Exact Sciences (EXAS) (EXAS or the "Company") is the Wisconsin-based company behind the overhyped soon to fail colon cancer screening test known as Cologuard. As such, we believe that the fair value for this persistently cash flow negative, value-destroying Company is at best equal to the value of the cash on its balance sheet, or less than $3 per share. We will demonstrate over the course of this series, why we feel that the stock should be aggressively sold or shorted. This article is the first in a five part series which will cover:
- Why the Cologuard test is actually worse than currently available tests
- Why the economics of the Cologuard test do not support reimbursement
- Why almost nobody will use the Cologuard test even if it gets approved
- Why the Cologuard test is no Pap Smear (the Bull case deconstructed)
- How Exact Sciences has misled investors (for the third time)
In order to be successful, Exact Sciences will need to show that Cologuard is an effective test, gain reimbursement, convince doctors to prescribe the test, and make a simple enough test that patients will use it. We believe that they will fail at each of these tasks. In this article, we will explain why we believe that the Cologuard test has a high probability of being rejected by the FDA and why, in the unlikely event that it is even approved, it will almost certainly be a commercial failure due to its lack of efficacy.
Exact Sciences would like investors to believe that Cologuard will revolutionize the detection of colon cancer. The Company believes that Cologuard will replace the colon cancer screening Fecal Immunochemical Test (FIT), just as the FIT has progressively replaced the Fecal Occult Blood Test (FOBT) in recent years. To support this contention, the Company has used a highly promotional slide deck in conversations with institutional investors. Curiously (perhaps because Exact knows it is misleading?), a copy of this slide deck has never been filed with the SEC. To aid investors, we have posted a copy of it here.
Cologuard Is Actually Less Effective And More Expensive Than FIT
Exact Sciences highlights its contention that the Cologuard test is superior to the current standard of care, the FIT, by touting Cologuard's sensitivity and specificity. The Company reports the results of the Cologuard test as having a cancer sensitivity of 92%, a pre-cancer sensitivity of 42%, and a specificity of 87%. In contrast, Exact Sciences claims that the FIT has a cancer sensitivity of just 66%, a paltry pre-cancer sensitivity of 20%, and a specificity of 95%. A higher sensitivity and a higher specificity are better. A copy of the Company's slide with the FIT's performance is below:
In order to see how EXAS has misrepresented Cologuard, we need to review two important statistical concepts when evaluating a medical test: sensitivity and specificity. According to the Company's recent press release conveying its results, "Sensitivity measures a test's ability to correctly identify positive results. It is the percentage of patients who were determined by colonoscopy to have pre-cancerous polyps or cancer that had a positive Cologuard test result. Specificity measures a test's ability to correctly identify negative results. It is the percentage of patients who were determined by colonoscopy not to have pre-cancerous polyps or cancer that had a negative Cologuard test result."
While it seems from the data that the Cologuard test is better than the FIT, nothing could be further from the truth. In fact, the FIT, when properly administered, is superior to Cologuard. Exact Sciences hides this statistical fact by not providing an apples-to-apples comparison. When compared over the same number of years, the FIT proves to be more effective in identifying colon cancer. This article will clearly explain why.
Let's start by reviewing in a table format what Exact Sciences would like investors to believe:
First, it is important to consider the vastly different frequency with which these tests are meant to be administered. According to Exact, Cologuard is designed to be administered every three years and is expected to cost $400, per Canaccord Genuity's May 27, 2017 EXAS research report (see excerpt here). In contrast, the FIT is designed to be administered annually and costs $23 ($69 over three years) per the same report. This difference in frequency is incredibly important. The sensitivity and specificity of the Cologuard test should be compared to the sensitivity and specificity of the FIT administered each year for three years (the cumulative efficacy of the FIT). We know that even Exact Sciences believes it is important to consider the cumulative efficacy of a test, because they do so in their presentation on slide 20 as can be seen here:
When you normalize the FIT and Cologuard tests for their frequency, the FIT shows itself to be superior to Cologuard in nearly every aspect. Over a three-year time period, the FIT is better at finding cancer and misses fewer potential cases.
Below is the expected performance of the Cologuard test in a 10,000 patient group, identical to what the company used in its DeepC test.1 This assumes that the test is given once every three years, as suggested by the Company. Below it, we show the FIT being administered to those same patients every year for three years.
Even based on the data that the Company highlights, investors should clearly understand that Cologuard significantly underperforms FIT:
- FIT finds 61 cancerous tumors while Cologuard finds 59
- FIT finds 367 pre-cancerous lesions while Cologuard finds 316
- FIT misses 388 patients in need of a colonoscopy while Cologuard misses 441
- FIT results in 1,310 unnecessary colonoscopies while Cologuard results in 1,194
When compared on an apples-to-apples basis, the FIT outperforms Cologuard on the three most important metrics. The lone exception is unnecessary colonoscopies, where it is a mere 10% worse.
Exact Sciences Uses Old, Obsolete Data To Show A More Favorable Result For Cologuard
To make things worse, the Company chose data from an outdated FIT study that happened to be the single worst study we could find from the last 10 years. We identified 6 other trials that on average, inclusive of the trial that Exact used in its presentation (a total of 7 trials), shows that the FIT has 77% sensitivity to cancerous tumors, 39% sensitivity to pre-cancerous lesions, and 95% specificity.2 When these average representations of the FIT are used, the Cologuard test looks even worse in comparison:
Here is the expected performance of the FIT based on the average performance in the trials we identified:
Using this average data, we can now see that in all cases the FIT appears to be even more effective than originally thought:
- FIT finds 63 cancerous tumors while Cologuard finds 59
- FIT finds 579 pre-cancerous lesions while Cologuard finds 316
- FIT misses 174 patients in need of a colonoscopy while Cologuard misses 441
- FIT results in 1,277 unnecessary colonoscopies while Cologuard results in 1,194
In fact, after only two years of testing the FIT is already outperforming the Cologuard test:
- After two years, FIT finds 60 cancerous tumors while Cologuard finds 59
- After two years, FIT finds 469 pre-cancerous lesions while Cologuard finds 316
- After two years, FIT misses 286 patients in need of a colonoscopy while Cologuard misses 441
- After two years, FIT results in 872 unnecessary colonoscopies while Cologuard results in 1,194
Exact Sciences Ignores the 800lb Gorilla In The Room: The InSure FIT
Now we would like to introduce the InSure FIT. While the FIT is a good test, the InSure FIT is even better and is the current leader in take-home FIT testing. Importantly, because of the quantity of stool needed for Cologuard, it will be almost exclusively a take-home test - more on this in part 3. The InSure FIT has 88% sensitivity to cancerous polyps, 43% sensitivity to precancerous polyps, and 97% specificity, which can be seen here. A comparison between Cologuard and InSure would be as follows:
These are the results that we would expect from the InSure FIT using the identical 10,000 patient group:
Using the InSure FIT as the comparator, Cologuard underperforms every single end point:
- InSure finds 64 cancerous tumors while Cologuard finds 59
- InSure finds 610 pre-cancerous lesions while Cologuard finds 316
- InSure misses 142 patients in need of a colonoscopy while Cologuard misses 441
- InSure results in 905 unnecessary colonoscopies while Cologuard results in 1,194
In fact, the InSure FIT outperforms the Cologuard test only two years into the three-year testing cycle:
- After two years, InSure finds 63 cancerous tumors while Cologuard finds 59
- After two years, InSure finds 504 pre-cancerous lesions while Cologuard finds 316
- After two years, InSure misses 249 patients in need of a colonoscopy while Cologuard misses 441
- After two years, InSure results in 614 unnecessary colonoscopies while Cologuard results in 1,194
Exact Sciences Hid Data From Its Pivotal Trial
The mystery of the FIT comparison does not stop there. There are two other oddities regarding the Cologuard trial. First, it appears that Exact administered a FIT in its DeepC trial in order to compare the two tests in an identical patient population. Exact clearly states a FIT was administered as a comparison here. Exact then chose not to disclose the data generated in its trial and chose to instead use an outdated 2005 FIT study done in Japan by Morikawa as the comparator test. In our experience, this type of sleight of hand is performed when a company wants to distract investors from potentially troubling results. Further, in order to make Cologuard look better, the Company chose the FIT-CHEK test as the FIT comparator in the DeepC trial, even though FIT-CHEK only uses one fecal sample and is primarily administered in doctor's offices, which is a different setting from Cologuard's. If the Company had instead used the InSure FIT, which like Cologuard is a take-home test, it would probably have doomed Cologuard.
In addition to choosing an irrelevant FIT as the comparator for Cologuard, the Company also chose not to report the results of the FIT included with Cologuard. As can be seen on see slide 14 posted below, Cologuard actually includes a FIT, the results of which can be separated from the Cologuard test as a whole to see whether the expensive DNA testing actually adds any value above and beyond the FIT included with Cologuard. Bizarrely, we don't know which FIT is included with Cologuard, so the supposedly positive DeepC trial results may simply reflect a better FIT included with Cologuard. We would be shocked if these data are not available, and believe that their exclusion suggests there is insignificant value to the DNA testing, not to mention how little value it seems to add for an extra 17x the cost of the FIT alone.
Making Up Numbers To Make Cologuard Look Better
We now again call attention to slide 20, which contains the unusual language "Not based on clinical evidence." We find this a curious, weird, and misleading way to tout the efficacy of a test - tell everyone how great it will be over a 9-year period and then back away from your claims with the disclaimer that they're not based on clinical evidence. If the claims are not coming from clinical evidence, then where are they coming from? Hopes and dreams? Here is the relevant portion of the slide once again:
If You Thought That It Couldn't Get Any Worse…
We want to leave with one last telling sleight of hand in the information that Exact has been disseminating. We have already posted slide 18 from the June 2013 presentation above. We now highlight slide 16 from the May 2013 presentation, which can be found here. At first glance, it looks like the same slide. However, after taking a closer look, we can see that for the FIT statistics, Exact provided a range in May (with two sources). Now going back to slide 18 from the June presentation Exact drops the second source, which happens to be more recent and more favorable to the FIT, and instead simply reports the lower (and arguably outdated) end of that range. Here is a side-by-side comparison of the two slides:
We do not believe that the FDA is likely to be as easily fooled as investors. Instead, we suspect that they will ask the hard questions, such as which test is actually better, what do the true historical comparisons really show, and what did the Company's own FIT results show in the same population. We believe that as the true irrelevance of the Cologuard test comes to light, it has a high likelihood of being rejected by the FDA. In the unlikely event the test is approved, it will likely fail to gain adoption due to its poor predictive abilities. Further, we intend to submit a citizen's petition to the FDA outlining why Cologuard should not be approved.
More To Come
This article, the first in a five part series, began by illustrating that regardless of the set of data that you use, the Cologuard test is inferior to the FIT. Going forward, we are excited to continue our takedown of EXAS's lofty claims regarding its Cologuard test. Next we will discuss the economics of the Cologuard test, and why they don't support reimbursement. That will be followed by a detailed argument explaining why we feel that very few doctors will actually prescribe the Cologuard test even if it gets approved. Our fourth installment will deconstruct the bull case, and explain why the Cologuard test will never be the Pap Smear. Finally, we will dive further into the facts, and explain how Exact Sciences has misled investors (again). Stay tuned.
1. The test comparison makes the following assumptions:
- 10,000 patients over the age of 50 are tested
- Based on the company's DeepC trial, the expectations for positive tests for cancerous and precancerous polyps for this age group is approximately 0.64% and 7.52% respectively
- The Cologuard test is given once every three years, while the FIT is given every year
2. Below are links to FIT data that Exact Sciences failed to mention in its June 2013 Presentation:
- HemeSelect (Fujirebio, Inc., Tokyo, Japan), Hemoccult-ICT (Beckman Coulter, Inc., Primary Care Diagnostics, Los Angeles, CA), and Magstream (Tokyo, Japan)
- American Cancer Society (InSure Data used in Comparison)
- Graeme P. Young
- Wjikerslooth-found in May 2013 presentation above