Table 4

Predictive model comparisons at baseline

Model summary measureModelSCANSRUN DMCHARMONISATIONPRESERVEASPS-FamCADASIL
AICClinical283.451148.17304.77261.09496.22131.10
Clinical-DTI255.591121.22287.73243.55503.45124.44
Clinical-MRI204.451119.72303.27247.49493.74123.15
Clinical-MRI-DTI199.541113.17293.52246.02502.68122.64
Adj R2 Clinical0.4200.4220.3730.2290.5020.107
Clinical-DTI0.4780.4540.4560.3730.5160.235
Clinical-MRI0.5700.4590.3980.3650.5130.292
Clinical-MRI-DTI0.5640.4670.4470.3910.5230.311
  • Employing the AIC and the adjusted R2 variance (Adj R2), the model fits and the Adj R2 of the Clinical model (Clinical: age and sex and education/premorbid IQ), the Clinical-DTI model (age and sex and education/premorbid IQ+MD median), the Clinical-MRI model (age and sex and education/premorbid IQ and BV and WMH and lacune count and CMB) and the Clinical-MRI-DTI model (age and sex and education/premorbid IQ & BV and WMH and lacune count and CMB and MD median) were compared. AIC was lowest (reflecting best model fit) for the DTI+MRI + clinical model in all single-centre SVD cohorts (SCANS, RUN DMC and CADASIL) indicating that this is the best model dealing with the trade-off between the goodness of fit of the model and the simplicity of the model. The multimodal MRI’s Adj R2 ranged between 0.311 and 0.564 across the cohorts.

  • AIC, Akaike information criterion; BV, brain volume; CMB, cerebral microbleeds; DTI, diffusion tensor imaging; MD, mean diffusivity; Adj. R2, adjusted R2 variance; SVD, small vessel disease; WMH, white matter hyperintensities.