Linkage 8/13: The Headache of Brain Testing

26638mediumMany neurological disorders struggle with the same problem as their cousins, the psychiatric disorders: a fuzziness of diagnosis. Even well-known diseases such as Parkinson’s and Alzheimer’s are tricky for physicians to diagnose, since their hallmark symptoms (dementia, or movement issues) show up late and can reflect any number of conditions with different treatment strategies. Meanwhile, the subjective elements of diagnosis for a disease like autism can produce even more confusion, particularly in parents wanting the best care for their child.

The hope is that, someday, diagnosing Alzheimer’s or autism will be as simple as running a quick test that says Yes or No. Two studies this week produced excitement that physicians are nearer to such a goal, but also revealed how difficult it will be to obtain such a definitive answer. The coverage of the papers also revealed the importance of carefully reading the results of testing the test, knowing the difference between the testing terminology of “sensitivity” and “specificity.”

The most glaring example of this confusion was over a paper announcing a new biomarker test for Alzheimer’s Disease in the journal Archives of Neurology. Here, a group of scientists used an interesting technique to find the best way to predict Alzheimer’s disease from a spinal tap test, working from a data set of hundreds of patients with Alzheimer’s, mild cognitive impairment, or no dementia. With an unbiased analysis, the researchers determined thresholds for two markers involved in the pathology of Alzheimer’s (beta-amyloid and tau protein) that successfully predicted a diagnosis of Alzheimer’s from the spinal tap alone. This “sensitivity,” the ability to make the correct diagnosis in someone who has the disease and avoid false negatives, was reported at impressive levels of 94 to 100 percent.

However, news reports mistakenly said that the new biomarkers were a “100 percent accurate” test for Alzheimer’s. As many blogs have already pointed out, that’s just not, well, accurate, and in fact obscures the most interesting part of the study. As reported by the authors, the biomarkers misidentified about one-third of “control” subjects, that is patients who showed no signs of dementia or impairment, as being positive for Alzheimer’s. That’s a pretty low “specificity,” since it means the test generates significant false positives. But are they really false? The authors also report preliminary results that “there was a tendency for more progression to MCI in cognitively normal subjects with the AD feature,” meaning that people who were normal at baseline, but showed biomarkers for Alzheimer’s, may be on the cusp of developing symptoms of the disease. The “wrong” answers might just be the most useful answers, drawing attention to people on the verge of developing Alzheimer’s, a population that may be helped more by early treatment.

Another neurological condition where an early, clear diagnosis would be very helpful is autism. The current psychiatric means of diagnosing autism may be the main driver of its steadily increasing rates, a fact which has caused some to wonder whether those guidelines are specific enough, and whether autism is even one single disease at all. Rather than basing the diagnosis on interviews and psychiatric assessments, some are working towards genetic or imaging techniques that can give more definitive answers on autism.

There’s hope in this week’s Journal of Neuroscience, which features a paper from King’s College in London about using MRI images to diagnose autism. The test is not simple; the title of the paper is “Describing the Brain in Five Dimensions,” which gives you some idea of the complicated math behind the diagnosis process. But like the Alzheimer’s study, it shows promise, as using parameters like cortical thickness and “sulcal depth” (the space between the wrinkles of the brain) correctly diagnosed 90 percent of autistic adults. Once again, sensitivity vs. specificity is important to understand: the 90 percent sensitivity was accompanied by 80 percent specificity, meaning 20 percent of control subjects were mistakenly put in the autistic group. Even more importantly, the study was done in adults, not children, where most of the interest in autism diagnosis currently lies. More work will have to be done to prove the MRI test’s worth in younger patients, the authors admit.

Both tests are thus flawed, but show a potential for accuracy undreamed of even a decade ago for such difficult to diagnose diseases. Both papers (and accompanying editorials) argue that these tests will be supplements to traditional means of diagnosis, not a replacement, but for a field that relies heavier than most areas of medicine on judgment calls and subjectivity, they will be useful tools indeed. But even tests need tests, so proceed with caution.


The Discover blog NCBI ROFL usually finds papers worthy of mockery in the vast scientific library of PubMed, but I actually kind of liked this one: why baseball players with names that start with K are more likely to strike out.

A side effect of the amazing success in treating HIV/AIDS over the past decade is that such patients are now living long enough to die of other diseases, such as stroke or heart failure. A study on these new concerns for patients and physicians who deal with HIV/AIDS was recently presented at a conference in Vienna by University of Chicago professor of medicine Renslow Sherer, and was written up by the Chicago Tribune.

About Rob Mitchum (525 Articles)
Rob Mitchum is communications manager at the Computation Institute, a joint initiative between The University of Chicago and Argonne National Laboratory.
%d bloggers like this: