Abstract
Background:
Non-invasive biomarkers of cognitive impairment are needed. We aim to evaluate transcranial sonographic markers as predictors of cognitive impairment in a prospective cohort.
Objective:
To study the changes in the third ventricle diameter and the SN echogenicity between the baseline and the control visit, as well as its association with cognitive performance and the diagnosis of cognitive impairment in a prospective cohort.
Methods:
From the longitudinal population-based Asymptomatic Intracranial Atherosclerosis Study, we selected those subjects that received a complete transcranial sonography (TCS) and extensive cognitive testing, both at baseline and follow-up. We evaluated third ventricle (IIIv) width, echogenicity of substantia nigra (SN), and temporal changes of these parameters.
Results:
We included 289 participants with a median follow-up time of 7.16 years. Those subjects who developed cognitive decline (n = 23, 7.96%) had a larger IIIv at baseline than those who did not (0.54±0.14 cm versus 0.41±0.15 cm; p = 0.001). A cut-off point of 0.465 cm for the IIIv width was identified as an independent predictor of long-term cognitive impairment after adjustment for age, gender, educational level, and vascular risk score. Change in IIIv diameter after follow-up was not associated with diagnosis of cognitive impairment. The area of SN and the presence of hyperechogenicity of the SN remained stable over time and was not associated with the diagnosis of cognitive impairment.
Conclusion:
IIIv width assessed by TCS emerged as an independent predictor of long-term cognitive impairment.
INTRODUCTION
Cognitive decline is one of the major causes of disability and dependency among older people. Worldwide, around 50 million people have dementia, and there are nearly 10 million new cases every year [1]. Since neurodegenerative diseases present a long asymptomatic phase, the search for premorbid disease markers has increased in the last two decades. The detection of subjects in preclinical stages of ongoing neurodegeneration is crucial for an early treatment and for the application of new drugs with neuroprotective potential. Until date, biological and imaging markers have been developed in neurodegenerative diseases such as Alzheimer’s disease (AD). Cerebrospinal fluid biomarkers have shown to predict the risk of AD [2, 3], and amyloid positron emission tomography has also demonstrate to have a high specificity for AD plaques [4–6]. The development of disease biomarkers in other forms of neurodegenerative cognitive decline and the usefulness of less invasive biomarkers needs further investigation.
Transcranial sonography (TCS) has been extensively studied and applied in the field of movement disorders in the last decades [7–10] but its role in the diagnostic work up of patients with dementia has not been thoroughly explored [11]. The width of the third ventricle (IIIv) is a surrogate marker of brain atrophy [12] and some studies related this sonographic finding with cognitive decline [13–14]. Other deep brain structures, such as the substantia nigra (SN) have been rarely assessed by means of TCS in cognitive impairment [15–17].
Therefore, we aimed to evaluate the role of IIIv width, SN hyperechogenicity, and other sonographic measures, and temporal changes of these parameters in the prediction of long-term cognitive impairment.
METHODS
Setting and study participants
The current study was conducted between March 2007 and May 2017 at the Hospital Universitari Germans Trias i Pujol, Badalona, Spain. This study is included in the framework of the Barcelona-Asymptomatic Intracranial Atherosclerosis (AsIA) study, a prospective population-based study aimed to determine the prevalence of asymptomatic intracranial atherosclerosis, its associated clinical, molecular and genetic risk factors, and its prognostic impact. The complete study protocol has been reported in detail elsewhere [18]. Briefly, Barcelona-AsIA study included from March 2007 to June 2010 a random population sample of 933 Caucasian subjects older than 50 years with a moderate-high vascular risk and without history of stroke, coronary disease, neurodegenerative disease, or severe disability. At baseline, subjects received a complete transcranial and cervical neurosonological study and most of them also underwent a complete cognitive evaluation [19]. Barcelona-AsIA cohort was followed-up until May 2017 to evaluate incident long-term vascular events [20] and/or cognitive impairment. This study was approved by the Ethics Committee of our Institution (Germans Trias i Pujol University Hospital, Barcelona), and all patients gave their written consent to participate in the study.
In the present study we included subjects that underwent two in-person visits (baseline and follow-up), receiving both TCS and neuropsychologic studies at each visit.
Neuropsychological evaluation
At baseline and at follow-up visits, all participants underwent an extensive neuropsychological assessment, including different tests in four main cognitive domains: verbal memory, visuospatial skills and visual memory, speed, and language and verbal fluency. Verbal memory was assessed by means of the Word List Learning subtest of the Wechsler Memory Scale (WMS-III) [21] (immediate and delayed scores); visuospatial skills and visual memory were assessed by means of several subtests of the WMS-III [22] (Visual Reproduction I-Immediate and II-Delayed subtests and Visual Reproduction Copy); speed was evaluated with part A of the Trail Making Test [23], the Grooved Pegboard Test [24], and with the Digit Symbol subtest of the Wechsler Adult Intelligence Scale–Third Edition [25]; finally, verbal fluency and language were assessed by phonological fluency (letter P in 1 minute), semantic fluency (animals in 1 minute) [26], and the short form of the Boston Naming Test (BNT 15 items) [27] to assess visual naming ability.
The Mini-Mental State Examination (MMSE) was used as a measure of global cognitive function [28]. Subjective cognitive complaints were recorded by the patient and/or their family with the help of a trained neuropsychologist with the short version of the Spanish Informant Questionnaire on Cognitive Decline in the Elderly [29]. The Geriatric Depression Scale (GDS) was also applied [30].
The tests were administered by trained neuropsychologists blinded to clinical data in a single 1 h session the same day of the ultrasound examination.
Transcranial sonography
TCS was performed in all participants using a 2.5-MHz transducer (General Electric Vivid/Pro equipment (Horten, Norway) by two experienced ultrasound examiners (ELC, MH). The penetration depth was 14 cm and the dynamic range was 45–55 db. The examination was done from both sides using the temporal approach. Parenchyma images at the baseline visit were obtained by eliminating the vascular structures of the Willis polygon in the mesencephalic and thalamic planes. Images were digitally stored for further off-line evaluation (EchoPAC workstation, GE Healthcare). Digitally stored images were assessed by one experienced sonographer (AMCC) blinded to the neuropsychological tests. The following sonographic parameters were assessed: IIIv width, size of the frontal horn of the lateral ventricles, and echogenicity of the SN and the brainstem raphe nuclei. The minimal transverse diameter in the axial scanning plane of the IIIv, and the right and left frontal horn of the lateral ventricles were measured in the thalamic plane. Area of echogenicity in the SN was manually encircled and measured. In accordance to previously reported cut-off values, hyperechogenicity of the SN (SN+) was defined as an area of echogenic signal equal or greater than 0.20 cm2 [31]. Areas of echogeniticy of less than 0.20 cm2 were classified as normoechogenic. If one or both sides of the SN were hyperechogenic, the structure was classified as SN+. Echogenicity of the brainstem raphe was rated as hypoechogenic when this midline structure in the midbrain was interrupted or not visible [32]. Visible midbrain raphe was labeled as normoechogenic.
We excluded those subjects without optimal acoustic bone window that precluded analysis of at least one TCS parenchymal parameter.
Follow-up and definition of cognitive decline
Between April 2016 and May 2017, participants were invited to come to our center to undergo an in-person follow-up visit, where they underwent a new TCS and a new neuropsychological study, following the same methodology as that in the baseline visit.
In addition, a structured interview was done to specifically to assess their cognitive state. Those subjects in whom a possible cognitive impairment was detected during the structured interview, based on the patient’s and informant responses, and in the MMSE score changes from baseline visit, were later assessed by an experienced neurologist specialized in cognitive decline, who was aware of complete neuropsychological exams at baseline and follow-up visits. Additional ancillary tests, such as a computerized tomography or a magnetic resonance imaging (MRI) of the brain were done, if necessary, in order to characterize the cognitive decline and its possible etiology, following the regular procedures of the routine clinical practice.
To assess the severity of cognitive symptoms, the neurologist used the Clinical Dementia Rating (CDR score) [33]. Mild cognitive impairment (MCI) was defined as a change in cognition in comparison with the person’s previous level with preservation of independence in functional abilities and with a performance equal or lower than –1.5 standard deviations in at least one cognitive domain according to patients age and educational background, following the recommendations of the National Institute on Aging-Alzheimer’s Association [34]. Dementia was defined if the cognitive decline was significant to produce an impairment in social or occupational functioning. In those subjects who had already been diagnosed with cognitive decline or dementia in other center, the electronic medical records were checked in order to register the patient diagnosis.
Based on the scores of the neuropsychological tests, we also calculated the Global cognitive impairment index (GCII) [35] as a global measure of cognitive status, both at baseline and follow-up visits.
History of clinical depression was also assessed in the follow-up visit. Depression was considered when the patient had been diagnosed of depression by a physician or if they were taking antidepressants.
Statistical analyses
Descriptive demographical, clinical, and TCS data are given in means, standard deviations, number, and percentages. Comparisons between groups were done with chi-square test, Student’s t-test, and paired t-test, as appropriate. The scores obtained in the different neuropsychological tests were transformed into z scores calculated from the baseline data of the whole AsIA population with a baseline neuropsychological study (n = 747) and adjusted by age and education. A test was considered to be altered if the z was equal or inferior to –1.5 standard deviations. If one of the tests of each domain was altered, we considered that the domain was impaired. All neuropsychological variables were standardized using z-transformation and adjusted for sign according to performance direction (with higher values indicating better performance for all variables). To calculate the GCII, all z-transformed neuropsychological variables were added [35]. Binary logistic regression was used to compare the probability of having cognitive impairment in the follow up, adjusting for the covariates that were associated in the univariate analysis. Receiver operating characteristic (ROC) analysis was performed to evaluate the value of the IIIv width in the prediction of long-term cognitive decline. The linear relationship between the third ventricle size and the global cognitive impairment index was tested using the Spearmans rank coefficient. Statistical analyses were performed with SPSS version 18.0 (IBM, NYC, USA) and a Type I error of 5% was used in all analyses.
RESULTS
From the initial cohort of 933 subjects recruited at baseline, 463 performed the in-person follow-up visit after a median of 7.16 years. Sonographic measures and a complete neuropsychological evaluation were available at baseline and at follow-up visit for analysis in 289 participants (see flowchart in Fig. 1).
Demographic and clinical data of the baseline (n = 933) and the final cohorts (n = 289) are shown in Supplementary Table 1. The mean age at baseline of the final cohort was 64.14±6.69 years, 74% subjects were men. Demographic, clinical, and neuropsychological characteristics of the final cohort of subjects at follow-up visit are represented in Table 1. The third ventricle size was 0.42±0.15 cm at baseline visit, 41 subjects (16.87%) had SN+and 43 (17.70%) had hypoechogenicity of the brainstem raphe nuclei. The size of IIIv and the lateral ventricles (LV) increased from the first to the follow-up ultrasound study (IIIv: 0.42±0.15 versus 0.50±0.20 cm, right LV: 1.60±0.20 versus 1.68±0.27 cm, left LV 1.59±0.20 versus 1.70±0.26 cm).
Flowchart. IIIv, third ventricle; SN, substantia nigra.
Demographic, clinical, and neuropsychological variables at follow-up visit
MMSE, Mini-Mental State Examination; S-IQCODE, Spanish Informant Questionnaire on Cognitive Decline in the Elderly; TMT-A, Trail Making Test; WAIS-III, Wechsler Adult Intelligence Scale–Third Edition; GDS, Geriatric Depression Scale.
Cognition and ventricle size
After 7.16 years of follow-up, 23 (7.96%) subjects were diagnosed of cognitive decline, both MCI (18 subjects) or dementia (5 subjects). Differences between subjects with and without the diagnosis of cognitive impairment at follow-up are shown in Table 1. Etiologies of the cognitive decline (MCI or dementia) were attributed to AD in 3 subjects, 6 vascular dementia, 1 dementia with Lewy bodies, 2 mixed dementia, and 11 of undetermined etiology. Compared to subjects without cognitive impairment, those with cognitive decline had a larger IIIv width at baseline (0.54±0.14 cm versus 0.41±0.15 cm; p = 0.001) and at follow-up (0.65±0.22 cm versus 0.49±0.20 cm; p = 0.001) (Table 2). The size of the LV at baseline was similar in those subjects who developed cognitive decline during follow-up compared to those who did not (right LV 1.69±0.17 versus 1.59±0.20; p = 0.051; left LV 1.66±0.15 versus 1.59±0.20; p = 0.222). Change in IIIv diameter after follow-up was not associated with diagnosis of cognitive impairment (0.09±0.10 versus 0.10±0.12 cm; p = 0.424). Regarding the IIIv size at baseline, ROC analysis determined a cut-off point of 0.465 cm (sensitivity 90%, specificity 72.1%, positive predictive value (PPV) 20.40%, negative predictive value (NPV) 98.42%) in the prediction of long-term MCI/dementia with an area under the curve of 0.79 [0.65–0.92] (Fig. 2). In multivariable analyses adjusted by age, gender, educational level, and vascular risk score, IIIv diameter equal or higher than 0.465 cm at baseline was independently associated with the diagnosis of MCI/dementia with an odds ratio of 13.44 [3.59–50.37]. Although older age at baseline was associated with both larger IIIv diameter and with the diagnosis of cognitive impairment after follow-up (Table 1), there was no significant interaction between both variables, and IIIv diameter remained independently associated with incident cognitive impairment after adjustment for age and other variables.
ROC curve of the third ventricle width at baseline and the risk of long-term cognitive decline. A cut-off point of 0.465 cm presented a sensitivity of 90% and a specificity of 72.1% in the prediction of MCI/dementia, with an area under the curve of 0.79 [0.65–0.92].
Sonographic markers comparison between subjects with and without cognitive impairment at follow-up
SN, substantia nigra; SN+, hyperechogenicity of the SN; IIIv, third ventricle. *T-test; #chi-square/Fisher test.
We found a negative correlation between the IIIv width at baseline and the GCII at follow-up, showing that a higher IIIv width at baseline was associated with worse cognitive performance at follow-up visit (r =–0.232; p = 0.001) (Fig. 3).
Relationship between the third ventricle width and the global cognitive impairment index (GCII).
Other sonographic measures
The area of the SN was stable over time (Table 2) and was not associated with the diagnosis of MCI or dementia in our sample. The diameter of the frontal horns of the lateral ventricles increased at the follow-up visit, but this difference was not statistically significant (Table 2). Hypoechogenicity of the brainstem raphe nuclei was not associated with the presence of depression in the follow-up visit or with the score in the GDS scale (Supplementary Tables 2 and 3).
DISCUSSION
To our knowledge, this is the first population-based study to explore the predictive value of the IIIv size and other sonographic measures related to cognitive impairment. Our major finding was that IIIv width equal or higher than 0.465 cm was independently associated with the diagnosis of cognitive decline. We also found that the area of SN and the presence of SN+ remained stable over time and was not associated with the diagnosis of cognitive impairment in our sample. Finally, hypoechogenicity of the brainstem raphe nuclei was not associated with the presence of depression in our population.
Dementia is a slowly progressing medical condition in which pathogenic mechanisms start decades before the onset of clinical symptoms. However, there is still limited data of biomarkers in pre-clinical stages of dementia and MCI. In AD, cerebrospinal fluid biomarkers have shown to predict the risk of develop the disease [2, 3] and amyloid positron emission tomography has also demonstrated to have a high specificity for AD plaques [4–6]. Both biomarkers are highly concordant showing 80–90% of agreement across studies [36, 37], but their cost effectiveness as a screening tools needs to be evaluated [38]. The development of non-invasive disease biomarkers needs further investigation.
The IIIv width, measured by TCS, has been briefly studied in some neurological and systemic diseases [39–41]. The ventricular size have been proposed as promising marker of preclinical brain atrophy, can reliably be monitored by using TCS, and also seems to have an excellent interobserver reproducibility [12, 42–44]. Previous studies with MRI have demonstrated that serial measurement of ventricular volume and its enlargement over the time correlated with cognitive decline [44–47]. The measurement of the IIIv size by means of TCS closely match to that observed in MRI and computed tomography studies [12, 42], being a less-invasive, cheaper, and available technique that could be applicable to large populations. Similar to our results, Wollenweber et al. prospectively studied by means of TCS a cohort of subjects suggesting that the IIIv size differentiated between those subjects who developed a cognitive worsening after 5 years of follow-up [13]. Results from our study indicate that a larger IIIv diameter at baseline confer a greater risk of cognitive impairment with a high sensitivity and specificity, independently of age, sex and educational level. Also, the high negative predictive value of this marker could identify those subjects with a lower risk of developing cognitive decline. However, its low positive predictive value indicate that a second test should be applied in those subjects at risk of developing cognitive decline.
Thus, TCS could be a useful screening tool to detect populations at high-risk for the development of cognitive impairment. Further prospective studies, in combination with other biomarkers of cognitive decline, are needed to confirm the profitability of the technique in the work-up of cognitive decline.
A possible association between SN+ and the impairment in some cognitive tasks in healthy subjects has been suggested [15, 16]. Our results are not in this line since we found that SN+ was not associated with the diagnosis of cognitive decline in our population. However, the role of SN+ in cognitive decline needs further research, since SN+ could be related only to some etiologies of cognitive decline, such as dementia with Lewy bodies. Results from studies assessing the SN in AD and in dementia with Lewy bodies showed that the majority of AD patients have a normal echogenicity of the SN, but most patients with dementia with Lewy bodies have an enlargement of the SN echogenicity [48]. We also found that the area of SN and the presence of SN+ remained stable after 7 years of follow-up in this prospective cohort of subjects. Evidence that the area of SN hyperechogenicity does not change over time in the course of Parkinson’s disease was already provided in a longitudinal study [9]. However, these findings were not previously evaluated in the general population.
Hypoechogenicity of the brainstem raphe is a frequent feature in people with depression [49–51] and it is thought that reflects dysfunction of the serotonergic dorsal raphe nucleus. Our data showed that hypoechogenicity of the brainstem raphe nuclei was not associated with the presence of depression in our population nor with the development of cognitive decline. However, we did not assess in detail the presence of depression symptoms and we also did not apply any specific clinical criteria for the diagnosis of depression, so these findings should be interpreted with caution.
The strengths of our study include the longitudinal long-term population-based setting with a significant sample size, the assessment of the SN echogenicity and its association with cognitive decline, which has been rarely assessed in the literature. However, some limitations should be also considered when interpreting our results. First, the sample included in the present study have some differences with the original Barcelona-AsIA cohort, thus precluding conclusions from a population-based point of view. Second, parenchyma images at the baseline visit were obtained by eliminating the vascular structures of the Willis polygon in the mesencephalic and thalamic planes, and not explicitly stored to evaluate the parenchymal structures, so some measurements such as lateral ventricles width were only evaluated in a subset of patients. Finally, an inadequate bone window, present in around 25% of participants and which is a technical limitation of the technique, should also be remarked.
In conclusion, our results suggest that IIIv width assessed by TCS emerged as an independent predictor of long-term cognitive impairment. This biomarker may be used to identify those at risk of future cognitive decline helping to identify a high-risk population for trials with drugs for the prevention or change of the natural history of dementia.
Footnotes
ACKNOWLEDGMENTS
This work was supported by RETICS RD 16/0019/0020 from Instituto de Salud0020Carlos III and cofinanced by Fondo Europeo de Desarrollo Regional (FEDER).
The authors are grateful to the participants in this study.
This project was funded by the program of Promotion in the Biomedical Investigation and Health Sciences from the Institute of Health Carlos III (which is the main Public Research Entity funding, managing and carrying out biomedical research in Spain), and the Spanish Ministry of Science, Innovation and Universities (PI15/00605).
