Facebook YouTube Twitter Linkedin
Facebook YouTube Twitter Linkedin


5 December 2018

Normative data from linear and non-linear quantile regression in CANTAB: cognition in mid-to-late life in an epidemiological sample

Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring recently published the first study to derive normative CANTAB data from quantile regressions of cognitive test performance for a large epidemiological sample of adults aged 57-84 years.

The therapeutic aim of prevention trials in Alzheimer’s Dementia is to curtail disease progression in its early stages. Within this model of drug development, it is essential to separate out pathological decline from typical ageing to assess if preventative treatment is truly effective. With this in mind, a team of scientists from Cambridge Cognition, the University Hospital of Essen and Janssen recently published normative cognitive data from a large epidemiological sample of adults in mid-late-life in the journal Alzheimer’s & Dementia: Diagnosis, Assessment & Disease Monitoring.


What is normative data?

Normative data allows the comparison of the cognitive performance of an individual person relative to a large group of people of similar age, sex and education level.  This comparison allows us to determine whether an individual’s cognitive function is average, exceptional or impaired at a particular time-point in their lives. Re-examining performance over time can help to reveal if a person is showing changes in keeping with those expected during healthy ageing, or if they show a cognitive decline that is outside the norm. Sensitive normative data can help to improve detection of impairment and cognitive decline.


Which methods did the study use?

Regression based approaches allow a greater resolution of normative data, describing expected declines per year of age, which is not possible in more traditional approaches where data is stratified by defined age-ranges. Non-linear quantile regression modelling can fit discrete slopes to identify different trajectories of cognitive ageing across performance ranges.  

The study used linear and non-linear quantile regression modelling to derive normative cognitive data from a population sample of 1535 healthy adults, age 57-84 years, who were tested with CANTAB Paired Associates Learning (PAL), Spatial Working Memory (SWM), and Reaction Time (RTI) tasks as part of their participation in the Heinz Nixdorf Recall study.


What were the main study findings?

CANTAB measures were sensitive to the decline in cognitive performance seen with typical ageing i.e. increased errors on working and episodic memory tasks, and increased reaction times. Furthermore, the rates of age-related decline were heterogeneous between performance bands. Non-linear quantile regression was sensitive to this heterogeneity, providing more representative normative thresholds for impairment. Indeed, impaired PAL performance  was particularly well differentiated by the non-linear quantile regression approach.


What are the implications?

Previous research has shown that PAL performance is not only sensitive to the staging of Alzheimer’s Dementia, but also early physiological biomarkers of the disease. Therefore, the development of this sensitive normative data for older adults’ PAL performance holds promise for disambiguating normative decline from the early signs of clinically-meaningful cognitive impairment.


We were excited to be able to apply non-linear quantile regression methods to CANTAB data from a large epidemiological sample. These methods are more sensitive than conventional normative statistics as they allow for modelling of different trajectories of age-related decline. We believe that a more sophisticated understanding of normal ageing will enable better methods for detecting clinically-relevant memory impairment.

Corresponding Author, Dr Francesca Cormack 


Where can I access the publication?

“Normative data from linear and non-linear quantile regression in CANTAB: cognition in mid-to-late life in an epidemiological sample” by Rosemary A. Abbott, Caroline Skirrow, Martha Jokisch, Maarten Timmers, Johannes Streffer, Luc van Nueten, Michael Krams, Angela Winkler, Noreen Pundt, Pradeep J. Nathan, Philippa Rock, Francesca K. Cormack, Christian Weimar (DOI: 10.1016/j.dadm.2018.10.007) appears in Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoringpublished open access by Elsevier.

The publication is freely available for download, just follow the DOI link: DOI: 10.1016/j.dadm.2018.10.007


About Alzheimer’s & Dementia: Diagnosis, Assessment & Disease Monitoring

The mission of Alzheimer's & Dementia: Journal of the Alzheimer's Association is to bridge the knowledge gaps across a wide range of bench-to-bedside investigation. The journal publishes the results of studies in: behavior, biochemistry, genetics, molecular biology, pharmacology, physiology, protein chemistry, neurology, neuropathology, psychiatry, geriatrics, neuropsychology, epidemiology, sociology, health services research, health economics, political science and public policy. Content emphasizes interdisciplinary investigations, integrative/translational articles, related to: etiology, risk factors, early detection, disease modifying interventions, prevention of dementia and applications of new technologies in health services.


Find out more 

Tags : normative data | regression | epidemiology