“Colonoscopy? How About a Blood Test?”
by Paul Tullis
Bloomberg Businessweek, January 11, 2018
Predictive Medicine Is Finally Getting Real
Colleagues call him “The Brain.” He served in an elite unit of the Israeli Defense Forces’ Intelligence Corps, developing classified surveillance software. Later Nir Kalkstein was on the faculty of electrical engineering at his alma mater, Technion-Israeli Institute of Technology. From there, a couple of start-ups. After one of them fell apart, he and a former classmate from Technion created a company to use algorithms to trade in the financial markets. Final (Financial Algorithms) performed tens of thousands of derivatives trades a day and ended every trading day with a profit, over more than 13 years, generating a return for Kalkstein and his partners of 100,000%.
One of them bought a $30 million Manhattan penthouse, but Kalkstein, now 44, remains low-key. He’s not happy having enriched 30 shareholders to stratospheric levels of wealth–he wanted to create value for the public. Along with an organizational consultant at Final and a mathematician he knew, Kalkstein began to think of how to apply algorithms to medicine. Thus was launched Medial Research.
Big data and clever algorithms are finally putting us on the cusp of something public health people have been dreaming of and talking about for a decade or so: moving from reactive to predictive health care. “This is a pretty active area of research now, and will continue to be,” says Jeff Elton, managing director of Predictive Health Intelligence, a division of Accenture. Instead of diagnosing what you have, we’re getting to the point of telling people which conditions or diseases they’re on the road to developing, or are likely to be found with more invasive testing. Genetic testing is one element of it; Medial is at the forefront of the field of related field of cytomics. Think of it as CompStat for doctors: Just as police look at crime figures and other data to decide where to deploy resources, by looking for a variety of biomarkers and running the results through powerful analytic engines, clinicians can recommend preventive measures (“eat more broccoli”) or more invasive tests.
Kirkstein uses methods derived from information theory, signal processing, and machine learning to develop an algorithm he says can predict when a patient in intensive care or the emergency department is going to “crash,” giving medical staff precious seconds to intervene. The company is also working on mathematical predictors of seizures–wearable sensors will give epileptics an alarm when a seizure is likely to strike, allowing them to pull over to the side of the road or get out of the bathtub before they become disabled.
In the here and now, the company has clinically validated a blood test for colorectal cancer, the third most-diagnosed cancer in men and second-most in women, worldwide. Staff has published two peer-reviewed retrospective studies on its work. CRC is identified now by one of the more invasive diagnostic procedures in medicine. The Medi-Al test works not by looking for something particular in the blood, but by running 20 parameters in the sample through decision trees and cross-validation techniques to compare it with a database of millions of de-identified blood samples (some from patients known to have CRC) to give a statistical likelihood of the disease. A doctor then recommends a colonoscopy, or not. The studies showed few false positives or negatives, and a pilot program flagged 38% of patients who had the screen for colonoscopy; 100 of these 111 had precancerous lesions and 11 had active cancer. None had had the diagnostic tests in use for CRC. Similar test-runs with samples from Poland, the UK, and the US–where patients have different nutrition, genetics, and environments–yielded similar results.
“What Medial gives you is a data-based, analytically powerful approach to bring everything we have available in medicine to bear in order to handle much more effectively–or prevent–downstream complications of an illness,” said David M. Lawrence MD MPH, former CEO of Kaiser Health Plan and Hospitals. “It’s a tremendous tool for the health care system. I think we’re just beginning to see how powerfully Medial can apply these tools to healthcare, because we’re just at the beginning of this revolution.”
“More non-invasive ways of screening for colorectal cancer are needed,” says Dr. Matthew Kalady of the Cleveland Clinic’s dept. of colorectal surgery. “There’s a significant population that isn’t getting screened because of the stigma of colonoscopy. It’s one of those cancers that’s curable if you get it early, so identifying it is a big deal.” Medial’s screen, he said, “can make an impact because anything you can do to increase the population that gets screened is important.” Accenture’s Elton calls the CRC screen “significant” and “quite clever.”
Medial last week inked a deal with Kaiser to use Medial’s technology and its database of blood samples to screen patients in its health plan, and the company is working on similar tests for other cancers typically only diagnosed late-stage, and to develop algorithms to predict which pre-diabetic patients will become diabetic within 12 months, and which diabetics will experience complications such as nephropathy and cardiovascular disease (currently, no such standards or guidelines exist). The company recently received a $20m investment from Hong Kong billionaire Li Ka-shing’s Horizon Ventures; Kalkstein has pledged to re-invest any profits on his own investment.
“A doctor tells you what you have,” says Medial’s chairman, Ofer Ariely. “We tell you what you are likely to get in six months, two years, or five years.”
In 2500-3500 words I can tell Medi-Al’s story, drawing on visits to its labs and offices and interviews with its founders, chief medical officer, and senior clinical and bioinformatics staff, from conception through the development of its algorithms through its latest deal with Kaiser. Kalkstein is a bit of a recluse, and has only given one interview (in Hebrew), but he has agreed to an interview with me and Businessweek.