Timo Soeterik

146 CHAPTER 8 TABLE 1. Continued Overall pN0 pN1 Biopsy type N (%) TRUS-guided SB TRUS-SB+ MRI-TB 694 (65) 368 (35) 479 (63) 282 (37) 215 (71) 86 (29) Median (IQR) total nodes resected 20 (13 - 25) 17 (12 - 24) 20 (14 - 28) SAH: St. Antonius Hospital, HGT: Hospital Group Twente, CWH: Canisius Wilhelmina Hospital, SD: standard deviation, IQR: interquartile range, TRUS: transrectal ultrasonography, SB: systematic biopsy, MRI-TB: magnetic resonance image-guided target biopsy. Percentages may not total 100 due to rounding. When using mpMRI T-stage, discrimination in terms of AUC, increased to 0.72 (95% CI 0.71 – 0.73) for the MSKCC and 0.75 (95% CI 0.74 – 0.76) for the Briganti nomogram (Table S2). Mean predicted probability for LNI changed to 30% for the MSKCC and 27% for the Briganti nomogram (Table S2). As shown in the calibration plots, the agreement between predicted and observed probabilities was comparable (both moderate calibration) for both DRE T-stage and mpMRI T-stage. Calibration intercepts were closer to 0 when using mpMRI instead of DRE for both the MSKCC (0.04 [95% CI: 0 – 0.08] versus 0.08 [95% CI: 0.04 – 0.12]), and the Briganti nomogram (0.05 [95% CI: 0 – 0.08] versus 0.10 [95% CI: 0.06 – 0.13]); (Table S2). In a head-to-head comparison, calculating the LNI risk using mpMRI T-stage with the Briganti nomogram led to higher AUCs in all bootstrap samples, compared with Briganti DRE T-stage as well as both DRE T-stage and mpMRI T-stage with the MSKCC nomogram. Clinical usefulness Using mpMRI T-stage resulted a in higher true-positive rate and a higher false-positive rate for the detection of positive lymph nodes for all risk thresholds, compared to using DRE T-stage (Figure 2). Use of mpMRI T-stage led to increased sensitivity for the detection of LNI for all risk thresholds in both models, countered by a lower specificity, compared with DRE T-stage. In Tables S3 and S4, total numbers of missed LNI cases per risk threshold are presented, combined with rates of performed ePLND and number of positive LNI cases. For all thresholds, the number of missed LNI cases was lower when mpMRI T-stage was used, countered by higher rates of unnecessary ePLND (pN0). Decision curve analysis revealed that use of mpMRI T-stage in both nomograms resulted in higher net benefits, compared with DRE T-stage, for the risk thresholds between 5% and 20%. Net benefits for both the MSKCC and Briganti nomograms, using mpMRI T-stage, were comparable for this range of risk thresholds. For risk thresholds ranging from 20% and 30%, the combined use of mpMRI T-stage with the Briganti nomogram would lead to the highest net benefit (Figure 3).