Differential effects of body mass index on domain-specific cognitive outcomes after stroke

Although the obesity paradox is an important modifiable factor in cardiovascular diseases, little research has been conducted to determine how it affects post-stroke cognitive function. We aimed to investigate the association between body mass index (BMI) and domain-specific cognitive outcomes, focusing on the subdivision of each frontal domain function in post-ischemic stroke survivors. A total of 335 ischemic stroke patients were included in the study after completion of the Korean-Mini Mental Status Examination (K-MMSE) and the vascular cognitive impairment harmonization standards neuropsychological protocol at 3 months after stroke. Frontal lobe functions were analyzed using semantic/phonemic fluency, processing speed, and mental set shifting. Our study participants were categorized into four groups according to BMI quartiles. The z-scores of K-MMSE at 3 months differed significantly between the groups after adjustment for initial stroke severity (p = 0.014). Global cognitive function in stroke survivors in the Q1 (the lowest quartile) BMI group was significantly lower than those in Q2 and Q4 (the highest quartile) BMI groups (K-MMSE z-scores, Q1: − 2.10 ± 3.40 vs. Q2: 0.71 ± 1.95 and Q4: − 1.21 ± 1.65). Controlled oral word association test findings indicated that phonemic and semantic word fluency was lower in Q4 BMI group participants than in Q2 BMI group participants (p = 0.016 and p = 0.023 respectively). BMI might differentially affect cognitive domains after ischemic stroke. Although being underweight may negatively affect global cognition post-stroke, obesity could induce frontal lobe dysfunctions, specifically phonemic and semantic word fluency.


Introduction
Obesity is a growing public health concern and a well-recognized risk factor for cerebrovascular disorders and overall mortality 1 . However, an increasing body of evidence suggests that obesity is associated with favorable functional outcomes, lower mortality, and lower risk of stroke progression after stroke. 2-4 . Referred to as the 'obesity paradox', this paradoxical phenomenon has been observed in association with cerebrovascular disorders, cardiovascular diseases 5 , peripheral arterial disease 6 , diabetes mellitus 7 , malignancy 8 , and dementia 9,10 .
This paradox has also been reported in the correlation between body mass index (BMI) and cognitive function in the general population. A higher BMI in middle-to-late life predicts a lower risk of developing dementia 10 , whereas being underweight (BMI, < 20kg/m 2 ) is associated with an increased risk of dementia 9 . However, these reports did not explore the independent association between BMI and poststroke cognition. One Swedish study reported that BMI did not in uence cognitive impairment at 20 months after stroke in a sample of 149 stroke survivors 11 . However, that study used only the Mini-Mental Status Examination (MMSE) to assess cognition, which cannot be considered appropriate given the higher incidence of frontal dysfunction among stroke survivors 12 . To date, few studies have used detailed neuropsychological evaluations to determine how obesity affects post-stroke cognitive impairment.
Considering that a lesion's localization and pathogenesis can be estimated through the pattern of domain-speci c cognitive impairment, it would likely be of clinical signi cance to determine whether obesity affects particular domains of cognitive function after stroke.
In this context, we aimed to investigate the differential effects of BMI on domain-speci c and global cognitive function at 3 months post-ischemic stroke using the Korean Vascular Cognitive Impairment Harmonization Standards-Neuropsychological Protocol(K-VCIHS-NP), a comprehensive neuropsychological test developed for post-stroke survivors. We hypothesized that BMI might have distinct roles in each neural substrate associated with each cognitive domain, as previous studies have shown that high BMI is associated with executive dysfunction and frontal lobe atrophy in a generally healthy population [13][14][15] .

Study participants
Participants were enrolled from the Korean Vascular Cognitive Impairment Harmonization Standards (K-VCIHS) study, which has been previously described in detail 16 . The K-VCIHS cohort enrolled patients with ischemic stroke who had been consecutively admitted to 12 university hospitals from October 2007 to August 2008. Inclusion criteria for our study comprised the following: (i) a diagnosis of acute ischemic stroke with neurological de cits persisting for ≥ 24 hours, (ii) a relevant ischemic lesion observed on magnetic resonance imaging, (iii) admission within 7 days after symptom onset, and (iv) available data on admission concerning BMI and the results of K-VCIHS-NP performed at 3 months post-ischemic stroke. In total, 335 participants from the K-VCIHS cohort who ful lled the inclusion criteria were nally analyzed. We collected clinical variables and the results of K-VCIHS-NP with the consent of the principal investigator of the K-VCIHS-NP study group and the Institutional Review Board of Hallym University Sacred Heart Hospital. BMI was calculated as a participant's weight (kg) divided by their height (m) squared, measured at the rst day of admission according to the institutional protocol with an automatic weight and height machine. Body weight and height were measured to the nearest 0.1kg and 0.1cm respectively using BSM-330(InBody, South Korea). Those who were not able to stand were weighed using an electronic bed scale, SCB330-7(SOHWA Inc, South Korea), in the supine position. The participants were then strati ed into BMI quartiles 17 18 . Moreover, initial stroke severity represented as the National Institute of Health Stroke Scale and vascular risk factors were also collected. Hallym University Sacred Heart Hospital Institutional Review Board approved the study and waived the requirement for patient consent because of the retrospective nature of this study as well as the minimal risk that it posed to participants. The study protocols conformed to the guidelines of the Declaration of Helsinki.

Cognitive function assessment
Participants completed the Korean Mini-Mental Status Examination (K-MMSE) and a 60-minute neuropsychological test as described in the K-VCIHS study protocol 19 . The K-VCIHS assesses four cognitive domains: (i) executive/activation function, using the Korean version of the controlled oral word association test (COWAT) for phonemic and semantic uency, digit symbol coding, and trail making tests A and B; (ii) language, using the Korean-Boston naming test; (iii) visuospatial, using the Rey complex gure test (copy), and; (iv) memory, using delayed recall scores from the Seoul verbal learning test. The K-VCIHS study protocol also included the Informant Questionnaire of Cognitive Decline in the Elderly (IQCODE) for the premorbid history of cognitive dysfunctions. The speci c tests and scales which comprises the K-VCIHS protocol and relevant references are described in Supplemental Table I. We obtained standardized Z-scores for each domain and for the K-MMSE after adjustment for age, sex, and educational level to directly compare groups according to BMI quartiles.

Brain imaging
All participants underwent brain magnetic resonance imaging (MRI), performed using a 3T whole-body MRI system. Diffusion-weighted imaging was used to obtain the apparent diffusion coe cient and assess acute cerebral infarction. The location and number of acute ischemic stroke lesions were rated and quanti ed. We subdivided the stroke lesions as follows: (i) cortical vs. subcortical-only, (ii) left-sided vs. right-sided, and; (iii) single vs. multiple 2 .

Statistical analysis
In order to detect an effect η 2 p=0.04 with 80% power in a between-subjects ANCOVA(four groups, numerator df = 3, alpha = 0.05), a priori power calculation of G*Power 3.1.9.4 20 suggested that we would need 67 subjects in each group(n = 266). Pearson's chi-square test and an analysis of variance (ANOVA) were used, as appropriate, to compare demographic characteristics between the groups according to BMI quartiles. For the univariate analysis, an ANOVA was used to compare the Z-scores of each cognitive domain between the groups. The Levene's test and Kolmogorov-Smirnov tests were performed to check for equality of variances and normality of dependent variables, respectively. An analysis of covariance (ANCOVA) was performed adjusting for the covariates which had a p-value less than 0.10 in the univariate analysis. Age, sex, and education levels were considered in the z-score transformation process and were not included as covariates. The effect size of BMI quartiles on dependent variables in the ANCOVA were presented with partial eta squared (η 2 p). Two-sided p-values < 0.05 were considered to indicate statistical signi cance. All analyses were performed using IBM SPSS, version 26.

Results
The mean age of the participants was 64.8 ± 12.4 years, and women comprised 38.9% of the study population. The median "BMI was 23.89 ± 3.10 kg/m 2 (range, 15.82-32.93 kg/m 2 ). The BMI quartile groups did not differ in terms of age, sex, years of education, previous stroke history, or pre-stroke cognitive decline. There were no signi cant differences in initial stroke severity, stroke etiology according to the TOAST classi cation, or the number and location of ischemic lesions. However, the prevalence of hypertension was higher in the Q4 BMI group (the highest) than in the Q1 BMI group (the lowest). The prevalence of other vascular risk factors did not differ signi cantly between the groups ( Table 1).
The Z-scores of the K-MMSE differed signi cantly between the quartile groups at 3 months post-ischemic stroke onset after adjustment for hypertension and hyperlipidemia, showing a p-value < 0.1 in univariate analysis (p = 0.003, η 2 p=0.042, ANCOVA). A multiple comparison analysis indicated that stroke survivors in the Q1 BMI group showed signi cantly lower global cognitive functions (K-MMSE Z-score, -2.1 ± 3.4) than those in the Q2 and Q4 BMI groups (K-MMSE Z-score, Q2: -0.71 ± 1.95; Q4: -1.21 ± 1.65). The Zscores of the COWAT also signi cantly differed between the quartile groups. A multiple comparison analysis of the COWAT results indicated that participants in the Q4 BMI group had signi cantly lower phonemic and semantic word uencies than their counterparts in the Q2 BMI group, despite similar levels of global cognitive function after adjustment for a history of hypertension and hyperlipidemia (p = 0.012, η 2 p=0.035 and p = 0.012, η 2 p=0.034 respectively; ANCOVA). Other neuropsychological test ndings relating to frontal lobe functions, including processing speed and mental set-shifting, and other cognitive domains, did not differ signi cantly between the groups ( Table 2).

Discussion
In this study, we assessed the association between obesity phenotypes de ned using BMI quartiles at admission and cognitive function at 3 months post-ischemic stroke onset. To our knowledge, this study is the rst to evaluate BMI and domain-speci c cognitive outcomes using a comprehensive, standardized, neuropsychological protocol in relation to a multicenter cohort of stroke patients 16 . Our data analysis indicated that a lower BMI at admission was associated with a higher risk of global cognitive deterioration, while a higher BMI was associated with signi cantly worse frontal dysfunction postischemic stroke. These ndings suggest that BMI has differential effects on various cognitive domains following an ischemic stroke.
Previous epidemiologic studies concerning the relationship between BMI and cognitive impairment have yielded controversial ndings. The relationship between obesity and long-term cognitive outcomes has alternately been identi ed as direct, inverse, U-shaped, or even absent 21,22 . However, as previous studies only performed the MMSE to assess cognition, they might have overlooked the in uence of BMI on frontal/executive function, given that the MMSE is not sensitive to evaluating frontal lobe function 23 . Our results are in line with previous studies showing that a low BMI was signi cantly associated with a high risk of cognitive decline 9 .
Despite uncertainly concerning the precise mechanism underlying post-stroke worsening of cognitive function, decreased body weight has been identi ed as an early indication of declining health and even neurodegeneration 24 . Moreover, several studies have proposed that leptin, an adipokine produced by adipose tissue, exerts neuroprotective effects through anti-oxidative activity and its promotion of hippocampal progenitor cell proliferation. As underweight patients may have decreased levels of leptin, a lower BMI may result in less neuroprotection after neurological insult 25 .
Our results also showed that a higher BMI was signi cantly associated with worse frontal/executive function, speci cally in phonemic and semantic uencies. This nding is in agreement with those of previous studies that have linked obesity to temporal atrophy 13 . Moreover, one study that used highresolution 3D MRI scans reported that obese individuals had a signi cantly lower density of gray matter in the frontal lobe, post-central gyrus, and middle frontal gyrus than control group participants 14 . Obesity has also been reportedly related to executive dysfunction with other cognitive functions preserved, even in neurologically healthy adults without cognitive impairment 15 . Speci cally, obese adults perform worse on executive function tests, especially those testing verbal interference, than their counterparts with a lower BMI. These ndings may indicate vulnerability of the frontotemporal lobe in obese patients to acute stroke, regardless of lesion location.
Our study had some limitations. First, only participants' height and weight measurements were used to determine the BMI. Other adiposity data, such as abdominal circumference or waist-hip ratio, were unavailable; therefore, we could not address possible differences between leanness and being underweight and their relationships with long-term cognitive outcomes 26 . Future studies should consider replicating the approach of studies that have investigated the relationship between obesity and Alzheimer's disease through using the waist-to-hip ratio and waist circumference at multiple sites as a measure of central obesity. Second, we were unable to determine any causal relationships between BMI and cognition because this study spanned a relatively short observational period. Third, restrictions were inevitable concerning the number of participants studied and some of the explanatory variables because this study was the result of a secondary analysis of a multicenter study that aimed to investigate the prevalence of cognitive disorders post-stroke. Consequently, we did not include several image variables such as cerebral atrophy or white matter hyperintensities, which may have affected cognitive function post-stroke. Furthermore, we did not assess temporal changes in BMI from stroke onset. Nonetheless, the main strength of our study is that it is by far the rst study to investigate the effects of BMI on each cognitive domain after stroke. These ndings may provide additional evidence for pathophysiological aspects of BMI on each neural substrate after ischemic stroke.

Conclusions
Our research suggests that BMI may interact variably with cognitive domains post-ischemic stroke. Although being underweight might negatively affect global cognition post-stroke, obesity might also induce frontal lobe dysfunctions, speci cally in terms of phonemic and semantic word uency. Though exact pathomechanisms being unclear, these ndings may be attributed to the distinct roles of BMI on neural substrates, speci cally the frontal lobe area. A large prospective cohort study with a focus on neural substrates evaluated with brain imaging is necessary to elucidate further relationships between BMI and post-stroke cognitive outcomes and to validate our ndings.

Competing interests
The author(s) declare no competing interests.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.