Combination Antivascular Endothelial Growth Factor and Modified Panretinal Photocoagulation in Management of Proliferative Diabetic Retinopathy

Abstract

Purpose:

This work evaluates the effects of combined intravitreal antivascular endothelial growth factor (anti-VEGF) and modified panretinal photocoagulation (PRP) for management of proliferative diabetic retinopathy (PDR).

Methods:

This retrospective case series included 37 eyes of 33 patients with high-risk PDR. Anti-VEGF injections (≥ 2) were followed by modified, midperipheral PRP performed in 2 or more sessions. Visual and anatomic outcomes were tracked for 1 year after treatment. Regression analysis was performed for factors predictive of final outcomes.

Results:

Mean visual acuity (VA) at initial and final visit were 20/50 and 20/40 (P = .22), respectively, over a mean follow-up duration of 341.4 days. Central foveal thickness decreased from 321.8 µm to 258.6 µm (P = .01). Resolution of PDR was achieved in 94.6% of eyes, with 5.4% of eyes requiring additional anti-VEGF for persistent neovascularization. Final VA was significantly associated with baseline VA, VA at 1 month, and any adverse anatomical events. Treatment noncompliance was present in 24.3%; compliance decreased with increasing medical comorbidities, but was not significantly associated with final VA.

Conclusions:

Combination of anti-VEGF and modified PRP preserved VA and yielded PDR regression in the majority of eyes. This combination provides rapid PDR regression with anti-VEGF while achieving durable disease suppression in this real-world cohort without traditional PRP.

Introduction

Diabetic retinopathy (DR) is the leading cause of blindness in working-age adults in the United States, and the number of Americans affected is expected to double from 7.7 million in 2010 to 14.6 million by 2050. ¹ Among affected individuals with DR, those with high-risk proliferative diabetic retinopathy (PDR) are at greatest risk for vision loss if left untreated. ^{2 -4} The Diabetic Retinopathy Study (DRS) is a landmark study conducted in the late 1970s and early 1980s that established the role of panretinal photocoagulation (PRP) for treatment of high-risk PDR. ³ This work, along with the Early Treatment Diabetic Retinopathy Study (ETDRS), established treatment and screening guidelines for PDR that have since resulted in almost halving the incidence and progression of PDR. ⁴

Traditional PRP as performed in the DRS involves placement of 800 to 1600 500-µm scatter burns via argon, or alternatively xenon laser, spaced approximately one-half burn apart, and each approximately 0.1 seconds in duration, along with focal treatment of new vessels and microaneurysms. ² This protocol can be completed in 1 or 2 office visits and is relatively cost-effective. However, traditional PRP often results in irreversible laser damage to the peripheral retina (Figure 1) thus leading to poor peripheral and night vision, peripheral visual field loss, and increased risk of developing diabetic macular edema (DME). ⁵

Figure 1.

Widefield fundus images and fluorescent angiography of patient with proliferative diabetic retinopathy of both eyes who received traditional panretinal photocoagulation. (A and B) Fundus images and (C and D) fluorescent angiography were taken the same day (Optos). Photos courtesy of Aristomenis Thanos, MD.

Within the past decade, antivascular endothelial growth factor (anti-VEGF) therapies have taken over the treatment of retinal diseases including age-related macular degeneration and DME. The Diabetic Retinopathy Clinical Research Network (DRCR.net) conducted Protocol S, a prospective clinical trial comparing visual acuity (VA) outcomes in eyes treated with PRP against those treated with the anti-VEGF agent ranibizumab. That study showed ranibizumab was noninferior to PRP for VA change from baseline and superior to PRP in a lower incidence of DME and preservation of the visual field at 2-year follow-up. ⁶ VA was similar between both groups, whereas the difference in visual field was reduced at 5-year follow-up, suggesting that the effects of anti-VEGF therapy were at least partially sustainable in those patients who were not lost to follow-up (66% of enrollees). ⁷

Recently the PANORAMA study, a phase 3 clinical trial using aflibercept, also demonstrated the effect of anti-VEGF therapy with aflibercept on regression of DR. ⁸ Taken together, the Protocol S and PANORAMA trials have provided compelling evidence for use of anti-VEGFs as a treatment tool for PDR and an alternative to PRP without the well-known adverse effects of ablative laser.

The treatment of DR with anti-VEGF agents requires regular and frequent intravitreal injections per eye, which necessitates high patient compliance, carries a significant cost, and confers small but definitive risk of endophthalmitis and possibly systemic adverse effects. ^9,10 Beyond these issues, concerns about long-term durability of anti-VEGF therapy in DR remain as VEGF suppression with current agents requires continued therapy. Obeid et al recently demonstrated that patients who received anti-VEGF monotherapy for PDR but had unintended lapses in treatment follow-up had higher incidence of neovascularization of the iris (NVI) and tractional retinal detachment (TRD), as well as worse VA decline from before the unintended lapse in treatment compared with those who received PRP. ¹¹ Further, Wubben and colleagues studied a small cohort of patients with DR treated with anti-VEGF who experienced a median treatment interruption interval of 12 months. Seventy-seven percent of the eyes developed more than 3-line VA loss, and approximately half of those had hand motion or worse VA at final follow-up. ¹² Thus, the visual loss of patients with PDR managed with anti-VEGF monotherapy who become noncompliant and fall through the cracks is significant and potentially irreversible.

Despite these poor outcomes with treatment interruption, anti-VEGFs in a compliant patient have the potential to effectively manage PDR while preserving peripheral retina. Thus, the treatment of PDR in the diabetic population that is at risk for loss -to -follow-up continues to be a subject of ongoing controversy. Although questions remain about long-term durability of anti-VEGF agents that require ongoing injections in DR, the efficacy and durability of PRP is time-proven over 4 decades of clinical experience and published studies. Thus, PRP is poised to remain an important part of our treatment armamentarium for the foreseeable future despite the impressive results of anti-VEGF monotherapy in DR trials.

In several recent studies, combining anti-VEGFs and traditional PRP resulted in decreased number of PRP treatments, decreased time to clearance of vitreous hemorrhage (VH), and resolution of retinal neovascularization, with either improved or equivalent final VA compared with PRP alone. ^{13 -18} A study by Mansour et al demonstrated that combined anti-VEGF and a modified peripheral PRP, opposed to traditional PRP as described in the DRS, was able to decrease the number of necessary injections for DME compared with anti-VEGF monotherapy and could therefore benefit patients who had poor compliance. ¹⁹ The possible benefits of combining anti-VEGF and modified (“lighter”) PRP rather than monotherapy of either approach include rapid suppression of retinal and/or anterior segment neovascularization, faster visual rehabilitation with preservation of peripheral retinal function and visual field, and improved treatment durability. Anecdotally, a number of practicing retina specialists have already adopted this combination approach since the advent of Protocol S despite a lack of published studies establishing its efficacy or treatment algorithm.

While existing work on combined anti-VEGF and PRP approaches has largely used traditional PRP, there is very limited work on modified PRP in this approach. Additionally, the previous study on modified PRP in a combined treatment regimen was promising but was limited to a sample size of 10 eyes, ¹⁹ and further validation of combined injection and modified PRP, especially in a generalizable cohort, would strengthen the evidence regarding this regimen. In this study, we report real-world experience of using a combined approach with modified PRP for treatment of high-risk PDR and analyze visual and treatment outcomes.

Methods

Patients

This study was designed as a retrospective review of a series of consecutive eyes with high-risk PDR treated with a combination approach (multiple anti-VEGF injections followed by modified PRP laser). The study protocol was granted institutional review board approval by the Western Institutional Review Board (study number 1185536) and adhered to the tenets of the Declaration of Helsinki. Informed consent was not sought for the present study because it does not contain identifiable or personal patient information.

A database search of electronic medical records (IntelleChartPRO, Nextech) was undertaken to obtain a list of all eligible patients with a diagnosis of PDR at a single tertiary referral retina practice (Pennsylvania Retina Specialists, PC) from January 1, 2015 to October 29, 2018. Medical records were further reviewed to determine inclusion and exclusion eligibility. Inclusion criteria were as follows: diagnosis of treatment-naive high-risk PDR at time of treatment initiation, adequate medical records through last follow-up, a minimum of initial 2 anti-VEGF injections (range, 2-6 injections) for treatment of PDR (bevacizumab, ranibizumab, aflibercept, or a combination were allowed), a minimum of 2 PRP sessions (range, 2-3 sessions) performed after the anti-VEGF injections, a minimum follow-up period of 1 month following the last PRP, and a minimum of 6 months since initiation of treatment. Eyes were excluded if they had a diagnosis other than high-risk PDR (eg, nonproliferative diabetic retinopathy [NPDR], age-related macular degeneration, retinal vein occlusion, retinal artery occlusion, hypertensive retinopathy, uveitis, or history of ocular melanoma with or without radiation retinopathy), prior treatment with either PRP or anti-VEGF agents for any reason, and prior vitrectomy for any reason. As an example, a patient with PDR in one eye and NPDR in the other would have been included in the preliminary screening; however, the eye with NPDR would have been excluded from analysis. When applicable, patients identified through the database search who met inclusion and exclusion criteria but did not have adequate duration of follow-up were monitored to May 2019 and included in analysis once minimum follow-up requirements were achieved.

Measurement of Parameters

The IntelleChartPRO electronic record system was used for data collection. High-risk PDR was diagnosed through clinical examination by 6 fellowship-trained retina specialists based on clinical biomicroscopic fundus examination assisted by either widefield or conventional fluorescein angiography. Presence of DME was determined qualitatively and quantitatively using spectral domain–optical coherence tomography platform and central foveal thickness (CFT) measurements. Uncorrected and pinhole Snellen VA measurements were collected each visit by trained ophthalmic technicians. When these were discordant, the pinhole VA was used as the estimate of best habitual VA and included in data collection and analysis. Snellen VA measurements were converted to logarithm of the minimum angle of resolution (logMAR) equivalents for purposes of statistical analysis. ^{20 -22} Count fingers was assigned 1.98 logMAR and hand motion was assigned 2.28 logMAR. ²³ For representation and reporting, computed mean and median logMARs were converted to Snellen equivalents ^20,21 and changes in VA were converted to letters of improvement (ETDRS). ^20,24

Treatment Protocol

Eyes received intravitreal injections of bevacizumab (1.25 mg) (Genentech), ranibizumab (0.3 mg) (Genentech and Novartis), aflibercept (2 mg) (Regeneron), or a combination of these anti-VEGF agents at baseline and at intervals of at least 4 weeks. Treatment-naive patients received a minimum of 2 anti-VEGF injections for active high-risk PDR, and those with DME received additional injections until appropriate resolution of DME was achieved. The total number of anti-VEGF injections in eyes with or without DME and the choice of anti-VEGF agent were at the discretion of the treating retina specialist. Thereafter, modified PRP using an Iridex argon laser was performed. The PRP was targeted to achieve medium to light gray burns of approximately 500 µm in size, one-half to 1 burn width apart, and approximately 2 to 3 disc diameters outside the arcades, that extend peripherally to the far periphery. The treatment end point was complete peripheral retinal scatter photocoagulation, achieved in approximately 800 to 1600 spots (Figure 2). All laser treatment was performed as single-shot, nonpattern laser.

Figure 2.

Widefield fundus images and fluorescent angiography of study patient with proliferative diabetic retinopathy of both eyes. (A and B) Fundus images and (C and D) fluorescent angiography were taken the same day (Optos). (A and C) The posttreatment right eye was 20/25 at follow-up (20/25 at baseline), and (B and D) the untreated left eye was 20/40 at follow-up (20/20 at baseline).

Statistics

Data were analyzed using Microsoft Excel. GraphPad Prism 5 was used to perform simple linear regression, Wilcoxon rank sum test for comparison between 2 groups of data, and Kruskal-Wallis test for comparison between more than 2 groups of data. Data were reported as mean ± SD and as median (interquartile range). Statistical significance was assumed with a P value less than .05.

Results

Patient Demographics and Baseline Characteristics

A total of 1080 eyes of 540 patients with a diagnosis of PDR and an office visit between January 1, 2015 and October 29, 2018, were screened for eligibility. Following screening, 37 eyes of 33 patients met inclusion and exclusion criteria and were included in data collection and analysis. The reasons for exclusion were prior PRP (65.2%), alternate diagnosis (10.4%), prior vitrectomy (8.9%), insufficient number of injections prior to PRP (7.4%), insufficient PRP only (6.7%), and insufficient follow-up duration after final PRP (1.1%). Some eyes met multiple exclusion criteria. The mean age of patients who met inclusion and exclusion criteria was 57.6 ± 11.5 years, and 54.5% were female. The majority of patients were white (81.8%) and most had type 2 diabetes mellitus (93.9%). The average number of systemic medical comorbidities per patient was 4.7 ± 2.5, with 2.9 ± 1.3 of these being cardiovascular comorbidities, and 30.3% of patients had chronic kidney disease. The mean baseline best-corrected VA was 0.42 logMAR (Snellen equivalent, 20/50), and the mean CFT was 321.8 µm.

DME was not present, present but noncenter involving, and present and center involving in 27.0%, 18.9%, and 54.1% of included patients, respectively. Two patients presented with extramacular diabetic TRD. Two additional patients presented with neovascular glaucoma (NVG) at baseline and neither had received prior surgical intervention for NVG. One of the patients with NVG was already scheduled for cyclophotocoagulation with a glaucoma specialist at the time of the initial visit. Detailed baseline characteristics are summarized in Table 1.

Table 1.

Baseline Characteristics.

Patient characteristics		Value N = 33 patients
Age, y	Mean (SD)	57.6 (11.5)
	Median (IQR)	60 (16)
Sex	Female, No. (%)	18 (54.5)
	Male, No. (%)	13 (45.5)
Ethnicity, No. (%)	White	27 (81.8)
	Black/African American	3 (9.1)
	Asian	1 (3.0)
	Other	3 (9.1)
Type of diabetes	Type 1	2 (6.1)
	Type 2	31 (93.9)
Hemoglobin A1C	Mean (SD)	8.1 (1.9)
	Median (IQR)	7.5 (2.4)
Comorbidities	Totala, No. per patient, mean (SD), median (IQR)	4.7 (2.5), 4 (3)
	Cardiovascular, mean (SD), median (IQR)	2.9 (1.3), 3 (2)
	Chronic kidney disease, No. (%)	10 (30.3)
	Has not required dialysis	5 (15.2)
	Requiring dialysis/postrenal transplant	5 (15.2)
Ocular characteristics		N = 37 eyes
Diagnosis, No. (%)	Nonproliferative diabetic retinopathy	0 (0.0)
	Active proliferative diabetic retinopathy	37 (100.0)
Lens status, No. (%)	Phakic	28 (75.7)
	Pseudophakic	9 (24.3)
Visual characteristics		logMAR Snellen
Visual acuity	Mean (SD)	0.42 (0.40) 20/50
	Median (IQR)	0.30 (0.50) 20/40
CFT, µm	Mean (SD)	321.8 (119.6)
	Median (IQR)	281.0 (119.0)
Presence of DME, No. (%)	None	10 (27.0)
	Noncenter involving	7 (18.9)
	Center involving	20 (54.1)
Presence of VH, No. (%)	Any	10 (27.0)
Location of neovascularization, No (%)	Anterior segmentb	2 (5.4)
	Disc	14 (37.8)
	Elsewhere	29 (78.4)

Abbreviations: CFT, central foveal thickness; DME, diabetic macular edema; IQR, interquartile range; logMAR, logarithm of the minimum angle of resolution; VH, vitreous hemorrhage.

^a Total comorbidities included cardiovascular (hypertension, coronary artery disease, peripheral vascular disease, aortic stenosis, atrial fibrillation, congestive heart failure, myocardial infarction, cerebrovascular incident, diabetes, hyperlipidemia, and hypercholesterolemia) as well as noncardiovascular (anxiety, depression, seasonal allergies, asthma, chronic obstructive pulmonary disease, sleep apnea, osteoarthritis, gout, psoriatic arthritis, unspecified arthritis, Sjögren syndrome, hypothyroidism, Lyme disease, gastroesophageal reflux, unspecified gallbladder disease, Crohn disease, colorectal cancer, gastric cancer, basal cell cancer, unspecified cancer, trigeminal neuralgia, and carpal tunnel syndrome) comorbidities.

^b Both cases had neovascularization of the iris as well as neovascular glaucoma. One case was scheduled for cyclophotocoagulation with a glaucoma specialist at time of presentation.

Treatment Course

All eyes received a minimum of 2 anti-VEGF injections, with an average of 3.8 ± 1.5 injections per eye prior to the first PRP session. The anti-VEGF medications administered were bevacizumab (24.3%), ranibizumab (48.6%), aflibercept (21.6%), or in a minority of cases a combination of 2 different agents, mainly because of changes in medical insurance coverage. The average number of PRP sessions was 2.3 ± 0.5 per eye. Because this was a retrospective study with multiple treating physicians, patient follow-up schedules were not identical and were determined by clinical judgment and discretion of the treating retina specialist. However, 35 (94.6%) study eyes received follow-up within the first 2 months of PRP completion. Nine patients (24.3%) missed 1 or more appointments over the duration of follow-up. A total of 17 (45.9%) eyes required additional anti-VEGF treatment after PRP was completed. The most common indication for additional anti-VEGF treatment was persistent center-involving DME (15 of 17 eyes, 88.2%). Two of these 17 eyes (11.8%) required additional anti-VEGF for persistent neovascularization. The average number of days between PRP completion and first additional anti-VEGF injection was 56.1 and 375 days for DME and recurrent retinal neovascularization, respectively. The treatment course is summarized in Table 2.

Table 2.

Treatment Course and Outcomes.

Intravitreal anti-VEGF treatment		Value N = 37 eyes
Injections, No.	< 2	0 (0.0)
	≥ 2	37 (100.0)
	Mean (SD)	3.8 (1.5)
	Median (IQR)	4 (2)
Medication, No. (%)	Bevacizumab	9 (24.3)
	Ranibizumab	18 (48.6)
	Aflibercept	8 (21.6)
	> 1a	2 (5.4)
PRP sessions
	Mean (SD)	2.3 (0.5)
	Median (IQR)	2 (1)
Additional anti-VEGF treatment after PRP
	No. needing additional anti-VEGF (%)	17 (45.9)
Indication: DME	No. (%)	15 (40.5)
	Days to first additional injection, mean (SD)	56.1 (56.6)
	Days to first additional injection, median (IQR)	30 (38.5)
Indication: persistent neovascularization	No. (%)	2 (5.4)
	Days to first additional injection, mean (SD)	375 (174)
	Days to first additional injection, median (IQR)	375 (174)
Any missed appointments for duration of follow-up, No. (%)		9 (24.3%)

Abbreviations: anti-VEGF, antivascular endothelial growth factor; DME, diabetic macular edema; IQR, interquartile range; PRP, panretinal photocoagulation.

^a Medications were changed to ensure insurance coverage.

Outcomes of Treatment

The mean duration of follow-up was 341.4 days (range, 28-1208 days). At the final visit, mean VA was 0.32 logMAR (Snellen, 20/40) (Table 3). This was a +6.0 letter score improvement from the baseline 0.42 logMAR (Snellen, 20/50) but did not represent a statistically significant change in VA (P = .22). The CFT was significantly decreased from baseline 321.8 µm to 258.6 µm at the final visit (P = .01). DME was largely resolved or noncenter involving, with only 1 eye (2.7%) still exhibiting center-involving edema at last follow-up (Table 3). PDR was inactive in 35 (94.6%) eyes at the final visit (see example shown in Figure 1). Two eyes (5.4%) had active PDR at the final visit and were scheduled for ongoing additional treatment with fill-in PRP.

Table 3.

Visual and Anatomic Outcomes.

Outcome			N = 37 eyes
Follow-up period since last PRP, d	Mean (SD)		341.4 (348.2)
	Median (IQR)		183 (459)
Lens status, final, No. (%)	Phakic		25 (67.6)
	Pseudophakic		12 (32.4)
Visual acuity		logMAR	Snellen	Letter change
Initial visit	Mean (SD)	0.42 (0.40)	20/50
	Median (IQR)	0.30 (0.50)	20/40
Final visit	Mean (SD)	0.32 (0.45)	20/40	+6.0 (20.0)
	Median (IQR)	0.20 (0.48)	20/30	0 (19)
CFT			Measured	Change
Initial visit, µm	Mean (SD)		321.8 (119.6)
	Median (IQR)		281.0 (119.0)
Final visit, µma	Mean (SD)		258.6 (50.1)	–78.7 (134.5)
	Median (IQR)		257.0 (59.0)	–15 (148.5)
DME at final visit, No. (%)	None		20 (54.1)
	Noncenter involving		16 (43.2)
	Center involving		1 (2.7)
PDR activity at final visit	Active, No. (%)		2 (5.4)
Anatomic outcomes, No. (%)	Vitreous hemorrhage		12 (32.4)
	Vitrectomy for any reasonb		4 (10.8)
	IOP elevationc		2 (5.4)
	Neovascular glaucomac		1 (2.7)
	Diabetic TRDd		1 (2.7)
	Macula sparing		1 (2.7)
	Hyphema		0 (0.0)
	Tube/glaucoma surgery		0 (0.0)
	Endophthalmitis		0 (0.0)

Abbreviations: CFT, central foveal thickness; DME, diabetic macular edema; IOP, intraocular pressure; IQR, interquartile range; logMAR, logarithm of the minimum angle of resolution; PDR, proliferative diabetic retinopathy; PRP, panretinal photocoagulation; TRD, tractional retinal detachment.

^a Central foveal thickness was available for only 31 eyes at final visit.

^b Primary indications for vitrectomy: vitreous hemorrhage (3), tractional retinal detachment (1).

^c One eye in this subset had neovascular glaucoma at initial presentation and was already scheduled with a comprehensive ophthalmologist for cyclophotocoagulation. This eye required antivascular endothelial growth factor (anti-VEGF) treatment between PRP sessions for management of neovascular glaucoma (was not included in number of eyes requiring additional anti-VEGF treatment after PRP sessions).

^d Two eyes presented with macula-sparing TRD at initial presentation that did not progress throughout treatment and follow-up; therefore, these are not reported as complications in this table.

Analysis was also performed on the subset of cases (24 of 37, 64.9%) that had final follow-up at approximately 3 months or more (mean duration of follow-up 491.5 days; range, 84-1208 days). Mean baseline VA for this subset was 0.50 logMAR (0.41 SD; Snellen, 20/60), and final VA was 0.41 logMAR (0.52 SD; Snellen, 20/50). This was a +5.0 letter score improvement from baseline but did not represent a statistically significant change in VA (P = .25). The CFT was significantly decreased from baseline 331.4 µm (97.4 SD) to 258.7 µm (50.4 SD) at the final visit (P = .01). Overall, the data from this subset were comparable to that of the total cohort.

The most common adverse event during the treatment and follow-up course was development of VH, which occurred in 12 eyes (32.4%) (see Table 3). In that group, 4 eyes (10.8%) underwent vitrectomy. Three were performed for visually significant nonclearing VH and 1 was performed for progressive TRD that developed during the treatment and subsequent follow-up. The 2 cases of preexisting extramacular TRD did not progress during the course of the study and thus did not require surgical treatment. Cataract extraction occurred in 3 eyes (8.1%). Two eyes (5.45%) experienced elevation in intraocular pressure; however, one of those eyes had preexisting NVG that was already scheduled for cyclophotocoagulation by an outside glaucoma specialist at the initial visit. Of note, the other study eye that presented with NVG did not progress to needing any glaucoma intervention following the combination treatment in this study and was effectively managed medically. There were no cases of hyphema, rhegmatogenous retinal detachment, or endophthalmitis.

Factors Associated With Visual Outcome

Baseline VA was significantly associated with VA at the final visit (P = .02) (Table 4). VA at 1 month after the final PRP session could not be collected on all eyes because of the retrospective nature of the study, as not all patients had scheduled follow-up at 1 month. Seventeen of 37 study eyes had 1-month follow-up after the last PRP session. For these 17 eyes, VA at the 1-month visit was significantly associated with final VA (P < .001). The occurrence of any adverse anatomic outcomes, defined as nonclearing VH, intraocular pressure elevation, NVG, TRD, or need for vitrectomy for any of these adverse anatomic outcomes, also had significant association with final VA (P = .01). Subcategory analysis of any association between VH or vitrectomy for any reason with final VA revealed no statistically significant association (P = .07, P = .67, respectively), whereas other subcategories were not analyzed because of their very low incidence.

Table 4.

Analysis of Factors Associated With Visual Acuity Outcome.

Factor	P e
Baseline patient characteristics
Age, ya	.27
Sexb	.22
Ethnicityb	.68
Type of diabetesc	NA
Hemoglobin A1C a	.58
No. comorbidities, totala	.98
No. comorbidities, cardiovasculara	.39
History of CKDb	.86
Initial ocular characteristics
Lens statusb	.42
VAa	.02
CFTa	.8
DMEb	.14
Vitreous hemorrhageb	.18
Neovascularization of discb	.73
Neovascularization elsewhereb	.75
Neovascularization of anterior segmentc	NA
Procedure
Choice of medicationb	.31
No. of injectionsa	.84
No. of PRPa	.13
Requiring any additional injectionsb	.91
Indication: DMEb	.93
Indication: persistent neovascularizationc	NA
Any missed appointmentsb	.24
Any complicationsb	.01
Development of vitreous hemorrhageb	.07
Underwent vitrectomy for any reasonb	.67
Neovascular glaucomac	NA
Diabetic TRDc	NA
Days since last PRPa	.66
VA at 1 moa,d	< .001

Abbreviations: CFT, central foveal thickness; CKD, chronic kidney disease; DME, diabetic macular edema; NA, not available; PRP, panretinal photocoagulation; TRD, tractional retinal detachment; VA, visual acuity.

^a The effect of these continuous variables on final VA was analyzed using simple linear regression.

^b Wilcoxon rank sum test or Kruskal-Wallis test was used to compare between 2 groups and greater than 2 groups, respectively. Comparisons within each characteristic or procedure were performed as follows: sex (male, female); race (white, Asian, African American, other); history of CKD (no, yes); lens status (phakic, pseudophakic); DME (none, noncenter involving, center involving); vitreous hemorrhage (not present, present); neovascularization of disc (not present, present); neovascularization elsewhere (not present, present); choice of medication (ranibizumab, bevacizumab, aflibercept); requiring any additional injections (no, yes); additional injections were indicated for DME (no, yes); complications (not present, present); development of vitreous hemorrhage (no, yes); underwent vitrectomy for any reason (no, yes).

^c These subgroup analyses were not performed because the number of cases within at least one of the groups for comparison was fewer than 3.

^d VA at 1 month following last PRP was analyzed for patients who were scheduled and arrived for 1-month follow-up and for whom the 1-month visit was distinct from their final visit. N = 17 eyes.

^eBold values signify statistical significance (P < .05).

All other factors, including age, sex, ethnicity, hemoglobin A_1C, number of all comorbidities, number of cardiovascular comorbidities, current or history of chronic kidney disease, initial ocular characteristics other than VA, number of injections, number of PRP sessions, need for any additional anti-VEGF injections, any missed appointments, and interval between last PRP session and final follow-up visit did not exhibit a statistically significant correlation (see Table 4). The choice of anti-VEGF medication also did not yield statistically significant difference in final VA (P = .31).

Effect of Missed Appointments

Patients who had missed appointments, which we defined as “no shows,” as opposed to canceled visits with prior notification, did not have significantly worse final VA compared with those who did not have any missed appointments (see Table 4). However, patients who missed appointments developed a significantly higher incidence of center-involving DME at the final visit (P = .02) and VH (P = .03) compared with those who did not miss appointments. There was no statistically significant difference between the “no missed appointments” and “missed appointments” groups in terms of final CFT, occurrence of any complications, or need for vitrectomy for any reason (Table 5). Further comparison between the 2 groups of patients showed that the total number of medical comorbidities was significantly higher in those who missed appointments than those who did not (P = .02). There was no statistically significant correlation between whether patients missed appointments and the number of cardiovascular comorbidities they had (P = .23) (see Table 5).

Table 5.

Comparison of Visual and Anatomical Outcomes by Whether Any Appointments Were Missed.

		No missed appointments		Missed appointments
Outcome	P	Mean (SD)	Median (IQR)	Mean (SD)	Median (IQR)
Final VA	.24
Final CFT	.13
Final DMEa	.02	0.4 (0.5)	0 (1)	0.9 (0.6)	0.5 (1)
Any complications	.09
Vitreous hemorrhageb	.03	0.3 (0.4)	0 (0.8)	0.7 (0.5)	1 (1)
Vitrectomy, any reason	.22
TRDc	NA
IOP elevationc	NA
NVGc	NA
No. of comorbidities (any)	.02	3.9 (1.6)	4 (2.8)	7.1 (3.9)	6 (7.5)
No. of comorbidities (cardiovascular)	.23

Abbreviations: CFT, central foveal thickness; DME, diabetic macular edema; IOP, intraocular pressure; IQR, interquartile range; NA, not available; NVG, neovascular glaucoma; TRD, tractional retinal detachment; VA, visual acuity.

^a This noncontinuous outcome was assigned numerical representation for analysis (none = 0, noncenter involving = 1, center involving = 2).

^b This noncontinuous binary outcome was assigned numerical representation for analysis (vitreous hemorrhage not present = 0, vitreous hemorrhage present = 1)

^c These subgroup analyses were not performed because the incidence of these complications was fewer than 3. The preexisting NVG and TRD cases were not included in this analysis of complications.

Conclusions

For the past 4 decades, treatment of high-risk PDR has traditionally involved DRS-style PRP. The advent of anti-VEGF agents has introduced new alternatives by targeting a key downstream effector of PRP therapy—VEGF secreted by retinal tissue in response to prolonged ischemia or hypoxia. Results of phase 3 clinical trials have provided compelling data on safety and efficacy of anti-VEGFs in producing DR regression while preserving peripheral retina and visual field and affecting superior VA outcomes compared with traditional PRP. ^{6 -8} Despite this paradigm shift, anti-VEGFs require ongoing therapy and high compliance on the part of patients. Thus, patients with severe DR on anti-VEGF monotherapy who are lost to follow-up may experience worse visual outcomes compared with patients who had undergone traditional PRP. ^11,12

The long-term effects of repeated anti-VEGF suppression in patients with diabetes are to be determined but may include variable response to therapy in part secondary to VEGF diversity, changes in ligand, and receptor interactions in retinal signaling pathways, some of which are involved in survival response to injury, systemic availability, and risk of endophthalmitis. ^{25 -27} For all these reasons, PRP alone or in combination with anti-VEGFs may be more suitable for certain patient populations.

Our protocol combined both anti-VEGF and modified PRP with the goal of decreasing the need for chronic, short-interval anti-VEGF injections to treat persistent neovascularization. Furthermore, the modified PRP preserved more peripheral retina compared with traditional PRP. We have shown that the combination of anti-VEGF followed by modified PRP prevents decline in VA and may even improve vision (mean initial VA: 20/50 to mean final VA: 20/40, +6 ETDRS letter improvement, P = .22) in an average follow-up period of 341 days. For reference, the baseline and final VA of eyes enrolled in Protocol S was 20/32 both in ranibizumab and PRP groups. ⁶ Over the study period, 17 (45.9%) eyes required additional anti-VEGF injections; however, only 2 of these cases (5.4%) were for persistent retinal neovascularization, and these were performed at 375 ± 175 days after completion of the treatment course (see Table 2). The number of cases requiring additional injections for this indication was low, but the interval was comparable to that of the PRP arm of Protocol S, which was approximately 7 months. ⁶ Of note, 45% of the Protocol S PRP group required additional PRP.

The present work demonstrates several findings with potentially important applications to clinical practice. Poor baseline VA and incidence of any of a collective selection of adverse anatomical events correlated with poor final VA. We did not demonstrate a statistically significant effect of any specific adverse anatomical event. The incidences of TRD in our study (1, 2.7%) and NVI (1, 2.7%) were comparable to those of the PRP arm of the recent study by Obeid et al, which were 1 (2.1%) and 1 (2.1%), respectively. ¹¹ This was markedly lower than the incidences of TRD (10, 33%) and NVI (4, 13.3%) in the anti-VEGF monotherapy arm, and indicates that the combined approach is able to yield durable anatomical and functional outcomes similar to treatment with PRP.

Interestingly, although we showed that missing appointments resulted in a higher incidence of VH and final DME, it did not have a statistically significant effect on final VA (see Table 5). This may indicate the robustness of the combined treatment approach toward patient populations of varying levels of compliance who otherwise are at high risk for vision loss and poor anatomic outcomes. ^11,12 Additionally, monitoring the treatment response after 1 to 2 months is important to predict final VA (see Table 4) and need for additional anti-VEGF for DME (see Table 2). Overall, both types of monotherapy as well as our combination approach are reasonable management strategies for PDR, and the choice of treatment should take into account the likelihood of patient compliance with the prescribed treatment course and factors that may affect that compliance, such as the number of medical comorbidities. Interestingly, 24.3% of our cases missed appointments, which is more representative of the real-world PDR patient population than controlled clinical trial study populations.

Much of our patient population are likely poor candidates for anti-VEGF monotherapy because they have factors, such as multiple comorbidities (see Table 5), that correlate with noncompliance. We expect that patients who can return to the clinic several times during an initial interval but may fail long-term follow-up may benefit from the combined approach. Patients with a high likelihood to never follow up after the initial visit may benefit from immediate full PRP.

Owing to the retrospective nature of the study, we were unable to randomly assign the participants or blind the investigators. This also limited the outcome measures available for analysis. In particular, visual field and scotopic vision were not readily available for the majority of the patients in this retrospective analysis. The number of cases that met the enrollment criteria was limited, and the study may have been insufficiently powered to reveal differences in our baseline, procedural, and outcome parameters. Additionally, the study cohort comprised patients treated by different retina specialists in the practice and, although the clinicians aimed to preserve consistency, variations in the laser technique may exist. However, this interindividual variation likely reflects the real-world application of PRP ²⁸ and contributes to the robustness and generalizability of our data.

Future work that would allow testing of visual field and scotopic vision would provide evidence regarding the functionality of the regions of the retina that are spared in our modified PRP. In particular, given recent work showing a reduced difference in visual field between PRP and anti-VEGF groups during a 5-year follow-up, the effect of our modified protocol on peripheral vision may benefit from multiyear follow-up. Furthermore, larger cohort studies—preferably prospective, randomized clinical trials—would clarify whether the combination anti-VEGF and midperipheral PRP protocol is noninferior, and help further identify subpopulations that may benefit from our approach over either monotherapy.

Advances in anti-VEGF therapy have given rise to new approaches of treating high-risk PDR. Our study describes a combination therapy involving anti-VEGF followed by modified PRP that is able to preserve VA and regress neovascularization in 94.6% of study eyes with high-risk PDR and variable levels of compliance. Although larger, prospective trials are needed to further validate our findings, the combination method may be a promising avenue toward durable resolution of PDR while preserving more peripheral retinal function and achieving sustained and durable disease suppression in patients with diabetes at high risk for vision loss.