Relationships of birthweight and cortical characteristics across LCBC, ABCD, and UKB samples.

Age, sex, time (interval since baseline) and scanner site (as well as ethnicity in the ABCD) were controlled for. Significant relationships are shown for area, thickness, and volume for each sample, from left to right: lateral view and medial view, right hemisphere.

Interactions of BW and time on cortical characteristics across LCBC, ABCD, and UKB samples.

Age, sex, scanner site, time, and birth weight (as well as ethnicity in the ABCD) were controlled for. Significant relationships are shown, from left to right: lateral view, right and left hemisphere, and medial view, right and left hemisphere.

Spatial correlation of birth weight effects on brain structure across datasets for cortical a) area, b) thickness, and c) volume.

Spatial correlation of birth weight effects on brain structure across datasets. For each panel, the upper triangular matrix shows Pearson’s (r) pairwise spatial correlation between the different cohorts’ cortical maps. Data is shown as a color-density plot. The red line represents the fitting between the two maps. The lower triangular matrix shows the significance testing. The dashed-grey line shows the empirical correlation, while the orange histogram represents the null distribution based on the spin test. The diagonal shows the effect of birth weight on cortical structure (right hemisphere shown only). Note that the βeta-maps are shown as a percentile red-green-blue scale, where red represents a lower (or more negative) effect of birth weight on cortical structure and vice versa. See Supplementary Table 2 for stats. Units in the density maps represent birth weight effects as mm/g, mm2/g, and mm3/g (10e-5) for cortical thickness, area, and volume, respectively.

Effects of birthweight discordance on cortical area in the sample of monozygotic (MZ) twins.

Significant relationships are shown from left to right: lateral view, right and left hemisphere, and medial view, right and left hemisphere. Plots are showing -for illustrative purposes – individual data points and expected trajectories for cortical area in mm (Y-axes) within the significant regions according to birth weight (BW) discordance (left panel) and BW (right panel) in kilograms (X-axes).

Effects of birthweight discordance on cortical thickness in the sample of monozygotic (MZ) twins.

Relationships significant corrected with cluster-forming threshold of 2.0 (p< .01) are shown from left to right: lateral view, right and left hemisphere, and medial view, right and left hemisphere. Plots are showing – for illustrative purposes - individual data points and expected trajectories for cortical thickness in mm (Y-axes) within the significant regions according to birth weight (BW) discordance (left panel) and BW (right panel) in kilograms (X-axes).

Descriptive statistics for the longitudinal samples.

F = number of females in the sample, M = Mean, SD= Standard deviation. Numbers are given in years for baseline age, time since baseline and education, birthweight is given in kilograms. For LCBC, only 584 participants had information on education. Parental education was used in ABCD, and in LCBC when the participant was below 18 years of age, and also if no other education information was available for participants up to 21 years.

MR acquisition parameters

TR: Repetition time, TE: Echo time, TI: Inversion time, FoV: Field of View, iPat: in-plane acceleration, GRAPPA: GRAPPA acceleration factor. *Customized

Relationships of birthweight and cortical area across LCBC, ABCD, and UKB samples when controlling for age, sex, time (interval since baseline) and scanner site (as well as ethnicity in the ABCD).

Significant relationships are shown from left to right: lateral view, right and left hemisphere, and medial view, right and left hemisphere.

Relationships of birthweight and cortical characteristics across LCBC, ABCD, and UKB samples when controlling for education, age, sex, time (interval since baseline) and scanner site (as well as ethnicity in the ABCD).

Significant relationships are shown from left to right: lateral view, right and left hemisphere, and medial view, right and left hemisphere.

Relationships of birthweight and cortical characteristics across LCBC and ABCD samples when controlling for gestational length in weeks (LCBC) or weeks born prematurely (ABCD), age, sex, time (interval since baseline) and scanner site (as well as ethnicity in the ABCD).

Significant relationships are shown from left to right: lateral view, right and left hemisphere, and medial view, right and left hemisphere.

Relationships of birthweight and cortical characteristics across LCBC, ABCD, and UKB when restricting the samples to participants with birth weights between 2.5 and 5.0 kg, controlling for age, sex, time (interval since baseline) and scanner site (as well as ethnicity in the ABCD).

(Compare to Figure 1.) Significant relationships are shown from left to right: lateral view, right and left hemisphere, and medial view, right and left hemisphere.

Relationships of birthweight and cortical characteristics across LCBC, ABCD, and UKB samples when controlling for intracranial volume (ICV), age, sex, time (interval since baseline) and scanner site (as well as ethnicity in the ABCD).

Significant relationships are shown from left to right: lateral view, right and left hemisphere, and medial view, right and left hemisphere.

Interactions of BW and time on cortical characteristics across LCBC, ABCD, and UKB samples when controlling for age, sex, scanner site, time, birth weight, and the interaction of baseline age and time (as well as ethnicity in the ABCD).

Significant relationships are shown from left to right: lateral view, right and left hemisphere, and medial view, right and left hemisphere.

Plots showing individual data points and expected trajectories for cortical area within the significant regions (refer to Figures 1 and 2) of each sample split in two based on BW (higher BW in red color= upper half, lower BW in blue color= lower half of BW distribution) are shown. Y-axis: cortical area: mm2, thickness, X-axis: time in years.

Plots showing individual data points and expected trajectories for cortical thickness within the significant regions (refer to Figures 1 and 2) of each sample split in two based on BW (higher BW in red color= upper half, lower BW in blue color= lower half of BW distribution) are shown. Y-axis: cortical thickness: mm, X-axis: time in years.

Plots showing individual data points and expected trajectories for cortical volume within the significant regions (refer to Figures 1 and 2) of each sample split in two based on BW (higher BW in red color= upper half, lower BW in blue color= lower half of BW distribution) are shown. Y-axis: cortical thickness: mm, X-axis: time in years.

Spatial correlation of birth weight effects on brain structure change across datasets.

For each panel, the upper triangular matrix shows Pearson’s (r) pairwise spatial correlation between the different cohorts’ cortical maps. Data is shown as a color-density plot. The red line represents the fitting between the two maps. The lower triangular matrix shows the significance testing. The dashed-grey line shows the empirical correlation, while the orange histogram represents the null distribution based on the spin test. The diagonal shows the effect of birth weight on cortical structure (right hemisphere shown only). Note that the βeta-maps are shown as a percentile red-green-blue scale, where red represents a lower (or more negative) effect of birth weight on cortical structure and vice versa. See Supplementary Table 2 for stats. The different panels show the spatial correlation of birth weight effects on cortical a) area, b) thickness, and c) volume. Units in the density maps represent birth weight effects as mm/g, mm2/g, and mm3/g (10e-5) for cortical thickness, area, and volume, respectively.

Spatial correlation of birth weight effects on brain structure across datasets. Pearson’s correlation and significance for pairwise spatial correlations between the different cohorts’ birth weight, with and without correction for ICV, and birth weight x time (years) -related cortical maps; i.e. effects of birth weight on cortical structure and change in cortical structure. Significance was assessed using the spin test. Spatial correlations were assessed using the Beta maps (and -log10(p) for birth weight). See Figure 3 and Supplementary Figure 9 and 10 for a visual representation. P-values are FDR-corrected (n = 9).

The degree of within-sample replicability of birth weight effects on cortical structure for LCBC, ABCD, and UKB.

Brain images show the exploratory replicability analyses; i.e. the proportion of results (refer to Figure 1) that are significant across different subsamples (50% of the original sample; |N| = 500 subsamples). The overlay is thresholded at replicability (p) = .1. For visual purposes data has been up-sampled and smoothed (FWHM = 5) on fsaverage space. Left-side violin plots show the exploratory replicability; i.e., the proportion of significant results across the different subsamples (unit = voxel). Right-side plots show the confirmatory replicability; i.e. for each sample, the proportion of results (p < 0.01; FWE-corrected) that also pass significance criteria (p < 0.05) in a test subsample (|N| = 500 subsamples) (unit = analysis). Embedded boxplots display median, interquartile range, and range.

Interaction effects of birth weight discordance and time on cortical thickness in the sample of monozygotic (MZ) twins.

Significant relationships are shown from left to right: lateral view, right and left hemisphere, and medial view, right and left hemisphere. Plots showing individual data points and expected trajectories for cortical thickness in mm (Y-axes) within the significant regions according to BW discordance (left panel) and BW (right panel) in kilograms (X-axes) are shown.

Comparison between spline (GAMM) and linear (LME) models on the effect of birth weight on cortical characteristics.

Age was fitted either as a smoothing spline using generalized additive mixed models (GAMM, mgcv r-package) or a linear regressor with a linear mixed models (LME, lmer r-package) framework. The analyses were performed on ROI-based using the Desikan-Killiany atlas. Significance was considered at a FDR corrected threshold of p < 0.04. All the remaining parameters were comparable to the main analyses shown in Figure 1. The viridis-yellow scale represents the lower-higher Beta regressors. Red contour displays regions showing significant effects of birth weight. Note the high correspondence with both fitting models. Differences are only noticeable in the LCBC sample due to the datasets’ wider age range (i.e., lifespan dataset).

Comparison between spline (GAMM) and linear (LME) models on the effect of birth weight on cortical change.

Age was fitted either as a smoothing spline using generalized additive mixed models (GAMM, mgcv r-package) or a linear regressor with a linear mixed models (LME, lmer r-package) framework. The analyses were performed on ROI-based using the Desikan-Killiany atlas. Significance was considered at a FDR corrected threshold of p < 0.04. All the remaining parameters were comparable to the main analyses shown in Figure 1. The viridis-yellow scale represents the lower-higher Beta regressors. Red contour displays regions showing significant effects of birth weight. Note the high correspondence with both fitting models. Differences are only noticeable in the LCBC sample due to the datasets’ wider age range (i.e., lifespan dataset).

Exploratory and confirmatory replicability of birth weight on cortical change within datasets. Units represent median and interquartile range.

Exploratory and confirmatory replicability across datasets. “--" denotes no significant clusters in the right-hand dataset. a → b and b → a denotes directionality of the replicability analyses being “a” and “b” in the left and right hand of the “Compared Datasets” column. BW = Birth weight.