Comparing neuropsychological test results against age-adjusted norms is one way of determining whether an individual might be facing impending dementia. However, “normal” values can vary considerably, even after accounting for variables such as age and education. An alternative, and possibly more sensitive, method is to look for within-person inconsistencies across a range of neuropsychological tests. In yesterday’s JAMA, Roee Holtzer and colleagues from the Albert Einstein College of Medicine, Yeshiva University, New York, report that just such an approach can predict future dementia, even after a single baseline session of testing. “That is a fairly novel finding,” said Holtzer in an interview with Alzforum, adding that it has the potential to improve current diagnostic tests. “The advantage is that this strategy simply complements existing methods and can be used without further taxing the patient and the clinician because it doesn’t require any additional measures and it doesn’t require any additional time,” he said.

The rationale behind the idea is that different tests have different sensitivities to age-related disease, resulting in greater inconsistency or variability across test outcomes in affected subjects. In dementia, for example, memory is particularly vulnerable. “If there is a decrement in memory but other cognitive abilities that are not as sensitive to age-related changes remain fairly intact, then by definition you would also expect the [across test] variability within that person to increase,” explained Holtzer.

To test if this is true, Holtzer and colleagues looked at test variability in 897 patients who were enrolled in the Einstein Aging Study, a population-based longitudinal study of older adults (aged 73 or older) in Bronx County, New York. The researchers looked at baseline scores on three neuropsychological tests—the Free and Cued Selective Reminding Test, and the Digit Symbol Substitution and Vocabulary subtests of the Wechsler Adult Intelligence Scale Revised. Holtzer said those tests were carefully chosen because they are commonly used, were available for all patients, and because they test three different domains—memory, attention/executive function, and verbal IQ, the last not being sensitive to the effects of aging and age-related diseases.

The researchers calculated the within-person, across-test variability for each subject and then compared the results with rates of incident dementia (all patients in the Einstein Aging Study are followed up with every 12 to 18 months). The number of incident dementia cases (12 percent) in the quartile with the highest across-test variability was significantly higher than the other three quartiles. For example, only 1 percent (three out of 225 subjects) of people in the lowest quartile developed dementia. Importantly, the results were still significant even after correcting for the performance level on each test alone. “That is very important because without that particular analysis one could argue, and rightly so, that the variability we’ve calculated and demonstrated to be predictive of future dementia is mediated entirely by simply poor memory scores or by poor scores on measures of attention and executive function. We have controlled for that and we have shown that above and beyond performance level on individual tests, the degree of within-person variability remained a significant predictor of future dementia.”

As the authors point out, this study needs to be replicated in an independent population, but should it stand up to that test, it may help researchers detect those earliest changes that are harbingers of future dementia. “If we really want to detect the earliest declines in performance in people who are on the path to dementia, we need to focus on intra-individual decline, and that is exactly what this group has done in a very well carried out study,” said John Morris, Washington University School of Medicine, St. Louis.

Within-person, across-test variability might also fluctuate. “It will be interesting to look at changes in variability over time to see how those are also related to prediction of dementia,” said Holtzer. It might also be attractive to those looking for early changes in randomized clinical trials, given that the method does not require any additional testing beyond that already incorporated in trial design. “In trials we want measures that are sensitive to change and intervention, and in most trials we want changes that are economic. The advantage of this method of within-person variability is that you don’t need to augment an already standardized neuropsychological assessment. You can use values from existing tests and simply calculate the variability within persons. Potentially it can be used within trials as a measure sensitive to change over time, or intervention, or both,” said Holtzer.—Tom Fagan.

Holtzer R, Verghese J, Wang C, Hall CB, Lipton RB. Within-person across-neuropsychological test variability and incident dementia. JAMA. 2008 August 20;300:823-830. Abstract


Make a Comment

To make a comment you must login or register.

Comments on this content

No Available Comments


Paper Citations

  1. . Within-person across-neuropsychological test variability and incident dementia. JAMA. 2008 Aug 20;300(7):823-30. PubMed.

Further Reading


  1. . Within-person across-neuropsychological test variability and incident dementia. JAMA. 2008 Aug 20;300(7):823-30. PubMed.

Primary Papers

  1. . Within-person across-neuropsychological test variability and incident dementia. JAMA. 2008 Aug 20;300(7):823-30. PubMed.