Factors Associated With Changes in Visual Acuity and Central Subfield Thickness at 1 Year After Treatment for Diabetic Macular Edema With Ranibizumab

Objective To identify factors that predict the success or failure of treatment with intravitreal ranibizumab for patients with diabetic macular edema.

Methods A total of 37 baseline demographic, systemic, ocular, optical coherence tomographic, and fundus photographic variables were assessed for association with change in visual acuity or central subfield thickness between baseline and 1 year in 361 eyes that were randomly assigned to intravitreal ranibizumab with prompt or deferred laser treatment within a trial of ranibizumab, triamcinolone acetonide, and laser treatment for center-involved diabetic macular edema. A categorical variable describing follow-up anatomic responses to therapy was added to the visual acuity outcome model.

Results After adjusting for baseline visual acuity, a larger visual acuity treatment benefit was associated with younger age (P < .001), less severe diabetic retinopathy on clinical examination (P = .003), and absence of surface wrinkling retinopathy (P < .001). The reduction in central subfield thickness during the first treatment year also predicted better visual acuity outcomes (P < .001). After adjusting for baseline central subfield thickness, the presence of hard exudates was associated with more favorable improvement on optical coherence tomographic scan (P = .004). Because only 11 eyes experienced vision loss and 6 eyes experienced an increase in central subfield thickness, factors for poor outcomes could not be evaluated.

Conclusions A review of baseline factors and anatomic responses during the first year of ranibizumab therapy for association with visual acuity outcome did not identify any features that would preclude ranibizumab treatment. However, baseline central subfield thickness is the strongest predictor of anatomic outcome, and reduction in central subfield thickness during the first treatment year is associated with better visual acuity outcomes.

Diabetic macular edema (DME) has been the leading cause of moderate vision loss in people with diabetes mellitus.¹ The prevalence of diabetes is expected to increase as the prevalence of obesity continues to increase markedly; therefore, improving treatment of DME will become increasingly important.² Based on findings from the Early Treatment Diabetic Retinopathy Study (ETDRS), focal/grid laser photocoagulation had been the mainstay of treatment for DME since 1985.³ However, recent studies^3-6 suggesting a role for vascular endothelial growth factors (VEGFs) in the pathogenesis of DME have prompted evaluation of anti-VEGF drugs, such as ranibizumab (Lucentis; Genentech), bevacizumab (Avastin; Genentech), and aflibercept (Eylea; Regeneron Pharmaceuticals, Inc, and Bayer Healthcare Pharmaceuticals) for the treatment of DME.

A multicenter-randomized Diabetic Retinopathy Clinical Research Network (DRCR.net) trial found that intravitreal ranibizumab, either with prompt or deferred focal/grid laser treatment, resulted in superior visual acuity and central retinal thickness outcomes compared with focal/grid laser treatment alone or triamcinolone acetonide with laser treatment at both 1 year and 2 years of follow-up.^7,8 The 2 ranibizumab groups, with a median baseline best-corrected visual acuity of 20/50 (approximate Snellen equivalent), had nearly identical outcomes at 1 year. Both groups averaged nearly 2 lines of visual acuity improvement, almost 30% of study participants improved 3 or more lines of visual acuity, and fewer than 5% of study participants lost 2 or more visual acuity lines. Findings from this clinical trial support the use of ranibizumab for the management of center-involved DME with vision impairment. Identifying risk factors that predict the success or failure of treatment could help investigators to make informed decisions as to which patients should be treated with intravitreal ranibizumab. Therefore, we are presenting additional analyses of 361 eyes of study participants who were randomly assigned to either 0.5 mg of ranibizumab and prompt laser treatment (n = 180) or 0.5 mg of ranibizumab and deferred laser treatment (≥24 weeks) (n = 181).

All data used for this analysis were collected from baseline to 1 year by clinical sites.⁷ The protocol for this trial is available at the DRCR.net website (http://www.drcr.net).

Major eligibility criteria for participation in this trial included (1) a best-corrected Electronic ETDRS (E-ETDRS) visual acuity (using the E-ETDRS Visual Acuity Test) letter score of 78 to 24 (approximate Snellen equivalent, 20/32 to 20/320), (2) definite retinal thickening from DME involving the foveal center on clinical examination as the cause of vision loss, and (3) confirmation of foveal edema with a central subfield thickness of 250 μm or greater by use of time-domain optical coherence tomography (OCT).

At baseline and at each follow-up visit, the best-corrected visual acuity letter score was measured at 3 m by a certified masked examiner using the E-ETDRS Visual Acuity Test, and OCT images were obtained by a certified operator using the Zeiss Stratus OCT machine (Carl Zeiss Meditec, Inc). Additional testing at baseline included slitlamp examination, fundus examination, and standard ETDRS 7-field color stereoscopic fundus photographs. Fundus photographs were graded at the Fundus Photograph Reading Center (University of Wisconsin, Madison).

During the first year, follow-up visits occurred every 4 weeks (±1 week), with the potential to receive ranibizumab at each visit. Eyes assigned to either ranibizumab arm were required to have 4 consecutive monthly ranibizumab injections. Treatment was at the investigator's discretion at any visit that fulfilled the “success” criteria (defined as a Snellen equivalent of 20/20 or a central subfield thickness of <250 μm) beginning with the 16-week study visit. Two additional ranibizumab injections were required through week 20 if success had not been met. Beginning at week 24, treatment was at the investigator's discretion if the participant's treatment met the criteria for “failure”; however, treatment could be deferred for “futility” criteria beginning at week 52. Unless failure or futility criteria were met, investigators were encouraged to continue intravitreal injections when edema persisted or recurred.

Study participants assigned to the ranibizumab plus prompt laser group received focal/grid laser treatment within 3 to 10 days of their baseline ranibizumab injection. Additional laser treatment was administered as often as every 13 weeks if persistent DME involved or threatened the fovea and if complete laser treatment had not been previously administered. Study participants assigned to the ranibizumab plus deferred laser group first became eligible for laser treatment anytime at week 24 or beyond if DME persisted or threatened the fovea and if the “futility” criteria were met.

Additional details of the treatment protocol can be found in the article by the DRCR.net that summarizes the primary outcome of the trial.⁷ In particular, in that article,⁷ Appendices 1 to 3 depict flowcharts of patient management, and their Table 1 provides important definitions used to guide management.

A total of 361 eyes of study participants randomly assigned at baseline to either of the 2 ranibizumab groups within this trial were eligible for these analyses. Thirty-seven baseline demographic, systemic, ocular, OCT, and fundus photographic factors were assessed for association with changes in visual acuity or central subfield thickness between baseline and 1 year (eTable 1). Change in visual acuity (using “letter score” as a continuous variable) was assessed as the primary dependent variable. The association between baseline visual acuity and the proportion of participants reaching 74 (approximate Snellen equivalent, 20/32) or better visual acuity, as well as the percentage reduction in visual acuity deficit at 1 year, was also explored. The baseline visual acuity deficit was defined as 84 minus the baseline visual acuity letter score; a letter score of 84 (approximate Snellen equivalent, 20/20) was considered “normal” visual acuity. The percentage reduction in visual acuity deficit at 1 year was defined as the change in visual acuity letter score from baseline to 1 year divided by the baseline visual acuity deficit and multiplied by 100%. We were unable to evaluate factors associated with poor visual acuity outcomes because the numbers of individuals in the ranibizumab treatment groups with vision loss were too small to analyze. In the ranibizumab-assigned treatment arms 1 year following treatment initiation, only 5 eyes (1%) lost 10 to 14 letters of visual acuity, and 6 eyes (2%) lost at least 15 letters.

Associations with anatomic change were also investigated using change in central subfield thickness (detected on OCT images and measured in microns) as the dependent variable. Other dependent variables included a central subfield thickness of less than 250 μm and a percentage reduction in excessive retinal thickness. The percentage reduction in excessive retinal thickness was calculated in a similar fashion to the percentage reduction in visual acuity deficit. Baseline excessive retinal thickness was defined as the baseline central subfield thickness measurement minus 201 μm (the mean value of a cohort of persons with diabetes with no or mild retinopathy and no clinical evidence of DME).⁹ The percentage reduction in excessive retinal thickness was defined as the change in central subfield thickness from baseline to 1 year divided by the baseline excessive retinal thickness and multiplied by 100%. We were unable to evaluate factors associated with poor central subfield thickness outcomes because the numbers of individuals in the ranibizumab treatment groups with an increase in central subfield thickness were too small to analyze. Central subfield thickness increased by 20% (1-step worsening of log OCT) in only 6 eyes receiving ranibizumab between study entry and the 1-year examination.

A composite outcome of both favorable functional and anatomic end points was also evaluated because the simultaneous presence of a reduction in retinal thickness at the time of vision improvement may provide a greater level of confidence that the observed vision changes are attributed to the biologic effects of the drug. The composite outcome was defined as at least a 10-letter improvement and at least a 20% reduction in central subfield thickness (the equivalent of a 1-step reduction of log OCT).¹⁰

Linear regression (visual acuity and central subfield thickness outcomes) and Poisson regression (composite outcome) models with robust variance estimation were used to assess potential risk factors for each of the 3 outcome variables specified.¹¹ Vision analyses were adjusted for baseline visual acuity, anatomic analyses were adjusted for baseline central subfield thickness, and the composite outcome was adjusted for both. Potential risk factors with P < .10 in the univariate models were included in multivariate models. Final multivariate models consist of factors with P < .01 following a backward selection process, and each eliminated covariate was tested by adding it to the final model, one at a time, to confirm P ≥ .01. When continuous or ordinal values were available for baseline features, these were used in the variable selection process; categories are shown for these variables for ease of interpretation, to report estimates and relative risks (and 95% CIs). In the multivariate models, missing values of covariates were handled by adding a separate missing category for discrete covariates and adding a missing indicator variable for continuous covariates.

A second visual acuity outcome model was constructed that included a categorical variable created to separate groups by their OCT behavior, or thickness evolution, at follow-up visits, in response to treatment during the first treatment year. All study eyes were differentiated into 1 of 4 separate categories based on whether they had at least a 20% reduction from baseline central subfield thickness at the 16-week, 32-week, and 1-year study visits. The “early and consistent” group of anatomic responders experienced at least a 20% reduction in central subfield thickness by the 16-week study visit and sustained at least this much improvement at the 32-week and 1-year study visits. The “early but inconsistent” group of anatomic responders experienced this threshold level of central subfield thickness reduction by the 16-week study visit but failed to sustain this anatomic improvement consistently at the 32-week or 1-year study visits, or both. The “slow and variable” group of anatomic responders did not manifest a 20% or greater reduction in central subfield thickness at the 16-week study visit but did so at either the 32-week or 1-year study visit, or both. Finally, the “nonresponder” group did not manifest this threshold level of central subfield thickness at the 16-week, 32-week, or 1-year study visits. Pairwise comparisons among subgroups regarding the evolution pattern of macular thickness were performed using linear contrasts. The number of injections received during year 1 was included as an additional adjustment covariate (continuous and categorical in separate models) to confirm that the relationship seen was not due to differences in the number of injections among the subgroups.

Univariate analyses were performed initially on each of the 2 groups (ranibizumab and prompt or deferred laser treatment). Results were consistent between groups; therefore, we have only reported the combined analyses. All factors, with one exception (prior DME treatment), showing evidence of strong association (P < .001) in one of the treatment groups also showed evidence of association (P < .10) in the other treatment group. The observation that 1 factor was associated with visual acuity outcome in the deferred laser treatment group but was not associated with visual acuity outcome in the prompt laser treatment group may be due to chance in view of the large number of factors evaluated.

Data to assess baseline predictors for the visual acuity and OCT outcomes were available for 338 eyes (94%) and 334 eyes (93%), respectively. A complete set of central subfield thickness measurements at baseline, 16 weeks, 32 weeks, and 1 year were available for 288 study eyes (80%) to explore the effect of anatomic change during treatment on the 1-year visual acuity outcome.

Results for analyses limited to participant demographics (3 variables), systemic conditions (9 variables), and medication use (3 variables) are shown in Table 1. Results for analyses limited to prior ocular history (8 variables) are shown in Table 2. Results for analyses limited to ocular examination characteristics at study entry (5 variables) are shown in Table 3. Results for analyses limited to ocular characteristics obtained from the initial visit OCT (5 variables) and fundus photographs (4 variables) are shown in Table 4.

The average visual acuity letter score at baseline was 62 (approximate Snellen equivalent, 20/63). Eyes that started with lower visual acuity scores (poorer levels of acuity) were more likely to realize larger gains in visual acuity over time (Table 3) but less likely to reach near-normal visual acuity (letter score ≥74; approximate Snellen equivalent, 20/32 or better; eTable 2). The overall median percentage reduction in visual acuity deficit between baseline and 1 year was 52% and did not significantly differ across the range of baseline visual acuities (P = .92). For example, the median percentage reduction in the deficit was 55% among those with a baseline visual acuity Snellen equivalent of 20/80 or better compared with 36% for those with a Snellen equivalent visual acuity of worse than 20/80 (P = .64).

After adjusting for baseline visual acuity, we found that 3 factors were associated with a larger magnitude of treatment benefit on visual acuity at 1 year: younger age (P < .001; Table 1), less severe diabetic retinopathy on clinical examination (P = .003; Table 3), and absence of surface wrinkling retinopathy on fundus photographs (P < .001; Table 4). There was a mean increase of 2.2 letters (95% CI, 1.1-3.3 letters) in visual acuity gains at 1 year for every 10 years of younger age across participants. Hence, a person aged 50 years would be predicted to have an additional 4 letters of vision improvement compared with an individual aged 70 years. Compared with eyes having proliferative diabetic retinopathy or prior panretinal photocoagulation, eyes with severe nonproliferative diabetic retinopathy or moderate nonproliferative diabetic retinopathy or less were estimated to have a mean improvement of 4 more letters (95% CI, 1-7 letters and 2-7 letters, respectively). Eyes that showed no evidence of surface wrinkling retinopathy were estimated to have an additional mean improvement of 4 letters (95% CI, 1-7 letters) compared with eyes that had either questionable or definite wrinkling on fundus photographs.

At baseline, the mean central subfield thickness of eyes treated with ranibizumab was 406 μm. Eyes that started with a greater central subfield thickness were more likely to realize greater reductions in thickness over time (Table 4) but less likely to reach normal or near-normal thickness (central subfield thickness of <250 μm; data not shown).

This relationship between greater reduction in central subfield thickness and greater baseline thickness was confirmed when the median percentage change in excessive retinal thickness was explored among groups. For example, the median percentage decrease in excessive retinal thickness was 46% for eyes with a baseline thickness of less than 300 μm and 82% for eyes with a baseline thickness of 500 μm or greater (P = .003).

After adjusting for baseline central subfield thickness, hard exudates within the 6-mm ETDRS grid on fundus photographs were found to be associated with change in central subfield thickness (P = .004; Table 4). Eyes with questionable or definite hard exudates present in the macular region at baseline were estimated to have a reduction in thickness of 31 μm or greater (95% CI, 8-53 μm) than eyes without any hard exudates.

The composite 1-year outcome of an improvement of 10 letters or more in visual acuity and a 20% or more reduction in central subfield thickness was observed in 37% of the overall cohort. After adjusting for baseline visual acuity and central subfield thickness, age and hard exudates within the 6-mm ETDRS grid on fundus photographs were found to be associated with this composite outcome. For every 10 years of younger age, there was a 27% increase in the relative probability of this composite outcome (relative risk, 1.27 [95% CI, 1.09-1.48]; P = .003). Eyes with questionable or definite hard exudates present in the macular region at baseline were twice as likely (43% vs 20%; relative risk, 2.01 [95% CI, 1.29-3.14]; P = .002) to have this favorable combined functional and anatomic outcome than eyes without any hard exudates.

The 4 patterns of the evolution of central subfield thickness during the first year of ranibizumab treatment showed differences in visual acuity outcome at 1 year (eTable 3). Early and consistent anatomic responders had a mean (SD) improvement of 13 (9) letters, early but inconsistent responders had a mean (SD) improvement of 9 (9) letters, slow and variable responders had a mean (SD) improvement of 7 (11) letters, and nonresponders had a mean (SD) improvement of 4 (9) letters (P < .001). These differences were not explained by differences in the number of intravitreal injections administered to these subgroups (P < .001 after adjustment for number of injections). The number of mean (SD) injections varied from 8 (3) to 10 (3) among these subgroups; the group with the most robust OCT evolution pattern had the least number of injections.

When the evolution pattern of central subfield thickness was added to the change in the visual acuity outcome regression model, eyes that demonstrated an early and consistent anatomic response had, on average, a vision improvement that was approximately 6 letters (95% CI, 4-9 letters) greater than that seen in eyes that did not manifest anatomic improvement at any visit and that was 4 letters (95% CI, 0.4-7 letters, approximately 1 line) greater than that seen in eyes that had a late (at 16-week and/or 1-year study visits) anatomic response (P < .001; Table 5). Age and surface wrinkling retinopathy also remained associated with change in visual acuity, similar to the model containing only baseline factors. Our Figure shows the effect of the evolution pattern of central subfield thickness on the median change in visual acuity between baseline and 1 year within subgroups with worse and better baseline visual acuity.

Several clinical trials^5-8,12-14 have recently demonstrated that intravitreal ranibizumab therapy, with either prompt or various degrees of deferred laser treatment, results in improved visual acuity outcomes at 1 and 2 years compared with focal/grid laser treatment alone. Identification of factors associated with relatively good or poor outcomes can help inform treating ophthalmologists and patients as to what they can expect, on average, when choosing intravitreal ranibizumab as a treatment for DME. We were unable to evaluate factors associated with poor visual acuity outcomes or poor central subfield thickness outcomes because the numbers of individuals in the ranibizumab treatment groups with vision loss or an increase in central subfield thickness were too small to analyze. Thus, our attention was focused on predictors of improvement in visual acuity, retinal thickness, or the simultaneous presence of both outcomes when treated with ranibizumab for center-involved DME.

A large number of baseline features (n = 37) were evaluated for their relationship with vision and anatomic improvement at 1 year (eTable 1). The selected variables covered a wide range of features: demographics, medical history, medication use, past ocular history with particular emphasis on past interventions for diabetic retinopathy, present ocular manifestations of diabetic eye disease, and morphologic features identified on retinal images as interpreted by a reading center. No features were identified that would predict groups for whom treatment appeared to be harmful at 1 year. Some features were identified that were associated with a relatively better outcome. These analyses suggest that intravitreal ranibizumab could be considered for all patients with center-involved DME. This additional information on relative risks and benefits could help patients and their physicians in their choice of treatment and their expectations.

Four factors were identified that were associated with a larger magnitude of increase in visual acuity at 1 year: visual acuity at the time of treatment initiation, younger age, less severe retinopathy level as assessed by the treating ophthalmologist, and absence of surface wrinkling retinopathy. The observed association between worse visual acuity at the time of treatment and an increased degree of visual acuity improvement may be at least partially affected by “ceiling effects” on the degree of improvement possible for those with better visual acuity at the time of treatment. Our secondary analysis, which substituted percentage reduction in visual acuity deficit at 1 year for change in visual acuity, represented an attempt to eliminate the ceiling effect of baseline visual acuity. In this analysis, the association between baseline visual acuity and visual acuity outcomes was not confirmed. The association with younger age and surface wrinkling retinopathy remained, which suggests that the ceiling effect may be largely responsible for the observed association between 1-year visual acuity outcome and baseline visual acuity.

It is unknown why younger age is associated with superior visual acuity outcomes. Younger age has also been found to be associated with superior visual acuity outcomes when treating patients who have neovascular, age-related macular degeneration with ranibizumab.¹⁵ The present analysis found a median difference of 3 letters for those younger than 60 years of age compared with those 60 years of age or older. However, at the extremes of the age distribution of our participants, the median difference was as large as 7 letters (ie, participants aged 47 years or younger [5th percentile] vs participants aged 78 or older [95th percentile]).

Eyes graded by the participating investigators as having nonproliferative diabetic retinopathy, rather than proliferative diabetic retinopathy or prior panretinal photocoagulation, had a greater magnitude of visual acuity improvement, but this was not confirmed in the analysis in which retinopathy level at baseline was assessed by the independent reading center. This association may have been a chance observation, although it is plausible that eyes with the most advanced grades of retinopathy may have more ischemia or permanent damage/scarring that would limit their potential for vision improvement.

With respect to superior visual acuity outcomes in the absence of surface wrinkling retinopathy, persons with DME who show evidence of surface wrinkling retinopathy might be anticipated to do less well than those with DME who had shown no evidence of surface wrinkling because the former could have an effect on the macular anatomy, which could retard functional and anatomic improvement. As part of the study inclusion criteria, investigators were asked to include only those eyes that had definite retinal thickening due to DME on clinical examination that involved the center of the macula; central subfield thickening due to DME is believed to be the main cause of visual acuity loss. Therefore, the enrollment of persons with eyes that have some degree of abnormal vitreoretinal interface anatomy was permitted as long as the investigator did not interpret that finding to have a major impact on visual function. Thus, the 17% of individuals in this trial with surface wrinkling retinopathy, identified as questionable or definite on color fundus photographs at a reading center, likely represent eyes in the mild range of the spectrum of abnormal vitreoretinal relationships. As noted in Table 4, there was a positive gain in median vision and a reduction in thickness when these eyes were treated with ranibizumab. However, the difference in outcome was roughly an average 4-letter increase in visual acuity for the group without surface wrinkling retinopathy compared with the group with surface wrinkling. In this cohort, this corresponds to approximately 15% fewer study participants experiencing at least 10- or 15-letter gains in visual acuity in the presence of surface wrinkling retinopathy when receiving treatment of DME with ranibizumab.

An additional variable associated with 1-year visual acuity outcomes was the evolution pattern of central subfield thickness during the first year of treatment. The finding that eyes that rapidly and consistently demonstrate a favorable anatomic response are more likely to have superior visual acuity outcomes would be anticipated. However, the treatment protocol instructed investigators to continue monthly injections until the criteria for “success” were met or until stabilization (successive treatments not yielding incremental improvements in vision or central subfield thickness). When investigators applied this protocol, a larger number of injections were administered to the groups of individuals with the less favorable OCT evolution patterns. This suggests that some eyes are simply less responsive (at the anatomic level) than others to the intervention. However, we were unable to identify, in advance, predictors of who would or would not have a rapid and consistent response to treatment, and even those subgroups that responded late or appeared to be refractory still had positive changes in both mean and median visual acuity relative to baseline. In other words, a less favorable OCT response pattern is not a reason to discontinue treatment, and the early nonresponders may be converted to later responders.

One baseline factor, other than the initial thickness value, that was associated with more favorable improvement in the central subfield at 1 year, was the presence of hard exudates within 6 mm of the foveal center. Hard exudates may be a marker for hyperpermeability and fluid turnover, as well as a marker of areas of retina with an intact “blood-retinal barrier.” These are the pathophysiologic characteristics that ranibizumab would be anticipated to most favorably affect. In addition, eyes without hard exudates might have additional underlying mechanisms for central retinal thickening, including ischemia, traction, or cystoid degeneration; these mechanisms would not be anticipated to be altered by the administration of ranibizumab. The relationship of initial thickness value with a more favorable OCT outcome may be influenced by statistical issues of floor effects and regression to the mean, rather than representing a true biologic relationship; however, our analysis of change in percentage of excessive retinal thickness suggests that it may be a true relationship.

When a composite outcome was constructed that consisted of both a favorable OCT change and visual acuity change at 1 year, the predictors of this outcome were baseline visual acuity, younger age, and the presence of hard exudates in the macula at the time of treatment initiation. This result is consistent with the results already discussed because each of these factors had been identified with either a more favorable visual acuity outcome or a more favorable OCT outcome.

In summary, multiple studies^5-7 have demonstrated that, for eyes with center-involved DME, intravitreal ranibizumab improves both functional and anatomic outcomes compared with laser treatment alone. An extensive post hoc search of baseline factors to differentiate these outcomes among these individuals being treated with ranibizumab, with a median number of 8 injections in the first year, did not identify any factors that could be considered to be contraindications to treatment. It did identify some factors that may be associated with a greater improvement, and this can be helpful in discussions with patients when balancing the risks and benefits of treatment for center-involved DME.

Correspondence: Haijing Qin, MS, Jaeb Center for Health Research, 15310 Amberly Dr, Ste 350, Tampa, FL 33647 (drcrstat3@jaeb.org).

Submitted for Publication: December 19, 2011; final revision received March 12, 2012; accepted March 12, 2012.

Published Online: May 14, 2012. doi:10.1001/archophthalmol.2012.1107

Funding/Support: This work was supported through a cooperative agreement from the National Eye Institute and the National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Department of Health and Human Services (grants EY14231, EY14229, and EY018817).