Data are from a 36-month, randomized, double-masked, placebo-controlled, parallel, multicenter trial of once-daily ruboxistaurin for vision loss in patients with diabetic retinopathy. The clinical findings and characteristics of this trial have been reported previously 22. A total of 684 patients were randomized at 70 clinical sites in the United States. Randomized patients had type 1 or type 2 diabetes and HbA1c values ≤ 13%. To be eligible, patients had to have at least one eye that met ocular entry criteria: (a) Early Treatment Diabetic Retinopathy Study (ETDRS) retinopathy levels of > 47A and < 53E; (b) test-corrected visual acuity score of > 45 letters as measured by the ETDRS visual acuity protocol 23; (c) no history of panretinal photocoagulation for diabetic retinopathy; (d) no evidence of glaucoma; and (e) no history of conditions affecting diabetes retinopathy progression. Ethical review boards provided written approval of the study protocol and the informed consent document. The study was initiated at each site after the principal investigator and the study sponsor obtained approval documents. The principal investigator at each site was responsible for obtaining informed consent from each patient or legal representative and for obtaining the appropriate signatures on the informed consent document prior to performing any procedures or administering any study drug.

Visual acuity was determined as part of an ophthalmologic examination (including slit-lamp biomicroscopy, intraocular pressure, and ophthalmoloscopy) at screening and each study visit. Visual acuity was quantified in terms of number of letters as measured by the Early Treatment Diabetic Retinopathy Study (ETDRS) visual acuity protocol 23. These scores have a possible range of 0 to 100 letters, with higher scores indicating better visual acuity. For the current analysis, visual acuity was defined at each time point (i.e., baseline and month 18) as each individual's ETDRS rating for his/her better eye (referred to as "best eye ETDRS"), regardless of which eye is used at other time points, and regardless of whether the best eye met criteria for being included as a "study eye." Previous research has found that binocular acuity can be inferred from measures of monocular acuity in the better eye, particularly when assessing functional impairment associated with visual acuity, and separate assessment of binocular acuity is not required 32.

The health outcomes measures used in the current analysis were administered at baseline, month 18, and month 36. The current analysis focused on change from baseline to month 18 because of the sample size limitations associated with health outcomes data at month 36. Baseline analyses, including descriptive statistics, were performed using the sample of all enrolled patients. Analyses of change in visual acuity and/or HRQL were performed with the subset of patients who had at least one VFQ-25 subscale present at both baseline and month 18 (this subset is called the "change analysis sample"). Analyses were conducted to assess whether patients meeting this criterion differed from patients who were excluded. The primary measure of HRQL in the study was the condition-specific VFQ-25. Secondary analyses examined group differences in the generic SF-36.

Descriptive analyses were conducted to summarize demographic/clinical variables, visual acuity (ETDRS), and HRQL. The scoring of all questionnaires followed the guidelines set forth by the instrument developers. Categorical variables were summarized in terms of frequencies and percentages, and continuous variables were summarized in terms of means and standard deviations. Preliminary analyses were conducted to examine reliability and validity of the VFQ-25 to ensure that this instrument performed adequately in patients with diabetic retinopathy. Cronbach's alpha for each VFQ-25 scale was computed to assess internal consistency reliability. Construct validity of the VFQ-25 was examined with Spearman correlations between the VFQ-25 scales and visual acuity at baseline and month 18. In this study, correlations were interpreted as small (0.10), moderate (0.30), or large (0.50) following the guidelines proposed by Cohen 33.

The primary analyses focused on the link between visual acuity (i.e., ETDRS) and HRQL (i.e., the VFQ-25 and to a lesser extent, the SF-36) over time. This link was analyzed with change in visual acuity defined both categorically and continuously. First, categorical analyses were conducted to characterize the HRQL decreases/gains associated with various degrees of change in visual acuity. Patients were categorized into five groups based on ETDRS change from baseline to month 18: Improved by ≥ 10 letters, Improved by 5 to 9 letters, No change (-4 to +4 letters), Worsened by 5 to 9 letters, and Worsened by ≥ 10 letters. A series of 12 general linear models (GLMs) with Scheffe's post hoc pairwise comparisons were conducted to compare the VFQ-25 change scores of the five groups, when controlling for age, gender, and baseline visual acuity. In these models, the change group is a five-level independent variable. Age and baseline visual acuity are continuous covariates, and gender is a categorical covariate. The dependent variables were change scores from baseline to month 18. A parallel set of 12 models was conducted with the SF-36 subscales as the dependent variables.

To examine the relationship between the VFQ-25 subscales and visual acuity defined as a continuous variable, 12 additional GLMs were conducted. These models are the same as those described above, except visual acuity was entered as a continuous variable rather than a five-level categorical variable.

Finally, as an exploratory descriptive analysis, change scores for each VFQ-25 item were calculated for each of the five ETDRS change groups. These scores are presented to provide a rough indication of the precise areas of HRQL and functioning that may tend to change along with visual acuity in patients with diabetic retinopathy.

SAS statistical software version 8.2 was used for all analyses. All statistical tests were two-tailed and conducted with significance level for testing fixed at 0.05.

Baseline demographic and clinical characteristics for the total sample (N = 684), change analysis sample (N = 535), and patients excluded from the change analysis sample (N = 149) are presented in Table 1. There were no statistically significant differences between the change analysis sample and the excluded sample in age, gender, type of diabetes, duration of diabetes, body mass index (BMI), or baseline ETDRS. The only difference between the two samples was in ethnicity, as the change analysis sample had a somewhat larger percentage of Caucasian patients. The change analysis sample was primarily male (64.1%) and Caucasian (81.9%), with a mean age of 59.3 years. This sample was primarily diagnosed with type 2 diabetes (88.0%), and the sample had a mean BMI of 32.8 at baseline.

Baseline descriptive statistics for the VFQ-25 are presented in Table 2. Mean subscale scores ranged from 69.5 (general vision) to 94.6 (color vision), and the total score was 84.1. At baseline, there was substantial heterogeneity in the sample, with scores ranging from 0 to 100 in several subscales. The multi-item subscales of the VFQ-25 generally had adequate internal consistency reliability. Cronbach's alpha was 0.68 for the ocular pain subscale, 0.64 for the social function subscale, and at least 0.72 for all other subscales. Correlations with ETDRS visual acuity rating were statistically significant (p < 0.01) for all subscales except ocular pain. The correlation coefficients were in the moderate range (i.e., ≥ 0.30) for the total score and the general vision, near vision, distance vision, role limitations, and driving subscales. In the change analysis sample, mean best eye ETDRS visual acuity was 81.3 letters at baseline (Table 1) and 81.4 letters at month 18.

The five change groups that are examined in the subsequent analyses were compared in terms of age and baseline visual acuity. There were no significant differences among the groups in age. However, there were some group differences in baseline visual acuity. Baseline ETDRS visual acuity ratings for the five groups were as follows: 72.2, Improved by ≥ 10 letters; 79.0, Improved by 5 to 9 letters; 83.0, No change (-4 to +4 letters); 81.5, Worsened by 5 to 9 letters, and 79.9, Worsened by ≥ 10 letters. The group that improved by ≥ 10 letters had significantly worse baseline visual acuity than the other four groups (p < 0.01). The only other significant difference was between the group that improved by 5 to 9 letters and the no change group (p < 0.01).

Results of GLMs examining the link between best eye ETDRS visual acuity and the VFQ-25 are presented in Table 3. Change in visual acuity was generally associated with corresponding changes in most VFQ-25 scores. However, pairwise comparisons revealed no statistically significant differences in mean VFQ-25 change scores among the two improvement groups and the no change group. Compared with the improved by ≥ 10 letters group, the worsened by 5 to 9 letters group had significantly different change scores in the VFQ-25 dependency subscale and total score. The group that worsened by ≥ 10 letters had significantly different change scores than most of the other groups, including the no change group, in all VFQ-25 scales except ocular pain. There were no differences among groups in the ocular pain.

A parallel set of models was conducted to assess whether the five change groups differ in HRQL change as assessed by SF-36 subscales (Table 4). Results suggest that change in the SF-36 did not correspond as closely as the VFQ-25 to change in visual acuity. For example, the SF-36 reflected slight declines in overall quality of life from baseline to month 18, even among the groups that improved in visual acuity. The only statistically significant pairwise comparison between groups was between the no change group and the worsened by ≥ 10 letters group in the mental health subscale.

In addition, 12 GLMs were conducted to examine the relationship between change in visual acuity defined as a continuous variable and change in the VFQ-25 scales, controlling for age, gender, and visual acuity at baseline. In all models, except the one with the VFQ-25 ocular pain subscale as the dependent variable, change in visual acuity was significantly associated with change in the VFQ-25 scale (p < 0.0001). Baseline visual acuity was also associated with the following four dependent variables: Distance Activities, Dependency, Driving, and the VFQ-25 total score (p < 0.05). Age and gender were not significantly associated with the VFQ-25 scales in any of the models.

Mean change in each individual item of the VFQ-25 was computed for each of the five groups from baseline to month 18 (Table 5). In general, most individual items tended to reflect improvement among the groups that improved in ETDRS and worsening among the groups that decline in ETDRS. Items that followed this logical pattern included (item 2) present eyesight rating, (item 5) difficulty reading newsprint, (item 6) seeing well up close, (item 7) difficulty finding objects on a crowded shelf, (item 8) difficulty reading street signs, (item 17) accomplish less due to vision, (item 18) work less due to vision, and (item 21) frustrated due to vision.