Abstract
The choice of test for colorectal cancer screening is currently a major public health issue. We entirely agree with Burch et al. 1 asserting in the Journal that ‘well-designed diagnostic cohort studies directly comparing guaiac-and immunochemical-faecal occult blood tests (FOBTs) are required to determine the true comparative accuracy of these different testing strategies’. In their review, Burch et al. 1 identified only two such studies.
We have ourselves set-up such a study. Since June 2004, all average-risk individuals aged 50-74 years in the geographic area of Calvados (Normandy, France) are offered the opportunity to simultaneously perform a G-FOBT (non-rehydrated Haemoccult II, with two faecal samples each from three consecutive stools, with no particular dietary restriction), and one I-FOBT (Magstream, with two samples on two different days). Individuals with at least one positive test are examined further by colonoscopy. The two FOBTs are processed independently, both patients and physicians (including gastroenterologists) being blinded to each individual test result. After testing, individuals negative for both tests or non-compliant with colonoscopy are assessed for the presence of invasive cancer two years after the test using the local digestive cancer registry.
We compared the preliminary results of our diagnostic cohort study obtained on the first 10,673 pairs of tests, 2 with those of four studies (two diagnostic cohort studies3,4 and two case-control studies5,6) reporting direct comparison between one of two G-FOBTs (Haemoccult or Haemoccult SENSA) and an I-FOBT for the detection of colorectal cancer identified by Burch et al. 1 using ratio of sensitivities (RSN) and ratio of false-positive rates (RFP). 7 A RSN > 1 implies that sensitivity of the I-FOBT is greater than that of the G-FOBT, and a RFP >1 implies that the specificity of the I-FOBT is lower than that of the G-FOBT. Figure 1 provides a relative receiver operating characteristics (ROC) curve plotting RSN and RFP for each comparison presented in these studies. Several points may correspond to a single study since some studies evaluated several thresholds for a given quantitative immunochemical FOBT, 2 whereas others evaluated several numbers of samples for a given FOBT, 5 or several FOBTs.3,5 With new generation I-FOBT, involving automated reading and quantitative outcome, the issue of test choice is addressed through the choice of a positivity threshold and the number of samples providing a choice of the ideal balance between sensitivity and specificity as suggested in our results 2 and in other studies evaluating the diagnostic accuracy of I-FOBTs, without any comparison with G-FOBT.8,9 Compared with G-FOBT, a certain range of positivity threshold for I-FOBT provides a gain in both sensitivity and specificity. Moreover, costs associated with newly automated I-FOBT (such as Magstream) are reduced in comparison with non-automated tests, hence, the foreseeable change towards the use of I-FOBT for colorectal screening purposes. 10 More studies are required to evaluate cost-effectiveness of alternative screening strategies using I-FOBT according to number of samples and positivity threshold.
Results of studies for the detection of colorectal cancer plotted in relative ROC space (case-control and cohort studies). RSN are ratio of sensitivities and RFP are ratio of false positive rates comparing an immunochemical-faecal occult blood test (I-FOBT), with a guaiac-faecal occult blood test (G-FOBT). Black symbols are results from cohort studies, and white symbols from case-control studies. Squares (■) are results from studies using the Hemoccult SENSA, and lozenges (♦) are results from studies using the Haemoccult. Points linked with dotted lines (…) are results from the same study comparing sensitivities and specificities of a G-FOBT and an I-FOBT analysed using several numbers of samples. Points linked with continuous lines (-) are results from the same study comparing sensitivities and specificities of a G-FOBT and an I-FOBT analysed using several positivity thresholds