The true-positive rates observed in this study for those found to have cancer in either population are likely to overestimate the true sensitivity of screen-detected cancers in the general population because the diagnostic cohort was included. The design is powerful for determining relative sensitivity, as results are paired for these tests. Better estimates of true sensitivity come from the screening population (Tables 3 and 4), although because not everyone had a colonoscopy, results shown in Table 4 represent the upper limit of actual sensitivity within the limitation of the number of detected cases."
In other words, InSure Fit's 88% sensitivity for cancer and 43% sensitivity for pre-cancer are not realistic, real-world measures. They're over-estimations, according to the study authors.
What does a more realistic evaluation of Insure Fit's accuracy in detecting cancer and pre-cancer look like? Thankfully, the
paper tells us:
An estimate of true sensitivity can be obtained from results in those screening cases where colonoscopy was performed (Table 4). Although numbers were relatively small, InSure was significantly more sensitive for cancer (75% vs. 37.5%) and significant adenomas (27% vs. 15%).
Oh! When the study authors looked only at patients where colonoscopy was performed -- the most accurate method to diagnose colon cancer -- the performance metrics of Insure Fit are downgraded significantly.
Cancer sensitivity is really 75%, not 88%. Pre-cancer sensitivity is 27%, not 43%, as "Alpha Exposure" claims.
The hedge fund doesn't include these more real-world Insure Fit numbers in its
article because they're numerically inferior to Cologuard -- and that just ruins the entire premise of the short thesis.
The remaining four parts of the "Alpha Exposure" short thesis on Exact Sciences might be more convincing. Part one was not.
-- Reported by Adam Feuerstein in Boston.