Factors Associated with Proliferative Diabetic Retinopathy Worsening

Diabetic Retinopathy Clinical Research Network Factors Associated with Worsening Proliferative Diabetic Retinopathy in Eyes Treated with Panretinal Photocoagulation or Ranibizumab (Protocol S) Susan B. Bressler, M.D. 1

Background: PRP vs. Anti-VEGF for PDR as reported in primary paper* When managing PDR, anti-VEGF injections (ranibizumab) resulted in visual acuity at 2-years that was no worse than (non-inferior to) PRP (+2.2 letters favoring ranibizumab 95% CI: -0.5, +5.0) Vision during 2-year course was superior with ranibizumab (area under the curve) Eyes randomly assigned to ranibizumab were less likely to undergo vitrectomy Numerically lower rates of vitreous hemorrhage and retinal detachment with ranibizumab compared to PRP (not statistically significant) *Protocol S Primary Manuscript. JAMA. 2015;314(20)2137-2146 5-letter non-inferiority margin 2

Prevalence of Vitrectomy and PDR-Worsening Events at 2 Years as reported in primary paper Ranibizumab Group N=191 4% PRP Group N=203 15% P-value < 0.001 Vitrectomy Vitreous hemorrhage Any retinal detachment Neovascular glaucoma Neovascularization of the iris 27% 34% 0.09 6% 10% 0.08 2% 3% 0.50 1% 1% 0.96 3

Objectives 1. Compare the rates and timing of PDR-worsening events individually and as a composite outcome (VH, RD, NVI/NVA, NVG) by treatment group 2. Explore differences in severity of these events between treatment groups 3. Identify predictive factors for PDR-worsening events 4

Objective #1: Compare Rates and Timing of PDR-Worsening Events Composite Outcome: First occurrence of any one of the following: vitreous hemorrhage, any retinal detachment, NVI/NVA or NVG 5

Composite Outcome* 34% N = 191 *First occurrence of VH, RD, or NVI/NVG 6

Composite Outcome* Ranibizumab vs. PRP P=0.063 42% N = 203 N = 191 34% *First occurrence of VH, RD, or NVI/NVG 7

Composite Outcome*: Eyes Without Baseline CI-DME + Vision Loss N = 155 31% *First occurrence of VH, RD, or NVI/NVG 8

Composite Outcome*: Eyes Without Baseline CI-DME + Vision Loss Ranibizumab vs. PRP P=0.008 45% N = 147 N = 155 31% *First occurrence of VH, RD, or NVI/NVG 9

Objective #3: Identify Predictive Factors for PDR-Worsening Events 10

19 Baseline Characteristics Evaluated Gender Age Race/Ethnicity Diabetes type Diabetes duration HbA1c Hypertension Best-corrected visual acuity Central subfield thickness Vision-Impairing CI-DME Lens status (phakic vs. pseudophakic) Vitreous hemorrhage on clinical exam Neovascularization on clinical exam NVD or NVE only vs. NVD+NVE *Diabetic retinopathy severity on fundus photographs (ETDRS) *Vitreomacular traction *Epiretinal membrane ^Laser delivery type (single- spot vs. pattern scan) ^Total number of PRP spots (controlling for laser type) ^Number of PRP sittings performed (1 vs. 2 or 3) *Graded by reading center ^PRP group only, subject to inv discretion Yellow = Subject-level factor White = Eye-level factor 11

Evaluating Predictive Factors Step 1: Evaluated whether each factor had similar effects between groups (testing for an interaction) Ex) Did the treatment effect comparing the ranibizumab and PRP groups differ according to gender? There was no strong statistical evidence to suggest that individual factors behaved differently by treatment groups Therefore, the ranibizumab and PRP groups were combined for evaluation of predictive factors Analyses controlled for the effect of treatment 12

Evaluating Predictive Factors Step 2: Each predictive factor was evaluated by itself (univariate analysis) Step 3: Predictive factors with P < 0.10 were included simultaneously in a model selection process to determine which factors may be the most important (multivariate analysis) 13

Predictive Factors for Composite Outcome* Univariate Analyses DR Severity Level NVD+NVE Vitreous Hemorrhage Epiretinal Membrane Age Race/Ethnicity PRP Laser Delivery Type Gender Visual Acuity P < 0.01 0.01 < P < 0.05 0.05 < P < 0.10 *VH, RD, or NVI/NVG PRP group only, subject to investigator discretion 14

Predictive Factors for Composite Outcome* Multivariate Analysis No. Eyes Group DR Severity Level Moderate PDR (level 65) or better High risk PDR (71) or worse Ran PRP GroupP-valueHazard Ratio < 0.001 Variable (99% CI) 242 15% 25% - 3.97 146 57% 58% (2.48, 6.36) *VH, RD, or NVI/NVG P-value for ranibizumab-PRP comparison = 0.024 when adjusting for DRSL (compared with P = 0.063 unadjusted) 15

Predictive Factors for Composite Outcome* Multivariate Analysis No. Eyes Group DR Severity Level Moderate PDR (level 65) or better High risk PDR (71) or worse PRP Laser Delivery Single-Spot 164 Ran PRP GroupP-valueHazard Ratio < 0.001 Variable (99% CI) 242 15% 25% - 3.97 146 57% 58% (2.48, 6.36) 0.008 - 34% - 2.04 Pattern scan 39 - 49% (1.02, 4.08) PRPgroup only, subject to investigator discretion 16

Summary In eyes with PDR, rates of PDR-worsening were lower with ranibizumab compared to PRP Especially among eyes not required to receive ranibizumab at baseline for vision-impairing CI-DME and when adjusting for baseline DR severity Clinical application of these findings: While anti-VEGF requires compliance to a more frequent visit schedule than PRP, these findings provide additional evidence supporting the use of ranibizumab as an alternative therapy to PRP for PDR, at least through 2 years of follow-up 17

Summary The ETDRS Diabetic Retinopathy Severity Scale remains an important predictor for worsening of diabetic retinopathy Well-known for decades, but still true in this era of anti-VEGF therapy and modern PRP May support intervention prior to high-risk PDR characteristics Pattern scan PRP may not be as effective as conventional single-spot PRP in limiting PDR-worsening Caution: Laser type not assigned randomly, although 68% of investigators that used pattern scan used it exclusively 18