Clarifying relationship between tests¡¯ performance, bleeding pattern and observation level.
The data of 32,225 average-risk subjects who performed both Hemoccult II (guaiac) and Magstream (immunochemical) tests were re-analysed by varying the cutoff and number of samples of Magstream.
The identical performances obtained using one or two samples of Magstream (lower cutoff for one sample) at the population level were explained by opposite patterns of bleeding according to the presence of advanced neoplasias. They translated into discrepancy at the individual level: for example a 60 % increase in sensitivity and 20 % in specificity observed with one (39 ng Hb/ml cutoff) or two samples (63 ng Hb/ml cutoff) Magstream compared with Hemoccult II meant that 28.5 % of the subjects testing positive with one sample (18.0 % in subjects with advanced neoplasias) would have been considered negative by using two samples of Magstream at a higher cutoff (and reciprocal).
The identical performance of immunochemical tests using one or two samples (different cutoff), explained by opposite pattern of bleeding according to advanced neoplasias is true only at the population level, the appropriate level for mass screening.