Multivariable models for advanced colorectal neoplasms in screen-eligible individuals at low-to-moderate risk of colorectal cancer: towards improving colonoscopy prioritization

Table 4 Performance of sequential modelling strategy at different sensitivity thresholds for CRC and HRA detection

Performance characteristic	Sensitivity of CRC detection in complete cohort (Model #1)			Sensitivity of CRC or HRA detection in residual cohort (Model #2) (%)
Performance characteristic	100%	99%	95%
% missed CRC	0	0	0.58	80
% missed HRA	6.2	7.2	13.7
% colonoscopies avoided	16.4	18.7	31.3
% missed CRC	0	0.58	0.58	70
% missed HRA	9.5	10.5	20.7
% colonoscopies avoided	23.8	25.5	40.9
% missed CRC	0	0.58	0.58	60
% missed HRA	12.7	14.4	27.6
% colonoscopies avoided	27.5	30.5	48.7

Interpretation (100% Column, 80% row): At minimum probability cut points to permit 100% CRC detection in Model #1 and 80 CRC/HRA detection in Model #2, application of the sequential modelling strategy predicts a miss rate of 0% for CRC and 6.2% for HRA and reduces colonoscopy burden by 16.4%

ISSN: 1471-230X