Insights into the role of extracellular-to-total body water ratio on neurocognitive decrements over time in Asians with type 2 diabetes

Abstract

Background

Extracellular water-to-total body water (ECW/TBW) ratio reflects volume overload. It promotes aortic stiffness which predicts Alzheimer's disease and related dementias (ADRD) development.

Objective

We studied longitudinal relationship between ECW/TBW ratio and neurocognitive decrements in type 2 diabetes (T2D) and mediating role of aortic stiffness.

Methods

In a prospective Asian cohort, we measured ECW/TBW ratio using bioimpedance analysis (BIA) and pulse wave velocity (PWV) using applanation method. The Repeatable Battery for Assessment of Neuropsychological Status (RBANS) was used to assess neurocognitive performance. Linear mixed model was applied to examine if ECW/TBW was associated with neurocognitive performance

Results

The study comprised 1412 patients at baseline. Over maximum of 9.4 years’ follow-up, 823 patients had repeated neurocognitive assessment. We observed associations between higher ECW/TBW ratio (continuous variable) and edema group with lower total RBANS score. The adjusted coefficients were −1.43 (95%CI −2.31, −0.56) and −5.88 (95%CI −9.00, −2.75) respectively. There were similar observations for delayed memory, visuospatial/constructional and attention. Based on cross-lagged panel model, there was larger standardized effect size from baseline ECW/TBW ratio to follow-up total RBANS score (β = 0.053, p = 0.026) compared to that from baseline total RBANS score to follow-up ECW/TBW ratio (β = −0.020, p = 0.461). PWV mediated 13.3% of the relationship.

Conclusions

Our results demonstrated a relationship between volume overload and cognitive decrements, with potential mediation from aortic stiffness. There was temporal precedence of ECW/TBW over neurocognitive performance. BIA-determined ECW/TBW ratio is a promising biomarker of neurocognitive decrements in T2D.

Keywords

Alzheimer's disease arterial stiffness extracellular water neurocognitive performance type 2 diabetes

Introduction

Diabetes confers an elevated risk for cognitive decline, mild cognitive impairment, Alzheimer's disease and related dementias (ADRD).^1,2 This may be attributed to Insulin resistance, advanced glycation end-products accumulation, and mitochondrial dysfunction which contribute to neuroinflammation, oxidative stress resulting in tau phosphorylation, and neurovascular dysfunction.^3–5 There is increasing research spotlight on diabetes-related dementia, an emerging diabetic complication, as evidenced by the proliferation of journal publications on this topic.^1,5 Individuals with both type 2 diabetes (T2D) and higher dementia risk were found to have poorer quality of life.⁶ However, there remains a paucity of knowledge on the possible mechanisms underpinning the biological link between diabetes and cognitive decline that precedes ADRD. This warrants a continual pursuit for early biomarkers that will offer insights on the biological mechanisms, facilitate future therapeutic development and serve as monitoring targets.

The ratio of extracellular water-to-total body water (ECW/TBW) is a body composition marker indicating fluid balance that is maintained by an osmotic gradient.⁷ In diabetes, elevated glucose levels induce a hyperosmotic effect that promotes water movement from intracellular compartment to extracellular compartment, thereby increasing ECW/TBW ratio.^8,9

Increased ECW/TBW ratio reflects volume overload¹⁰ which promotes aortic stiffness.¹¹ Aortic stiffness, measured by applanation tonometry,¹² is an established risk factor for ADRD such as AD and vascular dementia.¹³ It exerts a high pulsatile force on the brain microvasculature, resulting in microvascular ischemia, white matter lesions and brain atrophy.¹³ These brain changes may contribute to cognitive decline.¹⁴ We have reported that ECW/TBW was inversely linked to neurocognitive performance in subjects with T2D.¹⁵ It is also possible that poor neurocognitive performance could lead to reduced fluid intake with resultant dehydration, reduction in TBW and increase in ECW/TBW ratio.¹⁶ However, this temporal relationship is yet to be established. Research on the biological mechanisms underpinning the relationship between ECW/TBW ratio and neurocognitive decrements over time is also lacking.

We therefore investigated if ECW/TBW was related to neurocognitive decrements over time and determined the mediating action of aortic stiffness on the relationship. The study was conducted on patients who took part in the Singapore Study of Macroangiopathy and Microvascular Reactivity Type 2 Diabetes (SMART2D).

Methods

Subject recruitment

We analyzed data from SMART2D which prospectively recruited patients with T2D from dedicated diabetes clinics in a tertiary referral hospital and general primary care clinics. SMART2D aims to study traditional and novel biomarkers which predict development and progression of diabetic complications. The details on the eligibility criteria for the study are found in our earlier publications.¹⁷ We received ethics approval from the National Healthcare Group Domain Specific Review Board (NHG DSRB) in Singapore. All the subjects who joined the study had provided informed consent. We complied with the ethical guidelines and standards from the Helsinki Declaration of 1975 and NHG DSRB. The final analysis included 1412 subjects after excluding 133 subjects younger than 45 years and 233 subjects without neurocognitive measurement.

We collected information on demographics such as age, gender and ethnicity, medical history and medications from questionnaire with verification from medical records. We checked for history of cardiovascular disease based on the presence of acute myocardial infarction, stroke, blockade of arteries, balloon angioplasty and/or cardiac bypass. Mean arterial pressure (MAP) was computed with the formula 1/3×((2×diastolic blood pressure) + systolic blood pressure).¹⁸ We used Geriatric Depression Scale (GDS) for depression screening. Potential depression is indicated by a score of 5 and above.¹⁹

The tertiary hospital laboratory performed the following laboratory measurements from the fasted blood and urine samples: hemoglobin A1c (HbA1c), low-density lipoprotein-cholesterol (LDL-C) and urine albumin with Tina-Quant Hemoglobin AIc Gen. test, enzymatic colorimetric test and immunoturbidimetric test respectively (Roche cobas^® c 501, Germany). The Chronic Kidney Disease-Epidemiology Collaboration (CKD-EPI) equation was applied to calculate estimated glomerular filtration rate (eGFR).²⁰ We determined Apolipoprotein E (APOE) ε4 allele positivity (i.e., presence of at least one allele) with OmniExpress-24 array.²¹

There were two timepoints of measurements. We measured exposure and outcome between 2014 and 2017, and repeated the measurements in the subsequent phase 2019 and 2025 as follows:

ECW/TBW measurement

We determined ECW/TBW ratio using bioimpedance analysis (BIA). BIA is a technology which quantifies electrical impedance and resistance of different body tissues such as body water, muscle and fat. This allows BIA to estimate body water distribution including extracellular, intracellular and total body water.²² BIA has been increasingly used to determine water distribution in clinical and research settings.²³ It is a convenient and non-invasive device to assess body composition, and has been well validated with deuterium dilution, the gold standard approach to measure body water.^24,25 To reduce measurement bias, all the subjects had to undergo fasting for at least 8 h before BIA measurement. In normal conditions, ECW/TBW ratio is maintained at 0.30 to 0.38. The ratio ranges 0.39 to 0.40 for slight edema. Edema is present if ECW/TBW ratio exceeds 0.40.^10,15,26,27

Cognitive performance measurement

The Repeatable Battery for Assessment of Neuropsychological Status (RBANS) was utilized to quantify neurocognitive performance.²⁸ This locally validated test contains 12 subtests to measure domains of immediate memory, visuospatial/constructional, attention, language, and delayed memory.^28,29

Aortic stiffness measurement

The criterion measure of aortic stiffness is carotid-femoral pulse wave velocity (PWV).³⁰ We quantified PWV by placing applanation tonometer over carotid and femoral arteries and measuring time delay between the waveforms reaching these two sites. We then calculated PWV as the length between the two sites over pulse transit time (m/s).¹² The procedure was carried out non-invasively with SphygmoCor ^® (AtCor Medical, Sydney, Australia). The PWV data in 2014–2017 was analyzed in the mediation analysis.

Statistical analysis

STATA 14 was used for statistical analysis. Comparison of continuous baseline characteristics by no edema, slight edema and edema groups was made with one-way ANOVA or Kruskal-Wallis test. Comparison of categorical baseline characteristics by the same groups was done with Chi-Square test. At baseline, we analyzed the association between exposure and outcome using linear regression which included demographics, education < 7 years, APOE ɛ4 positivity, body mass index (BMI), MAP, HbA1c, LDL-C, eGFR, log-transformed urine albumin-to-creatinine ratio (uACR), GDS ≥ 5, history of cardiovascular disease and medications (metformin, diuretic (e.g., frusemide and hydrochlorothiazide) and sodium-glucose cotransporter 2 inhibitor (SGLT2i)) as adjusting covariates. We considered the biological plausibility and results on univariable association with neurocognitive performance in the selection of these covariates. ECW/TBW ratio was included as continuous and categorical exposure variable in the analysis. Analysis of ECW/TBW ratio as a continuous variable enables us to retain statistical power and precision of information.³¹ Analysis of ECW/TBW ratio as a categorical variable facilitates easier interpretation of results and clearer communication of findings^31,32 as the cut-off points for no edema, slight edema and edema were also used in previous studies.^15,26 Linear mixed model was applied to determine the longitudinal relationship between ECW/TBW ratio and neurocognitive decrements over time. We included the same covariates for adjustment.

Unlike linear mixed model which only informs whether ECW/TBW ratio is associated with neurocognitive decrements over time, cross-lagged panel model (CLPM) allows us to examine the bi-directional relationship between ECW/TBW ratio and neurocognitive performance over time (i.e., whether ECW/TBW ratio precedes neurocognitive performance or vice versa). CLPM is an analytical approach based on structural equation modelling³³ and is key to understanding the temporality of the relationship. We used CLPM with fixed intercept rather than random intercept as we are more interested in understanding prospective inter-individual neurocognitive effects from baseline ECW/TBW ratio rather than inter-individual effects. Nonetheless, we acknowledge that the results from fixed intercept CLPM could be biased. However, with only two time points for our exposure (ECW/TBW ratio) and outcome (RBANS), we could only use fixed intercept CLPM as random intercept CLPM requires more than two time points.³⁴ The median duration of follow-up is 4.9 years (IQR 4.2–5.9, indicating that time interval is around 5 years for most of the study participants. Previous research using CLPM has also taken a 5-year duration to observe lagged effects.³⁵ The results of CLPM may not be reliable if the time interval is too long.³⁶ Although time interval of <5 years will be better for the assumptions of CLPM to hold, the duration may not suffice for ECW/TBW ratio effect on RBANS to be detected. Model fit statistics were used for model evaluation, with Root Mean Square of Approximation <0.06, Tucker Lewis Index >0.90 and Comparative Fit Index >0.90 indicating that the model was well fitted.³⁷

We applied Kenny and Baron's approach³⁸ to assess mediation by PWV. The approach required us to determine the following relationships: (1) edema and PWV; (2) PWV and cognitive decrements; and (3) edema and cognitive decrements. Additionally, (4) the edema–cognitive decrements relationship must be attenuated by the adjustment for PWV in the model.

Results

The study comprised 1412 subjects. Their mean age was 61.5 (SD 8.1). In Table 1, there were slightly more males than females. The corresponding proportions of Chinese, Malay and Indian ethnicities were 51.4%, 19.2% and 22.8%. More than 30% subjects received < 7 years’ education. There were 19.4% of subjects with no edema, 68.6% with slight edema and 12.0% with edema. There was an increase in age and proportion of Malay ethnicity from non-edema to slight edema and edema groups. Compared to the non-edema group, the slight edema and edema groups had poorer control of BMI, HbA1c, MAP, uACR and eGFR. They also had higher PWV and lower global and multi-domain neurocognitive performance.

Table 1.

Baseline characteristics stratified by ECW/TBW ratio categories.

	All	No edema	Slight edema	Edema	p
Number	1412	274 (19.4)	968 (68.6)	170 (12.0)
ECW/TBW ratio	0.39 ± 0.01	0.38 ± 0.00	0.39 ± 0.01	0.41 ± 0.01	<0.001
Age (y)	61.5 ± 8.1	57.0 ± 6.9	62.0 ± 7.7	65.6 ± 8.7	<0.001
Male (%)	739 (52.3)	196 (71.5)	456 (47.1)	87 (51.2)	<0.001
Ethnicity (%)					<0.001
Chinese	763 (54.1)	155 (56.8)	532 (55.0)	76 (44.7)
Malay	270 (19.2)	38 (13.9)	174 (18.0)	58 (34.1)
Indian	321 (22.8)	69 (25.3)	221 (22.9)	31 (18.2)
Other	56 (4.0)	11 (4.0)	40 (4.1)	5 (2.9)
Education < 7 y (%)	455 (32.3)	53 (19.3)	339 (35.1)	63 (37.1)	<0.001
GDS ≥ 5 (%)	61 (4.3)	8 (2.9)	40 (4.2)	13 (7.7)	0.052
Diabetes duration (y)	15.2 ± 9.1	11.7 ± 7.2	15.5 ± 9.1	19.1 ± 10.2	<0.001
BMI (kg/m²)	27.2 ± 4.9	26.8 ± 3.9	27.0 ± 4.8	29.0 ± 6.3	<0.001
MAP (mmHg)	99.6 ± 10.4	98.4 ± 8.9	99.4 ± 10.1	103.0 ± 12.9	<0.001
HbA1c (%)	7.9 ± 1.6	7.6 ± 1.2	7.9 ± 1.6	8.1 ± 2.0	0.002
LDL-C (mmol/l)	2.7 ± 0.8	2.7 ± 0.7	2.7 ± 0.8	2.7 ± 1.0	0.852
eGFR (ml/min/1.73m²)	78.5 ± 27.6	88.9 ± 18.5	79.8 ± 26.4	54.4 ± 32.6	<0.001
uACR (mg/g)	28.1 (8.0–137.0)	13.5 (5.7–49.0)	28.0 (8.0–118.5)	250.1 (42.8–1207.0)	<0.001
PWV (m/s)	10.6 ± 3.7	9.2 ± 2.4	10.7 ± 3.9	12.2 ± 3.8	<0.001
Presence of APOE ε4 allele (%)	204 (17.7)	36 (15.5)	137 (17.5)	31 (22.6)	0.217
History of cardiovascular disease (%)	258 (18.9)	34 (12.6)	168 (18.0)	56 (34.4)	<0.001
Use of metformin (%)	1154 (82.0)	237 (87.1)	805 (83.4)	112 (65.9)	<0.001
Use of diuretics (%)	365 (41.2)	30 (21.1)	239 (39.8)	96 (67.6)	<0.001
Use of SGLT2i (%)	385 (34.1)	86 (39.5)	266 (34.5)	33 (23.6)	0.008
RBANS
Total	100.3 ± 8.8	104.3 ± 7.2	100.1 ± 8.6	94.9 ± 9.3	<0.001
Immediate memory	101.3 ± 11.7	105.4 ± 10.5	101.0 ± 11.8	96.5 ± 11.4	<0.001
Delayed memory	100.0 ± 12.1	103.8 ± 9.5	99.7 ± 12.3	95.2 ± 12.8	<0.001
Visuo-spatial/ Constructional	100.0 ± 11.9	105.0 ± 7.4	99.7 ± 11.4	93.7 ± 16.7	<0.001
Language	100.0 ± 11.4	101.6 ± 11.1	100.4 ± 10.8	95.7 ± 14.0	<0.001
Attention	100.1 ± 12.1	105.9 ± 10.3	99.6 ± 11.9	93.1 ± 11.6	<0.001

ECW/TBW: extracellular-to-total body water; GDS: Geriatric Depression Scale; BMI: body mass index; MAP: mean arterial pressure; HbA1c: hemoglobin A1c; LDL-C: low-density lipoprotein cholesterol; eGFR: estimated glomerular filtration rate; uACR: urinary albumin-to-creatinine ratio; PWV: pulse wave velocity; APOE: Apolipoprotein E; SGLT2i: sodium-glucose cotransporter 2 inhibitor; RBANS: Repeatable Battery for the Assessment of Neuropsychological Status

In Table 2, negative correlation was noted between ECW/TBW ratio with eGFR and RBANS score (total and all domains). There were positive correlations between ECW/TBW ratio with age, diabetes duration, MAP, uACR, BMI and PWV.

Table 2.

Spearman correlation between ECW/TBW ratio and baseline characteristics (N = 1412).

Variable	Correlation coefficient	p
Age	0.333	<0.001
Diabetes duration	0.261	<0.001
BMI	0.133	<0.001
MAP	0.090	<0.001
HbA1c	0.036	0.178
LDL-C	−0.051	0.056
eGFR	−0.281	<0.001
uACR	0.295	<0.001
PWV	0.316	<0.001
RBANS
Total	−0.337	<0.001
Immediate memory	−0.236	<0.001
Delayed memory	−0.225	<0.001
Visuo-spatial/ Constructional	−0.310	<0.001
Language	−0.143	<0.001
Attention	−0.344	<0.001

ECW/TBW: extracellular-to-total body water; BMI: body mass index; MAP: mean arterial pressure; HbA1c: hemoglobin A1c; LDL-C: low-density lipoprotein cholesterol; eGFR: estimated glomerular filtration rate; uACR: urinary albumin-to-creatinine ratio; PWV: pulse wave velocity; RBANS: Repeatable Battery for the Assessment of Neuropsychological Status

For each SD increment of ECW/TBW ratio at baseline, there was decrement of total RBANS score by −2.79 (95%CI −3.23, −2.35) in the unadjusted Model 1 (Table 3). When we included age, gender, education, APOE ɛ4 positivity, BMI, MAP, HbA1c, LDL-C, eGFR and uACR in Model 2, ECW/TBW ratio remained inversely associated with total score (coefficient −1.26, 95%CI −1.97, −0.55). In Model 3 that additionally included GDS, history of cardiovascular disease and medications, the finding was similar with coefficient −1.17 (95%CI −1.90, −0.44). With no edema as reference group, the slight edema and edema groups experienced decreased total score in the unadjusted and adjusted models. In Model 3, there were total score decrements by −2.21 (95%CI −3.90, −0.52) and −4.52 (95%CI −6.98, −2.05) in the slight edema and edema groups respectively.

Table 3.

Association between ECW/TBW ratio and RBANS total score at baseline (N = 1412).

	Coefficient (95% CI) p
	Model 1^a	Model 2^b	Model 3^c
ECW/TBW ratio (per SD increment)	−2.79 (−3.23 to −2.35) p < 0.001	−1.26 (−1.97 to −0.55) p = 0.001	−1.17 (−1.90, −0.44) p = 0.002
ECW/TBW ratio categories
No edema	Reference	Reference	Reference
Slight edema	−4.24 (−5.37 to −3.10) p < 0.001	−1.52 (−2.63 to −0.41) p = 0.007	−2.21 (−3.90, −0.52) p = 0.011
Edema	−9.47 (−11.09 to −7.86) p < 0.001	−4.27 (−6.02 to −2.52) p < 0.001	−4.52 (−6.98, −2.05) p < 0.001

Model 1: unadjusted

Model 2: age, gender, ethnicity, education, presence of APOE ε4 allele, body mass index, mean arterial pressure, hemoglobin A1c, low-density lipoprotein-cholesterol, estimated glomerular filtration rate and log-transformed urinary albumin-to-creatinine ratio

Model 3: Model 2 + geriatric depression scale ≥ 5, history of cardiovascular disease, use of metformin, use of diuretic and use of sodium-glucose cotransporter 2 inhibitor

The findings of linear mixed model are found in Table 4. Baseline ECW/TBW ratio (per SD) was associated with decrement of total score at follow-up by 3.13 (95%CI −3.70, −2.56) in Model 1 which included follow-up period. When we additionally controlled for the covariates similar to those for the baseline analysis in Model 2, the relationship between ECW/TBW ratio and neurocognitive decrements persisted. In Model 3, the observation was similar with adjusted coefficient −1.43 (95%CI −2.31, −0.56). We also observed total score decrements by 3.71 (95%CI −4.94, −2.48) in the slight edema group and 9.70 (95%CI −11.81, −7.59) in the edema group in Model 1. Subjects with edema experienced decrements by 5.88 (95%CI −9.00, −2.75) in Model 3.

Table 4.

Association between ECW/TBW ratio and RBANS total score at follow-up.

	Coefficient (95% CI) p
	Model 1^a	Model 2^b	Model 3^c
ECW/TBW ratio (per SD increment)	−3.13 (−3.70 to −2.56) p < 0.001	−1.28 (−1.94 to −0.63) p < 0.001	−1.43 (−2.31, −0.56) p = 0.001
ECW/TBW ratio categories
No edema	Reference	Reference	Reference
Slight edema	−3.71 (−4.94 to −2.48) p < 0.001	−1.04 (−2.25 to 0.17) p = 0.092	−1.53 (−3.29, 0.23) p = 0.088
Edema	−9.70 (−11.81 to −7.59) p < 0.001	−5.01 (−7.35 to −2.68) p < 0.001	−5.88 (−9.00, −2.75) p < 0.001

Model 1: follow-up period

Model 2: Model 1 + age, gender, ethnicity, education, presence of APOE ε4 allele, body mass index, mean arterial pressure, hemoglobin A1c, low-density lipoprotein-cholesterol, estimated glomerular filtration rate and log-transformed urinary albumin-to-creatinine ratio

Model 3: Model 2 + geriatric depression scale ≥ 5, history of cardiovascular disease, use of metformin, use of diuretic and use of sodium-glucose cotransporter 2 inhibitor

ECW/TBW ratio (per SD) was associated with neurocognitive decrements longitudinally in all the neurocognitive domains in Model 1 (Table 5). Similar findings for ECW/TBW ratio in categories were observed in Model 1 except for language for the slight edema group (i.e., no evidence of statistical association). In Model 3, for each SD increment of ECW/TBW ratio, there were RBANS score decrements by 2.04 (95%CI −3.39, −0.70) in delayed memory, 1.84 (95%CI −3.02, −0.67) in visuospatial/constructional and 1.48 (95%CI −2.60, −0.35) in attention. The slight edema group was associated with lower score in delayed memory (coefficient −3.13, 95%CI −5.85, −0.42) in Model 3. The inverse associations were noted in more domains for the edema group – immediate memory (coefficient −3.84, 95%CI −7.48, −0.19), delayed memory (coefficient −7.25, 95%CI −12.08, −2.41), visuospatial/constructional (coefficient −7.07, 95%CI −11.29, −2.84), language (coefficient −6.10, 95%CI −10.42, −1.77) and attention (coefficient −5.30, 95%CI −9.34, −1.25) in Model 3.

Table 5.

Association between ECW/TBW ratio and RBANS domain scores at follow-up.

	Coefficient (95% CI) p
	Model 1^a	Model 2^b	Model 3^c
Immediate memory
ECW/TBW ratio (per SD increment)	−2.94 (−3.72 to −2.15) p < 0.001	−0.97 (−1.91 to −0.02) p = 0.045	−1.06 (−2.31, 0.20) p = 0.098
ECW/TBW ratio categories
No edema	Reference	Reference	Reference
Slight edema	−3.72 (−5.37 to −2.06) p < 0.001	−1.10 (−2.85 to 0.64) p = 0.216	−1.95 (−3.99, 0.09) p = 0.061
Edema	−8.20 (−11.10 to −5.29) p < 0.001	−2.91 (−6.28 to 0.46) p = 0.091	−3.84 (−7.48, −0.19) p = 0.039
Delayed memory
ECW/TBW ratio (per SD increment)	−3.07 (−3.88 to −2.26) p < 0.001	−1.48 (−2.47 to −0.49) p = 0.003	−2.04 (−3.39, −0.70) p = 0.003
ECW/TBW ratio categories
No edema	Reference	Reference	Reference
Slight edema	−3.74 (−5.45 to −2.03) p < 0.001	−1.45 (−3.28 to 0.39) p = 0.123	−3.13 (−5.85, −0.42) p = 0.023
Edema	−9.30 (−12.31 to −6.29) p < 0.001	−5.35 (−8.89 to −1.81) p = 0.003	−7.25 (−12.08, −2.41) p = 0.003
Visuospatial/ Constructional
ECW/TBW ratio (per SD increment)	−3.88 (−4.60 to −3.15) p < 0.001	−1.74 (−2.62 to −0.86) p < 0.001	−1.84 (−3.02, −0.67) p = 0.002
ECW/TBW ratio categories
No edema	Reference	Reference	Reference
Slight edema	−5.23 (−6.76 to −3.70) p < 0.001	−1.88 (−3.50 to −0.26) p = 0.023	−2.00 (−4.36, 0.37) p = 0.098
Edema	−12.42 (−15.13 to −9.71) p < 0.001	−6.97 (−10.10 to −3.83) p < 0.001	−7.07 (−11.29, −2.84) p = 0.001
Language
ECW/TBW ratio (per SD increment)	−1.64 (−2.39 to −0.90) p < 0.001	−0.55 (−1.47 to 0.38) p = 0.247	−0.76 (−1.98, 0.46) p = 0.222
ECW/TBW ratio categories
No edema	Reference	Reference	Reference
Slight edema	−1.23 (−2.81 to 0.34) p = 0.124	−1.01 (−0.68 to 2.70) p = 0.239	0.48 (−1.94, 2.90) p = 0.698
Edema	−5.10 (−7.86 to −2.34) p < 0.001	−4.41 (−7.67 to −1.15) p = 0.008	−6.10 (−10.42, −1.77) p = 0.006
Attention
ECW/TBW ratio (per SD increment)	−4.65 (−5.39 to −3.91) p < 0.001	−1.69 (−2.56 to −0.82) p < 0.001	−1.48 (−2.60, −0.35) p = 0.010
ECW/TBW ratio categories
No edema	Reference	Reference	Reference
Slight edema	−5.56 (−7.16 to −3.96) p < 0.001	−1.81 (−3.42 to −0.20) p = 0.027	−1.10 (−3.37, 1.17) p = 0.342
Edema	−14.06 (−16.83 to −11.29) p < 0.001	−5.48 (−8.58 to −2.38) p = 0.001	−5.30 (−9.34, −1.25) p = 0.010

Model 1: follow-up period

Model 2: Model 1 + age, gender, ethnicity, education, presence of Presence of APOE ε4 allele, body mass index, mean arterial pressure, hemoglobin A1c, low-density lipoprotein-cholesterol, estimated glomerular filtration rate and log-transformed urinary albumin-to-creatinine ratio

Model 3: Model 2 + geriatric depression scale ≥ 5, history of cardiovascular disease, use of metformin, use of diuretic and use of sodium-glucose cotransporter 2 inhibitor

The results of CLPM are found in Figure 1. There was a larger standardized effect size from baseline ECW/TBW ratio to follow-up total RBANS score (β = −0.053, p = 0.026) compared to that from baseline total RBANS score to follow-up ECW/TBW ratio (β = −0.020, p = 0.461). The criteria for mediation by PWV in the inverse relationship between edema and follow-up RBANS total score were met and described in Figure 2. PWV accounted for 13.3% of the relationship (p = 0.004).

Figure 1.

Cross-lagged panel model (CLPM) of ECW/TBW ratio (per SD) and cognitive function.

Figure 2.

Mediation of the association between edema and follow-up RBANS total score by PWV.

Discussion

In our current study, we observed that higher ECW/TBW ratio is associated with neurocognitive decrements over time in patients with T2D. This association was evident in global and multiple neurocognitive abilities involving delayed memory, visuo-spatial/constructional and attention. For the first time, our study unraveled the temporal relationship which was unidirectional from ECW/TBW ratio at baseline to neurocognitive decrements at follow-up. The mediation analysis provided further insights into the role of arterial stiffness as a mediating factor in the relationship between edema and neurocognitive decrements.

Although ECW/TBW ratio has been studied in various medical conditions, including malignancy, cardiovascular disease, renal disease and hepatic disease,²⁶ its association with cognitive impairment has been underexplored. We have earlier observed a cross-sectional relationship between ECW/TBW ratio and lower neurocognitive performance in subjects with T2D.¹⁵ However, we were unable to draw causal inference from the cross-sectional study. The CLPM has provided an additional lens to understand the temporal sequence of the relationship. Our findings revealed that ECW/TBW ratio preceded neurocognitive decrements and not vice versa; there was no evidence of bi-directional relationship.

Some studies suggested that ECW/TBW is an inflammatory marker.³⁹ Inflammation may drive cellular dysfunction, allowing fluid release from the intracellular to the extracellular compartments with resultant increase in ECW/TBW ratio or volume overload.^40,41 ECW/TBW ratio reflects volume status as excess volume builds up largely in ECW.¹⁰ Brain natriuretic peptide (BNP) is released in volume overload, and its level is elevated in individuals with high ECW/TBW ratio.⁴² Volume overload may increase hydrostatic pressure within the intestinal walls, thereby increasing intestinal permeability and promoting the release of bacteria and toxins into the systemic circulation.⁴³ The proinflammatory cytokines may enter the brain via blood-brain barrier (BBB), leading to neuroinflammation and cognitive decline.⁴⁴

ECW/TBW ratio may raise blood pressure and exert tension on aortic wall.⁴⁵ Over time, these hemodynamic forces change the aortic wall's constituents with increase in collagen fibers and breakdown in elastic tissue.⁴⁶ This process results in aortic stiffness which may exert a pulsatile force on the delicate microvasculature in the brain, contributing to cerebral hypoperfusion, disruption of BBB and accumulation of tau and amyloid-beta proteins.^13,46–49 Our mediation analysis supports the mediating role of aortic stiffness in the pathway linking edema to neurocognitive decrements.

We observed that volume overload affected several neurocognitive areas including delayed memory, visuo-spatial/constructional and attention. Mapped to these neurocognitive domains are medial temporal lobe (delayed memory),⁵⁰ parietal lobe (visuo-spatial/ constructional),⁵¹ and frontal and parietal lobes (attention)⁵² in the brain. Our results suggested that volume overload had a widespread impact on the brain, affecting multiple regions. This makes volume overload, reflected by elevated ECW/TBW ratio, a high-priority target for clinical intervention.

Volume overload can be modified by medications (e.g., diuretics), dietary control (e.g., low sodium), fluid control and appropriate management of medical conditions such as renal impairment.⁵³ Given that BIA is a convenient and non-invasive tool to assess water balance, we can potentially include BIA in the routine monitoring of patients for volume overload. The gold-standard method for fluid status assessment is isotope dilution which includes administration of deuterium oxide for measuring TBW and sodium bromide for measuring ECW.⁵⁴ However, this method requires blood sampling and is difficult to execute.⁵⁴ BIA, on the other hand, is non-invasive, fast, convenient and inexpensive.^54,55 It has been suggested to estimate fluid volume as a bedside tool in hospital.⁵⁵ However, the use of BIA may be limited by hydration changes.⁵⁶ Vigorous physical activity, alcohol or fluid consumption prior to measurement may impede BIA's ability to assess fluid status.⁵⁶ Clinicians should also be educated on the clinical importance of controlling and monitoring volume overload as part of prevention strategy against cognitive impairment in patients with T2D. Awareness should be raised on the multiple neurocognitive abilities affected by volume overload as they can impact diabetes self-care including medication compliance, appointment attendance, doctor-patient communication and regular blood glucose monitoring. We can also consider “de-stiffening strategy” to reduce aortic stiffness through aerobic exercise and appropriate anti-hypertensives.⁵⁷

In our longitudinal study, the availability of ECW/TBW ratio and neurocognitive performance measurements at baseline and follow-up allowed us to examine the temporal sequence of the relationship. Our mediation results also improved our understanding of the possible biological mechanism (i.e., aortic stiffness) underpinning the relationship. This study was conducted on a large cohort followed up over a relatively long period. We adjusted for an extensive array of potential cofounders including demographics, education, depression, clinical and genetic markers, medical history and medications. We also used carotid-femoral PWV, the criterion measure of aortic stiffness,³⁰ in our study. Nevertheless, we acknowledged some weaknesses in our current study. There was a lack of imaging tools (e.g., computed tomography or magnetic resonance imaging) to visualize affected brain structures. We also lacked information on cardiac failure and liver failure which could potentially confound the relationship. As our cohort only comprised subjects with T2D, our findings could not be applied to individuals without diabetes or from other settings (e.g., community and inpatient).

Conclusion

Volume overload, as indicated by high ECW/TBW ratio, contributed to neurocognitive decrements in patients with T2D. ECW/TBW ratio preceded neurocognitive decrements. It affected global and several neurocognitive domains. Aortic stiffness formed part of the causal pathway that linked volume overload and neurocognitive decrements. ECW/TBW ratio is a promising biomarker for monitoring and interventions to prevent ADRD in T2D.

Footnotes

Acknowledgements

We would like to thank the SMART2D research coordinators and Research Officer for their recruitment efforts.

ORCID iDs

Serena Low

Su Chi Lim

Ethical considerations

The study received approval from the National Healthcare Group Domain Specific Review Board (NHG DSRB) in Singapore

Consent to participate

All participants provided informed consent for the recruitment.

Author contribution(s)

Serena Low: Conceptualization; Formal analysis; Methodology; Writing – original draft.

Angela Moh: Methodology; Writing – review & editing.

Huili Zheng: Conceptualization; Formal analysis; Writing – review & editing.

Theedaraj Bun Chuan: Investigation; Writing – review & editing.

Keven Ang: Project administration; Writing – review & editing.

Wern Ee Tang: Investigation; Writing – review & editing.

Ziliang Lim: Conceptualization; Formal analysis; Writing – review & editing.

Tavintharan Subramaniam: Investigation; Writing – review & editing.

Chee Fang Sum: Investigation; Writing – review & editing.

Su Chi Lim: Conceptualization; Formal analysis; Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research was supported by Singapore National Medical Research Council (NMRC) (MOH-001772-01, MOH-001688-00 and DYNAMO II: MOH-001327-02). The first author was supported by Singapore NMRC (MOH-000714-01) and National Healthcare Group-Lee Kong Chian School of Medicine (NHG-LKCMedicine) (CSCS-24-01-01).

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The data is not publicly available as there are ethical or privacy constraints. The data that support the findings of this study, however, may be available from the corresponding author upon reasonable request.

References

Cao

Yang

, et al. The relationship between diabetes and the dementia risk: a meta-analysis. Diabetol Metab Syndr 2024; 16: 101.

Ab-Hamid

Omar

Ismail

CAN

, et al. Diabetes and cognitive decline: challenges and future direction. World J Diabetes 2023; 14: 795–807.

Arnold

Arvanitakis

Macauley-Rambach

, et al. Brain insulin resistance in type 2 diabetes and Alzheimer disease: concepts and conundrums. Nat Rev Neurol 2018; 14: 168–181.

Biessels

Reagan

. Hippocampal insulin resistance and cognitive dysfunction. Nat Rev Neurosci 2015; 16: 660–671.

Ehtewish

Arredouani

El-Agnaf

. Diagnostic, prognostic, and mechanistic biomarkers of diabetes mellitus-associated cognitive decline. Int J Mol Sci 2022; 23: 6144.

Bello-Chavolla

Aguilar-Salinas

Avila-Funes

. The type 2 diabetes-specific dementia risk score (DSDRS) is associated with frailty, cognitive and functional status amongst Mexican community-dwelling older adults. BMC Geriatr 2020; 20: 363.

Bhave

Neilson

. Body fluid dynamics: back to the future. J Am Soc Nephrol 2011; 22: 2166–2181.

Buffa

Saragat

Succa

, et al. Elderly subjects with type 2 diabetes show altered tissue electrical properties. Nutrition 2013; 29: 132–137.

Buscemi

Blunda

Maneri

, et al. Bioelectrical characteristics of type 1 and type 2 diabetic subjects with reference to body water compartments. Acta Diabetol 1998; 35: 220–223.

10.

Pérez-Morales

Donate-Correa

Martín-Núñez

, et al. Extracellular water/total body water ratio as predictor of mortality in hemodialysis patients. Ren Fail 2021; 43: 821–829.

11.

Czyżewski

Wyzgał

Czyżewska

, et al. Contribution of volume overload to the arterial stiffness of hemodialysis patients. Ren Fail 2017; 39: 333–339.

12.

Low

Moh

Zheng

, et al. Growth-differentiation factor 15 (GDF-15) is associated with arterial stiffness and impaired endothelial function with mediation by albuminuria in the Singapore study of macroangiopathy and microvascular reactivity in type 2 diabetes (SMART2D) cohort. Endocrine 2025; 89: 736–744.

13.

Pase

Beiser

Himali

, et al. Aortic stiffness and the risk of incident mild cognitive impairment and dementia. Stroke 2016; 47: 2256–2261.

14.

Kalaria

Akinyemi

Ihara

. Stroke injury, cognitive impairment and vascular dementia. Biochim Biophys Acta 2016; 1862: 915–925.

15.

Low

Lim

, et al. Higher ratio of extracellular water to total body water was associated with reduced cognitive function in type 2 diabetes. J Diabetes 2021; 13: 222–231.

16.

Kehayias

Ribeiro

Skahan

, et al. Water homeostasis, frailty and cognitive function in the nursing home. J Nutr Health Aging 2012; 16: 35–39.

17.

Low

Zheng

Liu

, et al. Longitudinal profiling and tracking stability in the Singapore study of macro-angiopathy and microvascular reactivity in type 2 diabetes cohort. Diab Vasc Dis Res 2023; 20: 14791641231218453.

18.

Nation

Wierenga

Delano-Wood

, et al. Elevated pulse pressure is associated with age-related decline in language ability. J Int Neuropsychol Soc 2010; 16: 933–938.

19.

Nyunt

Fones

Niti

, et al. Criterion-based validity and reliability of the geriatric depression screening scale (GDS-15) in a large validation sample of community-living Asian older adults. Aging Ment Health 2009; 13: 376–382.

20.

Levey

Stevens

Schmid

, et al. A new equation to estimate glomerular filtration rate. Ann Intern Med 2009; 150: 604–612.

21.

Low

Moh

Goh

, et al. Association between kidney disease index and decline in cognitive function with mediation by arterial stiffness in Asians with type 2 diabetes. J Alzheimers Dis Rep 2024; 8: 1199–1210.

22.

Campa

Toselli

Mazzilli

, et al. Assessment of body composition in athletes: a narrative review of available methods with special reference to quantitative and qualitative bioimpedance analysis. Nutrients 2021; 13: 1620.

23.

Mehra

Starkoff

Nickerson

. The evolution of bioimpedance analysis: from traditional methods to wearable technology. Nutrition 2025; 129: 112601.

24.

Haas

Schütz

Engeli

, et al. Comparing single-frequency bioelectrical impedance analysis against deuterium dilution to assess total body water. Eur J Clin Nutr 2012; 66: 994–997.

25.

Kushner

Schoeller

. Estimation of total body water by bioelectrical impedance analysis. Am J Clin Nutr 1986; 44: 417–424.

26.

Nishikawa

Yoh

Enomoto

, et al. Extracellular water to total body water ratio in viral liver diseases: a study using bioimpedance analysis. Nutrients 2018; 10: 1072.

27.

InBody Co Ltd K. InBodyS10 user's manual. Korea. 1996.

28.

Randolph

Tierney

Mohr

, et al. The repeatable battery for the assessment of neuropsychological Status (RBANS): preliminary clinical validity. J Clin Exp Neuropsychol 1998; 20: 310–319.

29.

Phillips

Cheung

Collinson

, et al. The equivalence and difference between the English and Chinese language versions of the repeatable battery for the assessment of neuropsychological Status. Clin Neuropsychol 2015; 29: 1–18.

30.

Van Bortel

Laurent

Boutouyrie

, et al. Expert consensus document on the measurement of aortic stiffness in daily practice using carotid-femoral pulse wave velocity. J Hypertens 2012; 30: 445–448.

31.

Gupta

Day

Tobin

, et al. Understanding the effect of categorization of a continuous predictor with application to neuro-oncology. Neurooncol Pract 2022; 9: 87–90.

32.

Bennette

Vickers

. Against quantiles: categorization of continuous variables in epidemiologic research, and its discontents. BMC Med Res Methodol 2012; 12: 21.

33.

Allison

Williams

Moral-Benito

. Maximum likelihood for cross-lagged panel models with fixed effects. Socius 2017; 3: 2378023117710578.

34.

Usami

. On the differences between general cross-lagged panel model and random-intercept cross-lagged panel model: interpretation of cross-lagged parameters and model choice. Struct Equ Modeling 2021; 28: 331–344.

35.

Shelleby

Ogg

. Longitudinal relationships between parent involvement, parental warmth, ADHD symptoms, and Reading achievement. J Atten Disord 2020; 24: 737–749.

36.

Singh

MA-O

Verhulst

BA-O

Vinh

PA-O

, et al. Using instrumental variables to measure causation over time in cross-lagged panel models. Multivariate Behav Res 2024; 59: 342–370.

37.

Zhang

Jia

, et al. The longitudinal bidirectional association between sarcopenia and cognitive function in community-dwelling older adults: findings from the China health and retirement longitudinal study. J Glob Health 2023; 13: 04182.

38.

Baron

Kenny

. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol 1986; 51: 1173–1182.

39.

Kamrani

Mohammadzadeh

Sobhani

, et al. Phase angle and water cell distribution as inflammation indicators linked to hematological markers across BMI categories. Sci Rep 2025; 15: 16147.

40.

da Silva

Orsso

Gonzalez

, et al. Phase angle and cellular health: inflammation and oxidative damage. Rev Endocr Metab Disord 2023; 24: 543–562.

41.

Zhu

Gan

Jiang

, et al. Analysis of factors influencing the increase of extracellular water ratio in tumor patients without edema signs. Front Med (Lausanne) 2025; 12: 1642980.

42.

Ohashi

Saito

Yamazaki

, et al. Brain natriuretic peptide and body fluid composition in patients with chronic kidney disease: a cross-sectional study to evaluate the relationship between volume overload and malnutrition. Cardiorenal Med 2016; 6: 337–346.

43.

Yajima

. Association of extracellular water/total body water ratio with protein-energy wasting and mortality in patients on hemodialysis. Sci Rep 2023; 13: 14257.

44.

Gryka-Marton

Grabowska

Szukiewicz

. Breaking the barrier: the role of proinflammatory cytokines in BBB dysfunction. Int J Mol Sci 2025; 26: 3532.

45.

Lafta

Ukrainetz

Davison

, et al. Inter- and intradialytic fluid volume changes and vascular stiffness parameters in patients on hemodialysis. PLoS One 2022; 17: e0262519.

46.

Kim

. Arterial stiffness and hypertension. Clin Hypertens 2023; 29: 31.

47.

Cooper

O'Donnell

Beiser

, et al. Association of aortic stiffness and pressure pulsatility with global amyloid-β and regional tau burden among Framingham Heart Study participants without dementia. JAMA Neurol 2022; 79: 710–719.

48.

Hughes

Wagenknecht

Craft

, et al. Arterial stiffness and dementia pathology: atherosclerosis risk in communities (ARIC)-PET study. Neurology 2018; 90: e1248–e1256.

49.

Singer

Trollor

Baune

, et al. Arterial stiffness, the brain and cognition: a systematic review. Ageing Res Rev 2014; 15: 16–27.

50.

Jeneson

Squire

. Working memory, long-term memory, and medial temporal lobe function. Learn Mem 2012; 19: 15–25.

51.

Bai

Liu

Guan

. The visuospatial and sensorimotor functions of posterior parietal cortex in drawing tasks: a review. Front Aging Neurosci 2021; 13: 717002.

52.

Ptak

. The frontoparietal attention network of the human brain: action, saliency, and a priority map of the environment. Neuroscientist 2012; 18: 502–515.

53.

Faucon

Leffondré

Flamant

, et al. Trajectory of extracellular fluid volume over time and subsequent risks of end-stage kidney disease and mortality in chronic kidney disease: a prospective cohort study. J Intern Med 2021; 289: 193–205.

54.

Jaffrin

Morel

. Body fluid volumes measurements by impedance: a review of bioimpedance spectroscopy (BIS) and bioimpedance analysis (BIA) methods. Med Eng Phys 2008; 30: 1257–1269.

55.

Powers

Choi

Bitting

, et al. Rapid measurement of total body water to facilitate clinical decision making in hospitalized elderly patients. J Gerontol A Biol Sci Med Sci 2009; 64: 664–669.

56.

El Dimassi

Gautier

Zalc

, et al.

Body water volume estimation using bio impedance analysis: where are we?

IRBM 2024; 45: 100839.

57.

Chen

Shen

Liu

, et al. Arterial stiffness and stroke: de-stiffening strategy, a therapeutic target for stroke. Stroke Vasc Neurol 2017; 2: 65–72.