Introduction

Diabetic retinopathy (DR) is the leading cause of new onset vision loss in adults 20–74 years of age in the United States and a leading cause worldwide [1, 2]. Among the US population 40 years and older with diabetes, 28.5 % have some degree of retinopathy, and 1.5 % have proliferative diabetic retinopathy (PDR) [1]. A meta-analysis showed that the global prevalence is even higher with 34.6 % of individuals 20–79 years old with any DR and 6.96 % with PDR [3]. Although recent advances in management have helped to lessen progression to PDR and vision-threatening levels of retinopathy [4, 5], the risk of retinopathy progression is still alarmingly high. Six-year follow-up of an African American cohort found that 15.0 % had developed PDR over that time period, and a study in an English population reported a similar but slightly lower incidence of 11.0 % over 10 years [6, 7]. Twenty-five year follow-up of type 1 diabetics from the Wisconsin Epidemiologic Study of Diabetic Retinopathy showed that 42 % had progressed, although the young age at diagnosis (<30 years old) may inflate the progression rate relative to the overall diabetic population [8]. Nevertheless, these studies have shown that more severe retinopathy at baseline is associated with greater risk of progressing to PDR, which is particularly concerning given the significant underutilization of vision care services by individuals diagnosed with diabetes, especially African Americans [9, 10]. Furthermore, worsening disease severity is significantly associated with negative impact on health-related quality of life, higher risk of cardiovascular death and non-fatal events, and greater cognitive decline [11–13].

A number of randomized clinical trials have recently established the effectiveness of anti-vascular endothelial growth factor (VEGF) therapy, particularly ranibizumab, in the treatment of diabetic macular edema (DME) [14••, 15, 16, 17]. Once PDR develops, panretinal photocoagulation (PRP) remains the standard of care established by the Diabetic Retinopathy Study (DRS) [18]. While efficacious, PRP has a number of potential adverse effects, including exacerbation of macular edema, peripheral visual-field defects, and nyctalopia [19], and its administration can be difficult in the presence of vitreous hemorrhage (VH), a complication of retinal neovascularization [20, 21]. Other sight-threatening complications of PDR are tractional retinal detachment (TRD), which, along with non-clearing VH, can be managed by vitrectomy, and neovascular glaucoma [22]. There are currently no pharmacologic agents FDA approved for the treatment of PDR, but there is increasing interest in their utility considering the large public health burden PDR poses, the shortcomings and risks of current treatments, and the efficacy of anti-VEGF therapy for DME. This review will summarize the current literature on the use of pharmacotherapy for the prevention and treatment of PDR.

Anti-VEGF Therapy

VEGF in Proliferative Diabetic Retinopathy

Chronic hyperglycemia in diabetes activates multiple aberrant biochemical and cellular pathways resulting in microvascular dysfunction and retinal ischemia. In the absence of oxygenation, tissues upregulate VEGF, a potent signaling molecule that plays an integral role in angiogenesis by inducing endothelial cell proliferation, migration, and tube formation. Even before retinal neovascularization develops, increased VEGF in DR leads to degradation of endothelial cell–cell adhesions, which causes vessel leakage and macular edema [23].

Samples from human eyes have established the importance of VEGF in PDR. Significantly elevated levels of VEGF were demonstrated in vitreous and ocular-fluid samples collected at the time of intraocular surgery from patients with active PDR when compared to individuals with quiescent PDR, NPDR, or non-diabetic patients [24, 25]. Injection of VEGF165, an abundant splice variant of VEGF, into the eyes of non-human primates caused development of phenotypic features of NPDR and PDR, including pre-retinal and iris neovascularization [26–28]. Control eyes did not develop any DR-related pathology, and a dose–response relationship with VEGF was observed. Finally, murine and non-human primate models of ischemic retinopathy were used to demonstrate that VEGF-neutralizing antibodies reduced retinal and iris neovascularization [29, 30]. Taken in aggregate, these studies provide a strong basis for investigating anti-VEGF agents for PDR.

Diabetic Retinopathy Progression

As previously mentioned, ranibizumab has been extensively studied for the treatment of DME, and while changes in visual acuity were the primary outcomes of these studies, several randomized clinical trials have included DR progression as a secondary outcome. These results provide insight into the efficacy of anti-VEGF agents in preventing or delaying onset of PDR. RIDE and RISE [14••], two parallel phase III multicenter trials, randomized 759 participants with DME and study eye BCVA 20/40-20/320 into one of three treatment arms: sham injections, 0.3-mg intravitreal ranibizumab, and 0.5-mg intravitreal ranibizumab. Injections were given monthly, and starting at month 3, all patients were evaluated monthly for as-needed macular laser based on protocol-specified criteria. One outcome for evaluating DR progression was the ETDRS severity scale level that was assigned based on grading of color fundus photos. At the 2-year primary endpoint, 15.0 and 11.5 % of participants in the sham groups of RISE and RIDE had progressed to PDR, while 1.6–5.6 % in the ranibizumab groups of both trials progressed (p = 0.0001–0.0206 for comparisons with the sham groups).

A separate analysis by Ip et al. [31••] found that the ranibizumab groups were significantly less likely to worsen and significantly more likely to improve by ≥2 and ≥3 steps on the severity scale. In addition, the median severity level decreased in both ranibizumab groups while it remained at “moderately severe NPDR” in the sham groups, and the percent of patients with mild NPDR increased more substantially in the ranibizumab groups compared to the sham group. Other outcomes indicating progression to PDR were also considered, including receiving PRP, occurrence of VH, performance of vitrectomy for PDR-related sequelae, and development of iris or retinal neovascularization-related complications. Combining these other outcomes resulted in 33.8 % of sham-treated eyes versus 11.2–11.5 % of ranibizumab-treated eyes developing PDR (p < 0.0001). Sham-treated patients were more likely to receive PRP, develop VH, or undergo vitrectomy.

RIDE and RISE were only randomized for the first 2 years since patients in the sham groups were allowed to crossover to receive 0.5 mg ranibizumab. Nevertheless, results at the end of year 3 showed that in both ranibizumab groups, 37.8–40.9 % of eyes improved by ≥2 steps and 11.3–15.4 % by ≥3 steps [15]. This is compared with 23.4 and 24.3 % in RIDE and RISE, respectively, who had an improvement of ≥2 steps and 2.6 and 4.0 % for ≥3 step improvements among the sham groups. Finally, the sham crossover group demonstrated improvements in DR severity after receiving ranibizumab.

The DRCR.net has also investigated the efficacy of ranibizumab for preventing retinopathy progression. A study of 854 eyes of 691 participants with DME compared 0.5 mg intravitreal ranibizumab plus prompt (3–10 days after injection) or deferred (≥24 weeks after injection) focal/grid photocoagulation and 4 mg intravitreal triamcinolone plus prompt macular laser with sham injection plus prompt laser [16]. Injections were given every 4 weeks for the first 12 weeks and then were administered PRN as often as every 4 weeks based on a specified protocol for retreatment. At 1 year, ranibizumab-treated eyes were less likely to demonstrate progression of retinopathy compared to the sham group based on color fundus photo grading. Furthermore, the ranibizumab group was less likely to have a VH or receive PRP compared to the sham group.

Follow-up of the initial study population was continued for 3 years, and an exploratory analysis was performed to evaluate the risk of worsening of DR [32•]. The criteria for progression to PDR were expanded to include: worsening by ≥2 steps on the ETDRS severity scale, receiving PRP, development of VH, or vitrectomy for PDR. There were 538 of 792 eyes (67.9 %) that did not have PDR at baseline, and among these eyes, the ranibizumab groups had significantly lower cumulative probabilities of progressing to PDR at the end of years 1, 2, and 3. Similar results were observed in the cohort with baseline PDR, and statistically significant hazard ratios of 0.43 and 0.42 were calculated for ranibizumab plus prompt or deferred laser, respectively, compared to the sham group. A major limitation to this study is that fundus photographs at 2 and 3 years were available for only 6 and 50 % of the eyes, respectively, due to a change in the protocol related to the beneficial effect of ranibizumab on DME. As such, the proportion of individuals at these time points with worsening of DR are likely underestimated.

Adjunct to PRP

PRP is hypothesized to exert its effect by improving oxygenation of the inner retina, subsequently decreasing VEGF production and inducing regression or stabilization of neovascularization [33]. Anti-VEGF agents have also been reported to induce regression of retinal neovascularization in PDR. Uncontrolled case series have demonstrated that bevacizumab could induce complete regression of neovascularization, as assessed by ophthalmoscopy and/or fluorescein angiography, in 61.4–100 % of participants. Regression could occur as early as 1 week and as late as 1 month post-injection, with cases of recurrence 2 weeks to 3 months after initial treatment [34–40]. Filho et al. [41] conducted a randomized controlled trial comparing PRP to PRP with ranibizumab given as a single 0.5 mg injection after the first of two laser sessions; retreatment with ranibizumab in the combined group and additional PRP in the laser only group were given at 16 and 32 weeks after baseline. At 48 weeks, total area of fluorescein leakage, the primary outcome, was reduced in both groups compared to baseline, but the reduction was significantly greater with PRP plus ranibizumab (p = 0.0291). This group also demonstrated better visual acuity outcomes and a reduction in central subfield macular thickness (CSMT) compared to the PRP only group, which showed worsening of visual acuity and an increase in CSMT. A similar study with bevacizumab demonstrated a significant reduction in neovascularization but no differences in BCVA compared to PRP alone [42].

Another potential role for anti-VEGF as an adjunct to PRP is for the synergistic effect it may have on reducing exacerbations of macular edema. One of the largest studies, conducted by the DRCR.net, compared ranibizumab and triamcinolone with sham injections in patients with severe NPDR or PDR and existing DME who received focal/grid photocoagulation at baseline and PRP at least 2 weeks later [43•]. At the 14-week primary endpoint, the ranibizumab group (n = 113) had a mean visual acuity change of +1 letter compared to −4 letters for the sham group (p < 0.0001), and the mean CSMT decreased by significantly more in eyes treated with ranibizumab (−39 µm vs. −5 µm for sham, p = 0.01). A retrospective review by Mason et al. [44] found similar results for bevacizumab in 60 eyes with severe PDR receiving two sessions of PRP ± bevacizumab, showing a significantly higher BCVA and significantly lower foveal thickness for the combination group at 24 weeks follow-up (p ≤ 0.0001 for both comparisons).

Regression of Rubeosis

In addition to retinal neovascularization, rubeosis in the context of PDR has been noted to respond to anti-VEGF therapy. A case series of 28 eyes noted that 20 (71.4 %) had rubeosis that responded to bevacizumab, while smaller studies reported higher response rates. Regression could be observed as early as 1 day post-treatment, although some patients required 2 or 3 monthly injections before any effect was seen. Studies that had long enough follow-up reported recurrence in some patients, but 4 patients followed for 12 months did not have any recurrence after receiving 3 monthly ranibizumab injections. Finally, the effect on IOP was variable, with some case series finding that it was well controlled during short-term follow-up and others requiring medical and/or surgical interventions [45–49].

Vitreous Hemorrhage

Anti-VEGF treatments have also been investigated for their use in resolving VH, which can prevent administration of PRP and may necessitate vitrectomy. Uncontrolled case series have reported positive results, with hemorrhage partially clearing as early as 1 week after treatment and remaining clear for 1–9 months post-treatment, with some patients requiring multiple injections [34–37]. A larger study of 40 eyes with persistent VH was compared to a historical control group after receiving one or two intravitreal ranibizumab injections; the treatment group demonstrated a significantly shorter vitreous clear-up time and a significantly lower rate of vitrectomy [50]. The DRCR.net recently carried out a double-masked, randomized controlled trial comparing 0.5 mg ranibizumab to saline since it has been hypothesized that clearance of VH may be due to the intravitreal injection itself and not anti-VEGF activity [51]. At 16 week follow-up, the cumulative probabilities of vitrectomy were not significantly different between the two groups (ranibizumab 12 % vs. saline 17 %, p = 0.37), although the ranibizumab group was more likely to have completed PRP (44 vs. 31 %, p = 0.05) and to experience a greater increase in mean visual acuity (22 letters vs. 16 letters, p = 0.04). At 1 year follow-up, both the cumulative probability of vitrectomy and the mean change in visual acuity did not demonstrate significant differences between the two groups [52]. One limit to this study is that the rates of vitrectomy were lower than expected for both groups, so it may have been underpowered to detect a treatment difference.

Adjunct to Vitrectomy

Anti-VEGF therapy prior to vitrectomy has been investigated for its potential in improving surgical outcomes and reducing postoperative VH. A study by di Lauro et al. [53] randomized 72 eyes of 68 patients to receive a sham injection or 1.25 mg of bevacizumab 7 or 20 days prior to vitrectomy for VH or TRD. Intraoperative management was improved for bevacizumab administered at either time point before surgery compared to the sham group in terms of intraoperative bleeding (8.3 % and 12.5 vs. 79.1 %, p < 0.0001 for both comparisons), use of endodiathermy (8.3 % and 12.5 vs. 54.1 %, p < 0.0001 for both comparisons), and operating time (65 min and 69 vs. 84 min, p = 0.025 and 0.031). Zhang et al. [54] found similar results by conducting a meta-analysis of 8 randomized controlled trials, including the di Lauro et al. study, comparing vitrectomy with and without preoperative bevacizumab. The authors found a shorter overall surgical time (mean difference = −26.89 min, p < 0.0001), smaller number of endodiathermy applications (mean difference = −3.46, p = 0.02), less intraoperative bleeding risk (OR = 0.10, p = 0.003), and lower risk of recurrent VH within the first month (OR = 0.35, p < 0.0001) for preoperative treatment with bevacizumab. No difference in the VH recurrence was observed after the first month. In contrast to this meta-analysis, Ahn et al. [55] found no significant decrease in risk of recurrent VH within 4 weeks of vitrectomy among 107 eyes randomized to preoperative or post-vitrectomy bevacizumab compared with vitrectomy alone (22.2 % and 10.8 vs. 32.4 %, p = 0.087). No significant differences in recurrence rates were observed more than 4 weeks after vitrectomy, either, with the post-vitrectomy bevacizumab group actually experiencing a higher risk of recurrence.

Safety

One of the main safety concerns that has arisen in the context of anti-VEGF use for PDR is the potential for development or progression of TRD, possibly due to rapid neovascular involution with accelerated fibrosis. A retrospective review of 211 patients who had received intravitreal bevacizumab for PDR refractory to PRP found that 11 (5.2 %) developed or had progression of TRD 3–31 days post-treatment, although the authors acknowledged this could have happened by natural history [56]. Van Geest et al. [57] measured VEGF and connective tissue growth factor (CTGF) in vitreous samples from patients with PDR, some of whom received bevacizumab preoperatively; they found that VEGF was negatively correlated and CTGF positively correlated with fibrosis and that patients receiving bevacizumab had more fibrosis (p < 0.0001). There is also a concern about administering anti-VEGF therapy to diabetic patients who are already at increased risk of cardiovascular events and mortality. Large clinical trials have found that ocular adverse events (AEs) are similar in ranibizumab and sham-treated groups, including incidence of TRD, although event rates are typically low making it difficult to rule out a small difference in risk [14••, 15, 16, 17]. The 36-month data from RISE and RIDE also revealed a potentially increased risk of non-fatal cardiovascular events and overall incidence of death in patients receiving the 0.5 mg dose, but this was not found with the FDA-approved dose of 0.3 mg.

Based on current available evidence, it appears that anti-VEGF therapy may have a role in preventing and treating PDR (Table 1). Treated eyes in large clinical trials of DME showed a decrease in the risk of retinopathy progression and even demonstrated improvements on the ETDRS severity scale. These studies, however, were not designed with the primary objective of determining who would benefit most from anti-VEGF injections as prevention for PDR, and any conclusions about disease progression and regression are limited to a patient population with DME at baseline. Once PDR has developed, anti-VEGF therapy can induce regression of retinal and iris neovascularization, although the evidence is largely based on case series. A search of ClinicalTrials.gov {performed August 2014} showed several trials, at various stages of completion, that are evaluating the effects of different combinations of ranibizumab and prompt or deferred PRP on regression of neovascularization or reduction of macular edema (NCT01280929, NCT01941329, NCT01594281, NCT01489189). Pending the results of these studies, anti-VEGF as an adjunct to PRP, or perhaps a first-line treatment, may gradually become a standard of care for high-risk PDR, although the potential increased risk of TRD from inducing rapid regression must be studied further. Finally, current evidence does not support a role for anti-VEGF therapy in treating VH or for reducing recurrence of VH when used preoperatively for vitrectomy, although it may improve intraoperative outcomes.

Table 1 Anti-VEGF therapy in proliferative diabetic retinopathy
Full size table

Anti-Inflammatory Therapy

Rationale

Inflammation has a central role in the pathogenesis of DR. Rat models of diabetes have demonstrated that leukostasis, attraction and adhesion of leukocytes to the vascular wall, is temporally and spatially related to capillary nonperfusion in the retina, and ICAM-1, the molecule mediating leukocyte adherence to the endothelium, is upregulated in diabetic humans and animals [58, 59]. Numerous proinflammatory enzymes (i.e., iNOS, COX-2), transcription factors (i.e., NF-κβ), cytokines (i.e., IL-1β, TNFα), and lipid mediators (i.e., prostaglandins, leukotrienes) are also upregulated. This leads to retinal capillary degeneration and blood-retinal barrier breakdown, which results in edema. VEGF is also proinflammatory, and its expression is increased by cytokine release, suggesting that inhibiting VEGF works on the inflammatory component of DR while anti-inflammatory agents would further reduce VEGF [60].

Corticosteroids

The DRCR.net has performed exploratory analyses of clinical trials focusing on DME to investigate the effect, if any, that triamcinolone has on DR progression. One study in which patients were randomized to 1 or 4 mg of intravitreal triamcinolone or focal/grid photocoagulation calculated the cumulative probability of progressing to PDR based on reading center grading, receiving PRP, developing VH, or worsening by ≥2 steps on the ETDRS scale [61]. At 3 years of follow-up, the probabilities were 35, 30, and 37 % for the 1 mg, 4 mg, and laser groups, respectively, with the only significant difference between the 4 mg and laser groups (p = 0.02). Another study, previously discussed, examining ranibizumab plus prompt or deferred macular laser also compared 4 mg of triamcinolone plus prompt laser [16]. An exploratory analysis showed that among eyes without PDR at baseline, the cumulative probability of worsening retinopathy was significantly lower at 1 year for the triamcinolone group compared to sham plus prompt laser, but this difference was not maintained at 2 and 3 years, with the triamcinolone group ultimately having a higher rate of progression (37 vs. 23 %) [32•]. The subgroup analysis among eyes with PDR at baseline demonstrated a significantly lower probability of progression for triamcinolone at all follow-up visits.

Corticosteroids have also been investigated as an adjunct to PRP to reduce macular edema in eyes with pre-existing DME. The DRCR.net found that during short-term (14 weeks) follow-up, eyes receiving triamcinolone plus macular laser had a significant improvement in mean visual acuity and a significant decrease in mean central subfield thickness compared to laser alone [43•]. Mirshahi et al. [62] did not find any difference in BCVA or central macular thickness from baseline to 6-month follow-up comparing eyes receiving PRP and macular laser ± triamcinolone pre-treatment 1 week earlier. Finally, one study examined posterior sub-Tenon triamcinolone for patients with PDR undergoing vitrectomy for retinal detachment and found no differences in reattachment rate, visual acuity, or postoperative complications compared to a non-randomized control group at 6 months follow-up [63]. Overall, the results regarding corticosteroids are mixed, and given the significant ocular AEs of cataract formation and IOP elevation [16, 43, 61], as well as the efficacy of anti-VEGF therapy for the same indications, routine steroid use does not seem warranted at this time.

Tetracyclines

Microglia are the resident immune cells of the retina and become activated in DR, contributing to the inflammatory process [64–66]. Tetracyclines, best known as antibiotics, have demonstrated anti-inflammatory and neuroprotective effects in animal models and cell culture, including inhibition of microglial morphologic changes at low doses [67, 68]. Scott et al. [69] carried out a small proof-of-concept trial in which 30 patients with severe NPDR or non-high-risk PDR were randomized to 24 months of 50 mg doxycycline or placebo once daily. They found no significant differences between the two groups with respect to change in visual acuity, change in CSMT, or disease progression. The drug was well-tolerated. An uncontrolled Phase I/II study of 5 patients treated with minocycline for DME showed modest improvements in BCVA and CSMT in almost all eyes over 6 months without the need for macular laser and/or anti-VEGF therapy [70]. These results are encouraging and warrant further investigation given the potential for an oral medication to reduce the number of intravitreal injections required.

Pharmacologic Vitreolysis

Retinal neovascularization requires a scaffold to grow onto the posterior surface of the vitreous [71]. Eyes that have a complete posterior vitreous detachment (PVD) are less likely to develop neovascular disease and experience PDR progression [72]. The period when a PVD is evolving, however, is a time of increased risk for the development of complications from PDR. If vitreous liquefaction and proteolysis releasing the posterior vitreous cortex from the internal limiting membrane do not occur in concert, excessive traction on the retina can lead to VH, TRD, and/or DME [73]. Ocriplasmin is a recombinant serine protease FDA approved for one-time use in vitreomacular traction syndrome. The most relevant outcome measure from clinical trials is the rate of total PVD since partial PVD in patients with PDR could still facilitate neovascularization. In the Phase II trials, total PVD was induced in 24 and 31 % of eyes receiving the FDA-approved dose of 125 µg compared with 8 and 10 % in the sham groups; the Phase III trial showed 13.4 % of drug-treated eyes and 3.7 % of sham-treated eyes had total PVD at day 28 [74–76]. Few studies have investigated pharmacologic vitreolysis in DR, and most have used autologous plasmin enzyme, which is larger and less stable than ocriplasmin [77]. Two case series found 38 and 41.7 % of eyes had total PVD after one injection, 50 % with PDR showed significant neovascularization regression, and success of PVD was significantly correlated with a thinner, less reflective posterior vitreous face [78, 79]. Pharmacologic vitreolysis may be more effective in diabetic eyes than in non-diabetic eyes with vitreomacular traction, but larger double-masked, placebo-controlled trials are required to obtain more definitive evidence of efficacy. Further consideration could be warranted in patients with complications of PDR who are not good candidates for vitrectomy.

Future Directions

While targeting VEGF has shown great efficacy in treating DME and promise of preventing PDR or treating its complications, there is increasing interest in other potential targets for drug design. A study of VEGF-independent cytokines in vitreous of PDR patients found that 16 cytokines other than VEGF were significantly associated with the presence of PDR [80]. Topical ketorolac given to patients undergoing vitrectomy for complications of PDR showed an ability to significantly reduce vitreous levels of IL-8 and PDGF-AA, cytokines that may contribute to PDR [81]. The PI3K/Akt/mTOR pathway has been implicated in angiogenesis, and sirolimus has been well-tolerated in small studies for DME, suggesting that it could also be investigated in PDR [82, 83]. Finally, blockade of the renin-angiotensin system with enalapril and losartan in patients with type 1 diabetes produced significant reductions in retinopathy progression by ≥2 steps on the ETDRS scale [84].

Conclusion

FDA approval of pharmacologic agents for prevention or treatment of PDR seems a distant goal, but results of completed and continuing studies may soon advocate for the use of pharmacotherapy for select indications. Anti-VEGF medications appear to be the most promising treatments, but the pathogenesis of DR is complex and involves a plethora of inflammatory mediators, signaling molecules, and cell types (Table 2). Exploring other targets for drug design may expand the PDR armamentarium to maximize visual outcomes and provide greater variety in the methods of administration.

Table 2 Non-anti-VEGF therapies in proliferative diabetic retinopathy
Full size table