Mieke Bus

5 77 Figure 4: Boxplot of low grade (n=10) vs. high grade (n=15). No reliable attenuation coefficient could be calculated fromCIS. Each used attenuation value is an average of 5 individual measurements. Median (interquartile range) attenuation coefficient is 2.1 (1.7-2.3 mm-1) in low grade tumors compared to 3.0 (2.9-3.7 mm-1) in high-grade tumors (p-value <0.006). The green line is indicating the cut-off value of 2.4mm-1. ROC curve of low grade (n=10) vs. high grade (n=15). Each used attenuation value is an average of 5 individual measurements. The area under curve is 0.92 (95% confidence interval 0.7-1.0%, p-value <0.0001). When using a µoct cut-off value of 2.4 mm-1 this results in a sensitivity of 87% (60- 98%) and a specificity of 90% (56-100%). ROC-analyses of low-grade compared to high-grade showed an area under the curve of 0.92 (CI 95% 0.8-1.0, p-value <0.01) (fig 4B). This results in a sensitivity of 87% with a confi- dence interval of 60-98%, specificity of 90% (56-100%), PPV of 93% (66-100%), and PPV of 82% (48-98%) when using µ oct cut-off value of 2.4 mm -1 (table 3). Table 3: Statistical results OCT stage Biopsy stage OCT grade (µoct ± 2.4 mm-1) biopsy grade Sensitivity (95%CI) 100% (72-100%) 29% (4-71%) 87% (60-98%) 75% (43-95%) Specificity (95% CI) 92% (62-100%) 100% (63-100%) 90% (56-100%) 100% (63-100%) PPV (95% CI) 92% (62-100%) 100% (16-100%) 93% (66-100%) 100% (66-100%) NPV (95% CI) 100% (72-100%) 62% (32-86%) 82% (48-98%) 73% (39-94%) Prevalence (95% CI) 48% (27-70%) 47% (21-73%) 60% (39-79%) 60% (36-81%) AUC (95%) n.a. n.a. 0.92(0.7-1) n.a. P n.a. n.a. <0.0001 n.a. CI: Confidence interval, PPV: positive predictive value, NPV: negative predictive value, AUC: Area under the curve, P: p-value, n.a.: not available.