Table 4:

Overview of prospective observation studies investigating level-change or correlated change associations of whole brain volume/CSF and cognitive abilities.

Brain measure/author, study	Sample size at baseline (age range in years)	Measurement occasions	Time intervals in years	Healthy cognitive aging:screening instruments at baseline	Statistical method (change measure)	Covariates	Neuroanatomical measure (cf. Table 1)	Brain regions	Cognitive variables	Brain-cognition relations (+= positive, −=negative)
Whole brain
Schmidt et al. (2005), Austrian Stroke Prevention Study	329 (50–75)	2–3	3	Nonea	Repeated measures regression analysis (change ratiobrain)	Age Sex Education (years) VR Diabetes Cardiac disease Smoking Cholesterol	Whole brain volume (NBV)	-	Episodic memory Lern- und Gedächtnistest (immediate, delayed recall) Reasoning WCST Attention and speed Alters-Konzentrations-Test TMT B Digit span Complex reaction time test	Level-change Not reported Correlated change ΔNBV + Δepisodic memory, Δattention and speed
Charlton et al. (2010), GENIE	106 (50–90)	2	2	Nonea	Correlation (change ratiobrain/cog)	Age	Whole brain volume (NBV)	-	Working memory Digit span backwards, Letter-number sequencing	Level-change No effect Correlated change No effect
Grimm et al. (2012), BLSA	149 (56–86)	6.8 ± 2.8	1	Nonea	Bivariate LCSbbrain/cog	None	Ventricular volume (CSF)	Lateral ventricles	Episodic memory CLVT	Level-change No effect (in final model) Correlated change (lagged) ΔLateral ventricle volume (previous) − ΔCLVT (subsequent)
Mak et al. (2015a)	33 normal controls (76.7 ± 5.3)	2	1	No history of dementia or CAMCOG score <80	Correlation (change ratiobrain/cog)	Age Sex	Whole brain volume (TBV)	-	Global cognition CAMCOG	Level-change No effect Correlated change No effect
Ritchie et al. (2015b), Lothian Birth Cohort	657 (M age 73)	2	3	MMSE ≥ 24	Multivariate LCScbrain/cog	Age, Sex	Whole brain volume (TBV)	-	Fluid intelligence Matrix reasoning Block design Digit span backward Letter-number sequencing Episodic memory Logical memory (immediate, delayed recall) Verbal paired associates Spatial span (forward, backward) Processing speed Digit-symbol Symbol search Simple-/four-choice reaction time task Inspection time task	Level-change Episodic memory T1 + ΔTBV Correlated change No effect
Leong et al. (2017), Singapore Longitudinal Aging Brain Study	111 (56.1–83.1)	2–4	1.5–2	MMSE ≥ 26	Correlation (MLM slope estimatesbrain/cog)	Age Sex Education VR BMI ICV	Whole brain volume (TBV) Ventricular volume (CSF)	-	Global cognition Composite of all abilities below EF Category fluency (three categories) Design fluency TMT B Processing speed TMT A Symbol search Symbol-digit modalities Visuospatial memory Visual paired associates test (immediate, delayed recall) Verbal memory AVLT (immediate, delayed recall) Attention Digit span Spatial span (forward and backward)	Level-change Not reported Correlated change ΔTBV+Δglobal cognition, Δverbal memory ΔTotal ventricular volume−Δverbal memory, ΔEF

brain, brain measurements; cog, cognitive measurements; GENIE, St GEorge’s Neuropsychology and Imaging in the Elderly; LCS, latent change score model; MLM, multilevel model; change ratio, a measure of change proportionalized on the baseline level and in some cases also on the time interval (for exact formula consult original publications); difference score, value at T1 – value at T2; CSF, cerebrospinal fluid; ICV, intracranial volume; TBV, total brain volume; NBV, normalized brain volume; TMT, trail making test; AVLT, Rey auditory verbal learning test; EF, executive functions; MMSE, mini-mental-state-examination; CAMCOG, Cambridge Cognitive Examination; VR, vascular risk (for details refer consult original publications); APOE-ε4, apolipoprotein E allele 4; BMI, body mass index; Δ, change.

^aNo information provided on screening instruments and/or cut-offs used (for description of healthy aging criteria consult original publication).

^bChange was modeled over age (instead of time; consult original publication for details).

^cAll cognitive variables were included in the same model.