Accuracy for Predicting Deep Submucosal Invasion

A video with the lesion assessment, without any data on the patient, will be recorded in a
device connected to the processor provided by the Principal Investigator. The name of the
file will be the record ID. All the lesions will be tested by the same endoscopist in vivo
and an assistant will fulfill the data collection sheet during the colonoscopy.

First, the lesions will be cleaned and observed in a stable position. Size, location,
morphology, demarcated areas, and gross morphological malignant features will be evaluated.
Based on these WLE characteristics, a deep invasion prediction will be performed (control
test). Second, the lesion will be assessed using NBI with near focus or magnification or BLI
with magnification. A second cleaning with pronase (or N-acetylcysteine if pronase is not
available) if the surface cannot be clearly observed because of the presence of mucus or if
crystal violet is going to be used. Crystal violet 0.05% will be used in case of polyps type
2B in the JNET classification or lesions with a demarcated area. A non-traumatic catheter (or
spray catheter) will be used to spray the crystal violet over the lesion. A final prediction
of deep invasion will be performed for NBI or BLI ± CE (test evaluated).

The use of a cap to observe the bottom of the lesion, fix the lesion close to the endoscope
or to observe the lesion underwater immersion is strongly recommended.

The resection technique will be decided upon according to the local experience. In case of
endoscopy resection (cold snare, EMR, ESD, full thickness), lesions will be removed via the
anus (not through the endoscopy channel) in order to preserve their integrity. Although EMR
is performed, if possible, lesions will be referred to the pathologist well oriented and
pinned out on a cork based, as is standard procedure in ESD.

In order to ensure that endoscopic assessment is performed before the histology evaluation,
both diagnostic assessments (control test and test evaluated) will be recorded on the REDCap
database on the day of the colonoscopy. REDCap records the time and date of all changes in
the variables' results. The remaining variables (demographic data, etc.) will be recorded on
the data collection sheet and copied later into REDCap.

Videos of the lesion assessments will be sent to the Principal Investigator. Centralized
visualization will be conducted to detect protocol violations and to exclude lesions from the
study.

A blinded histology assessment will be conducted by the local pathologist and if a carcinoma
with submucosal invasion is diagnosed, histology slides will be referred for an additional
blinded and centralized histology evaluation at the end of the study.

Pathologists participating in the histological phase will assess all the slides with
submucosal invasion and will collect the histological factors associated with lymph node
metastasis.

Finally, investigators participating in the translational phase will refer paraffin blocks of
10 lesions of each JNET category (2A, 2B and 3) for genetic tests (sequencing of a panel of
45 genes and analysis of alterations in the number of copies of the genome).

Inclusion criteria:

- Non-pedunculated type 0 lesions in Paris classification (not obvious cancers)

- Lesions larger than 10 mm

Exclusion criteria are:

- Lesions assessed as JNET 1 by the endoscopist or serrated by the pathologist

- Previous biopsy or resection attempt

- Previous CT, MR or USE

- Unavailable histology

- Inflammatory bowel disease

- Informed consent not obtained

- Protocol violation