The relationship between duration of severe DME (ETDRS DME level ≥ 4) and VA was assessed by analyzing the distribution of baseline-to-endpoint change in VA by duration of severe DME during the study among eyes with baseline VA ≥ 70 letters. The baseline distributions of ocular factors that might influence this relationship were similar in the placebo and RBX groups
(Table 1) . In placebo-treated patients
(Fig. 2A) , there was progressively greater decline in VA with increasing duration of severe DME. There was no decrease in median VA in the 304 eyes that never exhibited severe DME at any visit during follow-up in the trial (8 of these eyes had severe DME at baseline but not at any follow-up visit, as did 14 of the corresponding eyes in the RBX group). In the 161 eyes in the placebo group that had severe DME at one or more follow-up visits, the decline in VA grew progressively greater with longer duration of DME of this severity (−0.63 letters per month,
P for trend = 0.0002). Among RBX-treated eyes
(Fig. 2B)there were 124 that had severe DME at one or more follow-up visits and in them there appears to have been less decline in VA for a given duration of severe DME (−0.30 letters per month,
P for trend = 0.0332;
P = 0.010 for difference in slopes between
Figs. 2A 2B ). In similar analyses in eyes with baseline VA < 70 letters, no relationship between duration of severe DME and VA loss was apparent in either treatment group (data not shown).
The results shown in
Figure 2and just described were not adjusted for baseline factors that might be predictive of baseline-to-endpoint change in VA. In a regression analysis of this outcome that included such factors, as well as duration of severe DME and treatment, there was a strong interaction (
P = 0.0019) between treatment and duration of severe DME. Accordingly, similar analyses, which included baseline levels for ETDRS DR severity (≥47B vs. <47B), ETDRS DME severity (≥4 vs. <4), VA (letters), and history of prior photocoagulation for DME, in addition to duration of severe DME, were performed separately in each treatment group. In the placebo group, the estimate for the effect of duration of severe DME was 0.67 letters lost per month of severe DME (
P < 0.0001), corresponding to 24 letters over the 36-month duration of the trial (74% of patients completed at least 36 months of follow-up). From baseline to endpoint, on average, eyes with prior photocoagulation for DME lost 2.1 fewer letters than those without (
P = 0.073); eyes in ETDRS DR severity level ≥ 47B lost 1.7 more letters than those in level ≤ 47A (
P = 0.13), and for every 5-letter decrease in baseline VA, eyes lost 0.8 fewer letters (
P = 0.093).
In the RBX group, the estimate for the effect of duration of severe DME was 0.47 letters lost per month of severe DME (P < 0.0001), corresponding to 17 letters over the 36-month duration of the trial. P for difference between the placebo and RBX groups in the effect of duration was 0.022. In this analysis, prior photocoagulation was not prognostic, whereas results for DR severity and baseline VA were similar to those in the placebo group analysis: eyes in level ≥ 47B lost 2.0 more letters (P = 0.024) and for every 5-letter decrease in baseline VA, eyes lost 1.1 fewer letters (P = 0.0039). In the RBX analysis, unlike the analysis in the placebo group, baseline DME severity level appeared to be a risk factor; eyes in DME level ≥ 4 at baseline lost 5.7 fewer letters than those in level < 4 (P = 0.0002).