Neutralization of Vascular Endothelial

n

PhDObjective: To determine the effect of suppression of vascular endothelial growth factor (VEGF) by monthlyinjection of ranibizumab on posterior retinal nonperfusion (RNP) in patients with diabetic macular edema (DME).

Design: Unplanned retrospective analysis of prospectively collected data from 2 randomized, shaminjection-controlled, double-masked, multicenter clinical trials.

Participants: Six hundred sixty-six patients with DME.Methods: An independent reading center measured the area of RNP on uorescein angiograms obtained in

the phase 3 RISE and RIDE trials.Main Outcome Measures: The percentage of patients with no posterior RNP.Results: The percentage of patients with no posterior RNP decreased in the sham group between baseline

and month 24, but remained relatively stable in the 2 ranibizumab groups. After month 24, the sham groupcrossed over to receive monthly injections of ranibizumab 0.5 mg, and the differences between the sham andranibizumab groups were reduced. The percentage of patients who showed an increase in posterior RNP frombaseline increased over time in all 3 groups, but at a faster rate in the sham group, resulting in statistically sig-nicant differences at every time point between months 3 (9.6% vs. 18.5%; P 0.016) and 24 (16.1% vs. 37.6%;P

Although glucose-induced damage to retinal vessels Efcacy and Safety trials, uorescein angiograms were ob-

Ophthalmology Volume -, Number -, Month 2014almost certainly contributes to DME, evidence has emergedthat damage to retinal vessels also could lead to other changesthat may exacerbate leakage. A relatively simple clinicalexperiment showed that retinal hypoxia contributes to DME,suggesting that reduced perfusion (a consequence of glucose-induced retinal vascular damage) may be such an exacer-bating factor.6 Results from this experiment suggested apotential role for hypoxia-regulated gene products in thedevelopment of DME and led to a clinical trial testing anonspecic antagonist of vascular endothelial growth factor(VEGF) receptors that was found to reduce DME.7 Thedevelopment of ranibizumab (Lucentis; Genentech, Inc,South San Francisco, CA), a Fab antibody fragment thatspecically binds all isoforms of VEGF-A, made it possibleto test the role of VEGF directly. Even in a relatively smalltrial, it was clear that VEGF plays a major role,8 andsubsequent clinical studies conrmed the pathophysiologicrole of VEGF in DME.9,10 The extent of that role was bestshown by the RISE and RIDE trials, in which patients withDME were randomized to receive an intraocular injection ofranibizumab or a sham injection every month for 24months.11 At the month 24 primary end point, patients inboth ranibizumab arms showed a marked decrease in DMEand a mean improvement from baseline best-correctedvisual acuity of 11 to 12.5 letters, compared with 2.3 to 2.6letters in the sham groups. During a subsequent year offollow-up, the ranibizumab treatment groups continued toreceive monthly injections of their assigned dose of ranibi-zumab, and the sham groups received monthly injections ofranibizumab 0.5 mg. Visual acuity benets were maintainedin the ranibizumab groups and improved in the sham cross-over group, but visual outcomes in the sham crossover groupwere still notably inferior compared with eyes that hadreceived ranibizumab from the beginning of the study.12

Retinal vein occlusion (RVO), consisting of central RVO(CRVO) and branch RVO (BRVO), is a retinal vasculardisease in which the occlusion of outow vessels from theretina leads to hemorrhages, cotton wool spots, macularedema, and variable amounts of retinal nonperfusion (RNP).The presence of retinal ischemia in some patients leads toretinal neovascularization and vitreous hemorrhages, similarto what is seen in the advanced stages of diabetic retinop-athy; both of these diseases are classied as ischemic reti-nopathies. Similar to DME, a small trial demonstrated thatVEGF plays a major role in macular edema in patients withRVO.13 This nding was conrmed subsequently by largemulticenter trials.14e17

Despite the clinically signicant benets seen withintraocular injections of ranibizumab in patients with DMEand RVO, many retina specialists have voiced concerns thatthe neutralization of VEGF, a survival factor for endothelialcells, could damage retinal vessels and worsen RNP. In fact,several case reports have suggested that a single injection ofan anti-VEGF agent can cause acute worsening of retinalischemia.18e24 To address this issue in the Ranibizumab forthe Treatment of Macular Edema after Central Retinal VeinOcclusion Study: Evaluation of Efcacy and Safety andthe Ranibizumab for the Treatment of Macular EdemaFollowing Branch Retinal Vein Occlusion: Evaluation of

2vital signs, safety assessments, measurement of best-correctedvisual acuity with the Early Treatment Diabetic Retinopathy Study(ETDRS) protocol,27 contrast sensitivity, intraocular pressure, slit-lamp examination, indirect ophthalmoscopy, and optical coherencetomography. Fluorescein angiography and fundus photographywere performed at baseline and months 3, 6, 9, 12, 18, 24, 30, and36. Beginning at month 3, all patients were evaluated monthly forthe need for macular laser according to protocol-specied criteria.The primary end point was at month 24. Patients were followed upfor an additional 12 months, during which those randomized to oneof the ranibizumab arms received monthly injections of theirassigned dose and those assigned to sham had the option to crossover and receive monthly injections of ranibizumab 0.5 mg.

Measurement of Retinal Nonperfusion

As part of an exploratory analysis, the area of posterior RNP (alsoreferred to as retinal capillary nonperfusion) was measured onuorescein angiography by masked certied graders at theUniversity of Wisconsin Fundus Photograph Reading Center,Madison, Wisconsin, using a modication of the methodology ofthe ETDRS uorescein angiogram grading protocol28 as previouslytained at baseline and at months 3, 6, 9, and 12, and an in-dependent reading center measured the area of posterior RNPmasked with respect to treatment group. The percentage ofpatients with CRVO and no RNP increased slightly in theranibizumab groups and decreased in the sham group,resulting in a signicant difference at month 6.25 After month6, the sham group was able to receive ranibizumab, and thedifference between groups was eliminated. This nding wasreplicated in patients with BRVO.

As noted previously, diabetic retinopathy is an ischemicretinopathy and shares some features with RVO; however indiabetic retinopathy, the onset and progression of RNP aremore gradual. Furthermore, another reason exists for prog-ression of RNP in diabetic retinopathy: damage to pericytesand endothelial cells from sustained high levels of glucosein the retina. Therefore, it was uncertain as to whetherneutralization of VEGF would have an ameliorative effecton progression of RNP in diabetic retinopathy, as was thecase in RVO. This analysis was designed to address thatquestion.

Methods

The RISE and RIDE trials (registered on ClinicalTrials.gov asNCT00473330 and NCT00473382, respectively) were parallel,phase 3, multicenter, double-masked, randomized studies identi-cally designed to assess the efcacy and safety of ranibizumab inpatients with DME.26 The trials adhered to the tenets of theDeclaration of Helsinki and the Health Insurance Portability andAccountability Act. The protocol was approved by institutionalreview boards or ethics committees, and patients provided writteninformed consent. The design of the studies has been describedpreviously.11 Eligible participants (target, approximately 366 perstudy) were 18 years of age or older with diabetes mellitus (type1 or 2), decreased vision from DME (study eye best-correctedvisual acuity, 20/40e20/320 Snellen equivalent), and macularedema (time-domain optical coherence tomography central subeldthickness, 275 mm). Patients were randomized to monthly intra-vitreal injections of ranibizumab 0.3 or 0.5 mg or to sham in-jections. Study visits were scheduled every 307 days and included

ranibizumab treatment groups. All but one of the baseline patientcharacteristics was well balanced between the various treatmentgroups with 0 RNP and those with more than 0 RNP, with theone exception being history of smoking (Table 2, available atwww.aojournal.org). In patients with more than 0 RNP, historyof smoking was signicantly more prevalent for the 0.3-mg rani-bizumab group versus the sham group. We do not believe that thisimbalance had an impact on the reported comparisons because the

Table 1. Distribution of Study Eye Retinal Nonperfusion Area at Baseline: Pooled Data from RIDE and RISE

.3 mg (n [ 213) Ranibizumab 0.5 mg (n [ 225) Total (n [ 666)

(71.8) 167 (74.2) 488 (73.3)(23.0) 42 (18.7) 139 (20.9)(1.4) 8 (3.6) 18 (2.7)(2.8) 3 (1.3) 14 (2.1)(0.5) 3 (1.3) 4 (0.6)(0) 0 (0) 0 (0)(0) 1 (0.4) 1 (0.1)(0) 1 (0.4) 1 (0.1)

7e0e

Figure 2. The distribution of study eye retinal nonperfusion (RNP) area.Bars show the percentage of patients with no RNP, more than 0 to less than1 disc areas (DA) of RNP, 1 to less than 2 DA of RNP, or 2 DA or more ofRNP at several time points in the study in patients randomized to (A)monthly sham therapy (with crossover to ranibizumab 0.5 mg after 24months), (B) monthly ranibizumab 0.3 mg, or (C) monthly ranibizumab0.5 mg. *Percentages may not add to 100 owing to rounding. BL baseline.

Ophthalmology Volume -, Number -, Month 2014

4patients who showed no progression of RNP throughout the studyversus those in patients who showed progression of RNP is shownin Table 4 (available at www.aaojournal.org). The only signicantdifference was that patients who had progression of RNP wereyounger by a mean of 3.6 years, and there is no reason tobelieve that this would have any inuence on progression ofRNP. There was a trend for a greater percentage of patients whoshowed progression of RNP to have a glycosylated hemoglobinof more than 8%.

The major shift in the sham group occurred in the no posteriorRNP group and the more than 0 to less than 1 DA group, with theformer decreasing from 74% at baseline to 57% at month 24 andthe latter increasing from 21% at baseline to 35% at month 24 (Fig2A). Injections of ranibizumab 0.3 or 0.5 mg forestalled, but didnot completely prevent, this shift between the no posterior RNPand the more than 0 to less than 1 DA groups, which becamemanifest beyond month 24 (Fig 2B and C).

The percentage of patients showing an increase in posteriorRNP from baseline increased over time in all 3 groups, but at afaster rate in the sham group between months 3 and 24. Thisresulted in statistically signicant differences at every time point

Figure 3. The percent of patients with worsening study eye posteriorretinal nonperfusion. The symbols represent the percentage of patients withworsening posterior retinal nonperfusion in the study eye compared with(A) any baseline value and (B) no posterior retinal nonperfusion atbaseline in the sham group and the 2 ranibizumab (RBZ) treatment groupsat several time points throughout the study. Statistical comparisons weremade by the chi-square test for contingency tables. *P

al csurecauseedb

Campochiaro et al VEGF Promotes Retinal Nonperfusion in DMEbetween months 3 and 24 for sham versus ranibizumab 0.5 mg andbetween months 6 and 24 for sham versus ranibizumab 0.3 mg(Fig 3A). Crossover to treatment with ranibizumab 0.5 mg in thesham group starting after month 24 was followed by a reductionin the percentage of patients with an increase in posterior RNPfrom baseline at months 30 and 36, whereas the ranibizumabgroups continued their gradual rise. The curves were very similarfor the percentage of patients that increased from no posteriorRNP at baseline and showed the same differences between thesham and ranibizumab groups (Fig 3B).Figure 4. Schematic representation of the events leading to closure of retinlevels in the retina, which causes damage to retinal endothelial cells and cloreleases high levels of vascular endothelial growth factor (VEGF), whichfeedback loop. Ranibizumab neutralizes VEGF and interrupts the positive fDiscussion

In patients with DME, there is an increase in posterior RNPover time that is blunted signicantly by the neutralization ofVEGF. As soon as the control group was treated with intra-ocular injections of ranibizumab 0.5 mg beginning after themonth 24 visit, the progression of RNP was halted andreversed, providing additional evidence implicating VEGF.These results are very similar to those previously reported inpatients with CRVO and BRVO. In all 3 patient groups, therewas macular edema and RNP that worsened over time.Neutralization of VEGF improved macular edema andreversed the worsening in RNP. Therefore, these data supportthe hypothesis that macular edema and worsening of RNP areboth consequences of high intraocular levels of VEGF. Thissuggests that regardless of the mechanism by which VEGFbecomes elevated, as soon as high levels of VEGF are present,further disease progression occurs and a positive feedbackloop is initiated that accelerates disease progression. Statedanother way, retinal ischemia leads to increased intraocularlevels of VEGF, which in turn cause worsening of retinalischemia (Fig 4). Ranibizumab interrupts the VEGF positivefeedback loop, slowing or halting disease progression.Because elevated levels of VEGF occur in diseases inwhich retinal vessel closure occurs, the underlying diseaseprocesses have made it difcult to discern the role of VEGFin the vessel-closing activity. For instance, although it hasbeen recognized that worsening of ischemia often occurs inpatients with CRVO or BRVO, these ischemic changesoften are attributed to worsening of the underlying occlu-sion. This assumption is made even though there is noanatomic evidence that any occlusion of the central retinalvein or a branch retinal vein is less than total, and no evi-

apillaries in patients with diabetes. Hyperglycemia causes elevated glucoseof some vessels. Vessel closure causes retinal ischemia and ischemic retinaes worsening of retinal nonperfusion (RNP) and hence a VEGF-positiveack loop.dence that worsening of occlusion ever occurs. In theabsence of an understanding of the cause of worseningischemia, clinicians attempt to explain its occurrence as aconversion between 2 distinct entities, nonischemic andischemic RVOs. Measurements of RNP have demonstratedthat in the acute phase after an RVO, there is a broadspectrum of RNP and progression of RNP is the rule, ratherthan the exception. However, RNP progression often goesundetected unless it is at the severe end of the spectrumwhen the clinical picture may deteriorate rapidly. At thatpoint, the clinical change becomes obvious and appears tobe abrupt. Hence, it is described as a conversion fromnonischemic to ischemic RVO, although the worsening ofRNP has been occurring all along.

The masking of VEGF vaso-occlusive activity is evengreater in diabetic retinopathy in which diabetes itself isknown to damage retinal vessels. In fact, although our datademonstrate VEGF-induced closure of retinal vessels, thisclosure occurs on a background of gradual vaso-occlusion thatis not responsive to neutralization of VEGF by monthly in-jections of ranibizumab and presumably is the result of highglucose-induced damage to retinal vessels (Fig 4). Vascularendothelial growth factor may contribute to other aspects ofdiabetic retinopathy in addition to RNP. Grading of fundus

5

photographs from the RISE and RIDE studies for severity ofretinopathy based on the Diabetic Retinopathy Severity

30

3. Engerman R, Bloodworth JM Jr, Nelson S. Relationship ofmicrovascular disease in diabetes to metabolic control. Dia-

Ophthalmology Volume -, Number -, Month 2014Scale showed that eyes treated with ranibizumab weresignicantly less likely to worsen by 2 steps or more or 3steps or more and were signicantly more likely to improveby 2 steps or more or 3 steps or more compared with sham-treated eyes.31 Thus, high levels of VEGF may exacerbatemany of the lesions that constitute diabetic retinopathy andhence may accelerate the disease process.

The mechanism by which VEGF promotes closure ofretinal vessels in patients with DME is uncertain, but VEGFis a known monocyte chemoattractant, and repeated intra-ocular injections of VEGF in primates cause closure ofretinal vessels by leukostasis.32 Therefore, it is reasonable tohypothesize that leukostasis is involved in the VEGF-mediated progression of RNP in patients with DME. Italso is possible that VEGF-induced leukostasis participatesin the progression of retinopathy reected by increases inthe ETDRS diabetic retinopathy severity score, becausepatients with DME treated with intraocular steroids (whichalso should suppress leukostasis) are more likely to show a2-step or more or 3-step or more improvement than appro-priate control groups.33

A strength of this study is that it is based on data obtainedby masked graders in an independent reading center as part of2 closely monitored large, multicenter, sham injection-controlled trials. Another strength is that the results arevery similar to those seen in 2 large trials in RVO, anotherdisease process with high intraocular levels of VEGF, addingto the weight of the ndings. A potential limitation of thestudy is that RNP was measured in ETDRS elds 1 through3, which constitutes the posterior pole of the eye. It isreasonable to assume that observations made in this sampleof posterior retina reect what is occurring throughout theremainder of the retina, but this is not known with certainty.

In addition to providing new insight regarding the exac-erbating effect of VEGF in progression of RNP, the currentstudy has potential clinical implications. Monthly injectionsof ranibizumab suppress macular edema, progression ofRNP, and progression of retinopathy in diabetic patients. Thissuggests that sustained neutralization of VEGF may remedyseveral aspects of the disease and minimize worsening ofdiabetic retinopathy. Therefore, sustained delivery of a VEGFantagonist should be explored as a comprehensive approachto treatment of diabetic retinopathy.

References

1. Klein R, Klein BE. Vision disorders in diabetes. In: Harris MI,Cowie CC, Stern MP, et al., eds. Diabetes in America. 2nd ed.Bethesda, MD: National Institute of Diabetes and Digestive andKidney Disease, National Institutes of Health; 1995:293338.NIH publication no. 95-1468. Available at: http://diabetes.niddk.nih.gov/dm/pubs/america/contents.aspx. Accessed March4, 2014.

2. Klein R, Knudtson MD, Lee KE, et al. The WisconsinEpidemiologic Study of Diabetic Retinopathy XXIII: thetwenty-ve-year incidence of macular edema in persons withtype 1 diabetes. Ophthalmology 2009;116:497503.

6betes 1977;26:7609.4. Diabetes Control and Complications Trial Research Group.

Progression of retinopathy with intensive versus conventionaltreatment in the Diabetes Control and Complications Trial.Ophthalmology 1995;102:64761.

5. Robison WG Jr, McCaleb ML, Feld LG, et al. Degeneratedintramural pericytes (ghost cell) in the retinal capillaries ofdiabetic rats. Curr Eye Res 1991;10:33950.

6. Nguyen QD, Shah SM, Van Anden E, et al. Supplementaloxygen improves diabetic macular edema: a pilot study. InvestOphthalmol Vis Sci 2004;45:61724.

7. Campochiaro PA. C99-PKC412-003 Study Group. Reductionof diabetic macular edema by oral administration of the kinaseinhibitor PKC412. Invest Ophthalmol Vis Sci 2004;45:92231.

8. Nguyen QD, Tatlipinar S, Shah SM, et al. Vascular endothelialgrowth factor is a critical stimulus for diabetic macular edema.Am J Ophthalmol 2006;142:9619.

9. Diabetic Retinopathy Clinical Research Network, Elman MJ,Aiello LP, Beck RW, et al. Randomized trial evaluatingranibizumab plus prompt or deferred laser or triamcinoloneplus prompt laser for diabetic macular edema. Ophthalmology2010;117:106477.

10. Mitchell P, Bandello F, Schmidt-Erfurth U, et al; RESTOREStudy Group. The RESTORE study: ranibizumab mono-therapy or combined with laser versus laser monotherapy fordiabetic macular edema. Ophthalmology 2011;118:61525.

11. Nguyen QD, Brown DM, Marcus DM, et al; RISE and RIDEResearch Group. Ranibizumab for diabetic macular edema:results from 2 phase III randomized trials: RISE and RIDE.Ophthalmology 2012;119:789801.

12. Brown DM, Nguyen QD, Marcus DM, et al; RIDE and RISEResearch Group. Long-term outcomes of ranibizumab therapyfor diabetic macular edema: the 36-month results from twophase III trials: RISE and RIDE. Ophthalmology 2013;120:201322.

13. Campochiaro PA, Haz G, Shah SM, et al. Ranibizumab formacular edema due to retinal vein occlusions: implication ofVEGF as a critical stimulator. Mol Ther 2008;16:7919.

14. Campochiaro PA, Heier JS, Feiner L, et al; BRAVO In-vestigators. Ranibizumab for macular edema following branchretinal vein occlusion: six-month primary endpoint results of aphase III study. Ophthalmology 2010;117:110212.

15. Brown DM, Campochiaro PA, Singh RP, et al; CRUISE In-vestigators. Ranibizumab for macular edema following centralretinal vein occlusion: six-month primary end point results of aphase III study. Ophthalmology 2010;117:112433.

16. Campochiaro PA, Brown DM, Awh CC, et al. Sustainedbenets from ranibizumab for macular edema following cen-tral retinal vein occlusion: twelve-month outcomes of a phaseIII study. Ophthalmology 2011;118:20419.

17. Brown DM, Campochiaro PA, Bhisitkul RB, et al. Sustainedbenets from ranibizumab for macular edema followingbranch retinal vein occlusion: 12-month outcomes of a phaseIII study. Ophthalmology 2011;118:1594602.

18. Sabet-Peyman EJ, Heussen FM, Thorne JE, et al. Progressionof macular ischemia following intravitreal bevacizumab.Ophthalmic Surg Lasers Imaging 2009;40:3168.

19. Mansour AM, Bynoe LA, Welch JC, et al. Retinal vascularevents after intravitreal bevacizumab. Acta Ophthalmol2010;88:7305.

20. Shimura M, Yasuda K. Macular ischaemia after intravitrealbevacizumab injection in patients with central retinal vein