Author Response: Effect of Ruboxistaurin on the Visual Acuity Decline Associated with Long-standing Diabetic Macular Edema

We appreciate the comments and requests for clarification from Sahu and Majji on our article, “Effect of Ruboxistaurin on the Visual Acuity Decline Associated with Long-standing Diabetic Macular Edema.” ¹ Regarding their first question of baseline comparability between groups, the demographic factors they mention were shown to be equivalent between treatment groups in the original PKC-DRS2 study. ² The analysis in our recent article that included eyes with a baseline visual acuity (VA) >70 letters excluded only approximately 20% of the patients, and so baseline metabolic characteristics for the roughly 80% remaining would not be expected to differ substantially from the study population as a whole. In addition, as described in the article, we performed an adjusted analysis taking into consideration baseline eye characteristics that may affect DME progression, such as visual acuity, DME severity, prior focal photocoagulation, and DR severity. The results were consistent with our unadjusted results. 

Regarding their remaining concerns and questions about Figure 2, indeed, 75% of the subjects in the first PKC-DRS2 study completed the entire 36 months of follow-up, and there was no difference between treatment groups in the percentage of subjects who withdrew early. Additional valuable data were also obtained from subjects with shorter follow-up times, however, in that we were able to assess their duration of central macular involvement and to correlate it with their baseline-to-endpoint change in visual acuity, as well. 

Regarding the details of the analysis and presentation of data in Figure 2, to give a visual presentation of the distribution of the data (the VA change from baseline broken down by duration of severe DME by treatment group), we used box plots. Within each box, the dot represents the mean, the dark line represents the median, and the main box shows the interquartile range (IQR, the difference between the 3rd and 1st quartiles). The two whiskers show the upper adjacent value (UAV, the largest observation that is less than or equal to the 3rd quartile plus 1.5 × IQR) and the lower adjacent value (LAV, the smallest observation that is greater than or equal to the 1st quartile minus 1.5 × IQR). No confidence interval is presented in Figure 2. 

The trend analysis results mentioned in this article (relationship between duration of severe DME and VA in the results section) were based on the following analysis: Within each of five duration categories (Fig. 2), the mean and SE of the VA change from baseline of each treatment group were calculated. The duration of severe DME was calculated as either the exact time (for categories 0 month and 3 months) or the middle point of the category—for example, for category ≥6 to <18, the middle point is (6 + 18)/2 = 12. For the unadjusted analysis, a weighted linear regression analysis was performed, using the mean VA change as the response variable, the duration of severe DME as the dependent variable, and the inverse of the SE as the weight (a category containing a large number of eyes would have a smaller SE and hence a larger weight in such an analysis). This analysis was performed for each treatment group to get the estimated rate of VA change by month and its associated P-value. To compare the rates from two treatment groups, a likelihood-ratio test was used (a P-value of 0.010 was obtained). Based on that test result, we concluded that the rates of VA change were different between the two groups. With inclusion of baseline factors in this model, results related to the adjusted rate of VA change were obtained and were consistent with the unadjusted results. 

Finally, given that the analyses presented in the article were post hoc and hypothesis-generating, a mention of the power of the study to detect a difference between treatment groups would not be appropriate.