The learning curve for in-vivo characterisation of small colonic

polyps: number needed to train (NNT) is 200 polyps

PJ Basford, GR Longcroft-Wheaton, P Bhandari
Portsmouth Hospitals NHS Trust, Portsmouth, UK

Introduction & Aims Results cont.

• Indigocarmine (IC), narrow-band imaging (NBI) and
Flexible Spectral Imaging Colour Enhancement (FICE)
have been shown to be effective in the in-vivo diagnosis
of small colonic polyps.
• The learning curve for achieving high level of accuracy
with a new technology for real-time diagnosis of small
colonic polyps has not been determined.
Methods
• We aimed to assess the learning curve of a novel
electronic in-vivo diagnosis technology (Pentax iScan) for
an expert endoscopist.
• Patients presenting for screening colonoscopy through
the UK Bowel Cancer Screening Programme were
• 154 polyps were non-neoplastic, 245 were adenomatous
and 1 contained adenocarcinoma.
• Sensitivity and overall accuracy improved significantly for
both imaging modalities in the 3rd set of polyps as
compared to sets 1 and 2 (p<0.05). In Set 3&4 combined
WL overall accuracies of 93.5% and 95.0% were achieved
with WL and iScan respectively. There were no significant
differences in overall accuracy between the 2 modalities
in Sets 3 and 4.
• Negative predictive values for adenomatous histology of
recto-sigmoid polyps ≤5mm for the entire study were
97.4% and 95.1% for WL and iScan respectively.
iScan Overall Accuracy - Learning Curve
100

Accuracy of In-vivo predicted histology %

90
prospectively recruited.
80
• All colonoscopies were performed by a single expert
70
endoscopist, with extensive experience in in-vivo 60
diagnosis, using Pentax EC-3890Li 1.2 Megapixel HD 50
colonoscopes and EPKi processor. 40
• Polyps <10mm in size were assessed sequentially using 3 30
modalities 1) White light HD endoscopy (WL), 2) Pentax 20
iScan surface, contrast and tone enhancement. Optical 10
magnification was not used. 0
1 to 100 101-200 201-300 301-400
• Predicted histology (non-neoplastic, adenoma, cancer)
WL 75 83 92 95
was recorded for each modality and compared to the WL (%) iScan (%)
iScan 82 82 93 97
final histopathological diagnosis. Set 1 Sensitivity 78.8 86.8
• Results were analysed for sensitivity and specificity for (Polyps 1-100) Specificity 70.8 76.6
neoplasia, and overall accuracy. Accuracy 75.0 82.0
• To assess any learning effect results were analysed in 4 Conclusions
sets of 100 consecutive polyps. Set 2 Sensitivity 86.6 85.1 1. Even in expert hands there is a significant learning curve
(Polyps 101-200) Specificity 75.8 75.8 for using a new technology for the in-vivo diagnosis of
Accuracy 83.0 82.0 small colonic polyps, with improvement in performance
over the first 200 polyps assessed.
Results Set 3 Sensitivity 96.2 98.7 2. Excellent results can be achieved once the new
• A total of 400 polyps in 168 patients were eligible for (Polyps 201-309) Specificity 77.3 72.7 technology has been mastered.
inclusion in the study. Accuracy 92.0 93.0 3. This is the first report of results achieved with high-
• Mean polyp diameter was 4.0mm, median 3mm. 169 definition white light endoscopy which are comparable
polyps were in the proximal colon and 231 in the distal Set 4 Sensitivity 92.3 94.2 with electronic chromoendoscopy and IC
(Polyps 301-400) Specificity 97.9 100.0
colon. chromoendoscopy.
Accuracy 95.0 97.0