It would seem difficult to put up worse numbers than experimental Alzheimer’s drugs, 99 percent of which have failed in clinical trials since 2002. But another corner of Alzheimer’s research has managed it: blood tests to either diagnose the disease in asymptomatic patients or predict which healthy people will develop it years in the future. Although you wouldn’t know it from frequent headlines proclaiming, “Blood test can predict Alzheimer’s,” the percentage of tests that looked promising in a (usually small) study but eventually fell flat is … 100 percent.
Despite the two dozen such failures, scientists aren’t giving up. On Wednesday, researchers in Europe and Australia reported in the journal Science Advances that blood levels of 10 proteins did a pretty good job of identifying which cognitively unimpaired people had high enough brain levels of beta-amyloid, a marker of the disease, to be classified as having preclinical (meaning, without symptoms) Alzheimer’s. The test isn’t accurate enough to make diagnoses as part of medical care, its creators say, but if it’s validated in additional studies, it could give drug companies a desperately needed tool: a cheap, easy way to identify preclinical Alzheimer’s.
Preclinical patients have become as valuable to Alzheimer’s drug developers as glue-fingered tight ends are to football teams. That’s because more and more experts blame the failure of drug after drug on poor timing: Clinical trials included patients whose Alzheimer’s had progressed too far, their brain neurons killed and their synapses destroyed, to be helped by drugs that remove the molecules, such as amyloid, believed to be responsible for that carnage. Blowing out the match doesn’t help if the house is already ablaze.
About the Author Reprints
Send me more information.]
Bravo! It’s so refreshing to have the statistical aspects of the test, including PPV, NPV, and sample size and makeup limitations, clearly explained. The world would be a better place if every article about medical research “breakthroughs” followed this example. Thank you!