This past Sunday, a group of researchers reported in the journal, Nature Medicine, a preliminary technique that uses variation in blood levels of 10 fats to predict the likelihood that elderly individuals would develop mild cognitive impairment (MCI) or Alzheimer’s Disease in the following 2-3 years. The sample size was small and the results may not generalize beyond the narrow age-range and demographics of the study group (i.e. the assay is far from ready for “prime time”), but the study is an important first step towards a lower cost (vs PET imaging) and less invasive (vs spinal tap) predictive biomarker of cognitive decline*. Its publication has also triggered a flurry of discussion on possible ethical ramifications of this sort of blood biomarker. I will not attempt to address these ethical issues specifically here. Rather, I seek to highlight that how ethically troubling one views the technology to be may depend partly on the sort of knowledge one thinks these biomarkers reveal (applied epistemology at its best).
One way to understand these sorts of biomarkers are as diagnostics of Alzheimer’s Disease (or MCI) before any symptoms appear. A reaction that follows quite directly from this thinking is illustrated by this commenter on NPR:
“Are you aware that a test that is wrong ten percent of the time means that 100’s of thousands of people would be misdiagnosed with a horrible life ending illness? Since we don’t understand the pathophysiology of Alzheimer, this test is like many other indirect, “statistical” tests, are not something to base life changing decisions on.”
Viewed as a diagnostic, this particular set of fatty biomarkers has 90% specificity and 90% sensitivity, which means that 9 out of 10 people who “test positive” will develop cognitive impairment and only 1 out of 10 people who “test negative” will develop cognitive impairment (for a useful calculator of specificity and sensitivity, go here). One might decry, as this commenter did, the harm to those who would test positive but would remain cognitively intact or lament for those would test negative but are actually poised on the edge of cognitive decline.
But a second (and unfortunately less common) way of understanding this type of technology has been promoted by Jason Karlawish through an analogy with cardiovascular disease: that is, indicating a state of sufficiently high risk of complication to merit clinical attention. It is not about false positives and negatives of that complication, but of a state of altered biology that we care about because of its relation to those complications. I like to think about this second epistemological stance as understanding biomarkers as warnings of risky situations. Knowing that a hurricane is coming into town, for example, is not the same as knowing that my home will be destroyed (some houses, after all, will avoid damage), but the news is still serious.
But then again, the weatherman is not claiming to provide a diagnostic test for future home destruction. And if we would object on moral grounds to his doing so, perhaps we should move away from terms like preclinical Alzheimer’s disease.
* note that some news sources, 1, 2, report the Nature Medicine study as a blood test for Alzheimer’s disease, but since the method did not distinguish between MCI and Alzheimer’s, the authors’ own choice of “memory impairment” is more accurate