Re: Fecal Occult Blood Screening in the Minnesota Study

Extract

In a recent issue of the Journal, Church et al. (1) reanalyzed data from the Minnesota study to estimate the sensitivity of the screening test (fecal occult blood test composed of six slides) and the sensitivity of the screening program. Developed approaches agreed with the crude estimate of test and program sensitivities of about 90%. We agree with Church et al. with regard to the great importance of distinguishing screening test and screening program sensitivity. However, we believe that the study by Church et al. has some limitations with regard to the methodologies proposed and the public health implications.

Methodological limitations. As the investigators (1) suggested, the proportion of cancers detected by a particular test among all cancers diagnosed in complying people within 1 year of their screen in the Minnesota study provides a proper estimate for program sensitivity but not for screening test sensitivity, since it would suppose that all cancers existing at a screen emerge within 1 year after the screen and that all cancers emerging 1 year after the screen existed at the screen. To distinguish cancers entering the detectable preclinical phase after the screen from those existing at screen, Church et al. (1) suggested an adjustment for the incidence of cancers occurring after a false-positive test, assuming that all prevalent cases among positive tests have been detected. This method would be completely relevant if all positive case patients had undergone a complete colonoscopy. In fact, only 827 (83.9%) of 986 positive case patients underwent a complete colonoscopy so that some prevalent cases remained undiscovered among the positive case patients (2). Thus, such an adjustment underestimates the number of false-negatives and consequently overestimates the test sensitivity.