We read with interest the correspondence of Huisingh and McGwinn regarding the use of the statistical tests in the analysis of our recent study. In the analysis, we considered the reinverted eyes as the treatment group and used fellow eyes as the control group. We treated the data as unpaired and used the nonparametric Mann–Whitney U test for comparison between the 2 groups at various time points. The Wilcoxon signed-rank test is a frequently used nonparametric test for paired data (eg, consisting of pretreatment and posttreatment measurements) based on independent units of analysis. Hence, we used it for the comparison of outcomes in the phakic intraocular lens reinverted eyes (before and after viscocannula reinversion). It also was reported that this test cannot be used for paired comparisons arising from clustered data (eg, if paired comparisons are available for each of 2 eyes of an individual). Although the Mann–Whitney U test is ubiquitous in statistical practice for the comparison of measures of location for 2 samples where the assumption of normality is questionable, there have been reports in the literature in which the Mann–Whitney U test has been used comfortably for fellow eye comparisons in which the fellow eye was taken as the control.
We thank the authors for bringing out the effect of data clustering that can happen with the Mann–Whitney U test. Rosner and associates used extension of the signed-rank test for clustered data. They proposed a corrected variance formula of the Wilcoxon signed-rank test. Kitaya and associates used the Mann–Whitney U test to compare choroidal blood flow in eyes with central serous chorioretinopathy and their fellow eyes. Choroidal blood flow measurements were obtained in 22 eyes of 11 patients with central serous chorioretinopathy and were compared with fellow eyes. Contreras and associates reported the comparison of unaffected and affected eyes with nonarteritic anterior ischemic optic neuropathy using the Mann–Whitney U test. They compared the optic nerve head changes, namely, the disc area and cup-to-disc ratio, at various time points, such as the acute phase, 3 months, and 6 months, in affected eyes and fellow eyes.
We admit that the clustering effect can occur in data comparisons like this involving fellow eye comparisons in nonparametric tests. However, both the Wilcoxon and Mann–Whitney U tests can be used with debatable adjustments for clustered data in ophthalmology.