Combined vitrectomy, near-confluent panretinal endolaser, bevacizumab and cyclophotocoagulation for neovascular glaucoma — a retrospective interventional case series

Background: Neovascular glaucoma (NVG) is a severe, potentially blinding disease and a therapeutic challenge. The purpose of this study was to evaluate the safety and efficacy of an integrative surgical approach to neovascular glaucoma. Methods: Retrospective analysis of a one-year follow-up of a consecutive interventional case series of NVG. Eyes underwent transscleral cyclophotocoagulation, pars plana vitrectomy, near-confluent panretinal photocoagulation, and intravitreal bevacizumab. Phakic eyes underwent concomitant cataract surgery. Best-corrected visual acuity (BCVA, logMAR), intraocular pressure (IOP, mmHg), number of glaucoma medication, visual analog pain scale (VAPS, 0-10) were recorded at baseline, and 1, 3, 6, and 12 months. Blind eyes were excluded. Results: Seventy-seven eyes of 77 patients (45 male, 32 female, mean age 73.6±12.2 years) were included. NVG underlying conditions included retinal vein occlusion (41.6%), proliferative diabetic retinopathy (35.1%), central retinal artery occlusion (19.5%), and ocular ischemic syndrome (3.9%). Mean IOP decreased postoperatively from 46.3±10.1 mmHg to 14.5±7.9 mmHg (p<0.001), glaucoma medication from 4.7±1.3 to 1.8±1.8 (p<0.001), and VAPS from 6.0±1.8 to 0. BCVA remained unchanged. Postoperative intraocular inflammation had resolved in all eyes at the one-month follow-up. 71.4% (55/77) eyes did not require additional major interventions during follow-up. Conclusions: A single, comprehensive surgery session lowered IOP significantly, reduced GMS, and controlled pain.


Introduction
Neovascular glaucoma (NVG) is a serious complication of a variety of ocular and systemic conditions. Neovascularization is the formation of abnormal blood vessels in an abnormal location triggered by an imbalance of anti-angiogenic and proangiogenic factors caused by retinal ischemia 1 . NVG accounts for 3.9 to 9.2% of all new glaucoma diagnoses 2-4 . According to the Federal Statistical Office of Germany, NVG's age-specific incidence in Germany in the age group from 45 to 64 years is 8 per 100,000. It increases to 24 per 100,000 in subjects older than 64 years 5 . The incidence of NVG varies depending on the etiology of retinal ischemia. For central retinal vein occlusion (CRVO) the reported incidence is 16% 6 , for proliferative diabetic retinopathy (PDR) 21.3% 7 , for central retinal artery occlusion (CRAO) 14.5% 8 , and ocular ischemic syndrome (OIS) 12.9% 9 , respectively. Carotid artery obstructive disease and fistulas are additional, extraocular vascular causes of retinal ischemia 10 . Early recognition and treatment of NVG are imperative to prevent aggressive evolution with severe vision loss and intractable pain that can require enucleation within a few months 11 . NVG also carries a poor prognosis for general health: remarkably, the expected lifespan of patients with NVG decreased by 52% compared to an age-correlated normal population, which corresponds to 6.5 years. In diabetics with NVG, the expected lifespan was reduced even more significantly by 72% (5.1 years in this subgroup) 12 . In 1994, Miller et al. showed that retinal laser coagulation of primates could lead to CRVO and subsequent retinal ischemia with iris neovascularization 13 . In this model, neovascularization was mediated by vascular endothelial growth factor (VEGF) and correlated to its concentration. Several members of the VEGF family have since been identified, including VEGFA, VEGFB, VEGFC, VEGFD, and the placental growth factor (PLGF) with specific receptor-binding patterns. VEGFA primarily binds to VEGFR2, the activation of which stimulates neovascularization, relevant to NVG, and angiogenesis, the formation of normal blood vessels in normal development. In the eye, VEGF-A is produced by the retinal pigment epithelium, retinal ganglion cells, astrocytes, the endothelium, photoreceptors, and Müller cells [14][15][16][17] . Retinal hypoxia upregulates VEGF primarily via the hypoxiainducible factor (HIF-1α and HIF-2α). The concentration of HIF increases when hydroxylases are inhibited during hypoxia so that HIF is not degraded by the proteasome 18 . HIF binds to the hypoxia-responsive element (HRE) of the VEGF gene in the nucleus leading to its upregulation 19 .
Although the mechanism by which NVG emerges is hence relatively well understood, there is no consensus on how to best initiate treatment to address the different aspects. Treatment steps include lowering of IOP (topical and systemic glaucoma medications, glaucoma drainage implants, cyclodestruction) 20-24 , anti-inflammatory treatment (topical or intraocular steroids), reduction of retinal ischemia (panretinal photocoagulation (PRP)) 25,26 , and inhibition of VEGF 27-29 . Vitrectomy with PRP and silicone oil tamponade may also reduce IOP in eyes with NVG 30,31 .
Here, we evaluated the safety and efficacy of an integrative combined surgical approach for evolving NVG that combined pars plana vitrectomy, near-confluent full-scatter panretinal photocoagulation, off-label use of intravitreal bevacizumab, and transscleral cyclophotocoagulation in a single surgical session. We hypothesized that such a combined surgical approach for NVG would reduce IOP, medication, pain, reduce the number of necessary outpatient visits, and prevent the necessity of further surgeries.

Ethical statement
Our retrospective study was reviewed and approved by the Ethics Committee of the University of Würzburg (reference: 9/17-sc, dated February 3, 2017) and a clearance certificate for retrospective data evaluation was issued (reference: 20180108 02, dated January 30, 2018). For retrospective anonymized data analysis, patient's consent was not necessary. The Declaration of Helsinki, the Good Clinical Practice (GCP) guidelines and applicable data protection regulations were complied with.

Study design
We included all patients between October 2014 and August 2019 with NVG who met the inclusion criteria in this consecutive interventional case series.
All patients were treated at the University Eye Hospital in Würzburg, Germany. The combined intervention was part of routine patient care at our clinic and all patients with decompensated neovascular glaucoma and preserved visual

Amendments from Version 1
We are grateful for the opportunity to improve our manuscript. We appreciate the helpful comments and have provided our replies below. We trust that we were able to address the concerns that were raised.
In the Introduction, second paragraph, we changed the phrasing. " In 1994, Miller et al. showed that laser occlusion of all branch retinal veins in a primate could lead to CRVO and subsequent retinal ischemia." In the Results "arterial hypertension" was changed to hypertension as this is the currently accepted terminology.
In Table 1 and in the manuscript-we changed "R stadium" to "NVI stage" We clarified in the results, that all 3 patients with a retinal detachment consecutively developed a painless phthisis bulbi.
To provide comparison to our study we added a literature overview (Table 3) of different surgical approaches.
In the Discussion regarding NVI stage and interventions, we added, that "there was no difference between stage 2 and stage 3". Further we added more details on trabeculectomy in NVG.
Any further responses from the reviewers can be found at the end of the article REVISED function received this intervention. BCVA, intraocular pressure (IOP, mmHg), the number of glaucoma medication, visual analog pain scale (VAPS, 0-10) was recorded at baseline and at follow-up visits at one, three, six, and 12 months as part of routine care. Iris and anterior chamber neovascularization were graded using the Weiss and Gold rubeosis grading system 32 . Hypotony was defined as IOP ≤ 5 mmHg with hypotonous maculopathy, choroidal folds, or optic neuropathy 33 . Phthisis bulbi 34 was defined as IOP ≤ 5 mmHg in a shrunken eye with worse than hand motion vision with or without pain containing atrophic and disorganized intraocular structures. Success was defined as lOP ≤ 21 mmHg or IOP reduction ≥ 30% from baseline, with or without glaucoma medication and without vision loss 35 .
We retrospectively analyzed all patients treated with neovascular glaucoma at our clinic via the electronic hospital information system SAP®. The medical records of the treated patients were analyzed individually. The initial diagnosis of neovascular glaucoma and hospitalization served as the starting point. The cardiovascular risks were taken from the anesthesia premedication form.
Patient treatment and follow-up Surgical technique. All eyes underwent transscleral cyclophotocoagulation (810 nm diode laser, 360 degrees treatment to pop threshold with 20 spots and leaving out 3 and 9 o'clock), standard 3-port 23 gauge pars plana vitrectomy with a detachment of the posterior vitreous if not already present, near-confluent full-scatter panretinal photocoagulation (PRP) applied under indentation in all four quadrants from the vascular arcades to the ora serrata, intravitreal application of 0,1 mL of bevacizumab (Avastin® 25 mg/1 mL, Roche Pharma, Switzerland), and air tamponade. Phakic eyes underwent concomitant cataract surgery. The operation was carried out either with a retrobulbar block or general anesthesia.

Retreatment.
At follow-up, all eyes with elevated IOP were treated following an escalation scheme. First, glaucoma medications were increased to what was maximally tolerated. Eyes were then treated with transscleral cyclophotocoagulation (810 nm diode laser, 360 degrees treatment to pop threshold with 20 spots). Eyes that failed to respond to cyclophotocoagulation with a significant IOP reduction and had retained ambulatory visual acuity underwent tube shunt surgery. Repeated vitrectomy, including fill-in PRP, transscleral cyclophotocoagulation, and off-label use of intravitreal bevacizumab was applied to eyes with elevated IOP and dense vitreous hemorrhage. Fill-in PRP was applied when PRP could not be completed during surgery in eyes with extensive intraretinal hemorrhage. Further intravitreal injections of VEGF inhibitors were only applied for clinically significant non-ischemic macular edema and BCVA ≥1.1 logMAR.

Patients lost to follow-up.
All patients who failed to attend follow-up visits completed a standardized telephone survey that contained the following questions: What was the main reason for not following up? What is your current vision function? How many glaucoma drops are you using? Do you have any eye pain? Are you overall satisfied with the outcome of the treatment?
Statistical methods Data analysis was performed using Statistica 13.1 (Tulsa, Oklahoma, United States). The frequency of observations described categorical variables. Continuous variables were described as mean with standard deviation (SD) or median with range (minimum-maximum). Friedman test and Wilcoxon signed-rank test were used to compare data measured on an ordinal scale and continuous variables with non-normal distribution. Evaluation of data normality was performed using the Shapiro-Wilk test. Welch's t-test for unequal variances was used for IOP between pain versus no-pain group comparison. Categorical variables of the relationship between neovascularization stage and type of intervention were compared using the χ2 test with odds ratio (OR) measurement. Kaplan-Meier curve and log-rank test were used for success analysis. P-values <0.05 were considered significant.
Lost to follow-up 21 patients were lost to follow-up. 10 patients died during the one-year follow-up period. All 21 patients lost to follow-up completed a telephone survey. The main reasons for missed visits were general health issues in 52.4% (11/21), long-distance to our clinic 38.1% (8/21), and family-related reasons in 9.5% (2/21  reported overall satisfaction with the treatment. All patients were pain-free.

Discussion
By 1871, NVG was known as "glaucoma haemorrhagicum et apoplecticum" and feared as a consequence of ischemia that quickly led to enucleation due to "hefty ciliary neuralgia" 39 . In 1963, using improved equipment, Weiss et al. found that emerging neovascularization and fibrovascular membranes of the iris and the angle could be observed well before the onset of advanced NVG 40 , hinting at a window to initiate treatment. Today, the ability to detect neovascularization early is complemented by improved interventions that address both the underlying pathology and elevated IOP. However, these interventions need to be implemented during the early stages of the disease because NVG continues to have a rapid evolution and a poor prognosis for both the eye and the patient 12 . One study showed that the expected lifespan of patients with NVG decreased by 52% compared to an age-correlated normal population, corresponding to a loss of 6.5 years. In diabetics with NVG, the lifespan was reduced even more by 72% 12 . In our study, 74% (57/77) of patients had multiple cardiovascular risk factors and were overweight or obese. During follow-up, ten patients died. The high mortality rate is a reminder that any treatment strategy of NVG should aim at keeping the number of necessary follow-up visits and additional interventions at a minimum. We tried to address this need by combining several interventions in a single surgical session.
Retinal ischemia is the primary cause of NVG 10,41 . Accordingly, PRP is the standard of care to reduce posterior pole oxygen demand and angiogenic drive while vitrectomy is performed to increase the partial pressure of vitreous oxygen 42,43 . We performed lens extraction, pars plana vitrectomy, and endoscopic PRP because of significant media opacities and because endoscopic laser through the pars plana approach facilitates the delivery of 360° near-confluent peripheral retinal laser treatment out to the ora serrata. In our clinical experience, even relatively small areas of retinal ischemia may cause further NVG progression. Therefore, we took care to apply 360° PRP to near-confluence up to the ora serrata. Such extensive treatment would be challenging or impossible using standard PRP with an indirect ophthalmoscope or at the slit lamp.
In the healthy eye, the vitreous body and the iris-lens diaphragm form a relative diffusion barrier that maintains a higher panretinal photocoagulation prior to the surgery and in contrast to our study IOP ≤ 21 mmHg was the only success criteria.
Neovascular glaucoma is a major risk factor for failure of trabeculectomy 56,57 . Chronic blood-retinal barrier insufficiency, for example in advanced diabetic retinopathy, in which endothelial damage and the subsequent release of serum proteins occurs, plays an important role 58 . Retinal ischemia also leads to the production of inflammatory mediators 59 , which can lead to filtering bleb scarring and postoperative failure of the trabeculectomy 58 .
Our integrative surgical approach avoided tube-specific complications (e.g., tube exposure, retraction, corneal touch, obstruction), ranging from 13% to 26% in NVG over five years 24,48-51 . It delivers both retina and glaucoma treatment in a single surgical session and reduces the patient and health care system burden by simplifying postoperative care and follow-up. Our telephone survey with all patients lost to follow-up revealed that the main reasons for missed visits were other (general) health issues in 52.4% and long distance to the clinic in 38.1% but a high subjective satisfaction rate.
Patients with NVI stage 4 needed significantly more major interventions than stage 3 or stage 2, but there was no difference between stage 2 and stage 3. This finding suggests that early and comprehensive treatment may be more beneficial than a stepwise treatment strategy.
It is worth noting that IOP reduction can also be achieved without cyclodestruction applying only pars plana vitrectomy, lensectomy with a preserved anterior capsule, and panretinal endophotocoagulation as reported by Kinoshita et al. 31 . However, this study included only 13 eyes with a lower mean preoperative IOP of 29 mmHg as in our study. By contrast, other oxygen partial pressure in the posterior chamber than the vitreous overlying the posterior pole. Concurrently, it reduces the diffusion of angiogenic mediators. Recreating a relative diffusion barrier after vitrectomy 44 is beneficial and reduces NVI occurrence 45 . A relative barrier can be achieved with silicone oil 46 that reduces the incidence of NVG 46,47 . Following this concept, Bartz-Schmidt et al. treated 32 NVG patients with pars plana vitrectomy, retinal, and ciliary body photocoagulation, and silicone oil tamponade as eyes were left aphakic. This approach controlled IOP (defined as an IOP between 8 and 21 mmHg) in 72% of patients for at least one year 30 . In our study we achieved an IOP control in 77.9% and all eyes were pseudophakic and without silicone oil at the conclusion of the surgery, thereby suggesting that silicone oil as a diffusion barrier may not be necessary 30 .
Our success rate of 65%, defined as an IOP <22 mmHg, with or without glaucoma medication and without vision loss after a oneyear follow-up, is similar to success rates reported for glaucoma drainage devices, which range from 62% to 66.7% 24,48-50 . One study reported a 73% success rate in 38 eyes that received a glaucoma drainage device and had relatively few postoperative complications 51 . However, this report is the exception as others have found 24,48-50 . (Table 3).
Trabeculectomy has been one of the most important intraocular pressure lowering operations in glaucoma since the 1960s, so this surgical technique was also used in NVG. Different studies have shown a failure rate of up to 80% 52,53 . The use of mitomycin C (MMC) and 5-fluorouracil (5-FU) led to a higher success rate and was 54% after 18 months for MMC used intraoperatively and 55% after 35 months for 5-FU 54 . A combined trabeculectomy and retinal cryotherapy did not seem to improve the surgical success outcome for NVG in diabetic patients 55 . However, in this study patients had received References studies that only used anti-VEGF agents as primary treatment 28,62 reported failure rates up to 88% 62 . Anti-VEGF agents may be best used as an adjuvant treatment 29 . Consistent with our findings, a study that used pars plana vitrectomy, endoscopic peripheral panretinal photocoagulation, and endocyclophotocoagulation (ECP) also achieved an IOP reduction which was more effective than panretinal photocoagulation, intravitreal bevacizumab, pars plana vitrectomy, and trabeculectomy with mitomycin C or Ahmed valve placement. However, the authors reported a higher phthisis rate of 7.4% 63 .
Because of the underlying ocular disease, visual function remained low in most eyes in our study.
There are limitations to our study. Because the integrative surgical approach described here was the primary practice pattern, there was no control group. This limited us to an intragroup comparison of before versus after treatment data. As a retrospective study, it can only inform on future prospective studies' parameters and design and help formulate, but not answer, hypotheses about associations between treatment and outcomes.
In conclusion, this study shows that NVG can be controlled in most cases by an integrative surgical approach delivered in a single surgical session that combines transscleral cyclophotocoagulation, cataract removal, pars plana vitrectomy, near-confluent full-scatter panretinal photocoagulation, and intravitreal bevacizumab. Patients with advanced iris neovascularization required significantly more additional interventions. The described approach lowered IOP significantly, reduced the number of glaucoma medications, and controlled pain. Open Peer Review glaucoma. The treatment of this clinical entity is difficult and often unsatisfactory. The authors report on a convincing surgical approach and present retrospective results in a relatively large patient group. The presentation of the results is good and the manuscript is well written.

Data availability
Taking into account the reviewer report by Andrew W. Eller, I have no further comments to make.

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes

Are the conclusions drawn adequately supported by the results? Yes
center's experience with aggressive upfront treatment of neovascular glaucoma (NVG) with combined vitrectomy, near confluent pan retinal endolaser, bevacizumab and transscleral cyclophotocoagulation.
The authors hypothesize that upfront combined treatment would improve outcomes in terms of IOP, antiglaucoma medications, pain, necessity of further surgeries and help reduce follow-up visits. They report baseline characteristics and one-year outcomes for 77 patients with neovascular glaucoma that underwent the combined procedure. We congratulate them on implementing a success strategy in treating these complex cases. Often patients with NVG have disastrous outcomes with poorly controlled IOP and significant loss of vision. In this series, a number of patients were able to maintain reasonable vision. It can often be difficult to "titrate" treatment of IOP, often resulting in hypotony. There were three patients with phthisis bulbi and another six with hypotony for a total of nine or 11.7%. Three patients developed retinal detachment. Perhaps these were the same three patients who developed phthisis bulbi.
The major drawback of this study is the lack of alternate treatments for randomization. This limits our ability to draw any conclusions about optimal management strategies and only hypotheses can be generated. This is appropriately acknowledged by the authors who are careful not to overstate their conclusions.
The authors seemed to have corelated surgical success with the stage of iris neovascularization. In other words, there were more surgical interventions with stage 4, then stages 2 or 3. It is the experience of this reviewer that one can corelate treatment success of NVG with presenting IOP and the immediate response of IOP to medical therapy. In clinic, if the initial IOP is less than 40 mmHg and medical treatment can reduce the IOP into the low 20's, these patients can often be treated with an initial injection of an anti-VEGF medication, followed by PRP laser. In the long term, any residual glaucoma can usually be treated with medical therapy alone. Those patients presenting with an IOP greater than 40 mmHg, and not responding to medical therapy will require more aggressive treatment. If the visual potential is reasonable, this treatment may include anti-VEGF, PRP laser (with vitrectomy if visualization of the retina is poor), and glaucoma surgery. If the visual potential is very poor, then not invasive glaucoma treatment such as transscleral CPC) is advised. Using their data, are the authors able to confirm our observations regarding treatment of NVG.
As the authors have stated, there are a number of different underlying causes of neovascular glaucoma and they have been combined in this series. This is a fairly large series and we question whether the severity of the NVG and response to treatment may have differed according to the underlying pathology. For example, the iris neovascularization and glaucoma in central retinal vein occlusions (CRVO) tends to progress more rapidly then seen with diabetic retinopathy. Once the retinal ischemia has been addressed in CRVO, it seems these eyes are more likely to develop hypotony. What was the experience of the authors in this study?
The authors cite two studies from the mid-1980's when the use of silicone oil was very much in its early stages of clinical usage. At that time, it was felt that silicone oil reduced or perhaps prevented the flow of VEGF into the anterior chamber and therefore decreased the risk of iris neovascularization and glaucoma. We believe this theory has been discounted as we know that there is a flow of aqueous humor from the posterior segment into the anterior chamber. Retinal detachment alone increases VEGF production and it occurs in cases of proliferative diabetic retinopathy there can be an even more profound angiogenic effect. An alternative explanation for this observation is that silicone oil improves the outcome in these eyes by simply improving the chances for retinal re-attachment.
In the Introduction, the authors hypothesize that the combined approach would help decrease outpatient visits. This is again emphasized in the Discussion where authors report that patients with neovascular glaucoma have significant comorbidities and lower life expectancy and would therefore benefit from decreased outpatient visits. The counter-argument to be made is that combined aggressive surgical intervention would increase risk of complications and require closer monitoring. No data is provided on the number of follow up visits. Consider adding to results if available. The authors could consider performing a regression analysis to determine which baseline characteristics predicted success with this combined technique. This analysis will help identify patients in whom this technique would be most helpful. It could also show when this technique may be more aggressive than necessary in some cases, and could be more aggressive in others when the addition of a glaucoma drainage device may be beneficial.

Specific comments for revision:
In the Introduction, second paragraph, suggest following change. " In 1994, Miller et al. showed that laser occlusion of all branch retinal veins in a primate could lead to CRVO and subsequent retinal ischemia." 1.
In the Results, consider rephrasing "arterial hypertension" to hypertension as this is the currently accepted terminology.

2.
In Table 1 -consider rephrasing "R stadium" to read "NVI Stage" or Weiss and Gold Rubeosis grading to ensure clarity.

3.
In the Discussion, it would be beneficial to see outcomes with staged surgical approach (Incisional surgical management with trabeculectomy/tubes, followed by cryotherapy if poor response) from existing literature to enable some comparison with the combined approach reported in this manuscript.

4.
In the Discussion, the authors state "Patients with stage 4 anterior chamber neovascularization needed significantly more major interventions than stage 3 or stage 2, but there was a difference between stage 2 and stage 3". The data reports NO significant difference between stage 2 and 3. Please clarify.

5.
Overall, a well-organized manuscript with a practical approach to the treatment of neovascular glaucoma with novel data.

Is the work clearly and accurately presented and does it cite the current literature? Yes
Is the study design appropriate and is the work technically sound? Yes

Are sufficient details of methods and analysis provided to allow replication by others? Yes
If applicable, is the statistical analysis and its interpretation appropriate? Yes R1, 2: No data is provided on the number of follow up visits. Consider adding to results if available.
Authors: Information about the number of follow up visits is provided in section "study design": All patients were treated at the University Eye Hospital in Würzburg, Germany. The combined intervention was part of routine patient care at our clinic and all patients with decompensated neovascular glaucoma and preserved visual function received this intervention. BCVA, intraocular pressure (IOP, mmHg), the number of glaucoma medication, visual analog pain scale (VAPS, 0-10) was recorded at baseline and at follow-up visits at one, three, six, and 12 months as part of routine care.

R1, 3:
The authors could consider performing a regression analysis to determine which baseline characteristics predicted success with this combined technique. This analysis will help identify patients in whom this technique would be most helpful. Authors: Regression analysis was not included in this study on purpose because this analysis is part of our ongoing study.

R1, Specific Comments:
In the Introduction, second paragraph, suggest following change. "In 1994, Miller et al. showed that laser occlusion of all branch retinal veins in a primate could lead to CRVO and subsequent retinal ischemia." ○ In the Results, consider rephrasing "arterial hypertension" to hypertension as this is the currently accepted terminology.

○
In Table 1 -consider rephrasing "R stadium" to read "NVI Stage" or Weiss and Gold Rubeosis grading to ensure clarity.

○
In the Discussion, it would be beneficial to see outcomes with staged surgical approach (Incisional surgical management with trabeculectomy/tubes, followed by cryotherapy if poor response) from existing literature to enable some comparison with the combined approach reported in this manuscript.

○
In the Discussion, the authors state "Patients with stage 4 anterior chamber neovascularization needed significantly more major interventions than stage 3 or stage 2, but there was a difference between stage 2 and stage 3". The data reports NO significant difference between stage 2 and 3. Please clarify.

Authors:
We have changed the manuscript to include these suggestions.

Competing Interests:
No competing interests were disclosed.