Built to Say No

Every headline about Alzheimer's blood tests is about catching the disease. The family in the waiting room is usually asking a different question. Mom forgot the stove twice, Dad repeated a story at dinner, and what they want to know at two in the morning is not the sensitivity of an assay. It is whether they can stop bracing.

The strongest numbers in this research are quietly the negative ones. In the landmark Swedish work published in JAMA, the blood test's negative predictive value ran 87 to 92 percent across care settings, meaning roughly nine times in ten, a no is a no a family can believe. A newer test described in Alzheimer's & Dementia goes further and is built for exactly this, tuned to a negative predictive value of about 95 percent while accepting weaker sensitivity. It is engineered to say no reliably, at the cost of sometimes missing a yes. The engineering choice is the tell. Someone finally designed for the most common person in the clinic, the one who probably does not have it.

The honesty that has to ride along. A no means probably not Alzheimer's, it does not mean nothing is wrong, and the doctor who noticed something still owns that something. These tests are for people with symptoms, not for the worried well. And in the US, Medicare currently calls them screening and does not pay, so believing a no costs a family fifty to a hundred dollars out of pocket.

Worldwide, dementia is headed from 55 million people toward 139 million by 2050, which means the waiting rooms fill in every language, and most of the people in them will not have it. Months of a family's dread, resolved at the kitchen table for the price of a dinner. Almost nobody is writing that story, and it might be the best one this technology has.