There are lies, damned lies and statistics

Alzheimer’s disease (AD) is a horrible illness. A progressive neurodegenerative disorder that destroys memory, abstract thinking and cognitive function, it’s the most common cause of dementia in humans, affecting as many as five million Americans every year and being the sixth leading cause of death. The risk of getting affected by AD doubles every five years after the age of 65, meaning that with 10,000 baby boomers turning 65 every day, the number of Americans with AD may reach 14 million by 2050. The havoc AD wreaks on the life of affected people is enormous, and because AD patients require 24/7 care, it also profoundly changes the lives of their families. More to that, AD is damned expensive, adding more than $220 billion to the U.S. healthcare bill in 2013 alone.

There is no real cure for AD. Five drugs that were approved in the U.S. for AD treatment (the last in 2004) can only offer a brief respite from some symptoms in some people. Even more worrisome, the development of new AD therapies has so far been a total disaster: over the period of 2002-2012, 244 drug candidates have been assessed in 413 clinical trials, but only one has been approved by the FDA. This is a 0.4% success rate or, if you prefer, a 99.6% failure.

The main problem is AD diagnosis: similar to cancer, AD is often detected only after it has already done an irreparable damage to the patient. For this reason, many experts believe that the key to finding effective AD cure is in identifying reliable biomarkers, molecules that could signal the imminent onset of the disease before the pathological symptoms of it became evident. Obviously, to be useful in clinical setting, such molecules should be present in easily available body fluids, such as blood. Not surprisingly therefore, the quest for a “blood test for Alzheimer’s disease” has become one of the holy grails of the medical field.

This explains the close attention to a recent article published in the journal “Alzheimer’s & Dementia.” A team of 27 researchers claim having identified 10 plasma proteins that could predict the disease progression from a pre-AD condition to a full-blown AD within a year of blood sampling. Promptly, BBC jumped in and declared the study a “major step” towards AD blood test.

It now appears that the jubilation has been somewhat premature. DrugBaron, an influential blog covering biopharmaceutical industry, has posted a piece that tears the above study apart. Having pointed out to some minor deficiencies in the study design, DrugBaron focuses its criticism on the statistical treatment of the study results, arguing that the authors’ use of multivariate statistical analysis is questionable at best and outright wrong at worst. The piece’s conclusion is that the proposed cohort of protein biomarkers has no predictive power whatsoever.

Given the complexity of multivariate statistics, DrugBaron doesn’t blame the BBC or other news organizations for running “breakthrough” stories based on shaky science. But it reserves harsh words for the authors of the study, “Alzheimer’s & Dementia” reviewers and “AD experts” interviewed by the BBC, who should have known better and yet chose to ignore obvious deficiencies of the study.

As for those waiting for a “blood test for Alzheimer’s disease,” the wait still continues. And the wait for AD cure continues too.