AIM Selection of patients for investigation of suspected colorectal cancer is difficult. One possible improvement may be to measure DNA isolated from exfoliated cells collected from the rectum. METHOD This was a cohort study in a surgical clinic. Participants were aged ≥40 years and referred for investigation of suspected colorectal cancer. Exclusion criteria were inflammatory bowel disease, previous gastrointestinal malignancy, or recent investigation. A sample of the mucocellular layer of the rectum was taken with an adapted proctoscope (the Colonix system). Haemoglobin, mean cell volume, ferritin, carcino-embryonic antigen and faecal occult bloods were tested. Analysis was by logistic regression. RESULTS Participation was offered to 828 patients, of whom 717 completed the investigations. Three were lost to follow up. Seventy-two (10%) had colorectal cancer. Exfoliated cell DNA was higher (P<0.001) in cancer (median 5.4 μg/ml [inter-quartile range 1.8,12]) compared with those without cancer (2.0 μg/ml [IQR 0.78,5.5]). Seven variables were independently associated with cancer, including age (odds ratio [OR], 1.05; 95% confidence interval [CI], 1.02,1.08; P<0.001) DNA (OR, 1.05; CI, 1.01,3.6; P=0.01), mean cell volume (OR, 0.93; CI, 0.89,0.97; P=0.001), carcino-embryonic antigen 1.02 per μg/l (CI, 1.00,1.04; P=0.02), male sex (OR, 2.0; CI, 1.1,3.6; P=0.02), rectal bleeding (OR, 2.4; CI, 1.3,4.5; P=0.007) and positive faecal occult blood (OR, 6.7; CI, 3.4, 13; P<0.001). The area under the receiver-operating characteristic curve for the DNA score was 0.65 (0.58-0.72) and for the seven variable model 0.88 (CI, 0.84-0.92). CONCLUSION Quantification of exfoliated DNA from rectal cellular material has promise in the diagnosis of colorectal cancer, but this requires confirmation in a larger study.