Morphometric similarity deviations in stimulant use disorder point towards abnormal brain ageing

Abstract

Chronic drug use negatively impacts ageing, resulting in diminished health and quality of life. However, little is known about biomarkers of abnormal ageing in stimulant drug users. Using morphometric similarity network mapping, a novel approach to structural connectomics, we first mapped cross-sectional morphometric similarity trajectories of ageing in the publicly available Rockland Sample (20-80 years of age, n = 665). We then compared morphometric similarity and neuropsychological function between non-treatment-seeking, actively using patients with stimulant use disorder (n = 183, mean age: 35.6 years) and healthy control participants (n = 148, mean age: 36.0 years). The significantly altered mean regional morphometric similarity was found in 43 cortical regions including the inferior and orbital frontal gyri, pre/postcentral gyri and anterior temporal, superior parietal and occipital areas. Deviations from normative morphometric similarity trajectories in patients with stimulant use disorder suggested abnormal brain ageing. Furthermore, deficits in paired associates learning were consistent with neuropathology associated with both ageing and stimulant use disorder. Morphometric similarity mapping provides a promising biomarker for ageing in health and disease and may complement existing neuropsychological markers of age-related cognitive decline. Neuropathological ageing mechanisms in stimulant use disorder warrant further investigation to develop more age-appropriate treatments for older people addicted to stimulant drugs.

Keywords: cocaine; human connectome; morphometric similarity; orbitofrontal cortex; paired associates learning.

Graphical Abstract

Abnormal ageing in stimulant use disorder.

Figure 1

Cross-sectional trajectories of MS in a normative sample (NKI-RS, Nathan Kline Institute, Rockland Sample). Increasing (blue), decreasing (red), convex (black) and concave (green) trajectories were observed (A), with regions showing particularly high and low mean similarity scores tending towards zero with increasing age. All trajectories are plotted in the background in grey. The anatomical distribution of increasing, decreasing, convex and concave trajectories is shown in (B).

Figure 2

Case–control differences in SUD. Mean MS in healthy control participants (A) was significantly different from the mean MS in patients with SUD in 43 regions (C) (Supplementary material, Table 3). Orange colours indicate greater MS in controls and blue colours indicate greater MS in SUD (C). Significant differences between the control and SUD groups were observed in those regions with particularly high or low MS scores (D, golden circles show significant differences); the SUD group tended to have MS scores closer to zero (D, P_PERMUTATION = 0.002). Mean MS can be conceptualized mathematically as weighted regional degree as illustrated in (B).

Figure 3

Significant deviation from healthy brain development in SUD. Significant mean MS deviations from normative development in five age categories in patients with SUD. Significance was determined using one-sample t-tests, FDR corrected at P < 0.01. Five age categories comprised 20–30-, 30–35-, 35–40-, 40–45- and 45–60-year-old participants. Those regions for which the quadratic trajectory provided a poor fit to the data (adjusted r²<0, see Supplementary material, Fig. 6) were excluded. Deviations (absolute t-statistics) consistent with accelerated ageing are shown in blue, whereas those inconsistent with ageing are shown in red (see Supplementary material, Fig. 7 for the deviations in the Cambridge control group).

Figure 4

Brain-cognition association between MS and paired associates learning. Relationship between PAL, age, MS and stimulant drug use. The first two components of the PLS regression jointly explained 43.4% of variance in PAL total errors and age, and 4.35% of variance in regional MS in all 360 regions of interest. PLS2 explained a significant amount of variance in PAL performance and age (15.5%, P_perm = 0.02) but showed a strong positive association with PAL errors and a weak negative association with age (A). PLS1 also explained a significant amount of variance in PAL performance and age (27.9%, P_perm = 0.02) and was positively associated with both of these variables (B). The pattern of regional MS loadings on PLS1 (r = 0.385, P_perm = 0.37) and PLS2 (r = 0.165, P_perm = 0.44) was distinct from the pattern of the control versus stimulant use disorder group difference t-statistics (see Fig. 1C).