Exact Sciences braces for fallout from Cologuard’s precancer miss


Long-awaited data from the Deep-C study of Exact Sciences’ colorectal cancer diagnostic Cologuard have delivered an unexpectedly low result against precancerous lesions – a measure the company had earlier touted as a key selling point.

All key metrics worsened compared with the most recent trials, although most can be explained by the prospective design of Deep-C – for various reasons prospective studies tend to give worse results than those done on a case-control basis (see tables below). But the very poor sensitivity for precancers sticks out like a sore thumb, and could seriously undermine Cologuard’s place in the market.

Not unexpectedly the stock market did not take kindly to the news. Exact had been up over 10% in the pre-market, but by the time topline results from Deep-C were disclosed today the stock opened nursing a crash of over 30%.

10,000 patients

Deep-C was one of the largest prospectively designed clinical studies ever conducted in colorectal cancer screening, and final analysis comprised 10,000 patients, of whom 64 turned out to have colorectal cancer and 752 to have precancerous lesions (Event – Exact Sciences goes deep with Cologuard, November 26, 2012).

In the key diagnostic measures of sensitivity – the percentage of actual cases correctly identified – and specificity – negatives correctly given the all-clear – Cologuard did slightly worse than in earlier case-control studies. But efficacy across cancerous lesions did meet targets, and Cologuard was non-inferior to the currently used faecal immunochemical testing (FIT), a cheap but fairly unreliable diagnostic.

Comparison of Deep-C to earlier studies
Study Cancer sensitivity Precancer sensitivity Specificity
Case-control validation (2010) 85% 54% 88%
Case-control validation (2011) 98% 59% 91%
Case-control validation (2012) 98% 57% 90%
Prospective Deep-C (2013) 92% 42% 87%

Still, detecting just 42% of precancers is a big setback for Exact. Management had earlier touted Cologuard’s ability to detect precancerous lesions as an important advantage leading to timely intervention, and had set at least 50% as the meaningful target to hit, based on what it had seen in earlier studies.

Before the data were released analysts at Wedbush speculated that precancer sensitivity of 65% would represent a home run, while less than 50% was a “meaningfully negative scenario”. While Exact tried to gloss over its miss, saying it would “not make much of a difference clinically” and instead highlighting Cologuard’s 66% sensitivity for bigger precancerous polyps, the market clearly remained unconvinced.

It can also be argued that overall specificity slipping below 90% is a setback to Cologuard’s reliability; given the relatively low incidence of colorectal cancer this equates to over 20 falsely identified negatives for every actual cancer picked up, although US guidelines on non-invasive screening set an 85-90% specificity range.

Important or not?

Why are such apparent minutiae important if overall efficacy is good? In a word, pricing – an issue on which Exact insists its thinking has not altered despite the poor Deep-C data.

Cologuard is set to be priced at $200-400 per test versus FIT’s cost of just $30. FIT might be notoriously inaccurate but, at a time when healthcare budgets are stretched, to justify a tenfold difference Cologuard will need to blow FIT away in efficacy terms, or provide some other key differentiating factor.

In addition to this delicate pharmacoeconomic argument Cologuard’s ease of use – the test involves the patient passing a stool into a special pot that is then mailed to a lab – has yet to be proved fully. Exact plans to submit a US PMA next month, and expects Cologuard to face a US advisory panel sometime later this year before full Deep-C data are published in December.

The filing plans should not come as a surprise; Deep-C results certainly warrant this, and they are probably sufficient for approval. But Cologuard’s ability to shine in a real-world setting now looks far less certain than it did yesterday.

Study Detail Trial ID
Deep-C 10,000 subjects, prospectively screened for colorectal cancer NCT01397747

To contact the writer of this story email Jacob Plieth in London atjacobp@epvantage.comor follow@JacobEPVantageon Twitter

Share This Article