According to http://www.clinchem.org/content/47/4/624.full
> CEA as a Marker for Colorectal Cancer
> screening
> In screening for colorectal cancer, the aim should be to detect disease at either Dukes’ A or B stage. Malignancy detected at more advanced stages is unlikely to be more treatable than that detected through the usual course of events. Using an upper limit of normal of 2.5 μg/L, Fletcher (19) calculated that CEA has a sensitivity of 36% and a specificity of 87% in screening for Dukes’ A and B colorectal cancer. These findings, combined with the low prevalence of this malignancy in unselected populations, render the positive predictive value of CEA unacceptably low and thus of little value in screening healthy subjects. For the present, therefore, we must rely on fecal occult blood and endoscopy to screen for colorectal cancer (20).
What does 36% sensitivity and 87% specificity mean? Imagine that you test ten thousand 40-year olds and twenty-five of them do actually have colon cancer (this is more or less consistent with statistics telling that 0.23 of them will be diagnosed with colon cancer before turning 50). You will get a true positive result in 9 cases (out of 25) and a false positive in 1297 (out of 9975). So only 0.7% of the positive cases are real, and you missed two thirds of the cancers.