User's Manual

Indications, Contraindications, Warnings, Cautions
Accuracy of the Device—COLON 2 19
Analysis of Agreement
Location-based and size-based analyses of the agreement of CCE with OC were conducted. Two polyp
size thresholds, 6 mm and 10 mm, were considered for defining a subject as positive for a polyp. If one
or more polyps were identified on OC, the polyp with the largest estimated diameter was used for
comparison purposes and is referred to as the ‘reference’ polyp. The colon segment location of the
reference OC polyp was determined and recorded as the cecum, ascending colon, transverse colon,
descending-sigmoid colon or rectum. The largest polyp identified on CCE within the same or an
adjacent segment location of the “reference OC” polyp was used to determine agreement with OC based
on a size-matching algorithm. If there were two polyps of equal size that were “the largest” and located
in different segments of the colon, the location and size-matching algorithm was repeated on each of
these reference OC polyps to determine agreement of CCE with OC. In such instances, the reference
OC polyp chosen for the final determination was the one that was in favor of the device. A full
description of the location and size-matching algorithm is given in the Appendix.
For each subject, the CCE evaluation was classified into one of four categories of agreement with OC:
true positive (TP) agreement, false negative (FN) disagreement, false positive (FP) disagreement, or
true negative (TN) agreement. Letting TP, FN, FP, and TN denote the number of subjects that fall into
these categories (Table 2), the estimates of positive percent agreement (PPA) and negative percent
agreement (NPA) of CCE with OC are defined as follows:
PPA = 100 × TP / (TP+FN)
NPA = 100 × TN / (FP+TN)
Tables 3 and 4 display the observed 2 by 2 categorical data when, respectively, 6 mm and 10 mm are
used as the size thresholds for defining a subject as positive for a polyp. From these data tables, PPA and
NPA are estimated along with their 95% confidence intervals (Table 5).
Table 2: Format for reporting agreement results in a 2 x 2 categorical table
CCE
OC
negative positive
negative TN FN
positive FP TP
Table 3: Observed Categorical Data, 6 mm Threshold
CCE
OC
negative positive Total
negative 413 60 473
Positive 95 132 227
508 192 700