Research Article: Predicting cognitive resilience from midlife lifestyle and multi-modal MRI: A 30-year prospective cohort study

Date Published: February 19, 2019

Publisher: Public Library of Science

Author(s): Anya Topiwala, Sana Suri, Charlotte Allan, Vyara Valkanova, Nicola Filippini, Claire E. Sexton, Verena Heise, Enikő Zsoldos, Abda Mahmood, Archana Singh-Manoux, Clare E. Mackay, Mika Kivimäki, Klaus P. Ebmeier, Stephen D. Ginsberg.


There is significant heterogeneity in the clinical expression of structural brain abnormalities, including Alzheimer’s disease biomarkers. Some individuals preserve their memory despite the presence of risk factors or pathological brain changes, indicating resilience. We aimed to test whether resilient individuals could be distinguished from those who develop cognitive impairment, using sociodemographic variables and neuroimaging.

We included 550 older adults participating in the Whitehall II study with longitudinal data, cognitive test results, and multi-modal MRI. Hippocampal atrophy was defined as Scheltens Scores >0. Resilient individuals (n = 184) were defined by high cognitive performance despite hippocampal atrophy (HA). Non-resilient participants (n = 133) were defined by low cognitive performance (≥1.5 standard deviations (S.D.) below the group mean) in the presence of HA. Dynamic and static exposures were evaluated for their ability to predict later resilience status using multivariable logistic regression. In a brain-wide analysis we tested for group differences in the integrity of white matter (structural connectivity) and resting-state networks (functional connectivity).

Younger age (OR: 0.87, 95% CI: 0.83 to 0.92, p<0.001), higher premorbid FSIQ (OR: 1.06, 95% CI: 1.03 to 1.10, p<0.0001) and social class (OR 1 vs. 3: 4.99, 95% CI: 1.30 to 19.16, p = 0.02, OR 2 vs. 3: 8.43, 95% CI: 1.80 to 39.45, p = 0.007) were independently associated with resilience. Resilient individuals could be differentiated from non-resilient participants by higher fractional anisotropy (FA), and less association between anterior and posterior resting state networks. Higher FA had a significantly more positive effect on cognitive performance in participants with HA, compared to those without. Resilient individuals could be distinguished from those who developed impairments on the basis of sociodemographic characteristics, brain structural and functional connectivity, but not midlife lifestyles. There was a synergistic deleterious effect of hippocampal atrophy and poor white matter integrity on cognitive performance. Exploiting and supporting neural correlates of resilience could offer a fresh approach to postpone or avoid the appearance of clinical symptoms.

Partial Text

The term resilience is frequently used in a psychological context, referring to the ability to cope in the face of stressful life events. There is recent evidence that the concept of resilience is also meaningful and relevant for cognitive outcomes [1, 2]. Some individuals appear to maintain high memory function despite the presence of risk factors for impairment [3] or pathological brain changes, indicating resilience. For example, approximately one fifth of individuals positive for Alzheimer biomarkers have normal cognitive function, regardless of whether these are neuropathological findings on autopsy [4], amyloid on positron emission tomography (PET) imaging [5] or abnormal levels of cerebrospinal fluid (CSF) tau and amyloid [2, 6]. Similarly, the extent of functional deficits following stroke and traumatic brain injury can differ markedly [7, 8].

We combined multimodal neuroimaging and functional cognitive data to define a phenotype of resilience. Sociodemographic factors, but not lifestyle, were associated with resilience. Resilient subjects’ brains were characterized by higher structural integrity of white matter tracts and weaker inter-network functional connections.




Leave a Reply

Your email address will not be published.