Statistical analyses and plotting were performed using SAS for Windows (v 9.4; SAS Institute, Cary, NC) and GraphPad 5.0 (La Jolla, CA). In all tests,
P values less than 0.05 were considered statistically significant. ILS for the control and experimental groups are reported as mean ± standard deviation in millivolts as obtained from the lock-in amplifier. Unpaired
t-tests (two-tailed) were performed, at both slit widths, to analyze the difference in ILS between two observers, to measure interobserver variability (
Fig. 4A). For intraobserver variability, each observer recorded eight to 10 measurements in the same eye, and independent sample
t-tests were used to compare the means between two observers at both slit widths, with standard deviation used to describe variability (
Fig. 4B). A Kruskal-Wallis test was used to determine the statistical significance comparing ILS at different SUN grades, for each slit width, and a regression analysis evaluated the relationship of ILS on increasing SUN scores (
Fig. 5;
Table 2). A series of unpaired
t-tests were used to compare ILS in patients on postoperative days 1 and 4, as well as with healthy and PC-IOL patients (
Figs. 6A,
6B). Wilcoxon signed rank tests for paired data were applied to analyze the longitudinal change in ILS readings between postoperative days 1 and 4, for each slit width (
Fig. 7). Logistic regression models and receiver operating characteristic (ROC) analyses were performed to identify the slit size that provides maximum sensitivity and maximum specificity in discriminating SUN scores (grade 0, 1+, and 2+) from ILS. The diagnostic accuracy of ILS in predicting qualitative SUN scores was measured with the area under the curve (AUC; determined with 95% confidence interval) and cut-off values of ROC curves (
Fig. 8). The interaction term between ILS and slit size in the logistic regression models provided a significance test of whether ILS is more or less predictive of SUN scores between the two slit widths. Power analysis was performed using G*Power (Version: 3.1.9.2; downloaded from
http://www.gpower.hhu.de/) to determine the number of patients needed to detect a significant decrease in the mean ILS between two levels of aqueous flares with a finer granularity than the current subjective SUN scores. The signal to noise ratio (S/N ratio) of the measured ILS signal (i.e., the ratio of mean ILS to the standard deviation of ILS) was also calculated to assess the effect of instrument noise (
Table 1).