyes enrolled in this retrospective observational study. DR: Diabetic retinopathy; PPV: Pars plana vitrectomy; OCTA: Optical coherence tomography angiography; SO: Silicone oil; SOR: Silicone oil removal.

Vitrectomy procedure and follow-up

All surgeries were performed by a single experienced vitreoretinal surgeon. Preoperative ocular and systemic contraindications were excluded in all patients. Surgery was performed under local anesthesia using a standard 23-gauge three-port PPV technique, which included complete vitrectomy, meticulous removal of the preretinal proliferative membranes, and panretinal photocoagulation.

SO (Oxane 5700, Bausch + Lomb, Rochester, NY, United States) tamponade (4.5-5.0 mL) was used in eyes with significant tractional proliferation, or fibrovascular membranes. Sterile air was used as the intraocular tamponade in eyes without substantial traction or retinal detachment. Eyes with visually significant cataracts at the time of PPV or SOR underwent PPV combined phacoemulsification and intraocular lens implantation.

Postoperatively, patients in the Air group maintained facedown positioning for 1 week, whereas those in the SO group maintained facedown positioning for 2 weeks. All patients received topical antibiotic drops for 2 weeks and topical corticosteroid drops for 1 month. Routine follow-up visits were scheduled at 2 weeks, 1 month, 2 months, and 3 months after surgery. Each visit included measurement of BCVA and IOP, slit-lamp biomicroscopy of the anterior segment, wide-field fundus photography, and OCT or OCTA imaging.

Patients in the SO group maintained SO tamponade for 3 months. If the retina remained stably and no major complications were detected, SOR was subsequently performed. SO was removed via PPV through the pars plana, followed by intraocular filling with balanced salt solution.

OCTA imaging protocol

All patients underwent OCTA examinations 1 month and 3 months after PPV or SOR. Imaging was performed after pupillary dilation using a spectral-domain OCTA system (RTVue-XR Avanti with AngioVue software; OptoVue Inc., Fremont, CA, United States). The OCTA examiner and image analyst were blinded to group assignment. A 6 mm × 6 mm macular scan centered on the fovea was obtained for each eye. Split-spectrum amplitude-decorrelation angiography was used to generate the angiographic data.

All scans were manually reviewed against automaticallydefined segmentation boundaries, with manual boundary adjustments performed when necessary. The builtin software defined the foveal (0-0.5 mm), parafoveal (0.5-1.5 mm), and perifoveal (1.5-3.0 mm) regions as concentric circular/annular zones centered on the fovea. Automatic segmentation algorithms were used to visualize and quantify RT, the SCP, the DCP, and the FAZ (Figure 2). Only images with a signal strength of ≥ 7 and no motion artifacts were included in the analysis. No statistical difference in signal strength was observed between the groups. All acquired images were finally reviewed by the authors for adequate image quality.

Figure 2 Representative optical coherence tomography angiography images illustrating the measurement of retinal microcirculatory parameters. A: Measurement of retinal thickness; B: Measurement of the foveal avascular zone area; C: Measurement of vascular density (VD) in the superficial capillary plexus (SCP); D: Measurement of VD in the deep capillary plexus (DCP); E and F: Representative B-scan images showing the segmentation boundaries of the SCP and DCP.

Statistical analysis

Statistical analyses were performed using SPSS version 26.0 (IBM, Armonk, NY, United States). Categorical variables were analyzed using Fisher’s exact test. Comparisons of OCTA parameters and BCVA between the SO group and the Air group were performed using linear mixed-effects models, while paired linear mixed-effects models were adopted for comparisons between the SO group and the SOR group. All models were adjusted for the following clinical confounders: Duration of diabetes, blood glucose level, glycated hemoglobin A1c (HbA1c), and surgical modalities. Post-hoc pairwise comparisons were adjusted using the Bonferroni correction for multiple comparisons to mitigate type I error inflation. Pearson’s correlation analysis was conducted to evaluate the relationship between OCTA parameters and BCVA. A two-sided P value of < 0.05 was considered statistically significant.

RESULTS

General information

A total of 58 eyes with DR were enrolled: 34 in the Air group and 24 in the SO group. 8 eyes in the Air group and 4 patients in the SO group underwent combined PPV (PPV combined phacoemulsification), whereas the remaining patients underwent simple PPV (PPV alone). In the SO group, 20 eyes underwent planned SOR 3 months postoperatively and completed follow-up, forming the SOR subgroup. 3 eyes underwent phacoemulsification at the time of SOR (combined PPV).

The baseline patient characteristics are listed in Table 1. The three groups were comparable in terms of age, sex, IOP, axial length, and lens status (all P > 0.05). No significant differences were found in the prevalence of preoperative epiretinal membrane, preoperative retinal laser treatment, preoperative anti-vascular endothelial growth factor therapy (Lucentis^®, Novartis; 0.5 mg/0.05 mL), and types of PPV (all P > 0.05). There were also no significant differences among the three groups in the duration of DM or preoperative blood glucose levels or HbA1c (all P > 0.05).

Table 1 General characteristics of the patients in the three groups.

Variable	Air group (A, n = 34)	SO group (B, n = 24)	SOR group (C, n = 20)	P (A vs B)	P (B vs C)
Sex (M/F)	19/15	12/12	9/11	0.794	1.000
Age (years)	49.8 ± 8.1	50.6 ± 8.7	50.8 ± 10.3	0.734	0.944
Duration of DM (years)	9.2 ± 2.5	9.6 ± 3.2	9.5 ± 2.7	0.605	0.910
Blood glucose (mmol/L)	7.0 ± 0.5	7.3 ± 0.8	7.3 ± 0.9	0.113	0.702
Glycated hemoglobin A1c (%)	7.1 ± 0.7	7.2 ± 1.1	7.5 ± 1.2	0.665	0.384
IOP (mmHg)	15.8 ± 2.4	16.8 ± 2.8	16.1 ± 2.7	0.160	0.372
Axial length (mm)	23.8 ± 0.9	23.6 ± 0.7	23.8 ± 0.9	0.358	0.388
Lens status (clear lens/cataract/IOL)	14/12/8	7/10/7	5/5/10	0.645	0.339
Epiretinal membrane	8/34	6/24	4/20	1.000	0.737
Preoperative retinal laser	12/34	9/24	7/20	1.000	1.000
Preoperative anti-VEGF	4/34	5/24	4/20	0.470	1.000
PPV (simple/combined)	26/8	20/4	17/3	0.747	1.000

Data are presented as mean ± SD or n/total unless otherwise indicated. DM: Diabetes mellitus; F: Female; IOP: Intraocular pressure; IOL: Intraocular lens; M: Male; PPV: Pars plana vitrectomy; SO: Silicone oil; SOR: Silicone oil removal; VEGF: Vascular endothelial growth factor.

Changes in OCTA parameters between the SO and Air groups

Table 2 summarizes the comparative analysis of postoperative OCTA parameters between the SO and Air groups. Adjusted mean differences with 95% confidence intervals were presented in Supplementary Table 1. At 1 month, the SO group demonstrated significantly greater parafoveal RT (382.9 ± 48.4 µm vs 357.4 ± 25.9 µm, P = 0.016) and perifoveal RT (388.6 ± 46.3 µm vs 360.5 ± 26.1 µm, P = 0.036) compared with the Air group. These differences persisted at 3 months, with parafoveal RT of 380.4 ± 45.7 µm vs 355.6 ± 26.4 µm (P = 0.043) and perifoveal RT of 381.4 ± 44.6 µm vs 354.5 ± 25.8 µm in the SO and Air groups, respectively (P = 0.042). No significant difference in foveal RT was observed between the two groups at either time point (all P > 0.05).

Table 2 Retinal parameters and visual acuity in the silicone oil and Air groups.

Parameter	Region	Group	Group × time	1 month			3 months
		P value	P value	SO	Air	P value	SO	Air	P value
RT (μm)	Foveal	0.408	0.946	262.5 ± 25.4	265.9 ± 28.0	0.733	256.7 ± 24.5	260.5 ± 25.3	0.426
	Parafoveal	0.013	0.778	382.9 ± 48.4	357.4 ± 25.9	0.016	380.4 ± 45.7	355.6 ± 26.4	0.043
	Perifoveal	0.011	0.821	388.6 ± 46.3	360.5 ± 26.1	0.036	381.4 ± 44.6	354.5 ± 25.8	0.042
VD in the SCP (%)	Foveal	0.712	0.636	31.1 ± 2.9	30.8 ± 1.4	0.936	31.7 ± 2.2	31.9 ± 0.9	0.914
	Parafoveal	0.087	0.255	44.6 ± 4.9	42.8 ± 1.6	0.114	45.3 ± 1.9	44.5 ± 2.9	0.828
	Perifoveal	0.428	0.146	43.5 ± 3.2	42.7 ± 2.3	0.735	45.1 ± 5.2	46.3 ± 4.3	0.869
VD in the DCP (%)	Foveal	0.105	0.293	35.8 ± 2.7	34.3 ± 1.5	0.143	35.5 ± 2.4	35.1 ± 3.9	0.656
	Parafoveal	0.048	0.023	41.7 ± 3.2	43.2 ± 1.9	0.264	40.8 ± 2.3	46.5 ± 3.7	0.026
	Perifoveal	0.774	0.014	42.4 ± 7.5	42.8 ± 2.0	0.933	41.2 ± 6.1	45.4 ± 3.1	0.039
FAZ (mm2)		< 0.001	0.031	0.42 ± 0.14	0.35 ± 0.05	0.025	0.57 ± 0.07	0.38 ± 0.10	< 0.001
BCVA (logMAR)		< 0.001	0.538	0.82 ± 0.33	0.56 ± 0.32	0.022	0.71 ± 0.34	0.45 ± 0.26	0.005

BCVA: Best-corrected visual acuity; DCP: Deep capillary plexus; FAZ: Foveal avascular zone; RT: Retinal thickness; SCP: Superficial capillary plexus; SO: Silicone oil; VD: Vascular density.

Analysis of VD in the SCP revealed no significant differences between the SO and Air groups in the foveal, parafoveal, or perifoveal regions at 1 month or 3 months postoperatively (all P > 0.05). In contrast, VD in the DCP showed a distinct time-dependent pattern. At 1 month, no significant differences were observed between the groups in the foveal, parafoveal, or perifoveal regions (all P > 0.05). By 3 months, however, the SO group exhibited significantly lower VD in the DCP compared with the Air group in both the parafoveal (40.8 ± 2.3% vs 46.5 ± 3.7%, P = 0.026) and perifoveal (41.2 ± 6.1% vs 45.4 ± 3.1%, P = 0.039) regions.

The FAZ area was significantly larger in the SO group at both evaluated time points. The difference was evident at 1 month (0.42 ± 0.14 mm² vs 0.35 ± 0.05 mm², P = 0.025) and became more pronounced at 3 months postoperatively (0.57 ± 0.07 mm² vs 0.38 ± 0.10 mm², P < 0.001).

Changes in OCTA parameters before and after SOR

Twenty patients from SOR group with complete paired data before and after SOR were enrolled for comparative analysis of OCTA parameter changes, which was summarized in Table 3. Adjusted mean differences with 95% confidence intervals were presented in Supplementary Table 2. RT and VD in the SCP remained stable. No significant differences were observed between the SO and SOR groups in RT or VD in the SCP in any measured region (foveal, parafoveal, or perifoveal) at either 1 month or 3 months (all P > 0.05).

Table 3 Retinal parameters and visual acuity before and after silicone oil removal.

Parameter	Region	Status	Status × time	1 month			3 months
		P value	P value	SO	SOR	P value	SO	SOR	P value
RT (μm)	Foveal	0.720	0.911	261.9 ± 22.1	264.8 ± 32.2	0.721	254.1 ± 21.2	255.7 ± 25.3	0.841
	Parafoveal	0.448	0.952	388.8 ± 44.3	380.7 ± 25.5	0.450	387.1 ± 45.6	378.1 ± 25.1	0.401
	Perifoveal	0.206	0.404	397.0 ± 48.7	380.0 ± 34.3	0.210	376.2 ± 49.6	375.1 ± 30.7	0.933
VD in the SCP (%)	Foveal	0.484	0.718	31.0 ± 2.6	31.6 ± 2.3	0.486	31.8 ± 2.0	33.0 ± 4.6	0.230
	Parafoveal	0.664	0.734	45.0 ± 5.4	44.5 ± 4.1	0.665	45.1 ± 2.2	44.1 ± 1.6	0.363
	Perifoveal	0.458	0.498	43.5 ± 3.2	44.3 ± 3.2	0.460	45.3 ± 5.0	45.0 ± 2.1	0.830
VD in the DCP (%)	Foveal	0.714	0.667	35.2 ± 2.0	34.9 ± 4.3	0.715	35.7 ± 1.8	35.9 ± 3.1	0.810
	Parafoveal	0.991	0.029	41.2 ± 3.3	41.2 ± 5.2	0.991	40.6 ± 1.9	44.1 ± 3.1	0.003
	Perifoveal	0.243	0.020	42.7 ± 8.4	40.8 ± 2.7	0.247	40.5 ± 3.3	43.9 ± 3.7	0.037
FAZ (mm2)		0.001	0.050	0.43 ± 0.13	0.54 ± 0.15	0.002	0.56 ± 0.09	0.58 ± 0.07	0.636
BCVA (logMAR)		0.158	0.856	0.85 ± 0.32	0.73 ± 0.30	0.162	0.74 ± 0.31	0.55 ± 0.22	0.041

BCVA: Best-corrected visual acuity; DCP: Deep capillary plexus; FAZ: Foveal avascular zone; RT: Retinal thickness; SCP: Superficial capillary plexus; SO: Silicone oil; SOR: Silicone oil removal; VD: Vascular density.

In contrast, VD in the DCP showed evidence of recovery following SOR. Although no significant intergroup differences were present at 1 month, the SOR group demonstrated significantly higher VD in the DCP compared with the SO group at 3 months in both the parafoveal (44.1 ± 3.1% vs 40.6 ± 1.9%, P = 0.003) and perifoveal (43.9 ± 3.7% vs 40.5 ± 3.3%, P = 0.037) regions. At 1 month postoperatively, the SOR group also exhibited a significantly smaller FAZ area than the SO group (0.43 ± 0.13 mm² vs 0.54 ± 0.15mm², P = 0.021).

We also evaluated the longitudinal changes in OCTA parameters at four consecutive time points: 1 month and 3 months after SO tamponade and 1 month and 3 months after SOR. RT and VD in the SCP remained stable in all three regions (foveal, parafoveal, and perifoveal) throughout the follow-up period (Figure 3A and B). In the parafoveal region, VD in the DCP was lower at 3 months after SO tamponade than at 1 month. An increase in parafoveal VD in the DCP was observed only at 3 months after SOR, but not at 1 month (Figure 3C). Similarly, perifoveal VD in the DCP showed significant recovery at 3 months after SOR compared with that at 3 months after SO tamponade and at 1 month after SOR (Figure 3C). The FAZ area progressively increased from 1 month to 3 months after SO tamponade, with no significant change after SOR (Figure 3D).

Figure 3 Longitudinal changes in optical coherence tomography angiography parameters before and after silicone oil removal. A: Longitudinal changes in retinal thickness in three regions at four consecutive time points, with no significant differences; B: Longitudinal changes in vascular density (VD) in the superficial capillary plexus in the three regions, with no significant differences; C: Longitudinal changes in VD in the deep capillary plexus in the three regions; D: Longitudinal changes in the foveal avascular zone area; E: Longitudinal changes in best-corrected visual acuity. ^aP < 0.05. VD: Vascular density; DCP: Deep capillary plexus; FAZ: Foveal avascular zone; SCP: Superficial capillary plexus; BCVA: Best-corrected visual acuity; RT: Retinal thickness; SO: Silicone oil; SOR: Silicone oil removal.

BCVA and its correlation with OCTA parameters

At 1 month and 3 months postoperatively, the BCVA (logMAR) was significantly worse in the SO group than in the Air group (0.82 ± 0.33 vs 0.56 ± 0.32, P = 0.022, at 1 month; 0.71 ± 0.34 vs 0.45 ± 0.26, P = 0.005, at 3 months; Table 2). At 1 month after SOR, no significant difference in BCVA was found between the SOR and SO groups (P = 0.162). However, at 3 months after SOR, the SOR group exhibited significantly better BCVA compared with the SO group (0.55 ± 0.22 vs 0.74 ± 0.31, P = 0.041; Table 3), suggesting partial recovery of visual function after SOR.

Correlation analysis demonstrated that parafoveal and perifoveal RT were positively correlated with logMAR BCVA (r = 0.50, P < 0.001 and r = 0.54, P < 0.001, respectively), indicating that greater RT was associated with poorer visual acuity. In contrast, parafoveal and perifoveal VD in the DCP were negatively correlated with logMAR BCVA (r = -0.44, P = 0.016 and r = -0.49, P = 0.008, respectively), suggesting that lower VD in the DCP was linked to worse visual outcomes. Additionally, FAZ area was positively correlated with logMAR BCVA (r = 0.53, P < 0.001), implying that a larger FAZ area was associated with poorer vision (Table 4).

Table 4 Correlation between optical coherence tomography angiography parameters and best-corrected visual acuity.

Parameter	Foveal		Parafoveal		Perifoveal
	r	P value	r	P value	r	P value
RT	0.33	0.112	0.50	< 0.001	0.54	< 0.001
VD in the SCP	0.07	0.732	0.14	0.566	-0.22	0.175
VD in the DCP	-0.12	0.663	-0.44	0.016	-0.49	0.008
FAZ	0.53	< 0.001

DCP: Deep capillary plexus; FAZ: Foveal avascular zone; RT: Retinal thickness; SCP: Superficial capillary plexus; VD: Vascular density.

DISCUSSION

In this study, we investigated retinal morphological and microvascular changes using OCTA following vitrectomy for DR during SO or air tamponade and after SOR. Statistical analysis demonstrated significant alterations in parafoveal and perifoveal RT, parafoveal and perifoveal VD in the DCP, and FAZ area after SO tamponade. Only VD in the DCP showed partial increase after SOR. Furthermore, correlation analysis revealed that parafoveal and perifoveal RT, parafoveal and perifoveal VD in the DCP, and FAZ area were moderately correlated with postoperative BCVA.

Parafoveal and perifoveal RT was greater in the SO group than in the Air group, indicating a propensity for retinal edema in eyes with SO tamponade. After SOR, no significant changes in parafoveal and perifoveal RT were observed between the SO and SOR groups, suggesting that the retinal edema did not fully resolve after SOR. The difference in RT changes between the parafoveal and foveal regions may be attributed to structural disparities. The parafoveal retina is looser and more complex, rendering it more susceptible to intraoperative manipulation, postoperative inflammation, and other stressors that may be associated with persistent edema[13,14].

The most pronounced changes in retinal microcirculation were observed in parafoveal and perifoveal VD in the DCP at 3 months. These findings suggest progressive changes in the DCP vessels during follow-up in eyes with SO tamponade, which aligns with prior studies on SO tamponade in retinal detachment[15,16]. However, the decreased VD in the DCP exhibited partial recovery after SOR, indicating that microvascular changes in the DCP vessels may not represent complete vascular atrophy. Compared with the SCP, the DCP locates at a watershed zone with lower baseline oxygen saturation, which may correspond to more susceptible to hypoxic stress[16,17].

The FAZ area was enlarged at 1 month after SO tamponade and continued to increase at 3 months, with no significant change after SOR. The positive correlation between FAZ area and logMAR visual acuity indicates that a larger FAZ is associated with poorer visual outcomes. As an important indicator of retinal perfusion, persistent enlargement of the FAZ after SOR may suggest incomplete recovery of macular perfusion. These findings suggest that the retinal microcirculatory changes following SO tamponade include both short-term reversible alterations and long-term irreversible alterations[18,19].

At 3 months postoperatively, the SO group exhibited the poorest BCVA, potentially because of optical distortion induced by SO. At 3 months after SOR, BCVA improved compared with that at 3 months after SO tamponade. Correlation analyses demonstrated that changes in RT, VD in the DCP, and FAZ area were moderately associated with BCVA, suggesting the potential utility of OCTA for predicting visual prognosis. Theoretically, increased RT indicates retinal edema, whereas decreased VD in the DCP and enlarged FAZ suggest compromised retinal perfusion, all of which may contribute to suboptimal visual outcomes[20,21].

Existing studies investigating retinal microcirculatory alterations following SO tamponade primarily enroll patients with retinal detachment[22,23]. Longitudinal data tracking retinal microcirculatory changes after SO tamponade in eyes with DR remain scarce, which constitutes the core novelty of the present study. Prior OCTA studies focused on retinal detachment have proposed three potential pathways through which SO may influence retinal perfusion: Mechanical compression of retinal microvessels, chronic inflammatory injury to vascular endothelial cells, and aggravated hypoxia in the outer retina[24-26]. Whether these hypothesized pathways contribute to the microcirculatory changes identified in our cohort awaits further verification.

Current OCTA research on DR mostly focuses on screening subclinical microvascular damage in early DR, managing diabetic macular edema, automated image analysis with artificial intelligence, and so on. Few studies have explored retinal microcirculation after vitrectomy, especially in eyes with SO tamponade. Prior work has found more severe thinning of the peripapillary nerve fiber layer in diabetic eyes filled with SO[27,28]. Chronic intraocular inflammation related to diabetes also tends to trigger SO emulsification and accelerate epiretinal membrane formation[29]. Our study focuses on core OCTA parameters across multiple retinal regions and provides longitudinal sequential data to describe how retinal microcirculation changes throughout SO tamponade in complicated DR.

It should be noted that this is a retrospective non-randomized cohort study, significant selection bias exists. Although the duration of DM, preoperative blood glucose levels and HbA1c were comparable between the SO and Air groups, there were notable differences in overall preoperative disease severity. Moreover, patients requiring SO tamponade generally had more extensive fibrovascular proliferation or delayed presentation for treatment. All these factors inevitably introduced confounding bias into our results. Therefore, the poorer retinal microcirculation and BCVA observed in the SO group cannot be attributed solely to the effects of SO. The more severe disease at baseline and relevant confounding factors likely also contributed to these outcomes. We did not use propensity score matching to balance preoperative retinal lesions, limiting our ability to draw firm conclusions about the independent microvascular influence of SO. Future studies with larger sample sizes and more refined patient stratification are required to minimize such sources of bias.

This study had several other limitations. First, most patients underwent SOR 3 months after surgery, limiting the assessment of long-term retinal changes after SO tamponade. Future studies with variable SOR time points are needed to determine the effects of earlier or later SOR on retinal microcirculation. Second, our OCTA analysis focused only on VD, whereas perfusion density, a critical determinant of retinal function, was not evaluated. Third, most patients requiring vitrectomy presented with obvious vitreous hemorrhage, which hindered the acquisition of high-quality preoperative OCTA images. Finally, because of the small sample size and multiple statistical comparisons, the possibility of both type I and type II errors should be considered in this study. Given its exploratory retrospective design, our study is limited by the nonprespecified sample size. Future largerscale prospective studies are required to validate the betweengroup differences observed. These limitations also highlight important directions for future research.

CONCLUSION

Our findings show that SO tamponade in patients with complicated DR is associated with distinct alterations in retinal microcirculation. After SOR, only partial increase of VD in the DCP was observed, whereas no significant change was found in the FAZ area. OCTA not only enables rapid detection of morphological and perfusion changes in the retina but also provides quantitative parameters that correlate with postoperative visual prognosis, thereby facilitating postoperative follow-up and clinical management of patients undergoing vitrectomy.