Immunochemical faecal occult blood tests (I-FOBT) detect more effectively advanced neoplasia than guaiac tests (G-FOBT). The study aim was to compare the performance of an I-FOBT whilst varying the positivity threshold and considering four analysis modalities: one sample was performed (MG 1 ), two samples were performed and at least one sample was positive (MG 21 ), both samples were positive (MG 211 ) or the mean of the two samples' log-transformed haemoglobin contents exceeded the cutoff (MG 2m ). Screening for colorectal cancer using both G-FOBT and two samples' I-FOBT was performed by an average-risk population sample of 20,322 subjects. Among the 1,615 subjects with at least one positive test, 1,277 had a satisfactory colonoscopy result; 43 invasive cancers and 270 high-risk adenomas were detected. The I-FOBT was reinterpreted under each analysis modality (a random selection of one sample led to MG 1 ). For all modalities, increasing the positivity threshold decreased sensitivity and increased specificity. The relative ROC curves (in reference to G-FOBT) demonstrated similar performance for MG 1 and MG 21 , and improved performance for MG 2m . MG 211 sensitivity was limited within the range of positivity thresholds evaluated. For any specificity, MG 2m provided the highest sensitivity. For any sensitivity, MG 2m provided the highest specificity. For any positivity rate, MG 2m provided both the highest sensitivity and specificity. This study suggests the replacement of MG 21 by MG 1 or, for even better performance, by MG 2m provided that two samples are performed with similar participation (which should be explored). The targeted positivity rate could then be achieved by choosing the positivity threshold. '
UICCKey words: colorectal neoplasms; occult blood; mass screening; sensitivity and specificity; colonoscopy A recent systematic review of randomised trials 1 has confirmed that the screening of average-risk populations using the guaiac faecal occult blood test (G-FOBT) reduces specific mortality related to colorectal cancer. However, the low sensitivity of G-FOBT limits the extent of its benefit. Numerous studies have emphasised the relevance of immunochemical faecal occult blood tests (I-FOBT) in improving the sensitivity of screening.2-8 Several authors have concluded that evidence obtained in favour of testing with guaiac can be extended to the immunochemical test without new trials.9-13 However, screening is aimed at large samples of asymptomatic populations and a gain in sensitivity is not to be counterbalanced by reduced specificity. Some I-FOBT are automatically read, and provide a quantitative measure of faecal bleeding. They consequently allow a positivity threshold to be chosen to optimise the balance between test sensitivity and specificity. Furthermore, as for the G-FOBT, I-FOBT can involve one, two or more samples. Previous papers assessing the influence of the positivity threshold on test performance did not report the results of each sample, when available, in screening settings. 3,5,[1...