Grey and white matter metrics demonstrate distinct and complementary prediction of differences in cognitive performance in children: Findings from ABCD (N= 11 876)

Individual differences in cognitive performance in childhood are a key predictor of significant life outcomes such as educational attainment and mental health. Differences in cognitive ability are governed in part by variations in brain structure. However, studies commonly focus on either grey or white matter metrics in humans, leaving open the key question as to whether grey or white matter microstructure play distinct or complementary roles supporting cognitive performance. To compare the role of grey and white matter in supporting cognitive performance, we used regularized structural equation models to predict cognitive performance with grey and white matter measures. Specifically, we compared how grey matter (volume, cortical thickness and surface area) and white matter measures (volume, fractional anisotropy and mean diffusivity) predicted individual differences in cognitive performance. The models were tested in 11,876 children (ABCD Study, 5680 female; 6196 male) at 10 years old. We found that grey and white matter metrics bring partly non-overlapping information to predict cognitive performance. The models with only grey or white matter explained respectively 15.4% and 12.4% of the variance in cognitive performance, while the combined model explained 19.0%. Zooming in we additionally found that different metrics within grey and white matter had different predictive power, and that the tracts/regions that were most predictive of cognitive performance differed across metric. These results show that studies focusing on a single metric in either grey or white matter to study the link between brain structure and cognitive performance are missing a key part of the equation.


Summary of the variables
Supplementary Table 1.Characteristics of the variables used in the models.All the measures of grey and white matter share the same frequency and number of missing data per tissue.

Models per metric estimating cognitive factor from several regions in one metric
For every metric, we reported the table for the models with all the regions included as predictors and the models with only the regions that survived the regularization (the one that we used for the analyses).

Cortical Thickness (model with all the regions)
Supplementary Table 9

Models per tissue estimating cognitive factor from several regions in three metrics
For grey and white matter, we reported the tables for the models with all the regions included as predictors and the models with only the regions that survived the regularization (the one that we used for the analyses).

Grey matter metrics (model with all the regions)
Supplementary Table 33 Mean (sd) : -0.1 (1) min ≤ med ≤ max: -9.3 ≤ 0 ≤ 9.4 IQR (CV) : 1.2 (-8.2) 11107 distinct values 739 (6.2%) Supplementary Table 5. Regression estimates for the model estimating how grey matter volume of one region predict the cognitive factor Supplementary Figure 3. Standardized parameter estimates of how the surface area of each region of interest predict the cognitive factorGrey Matter VolumeSupplementary Figure 4. Standardized parameter estimates of how the grey matter volume of each region of interest predict the cognitive factorFractional AnisotropySupplementary Table 6.Regression estimates for the model estimating how fractional anisotropy of one region predict the cognitive factor

Cortical Thickness (model with the regularized regions)
Supplementary Figure 5. Standardized parameter estimates of how cortical thickness of each region of interest together predict the cognitive factorSupplementary Table10.Comparison of the model with free parameters and the model with parameters constrained to a same value for the cortical thickness model with all the regions.AIC/BIC are two information criterions.Pr(>Chisq): p-value of the chi-square ratio test.Supplementary Table 11.Regression estimates for the model estimating how cortical thickness of the regularized regions predict the cognitive factor Supplementary Figure 6.Standardized parameter estimates of how the cortical thickness of each regularized region of interest together predict the cognitive factorSupplementary Table12.Comparison of the model with free parameters and the model with parameters constrained to a same value for the cortical thickness model with the regularized regions.AIC/BIC are two information criterions.Pr(>Chisq): p-value of the chi-square ratio test.

model with all the regions)
Supplementary Table 13.Regression estimates for the model estimating how surface area of all the regions predict the cognitive factor Supplementary Figure 7. Standardized parameter estimates of how the surface area of each region of interest together predict the cognitive factorSupplementary Table14.Comparison of the model with free parameters and the model with parameters constrained to a same value for the surface area model with all the regions.AIC/BIC are two information criterions.Pr(>Chisq): p-value of the chi-square ratio test.Supplementary Table 15.Regression estimates for the model estimating how surface area of the regularized regions predict the cognitive factorSupplementary Figure8.Standardized parameter estimates of how the surface area of each regularized region of interest together predict the cognitive factor Supplementary Table16.Comparison of the model with free parameters and the model with parameters constrained to a same value for the surface area model with the regularized regions.AIC/BIC are two information criterions.Pr(>Chisq): p-value of the chi-square ratio test.

Grey Matter Volume (model with all the regions)
Supplementary Table 17.Regression estimates for the model estimating how grey matter volume of all the regions predict the cognitive factor Supplementary Figure 9. Standardized parameter estimates of how the grey matter volume of each region of interest together predict the cognitive factorSupplementary Table18.Comparison of the model with free parameters and the model with parameters constrained to a same value for the grey matter volume model with all the regions.AIC/BIC are two information criterions.Pr(>Chisq): p-value of the chi-square ratio test.

Grey Matter Volume (model with the regularized regions) SupplementaryTable 19 .
Regression estimates for the model estimating how grey matter volume of the regularized regions predict the cognitive factor Supplementary Figure10.Standardized parameter estimates of how the grey matter volume of each regularized region of interest together predict the cognitive factor Supplementary Table20.Comparison of the model with free parameters and the model with parameters constrained to a same value for the grey matter volume model with the regularized regions.AIC/BIC are two information criterions.Pr(>Chisq): p-value of the chi-square ratio test.

Degree of freedom AIC BIC Difference in degree of freedom Pr (>Chisq)
Supplementary Table 21.Regression estimates for the model estimating how fractional anisotropy of all the regions predict the cognitive factor Supplementary Table 23.Regression estimates for the model estimating how fractional anisotropy of the regularized regions predict the cognitive factor Supplementary Table22.Comparison of the model with free parameters and the model with parameters constrained to a same value for the fractional anisotropy model with all the regions.AIC/BIC are two information criterions.Pr(>Chisq): p-value of the chi-square ratio test.

Table 24 .
Comparison of the model with free parameters and the model with parameters constrained to a same value for the fractional anisotropy model with the regularized regions.AIC/BIC are two information criterions.Pr(>Chisq): p-value of the chi-square ratio test.

model with all the regions)
Supplementary Table26.Comparison of the model with free parameters and the model with parameters constrained to a same value for the mean diffusivity model with all the regions.AIC/BIC are two information criterions.Pr(>Chisq): p-value of the chi-square ratio test.
Supplementary Table 27.Regression estimates for the model estimating how mean diffusivity of the regularized regions predict the cognitive factor

Table 28 .
Comparison of the model with free parameters and the model with parameters constrained to a same value for the mean diffusivity model with the regularized regions.AIC/BIC are two information criterions.Pr(>Chisq): p-value of the chi-square ratio test.

model with all the regions)
Supplementary Table 29.Regression estimates for the model estimating how white matter volume of all the regions predict the cognitive factor Supplementary Table 31.Regression estimates for the model estimating how white matter volume of the regularized regions predict the cognitive factor Supplementary Table30.Comparison of the model with free parameters and the model with parameters constrained to a same value for the white matter volume model with all the regions.AIC/BIC are two information criterions.Pr(>Chisq): p-value of the chi-square ratio test.Supplementary Table32.Comparison of the model with free parameters and the model with parameters constrained to a same value for the white matter volume model with the regularized regions.AIC/BIC are two information criterions.Pr(>Chisq): p-value of the chi-square ratio test.

White matter metrics (model with the regularized regions)
. Regression estimates for the model estimating how cortical thickness, surface area and grey matter volume of all the regions predict the cognitive factor Supplementary Table38.Comparison of the model with free parameters and the model with parameters constrained to a same value for the white matter model with all the regions.AIC/BIC are two information criterions.Pr(>Chisq): p-value of the chi-square ratio test.Supplementary Table 39.Regression estimates for the model estimating how fractional anisotropy, mean diffusivity and white matter volume of the regularized regions predict the cognitive factor