Thanks, Kay, for your question and for the link to the Blue-C study. This paper is a prime example of obfuscation. Basics first: of course, any test is more sensitive for cancerous lesions than for pre-cancerous lesions. The former shed more of the biomarker that the test is looking for than the latter. Hence, a positive result is more probable for the cancer patient than for the ACN patient with precancerous or non-proliferative lesions.

What makes this study an example of obfuscation: It reports the sensitivity (92 of 98 cancer cases detected among the >20000 participants). But that is not what interests the indivdual who takes the test. It is the positive predictive value (PPV) that tells you how probable cancer is on a positive result. I would be able to calculate the PPV from the prevalence, sensitivity, and specificity values that are normally provided in studies. But this study plays a neat trick on us. It reports SENS and SPEC values for different subgroups, mixing up the case definitions in a way that makes it impossible to reconstruct the "magic table". The paper refers to tables in the supplement from which we could derive those values, but when you open the supplement, those tables are missing (tables S4-9).

Add to that the fact the Cologuard sponsored the study, and what you have to have is serious doubt about the claim that its test is any better than any of the others. Not for the end-user anyway.

BTW, the 6th test in my spreadsheet is about Cologuard, and the study was done independent of manufacturer's sponsorship