Abstracts
1997 Digestive Disease Week

De novo colonic ulcers: endoscopic diagnosis approached by decision analysis.

JM Sarmiento, W Mayoral, BG Wolff, CJ Goustot. Department of General Surgery, Colon & Rectal Surgery, and Digestive Endoscopy. Mayo Clinic, Rochester, MN.

Introduction: Previously undiagnosed colonic ulcers lack characterization regarding epidemiology, symptomatology, clinicopathological correlation and clinical outcome.

Aim: To define the epidemiology and the role of endoscopy in the approach to de novo colonic ulcers.

Methods: A review of all the patients with unexpected ulcerated lesions diagnosed during total colonoscopy in 1995 was carried out. Patients with erosions, flat or apthoid colonic ulcers, with biopsy, and without primary diagnosis were included. Histology and clinical outcomes were reevaluated a year later. Correlation of these findings with those on endoscopy was evaluated by McNemar chi-square. Also, prior and posterior probabilities for each disease were evaluated using Bayes' rule, in a decision analysis setting.

Results: We found 198 patients who met study criteria. Twenty-one (11%) had an ulcerated mass, 6 (3%) and ulcerated polyp, and 171 (86%) "simple" ulcer. The first two groups had confirmed histologic diagnosis in all cases; the latter group comprises the sample of our study. The predominant symptom was: hematochezia, 66 (38%); diarrhea, 42 (25%); abdominal pain, 44 (26%); anemia, 12 (7%). Multiple ulcers were found in 34 (20%) of the patients. The diagnostic probabilities, using Bayes' rule are:

A-Diag          B-Prior prob (%)    C-Pathol/Clin(%)     Post prob (%)
 Crohn's              30                   21                51.4
 Ulc. colitis         25                   11                22.4
 Undeterm.            20                    6                 9.8
 Infectious            9                   11                 8.1
 Ischemic              2                   12                 2.0
 Malignancy            9                    7                 5.1

After correlating endoscopic, clinical, and pathologic findings:

Diagnosis          Endosc.        Clin/Path.    Coincidental     p
 Crohn's              49               35            29         0.006
 Ulc. colitis         40               18            16         0.000
 Undeterm.            33                9             6         0.000
 Infectious           15               18            12         0.31
 Ischemic              3               19             3         0.000
 Malignancy           14               11            10         0.18

Conclusion: Endoscopic diagnoses of malignancy and infection in de novo ulcers of the colon is accurate. Crohn's, Ulcerative colitis and ischemia ulcers are difficult to diagnose by endoscopc observations alone. Bayes' rule allows identification of expected probabilities of disease, and should be based on local incidence rates. This type of analysis is simple, and straightforward, and is the basis for further clinical decision analysis.