Pichet_W/ Getty Images
In large, long-running studies of older adults, such as the Adult Changes in Thought (ACT) Study (participants = 5,546) and the Rush Memory and Aging Project (participants = 2,163), participants attend regular research visits over time and undergo different procedures including tests of thinking and memory. These studies generate large amounts of data that researchers can use to find associations between lifestyle or genetic factors and late-life cognitive functioning and impairment.
However, the specific types of memory and thinking tests that researchers used tend to vary from study to study. This lack of standardization across studies makes it difficult for researchers to compare individual results when using data pooled from multiple studies.
In what Paul Crane, MD, MPH, professor of medicine at the UW School of Medicine, calls a ‘Hello World” methods paper, a UW team shares the results of a major effort to make it easier for researchers to leverage data from multiple studies of older adults. They have succeeded in taking the cognitive test scores from 10 different studies and calibrating them according to the same metric. “What that means,” says Crane, “is that the scores from each study are now directly comparable to one another; they mean the same thing about a participant’s performance." So, if a person in Study A had been evaluated and studied in Study B, this person would be expected to have the same score.
These calibrated test scores for 10 different study datasets have many benefits to the Alzheimer’s field. A researcher can focus on their important scientific question without having to spend as much valuable time manipulating statistics. “Our science is really benefiting a lot from the recent moves towards data sharing and data availability,’ says Crane. “Our group is doing the work to make it so that those shared data are amenable to quick scientific analyses of relationships between cognition and whatever else is of interest, across multiple different datasets—and without anyone else needing to worry about the details that we worry about.”
The team is currently working to generate these co-calibrated memory scores for datasets from the UW ADRC Clinical Core, to make their participants’ data even more easily accessible to other researchers across the country.
This effort is part of the NIH-funded Alzheimer’s Disease Sequencing Project Phenotype Harmonization Consortium, a collaboration of four institutions attempting to pool data from 30+ research cohorts using data harmonization methods. The types of data include cognitive performance, genetics, neuroimaging, biomarker data, and autopsy neuropathology data. The goal of the Consortium is to produce a large-scale, racially diverse, standardized set of transparently defined data that will open new windows into the genetic basis of Alzheimer’s disease and resiliency.
Crane, who co-leads the cognitive data harmonization arm of this NIH Consortium, thinks that this larger data harmonization approach will ultimately lead to more confidence in scientific results in the Alzheimer’s field because researchers can check if an association found in one study can be confirmed in a different dataset. •
Paper:
Mukherjee S, Choi SE, Lee ML, Scollard P, Trittschuh EH, Mez J, Saykin AJ, Gibbons LE, Sanders RE, Zaman AF, Teylan MA, Kukull WA, Barnes LL, Bennett DA, Lacroix AZ, Larson EB, Cuccaro M, Mercado S, Dumitrescu L, Hohman TJ, Crane PK. Cognitive domain harmonization and cocalibration in studies of older adults. Neuropsychology. 2023 May. PMC9898463.
This research was supported in part by NIA grant AG066509.
