Background: Colorectal cancer cells are shed into the stool, providing a po
tential means for the early detection of the disease using noninvasive appr
oaches. Our goal was to develop reliable, specific molecular genetic tests
for the detection of colorectal cancer in stool samples. Methods: Stool DNA
was isolated from paired stools and primary tumor samples from 51 colorect
al cancer patients. Three genetic targets-TP53, BAT26, and K-RAS-were used
to detect tumor-associated mutations in the stool prior to or without regar
d to the molecular analyses of the paired tumors. TP53 gene mutations were
detected with a mismatch-ligation assay that detects nine common p53 gene m
utations. Deletions within the BAT26 locus were detected by a modified soli
d-phase minisequencing method. Mutations in codons 12 and 13 of K-RAS were
detected with a digital polymerase chain reaction-based method. Results: TP
53 gene mutations were detected in the tumor DNA of 30 patients, all of who
m had the identical TP53 mutation in their stools. Tumors from three patien
ts contained a noninherited deletion at the BAT26 locus, and the same alter
ations were identified in these patients' stool specimens. Nineteen of 50 t
umors tested had a KRAS mutation; identical mutations were detected in the
paired stool DNA samples from eight patients. In no case was a mutation fou
nd in stool that was not also present in the primary tumor. Thus, the three
genetic markers together detected 36 (71%) of 51 patients (95 % confidence
interval [CI] = 56% to 83%) with colorectal cancer and 36 (92%) of 39 pati
ents (95% CI = 79% to 98%) whose tumors had an alteration. Conclusion: We w
ere able to detect the majority of colorectal cancers by analyzing stool DN
A for just three genetic markers. Additional work is needed to determine th
e specificity of these genetic tests for detecting colorectal neoplasia in
asymptomatic patients and to more precisely estimate the prevalence of the
mutations and sensitivity of the assay.