Brain-age is associated with progression to dementia in memory clinic patients

Abstract

Background: Biomarkers for the early detection of dementia risk hold promise for better disease monitoring and targeted interventions. However, most biomarker studies, particularly in neuroimaging, have analysed artificially 'clean' research groups, free from comorbidities, erroneous referrals, contraindications and from a narrow sociodemographic pool. Such biases mean that neuroimaging samples are often unrepresentative of the target population for dementia risk (e.g., people referred to a memory clinic), limiting the generalisation of these studies to real-world clinical settings. To facilitate better translation from research to the clinic, datasets that are more representative of dementia patient groups are warranted.

Methods: We analysed T1-weighted MRI scans from a real-world setting of patients referred to UK memory clinic services (n = 1140; 60.2 % female and mean [SD] age of 70.0[10.8] years) to derive 'brain-age'. Brain-age is an index of age-related brain health based on quantitative analysis of structural neuroimaging, largely reflecting brain atrophy. Brain-predicted age difference (brain-PAD) was calculated as brain-age minus chronological age. We determined which patients went on to develop dementia between three months and 7.8 years after neuroimaging assessment (n = 476) using linkage to electronic health records.

Results: Survival analysis, using Cox regression, indicated a 3 % increased risk of dementia per brain-PAD year (hazard ratio [95 % CI] = 1.03 [1.02,1.04], p < 0.0001), adjusted for baseline age, age², sex, Mini Mental State Examination (MMSE) score and normalised brain volume. In sensitivity analyses, brain-PAD remained significant when time-to-dementia was at least 3 years (hazard ratio [95 % CI] = 1.06 [1.02, 1.09], p = 0.0006), or when baseline MMSE score ≥ 27 (hazard ratio [95 % CI] = 1.03 [1.01, 1.05], p = 0.0006).

Conclusions: Memory clinic patients with older-appearing brains are more likely to receive a subsequent dementia diagnosis. Potentially, brain-age could aid decision-making during initial memory clinic assessment to improve early detection of dementia. Even when neuroimaging assessment was more than 3 years prior to diagnosis and when cognitive functioning was not clearly impaired, brain-age still proved informative. These real-world results support the use of quantitative neuroimaging biomarkers like brain-age in memory clinics.

Keywords: Ageing; Brain-age; Dementia; Ecological validity; Electronic health records; Machine learning.

Fig. 1

Data pre-processing pipeline. (A) Memory clinic patients were classified as: “future Dementia Diagnosis” (orange) or “no Dementia Diagnosis” (grey). (B) Neuroimaging data preparation: brainageR was applied to T1-weighted MRI scans to obtain a brain-age estimate and normalised brain volume for each patient. Quality Control (QC) aimed to remove cases with image artefacts and occurred at two stages, before and after segmentation. (C) Clinical data preparation: clinical data was retrieved across three databases, South London and Maudsley Hospital Electronic Health Record (SLaM EHRs), Hospital Episodes Statistics (HES) and Office of National Statistics (ONS) which were accessed and linked to the neuroimaging data via the Clinical Record Interactive Search (CRIS). These databases provided diagnostic, demographic, cognitive and mortality data and facilitated labelling into “future Dementia Diagnosis” or “no Dementia Diagnosis”. Patients who were diagnosed with dementia before, or up to 3 months after the neuroimaging assessment, were excluded. (D) Merging the neuroimaging and clinical data: the final dataset (N = 1140) included only complete cases for the following variables: sex, age, brain-age, normalised brain volume, MMSE (Mini Mental State Examination) and scanner information. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

Kaplan-Meier plot for brain-PAD. This plot illustrates the proportion of patients who develop dementia based on a tertile split of brain-PAD score. At time 0 (time of the neuroimaging assessment, all patients are free of a dementia diagnosis. Over time, patients with higher brain-PAD scores (pink) are more likely to get a dementia diagnosis and more rapidly than the ones with lower brain-PAD scores (blue). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

Scatterplots of age, brain-age and brain-PAD. The scatterplots show on the x-axis, chronological age and on the y-axis either brain-age (top panel A) or brain-PAD (bottom panel B) split by group, “no Dementia Diagnosis” (noDD; grey and on the left) and “future Dementia Diagnosis” (futureDD; orange and on the right). (A) Chronological age vs brain-age: The identity line (green) shows the ideal case when chronological age matches the brain-age estimate, y = x. Lines of best fit (orange, grey) within each group are both positive showing that brain-age estimates tend to be larger than chronological age for both groups of patients and in particular for the “future Dementia Diagnosis” group. (B) Chronological age vs brain-PAD: The density plots on the right of each scatterplot illustrate the distribution of brain-PAD scores. The green line is set at brain-PAD = 0 (i.e., brain age matches chronological age). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Appendix C

Fig. D1

Ethnic background. Distribution of total sample split by 15 ethnic background groupings excluding two response categories for which there were no data available (“Not Stated” and “NULL”).