The model included patient and MRI characteristics including age at biopsy (as a continuous variable), race (as Asian, black, Hispanic, other, white, and unknown), MRI grade (as 3, 4, and 5), ROI location (as anterior or not), PSA (as a continuous variable), PSA density (as ≤0.15 and >0.15 ng/ml/cc), institution performing biopsy, prostate volume (as a continuous variable), ROI Density (as a continuous variable), and distance from ROI (as <10 and ≥10 mm). This multilevel model was an attempt to account for nested sources of variability (ie, cores within the same patient). To further evaluate the association of distance from ROI with csPCa, a multilevel regression with binomial distribution was performed at a per-core level of observation using PROC GLIMMIX in SAS 9.4 (SAS, Cary, NC, USA) with random intercept and patient as the subject. The impact of radiologist and urologist learning curve was investigated by evaluating CDR over progressing 2-yr increments ( Supplementary Fig. 1A). The band-specific CDR was stratified by institution performing biopsy, MRI grade, prostate-specific antigen (PSA) density, and prostate volume ( Fig. The CDR was evaluated across bands with a Cochran-Mantel-Haenszel test. A cancer detection rate (CDR) was also calculated by dividing prostate cancer (PCa) cores by total cores for each band ( Fig. To account for the nonindependence of cores within a patient, the 95% CIs were accounted for the correlated nature of the data. The cumulative distribution (and 95% confidence interval ) was presented for additive bands beyond the ROI ( Fig. 2A). The distribution of csPCa cores was calculated as csPCa cores for each band divided by total csPCa cores.
Each band was 5 mm in width, with the sixth band extending out beyond 25 mm from the ROI surface. Biopsy cores were grouped by distance from ROI surface into one of six concentric bands.
0 kommentar(er)
