More than chronic pain: behavioural and psychosocial protective factors predict lower brain age in adults with/at risk of knee osteoarthritis over two years

Abstract

The interplay between chronic musculoskeletal pain and brain ageing is complex. Studies employing machine learning models to assess relationships between brain age and chronic pain generally show that higher chronic pain severity associates with older brain age. Analyses to date have not considered individual and community-level socioenvironmental risk factors or behavioural/psychosocial protective factors as potential modifiers of cross-sectional and longitudinal brain age. This study aimed to elucidate the relationships between chronic pain, socioenvironmental risk, behavioural/psychosocial protective factors, and brain ageing. The sample comprised 197 adults (Men:Women = 68:129) from a prospective observational cohort study. Most individuals reported knee pain and were with/at risk of osteoarthritis. A subset of 128 participants (Men:Women = 41:87) completed a follow-up MRI session at 2 years and were included in the longitudinal analysis (Aim 2). Participants were 45-85 years of age and self-identified as non-Hispanic Black or non-Hispanic White. Data collected included demographics, health history, pain assessments, individual and community-level socioenvironmental factors (education, income, household size, marital and insurance status, and area deprivation index) coded as a summative socioenvironmental risk variable, and behavioural/psychosocial factors (tobacco use, waist circumference, optimism, positive and negative affect, perceived stress, perceived social support, sleep) coded as a summative behavioural/psychosocial protective factor variable. Structural MRI data were used to estimate brain age by applying a machine learning approach (DeepBrainNet). Cross-sectional analyses utilized regression and analysis of variance, while longitudinal analyses utilized a linear mixed model. Higher chronic pain stage and socioenvironmental risk are associated with an increased brain age gap (the difference between chronological age and predicted brain age). Participants who had higher socioenvironmental risk had brains that were about three years older than those of participants with lower risk. Having more behavioural/psychosocial protective factors correlated with a lower brain age gap; participants with higher behavioural/psychosocial protective factors had brains that were over three years younger than participants with fewer behavioural/psychosocial protective factors. Longitudinally, higher baseline behavioural/psychosocial protective factors are associated with lower brain age over the 2-year span, beyond the effects of chronic pain stage and socioenvironmental risk. Our findings show behavioural/psychosocial protective factors may counteract neurobiological ageing and help buffer the brain from chronic pain.

Keywords: behavioural/psychosocial protective factors; brain age; chronic musculoskeletal pain; machine learning; socioenvironmental risk.

Graphical Abstract

Figure 1

An overview of the MP-RAGE processing with DeepBrainNet and calculation of brain age gap. Caption: A summary of DeepBrainNet calculation of predicted brain age and brain age gap (BAG) from Magnetization Prepared Rapid Gradient Echo (MP-RAGE) Magnetic Resonance Imaging (MRI). The shown MP-RAGE axial slices are from one of our participants. Pre-processed MP-RAGE images for each participant were used for input into the pre-trained DeepBrainNet 2-Dimensional Convolutional Neural Network (2D CNN) model to yield a predicted brain age. The bottom scatter plot shows a representative relationship between predicted brain age and chronological age. The diagonal dotted line is a hypothetical perfect correlation between predicted brain age and chronological age. The difference between predicted brain age and chronological age (predicted brain age—chronological age) is the brain age gap. Positive values indicate an ‘older’ brain and negative values a ‘younger’ brain. This process was completed for all participants (n = 197).

Figure 2

Scatter plots depicting relationships between predictors of interest and brain age gap. Caption: All relationships shown are statistically significant (Pearson r correlations, P values < 0.05; n = 197). Individual dots indicate the adjusted BAG values for each participant. (A) (Hypothesis 1a) depicts the relationship between chronic pain stage and brain age gap (BAG). BAG is adjusted for sex, study site, health comorbidities, and image quality rating. (B) (Hypothesis 1b) depicts the relationship between socioenvironmental risk and BAG, with BAG adjusted for the same variables as in A. Panel C (Hypothesis 1c) depicts the relationship between behavioural/psychosocial protective factors and BAG, with BAG adjusted for the same variables as in A. Panel D (Hypothesis 1d) depicts the relationship between behavioural/psychosocial protective factors and BAG, with BAG adjusted for the same variables as in A plus chronic pain stage and socioenvironmental risk.

Figure 3

Post hoc behavioural/psychosocial protective factors median split group analysis violin plots. Caption: All groups are median split. Shapes denote socioenvironmental risk group membership, and colors denote chronic pain group membership. Individual circles and triangles are the adjusted BAG values for each participant. When adjusting for sex, study site, health comorbidities, image quality, chronic pain stage, and socioenvironmental risk, the group with higher behavioural/psychosocial protective factors had an adjusted brain age gap (BAG) 2.69 years ‘younger’ than the group with low behavioural/psychosocial protective factors (Model ANCOVA F(7189) = 7.18, P < 0.001; group t = −2.70, P  = 0.008; η_p² = 0.04 [0.01, 0.09], n = 197).