Comparison of Intraoperative Fluorescein Angiography–Guided Lavage Versus Conventional Lavage in Delayed Postoperative Vitreous Cavity Hemorrhage

Abstract

Purpose: To compare outcomes of vitreous lavage with and without intraoperative fluorescein angiography (FA) in eyes with delayed postoperative vitreous cavity hemorrhage. Methods: This prospective case-control study included patients with postoperative vitreous cavity hemorrhage who were randomized into 2 groups (intraoperative FA and lavage-only groups). Findings such as areas of untreated capillary nonperfusion, active neovascularization, and diffuse peripheral leakage in the retinal periphery were noted in the intraoperative FA group. Appropriate steps were performed, including additional endolaser and/or cryotherapy. In the lavage-only group, these steps were based on the surgeon’s discretion. Results: Twenty eyes of 19 patients were included in each group. All eyes showed unlasered areas of capillary nonperfusion on intraoperative FA, and targeted photocoagulation was performed. Six eyes showed diffuse peripheral leakage near the ora serrata in addition to areas of capillary nonperfusion, for which cryotherapy was performed. At 6 months, none of the patients in the intraoperative FA group had a recurrence of postoperative vitreous cavity hemorrhage, whereas 3 eyes in the lavage-only group developed recurrence. This difference was not statistically significant (P = .92). There was no significant difference in the best-corrected visual acuity between the groups at 6 months. Conclusions: There were no instances of recurrent hemorrhage in the intraoperative FA group. Preliminary data suggest the potential role of this technique to determine the causes of postoperative vitreous cavity hemorrhage as well as to reduce the rate of recurrence.

Introduction

Postoperative vitreous cavity hemorrhage, defined as dispersed vitreous cavity hemorrhage in eyes that have undergone pars plana vitrectomy (PPV) for vitreous hemorrhage due to proliferative diabetic retinopathy, retinal vein occlusion, or vasculitis, affects 6% to 75% of eyes undergoing vitrectomy and can occur in the early or late postoperative period. ¹ Early in the immediate postoperative period, hemorrhage of the vitreous cavity may be caused by the dissolution of blood clots from sites of membrane dissection or the vitreous base. Delayed hemorrhage occurs after the vitreous cavity has remained free of blood for at least 6 to 8 weeks after PPV or silicone oil (SO) removal. ² Late and recurrent postoperative vitreous cavity hemorrhage can be caused by progressive neovascularization, traction on residual fibrovascular stalks, or anterior hyaloidal fibrovascular proliferation at the sclerotomy ports. ² In mild cases, delayed postoperative vitreous cavity hemorrhage is observed; however, for dense cases in which there is a drastic reduction in vision, a vitreous lavage is necessary.

During vitreous lavage for postoperative vitreous cavity hemorrhage, steps including endolaser, meticulous base dissection for clearing clots lodged at the vitreous base, and cryotherapy to sclerotomy ports are performed to prevent recurrences of vitreous cavity hemorrhage. Currently, these steps are based on the retina surgeon’s discretion and are, hence, subjective. If the patient has had more than 1 episode of vitreous cavity hemorrhage, the retina surgeon may choose to inject gas or SO to prevent recurrences.

In this prospective study, the usefulness of intraoperative fluorescein angiography (FA) was analyzed to objectively assess the causes of delayed postoperative vitreous cavity hemorrhage. In addition, we compared the outcomes of vitreous lavage, with and without intraoperative FA guidance, in terms of best-corrected visual acuity and number of recurrences.

Methods

This prospective case-control study was performed at a tertiary care hospital from April 2024 to January 2025 in patients with proliferative diabetic retinopathy and retinal vein occlusion who had undergone vitrectomy with or without membrane surgery, SO removal, or vitreous lavage and had remained hemorrhage-free for at least 2 months before developing vitreous cavity hemorrhage. Patients with renal disease and allergies to fluorescein were excluded. Written informed consent was obtained from all patients. The study was approved by the institutional review board and adhered to the ethics of the Declaration of Helsinki.

Patients were instructed to avoid food and drink 3 to 4 hours before surgery to avoid nausea and vomiting caused by intravenous fluorescein. All surgeries were performed by a single surgeon (A.A.). None of the patients received intravitreal pharmacotherapy (either antivascular endothelial growth factor [anti-VEGF] or triamcinolone acetonide) before the vitreous lavage, at the end of lavage, or postoperatively.

Surgical Procedure

Digitally assisted vitreoretinal surgery was performed using the NGENUITY 3D visualization system (Alcon Laboratories, Inc). The surgeon wore passive 3D polarized glasses and was seated 1.2 meters away from the screen. Standard 25-gauge PPV was performed using the Constellation Vision System (Alcon Surgical). Vacuum-assisted suction with a cutter was sufficient for clearance of liquefied, dispersed hemorrhage. In cases with lodged blood clots in the retrolental area, vitrectomy with a low cut rate was done diligently, careful to avoid touching the lens. Blood clots lodged in the vitreous base were removed with a high cut rate and low vacuum.

Once hemorrhage was cleared and any remnant fibrous tissue removed, intraoperative FA was performed using the method described by Franklin et al. ³ A 485 nm bandpass filter, 14 mm in diameter (Edmund Optics), was placed into the filter holder of the accessory light source. A digital barrier filter was engineered using a specific color channel on the NGENUITY system that permitted only the yellow-green wavelength of light to be visualized. Intraoperative FA involved switching to the light source with the exciter filter and shifting to the preset color channel on the NGENUITY screen. The light source intensities were increased to 90 for the endoilluminator light pipe and 90 for the chandelier. Sodium fluorescein, 2.5 mL of 100 mg/mL concentration, was injected intravenously to obtain an intraoperative angiogram.

Intraoperatively, findings such as unlasered areas of capillary nonperfusion, active neovascularization of the disc or elsewhere, and diffuse peripheral leakage in the periphery were recorded. Endolaser was performed on the areas of capillary nonperfusion. The areas of diffuse leakage at the ora, near the ports, were diagnosed as anterior hyaloidal fibrovascular proliferation. In these cases, cryotherapy was performed on the ports. In the lavage-only group, targeted endolaser and cryotherapy were additionally performed at the retina surgeon’s discretion. SO injection was done in eyes that had already undergone lavage before the current episode of postoperative vitreous cavity hemorrhage as well as in eyes with poor vision in the contralateral eye.

Results

Preoperative Data

The intraoperative FA group included 20 eyes of 19 patients (14 males, 5 females). The mean ± SD age was 59.3 ± 7.7 years. The median best-corrected visual acuity (BCVA) at baseline was 1.6 logMAR. Eighteen eyes had proliferative diabetic retinopathy, 1 eye had branch retinal vein occlusion (BRVO), and 1 eye had central retinal vein occlusion (CRVO). Seventeen patients had previously undergone vitrectomy and endolaser. Of these 17, 9 eyes had also undergone fibrovascular membrane dissection. Three patients had undergone vitreous lavage and endolaser after the primary surgery. Postoperative vitreous cavity hemorrhage was seen after a mean of 270.5 ± 89.18 (median, 205 days) postoperative days.

The lavage-only group included 20 eyes of 19 patients (13 males, 6 females). The mean age was 54.5 ± 9.5 years. Eighteen eyes had proliferative diabetic retinopathy, 1 eye had BRVO, and 1 eye had CRVO. Eighteen patients had previously undergone vitrectomy and endolaser, of which 8 had undergone membrane surgery. Two patients had undergone vitreous lavage and endolaser after the primary surgery. Postoperative vitreous cavity hemorrhage was found in all eyes after a mean of 265.53 ± 119.4 (median, 225 days) postoperative days. All eyes had a visual acuity of hand movements close to face or less, with a median baseline BCVA of 1.9 logMAR. Both groups were evenly matched for age, sex, primary pathology, baseline BCVA, and primary surgery performed. There were no significant differences in duration between surgery and development of vitreous cavity hemorrhage in the 2 groups (Table 1).

Table 1.

Baseline Characteristics, Intraoperative Steps, and Final Outcomes in Both Groups.

Parameter		Intraoperative FA Group (n = 20)	Lavage-Only Group (n = 20)	P Value
Age (y), mean ± SD		59.3 ± 7.7	54.5 ± 9.5	.06
Male		14	13	.95
Female		5	6
PDR		18	18	.79
CRVO		1	1
BRVO		1	1
Baseline BCVA (median)		1.6	1.9	.52
Primary surgery	Vitrectomy	8	10	.79
	Vitrectomy + membrane surgery	9	8
	Vitreous lavage	3	2
Lens status	Phakic	2	0	.48
	Pseudophakic	18	20
Duration between primary surgery and vitreous cavity hemorrhage, d	Mean ± SD	270.5 ± 89.18	265.53 ± 119.4	.88
	Median	205	225	.34
Intraoperative steps Endolaser		20	20
Cryotherapy to ports		6	4
Fluid–air exchange		14	17
Gas tamponade (SF6)		0	0
Silicone oil		6	3
Recurrence at 6 months		3	0	.93
Median BCVA at 6 months		0.5	0.5	.33

Abbreviations: BCVA, best-corrected visual acuity; BRVO, branch retinal vein occlusion; CRVO, central retinal vein occlusion; FA, fluorescein angiography; PDR, proliferative diabetic retinopathy; SF₆, sulfur hexafluoride.

Intraoperative Data

Unlasered areas of capillary nonreperfusion were found in all eyes in the intraoperative FA group beyond what was clinically visible without the intraoperative FA filter (Figures 1 and 2). All eyes showed excellent delineation of unlasered areas of capillary nonreperfusion, and intraoperative FA guided-targeted photocoagulation was performed in all eyes (Figure 3). Six eyes showed diffuse peripheral leakage near the ora, at the sclerotomy ports (Figure 4), for which cryotherapy was performed. One eye showed extensive window defects in the early phase at the posterior pole that persisted until late phases of the angiogram, suggesting retinal and retinal pigment epithelium atrophy (Figure 5). Fourteen eyes were left filled with air at the end of surgery. Three eyes in this group had already undergone a vitreous lavage, and 3 had poor vision in the contralateral eye; therefore, these 6 eyes underwent SO injection (Table 1).

Figure 1.

(A) Unlasered areas seen without the intraoperative fluorescein angiography filter. (B) Well-delineated areas of capillary nonperfusion as seen on intraoperative fluorescein angiography.

Figure 2.

(A) Unlasered areas seen without the intraoperative fluorescein angiography filter, roughly annotated with yellow outlines (D). (B) Well-delineated areas of capillary nonperfusion, demarcated with a yellow outline (E). (C and F) A delay in filling of the inferotemporal branch retinal vein is demonstrated.

Figure 3.

Diffuse leakage seen with indentation at the ora near the vitrectomy ports.

Figure 4.

(A) Intraoperative screen grab shows lasered proliferative diabetic retinopathy with an unremarkable posterior pole. (B) Intraoperative fluorescein angiography in the same case shows early hyperfluorescence in the posterior pole extending to the midperiphery in the choroidal phase of the angiogram, suggestive of window defects due to retinal and retinal pigment epithelium atrophy.

Figure 5.

Intraoperative fluorescein angiography filter is used with a xenon lamp endoilluminator for intraoperative fluorescein angiography-guided endolaser therapy.

In all eyes in the lavage-only group, endolaser was performed on the clinically visible skip areas. Cryotherapy was performed on ports in 4 eyes where there were fewer skip areas. Seventeen eyes were left filled with air. Three eyes had previously had a vitreous lavage, 1 of which had poor vision in the contralateral eye; hence, these eyes had SO injection (Table 1).

For the 9 eyes across both groups, SO removal was done after 8 weeks. None of the patients in the intraoperative FA group had a recurrence of postoperative vitreous cavity hemorrhage at 6 months, whereas recurrences were seen in 3 eyes in the lavage-only group. The difference was not statistically significant (P = .92) (Table 1).

Conclusions

The incidence of postoperative vitreous cavity hemorrhage has decreased with the advent of smaller gauge vitrectomies and the use of preoperative anti‑VEGF therapy. ⁴ Early in the immediate postoperative period, vitreous cavity hemorrhage occurs due to the dissolution of uncut blood clots, especially at the inferior vitreous base. Vitreous cavity hemorrhage can be observed soon after surgery, but vitreous lavage surgeries are required for persistent and recurrent cases. ⁵ Delayed and recurrent postoperative cases of vitreous cavity hemorrhage can result from new-onset neovascularization elsewhere, with or without traction. Systemic risk factors for delayed cases of postoperative vitreous cavity hemorrhage include poor glycemic control and hypertension. ⁶ Recurrent cases have also been linked to coronary artery disease, use of antiplatelet and anticoagulant agents, and younger age. ⁷

Postoperative vitreous cavity hemorrhage can be prevented with measures such as preoperative/intraoperative injection of intravitreal anti‑VEGF, confluent endolaser, cryotherapy of sclerotomy sites, and indented base shaving. Tan et al ⁸ reported a reduced occurrence of postoperative vitreous cavity hemorrhage in their series, finding that with the transient switching off of the infusion and minimal globe indentation after vitrectomy (before initiating fluid–air exchange), oozing blood vessels can be identified and cauterized.

Identifying biomarkers such as persistent oozing blood vessels and delineation of untreated areas of nonperfusion has become possible with intraoperative FA. Early research on intraoperative FA-guided diabetic vitrectomy has shown reduced rates of postoperative vitreous cavity hemorrhage in the first 3 months. This was attributed to abnormal vascular leakage and ischemia being detected on intraoperative FA after the delamination of membranes. These residual areas of abnormal vascular leakage and ischemia received more confluent laser while sparing the better-perfused retina. ⁹

A previous study by Terasaki et al ¹⁰ used an ophthalmic endoscope system and an endoilluminator for the treatment of proliferative diabetic retinopathy, highlighting the value of intraoperative FA. All 12 eyes in their series were noted to have a wide avascular area anterior to the previous panretinal photocoagulation scars. Two eyes were found to have a fibrovascular ridge similar to a demarcation line in retinopathy of prematurity. One eye had been previously treated for neovascular glaucoma (NVG), and the other eye developed NVG in the subsequent weeks. It was purported that this ridge could be an early sign of anterior hyaloidal fibrovascular proliferation and an increased risk factor for NVG. Another finding in their study was fluorescein leakage from the sclerotomy sites, suggestive of fibrous ingrowth. In the current study, large areas of avascularity and peripheral leakage were found at the ora near the sclerotomy ports. The remnant uncut hyaloid at the ports may provide a nidus for fibrovascular growth. Therefore, cryotherapy was used for these very anterior areas.

Imai et al ¹¹ were the first to recognize the potential benefits of digitally assisted vitreoretinal surgery for intraoperative FA. By virtue of the image processing system and enhancement of the display in low illumination, the usefulness of 3D intraoperative FA in surgery for proliferative diabetic retinopathy was assessed. The authors used a 27-gauge light pipe connected to an excitation filter-mounted light source; the light pipe brightness was 60%. The excitation filter was placed in the sliding tray of a mercury vapor light source. An interference barrier filter was adapted to the surgical microscope. The display settings were modified to get high resolution visualization.

Cardamone et al ³ described placing of the excitation filter directly into an accessory holder in the Constellation vitrectomy system. Initially, an optical barrier filter was used, similar to the one used by Imai et al. ¹¹ Subsequently, a digital barrier filter was used through the NGENUITY visualization system. Using the accessory holder of the Constellation vitrectomy system eliminated the need for a special excitation filter-mounted light source. The digital filter reduces red and blue emissions and enhances only the green signal. Saturation and hue are modified to reduce the blue and green hues and to produce a gray-scale image similar to office-based FA. With this integrated FA system, the authors described clinical biomarkers such as vascular filling times, vascular occlusions, shunt vessels, capillary nonperfusion, and retinal and choroidal neovascularization. The 3D intraoperative FA system is highly efficient because it involves switching to the accessory light source with an exciter filter and merely changing the color channel on the digitally assisted vitreoretinal surgery system. Once intraoperative FA has been completed, it is equally easy to switch back to standard visualization by changing to the xenon light pipe endoilluminator.

Multiple vitreous lavage surgeries may be required when vitreous cavity hemorrhage develops after complex tractional retinal detachment surgeries and may involve extensive surgical steps such as endolaser, cryotherapy to ports, and SO tamponade to prevent recurrences. Early visual rehabilitation is the goal of surgery in diabetic vitrectomies, and recurrent hemorrhage can lead to disappointment in patients. Fluid–air exchange in the outpatient setting is a useful modality for treating patients with postoperative vitreous cavity hemorrhage. After being informed of the benefits of a lavage procedure, which involves base shaving, blood clot removal from the retrolental area, and cryotherapy to ports, the patients in our study made an informed decision to undergo a repeat surgery for a more thorough job. We acknowledge the benefits of an in-office procedure, and it will be interesting to have an additional arm for outpatient fluid–air exchange in future studies.

Ours is the first study to compare the outcomes of lavage with and without intraoperative FA guidance in cases of postoperative vitreous cavity hemorrhage secondary to proliferative diabetic retinopathy and retinal vein occlusion. In addition to untreated capillary nonperfusion, other novel findings were noted in our series, such as capillary leakage at the ora and extensive window defects in the posterior pole (suggestive of retinal and retinal pigment epithelium thinning). Moreover, targeted laser in areas of an ischemic retina, while relatively sparing areas of better perfusion, can theoretically preserve more peripheral retina and low light vision compared with indiscriminate placement of laser burns. When left to the surgeon’s discretion, there may be a tendency to use gas or oil tamponade as a blanket therapy to reduce recurrences. Prolonged tamponade with gas causes delay in visual rehabilitation. SO injection to prevent recurrence of postoperative vitreous cavity hemorrhage entails an additional surgery to remove SO after a few months, in addition to the risk of secondary glaucoma. Intraoperative FA can potentially reduce instances of undertreatment or overtreatment, enabling the surgeon to perform only the necessary steps. In our study, SO was used only in eyes that had already undergone lavage before the current episode of hemorrhage and in those eyes with poor vision in the contralateral eye. After 8 weeks, all eyes underwent SO removal.

Phototoxicity is a concern with intraoperative FA. All surgical light sources and bandpass exciter filters used are above the wavelength and the duration purported to be toxic (440 nm and 15 minutes). ¹² Moreover, the fluorescein signal fades after a mean of 7.5 minutes, limiting the risk of phototoxicity. So far, no evidence of phototoxicity has been reported. ⁹

A limitation of our study is the small sample size. Assuming a recurrence rate of 15% in the lavage-only group and 0% in intraoperative FA-guided group, a minimum of 37 eyes per group (74 total eyes) would be required to achieve 80% at a 5% significance level, resulting in a statistically significant difference. In our study, there were 20 eyes in each group. This small sample size was due to the stringent criteria of including only cases of postoperative vitreous cavity hemorrhage that persisted for 6 to 8 weeks and which were sufficiently dense to warrant vitreous lavage. Furthermore, advanced instrumentation in minimally invasive vitreoretinal surgery, near-complete removal of complex adherent membranes during primary surgery, and indented vitreous base shaving with wide-angle viewing systems have all drastically reduced the incidence of postoperative vitreous cavity hemorrhage. The recurrence rates between the intraoperative FA group (0%) and the lavage-only group (15%) did not reach statistical significance (P = .92). The 95% CI for the difference in recurrence rates ranged from -7.1% to 36.1%, indicating that while the sample maybe underpowered to detect a significant difference, the true effect could still favor intraoperative FA-guided lavage. We acknowledge the need for larger, adequately powered studies to validate these findings.

Another limitation of our study was the inclusion of cases of both proliferative diabetic retinopathy and retinal vein occlusion. Multiple diagnoses could affect the interpretation of results. However, there are no studies that compare the recurrence rates of postoperative vitreous cavity hemorrhage in patients with proliferative diabetic retinopathy and retinal vein occlusion. We chose to include these diagnoses in our study because similar systemic and ocular risk factors could be at play in the causation of postoperative vitreous cavity hemorrhage.

Postoperative vitreous cavity hemorrhage is visually debilitating, and episodes can reoccur despite 1 or more vitreous lavages. Intraoperative FA can help elucidate the underlying causes of recurrence and potentially enable the surgeon to tailor the steps needed for each case, avoiding extensive maneuvers such as cryotherapy to ports, fluid–gas exchange, and SO injection. Combined with other enhancements to digitally assisted vitreoretinal surgery, such as novel high-speed cutters, digital filters, and intraoperative optical coherence tomography, as well as the emerging use of artificial intelligence, intraoperative FA heralds a significant evolution and provides refinement to vitreoretinal surgical techniques and postoperative outcomes. Long-term studies with a larger sample size may reveal the true efficacy of this modality for vitreous lavage surgeries.