Abstract
Background It is important to establish criteria to define Vascular Cognitive Impairment (VCI) in India as VCI is an image-based diagnosis and MRI changes resulting from age with prevalent vascular risk factors may confound MRI interpretation.
Objective To establish normative community data for MRI volumetry including white matter hyperintensity volume (WMHV), correlated with age-stratified cognitive scores and vascular risk factors in adults ≥40 years old.
Materials and Methods We screened 2651 individuals without known neurological morbidity, living in Mumbai and nearby rural areas, using validated Marathi translations of Kolkata Cognitive Battery (KCB) and Geriatric depression score (GDS). We stratified 1961 persons with GDS ≤9 by age and cognitive score, and randomly selected ten percent from each subgroup for MRI brain volumetry. Crude volumes were standardized to reflect percentage of intracranial volume.
Results MRI volumetry studies were done in 199 individuals (F/M = 90/109; 73 with BMI ≥25; 44 hypertensives; 29 diabetics; mean cognitive score 76.5). Both grey and white matter volumes decreased with increasing age. WMHV increased with age and hypertension. Grey matter volume decreased with increasing WMHV. Positive predictors of cognition included standardized hippocampal volume (HCV), urban living, education, and BMI, while WMHV and age were negative predictors. Urban dwellers had higher cognitive scores than rural, and (paradoxically) smaller HCV.
Conclusions Further study is warranted into sociodemographic and biological factors that mutually influence cognition. Our findings contribute to baseline diagnostic criteria for VCI and could help in early diagnosis and control of cognitive decline and its key risk factors.
1. Introduction
India is a highly populated, rapidly ageing nation. The proportion of the Indian population aged over 60 is projected to increase from 8 percent in 2015 to 19 percent in 2050.1 This transition can significantly change the Indian morbidity profile, with chronic conditions such as age-related cognitive decline and comorbidities exacting a formidable toll on the healthcare apparatus. Hypertension and diabetes mellitus, and other cardiovascular risk factors such as smoking, are known risk factors for cognitive impairment.2–4 While definitional criteria exist for Alzheimer’s Disease and Vascular Dementia, the same is not true for Vascular Cognitive Impairment (VCI). There is a lack of clarity on what constitutes VCI. To diagnose VCI clinically, one must have some degree of clinical cognitive impairment (short of dementia), corroborated by imaging findings of vascular compromise. The latter would comprise subclinical lacunes, microhemorrhages, periventricular white matter hyperintensities (WMH), and at times even non-lacunar small embolic infarcts or bleeds which may not be clinically overt. However, it is well known that some of these imaging features, especially WMH, can coexist with vascular risk factors (VRFs) such as advancing age, hypertension, diabetes mellitus, obesity and hyperlipidemia, without overt clinical motor-sensory deficits. In regions with a high prevalence of undetected or untreated VRF, such as India, these factors can confound the definition of VCI.
Thus, it is necessary to know the baseline values of cognition in different age groups in populations without prior stroke or dementia in particular regions, in correlation with prevalent VRFs and brain volumetry, to lay down the ‘normative data’ specific to that population. Based on this information, the definition of VCI can be formulated for that population.
Hence, our objectives in this study were to determine cognitive scores in an age-stratified ‘normal’ population (with no known history of prior clinical strokes or cognitive/behavioral disorders) from the western part of Maharashtra state in India (both rural and urban areas), to obtain data on VRFs in this population (hypertension, diabetes mellitus, alcohol and tobacco use, body mass index), and after stratification for cognition in each age bin, to obtain volumetry data on MRI in each stratified bin as a representative sample of that bin. These baseline parameters would serve as ‘normative data’ of cognitive scores in relation to age, VRFs and brain volumetry (including volume of WMHs, if noted), and would hence enable a definition of VCI specific to this population, in further studies.
2. Materials and Methods
2.1. Study design
Initially, we translated the Kolkata Cognitive Battery (KCB), a partially modified Hindi cognitive screening battery used by Ganguli et al.5 and the Geriatric Depression Score (GDS),6 into Marathi and validated both for field use. The KCB is a screening battery consisting of eight tests designed to test different cognitive domains. Test 1 is for Verbal Fluency, 2 is for Object Naming, 3 is the Mini Mental State Examination (MMSE), 4 is a Calculation task, 5 tests Immediate Recall, 6 tests Visuo-Constructional Ability, 7 tests Delayed Recall, and 8 tests Delayed Word Recognition. We used these instruments to conduct a cross-sectional interview-based survey in several urban and rural communities in Maharashtra state. We selected the urban study population from Dharavi and other neighborhoods in Mumbai city. The rural population was chosen from villages that had ongoing primary health care projects under the community health department of the parent hospital, which made it feasible to obtain information on the number of households, their family size, and the number of people over 40 years of age in these areas. We conducted awareness programs about the study in these communities in the month preceding the survey, through pamphlets and doorstep campaigning with the help of community health workers.
2.2. Data collection
We invited all persons aged 40 and over, and living in the selected localities, to take part in the survey. We screened 2752 persons who consented to participate in our study, from March 2015 to July 2016, and excluded 19 with known intellectual developmental disorder, 31 with possible previous stroke, 36 with previous psychiatric illness, and 15 persons reported by family to have previously diagnosed dementia or significant memory impairment impeding daily activities. We interviewed the remaining 2651 participants and collected their sociodemographic data, noted current daily use of tobacco and alcohol, and measured their height, weight, blood pressure (BP) and fasting blood glucose. Participants were classified as having hypertension, diabetes mellitus, or other cardiovascular conditions based on self-report corroborated by medical records and prescriptions. Participants who had high blood glucose or high BP detected for the first time during examination, had repeat BP measurements, and fasting and post prandial blood sugar study on a later date, to confirm the diagnoses in accordance with standard criteria.7,8
Finally, trained personnel screened the participants using the validated Marathi KCB and GDS tools and identified 1961 individuals with GDS scores in the normal range (GDS ≤9). We stratified this group into four subgroups using ten-year age bins (40–49, 50–59, 60–69 and ≥70). Each age group was further sub-stratified into three cognitive classes based on their KCB score (<80, 80–100 and >100). Our final sample for MRI brain imaging (n=199) comprised ten percent of the individuals from each cognitive category within each age bin (12 subgroups in total), selected using computer generated random numbers.
2.3. Volumetry
We used a Siemens 1.5T MRI system (1.5 T Avanto, Siemens Healthcare, Germany), equipped with a 32-channel head coil, for MRI scanning. For each participant, we acquired 3D Fluid sensitive T2 space Inversion Recovery (FLAIR) images with repetition time = 6000 ms, echo time = 369 ms, flip angle = 90°, field of view = 260 mm × 208 mm × 174 mm, slice thickness = 1.14 mm, 128 slices per slab, matrix = 256 × 174, and voxel size = 1.01 mm × 1.19 mm × 1.14 mm.
2.4. Image processing
Anatomical images were segmented into grey matter, white matter and cerebrospinal fluid, followed by spatial normalization of images into stereotactic space using the voxel-based morphometry tool box in SPM12.9 The normalized images were smoothed with an 8 mm isotropic Gaussian kernel. The hippocampal regional volume was extracted by atlas based parcellation using automated anatomical labelling, AAL atlas.10,11 Automated white matter hyperintensity (WMH) detection was done using FLAIR images with the lesion segmentation tool box (LST) implemented for SPM12 package.12,13 We determined WMH volume using lesion prediction algorithm (LPA SPM12). The processing methodology is described in detail elsewhere.14 Post processing was done by an independent team with prior experience in volumetric studies of Indian populations,14,15 who were blinded to the neuropsychological and demographic data.
2.5. Statistical methods
In order to adjust for differences in cranial size, we standardized crude volumetric measurements and expressed them as a percentage of the total intracranial volume. We analyzed the data using IBM SPSS Statistics for Windows, Version 20.0. (Armonk, NY: IBM Corp) and JASP.16 We used the Mann Whitney U test, the Chi square test, and Welch’s unequal variances t-test, to study differences in sociodemographic and vascular risk factors. We used the Wilcoxon signed rank test to compare left and right hippocampal volumes, Spearman’s rank correlation coefficient for non-parametric correlations after testing for normality and equality of variances, and Pearson product-moment correlation to study correlation between variables after adjusting for confounders. Our analysis is primarily explorative, without a definite outcome variable. Therefore, we did not correct our obtained P values for multiple comparisons. While reporting results, we have replaced P values with estimates of effects or association and corresponding 95% confidence intervals. We used rank biserial correlation coefficients as indicators of effect size for Wilcoxon T and Mann Whitney U tests, Cramer’s V for the Chi square test, and Cohen’s d for Welch’s t test. We used linear regression to identify key predictors of volumetric variables and cognition scores. We entered area of residence, sex, age, education, body mass index (BMI), vascular risk factors, blood glucose and pressure readings, and brain regional volumes found to be associated in bivariate analysis, as independent factors for volumetric dependent variables. For cognitive dependent variables, we entered area of residence, sex, 10-year age groups, educational level, BMI, vascular risk factors, blood glucose and pressure readings, and standardized brain region volumes as independent factors. Each regression model was checked for multicollinearity, and independent variables with variance inflation factor (VIF) more than five, or tolerance less than 0.4 were removed from the final models. We have reported P values as an additional data point for predictor variables in the regression models. However, we have selected predictors to report based on 95% confidence intervals, in place of P value.
2.6. Glossary of abbreviations
Cognitive scores
KCB = Kolkata Cognitive Battery; KCB Memory = Memory component: Test scores 5, 7, 8 combined; KCB Executive = Executive component: Test scores 1, 2, 4, 6 combined.
Crude regional brain volumes (measured in milliliters)
GMV = Grey matter volume; WMV = White matter volume; CSFV = Cerebrospinal fluid volume; LHCV = Left hippocampal volume; RHCV = Right hippocampal volume; HCV = Total hippocampal volume; WMHV = White matter hyperintensity volume; TIV = Total intracranial volume.
Standardized volumes
All expressed with a % sign suffixed to the corresponding crude volume abbreviation (such as GMV for grey matter volume and GMV% for standardized grey matter volume).
Vascular risk factors
HTN = Hypertension; DM = Diabetes mellitus; IHD = Ischemic heart disease; BMI = Body mass index; SBP = Systolic BP; DBP = Diastolic BP; PPR = Pulse pressure; Tobacco = Current daily use of tobacco in any form; Alcohol = Current daily use of alcohol in any form.
Statistical tests
MWU = Mann Whitney U test; Wilcoxon T = Wilcoxon signed rank test; ρ = Spearman’s rank correlation coefficient; r = Pearson product-moment correlation; Χ2 = Chi square; Welch t = Welch’s unequal variances t-test; RC = Rank biserial correlation coefficient; OR = Odds ratio; 95% CI = 95% confidence intervals.
2.7. Standard Protocol Approvals, Registrations, and Patient Consents
The study design was approved by the Cognitive Science Research Initiative of the Department of Science & Technology under the Government of India, and ethical approval was granted by the institutional ethical committee of Lokmanya Tilak Municipal Medical College and General Hospital, Sion, Mumbai. Written informed consent was obtained from all participants in our study.
2.8. Data availability
The master dataset has not been made publicly available as it contains confidential sociodemographic and clinical data of the participants. Anonymized data will be shared by reasonable request from any qualified investigator.
3. Results
3.1. Demography
(Table 1) Our study sample comprised 199 individuals (109 men), three-fourths of whom (150/199) were between 40 and 59 years of age. A similar proportion (146/199) were poorly educated, having completed primary school or less. More than one third were overweight or obese (73/199; 36.9%).
3.2. Cognitive scores
The mean cognitive score (KCB) of our sample was 76.5 (range 21–125). The KCB score decreased with increasing age and increased with education. Average KCB score was notably lower in rural as compared to urban participants (Tables 2, 3).
3.3. Volumetry and its associations
On average, both crude and standardized volumes were smaller for the left hippocampus than the right (Table 2).
3.3.1. Age
GMV% and WMV% decreased with increasing age. CSFV and WMHV, both crude and standardized, increased with age (Table 3, Fig 1).
Diagonal cells show histogram distributions of their respective variables.
GMV% = Standardized Grey Matter Volume; WMHV% = Standardized White Matter Hyperintensity Volume
3.3.2. Sex
Crude volumes were lower in women than men, but after standardization, this difference disappeared for all variables save WMV% (higher in women) and CSFV% (higher in men); WMHV did not show a definite sex difference (Tables 2, 3).
3.3.3. Locality
Crude and standardized CSFV and HCV were on average larger in rural participants, while average WMV% was larger in city dwellers (Table 3).
3.3.4. Education
We did not note a uniform increase or decrease in regional brain volumes with years of education. However, compared to literate participants, illiterate participants (n = 29) had smaller mean GMV% (38.5 vs 40.9), higher mean CSFV% (16.3 vs 13.6) and higher mean WMHV% (0.3 vs 0.1). Illiterate participants also had higher average CSFV (198.9 vs 166.6 ml) and WMHV (3.8 vs 1.4 ml). Mean crude and standardized volumes of the right hippocampus were higher in illiterate participants than literate participants (3.5 vs 3.2 ml; and 0.29 vs 0.26 percent).
Among our literate participants, crude and standardized hippocampal volumes increased with years of education after controlling for age (Fig 2).
Results of correlation analyses controlling for age: LHCV: r = 0.17, 95% CI = 0.007 –0.32; RHCV: r = 0.22, 95% CI =0.06 –0.36; HCV: r = 0.19, 95% CI =0.04 –0.34; LHCV%: r = 0.18, 95% CI =0.05 –0.30; RHCV%: r = 0.23, 95% CI =0.10 –0.36; HCV%: r = 0.21, 95% CI = 0.08 – 0.34.
3.3.5. VRFs
Diastolic BP (DBP) correlated positively with crude and standardized hippocampal volumes. (Table 4, Fig 3). Correlation controlling for age showed both Systolic BP (SBP) and pulse pressure (PPR) to correlate positively with WMH volumes (Table 4).
Blood pressure readings (both systolic and diastolic values) were available for 164 participants. Diagonal cells show histogram distributions of their respective variables.
LHCV% = Standardized Left Hippocampal Volume; RHCV% = Standardized Right Hippocampal Volume; SBP = Systolic BP; DBP = Diastolic BP.
3.3.6. Cognitive tests
High scores in Test 8 of the KCB (Delayed Word Recognition) were correlated with higher hippocampal volumes (Table 4).
3.4. Inter-relationships of regional volumes, controlled for age
GMV% decreased with increasing WMHV% (r = -0.60, 95% CI =-0.77 – -0.26; see Fig 1). CSFV% correlated negatively with both GMV% (r = -0.41, 95% CI =-0.55 – -0.29) and WMV% (r = -0.49, 95% CI =-0.61 – -0.39).
3.5. Multivar iate analysis
(Table 5) Urban living, higher education and BMI predicted higher KCB scores, while age and WMHV% predicted reduced scores. We noted that HCV% and education predicted an increase in the executive component of the KCB, while BMI predicted increase in the memory component.
Female sex and WMHV were the key predictors of decreased GMV. Post standardization, however, female sex and hypertension predicted increase in GMV%, while WMHV%, HCV% and age predicted decrease. For both crude and standardized HCV, education predicted increase in volumes while urban living predicted decrease. Age, hypertension and SBP predicted increased WMHV.
4. Discussion
Image-based definition of VCI poses difficulties in India, with its high prevalence of vascular risk factors, often undetected and untreated, resulting in subclinical vascular findings on the MRI. It is difficult to determine whether these MRI findings indicate or predict a true VCI, or whether they are normal for age and associated vascular risk factors. Our study sample consisted of participants who had no prior history of stroke or clinical dementia, and who, after screening, did not suffer from depression. Our data would thus enable an evaluation of MRI volumetry in correlation with vascular risk factors on the one hand, and with age-stratified cognitive scores on the other.
Among our study participants, 22.11% were hypertensive and 14.57% were diabetic. Compared to rural participants, urban dwellers had a higher percentage of hypertensives (26.2% vs 15.6%) as well as diabetics (19.7% vs 6.5%; Χ2= 6.59, Cramer’s V = 0.18, OR = 0.28, 95% CIOR = 0.10– 0.78). This finding is consistent with the results of a recent nationally representative study, although our urban-rural differences are wider.17
The men in our sample had larger mean volumes than women for all brain regions. This can be explained by difference in cranial size, as standardized volumetric values were similar for both. The average right hippocampal volume (both crude and standardized) was larger than the left among our participants; similar findings have been reported by researchers from India and abroad.18–20
Our study shows an increase in hippocampal volume with diastolic BP. Lower DBP has been shown to be associated with smaller right hippocampal volume in healthy elderly individuals.21 Lower SBP, DBP, and mean arterial pressure (MAP) have also been associated with smaller volumes of the thalamus and putamen, with lower MAP associated with reduced HCV in elderly hypertensives, and it has been suggested that in order to ensure adequate perfusion, the target BP should not be too low in elderly hypertensives.22
Our observation of literate participants having a higher mean GMV% than illiterate, is consistent with reports of volume and metabolism increases in GMV with education in the healthy elderly,23 and similar volumetric findings by other researchers.24
Literature indicates a positive effect of education on hippocampal volume.25 It has been demonstrated that education may exert a “ dose-dependent” effect on HCV, with age-related decrease in volumes being steepest among those least educated, and less pronounced in individuals with higher educational attainment.26 The results of our bivariate and multivariate analyses (a positive correlation between HCV and years of education in literate participants, and education predicting increased HCV in linear regression) provide further evidence of this association. It is possible that, in our sample of healthy individuals, higher levels of education were able to counteract typical age-related changes in hippocampal structure. The differential rates of decline may be functionally significant, resulting in greater retention of the cognitive functions, including declarative memory performance, which relies on this structure.27
Hippocampal volume loss has been reported to be specifically associated with decline in delayed word recall.28 This is indirectly supported by our observation of the association of hippocampal volumes with the delayed word recognition score from the KCB.
Our illiterate participants were older than literate (mean age 60.7 vs 51.0 years, Welch t = 4.47, Cohen’s d = 0.96, 95% CId = 0.50 – 1.41), with higher percentages of alcohol (24.1% vs 0.6%) and tobacco (27.6% vs 2.9%) users. We also noted a greater proportion of hypertensives (27.6% vs 21.2%) and diabetics (24.1% vs 12.9%) among them. These factors may explain their higher observed WMHV.
Our correlations of WMHV with age, blood pressure, and illiteracy are consistent with literature establishing age, hypertension, diabetes mellitus, smoking and low educational level as important risk factors for WMHV.29–34
Urban versus rural living deserves close examination as a possible determinant of vascular and neurological morbidity. Our findings are concordant with several Indian and international studies which have reported association of urban residence with higher cognitive scores.35–41 Our finding of lower mean cognitive scores and HCVs in women could not have influenced this association, as women were equally represented in both subsets (45.5% rural vs 45.1% urban).
It is noteworthy that our urban participants had smaller volumes than rural for all volumetric variables including HCV, despite their higher cognitive scores, and in contradiction to our regression analysis showing that HCV% predicts high executive function. Urban living alters brain structures, and has been linked to mental disorders such as schizophrenia.42,43 Animal models44,45 and observational studies of post-traumatic stress disorder patients,46–48 demonstrate that hippocampal volume is inversely correlated with exposure to stress. Our urban participants, most of whom were of a low socio-economic status, would possibly have faced the challenges of underprivileged backgrounds in their daily lives, which could explain their comparatively smaller HCV. The discrepancy between volume and cognition in urban dwellers could also be explained by the fact that age, education, WMHV and BMI, along with urban residence itself, also independently predicted cognition in this study. Our urban participants were younger than their rural counterparts (mean age 51.41 vs 54.04 years), and had longer years in school (5.32 vs 3.03 Welch t = -4.94, Cohen’s d = - 0.711, 95% CId = -1.005 – -0.415), smaller mean WMHV (1.52 vs 2.00 ml, 0.13 vs 0.16 %), and higher mean BMI (24.22 vs. 23.62), all of which could override the effect of smaller HCV. Results from a study with a participant profile similar to ours (healthy, non-demented, relatively young), do not indicate any association of smaller HCV with cognitive decline.49 In another study, it was found that baseline hippocampal volume, albeit strongly related to baseline cognition, did not predict subsequent cognitive change.50 Researchers have noted that not all subjects with hippocampal decline will develop dementia, either due to death preceding symptoms, or because hippocampal volume decline can be part of healthy ageing.28 The apparent paradox of smaller HCV with higher cognition in city dwellers may be explained by the (possibly synergistic) effect of sociodemographic, behavioral, and environmental factors that compensate for possible stress-induced hippocampal volume loss.
4.1. Limitations
Persons aged over 70 years and rural dwellers were inadequately represented in our sample due to the logistical difficulties of transporting the elderly, and of bringing people over long distances from rural areas to our urban MRI center. Our study is cross-sectional. While we have established associations between cognition, brain regional volumes and vascular risk factors, we cannot comment on causality, or the longitudinal ramifications of observed volumetric differences on cognitive decline, without follow-up investigation.
5. Conclusions
Our population-based study of adults without neuropsychiatric morbidity attempts to place cognitive scores stratified by age and education, within the context of MRI brain volumetry findings. Our study shows that both GMV and WMV decrease, and WMHV increases, with increasing age. Our findings also suggest a shrinkage of GMV with increasing WMHV. Urban living, education, and standardized hippocampal volume predict higher cognitive scores, while WMHV and age predict lower scores. We noted higher cognitive scores in city dwellers with, paradoxically, smaller hippocampal volumes. Further study is warranted into sociodemographic and biological factors that mutually influence cognition. Our findings, particularly the correlation of KCB scores with WMHV and age, could help guide the definition of VCI in India.
Data Availability
Anonymized data will be shared by reasonable request from any qualified investigator.
Funding information
This study was fully funded by the Cognitive Science Research Initiative of the Department of Science & Technology, Government of India.
Declarations of interest
None.
Acknowledgment
The authors thank Dr Kuldeep Singh, Radiologist at Gurmeet Imaging, Mumbai, for his role in obtaining MRI volumetry sequences.
Footnotes
Study funding: Fully funded by the Cognitive Science Research Initiative, Department of Science & Technology, Government of India (Grant ID: CSRI 129/2012).
Ethical approval: Approved by the Institutional Ethical Committee of Lokmanya Tilak Municipal Medical College and General Hospital, Sion, Mumbai (approval no. 15/2015).
Declarations of interest: None.
6. References
