Table 7.
ColonFlag performance based on histopathology findings
| Study | Sensitivitya | Specificitya | PPVb | NPVb | ORc | AUCd |
|---|---|---|---|---|---|---|
| Ayling, 2018 | CRC=52.4% | CRC=71.3% | CRC=6.3% | CRC=97.6% | ||
| CRC+HRA=46.9% | CRC+HRA=72% | CRC+HRA=13.1% | CRC+HRA=93.8% | |||
| Hilsden, 2018 | CRC=5.1 | |||||
| Advanced adenoma/SSP=2.0 | ||||||
| Non-advanced adenoma/SSP=1.7 | ||||||
| Non-neoplastic polyp=1.2 | ||||||
| Schneider, 2020 | CRC=35.4% | CRC=17.7% | CRC=0.78 | |||
| Adenoma=3.8% | Adenoma=1.3% | Adenoma=0.57 | ||||
| Ayling, 2021 | CRC=81.8% | CRC=73.5% | CRC=8.3% | CRC=99.3% | ||
| CRC+HRA=42.8% | CRC+HRA =73.4% | CRC+HRA =13.7% | CRC+HRA=92.8% |
HRA: High-risk adenoma; OR: Odds ratio; SSP: Sessile serrated polyp;
a
Using the predicted outcomes from a binary classification model and comparing them to the true outcomes of the instances.
b
Using the predicted outcomes from ColonFlag and comparing them to the true outcomes of the instances.
c
Calculated by comparing the odds of the event in the exposed group to the odds of the event in the unexposed group using data.
d
Computed by plotting a Receiver Operating Characteristics (ROC) curve based on model predictions and true labels, then calculating the area under this curve.