1. Neuroscience

Rate of brain aging associates with future executive function in Asian children and older adults

  1. Susan F Cheng
  2. Wan Lin Yue
  3. Kwun Kei Ng
  4. Xing Qian
  5. Siwei Liu
  6. Trevor WK Tan
  7. Kim-Ngan Nguyen
  8. Ruth LF Leong
  9. Saima Hilal
  10. Ching-Yu Cheng
  11. Ai Peng Tan
  12. Evelyn C Law
  13. Peter D Gluckman
  14. Christopher Li-Hsian Chen
  15. Yap Seng Chong
  16. Michael J Meaney
  17. Michael WL Chee
  18. BT Thomas Yeo
  19. Juan Helen Zhou  Is a corresponding author
  1. Integrative Sciences and Engineering Programme, NUS Graduate School, National University of Singapore, Singapore
  2. Yong Loo Lin School of Medicine, National University of Singapore, Singapore
  3. Saw Swee Hock School of Public Health, National University of Singapore and National University Health System, Singapore
  4. Memory Aging and Cognition Centre, National University Health System, Singapore
  5. Singapore Eye Research Institute, Singapore National Eye Centre, Singapore
  6. Duke-NUS Medical School, Singapore
  7. Singapore Institute for Clinical Sciences (SICS), A*STAR Research Entities (ARES), Singapore
  8. Brain–Body Initiative Program, Agency for Science, Technology and Research (A*STAR), Singapore
  9. National University Health System, Singapore
  10. Liggins Institute, University of Auckland, New Zealand
  11. Douglas Mental Health University Institute, McGill University, Canada
  12. Strategic Research Program, A*STAR Research Entities (ARES), Singapore
  13. Department of Electrical and Computer Engineering, National University of Singapore, Singapore
  14. N.1 Institute for Health & Institute for Digital Medicine (WisDM), National University of Singapore, Singapore
https://doi.org/10.7554/eLife.97036.3
Share this article
Cite this article
  1. Susan F Cheng
  2. Wan Lin Yue
  3. Kwun Kei Ng
  4. Xing Qian
  5. Siwei Liu
  6. Trevor WK Tan
  7. Kim-Ngan Nguyen
  8. Ruth LF Leong
  9. Saima Hilal
  10. Ching-Yu Cheng
  11. Ai Peng Tan
  12. Evelyn C Law
  13. Peter D Gluckman
  14. Christopher Li-Hsian Chen
  15. Yap Seng Chong
  16. Michael J Meaney
  17. Michael WL Chee
  18. BT Thomas Yeo
  19. Juan Helen Zhou
(2025)
Rate of brain aging associates with future executive function in Asian children and older adults
eLife 13:RP97036.
https://doi.org/10.7554/eLife.97036.3
  2. Download BibTeX
  3. Download .RIS
8 figures, 20 tables and 1 additional file

Figures

Figure 1
Download asset Open asset
Study design schematic.

(A, B) T1 MRI scans were minimally preprocessed according to the simple fully convolutional network (SFCN) pipeline (Leonardsen et al., 2022). These were (a) directly input into the pretrained brain age model or (b) split into 10 cross-validation folds to finetune the model. The finetuned model transferred the weights from the pretrained model for initialization. All layers were then retrained. Age predictions were obtained on the test folds. BAG was calculated by subtracting chronological age from predicted age. Model interpretability was interrogated using guided backpropagation. (C) Cross-sectional and longitudinal association of BAG and cognitive performance were tested using multiple linear regression models in both elderly and children. Time intervals for BAG and cognition, based on data availability, are shown schematically. Annual rate of change was calculated from a linear regression with time for each participant. All models included chronological age and sex as covariates.:^ models for elderly also included years of education as a covariate;* models with (annual rate of) change in BAG also included baseline BAG as a covariate. EDIS, Epidemiology of Dementia in Singapore; SLABS, Singapore-Longitudinal Aging Brain Study; GUSTO, Growing Up in Singapore Towards healthy Outcomes; BAG, brain age gap.

Figure 2 with 1 supplement
Download asset Open asset
The pretrained brain age model performs well in elderly participants, while the finetuned model performs well in both elderly participants and children.

Black identity lines representing perfect prediction are included for reference. (A) Predicted brain ages from the pretrained model are plotted against chronological age. They are highly correlated for EDIS and SLABS (elderly), but not GUSTO (children). (B) Predicted brain ages from the finetuned model are plotted against chronological age. They are highly correlated in all three datasets. EDIS, Epidemiology of Dementia in Singapore; SLABS, Singapore-Longitudinal Aging Brain Study; GUSTO, Growing Up in Singapore Towards healthy Outcomes; N, number of participants; r, Pearson’s correlation coefficient; MAE, mean absolute error; NCI, no cognitive impairment; CIND, cognitive impairment no dementia.

Figure 2—figure supplement 1
Download asset Open asset
Variance of finetuned predicted ages by age group in Growing Up in Singapore Towards healthy Outcomes GUSTO.
Figure 3 with 1 supplement
Download asset Open asset
BAG from the pretrained model is negatively associated with executive function in elderly participants.

Bolded p-values indicate significance after Holm-Bonferroni correction (pcorr<0.05). All models include chronological age, sex, and years of education as covariates. Models with change in BAG also include baseline BAG as a covariate. Results are similar after finetuning (Figure 3—figure supplement 1). (A) Partial regression plot between baseline BAG and executive function in EDIS, colored by cognitive status. A significant negative association is observed. (B) Partial regression plot between baseline BAG and long-term rate of change in executive function (mean follow-up time = years) in SLABS. A negative association is observed, but it is not significant after correcting for multiple comparisons. (C) Partial regression plot of early longitudinal rate of change in BAG (mean follow-up time = years) when added to the model in (B). A significant negative association and increase in R2 is observed. (D) Partial regression plot as in (C), but with future rate of change in executive function (mean follow-up time = years), removing the overlap with early change in BAG. A significant negative association is again observed. N, number of participants; β, standardized regression coefficient; p, p-value for variable of interest (x-axis); ΔRadj2, change in adjusted R2 when adding variable of interest; BAG, brain age gap; NCI, no cognitive impairment; CIND, cognitive impairment no dementia; EDIS, Epidemiology of Dementia in Singapore; SLABS – Singapore-Longitudinal Aging Brain Study.

Figure 3—figure supplement 1
Download asset Open asset
Brain age gap from the finetuned model remains negatively associated with executive function in elderly.

Compare to Figure 3A–D. (A) EDIS dataset: baseline BAG relates to baseline execituve function. (B) SLABS dataset: Baseline BAG does not relate to future changes in executive function. (C, D) SLABS dataset: Change in BAQ relates to future changes in executive function (non-overlapping in C and overlaping in D).

Figure 4 with 2 supplements
Download asset Open asset
Longitudinal BAG from the finetuned model is positively associated with inhibition in children.

Bolded p-values indicate significance after Holm–Bonferroni correction (pcorr<0.05). All models include chronological age and sex as covariates. Models with change in BAG also include baseline BAG as a covariate. (A) Partial regression plot between baseline BAG (calculated from 4.5 or 6.0 years old) and future NEPSY-II inhibition scaled subscore (measured at 8.5 years old). No significant association is observed. (B) Partial regression plot of early longitudinal rate of change in BAG calculated from 4.5 to 7.5 years old (mean follow-up time = 2.4 ± 0.7 years) when added to the model in (A). A significant positive association and increase in R2 is observed. N, number of participants; β, standardized regression coefficient; p, p-value for variable of interest (x-axis); ΔRadj2 , change in adjusted R2 when adding variable of interest; BAG, brain age gap; GUSTO, Growing Up in Singapore Towards healthy Outcomes.

Figure 4—figure supplement 1
Download asset Open asset
Brain age gap (BAG) from the pretrained model is not associated with inhibition in children.

Estimated BAGs > 2 are not shown for visual clarity. Compare to Figure 4A and B.

Figure 4—figure supplement 2
Download asset Open asset
Baseline BAG is not associated with baseline IQ in children.

Both brainage and IQ were measured cross-sectionally at 4.5 years old. (A) Partial regression plot with baseline BAG from the pretrained model. Estimated BAGs > 2 are not shown for visual clarity. No significant relationship is observed. (B) Partial regression plot with baseline BAG from the finetuned model. No significant relationship is observed. N, number of participants; β, standardized regression coefficient; p, p-value for variable of interest (x-axis); Multiple R2, coefficient of determination; GUSTO, Growing Up in Singapore Towards healthy Outcomes; BAG, brain age gap; KBIT-2, Kaufman Brief Intelligence Test Second Edition.

Figure 5 with 1 supplement
Download asset Open asset
Finetuned brain age models focus on distinct features in children and elderly participants.

The top 10% of features are shown for four representative brain slices on the left. Relative contributions for gray and white matter features across the whole brain are shown on the right. Regions near the lateral ventricles are labeled in red. Features more prominent in elderly than children are labeled in magenta, while features more prominent in children than elderly are labeled in blue. Features and relative contributions are generally consistent between (A) EDIS and (B) SLABS, but key differences can be seen in (C) GUSTO. EDIS, Epidemiology of Dementia in Singapore; SLABS, Singapore-Longitudinal Aging Brain Study; GUSTO,Growing Up in Singapore Towards healthy Outcomes; MCP,–middle cerebellar peduncle; PCT, Pontine crossing tract; gCC, genu of corpus callosum; bCC, body of corpus callosum; sCC, splenium of corpus callosum; Fx, fornix (column and body); CST, corticospinal tract; ML, medial lemniscus; ICP, inferior cerebellar peduncle; SCP, superior cerebellar peduncle; CP, cerebral peduncle; ALIC, anterior limb of internal capsule; PLIC, posterior limb of internal capsule; RLIC, retrolenticular part of internal capsule; ACR, anterior corona radiata; SCR, superior corona radiata; PCR, posterior corona radiata; PTR, posterior thalamic radiation; SS, sagittal stratum; EC, external capsule; cingulum CG, cingulum (cingulate gyrus); cingulum HIP, cingulum (hippocampus); Fx/ST, fornix (cres)/stria terminalis; SLF, superior longitudinal fasciculus; SFO, superior fronto-occipital fasciculus; UF, uncinate fasciculus; TAP, tapetum; Vis, visual network; SomMot, somatomotor network; DorsAttn, dorsal attention network; SalVentAttn, salience/ventral attention network; Limbic, limbic network; Cont , control/frontoparietal network; Default, default mode network; Hip+Amy, hippocampus + amygdala; Put+Cau, putamen + caudate; Tha l, thalamus.

Figure 5—figure supplement 1
Download asset Open asset
Pretrained models focus on similar features as finetuned models in EDIS and SLABS, but not in GUSTO.

Compare to Figure 5A–C. Features more prominent in elderly than children are labeled in magenta, while features more prominent in children than elderly are labeled in blue. Features and relative contributions are generally consistent between (A) EDIS and (B) SLABS, but key differences can be seen in (C) GUSTO.EDIS, Epidemiology of Dementia in Singapore; SLABS, Singapore-Longitudinal Aging Brain Study; GUSTO, Growing Up in Singapore Towards healthy Outcomes.

Appendix 1—figure 1
Download asset Open asset
Example learning curves from (A) tuning the last layer only on Growing Up in Singapore Towards healthy Outcomes (GUSTO), showing underfitting; (B) tuning all layers on GUSTO, showing in stability; (C) using a cosine learning rate decay, showing a good fit (D) using the same parameters on Epidemiology of Dementia in Singapore (EDIS),showing ‘forgetting’; (E) using a lower initial learning rate(1e-4) on EDIS, showing a better fit; and (F) using an initial learning rate of 1e-6, showing underfitting.
Author response image 1
Download asset Open asset
Author response image 2
Download asset Open asset
Brain age predictions on unseen Caucasian sample of older adults.

Predictions from the (A) pretrained and (B) finetuned brain age models on ADNI participants. Compare to Figure 2 of the main text.

Tables

Table 1
Participant characteristics at baseline.

EDIS was cross-sectional, while SLABS and GUSTO were longitudinal. Reported as mean ± standard deviation (range). *GUSTO ethnicities were based on the mother. M/F, male/female; C/M/I/O, Chinese/Malay/Indian/Other; MMSE, Mini-Mental State Examination; EDIS, Epidemiology of Dementia in Singapore; SLABS, Singapore Longitudinal Aging Brain Study; GUSTO, Growing Up in Singapore Towards healthy Outcomes.

ElderlyChildren
CharacteristicEDIS (N=694)SLABS (N=215)GUSTO (N=678)
Age (years)69.91 ± 6.46 (60−88)68.17 ± 6.77 (55−85)5.85 ± 1.68 (4.2−11.3)
Sex (M/F)340/354101/114346/332
Ethnicity (C/M/I/O)276/184/234/0215/0/0/0370/187/120/1*
Education (years)6.18 ± 4.63 (0−22)12.02 ± 3.45 (0−21)N/A
MMSE score24.13 ± 3.59 (10−30)28.29 ± 1.27 (26−30)N/A
Imaging follow-up (years)N/A4.00 ± 3.33 (0−9.59)3.49 ± 2.41 (0−6.69)
Cognition sample sizeN=694N=81–212N=217–239
Appendix 1—table 1
Optimized initial learning rates for each dataset and fold.

EDIS, Epidemiology of Dementia in Singapore; SLABS, Singapore Longitudinal Aging Brain Study; GUSTO, Growing Up in Singapore Towards healthy Outcomes.

FoldEDISSLABSGUSTO
01e-41e-41e-3
11e-41e-41e-3
21e-41e-51e-3
31e-41e-51e-3
41e-51e-41e-3
51e-41e-41e-3
61e-51e-51e-3
71e-41e-51e-3
81e-41e-51e-3
91e-41e-41e-3
Appendix 2—table 1
Participant cognitive characteristics at baseline.

EDIS was cross-sectional, while SLABS and GUSTO were longitudinal. EDIS, Epidemiology of Dementia in Singapore; SLABS, Singapore Longitudinal Aging Brain Study; GUSTO, Growing Up in Singapore Towards healthy Outcomes; KBIT-2, Kaufman Brief Intelligence Test Second Edition; WCST, Wisconsin Card Sorting Test; NEPSY-II, A Developmental Neuropsychological Assessment Second Edition.

CharacteristicEDIS (N=694)
Global cognition z-score−2.55 ± 2.21 (−10.47 − 1.96)
Executive function domain z-score−1.57 ± 1.85 (−7.79 − 1.26)
Attention domain z-score−2.13 ± 2.63 (−13.69 − 2.65)
Language domain z-score−1.84 ± 1.70 (−11.72 − 2.85)
Visuomotor speed domain z-score− 1.64 ± 1.76 (−5.87 − 2.25)
Visuoconstruction domain z-score−2.50 ± 2.57 (−13.63 − 2.37)
Verbal memory domain z-score−1.40 ± 1.40 (−5.17 − 2.78)
Visual memory domain z-score−1.50 ± 1.48 (−9.10 −2.15)
CharacteristicSLABS (N=81)
Global cognition T-score51.82 ± 4.75 (41.21 − 61.65)
Executive function domain T-score51.98 ± 5.57 (42.39 − 64.08)
Attention domain T-score50.52 ± 6.11 (39.62 − 65.04)
Processing speed domain T-score52.96 ± 7.22 (36.13 − 71.84)
Verbal memory domain T-score52.18 ± 7.70 (32.38 − 67.04)
Visuospatial memory domain T-score51.48 ± 7.62 (35.00 − 65.80)
Cognition follow-up (years)7.83 ± 0.97 (5.58 − 9.59)
CharacteristicGUSTO (N=217–239)
KBIT-2 Composite IQ Standard Score (N=217, 4.5 years old)92.38 ± 14.24 (52 − 132)
WCST Total Errors Standard Score (N=220, 8.5 years old)99.45 ± 15.95 (64 − 136)
NEPSY-II Naming Scaled Score (N=239, 8.5 years old)10.45 ± 3.66 (3 − 19)
NEPSY-II Inhibition Scaled Score (N=239, 8.5 years old)10.31 ± 3.3 (2 − 18)
NEPSY-II Switching Scaled Score (N=239, 8.5 years old)9.26 ± 4.03 (1 − 19)
Appendix 2—table 2
Model equations for analyzing associations with cognition.

Cog, standardized cognitive score; BAG, brain age gap; bl, baseline; ∆, annual rate of change; EDIS, Epidemiology of Dementia in Singapore; SLABS, Singapore Longitudinal Aging Brain Study; GUSTO, Growing Up in Singapore Towards healthy Outcomes.

AgeDatasetLabelEquation
ElderlyEDISI. Baseline vs. baselineCogbl=BAGbl+age+sex+education
ElderlySLABSII. Baseline vs. changeΔCoglong=BAGbl+age+sex+education
ElderlySLABSIII. Change vs. change (overlapping)ΔCoglong=ΔBAGearly+BAGbl+age+sex+education
ElderlySLABSIV. Change vs. change (future)ΔCogfuture=ΔBAGearly+BAGbl+age+sex+education
ChildrenGUSTOI. Baseline vs. baselineCog4.5=BAG4.5+age+sex
ChildrenGUSTOII. Baseline vs. futureCog8.5=BAG4.5+age+sex
ChildrenGUSTOIII. Change vs. futureCog8.5=ΔBAG4.5−7.5+BAG4.5+age+sex
Appendix 3—table 1
Pretrained baseline vs. baseline results from EDIS.

p-values are bolded if less than α = 0.05. BAG, brain age gap; EDIS, Epidemiology of Dementia in Singapore; β, standardized regression coefficient; CIL, lower limit of 95% CI; CIU, upper limit of 95% CI; p, uncorrected p-value; pcorr, corrected p-value; ΔRadj2, change in adjusted R2 when adding variable of interest; R2, model coefficient of determination.

Variable of interestOutcomeβ (CIL, CIU)ppcorrΔ R2adjR2
Baseline BAG (pretrained)Baseline global cognition–0.1125 (–0.17, –0.06)<0.00010.00060.01060.5011
Baseline BAG (pretrained)Baseline executive function–0.1029 (–0.17, –0.04)0.00190.00760.00850.3297
Baseline BAG (pretrained)Baseline attention–0.0404 (–0.11, 0.03)0.24610.24610.00040.2562
Baseline BAG (pretrained)Baseline language–0.1145 (–0.18, –0.05)0.00090.00470.01070.2677
Baseline BAG (pretrained)Baseline visuomotor speed–0.0825 (–0.14, –0.02)0.00520.01360.00530.4684
Baseline BAG (pretrained)Baseline visuo-construction–0.0896 (–0.15, –0.03)0.00450.01360.00630.3911
Baseline BAG (pretrained)Baseline verbal memory–0.1096 (–0.17, –0.05)0.00060.00340.00990.3826
Baseline BAG (pretrained)Baseline visual memory–0.1395 (–0.20, –0.08)< 0.00010.00020.01650.3515
Appendix 3—table 2
Finetuned baseline vs. baseline results from EDIS.

p-values are bolded if less than α = 0.05. BAG, brain age gap; EDIS, Epidemiology of Dementia in Singapore; β, standardized regression coefficient; CIL, lower limit of 95% CI; CIU, upper limit of 95% CI; p, uncorrected p-value; pcorr, corrected p-value; ΔRadj2, change in adjusted R2 when adding variable of interest; R2, model coefficient of determination.

Variable of interestOutcomeβ (CIL, CIU)ppcorrΔ R2adjR2
Baseline BAG (finetuned)Baseline global cognition–0.1661 (–0.23, –0.10)< 0.0001< 0.00010.01710.5076
Baseline BAG (finetuned)Baseline executive function–0.1607 (–0.24, –0.08)<0.00010.00020.01580.3369
Baseline BAG (finetuned)Baseline attention–0.0632 (–0.14, 0.02)0.12230.12230.00150.2573
Baseline BAG (finetuned)Baseline language–0.1535 (–0.23, –0.07)0.00020.00070.01420.2712
Baseline BAG (finetuned)Baseline visuomotor speed–0.1071 (–0.17, –0.04)0.00200.00400.00660.4697
Baseline BAG (finetuned)Baseline visuoconstruction–0.1291 (–0.20, –0.06)0.00050.00150.00990.3946
Baseline BAG (finetuned)Baseline verbal memory–0.1722 (–0.24, –0.10)<0.0001<0.00010.01830.3910
Baseline BAG (finetuned)Baseline visual memory–0.2194 (–0.29, –0.15)<0.0001<0.00010.03020.3651
Appendix 3—table 3
Pretrained baseline vs. baseline results from SLABS.

p-values are bolded if less than α = 0.05. BAG, brain age gap; SLABS, Singapore Longitudinal Aging Brain Study; β, standardized regression coefficient; CIL, lower limit of 95% CI; CIU, upper limit of 95% CI; p, uncorrected p-value; pcorr, corrected p-value; ΔRadj2, change in adjusted R2 when adding variable of interest; R2, model coefficient of determination.

Variable of interestOutcomeβ (CIL, CIU)ppcorrΔR2adjR2
Baseline BAG (pretrained)Baseline global cognition–0.0574 (–0.18, 0.06)0.33761.0000–0.00020.3514
Baseline BAG (pretrained)Baseline executive function–0.0447 (–0.16, 0.07)0.46001.0000–0.00150.3379
Baseline BAG (pretrained)Baseline verbal memory–0.0089 (–0.14, 0.12)0.89251.0000–0.00380.2090
Baseline BAG (pretrained)Baseline visual memory–0.1299 (–0.27, 0.01)0.07160.42970.01050.0643
Baseline BAG (pretrained)Baseline attention–0.0187 (–0.15, 0.11)0.77501.0000–0.00350.2261
Baseline BAG (pretrained)Baseline processing speed–0.0036 (–0.12, 0.12)0.94921.0000–0.00290.4038
Appendix 3—table 4
Finetuned baseline vs. baseline results from SLABS.

p-values are bolded if less than α = 0.05. BAG, brain age gap; SLABS, Singapore Longitudinal Aging Brain Study; β, standardized regression coefficient; CIL, lower limit of 95% CI; CIU, upper limit of 95% CI; p, uncorrected p-value; pcorr, corrected p-value; ΔRadj2, change in adjusted R2 when adding variable of interest; R2, model coefficient of determination.

Variable of interestOutcomeβ (CIL, CIU)ppcorrΔR2adjR2
Baseline BAG (finetuned)Baseline global cognition–0.0674 (–0.19, 0.05)0.27841.00000.00060.3522
Baseline BAG (finetuned)Baseline executive function–0.0378 (–0.16, 0.09)0.54771.0000–0.00210.3373
Baseline BAG (finetuned)Baseline verbal memory–0.0548 (–0.19, 0.08)0.42451.0000–0.00140.2113
Baseline BAG (finetuned)Baseline visual memory–0.1243 (–0.27, 0.02)0.09740.58460.00810.0621
Baseline BAG (finetuned)Baseline attention0.0056 (–0.13, 0.14)0.93421.0000–0.00380.2258
Baseline BAG (finetuned)Baseline processing speed0.0229 (–0.14, 0.09)0.70081.0000–0.00250.4042
Appendix 3—table 5
Pretrained baseline vs. change results from SLABS.

p-values are bolded if less than α = 0.05. BAG, brain age gap; SLABS, Singapore Longitudinal Aging Brain Study; β, standardized regression coefficient; CIL, lower limit of 95% CI; CIU, upper limit of 95% CI; p, uncorrected p-value; pcorr, corrected p-value; ΔRadj2, change in adjusted R2 when adding variable of interest; R2, model coefficient of determination.

Variable of interestOutcomeβ (CIL, CIU)ppcorrΔR2adjR2
Baseline BAG (pretrained)Change in global cognition–0.1213 (–0.36, 0.12)0.32321.0000–0.00020.0211
Baseline BAG (pretrained)Change in executive function–0.2477 (–0.48, –0.01)0.04060.24330.04240.0711
Baseline BAG (pretrained)Change in verbal memory–0.1263 (–0.37, 0.11)0.29701.00000.00130.0490
Baseline BAG (pretrained)Change in visual memory0.0337 (–0.21, 0.28)0.78151.0000–0.01220.0338
Baseline BAG (pretrained)Change in attention0.0253 (–0.21, 0.26)0.83451.0000–0.01250.0457
Baseline BAG (pretrained)Change in processing speed–0.1032 (–0.35, 0.14)0.40151.0000–0.00390.0157
Appendix 3—table 6
Finetuned baseline vs. change results from SLABS.

p-values are bolded if less than α = 0.05. BAG, brain age gap; SLABS, Singapore Longitudinal Aging Brain Study; β, standardized regression coefficient; CIL, lower limit of 95% CI; CIU, upper limit of 95% CI; p, uncorrected p-value; pcorr, corrected p-value; ΔRadj2 , change in adjusted R2 when adding variable of interest; R2, model coefficient of determination.

Variable of interestOutcomeβ (CIL, CIU)ppcorrΔR2adjR2
Baseline BAG (finetuned)Change in global cognition–0.0480 (–0.30, 0.20)0.70411.0000–0.01160.0103
Baseline BAG (finetuned)Change in executive function–0.1165 (–0.36, 0.13)0.35311.0000–0.00170.0292
Baseline BAG (finetuned)Change in verbal memory–0.1431 (–0.39, 0.10)0.24911.00000.00450.0520
Baseline BAG (finetuned)Change in visual memory0.0409 (–0.21, 0.29)0.74311.0000–0.01180.0342
Baseline BAG (finetuned)Change in attention0.1232 (–0.13, 0.37)0.31931.00000.00010.0576
Baseline BAG (finetuned)Change in processing speed–0.0125 (–0.26, 0.23)0.92141.0000–0.01340.0066
Appendix 3—table 7
Pretrained change vs. change results from SLABS.

p-values are bolded if less than α = 0.05. BAG, brain age gap; SLABS, Singapore Longitudinal Aging Brain Study; β, standardized regression coefficient; CIL, lower limit of 95% CI; CIU, upper limit of 95% CI; p, uncorrected p-value; pcorr, corrected p-value; ΔRadj2, change in adjusted R2 when adding variable of interest; R2, model coefficient of determination.

Variable of interestOutcomeβ (CIL, CIU)ppcorrΔR2adjR2
Change in BAG (pretrained)Change in global cognition–0.1415 (–0.39, 0.11)0.26891.00000.00330.0370
Change in BAG (pretrained)Change in executive function–0.3807 (–0.61, –0.15)0.00170.01000.11000.1864
Change in BAG (pretrained)Change in verbal memory–0.0312 (–0.28, 0.22)0.80561.0000–0.01250.0498
Change in BAG (pretrained)Change in visual memory0.0917 (–0.16, 0.34)0.47181.0000–0.00640.0405
Change in BAG (pretrained)Change in attention–0.1872 (–0.44, 0.06)0.13710.68570.01640.0736
Change in BAG (pretrained)Change in processing speed–0.0868 (–0.34, 0.17)0.50011.0000–0.00740.0217
Appendix 3—table 8
Finetuned change vs. change results from SLABS.

p-values are bolded if less than α = 0.05. BAG, brain age gap; SLABS, Singapore Longitudinal Aging Brain Study; β, standardized regression coefficient; CIL, lower limit of 95% CI; CIU, upper limit of 95% CI; p, uncorrected p-value; pcorr, corrected pvalue; ΔRadj2, change in adjusted R2 when adding variable of interest; R2, model coefficient of determination.

Variable of interestOutcomeβ (CIL, CIU)ppcorrΔR2adjR2
Change in BAG (finetuned)Change in global cognition–0.2367 (–0.48, 0.01)0.05760.23050.03600.0570
Change in BAG (finetuned)Change in executive function–0.3861 (–0.62, –0.15)0.00140.00840.11900.1536
Change in BAG (finetuned)Change in verbal memory–0.1017 (–0.35, 0.14)0.40911.0000–0.00410.0607
Change in BAG (finetuned)Change in visual memory–0.0284 (–0.28, 0.22)0.82001.0000–0.01280.0349
Change in BAG (finetuned)Change in attention–0.2651 (–0.50, –0.03)0.02870.14340.04930.1163
Change in BAG (finetuned)Change in processing speed–0.0842 (–0.33, 0.17)0.50501.0000–0.00760.0125
Appendix 3—table 9
Pretrained baseline vs. future results from GUSTO.

p-values are bolded if less than α = 0.05. BAG, brain age gap; GUSTO, Growing Up in Singapore Towards healthy Outcomes; NEPSY-II, A Developmental Neuropsychological Assessment Second Edition; WCST, Wisconsin Card Sorting Test; β, standardized regression coefficient; CIL, lower limit of 95% CI; CIU, upper limit of 95% CI; p, uncorrected p-value; pcorr, corrected pvalue; ΔRadj2, change in adjusted R2 when adding variable of interest; R2, model coefficient of determination.

Variable of interestOutcomeβ (CIL, CIU)ppcorrΔR2adjR2
Baseline BAG (pretrained)Future WCST Standard Score–0.0585 (–0.24, 0.13)0.53340.9798–0.00280.0256
Baseline BAG (pretrained)Future Naming (NEPSY-II)0.1290 (–0.05, 0.31)0.15730.62920.00430.0149
Baseline BAG (pretrained)Future Inhibition (NEPSY-II)0.1074 (–0.07, 0.29)0.23870.71610.00170.0155
Baseline BAG (pretrained)Future Switching (NEPSY-II)0.0630 (–0.12, 0.24)0.48990.9798–0.00220.0106
Appendix 3—table 10
Finetuned baseline vs. future results from GUSTO.

p-values are bolded if less than α = 0.05. BAG, brain age gap; GUSTO, Growing Up in Singapore Towards healthy Outcomes; NEPSY-II, A Developmental Neuropsychological Assessment Second Edition; WCST, Wisconsin Card Sorting Test; β, standardized regression coefficient; CIL, lower limit of 95% CI; CIU, upper limit of 95% CI; p, uncorrected p-value; pcorr, corrected p-value; ΔRadj2 , change in adjusted R2 when adding variable of interest; R2, model coefficient of determination.

Variable of interestOutcomeβ (CIL, CIU)ppcorrΔR2adjR2
Baseline BAG (finetuned)Future WCST Standard Score–0.0046 (–0.20, 0.20)0.96421.0000–0.00460.0239
Baseline BAG (finetuned)Future Naming (NEPSY-II)–0.0108 (–0.21, 0.19)0.91431.0000–0.00420.0065
Baseline BAG (finetuned)Future Inhibition (NEPSY-II)0.0829 (–0.11, 0.28)0.40861.0000–0.00130.0125
Baseline BAG (finetuned)Future Switching (NEPSY-II)–0.0359 (–0.23, 0.16)0.72031.0000–0.00370.0091
Appendix 3—table 11
Pretrained change vs. future results from GUSTO.

p-values are bolded if less than α = 0.05. BAG, brain age gap; GUSTO, Growing Up in Singapore Towards healthy Outcomes; NEPSY-II, A Developmental Neuropsychological Assessment Second Edition; WCST, Wisconsin Card Sorting Test; β, standardized regression coefficient; CIL, lower limit of 95% CI; CIU, upper limit of 95% CI; p, uncorrected p-value; pcorr, corrected p-value; ΔRadj2, change in adjusted R2 when adding variable of interest; R2, model coefficient of determination.

Variable of interestOutcomeβ (CIL, CIU)ppcorrΔR2adjR2
Change in BAG (pretrained)Future WCST Standard Score–0.0639 (–0.21, 0.08)0.38771.0000–0.00110.0290
Change in BAG (pretrained)Future Naming (NEPSY-II)–0.0280 (–0.17, 0.12)0.70201.0000–0.00360.0156
Change in BAG (pretrained)Future Inhibition (NEPSY-II)0.0376 (–0.11, 0.18)0.60731.0000–0.00310.0166
Change in BAG (pretrained)Future Switching (NEPSY-II)0.0021 (–0.14, 0.15)0.97721.0000–0.00430.0106
Appendix 3—table 12
Finetuned change vs. future results from GUSTO.

p-values are bolded if less than α = 0.05. BAG, brain age gap; GUSTO, Growing Up in Singapore Towards healthy Outcomes; NEPSY-II, A Developmental Neuropsychological Assessment Second Edition; WCST, Wisconsin Card Sorting Test; β, standardized regression coefficient; CIL, lower limit of 95% CI; CIU, upper limit of 95% CI; p, uncorrected p-value; pcorr, corrected pvalue; ΔRadj2, change in adjusted R2 when adding variable of interest; R2, model coefficient of determination.

Variable of interestOutcomeβ (CIL, CIU)ppcorrΔR2adjR2
Change in BAG (pretrained)Future WCST Standard Score–0.0091 (–0.16, 0.14)0.90521.0000–0.00450.0239
Change in BAG (pretrained)Future Naming (NEPSY-II)0.0158 (–0.14, 0.17)0.84131.0000–0.00410.0067
Change in BAG (pretrained)Future Inhibition (NEPSY-II)0.2006 (0.05, 0.35)0.01030.04110.02370.0400
Change in BAG (pretrained)Future Switching (NEPSY-II)0.1795 (0.03, 0.33)0.02210.06630.01810.0311
Author response table 1
Linear relationship between pretrained baseline BAG and future cognitive score in ADNI.

Compare to Supplementary Tables S4 – S15 of the original text.

Variable of InterestOutcomeβ (CIL, CIU)pPcorrΔ R2adjR2
Baseline BAG (pretrained)Future Memory-0.1744 (-0.36, 0.01)0.06320.24420.01810.1278
Baseline BAG (pretrained)Future Executive Function-0.1368 (-0.32, 0.05)0.15220.30430.00810.0914
Baseline BAG (pretrained)Future Language-0.1758 (-0.36, 0.01)0.06110.24420.01850.1277
Baseline BAG (pretrained)Future Visuospatial Function-0.0284 (-0.26, 0.20)0.80530.8053-0.00980.0559
Author response table 2
Linear relationship between finetuned baseline BAG and future cognitive score in ADNI.

Compare to Supplementary Tables S4 – S15 of the original text.

Variable of InterestOutcomeβ (CIL, CIU)pPcorrΔ R2adjR2
Baseline BAG (finetuned)Future Memory-0.1260 (-0.33, 0.07)0.21550.86210.00400.1142
Baseline BAG (finetuned)Future Executive Function0.0244 (-0.18, 0.23)0.81351.0000-0.00720.0766
Baseline BAG (finetuned)Future Language-0.0823 (-0.28, 0.12)0.41931.0000-0.00250.1073
Baseline BAG (finetuned)Future Visuospatial Function0.0536 (-0.19, 0.30)0.66481.0000-0.00850.0572
Author response table 3
Linear relationship between pretrained change in BAG and future cognitive score in ADNI.

Compare to Supplementary Tables S4 – S15 of the original text.

Variable of InterestOutcomeβ (CIL, CIU)pPcorrΔ R2adjR2
Change in BAG (pretrained)Future Memory-0.1743 (-0.35, -0.00)0.04920.19680.02090.1549
Change in BAG (pretrained)Future Executive Function-0.0711 (-0.25, 0.11)0.43530.8705-0.00290.0959
Change in BAG (pretrained)Future Language-0.0577 (-0.23, 0.12)0.51800.8705-0.00420.1306
Change in BAG (pretrained)Future Visuospatial Function-0.1337 (-0.35, 0.08)0.21260.63770.00610.0718
Author response table 4
Linear relationship between finetuned change in BAG and future cognitive score in ADNI.

Compare to Supplementary Tables S4 – S15 of the original text.

Variable of InterestOutcomeβ (CIL, CIU)ppcorrΔ R2adjR2
Change in BAG (finetuned)Future Memory-0.2263 (-0.40, -0.05)0.01130.04510.04000.1674
Change in BAG (finetuned)Future Executive Function-0.1453 (-0.33, 0.04)0.11530.23050.01170.0966
Change in BAG (finetuned)Future Language-0.1737 (-0.35, 0.00)0.05430.16280.02040.1403
Change in BAG (finetuned)Future Visuospatial Function-0.1290 (-0.34, 0.08)0.22440.23050.00530.0722

Additional files

MDAR checklist
https://cdn.elifesciences.org/articles/97036/elife-97036-mdarchecklist1-v1.pdf

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Susan F Cheng
  2. Wan Lin Yue
  3. Kwun Kei Ng
  4. Xing Qian
  5. Siwei Liu
  6. Trevor WK Tan
  7. Kim-Ngan Nguyen
  8. Ruth LF Leong
  9. Saima Hilal
  10. Ching-Yu Cheng
  11. Ai Peng Tan
  12. Evelyn C Law
  13. Peter D Gluckman
  14. Christopher Li-Hsian Chen
  15. Yap Seng Chong
  16. Michael J Meaney
  17. Michael WL Chee
  18. BT Thomas Yeo
  19. Juan Helen Zhou
(2025)
Rate of brain aging associates with future executive function in Asian children and older adults
eLife 13:RP97036.
https://doi.org/10.7554/eLife.97036.3