Skip to main content
NCBI home page
Neurology logoLink to Neurology
. 2023 Dec 15;102(1):e207806. doi: 10.1212/WNL.0000000000207806

Association of Glycemic Variability With Imaging Markers of Vascular Burden, β-Amyloid, Brain Atrophy, and Cognitive Impairment

Hyemin Jang 1,*, Sungjoo Lee 1,*, Sungsik An 1,✉, Yuhyun Park 1, Soo-Jong Kim 1,✉, Bo Kyoung Cheon 1, Ji Hyun Kim 1,✉, Hee Jin Kim 1,✉, Duk L Na 1, Jun Pyo Kim 1,✉, Kyunga Kim 1,†, Sang Won Seo 1,†
PMCID: PMC10834128  PMID: 38165363

Abstract

Background and Objective

We aimed to investigate the association between glycemic variability (GV) and neuroimaging markers of white matter hyperintensities (WMH), beta-amyloid (Aβ), brain atrophy, and cognitive impairment.

Methods

This was a retrospective cohort study that included participants without dementia from a memory clinic. They all had Aβ PET, brain MRI, and standardized neuropsychological tests and had fasting glucose (FG) levels tested more than twice during the study period. We defined GV as the intraindividual visit-to-visit variability in FG levels. Multivariable linear regression and logistic regression were used to identify whether GV was associated with the presence of severe WMH and Aβ uptake with DM, mean FG levels, age, sex, hypertension, and presence of APOE4 allele as covariates. Mediation analyses were used to investigate the mediating effect of WMH and Aβ uptake on the relationship between GV and brain atrophy and cognition.

Results

Among the 688 participants, the mean age was 72.2 years, and the proportion of female participants was 51.9%. Increase in GV was predictive of the presence of severe WMH (coefficient [95% CI] 1.032 [1.012–1.054]; p = 0.002) and increased Aβ uptake (1.005 [1.001–1.008]; p = 0.007). Both WMH and increased Aβ uptake partially mediated the relationship between GV and frontal-executive dysfunction (GV → WMH → frontal-executive; direct effect, −0.319 [−0.557 to −0.080]; indirect effect, −0.050 [−0.091 to −0.008]) and memory dysfunction (GV → Aβ → memory; direct effect, −0.182 [−0.338 to −0.026]; indirect effect, −0.067 [−0.119 to −0.015]), respectively. In addition, increased Aβ uptake completely mediated the relationship between GV and hippocampal volume (indirect effect, −1.091 [−2.078 to −0.103]) and partially mediated the relationship between GV and parietal thickness (direct effect, −0.00101 [−0.00185 to −0.00016]; indirect effect, −0.00016 [−0.00032 to −0.000002]).

Discussion

Our findings suggest that increased GV is related to vascular and Alzheimer risk factors and neurodegenerative markers, which in turn leads to subsequent cognitive impairment. Furthermore, GV can be considered a potentially modifiable risk factor for dementia prevention.

Introduction

Alzheimer disease (AD) and cerebral small vessel disease (CSVD) are the 2 most common causes of dementia in the older people.1 AD and CSVD share common risk factors, including age and vascular risk factors such as hypertension, diabetes mellitus (DM), and obesity.2,3 These risk factors have insidiously and gradually affected the development of dementia over several decades. According to the amyloid cascade hypothesis, AD is characterized by beta-amyloid (Aβ) deposition, which in turn leads to the accumulation of neurofibrillary tangles, neuronal injury, and eventually, cognitive impairment, causing dementia.4 CSVD is characterized by extensive white matter hyperintensities (WMH), which are also associated with neuronal injury represented as cortical atrophy, and cognitive impairment, which leads to the development of vascular dementia.5,6 Our previous study also suggested that Aβ and WMH affected cognitive impairment with the mediation of brain atrophy, measured by cortical thickness and hippocampal volume, in specific regions.7

The association between DM and these neuroimaging biomarkers is inconsistent, although some epidemiologic studies have shown that DM increases the incidence of AD and CSVD.8-10 Recently, glycemic variability (GV), which refers to the dynamic change in blood glucose, has received increasing attention because it is considered a relevant risk factor of DM-related macrovascular and microvascular complications.11-14 Several studies have reported that GV may be associated with cognitive impairment.15,16 However, the association between GV and neuroimaging biomarkers of cognitive impairment, particularly related to each dementia subtype, remains to be elucidated. We hypothesized that GV would increase the CSVD burden, considering its detrimental effects on microvascular complications.17 In addition, GV could be a key to understanding the complex relationship between blood glucose and AD pathology. Furthermore, considering the relationship between WMH and Aβ uptake, and brain atrophy and cognition, it is reasonable to expect that WMH and Aβ uptake might mediate the relationship between GV and brain atrophy or cognition. Therefore, it is worthwhile to determine whether GV affects Aβ uptake or CSVD and whether it contributes to neuronal damage and cognitive impairment.

This study aimed to investigate the association between GV and Aβ uptake and CSVD in nondemented individuals in a memory clinic cohort. We also determined whether Aβ or WMH might mediate the relationship between GV and brain atrophy and cognition.

Methods

Study Participants

We included a total of 1,016 nondemented participants, including those who were cognitively unimpaired (CU) and those with mild cognitive impairment (MCI), who underwent Aβ PET, between August 2015 and March 2021, at the memory clinic of Samsung Medical Center (SMC) in Seoul, Korea.18 CU was defined to have normal cognition on neuropsychological tests, which is defined as scoring above −1.0 SD of age-matched and education-matched norms for memory domain and −1.5 SD for other cognitive domains.19 MCI was defined to meet the following criteria20: (1) subjective cognitive complaints by the participants or caregiver; (2) objective cognitive impairment in any cognitive domain (below the −1.0 SD of age-matched and education-matched norms in memory or −1.5 SD in other cognitive domains); (3) no impairment in activities of daily living; and (4) nondemented.21 All participants underwent comprehensive dementia evaluation including neuropsychological test,22 blood tests, and brain MRI. For this study, the exclusion criteria included the following: (1) WMH due to etiologies other than vascular pathology, such as radiation injury, leukodystrophy, or toxic/metabolic disorders; (2) traumatic brain injury; (3) normal pressure hydrocephalus; (4) territorial infarction (involving at least 2 or more middle cerebral artery subdivisions, anterior or posterior cerebral arteries, with a shortest length over 2 cm); (5) neurodegenerative disorders other than AD or ischemic etiologies such as corticobasal syndrome, progressive supranuclear palsy, frontotemporal dementia, or Lewy body/Parkinson disease dementia; and (6) rapidly progressive and treatable dementia.

Because GV was calculated from repetitive fasting glucose (FG) levels, which were obtained from electronic medical records in the SMC, we excluded 15 individuals whose FG levels were not measured at least once, between January 2008 and June 2021. We excluded 164 because they underwent the FG test at time intervals of >3 years, based on the measurement date of Aβ PET, which makes all FG measurements in time interval of −3 years to +3 years from the date of Aβ PET to be only included. Then 149 were excluded because they had FG levels measured less than 3 times. Finally, the cohort comprised 688 participants. Of these participants, 669 had Aβ values, 604 had cortical thickness values, 585 had both Aβ and cortical thickness values, and 643 had Aβ values and cognitive test results. The mean number of measurements for the FG level in each patient was 10.45. This study was approved by the institutional review board of the SMC.

Independent Variables and Covariate Assessment

GV was defined as the intraindividual visit-to-visit variability of FG levels, which were obtained in fasting blood samples. The intraindividual SD, which is the most widely used method,23 was calculated for each participant as an index of visit-to-visit glucose variability. The intraindividual mean values of FG levels were also considered because of the well-known linear relationship between the intraindividual mean values and SDs.23 DM was defined as self-reported diagnosis by specialist or current use of diabetic medications or insulin, and hypertension was defined as self-reported diagnosis or current use of antihypertensive medications. Hyperlipidemia and cardiac diseases were defined as self-reported diagnosis (hyperlipidemia and cardiac diseases including coronary heart disease, arrhythmia, or cardiac myopathy) or current use of any related medications.

Aβ PET Acquisition and Quantification Using Direct Comparison of Florbetaben and Flutemetamol Centiloid Values

All participants underwent either 18F-florbetaben (FBB) or 18F-flutemetamol (FMM) PET scans at SMC using a Discovery STe PET/CT scanner (GE Medical Systems, Milwaukee, WI). A 20-minute emission PET scan was performed in dynamic mode (consisting of 4 × 5-minute time frames) 90 minutes after each participant gest injection of 311.5 MBq FBB and 197.7 MBq FMM, respectively. Three-dimensional (3D) PET images were reconstructed in a 128 × 128 × 48 matrix with a 2 × 2 × 3.27-mm voxel size using the ordered-subset expectation maximization algorithm (FBB, iteration = 4 and subset = 20; FMM, iteration = 4 and subset = 20).

To standardize quantification of Aβ uptakes on PET images that used 2 different ligands, a direct comparison of the FBB-FMM centiloid (dcCL) method was used.24,25 Specifically, Aβ uptakes were quantified using BeauBrain Morph of BeauBrain Healthcare Co., Ltd., which performs fully automated image analysis of Aβ uptakes on PET images. In brief, the dcCL values were obtained in the following steps23: (1) preprocessing of PET images, (2) obtaining standardized uptake value ratio (SUVR) of FBB and FMM PET imagings, (3) determination of the dcCL global cortical target volume of interest, and (4) conversion of the SUVR to dcCL values using the dcCL conversion equation. Detailed methods were described in previous studies.24,25

MRI Acquisition and Definition of the Presence of Severe WMH

All participants underwent brain MRI, including 3D T1 and FLAIR using a 3.0T MRI scanner (Philips 3.0T Achieva; Best, the Netherlands). T1-weighted data were acquired with the following parameters: repetition time = 9.9 ms, echo time = 4.6 ms, field of view, 240 × 240 mm2; voxel size = 0.5 mm3 isotropic. For 3D FLAIR data acquisition, the same parameters as T1-weighted data acquisition were used except for repetition time (4,800 ms), echo time (300 ms), and inversion time (1,650 ms).

The presence of severe WMH was determined by an experienced neurologist, as reported in the literature.26-28 In brief, severe WMH was defined as a deep WMH lesion of ≥25 mm, based on the longest diameter of the lesions, and periventricular WMH lesion of ≥10 mm, based on the maximum length measured perpendicular (cap) and horizontal (band) to the ventricle.26

Measurement of Cortical Thickness and Hippocampal Volume

We measured cortical thickness and hippocampal volume as neurodegeneration outcome measures, which are important both in AD and vascular dementia. Previously, Aβ was found to be associated with the hippocampal atrophy or decreased cortical thickness in the temporal-parietal regions, which was further associated with memory impairment while WMH were associated with decreased cortical thickness in the frontal region, which was further associated with executive and memory dysfunctions.7

The images were processed using the CIVET anatomical pipeline (version 2.1.0).29 We registered native MR images to the MNI-152 template by linear transformation30 and corrected for intensity nonuniformities using the N3 algorithm.31 Then the images were divided into white and gray matter, CSF, and background. Then the inner and outer surfaces of the cortex were automatically extracted using the marching-cubes algorithm to obtain the cortical thickness, which indicates the Euclidean distance between the inner and outer surfaces.32,33 The intracranial volume was calculated by measuring the total volume of voxels within the skull-stripped brain mask. Hippocampal volume was measured using an automated hippocampus segmentation method that used a graph-cut algorithm in combination with atlas-based segmentation and morphologic opening.

Neuropsychological Tests

Participants underwent the Seoul Neuropsychological Screening Battery, which is a detailed neuropsychological test for attention, language, visuospatial function, verbal and visual memory, and frontal-executive function.19,34 In this study, we used a composite score for the memory and frontal-executive cognitive domains and the Mini-Mental State Examination (MMSE) score for general cognition. Memory domain scores were generated by adding the scores from verbal and visual memory tests: scores on the Seoul Verbal Learning Test immediate and delayed recall (both ranging 0–36) and recognition (ranging 0–24) and scores on the Rey–Osterrieth Complex Figure Test immediate and delayed recall (both ranging 0–36) and recognition (ranging 0–24). The frontal-executive domain score was generated by adding scores from a category word generation task (animal name), a phonemic word generation task, and the Stroop color reading test (ranging 0–120).35

APOE Genotyping

We extracted genomic DNA from peripheral blood leukocytes using the Wizard Genomic DNA Purification kit according to the manufacturer's instructions (Promega, Madison, WI). Two single-nucleotide variations (rs429358 for and rs7412) in the APOE gene were genotyped using TaqMan SNP Genotyping Assays (Applied Biosystems, Foster City, CA) on a 7500 Fast Real-Time PCR System (Applied Biosystems).

Statistical Analyses

The baseline characteristics are expressed as mean ± SD or number (percentage) according to their variable types. Multivariable logistic regression and linear regression were used to identify whether GV was associated with the presence of severe WMH and Aβ uptake, respectively. The variables with p values ≤0.2 in univariable analyses were chosen as covariates in multivariable analyses. These selected covariates included intraindividual mean values of FG levels, age, sex, hypertension, DM, and presence of APOE4 allele. Presence of APOE4 allele was included as a covariate to control (1) the possible impact of APOE4 on endothelial dysfunction, blood-brain barrier integrity, and cerebral perfusion36 when the outcome was severe WMH and (2) the well-known association between APOE4 and Aβ uptake when the outcome was Aβ uptake. The adjusted estimates (i.e., regression coefficients in linear regression and odds ratios in logistic regression) and 95% confidence intervals (CIs) were obtained through the multivariable analyses. When the analyzed outcome was Aβ uptake, log transformation, a type of Box-Cox transformation method,37 was used to revise the highly skewed distribution of amyloid. Results for Aβ uptake were reported in the original scale. To investigate whether the association between GV and imaging markers is different according to the presence of DM, we added the interaction term of GV and DM in the multivariable regression analyses. Furthermore, to investigate whether the association between GV and Aβ uptake is different according to cognitive status, we added the cognitive status as a covariate and its interaction term with GV in the multivariable linear regression analysis.

To investigate the mediating effect of severe WMH or Aβ uptake on the pathway from GV to brain atrophy and cognitive impairment, mediation analyses were performed using Vansteelandt's imputation method with a natural effect model based on the counterfactual-based framework using the R package medflex.38 We set GV as the exposure and severe WMH or Aβ uptake as the mediator and considered various outcomes related to brain atrophy (frontal thickness for WMH; hippocampal volume and parietal thickness for Aβ uptake) and cognitive impairment (frontal-executive domain score for WMH; memory domain score for Aβ uptake; and MMSE for both). The covariates used in the multivariable regression analyses were considered as confounders. The natural indirect effect (NIE) and the natural direct effect (NDE) of the exposure to the outcome were estimated. Then the total effect was obtained by summing the NIE and the NDE. The robust variance estimation was used to obtain 95% CIs for those effects and p values in the mediation analyses. The “proportion mediated” was calculated by dividing the NIE by the total effect to identify the proportion of the indirect effect in the total effect.

All statistical analyses were performed using R 4.1.0 (Vienna, Austria; R-project.org). Statistical significance was set at p < 0.05.

Data Availability

Anonymized data of the analyses presented in this study are available upon request from the corresponding author.

Results

Study Participants

Demographic and clinical characteristics of the study participants are summarized in Table 1. Among 688 participants, 245 were CU and 443 were MCI. The mean age of total study participants was 72.2 ± 8.1 years, and 51.9% (n = 357) were female participants. Regarding vascular risk factors, the prevalence of hypertension and DM in the patients was 50% and 25%, respectively. Among the 688 participants, 231 (33.6%) showed Aβ positivity on PET, while 45 (6.5%) had severe WMH.

Table 1.

Characteristics of Study Participants

Total (n = 688) CU (n = 245) MCI (n = 443)
Age 72.2 ± 8.1 71.5 ± 7.6 72.5 ± 8.4
Female sex 357 (51.9) 136 (55.5) 221 (49.9)
APOE 4 carriers 230 (34.4) 59 (25.0) 171 (39.6)
Vascular risk factors
 Hypertension 341 (50.2) 124 (51.7) 217 (49.3)
 DM 170 (25) 57 (23.8) 113 (25.7)
 Hyperlipidemia 285 (41.9) 112 (46.7) 173 (39.3)
 Cardiac diseases 103 (15.1) 40 (16.7) 63 (14.3)
Aβ
 Positivity 231 (33.6) 45 (18.4) 186 (42.0)
 Centiloid (total n = 669) 33.6 ± 44.3 17.2 ± 32.0 42.9 ± 47.6
Severe WMH
 Presence 45 (6.5) 0 (0.0) 45 (10.2)
MMSE 26.7 ± 3.0 28.3 ± 1.8 25.7 ± 3.1
Open in a new tab

Abbreviations: Aβ = beta-amyloid; CU = cognitively unimpaired; DM = diabetes mellitus; MCI = mild cognitive impairment; MMSE = Mini-Mental State Examination; WMH = white matter hyperintensities.

Values are presented as mean ± standard deviation or number (%).

Association of GV With Severe WMH or Aβ Uptake

Table 2 summarizes the association between GV and the presence of severe WMH and Aβ uptake. GV was associated with the presence of severe WMH (coefficient [95% CI]; p value: 1.032 [1.012–1.054]; p = 0.002) after controlling for age, sex, hypertension, presence of APOE4 allele, DM, and mean FG. However, DM (1.22 [0.533–2.718]; p = 0.631), hypertension (1.89[0.947–3.958]; p = 0.079), and mean FG (1.003 [0.982–1.023]; p = 0.792) were not significantly associated with the presence of severe WMH.

Table 2.

Association Between GV and the Presence of Severe WMH and Aβ Uptake

Presence of severe WMH Aβ uptake
Covariates Coeff (95% CI) p Value Coeff (95% CI) p Value
GV 1.032 (1.012–1.054) 0.002 1.005 (1.001–1.008) 0.007
Mean FG 1.003 (0.982–1.023) 0.792 0.996 (0.993–0.999) 0.021
Age 1.096 (1.043–1.155) <0.001 1.007 (1.003–1.012) 0.002
Sex 1.37 (0.707–2.703) 0.355 1.045 (0.972–1.124) 0.231
APOE4 1.354 (0.675–2.662) 0.383 1.591 (1.474–1.717) <0.001
Hypertension 1.89 (0.947–3.958) 0.079 0.971 (0.901–1.047) 0.444
DM 1.22 (0.533–2.718) 0.631 0.921 (0.829–1.022) 0.121
Open in a new tab

Abbreviations: Aβ = beta-amyloid; Coeff = coefficient; DM = diabetes mellitus; FG = fasting glucose; GV = glycemic variability (standard deviation of FG); N/A = not available; WMH = white matter hyperintensities.

GV was also associated with Aβ uptake (1.005 [1.001–1008]; p = 0.007) after controlling for age, sex, hypertension, DM, mean FG, and APOE4 presence. However, DM (0.921 [0.829–1.022]; p = 0.121) and hypertension (0.971 [0.901–1.047]; p = 0.139) were not associated with Aβ uptake, and mean FG was inversely associated with Aβ uptake (0.996 [0.993–0.999]; p = 0.021). The interaction effects of DM and GV on severe WMH (p = 0.850) and Aβ uptake (p = 0.709) were not significant, suggesting that the direction of association between GV and each imaging marker (WMH and Aβ) did not change according to the DM diagnosis.

As a result of examining whether the effect of GV on Aβ uptake is different according to the cognitive status, the effect of GV was consistent even after adjusting the cognitive status in multivariable linear regression on Aβ uptakes (1.0034 [1.0001–1.0066]; p = 0.042). In addition, the interaction effect between GV and the cognitive status on Aβ uptakes was not significant (p = 0.758), which suggests that the GV does not have different effects on Aβ uptake according to the cognitive status.

Relationship Between GV, Presence of Severe WMH, Aβ Uptake, and Brain Atrophy

While the causal mediation effect of severe WMH was identified for frontal thickness, that of Aβ uptake was identified for hippocampal volume and parietal thickness (Figure 1). When GV, presence of severe WMH, and frontal thickness were set as the exposure, mediator, and outcome, respectively, there was a significant NDE (−0.00100 [−0.00163 to −0.00038]; p = 0.002) of GV on frontal thickness, while an NIE was not significant (0.00004 [−0.00008 to 0.00016]; p = 0.500) (Figure 1A).

Figure 1. Diagram of Mediation Analyses Indicating the Relationship of GV With Brain Atrophy.

Figure 1

Open in a new tab

Results for mediation analyses of the association between (A) GV, severe WMH, and frontal thickness; (B) GV, Aβ uptakes, and hippocampal volume; and (C) GV, Aβ uptakes, and parietal thickness. Aβ = beta-amyloid; GV = glycemic variability; WMH = white matter hyperintensities. Direct and indirect effects are represented as coefficients (95% confidence interval) with p values.

When GV, Aβ uptake, and hippocampal volume were set as the exposure, mediator, and outcome, respectively, there was a significant NIE (−1.091 [−2.078 to −0.103]; p = 0.030) of GV on hippocampal volume, whereas NDE was not significant (0.460 [−2.727 to 3.648]; p = 0.777) (Figure 1B). In addition, when GV, Aβ uptake, and parietal thickness were set as the exposure, mediator, and outcome, respectively, there was a significant NDE (−0.00101 [−0.00185 to −0.00016]; p = 0.019) and NIE (−0.00016 [−0.00032 to −0.000002]; p = 0.048) of GV on parietal thickness. In this model, Aβ uptake explained 13.7% (proportion mediated; 95% CI [0.9%–14.7%]) of the association between GV and parietal thickness (Figure 1C).

Relationship Between GV, Presence of Severe WMH, Aβ Uptake, and Cognition

A causal mediation effect of severe WMH was identified for frontal-executive domain scores and MMSE scores, and a causal mediation effect of Aβ uptake was identified for memory domain scores and MMSE scores (Figure 2). When we set GV, presence of severe WMH, and frontal-executive function as the exposure, mediator, and outcome, respectively, significant NDE (−0.319 [−0.557 to −0.080]; p = 0.009) and NIE (−0.050 [−0.091 to −0.008]; p = 0.020) were observed. In this model, the presence of severe WMH explained 13.5% (proportion mediated; 95% CI [9.0%–14.1%]) of the association between GV and frontal-executive dysfunction (Figure 2A). Moreover, when the MMSE score was set as the outcome in the same mediation analysis, the NDE was significant (−0.030 [−0.049 to −0.011]; p = 0.002), but the NIE was insignificant (p = 0.559; Figure 2B).

Figure 2. Diagram of Mediation Analyses Depicting the Relationship of GV With Cognition.

Figure 2

Open in a new tab

Results for mediation analyses of the association between (A) GV, severe WMH, and frontal-executive function; (B) GV, severe WMH, and MMSE ; (C) GV, Aβ uptakes, and memory function and (D) GV, Aβ uptakes, and MMSE. Aβ = beta-amyloid; MMSE = Mini-Mental State Examination; WMH = white matter hyperintensities. Direct and indirect effects are represented as coefficients (95% confidence interval) with p values.

When GV, Aβ uptake, and memory function were set as the exposure, mediator, and outcome, respectively, there was a significant NDE (−0.182 [−0.338 to −0.026]; p = 0.022) and NIE (−0.067 [−0.119 to −0.015]; p = 0.012) of GV on memory dysfunction. In this model, Aβ uptake contributed to 26.8% (proportion mediated; 95% CI [26.0%–36.1%]) of the association between GV and memory dysfunction (Figure 2C).

When the MMSE score was set as the outcome in the same mediation analysis, there was significant NDE (−0.021 [−0.040 to −0.001]; p = 0.037) and NIE (−0.006 [−0.012 to −0.001]; p = 0.023) of GV on MMSE. In this model, Aβ explained 23.3% (proportion mediated; 95% CI [22.5%–41.3%]) of the association between GV and MMSE (Figure 2D).

Discussion

This study investigated the association between GV, major neuroimaging markers for CSVD, Aβ, and brain atrophy and cognition in a relatively large cohort comprising nondemented participants. The major findings of this study are as follows: first, increased GV was predictive of the presence of severe WMH and increased Aβ uptake. Second, increased Aβ uptake mediates the relationship between GV and brain atrophy. Finally, GV was associated with cognitive impairment, directly or indirectly, by mediating the presence of severe WMH and increased Aβ uptake. Altogether, our findings suggested that increased GV was more likely to be related to abnormal neuroimaging markers and subsequent cognitive impairment. Therefore, our findings provide clinical insights that glucose control for targeting GV could be helpful in dementia prevention, regardless of the presence of DM.

Our first major finding was that GV was associated with the presence of WMH and increased Aβ uptake. Our finding is consistent with that of a previous study showing that GV was associated with increased WMH, despite being found only in APOE4 carriers.39 Our finding could be explained by the fact that GV exacerbates microvascular and cardiovascular complications by generating excessive reactive oxygen species, which causes oxidative stress13,40-42 and affects cerebral microvascular endothelial dysfunction.43 We also demonstrated that GV was associated with increased Aβ uptake, regardless of the mean FG level or the presence of DM. A previous study including patients with DM demonstrated that GV is associated with an increased risk of clinically diagnosed AD.44 In other studies with participants without DM, GV was associated with poor cognitive performance.15,45 However, these studies used clinical manifestations as an outcome measure, which could not provide convincing evidence regarding the association between GV and Aβ pathologies, which we investigated in this study. This association could be explained by several biological factors such as hyperglycemia or hypoglycemia, compensatory insulin secretion, and increased insulin resistance. A previous study, including animal models, showed that acute hyperglycemia increased hippocampal lactate production (neuronal activity marker) and interstitial fluid Aβ, potentially leading to Aβ deposition.46 In addition, increased insulin resistance and subsequent insulin receptor desensitization has been associated with reduced synthesis of insulin-degrading enzymes, which play a role in Aβ degradation.47

Another interesting finding was that GV had greater impact than hypertension, DM, and mean FG level on WMH or increased Aβ uptake. Previously, hypertension and DM were recognized as risk factors of AD and vascular dementia. However, given that WMH and Aβ uptakes are characteristic imaging markers of vascular dementia and AD, respectively, GV may be considered a more significant risk factor of these conditions compared with hypertension and DM alone. We also demonstrated that the association between GV and imaging markers were not different according to the presence of DM. Thus, our findings suggest that a fluctuating FG level is more harmful than a stable FG level, regardless of whether FG is higher or if DM is present. This suggestion is supported by accumulating evidence that high GV could be more dangerous than persistently high glucose levels.

Our study revealed that increased Aβ uptake mediated the relationship between GV and brain atrophy in the hippocampus and parietal regions. This is consistent with a previous study suggesting that GV was associated with limbic and temporal-parietal atrophy.48 However, this study did not consider coexisting pathologies, such as Aβ and CSVD, affecting brain atrophy as possible mediators. Our study demonstrated that some proportion of the association between GV and brain atrophy was mediated through AD pathologies. We also found other pathways through which GV directly affected frontal thickness and parietal thickness even without significant mediational effects of the presence of severe WMH and Aβ uptake, respectively. This implied that GV may directly result in neuronal death other than AD or CSVD pathology. This finding could be explained by recent evidence that GV could directly contribute to brain atrophy through neurotoxicity caused by oxidative stress or neuroinflammation.12,17,40 Alternatively, higher GV might worsen brain atrophy through unmeasured vascular injury such as to normal-appearing WM microstructure, which was reported in a previous study investigating the relationship between DM and brain integrity in middle-aged adults.10 This study demonstrated that DM was associated with worse brain atrophy, which also supports the direct effect of abnormal glucose metabolism on structural measures even without AD pathology, considering that middle-aged people are less likely to AD pathology.10

Our final major finding was that GV was associated with cognitive impairment directly or indirectly by mediating the presence of severe WMH and increased Aβ uptake, which are early biomarkers for vascular dementia and AD. Although these 2 types of dementia can present with impairment in any cognitive domains, they are mostly characterized by frontal and memory dysfunctions, respectively5,49; we hypothesized that the presence of severe WMH and increased Aβ uptake may mediate the relationship between GV and cognition. As expected, our findings indicated that the presence of severe WMH and Aβ uptake partially mediated the relationship between GV and respective cognitive impairments. Therefore, GV could be a potentially modifiable risk factor to prevent Aβ deposition and CSVD development, and neurodegeneration, which could lead to cognitive impairment in older patients.

The strength of this study is that we investigated the association among GV, major imaging biomarkers of dementia, and cognition in a relatively large nondemented cohort. However, our study has several limitations. First, we used Aβ PET uptake, the presence of severe WMH on MRI, and cortical thickness for measuring Aβ, CSVD, and neurodegenerative pathologies because pathologic confirmation was not available. Therefore, other pathologies contributing to neurodegeneration, such as tau, transactive response DNA-binding protein (TDP-43), argyrophilic grain disease, and hippocampal sclerosis, could not be considered. Second, owing to the inherent challenges of a retrospective cohort study, data regarding the participants' neuroimaging biomarker status at baseline were not available. Therefore, we were unable to determine their causal relationships. However, a retrospective cohort study could be considered a realistic alternative because changes in neuroimaging biomarkers appear gradually, and the cost of their assessment is very expensive. Third, because GV data were retrospectively obtained from the clinical data warehouse, there were differences in the duration of follow-up among participants despite controlling for the duration of follow-up in the process of calculating GV changes. However, these results could reflect the clinical situation in real-world settings, which could be considered real-world evidence during health care decisions. Finally, we used assumptions regarding confounders in causal mediation analysis. Although it is impossible to exclude all potentially unmeasured confounders, we attempted to include all possible confounders that could significantly affect the abovementioned relationships.

In conclusion, GV is associated with major neuroimaging markers of dementia and cognition in a cohort of individuals without dementia. Our findings indicate that GV may be considered a potential modifiable risk factor in the prevention of cognitive impairment and subsequent dementia.

Glossary

AD

Alzheimer disease

Aβ

beta-amyloid

CSVD

cerebral small vessel disease

CU

cognitively unimpaired

DM

diabetes mellitus

FBB

18F-florbetaben

FG

fasting glucose

FMM

18F-flutemetamol

GV

glycemic variability

MCI

mild cognitive impairment

MMSE

Mini-Mental State Examination

NDE

natural direct effect

NIE

natural indirect effect

SMC

Samsung Medical Center

SUVR

standardized uptake value ratio

WMH

white matter hyperintensities

Appendix. Authors

Name Location Contribution
Hyemin Jang, MD, PhD Alzheimer's Disease Convergence Research Center, Samsung Medical Center; Department of Digital Health, Samsung Advanced Institute for Health Sciences & Technology, Sungkyunkwan University; Department of Neurology, Samsung Medical Center, Sungkyunkwan University School of Medicine; Neuroscience Center, Samsung Medical Center; Department of Neurology, Seoul National University Hospital, Korea Drafting/revision of the article for content, including medical writing for content; major role in the acquisition of data; study concept or design; and analysis or interpretation of data
Sungjoo Lee, MS Department of Digital Health, Samsung Advanced Institute for Health Sciences & Technology, Sungkyunkwan University, Seoul, Korea Drafting/revision of the article for content, including medical writing for content; major role in the acquisition of data; study concept or design; and analysis or interpretation of data
Sungsik An, MD Alzheimer's Disease Convergence Research Center, Samsung Medical Center, Seoul, Korea Major role in the acquisition of data
Yuhyun Park, PhD Alzheimer's Disease Convergence Research Center, Samsung Medical Center, Seoul, Korea Analysis or interpretation of data
Soo-Jong Kim Alzheimer's Disease Convergence Research Center, Samsung Medical Center, Seoul, Korea Analysis or interpretation of data
Bo Kyoung Cheon, PhD Alzheimer's Disease Convergence Research Center, Samsung Medical Center, Seoul, Korea Analysis or interpretation of data
Ji Hyun Kim Alzheimer's Disease Convergence Research Center, Samsung Medical Center, Seoul, Korea Analysis or interpretation of data
Hee Jin Kim, MD, PhD Alzheimer's Disease Convergence Research Center, Samsung Medical Center; Department of Neurology, Samsung Medical Center, Sungkyunkwan University School of Medicine; Neuroscience Center, Samsung Medical Center, Seoul, Korea Major role in the acquisition of data
Duk L. Na, MD, PhD Alzheimer's Disease Convergence Research Center, Samsung Medical Center; Happymind Clinic, Seoul, Korea Major role in the acquisition of data; study concept or design
Jun Pyo Kim, MD Alzheimer's Disease Convergence Research Center, Samsung Medical Center; Department of Neurology, Samsung Medical Center, Sungkyunkwan University School of Medicine; Neuroscience Center, Samsung Medical Center, Seoul, Korea Major role in the acquisition of data
Kyunga Kim, PhD Department of Digital Health, Samsung Advanced Institute for Health Sciences & Technology, Sungkyunkwan University; Biomedical Statistics Center, Research Institute for Future Medicine, Samsung Medical Center; Department of Data Convergence and Future Medicine, Sungkyunkwan University School of Medicine, Seoul, Korea Drafting/revision of the article for content, including medical writing for content; major role in the acquisition of data; study concept or design; and analysis or interpretation of data
Sang Won Seo, MD, PhD Alzheimer's Disease Convergence Research Center, Samsung Medical Center; Department of Digital Health, Samsung Advanced Institute for Health Sciences & Technology, Sungkyunkwan University; Department of Neurology, Samsung Medical Center, Sungkyunkwan University School of Medicine; Neuroscience Center, Samsung Medical Center, Seoul, Korea Drafting/revision of the article for content, including medical writing for content; major role in the acquisition of data; study concept or design; and analysis or interpretation of data
Open in a new tab

Study Funding

This research was supported by a grant from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI) funded by the Ministry of Health & Welfare and Ministry of Science and ICT, Republic of Korea (grant number: HU20C0111), and funded by the Ministry of HealthWelfare, Republic of Korea (grant number: HR21C0885). This research was supported by the “National Institute of Health” Research Project (2021-ER1006-021) and a National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (NRF-2020R1A2C1009778). This study was also supported by the Future Medicine 2030 Project of Samsung Medical Center [#SMX1210771]. This research was supported by the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI) and Korea Dementia Research Center (KDRC), funded by the Ministry of Health & Welfare and Ministry of Science and ICT, Republic of Korea (HU20C0414).

Disclosure

The authors report no relevant disclosures. Go to Neurology.org/N for full disclosures.

References

  • 1.Jellinger KA, Attems J. Prevalence of dementia disorders in the oldest-old: an autopsy study. Acta Neuropathologica. 2010;119(4):421-433. doi: 10.1007/s00401-010-0654-5 [DOI] [PubMed] [Google Scholar]
  • 2.Livingston G, Huntley J, Sommerlad A, et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet. 2020;396(10248):413-446. doi: 10.1016/S0140-6736(20)30367-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Kalaria RN, Maestre GE, Arizaga R, et al. Alzheimer's disease and vascular dementia in developing countries: prevalence, management, and risk factors. Lancet Neurol. 2008;7(9):812-826. doi: 10.1016/S1474-4422(08)70169-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Jack CR, Knopman DS, Jagust WJ, et al. Tracking pathophysiological processes in Alzheimer's disease: an updated hypothetical model of dynamic biomarkers. Lancet Neurol. 2013;12(2):207-216. doi: 10.1016/S1474-4422(12)70291-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Kalaria RN, Erkinjuntti T. Small vessel disease and subcortical vascular dementia. J Clin Neurol. 2006;2:1-11. doi: 10.3988/jcn.2006.2.1.1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Kim CH, Seo SW, Kim GH, et al. Cortical thinning in subcortical vascular dementia with negative 11C-PiB PET. J Alzheimers Dis. 2012;31(2):315-323. doi: 10.3233/JAD-2012-111832 [DOI] [PubMed] [Google Scholar]
  • 7.Ye BS, Seo SW, Kim GH, et al. Amyloid burden, cerebrovascular disease, brain atrophy, and cognition in cognitively impaired patients. Alzheimers Dement. 2015;11(5):494-503.e3. doi: 10.1016/j.jalz.2014.04.521 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Michailidis M, Moraitou D, Tata DA, Kalinderi K, Papamitsou T, Papaliagkas V. Alzheimer's disease as type 3 diabetes: common pathophysiological mechanisms between Alzheimer's disease and type 2 diabetes. Int J Mol Sci. 2022;23(5):2687. doi: 10.3390/ijms23052687 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Wang KC, Woung LC, Tsai MT, Liu CC, Su YH, Li CY. Risk of Alzheimer's disease in relation to diabetes: a population-based cohort study. Neuroepidemiology. 2012;38(4):237-244. doi: 10.1159/000337428 [DOI] [PubMed] [Google Scholar]
  • 10.Weinstein G, Maillard P, Himali JJ, et al. Glucose indices are associated with cognitive and structural brain measures in young adults. Neurology. 2015;84(23):2329-2337. doi: 10.1212/WNL.0000000000001655 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Gorst C, Kwok CS, Aslam S, et al. Long-term glycemic variability and risk of adverse outcomes: a systematic review and meta-analysis. Diabetes Care. 2015;38(12):2354-2369. doi: 10.2337/dc15-1188 [DOI] [PubMed] [Google Scholar]
  • 12.Smith-Palmer J, Brändle M, Trevisan R, Orsini Federici M, Liabat S, Valentine W. Assessment of the association between glycemic variability and diabetes-related complications in type 1 and type 2 diabetes. Diabetes Res Clin Pract. 2014;105(3):273-284. doi: 10.1016/j.diabres.2014.06.007 [DOI] [PubMed] [Google Scholar]
  • 13.Echouffo-Tcheugui JB, Zhao S, Brock G, Matsouaka RA, Kline D, Joseph JJ. Visit-to-visit glycemic variability and risks of cardiovascular events and all-cause mortality: the ALLHAT study. Diabetes Care. 2019;42(3):486-493. doi: 10.2337/dc18-1430 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Zhou JJ, Schwenke DC, Bahn G, Reaven P, VADT Investigators. Glycemic variation and cardiovascular risk in the veterans affairs diabetes trial. Diabetes Care. 2018;41(10):2187-2194. doi: 10.2337/dc18-0548 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Yu Z-B, Zhu Y, Li D, et al. Association between visit-to-visit variability of HbA1c and cognitive decline: a pooled analysis of two prospective population-based cohorts. Diabetologia. 2020;63(1):85-94. doi: 10.1007/s00125-019-04986-8 [DOI] [PubMed] [Google Scholar]
  • 16.Rawlings AM, Sharrett AR, Mosley TH, Ballew SH, Deal JA, Selvin E. Glucose peaks and the risk of dementia and 20-year cognitive decline. Diabetes Care. 2017;40(7):879-886. doi: 10.2337/dc16-2203 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Ceriello A, Esposito K, Piconi L, et al. Oscillating glucose is more deleterious to endothelial function and oxidative stress than mean glucose in normal and type 2 diabetic patients. Diabetes. 2008;57(5):1349-1354. doi: 10.2337/db08-0063 [DOI] [PubMed] [Google Scholar]
  • 18.Jang H, Kim HJ, Park S, et al. Application of an amyloid and tau classification system in subcortical vascular cognitive impairment patients. Eur J Nucl Med Mol Imaging. 2020;47(2):292-303. doi: 10.1007/s00259-019-04498-y [DOI] [PubMed] [Google Scholar]
  • 19.Ahn HJ, Chin J, Park A, et al. Seoul Neuropsychological Screening Battery-dementia version (SNSB-D): a useful tool for assessing and monitoring cognitive impairments in dementia patients. J Korean Med Sci. 2010;25(7):1071-1076. doi: 10.3346/jkms.2010.25.7.1071 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Albert MS, DeKosky ST, Dickson D, et al. The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from the National Institute on Aging‐Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimer's Dement. 2011;7(3):270-279. doi: 10.1016/j.jalz.2011.03.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.McKhann GM, Knopman DS, Chertkow H, et al. The diagnosis of dementia due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement. 2011;7(3):263-269. doi: 10.1016/j.jalz.2011.03.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Kang Y, Jahng S, Na DL. Seoul Neuropsychological Screening Battery, 2nd ed. Human Brain Research & Consulting Co.; 2012. [Google Scholar]
  • 23.DeVries JH. Glucose variability: where it is important and how to measure it. Diabetes. 2013;62(5):1405-1408. doi: 10.2337/db12-1610 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Klunk WE, Koeppe RA, Price JC, et al. The Centiloid Project: standardizing quantitative amyloid plaque estimation by PET. Alzheimer's Demen. 2015;11:1-15.e1-4. doi: 10.1016/j.jalz.2014.07.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Cho SH, Choe YS, Kim HJ, et al. A new Centiloid method for (18)F-florbetaben and (18)F-flutemetamol PET without conversion to PiB. Eur J Nucl Med Mol Imaging. 2020;47(8):1938-1948. doi: 10.1007/s00259-019-04596-x [DOI] [PubMed] [Google Scholar]
  • 26.Noh Y, Lee Y, Seo SW, et al. A new classification system for ischemia using a combination of deep and periventricular white matter hyperintensities. J Stroke Cerebrovasc Dis. 2014;23(4):636-642. doi: 10.1016/j.jstrokecerebrovasdis.2013.06.002 [DOI] [PubMed] [Google Scholar]
  • 27.Kim S, Choi SH, Lee YM, et al. Periventricular white matter hyperintensities and the risk of dementia: a CREDOS study. Int Psychogeriatr. 2015;27(12):2069-2077. doi: 10.1017/S1041610215001076 [DOI] [PubMed] [Google Scholar]
  • 28.Kang SH, Kim ME, Jang H, et al. Amyloid positivity in the alzheimer/subcortical-vascular spectrum. Neurology. 2021;96(17):e2201-e2211. doi: 10.1212/WNL.0000000000011833 [DOI] [PubMed] [Google Scholar]
  • 29.Zijdenbos AP, Forghani R, Evans AC. Automatic "pipeline" analysis of 3-D MRI data for clinical trials: application to multiple sclerosis. IEEE Trans Med Imaging. 2002;21(10):1280-1291. doi: 10.1109/TMI.2002.806283 [DOI] [PubMed] [Google Scholar]
  • 30.Collins DL, Neelin P, Peters TM, Evans AC. Automatic 3D intersubject registration of MR volumetric data in standardized Talairach space. J Comput Assist Tomog. 1994;18(2):192-205. doi: 10.1097/00004728-199403000-00005 [DOI] [PubMed] [Google Scholar]
  • 31.Sled JG, Zijdenbos AP, Evans AC. A nonparametric method for automatic correction of intensity nonuniformity in MRI data. IEEE Trans Med Imaging. 1998;17(1):87-97. doi: 10.1109/42.668698 [DOI] [PubMed] [Google Scholar]
  • 32.Lerch JP, Evans AC. Cortical thickness analysis examined through power analysis and a population simulation. Neuroimage. 2005;24(1):163-173. doi: 10.1016/j.neuroimage.2004.07.045 [DOI] [PubMed] [Google Scholar]
  • 33.Im K, Lee JM, Lee J, et al. Gender difference analysis of cortical thickness in healthy young adults with surface-based methods. Neuroimage. 2006;31(1):31-38. doi: 10.1016/j.neuroimage.2005.11.042 [DOI] [PubMed] [Google Scholar]
  • 34.Kang Y, Na DL. Seoul Neuropsychological Screening Battery (SNSB). Human Brain Research & Consulting; 2003. [Google Scholar]
  • 35.Banerjee G, Jang H, Kim HJ, et al. Total MRI small vessel disease burden correlates with cognitive performance, cortical atrophy, and network measures in a memory clinic population. J Alzheimers Dis. 2018;63(4):1485-1497. doi: 10.3233/JAD-170943 [DOI] [PubMed] [Google Scholar]
  • 36.Silva NCBS, Bracko O, Nelson AR, et al. Vascular cognitive impairment and dementia: an early career researcher perspective. Alzheimers Dement (Amst). 2022;14(1):e12310. doi: 10.1002/dad2.12310 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Box GE, Cox DR. An analysis of transformations. J R Stat Soc Ser B (Methodol). 1964;26(2):211-243. doi: 10.1111/j.2517-6161.1964.tb00553.x [DOI] [Google Scholar]
  • 38.Steen J, Loeys T, Moerkerke B, Vansteelandt S. Medflex: an R package for flexible mediation analysis using natural effect models. J Stat Softw. 2017;76(11):1-46. doi: 10.18637/jss.v076.i1136568334 [DOI] [Google Scholar]
  • 39.Livny A, Ravona-Springer R, Heymann A, et al. Long-term variability in glycemic control is associated with white matter hyperintensities in APOE4 genotype carriers with type 2 diabetes. Diabetes Care. 2016;39(6):1056-1059. doi: 10.2337/dc15-2331 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Watt C, Sanchez-Rangel E, Hwang JJ. Glycemic variability and CNS inflammation: reviewing the connection. Nutrients. 2020;12:3906. doi: 10.3390/nu12123906 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Nusca A, Tuccinardi D, Albano M, et al. Glycemic variability in the development of cardiovascular complications in diabetes. Diabetes Metab Res Rev. 2018;34(8):e3047. doi: 10.1002/dmrr.3047 [DOI] [PubMed] [Google Scholar]
  • 42.Nalysnyk L, Hernandez-Medina M, Krishnarajah G. Glycaemic variability and complications in patients with diabetes mellitus: evidence from a systematic review of the literature. Diabetes Obes Metab. 2010;12(4):288-298. doi: 10.1111/j.1463-1326.2009.01160.x [DOI] [PubMed] [Google Scholar]
  • 43.Li W, Maloney RE, Aw TY. High glucose, glucose fluctuation and carbonyl stress enhance brain microvascular endothelial barrier dysfunction: Implications for diabetic cerebral microvasculature. Redox Biol. 2015;5:80-90. doi: 10.1016/j.redox.2015.03.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Li TC, Yang CP, Tseng ST, et al. Visit-to-visit variations in fasting plasma glucose and HbA(1c) associated with an increased risk of Alzheimer disease: Taiwan Diabetes Study. Diabetes Care. 2017;40(9):1210-1217. doi: 10.2337/dc16-2238 [DOI] [PubMed] [Google Scholar]
  • 45.Bancks MP, Carnethon MR, Jacobs DR Jr, et al. Fasting glucose variability in young adulthood and cognitive function in middle age: the Coronary Artery Risk Development in Young Adults (CARDIA) Study. Diabetes Care. 2018;41(12):2579-2585. doi: 10.2337/dc18-1287 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Macauley SL, Stanley M, Caesar EE, et al. Hyperglycemia modulates extracellular amyloid-β concentrations and neuronal activity in vivo. J Clin Invest. 2015;125(6):2463-2467. doi: 10.1172/JCI79742 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Umegaki H. Type 2 diabetes as a risk factor for cognitive impairment: current insights. Clin Interv Aging. 2014;9:1011-1019. doi: 10.2147/CIA.S48926 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Cui X, Abduljalil A, Manor BD, Peng CK, Novak V. Multi-scale glycemic variability: a link to gray matter atrophy and cognitive decline in type 2 diabetes. PLoS One. 2014;9(1):e86284. doi: 10.1371/journal.pone.0086284 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Yoon CW, Shin JS, Kim HJ, et al. Cognitive deficits of pure subcortical vascular dementia vs. Alzheimer disease: PiB-PET-based study. Neurology. 2013;80(6):569-573. doi: 10.1212/WNL.0b013e3182815485 [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

Anonymized data of the analyses presented in this study are available upon request from the corresponding author.

Articles from Neurology are provided here courtesy of American Academy of Neurology

RESOURCES