GCA level: brain level analyses
Lifebrain and UKB p-value cluster maps for the relationship of GCA and cortical features controlled for training are shown in Figure 1. For cortical volume and area, there were widespread positive correlations of GCA bilaterally throughout the cortex mantle observed in all lobes. For area, significant effects were observed in 47.6% and 44.2% of the left and right hemisphere surface areas, respectively. For volume, similar figures were 37.4% and 19.5% for left and right, respectively. Figure 1 Level correlations: meta-analytic p-value cluster maps of correlations of general cognitive ability (GCA) and cortical features, controlled for training. Shown are meta-analytic p-value cluster maps (Lifebrain and UKB) of the relationships between GCA at baseline and cortical features when age at baseline, sex, time (since first scan), and education are controlled for ( p < 0.05, corrected using a clustering p-value threshold of p < 0.01). Relationships are shown, from left to right for each panel: right and left lateral view, right and left medial view. For cortical thickness, only small positive effects were observed, near the left central sulcus, covering only 1.1% of the area. The results of sample-wise, control, and no-control analyzes for training are shown in Supplementary Figs 1 and 2. Results were largely similar, although slightly more spatially restricted, in control than when no training was controlled. When ICV was added as a covariate, the cross-sectional effects for volume and cortical area in the meta-analysis shown in Figure 1 became non-significant, with only a very small effect on cortical thickness in the left hemisphere (see Supplementary Fig. 3 ), pointing out that these are broad effects based on larger neuroanatomical structures in general, rather than being region-specific. To show effect sizes, we calculated the effect of increasing GCA by 1 SD on cortical volume. Across Lifebrain and UKB, 1 SD higher GCA was associated with 1.0% greater cortical volume. Effect size maps for level-level analyzes showing regional variation in effect sizes for each sample separately are shown in Supplementary Figure 4. Effect sizes were numerically smaller in Lifebrain (0.6%) than in UKB (1 .3%). Restricting analyzes to regions where significant effects were observed, a 1 SD increase in GCA was associated with 2.0% greater cortical volume in Lifebrain and 1.6% greater volume in UKB, but note that these latter effect sizes are inflated due to being within important areas. Similar analyzes for cortical area showed that 0.8% larger area was associated with 1 SD higher GCA across the cortex, with effects of 0.6% in Lifebrain and 0.9% in UKB. Restricting analyzes to regions where significant effects were observed, a 1 SD increase in GCA was associated with 1.6% greater cortical volume in Lifebrain and 1.2% greater volume in UKB, with the same caveat as above. For thickness, the results were minimal: 0.06% across all studies (Lifebrain 0.04%; UKB 0.08%). Within significant clusters (UKB only), 1 SD higher GCA results were 0.8%.
GCA level: brain change analyses
Having confirmed the expected positive relations between GCA and cortical volume and area tested for training in terms of the intersection effect, we investigated the issue of slope effects. Correlations of GCA level at baseline and change in cortical features, controlling for training, are shown in Figure 2. As expected, the effects were more spatially restricted than those observed for the intercept models, with only limited regions showing significant relationships: Higher baseline GCA was associated with less regional reduction in cortical volume in the left middle gyrus, a medial region around the central sulcus, and a portion of the lingual gyrus. The most extensive effects were seen for thickness change, where higher baseline GCA was associated with less thinning in regions corresponding to volume effects, in addition to parts of the right frontal and lateral temporal cortices and a region in the most medial part of the slice between the central sulcus and superior frontal cortex. No correlations were observed with area change Overall, this means that the observed positive correlations of GCA with volume change mainly reflect less cortical thinning with higher GCA. (See Supplementary Fig. 5 for the result for each subsample separately). Figure 2 Correlations of level change: meta-analytic p-value cluster maps of correlations of general cognitive ability (GCA) with change in cortical features, controlling for the effect of training. The association is shown for the interaction of GCA at baseline and time (interval since baseline scan), when age, sex, time, GCA, education, and the interaction of education and time are controlled for (p < 0 .05, corrected using a clustering p-value threshold of p < .01). Important areas are shown, from left to right for each panel: right and left lateral view, right and left medial view. Correlations of GCA with cortical change were essentially unaffected by adding ICV as a covariate (Supplementary Figure 6). In order to illustrate GCA-cortical change relationships and to characterize the consistency of effects across samples (UKB and Lifebrain), we designed the generalized additive mixed model (GAMM) for the different GCA quintiles, from lowest to highest (Fig. 3). , depicting trajectories of change for mean cortical volume and thickness within regions showing significant GCA x time associations. In all samples, subgroups with higher GCA started with larger volume and had less volume loss over time. For example, on average, individuals with the highest cognitive score on the UKB would be expected to start with a regional average cortical volume of 1.72 mm3 that would be maintained over the next three years, while those with the lowest GCA would start with an average of 1 .64 mm3 and to decrease to 1.61 mm3 in the next three years. Thus, the largest differences in cortical volume associated with GCA are found across sections (plane), while differences in slope (change) are smaller over the follow-up period. For cortical thickness, change trajectories were also highly consistently ordered, but those with higher GCA did not uniformly have thicker cortex at the first time point in these regions. Instead, different rates of cortical thinning over time were critical for generating differences in cortical thickness in these regions during aging. This was evident in both samples, but particularly pronounced in the UKB. Figure 3 Cortical change trajectories according to general cognitive ability (GCA). Trajectories are shown by GCA quintile for illustrative purposes, for mean cortical volume and thickness change in the analysis model, and regionally significant sites of associations in each cohort (shown in Supplementary Figure 5). For UKB, quintiles refer to actual scores from min to max (0–13) on the test, while for Lifebrain, quintiles refer to z-scores (where the mean is zero) from min to max for the samples. To further assess effect sizes, we generated histograms of the peak distribution of effects of one SD higher GCA for each measure for absolute volume, area, and thickness, as well as their change, as shown in Fig. 4. (For cortical distributions of such sizes effect per sample, see Supplementary Fig. 7). As can be seen, almost all peaks show positive plane-to-plane relationships between GCA and volume and area. For thickness, the distribution is only slightly shifted to the right of zero, confirming the weak GCA thickness relationships. Regarding the GCA-brain level change, the histograms showed that for the region, the results were almost perfectly distributed around zero. For volume, there was a clear shift to the right, meaning that higher GCA tended to be associated with fewer volume reductions, but substantially less than for compensation effects. Cortical thickness showed the most right-skewed distribution, much greater than for displacement effects. Inspecting all histograms, it is clear that higher GCA is associated with greater cortical volume…
