95 Comparison of MRI response evaluation methods in rectal cancer: a multicentre and multireader validation study 6 Table 3 Diagnostic performance to detect a complete response with specified results demonstrating effects of reader experience level and image quality Sensitivity Specificity PPV NPV Accuracy mrTRG Average (all readers) 57% 64% 44% 77% 62% Expert readers Non-expert readers 55% 58% 78% 60% 54%* 41%* 80% 77% 71%* 59%* Optimal quality Below average quality 54% 60% 68%# 60%# 45% 43% 79% 76% 64%# 60%# modTRG Average (all readers) 40% 79% 50% 75% 67% Expert readers Non-expert readers 34% 42% 90%* 76%* 63%* 46%* 76% 75% 73%* 66%* Optimal quality Below average quality 34%# 46%# 85%# 74%# 52% 47% 76% 75% 70%# 65%# DWI patterns Average (all readers) 37% 82% 52% 75% 68% Expert readers Non-expert readers 36% 37% 90%* 79%* 67%* 48%* 77% 75% 74%* 67%* Optimal quality Below average quality 31%# 43%# 88%# 76%# 57%# 48%# 76% 74% 71%# 65%# Split scar Average (all readers) 36% 79% 46% 75% 66% Expert readers Non-expert readers 26% 39% 89% 76% 51% 45% 74% 75% 70% 65% Optimal quality Below average quality 36% 37% 84%# 74%# 50% 43% 77% 73% 70%# 63%# Notes: · Results were calculated using a complete response as the positive outcome and residual tumor as the negative outcome. · Expert readers (n=5) were MRI experts with ≥10 years dedicated experience in rectal MRI; non-expert readers (n=17) were abdominal radiologists or general radiologists with a specific interest in abdominal imaging. MR image quality was categorized as optimal in n=52 cases and as below average in the remaining n=38 cases. · Results printed in bold indicate a significant effect size as assessed using mixed model linear regression, with * indicating a significant difference in diagnostic performance between expert and non-expert readers, and # indicating a significant difference in diagnostic performance between scans with optimal and below average image quality. Effect sizes, confidence intervals and levels of statistical significance (p-values) are further detailed in Supplement 2.
RkJQdWJsaXNoZXIy MjY0ODMw