A Good News Story*

It’s not often that a story contains both good news and an asterisk. This article, about patients with life-threatening familial hypercholesterolemia (FH), is one such story.

The sad truth is that over 90% of the estimated 1.3 million patients in the US with the genetic disease do not know they have it. Often the first sign is a fatal heart attack; sometimes it is quadruple bypass surgery in a person in only the fourth decade of life. The FH Foundation, a patient-led charity, was founded to address these critical problems. Joshua Knowles, MD, PhD (assistant professor, Cardiovascular Medicine), serves as the Chief Medical Advisor for the FH Foundation, which is a major driving force behind a project funded by the American Heart Association, the Stanford Data Science Initiative, and Amgen that aims to identify patients with FH. The project is being led at Stanford by Knowles and Nigam Shah, MBBS, PhD (assistant professor, Biomedical Informatics).

This project combined the skills of Knowles and Shah to create an algorithm capable of scanning electronic medical records (EMRs) and picking out FH patients. The computer “learns” what an FH patient looks like by being shown examples of true positive patients. Then it picks out other patients with similar “patterns” in the EMR. Knowles explains: “We can scan all types of data in the EMR (lab results, clinic notes, text, etc.), which is in itself exciting. Because we don’t know the features of FH that it will identify as important, we also get insights into the disease process. Some will make a lot of sense (like LDL levels) while others will be head scratchers.”

Thus far, the algorithm performs very well. According to Shah: “The preliminary algorithm works; there’s no doubt about that. Now it’s a matter of improving it, validating it, and figuring out where we use it.”

Here the asterisk appears.

“We know that we can design an algorithm that can find most of the patients who have FH,” says Shah; “the problem is with our tolerance for false positives, patients identified as possibly having FH who do not have FH. If I label somebody incorrectly, how much testing, physician visits, money,and energy are we going to waste? We also don’t know physicians’ and patients’ tolerance level for a false positive diagnosis. These are the key issues we are working on now.”

“In an ideal world the algorithm would be perfect,” Shah continues, “but in the real world there are important trade-offs that need to be weighed. We hope that through a process of iteration with internal and external validation the algorithm will identify most FH patients while keeping false positives to acceptable levels.”

On the bright side, much progress has been made. Patients are being identified, beginning treatment, and entering a registry to follow them henceforth. Knowles comments: “The FH Foundation established a national patient registry called CASCADE FH in which Stanford is a leading participant. The registry is going like gangbusters, with over 2500 patients enrolled so far. An initial manuscript detailing the findings from the first 1400 people was recently submitted. The data are really eye opening.”

Knowles explains further: “Most people are not diagnosed until their mid 40s; by that time a high percentage already have established coronary disease, so the horse is out of the barn. Even after being treated at leading lipid clinics, most people will have an LDL of about 140 mg/dl, much higher than we would like.”

With the help of such compelling data, the FDA approved two drugs from a new class of cholesterol-lowering medications called PCSK9 inhibitors in August 2015. They are specifically targeted at patients with familial hypercholesterolemia, and that is a major step forward. Without an asterisk.