A recent study posted in The Lancets SSRN* preprint server evaluated the biological age of the brain as a disease stratification measure for Huntington's disease (HD).
Study: Brain Age as a New Measure of Disease Stratification in Huntington's Disease. Image Credit: Chinnapong/Shutterstock.com
*Important notice: SSRN publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.
Background
HD, a rare neurodegenerative disease, is caused by Huntingtin (HTT) gene mutations. It broadly impacts functional ability, leading to dysfunction in psychiatric, cognitive, and movement.
Despite recent advances in defining broad stages of disease progression, more accurate stratification methods are necessary to classify HD patients according to disease progression accurately.
Brain age is a biomarker predicting chronological age, which may better predict disease states than others. Brain age calculation methods estimate the discrepancy between chronological age and predicted brain age and provide insights into brain health and accelerated aging or neurodegeneration.
About the study
In the present study, researchers explored the potential use of brain age measures of HD patients to stratify disease progression states. They used structural magnetic resonance imaging (MRI) and clinical data from prior longitudinal studies (IMAGE-HD, TRACK-HD, and PREDICT-HD). Nine hundred fifty-three participants from the three cohorts were retained for analysis following quality control.
First, brain age was calculated using cytosine-adenine-guanine (CAG) and CAG-age product (CAP) scores, and patients were stratified into four groups. Patients with less than 36 CAG repeats were controls (group one), and those with CAP scores of 0-290, 290-368, and > 368 were stratified into groups two, three, and four, respectively.
White matter, grey matter, and cerebrospinal fluid (CSF) segmentations were checked for accuracy and quality, resulting in 82 exclusions. Brain age was calculated by a Gaussian process regression method. The brain age difference, i.e., predicted age difference (PAD), was estimated. Baseline brain age, chronological age, and brain-PAD were analyzed.
A linear mixed model analysis compared the capability of baseline brain-PAD and CAP score in explaining longitudinal disease progression. Further, longitudinal brain-PAD was evaluated as an outcome for each group, and the interactions between time and CAP groups were tested.
Finally, longitudinal k-means clustering, with six cluster sizes (two to seven), was performed using brain-PAD of HD patients to cluster longitudinal disease progression further.
Findings
The study included 669 HD patients and 284 healthy controls. The team found no significant differences between the controls' chronological and brain ages, although others exhibited substantial differences. Further, higher baseline brain-PAD was associated with increased disease severity.
The model with brain-PAD was top-ranked, with a significantly better model fit than the CAP score-based model.
The predicted linear trends in the longitudinal analysis of brain-PAD were only significant for controls and group three, suggesting that PAD may not capture the same linear patterns of progression observed in CAP groups. Nevertheless, meaningful interactions were evident between time and all CAP groups. Statistical analyses indicated that the optimal number of clusters was four, five, or six.
Baseline brain-PAD, volumetric, clinical, and cognitive measures were characterized according to the new clusters. CAG repeats, and brain-PAD were aligned in the five-cluster solution, but brain-PAD and CAP scores differed.
Further, the longitudinal progression of brain-PAD, putamen or caudate volume, Stroop (w), total motor score (TMS), and symbol digit modalities test (SDMT) was characterized for four- and five-cluster solutions.
This showed a cluster-time interaction for each biomarker, implying that other biomarkers follow distinct longitudinal trajectories for clusters defined on brain-PAD. The highest rate of change of cognitive markers (Stroop(w), TMS, and SDMT) was observed in cluster D in the four-cluster solution and cluster E in the five-cluster solution.
On the other hand, the highest rate of change in volumetric markers (caudate or putamen volume) was noted in cluster A for both.
Conclusions
In sum, the study showed a relationship between baseline brain-PAD and the progression of HD. In addition, there was an association between longitudinal changes in brain-PAD and HD progression, confirming the decline in brain health. Notably, the team identified five distinct states of HD progression based on longitudinal changes in brain-PAD.
The authors suggest that HD patients could be stratified into different disease progression states using clustering based on brain-PAD, where each group can have different progression trajectories for biomarkers. Nevertheless, the clustering approach needs independent validation before clinical application.
Overall, the findings indicate that brain-PAD can be used for improved HD patient stratification and to capture heterogeneous states of HD progression.
*Important notice: SSRN publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.
How brain connectivity and machine learning enhance understanding of human cognition