Fig. 4
The graph A shows a fitting curve about predicted probability between cutout and ED. The probability of cutout increased with ED increasing (the gray filling area indicates a 95% CI of cutout rates in each ED) and the probability of cutout increased dramatically when “ED ≥ 1.” The graph B shows the ROC analysis between ED value and cutout rates. The best cutoff value of ED was 1.022 (just a little bit more than 1/3 times of RFD, sensitivity = 73.3%, specificity = 86.1%, AUC = 0.867, p < 0.001)