[Diagnostic value of colonoscopy indication as predictor of colorectal cancer: is it possible to design a fast track diagnosis?]

Abstract

Background: Diagnostic delay in patients with colorectal cancer (CRC) is a quality indicator and its reduction could improve prognosis of the disease.

Objective: To analyze the diagnostic value of different colonoscopy indications in CRC and to select the signs or symptoms that, if prioritized in a rapid diagnostic circuit, would be most efficient.

Material and methods: A retrospective analysis of 2219 outpatients who underwent colonoscopy from 2000 to 2007 was performed. For each indication we calculated the sensitivity (S), positive predictive value (PPV), positive likelihood ratio (LR+), and number of colonoscopies needed to diagnose a case of CRC (NND).

Results: A total of 179 patients were diagnosed with CRC. The indications with greatest PPV were liver metastases (35.3%), suspicious radiological image (20.8%), and non-distal rectal bleeding (22%). Iron deficiency anemia (11%), constitutional syndrome (10%), any rectal bleeding (9.4%) and rectal syndrome (9%) had intermediate PPV. Constipation (6.3%), alternating constipation-diarrhea (3.3%), changes in bowel habits (3%), distal rectal bleeding (2.1%), diarrhea (1.8%) and abdominal pain (1.1%) had low PPV. The NND was 4 in liver metastases, 7 in non-distal bleeding and 8 in suspicious radiological image. Distal bleeding (13), diarrhea (14), abdominal pain (14), changes in bowel habits (15) and alternating constipation-diarrhoea (21) had negative NND. The subgroup of patients aged >or= 50 years showed lower NND in non-distal rectal bleeding (5), suspicious radiological image (5) and any rectal bleeding (16).

Conclusions: Patients with non-distal rectal bleeding should be prioritized over other indications in a strategy of rapid diagnosis of CRC. Age equal to or more than 50 years should also be considered because this factor seems to reduce NND. Distal bleeding, abdominal pain and changes in bowel habits had low PPV and were associated with other diagnoses than CRC. Consequently, prioritization of these factors would be inefficient.