Associations of novel and conventional anthropometric indices with domain-specific low cognitive performance in older adults

Abstract

Background

The comparative relevance of conventional and newer anthropometric indices for domain-specific low cognitive performance in older adults remains uncertain.

Design

and methods: We conducted a cross-sectional analysis of 2,384 adults aged 60 years or older from the National Health and Nutrition Examination Survey (NHANES) 2011-2014 with complete anthropometric, cognitive, covariate, and survey-design data. Eight anthropometric indices were standardized using survey-weighted means and standard deviations. Test-defined low cognitive performance was evaluated across the digit symbol substitution test (DSST), animal fluency test, immediate recall test, and delayed recall test. Low DSST performance was prespecified as the primary outcome. Survey-weighted logistic regression accounted for NHANES weights, strata, and primary sampling units, with false discovery rate correction across the eight primary DSST comparisons and an exploratory correction across all 32 index-domain combinations.

Results

In fully adjusted models, higher weight-adjusted waist index (WWI) and A Body Shape Index (ABSI) were associated with greater odds of low DSST performance (WWI: odds ratio [OR] 1.25, 95% confidence interval [CI] 1.07-1.46, p=0.011, q=0.045; ABSI: OR 1.24, 95% CI 1.07-1.43, p=0.009, q=0.045). Associations with other cognitive domains were weaker and did not remain significant under the broader 32-test correction.

Conclusions

WWI and ABSI showed modest, domain-specific associations with test-defined low processing-speed performance. The findings are hypothesis-generating and require longitudinal and external validation.

Significance for public health

This study addresses a public health question relevant to population aging: whether simple anthropometric measures can help characterize obesity-related differences in late-life cognitive performance. Using nationally representative NHANES data and survey-weighted standardized indices, we found that WWI and ABSI showed modest associations with test-defined low processing-speed performance. The findings should not be interpreted as clinical screening or prediction evidence, but they support further longitudinal research on body-shape-related adiposity and cognitive aging.

Introduction

With Alzheimer’s disease (AD) being the most common and severe form of dementia, ¹ low cognitive performance-which includes deficits in memory, attention, language, and executive function ² -is a major public health concern and an early indicator of dementia. The number of individuals living with AD in the United States is projected to rise dramatically over the next four decades, ³ highlighting the urgent need for effective preventive strategies. Among modifiable risk factors, obesity has emerged as a significant contributor to low cognitive performance, with neuroimaging studies showing that excess adiposity negatively affects brain structure and functional connectivity, potentially exacerbating cognitive impairment. This has prompted growing interest in identifying which measures of adiposity most accurately capture the risk of low cognitive performance.

Conventional anthropometric indicators such as body mass index (BMI) and waist circumference (WC) have been widely used to estimate obesity-related risks. ⁴ However, these measures have notable limitations: they cannot distinguish lean mass from fat mass and may not accurately reflect central fat distribution.^5,6 As a result, their utility for characterizing adiposity-related differences in cognitive performance may be limited. Importantly, prior work in psychiatry and gerontology has suggested that obesity-related metabolic and inflammatory processes may accelerate late-life low cognitive performance, underscoring the need for more refined adiposity measures that better capture risks relevant to cognition.^7,8

To address these gaps, several novel indices have been developed, including the weight-adjusted waist index (WWI), ⁹ relative fat mass (RFM), ¹⁰ waist-to-height ratio (WHtR), ¹¹ A Body Shape Index (ABSI), ¹² C-index, ¹³ body roundness index (BRI), ¹⁴ and Clinica Universidad de Navarra-Body Adiposity Estimator (CUN-BAE). ¹⁵ These indices are designed to capture different aspects of body composition and fat distribution. ¹⁶ However, evidence on their associations with cognitive outcomes remains mixed, and few studies have compared multiple indices across different cognitive domains on a standardized scale.

Leveraging data from the National Health and Nutrition Examination Survey (NHANES) 2011-2014, ¹⁷ this study examined the associations between eight obesity-related indices and test-defined low cognitive performance across multiple domains in a nationally representative sample of U.S. older adults. The analysis was designed as a simultaneous, domain-specific epidemiological comparison of conventional and novel anthropometric measures after standardization, with interpretation constrained to test-defined performance rather than clinical cognitive impairment.

The objective was to evaluate whether standardized anthropometric indices showed domain-specific associations with low cognitive test performance and whether any observed associations remained robust under prespecified and exploratory multiple-testing corrections.

Design and methods

Study population

The National Health and Nutrition Examination Survey (NHANES), conducted by the National Center for Health Statistics (NCHS), is a cross-sectional survey designed to assess the health and nutritional status of the U.S. population. ¹⁷ This analysis used data from the 2011-2012 and 2013-2014 survey cycles. Among 19,931 NHANES participants, 16,299 individuals aged <60 years were excluded, leaving 3,632 adults aged ≥60 years. Of these, 514 were excluded because of missing anthropometric measurements, leaving 3,118 participants with complete anthropometric data. A further 364 participants were excluded because of missing cognitive measures, leaving 2,754 participants with complete anthropometric and cognitive data. Finally (Figure 1), 370 participants were excluded because of missing covariate or survey-design information, yielding a final analytic sample of 2,384 participants. NHANES protocols were approved by the NCHS Research Ethics Review Board, and all participants provided written informed consent. Because this study was a secondary analysis of publicly available de-identified data, no additional institutional review board approval was required.

Figure 1.

Flow diagram of participant selection from NHANES 2011-2014.

Assessment of cognitive performance

Cognitive performance was assessed using data from the 2011-2014 NHANES cycles among adults aged 60 years and older. Four standardized neuropsychological tests were used: the digit symbol substitution test (DSST), the animal fluency test (AFT), the immediate recall test (IRT), and the delayed recall test (DRT). ¹⁸ DSST, a component of the Wechsler Adult Intelligence Scale-Third Edition, measures the ability to match numbers with symbols within 133 seconds and reflects processing speed, sustained attention, and executive efficiency. ¹⁹ AFT assesses semantic retrieval and verbal fluency. ²⁰ IRT was calculated by summing three immediate recall trials, yielding a total score from 0 to 30, ²¹ whereas DRT was administered after a delay/interference period created by intervening non-memory cognitive tasks in the NHANES testing sequence, with scores ranging from 0 to 10. ²² No study-specific interference task was introduced by the present analysis. In the absence of universally accepted diagnostic cutoffs for these NHANES cognitive tests, low cognitive performance was defined as a score below the 25th percentile for each test. The 25th percentile thresholds were calculated using the unweighted distribution before complete-case covariate exclusion. The corresponding cutoffs were 34 for DSST, 13 for AFT, 16 for IRT, and 5 for DRT. A distribution-based low-performance definition has been used in prior NHANES cognitive analyses,^23,24 but those references were not used as evidence of clinical diagnostic validity. Participants with scores below these test-specific thresholds were therefore classified as having test-defined low performance on the corresponding cognitive test, not clinically diagnosed mild cognitive impairment or dementia.

Assessment of anthropometric indices

Anthropometric measurements were obtained according to standardized NHANES body-measures procedures described by NCHS. ¹⁷ In brief, trained health technicians measured height, weight, and waist circumference in the Mobile Examination Center using calibrated equipment and standardized quality-control procedures. We computed eight adiposity indices, including body mass index (BMI), waist-to-height ratio (WHtR), weight-adjusted waist index (WWI), A Body Shape Index (ABSI), body roundness index (BRI), C-index, Clinica Universidad de Navarra-Body Adiposity Estimator (CUN-BAE), and relative fat mass (RFM). Detailed formulas and derivation methods are provided in Supplementary Table S1, with source references for BMI/WHtR, WWI, ABSI, BRI, C-index, CUN-BAE, and RFM.^{9,10,12,14,15,25–27} For the revised primary regression analyses, continuous anthropometric indices were standardized using survey-weighted means and standard deviations, so odds ratios represent the odds of test-defined low cognitive performance per 1-SD higher anthropometric index. Original-unit models were retained as supplementary analyses.

Covariates

The selection of covariates was guided by biological plausibility and prior literature on obesity and cognition.^28,29 Specifically, we adjusted for sociodemographic factors (race/ethnicity, educational attainment, marital status), lifestyle behaviors (smoking, alcohol intake), and major comorbidities (cardiovascular disease, diabetes, and hypertension), which are established determinants of both adiposity and cognitive outcomes. Inclusion of these variables aimed to minimize confounding and enhance the validity of the associations observed between anthropometric indices and cognitive performance. All covariate data were obtained from the NHANES database.

Statistical analysis

All statistical analyses were performed using R software (version 4.5.0 for the revision analysis). NHANES sampling weights and complex survey design variables were incorporated in all analyses. Because data from the 2011-2012 and 2013-2014 cycles were combined, 4-year MEC examination weights were calculated as WTMEC2YR/2. The survey design object used WTMEC4YR as the weight, SDMVSTRA as the strata, and SDMVPSU as the primary sampling unit, with nesting specified. Continuous variables were summarized as survey-weighted mean +/- SD and compared using survey-weighted t-tests, while categorical variables were summarized as unweighted counts with survey-weighted percentages and compared using Rao-Scott chi-square tests. Survey-weighted logistic regression models were used to examine associations between standardized anthropometric indices and test-defined low cognitive performance. Model 1 was unadjusted; Model 2 was adjusted for age and sex; and Model 3 was adjusted for age, sex, education, marital status, race/ethnicity, cardiovascular disease, diabetes, smoking, alcohol use, and hypertension. Low DSST performance was considered the primary cognitive outcome. For the primary DSST analysis, Benjamini-Hochberg FDR correction ³⁰ was applied across the eight standardized anthropometric indices. A conservative exploratory FDR correction across all 32 continuous index-domain combinations was also reported. Sensitivity analyses included survey-weighted 25th-percentile cognitive cutoffs and continuous standardized cognitive scores. Restricted cubic spline, subgroup, correlation, and VIF-like analyses were considered exploratory. Two-sided p values less than 0.05 were considered nominally significant.

Results

Baseline characteristics of the study population

The final analytic sample included 2,384 adults aged 60 years or older with complete anthropometric, cognitive, covariate, and survey design data. Cognitive performance was assessed using four tests-DSST, AFT, IRT, and DRT-and participants were classified into normal and low-performance groups based on each test-specific threshold. The weighted prevalence of test-defined low performance ranged from 12% for DSST to 20% for DRT.

Across cognitive tests, participants with low performance were generally older, had lower poverty-income ratio, had lower educational attainment, and had a higher burden of cardiometabolic conditions than those with normal performance (Table 1). Table 1 presents anthropometric indices as standardized z-scores for descriptive comparison. Exact p values for baseline comparisons are provided in Supplementary Table S9.

Table 1.

Baseline characteristics of study participants stratified by cognitive function status.

Characteristic	DSST		AFT		IRT		DRT
	Normal, N = 1837 (88%)	Low, N = 547 (12%)	Normal, N = 1865 (86%)	Low, N = 519 (14%)	Normal, N = 1880 (84%)	Low, N = 504 (16%)	Normal, N = 1807 (80%)	Low, N = 577 (20%)
Age (years)	68.35 ± 6.36***	72.74 ± 6.64***	68.38 ± 6.40***	72.00 ± 6.65***	68.21 ± 6.26***	72.42 ± 6.94***	67.98 ± 6.21***	72.51 ± 6.65***
PIR	3.34 ± 1.53***	1.93 ± 1.31***	3.30 ± 1.55***	2.36 ± 1.45***	3.30 ± 1.55***	2.48 ± 1.50***	3.28 ± 1.56***	2.72 ± 1.55***
Sex
Female	974 (54%)	236 (52%)	948 (53%)	262 (56%)	1,011 (55%)**	199 (45%)**	991 (56%)***	219 (43%)***
Male	863 (46%)	311 (48%)	917 (47%)	257 (44%)	869 (45%)**	305 (55%)**	816 (44%)***	358 (57%)***
Race
NH White	1,026 (85%)***	159 (55%)***	1,011 (84%)***	174 (63%)***	951 (82%)*	234 (75%)*	902 (82%)*	283 (78%)*
NH Black	363 (5.9%)***	190 (21%)***	368 (6.1%)***	185 (18%)***	438 (7.5%)*	115 (9.3%)*	407 (7.3%)*	146 (9.5%)*
Other Hisp	136 (2.3%)***	101 (12%)***	171 (2.9%)***	66 (6.8%)***	173 (3.0%)*	64 (5.6%)*	175 (3.2%)*	62 (4.4%)*
Other	178 (4.9%)***	21 (2.7%)***	144 (4.1%)***	55 (7.6%)***	161 (4.5%)*	38 (4.8%)*	171 (4.9%)*	28 (3.4%)*
Mex Am	134 (2.4%)***	76 (9.5%)***	171 (3.1%)***	39 (4.4%)***	157 (2.9%)*	53 (4.9%)*	152 (2.9%)*	58 (4.5%)*
Education
>=College	1,582 (90%)***	242 (52%)***	1,505 (88%)***	319 (68%)***	1,522 (88%)***	302 (71%)***	1,445 (87%)***	379 (77%)***
High School Grad	199 (8.0%)***	122 (22%)***	220 (8.6%)***	101 (17%)***	235 (8.7%)***	86 (15%)***	231 (8.9%)***	90 (13%)***
<High School	56 (2.1%)***	183 (26%)***	140 (3.4%)***	99 (16%)***	123 (3.4%)***	116 (14%)***	131 (3.9%)***	108 (9.8%)***
Marital
Married	1,133 (68%)***	270 (50%)***	1,128 (68%)**	275 (55%)**	1,117 (67%)*	286 (61%)*	1,074 (67%)	329 (63%)
Other	704 (32%)***	277 (50%)***	737 (32%)**	244 (45%)**	763 (33%)*	218 (39%)*	733 (33%)	248 (37%)
Smoking
Yes	929 (51%)	296 (53%)	956 (51%)	269 (53%)	976 (52%)*	249 (47%)*	930 (52%)	295 (47%)
No	908 (49%)	251 (47%)	909 (49%)	250 (47%)	904 (48%)*	255 (53%)*	877 (48%)	282 (53%)
Alcohol
No	1,314 (67%)***	465 (86%)***	1,342 (67%)***	437 (81%)***	1,373 (68%)**	406 (77%)**	1,344 (69%)	435 (68%)
Yes	523 (33%)***	82 (14%)***	523 (33%)***	82 (19%)***	507 (32%)**	98 (23%)**	463 (31%)	142 (32%)
Cardiovascular
No	1,732 (94%)***	487 (85%)***	1,751 (94%)*	468 (89%)*	1,763 (94%)*	456 (90%)*	1,703 (95%)***	516 (88%)***
Yes	105 (5.6%)***	60 (15%)***	114 (6.0%)*	51 (11%)*	117 (6.1%)*	48 (9.6%)*	104 (5.3%)***	61 (12%)***
Diabetes
No	1,466 (83%)***	376 (69%)***	1,470 (82%)	372 (77%)	1,473 (82%)	369 (78%)	1,407 (82%)	435 (79%)
Yes	371 (17%)***	171 (31%)***	395 (18%)	147 (23%)	407 (18%)	135 (22%)	400 (18%)	142 (21%)
Hypertension
Yes	1,095 (56%)***	381 (72%)***	1,121 (57%)**	355 (67%)**	1,148 (57%)	328 (65%)	1,107 (57%)	369 (63%)
No	742 (44%)***	166 (28%)***	744 (43%)**	164 (33%)**	732 (43%)	176 (35%)	700 (43%)	208 (37%)
BMI	0.00 ± 0.99	-0.03 ± 1.00	0.01 ± 0.99	-0.09 ± 1.00	0.02 ± 1.00**	-0.14 ± 0.94**	0.02 ± 1.00	-0.08 ± 0.96
WWI	-0.08 ± 0.96***	0.33 ± 1.02***	-0.06 ± 0.96**	0.18 ± 1.05**	-0.06 ± 0.98*	0.12 ± 0.97*	-0.08 ± 0.98***	0.16 ± 0.95***
ABSI	-0.00 ± 0.98***	0.27 ± 1.00***	0.01 ± 0.98*	0.19 ± 1.04*	-0.01 ± 1.00**	0.26 ± 0.91**	-0.04 ± 0.98***	0.31 ± 0.99***
BRI	-0.04 ± 0.97	0.12 ± 1.03	-0.02 ± 0.97	0.01 ± 1.03	-0.01 ± 0.98	-0.05 ± 0.99	-0.02 ± 0.99	0.00 ± 0.95
CUN-BAE	0.03 ± 0.97	0.04 ± 0.99	0.03 ± 0.97	0.03 ± 1.00	0.06 ± 0.97**	-0.13 ± 0.95**	0.07 ± 0.98***	-0.12 ± 0.95***
WHtR	-0.03 ± 0.98	0.12 ± 1.03	-0.02 ± 0.98	0.00 ± 1.03	-0.01 ± 0.98	-0.05 ± 0.99	-0.02 ± 1.00	0.01 ± 0.94
CI	-0.00 ± 0.99**	0.23 ± 0.99**	0.01 ± 0.99	0.11 ± 1.03	0.00 ± 1.00	0.16 ± 0.95	-0.02 ± 1.00***	0.24 ± 0.94***
RFM	0.03 ± 0.97	0.08 ± 0.99	0.03 ± 0.96	0.07 ± 1.01	0.06 ± 0.97*	-0.11 ± 0.96*	0.07 ± 0.97**	-0.10 ± 0.95**

DSST, Digit Symbol Substitution Test; AFT, Animal Fluency Test; IRT, Immediate Recall Test; DRT, Delayed Recall Test; NH White, non-Hispanic White; NH Black, non-Hispanic Black; Other Hisp, Other Hispanic; Mex Am, Mexican American; BMI, body mass index; WWI, weight-adjusted waist index; ABSI, A Body Shape Index; CI, conicity index; BRI, body roundness index; CUN-BAE, Clinica Universidad de Navarra-Body Adiposity Estimator; WHtR, waist-to-height ratio; RFM, relative fat mass. *p<0.05, **p<0.01, ***p<0.001 based on survey-weighted linear regression or Rao-Scott chi-square tests. Anthropometric indices are presented as standardized z-scores.

With respect to anthropometric measures, standardized WWI and ABSI values were generally higher among participants with low cognitive performance, especially for DSST and DRT, whereas BMI, WHtR, BRI, CUN-BAE, CI, and RFM showed less consistent differences across domains.

Logistic regression analyses of anthropometric indices and low cognitive performance

We examined the associations between standardized anthropometric indices and test-defined low cognitive performance using survey-weighted logistic regression models. In the revised primary analyses, odds ratios are interpreted per 1-SD higher anthropometric index, allowing more direct comparison across indices measured on different scales.

In the primary DSST analysis, higher standardized WWI and ABSI were associated with greater odds of low DSST performance after full adjustment and FDR correction across the eight anthropometric indices. Per 1-SD higher WWI, the odds ratio for low DSST performance was 1.25 (95% CI: 1.07-1.46; p = 0.011; q_primary = 0.045). Per 1-SD higher ABSI, the odds ratio was 1.24 (95% CI: 1.07-1.43; p = 0.009; q_primary = 0.045). Other standardized indices were not significant after DSST-specific FDR correction. Table 2 summarizes WWI and ABSI, and full results for all eight indices are provided in Supplementary Tables S3-S5.

Table 2.

Fully adjusted associations of WWI and ABSI with domain-specific test-defined low cognitive performance.

​	DSST			AFT		IRT		DRT
Characteristic	OR (95% CI)	p	q	OR (95% CI)	p	OR (95% CI)	p	OR (95% CI)	p
WWI	1.25 (1.07, 1.46)	0.011	0.045	1.11 (0.95, 1.31)	0.212	1.04 (0.87, 1.24)	0.660	1.19 (1.05, 1.34)	0.014
ABSI	1.24 (1.07, 1.43)	0.009	0.045	1.13 (0.97, 1.33)	0.141	1.09 (0.92, 1.31)	0.331	1.20 (1.04, 1.38)	0.025

Model 3 was adjusted for age, sex, education, marital status, race/ethnicity, cardiovascular disease, diabetes, smoking, alcohol use, and hypertension. Odds ratios represent the odds of test-defined low cognitive performance per 1-SD higher anthropometric index, standardized using survey-weighted means and standard deviations. Low DSST performance was considered the primary cognitive outcome. q indicates Benjamini-Hochberg FDR correction across the eight anthropometric indices within the DSST outcome. Conservative exploratory FDR results across all 32 index-domain combinations are reported in Supplementary Table S5.

For secondary exploratory cognitive outcomes, associations were weaker and less consistent. WWI and ABSI showed nominal associations with low DRT performance (WWI: OR = 1.19, 95% CI: 1.05-1.34, p = 0.014; ABSI: OR = 1.20, 95% CI: 1.04-1.38, p = 0.025), whereas associations with AFT and IRT were not statistically significant in fully adjusted standardized models. Under the conservative exploratory FDR correction across all 32 index-domain combinations, the DSST associations for WWI and ABSI were attenuated and no longer statistically significant (q_all_32 = 0.137 for both); DRT associations were also not significant after this broader correction.

Sensitivity analyses supported a cautious interpretation. When low cognitive performance was redefined using survey-weighted 25th-percentile cutoffs, the DSST cutoff increased to 42 and the DSST associations for WWI and ABSI were attenuated (WWI: OR = 1.15, 95% CI: 0.98-1.36, p = 0.110; ABSI: OR = 1.18, 95% CI: 1.03-1.36, p = 0.032, q_primary = 0.255). In continuous cognitive z-score analyses, WWI and ABSI were inversely associated with DSST scores (WWI: beta = -0.08, 95% CI: -0.14 to -0.03, p = 0.010; ABSI: beta = -0.07, 95% CI: -0.13 to -0.02, p = 0.011), but these associations were also attenuated under the 32-test FDR framework. Overall, the findings are best interpreted as suggestive and domain-specific.

Restricted cubic spline analyses

Restricted cubic spline analyses (Figure 2) were used to explore potential nonlinear associations of WWI and ABSI with low DSST performance. These analyses should be interpreted as exploratory assessments of association shape rather than as evidence of a diagnostic cutoff or causal inflection point.

Figure 2.

Restricted cubic spline analyses of WWI and ABSI in relation to low DSST performance. A: WWI; B: ABSI. Models were adjusted for age, sex, education, marital status, race/ethnicity, cardiovascular disease, diabetes, smoking, alcohol use, and hypertension. Knots were placed at the 5th, 35th, 65th, and 95th percentiles, and the median was used as the reference value. Predictors were modeled on their original scales, except ABSI, which was multiplied by 100. The figure presents exploratory nonlinear associations and should not be used to infer a diagnostic cutoff.

For WWI, the spline model showed modest overall evidence of association (P overall = 0.067) and evidence of nonlinearity (P nonlinearity = 0.035). For ABSI, the overall association was borderline (P overall = 0.050), with evidence of nonlinearity (P nonlinearity = 0.025). The spline knots were placed at the 5th, 35th, 65th, and 95th percentiles, and the median value was used as the reference. Predictors were modeled on their original scales, except for ABSI, which was multiplied by 100.

Exploratory subgroup analyses

Exploratory subgroup analyses were conducted for the primary DSST outcome (Figure 3). Overall, associations of WWI and ABSI with low DSST performance were generally directionally consistent across several demographic and clinical strata, but most interaction tests were not statistically significant. A potential interaction by race/ethnicity was observed for WWI (P interaction = 0.027), whereas the corresponding interaction for ABSI was not significant (P interaction = 0.291). Because subgroup-specific estimates were based on smaller strata and multiple interaction tests were performed, these findings should be interpreted cautiously and considered hypothesis-generating.

Figure 3.

Exploratory subgroup analyses of the associations of WWI and ABSI with low DSST performance. A: WWI; B: ABSI. Each subgroup label includes the stratum-specific sample size and the interaction p value. Models used the same covariates as Model 3 after removing the stratifying variable. Results should be interpreted cautiously because multiple subgroup comparisons and interaction tests were performed, and some strata had limited sample sizes.

Discussion

The principal finding of this cross-sectional NHANES analysis was that standardized WWI and ABSI showed modest associations with test-defined low DSST performance after full adjustment and FDR correction across the eight anthropometric indices within the prespecified primary DSST outcome. In contrast, associations with AFT and IRT were not robust, and DRT showed only secondary nominal associations. Under a more conservative exploratory correction across all 32 index-domain combinations, the DSST associations for WWI and ABSI were no longer statistically significant (q_all_32 = 0.137 for both), and DRT associations also did not remain significant. These DSST-focused findings are broadly consistent with prior NHANES-based studies linking central or body-shape-related adiposity indices with poorer cognitive performance. However, compared with studies focusing on a single index, the present analysis standardized eight anthropometric indices and applied multiplicity correction, which led to a more conservative interpretation. The attenuation of associations under the 32-test FDR framework suggests that these findings should be viewed as domain-specific rather than evidence of generalized cognitive impairment. Therefore, these results should be interpreted as hypothesis-generating rather than as definitive evidence of general cognitive impairment or predictive usefulness.

From a mechanistic perspective, WWI and ABSI may better capture body-shape-related adiposity patterns than BMI alone. Prior population and mechanistic studies have linked adiposity and central fat distribution to poorer cognitive performance, visceral adiposity, inflammatory signaling, insulin resistance, and vascular burden.^31–35 However, the present analysis did not directly measure visceral fat, inflammatory biomarkers, insulin resistance, neuroimaging markers, white matter injury, or blood-brain barrier function. These mechanisms should therefore be presented as plausible hypotheses rather than as mechanisms demonstrated by the current study.

The DSST-focused pattern may reflect the sensitivity of processing speed and executive efficiency to vascular and metabolic burden, but DSST is not a disease-specific or purely cognitive measure. DSST performance integrates psychomotor speed, attention, visual scanning, executive control, education, literacy, and other contextual factors. This is particularly important because the outcomes in this study were based on test-specific distributional cutoffs rather than clinical diagnoses. The findings should therefore be described as test-defined low performance rather than mild cognitive impairment or dementia.

The exploratory subgroup analyses suggested possible heterogeneity by race/ethnicity for WWI, but this finding should be interpreted cautiously. Differences in fat distribution, vascular risk profiles, social determinants, healthcare access, diet, and cumulative stress may all plausibly contribute to heterogeneous associations, but these factors were not directly tested in this analysis. In addition, several strata had limited sample sizes, and multiple interaction tests increase the likelihood of chance findings. Future studies should evaluate potential effect modification using larger, diverse cohorts with harmonized measures of adiposity, cognition, and contextual factors.

The intercorrelation analysis also supports a cautious interpretation of index-specific effects. Many anthropometric indices were mathematically correlated because they share common components such as waist circumference, height, weight, BMI, age, and sex. For example, BRI and WHtR, CUN-BAE and RFM, BMI and BRI, BMI and WHtR, WWI and CI, and ABSI and CI were highly correlated. VIF-like diagnostics indicated very high inter-index collinearity. For this reason, each anthropometric index was modeled separately rather than simultaneously in the primary regression models. The correlation matrix and VIF-like diagnostics were provided to transparently describe this issue.

It is also important to acknowledge the bidirectional and time-dependent nature of the relationship between adiposity and cognition. Central adiposity may contribute to poorer cognitive performance through vascular and metabolic pathways, but lower cognitive performance may also lead to reduced physical activity, poorer diet, and secondary changes in body composition. Longitudinal studies with repeated cognitive testing and repeated adiposity measures are needed to clarify temporal directionality and to evaluate potential mediation through inflammatory and metabolic pathways.

The main strengths of this study include the use of nationally representative NHANES data, simultaneous comparison of eight conventional and novel anthropometric indices, survey-weighted standardization to improve comparability across differently scaled indices, prespecified DSST-focused multiplicity correction, broader exploratory FDR correction across all index-domain combinations, and sensitivity analyses using survey-weighted cognitive cutoffs and continuous cognitive z-scores.

This study has several limitations. First, the cross-sectional design precludes causal inference and raises the possibility of reverse causation. Second, survivor bias may have influenced the observed associations among older adults. Third, residual confounding remains possible because not all relevant factors were available or included, such as detailed physical activity, diet quality, medication use, depression, sleep, APOE genotype, inflammatory biomarkers, and neuroimaging data. Fourth, anthropometric indices are indirect measures and may be affected by measurement error; they cannot replace imaging-validated measures of visceral adiposity. Fifth, the cognitive outcomes were based on relative test-score cutoffs rather than clinical diagnoses of mild cognitive impairment or dementia, and sensitivity analyses using survey-weighted cutoffs attenuated the DSST findings. Sixth, the 2011-2014 NHANES cycles may not fully represent contemporary patterns of obesity and cognitive health. Finally, we did not evaluate predictive discrimination, calibration, net reclassification improvement, or decision-curve performance, so readiness for individual-level screening cannot be inferred.

Conclusions

In conclusion, this cross-sectional NHANES analysis showed that standardized WWI and ABSI were associated with test-defined low DSST performance after full adjustment and FDR correction within the prespecified primary DSST analysis. Associations with other cognitive domains were weaker and exploratory, and the findings were attenuated under conservative correction across all index-domain combinations and under survey-weighted 25th-percentile sensitivity analysis. These results suggest that body-shape-related adiposity indices may help characterize adiposity-related differences in processing-speed performance in epidemiological research, but they should not be interpreted as evidence of readiness for individual-level screening or risk prediction. Future studies should validate these associations in longitudinal cohorts with repeated cognitive assessments, imaging-based measures of visceral adiposity, inflammatory and metabolic biomarkers, and external populations beyond the United States.

ORCID iD

Jingxu Chen https://orcid.org/0000-0003-2395-6120

Ethical considerations

NHANES protocols were approved by the National Center for Health Statistics Research Ethics Review Board. Because this study was a secondary analysis of publicly available, de-identified NHANES data, no additional institutional review board approval was required.

Consent to participate

Written informed consent was obtained from all participants by the National Center for Health Statistics.

Consent for publication

Not applicable; this study used publicly available de-identified data and contains no identifiable individual-level information.

Data Availability Statement

Publicly available data from the NHANES 2011-2012 and 2013-2014 cycles were analyzed in this study. The datasets are available from the National Center for Health Statistics NHANES website. The analytic code used to generate the current results is available from the corresponding author upon reasonable request.*