Prediction of Dementia Onset using the Short-Term Memory Recall Test

doi:10.21203/rs.3.rs-4935873/v1

The increased prevalence of dementia is serious that necessitates simple tests that can identify those at a high risk. This study aimed to investigate the predictive ability of the Short-Term Memory Recall Test (STMT) for the development of dementia. A total of 162 patients aged ≥40 years, examined twice at specific intervals ranging from 1 to 7 years, were included in the study. Herein, 96 of 162 patients were being treated for three known risk factors for dementia: hypertension, hyperlipidaemia, and diabetes. Binomial logistic regression analysis combined with receiver operating curve analysis was used to assess the likelihood of developing dementia. A prediction equation was developed using the STMT score and treatment statuses of the three aforementioned diseases as predictor variables. Finally, p-values calculated using the equation and actual prognoses were subjected to discriminant analysis to determine the hit rate, sensitivity and specificity. Patients with lower STMT scores, older patients, women, and patients undergoing treatment for the three aforementioned diseases were more likely to develop dementia; in 96 patients, the hit rate for prediction was 75.0, with a sensitivity of 81.5 and a specificity of 61.3. This study demonstrates the reliability of the STMT in predicting the onset of dementia.

Health sciences/Diseases

Health sciences/Medical research

Short-Term Memory Recall Test

dementia

prediction

risk factor

hit rate

sensitivity

specificity

There is a growing need to predict the onset of dementia and develop effective prevention methods [1, 2]. Methods utilizing brain imaging and blood and cerebrospinal fluid (CSF) biomarkers have attracted attention due to their potential for the early detection of dementia. However, CSF testing cannot be easily performed in routine outpatient care settings due to its invasive nature. Additionally, although imaging methods to visualise amyloid-β (Aβ) have been incorporated into the health insurance in Japan [3], challenges remain, such as the high cost of testing and the limited number of facilities that can perform the tests. Furthermore, there are patients whose diagnosis differ from the imaging findings [4].

To develop methods for dementia prevention, it is better to focus on mild cognitive impairment (MCI), identified by Petersen et al. [5] as a precursor stage to dementia and characterised by memory impairment. Therefore, an accurate and simple test for the early detection of memory impairment is desirable. The Montreal Cognitive Assessment (MoCA), which is commonly used in the field to confirm MCI, has high accuracy [6]. However, the Japanese version of the MoCA (MoCA-j) is difficult to incorporate into routine clinical practice due to its high time burden. In recent years, several tools have been developed to facilitate touchscreen-based measurements. However, they have limitations such as being dependent on a stable internet connection. Households of older patients are likely to lack an internet connection, making it difficult to provide the service easily during visits.

Therefore, there is a need to develop a questionnaire-based cognitive function tests that is easy to use in clinical practice, has high predictive power, and does not rely on biomarker measurements that impose a burden on patients. Kobayashi et al. [7] developed the Short-Term Memory Recall Test (STMT) that consists of three memory tasks and can be performed in a short time, focusing only on memory impairment. The STMT is a simple and easy to administer test that is effective in assessing cognitive function, and thus, meets the requirements for a test to predict the onset of dementia.

Various factors such as age, sex, hypertension, hyperlipidaemia, and diabetes have been reported to influence the development of dementia [8–10]. In particular, hypertension, hyperlipidaemia, diabetes and their combination in middle ages are known risk factors for Alzheimer’s disease and dementia in later life [11]. Thus, it is necessary to examine how the presence of hypertension, hyperlipidaemia, and diabetes (hereafter, referred to as the three diseases) affects the prediction of dementia development. Therefore, this study aimed to test the predictive ability of STMT for the development of dementia and evaluate the impact of the three diseases on the prediction.

The results of the ROC analysis for the prediction of dementia onset using STMT score, age, sex, and elapsed years for the cohort of 162 participants and using STMT score, age, sex, elapsed years and treatment status for the three diseases for 96 participants are shown in Table 3. The area under the ROC curve (AUC) is a measure of task performance results for the binary classification, with values ranging from 0.0 to 1.0, with values closer to 1.0 indicating better performance. Overall Model Quality (OMQ) is a measure of model quality, with values closer to 1 indicating better quality.

The analysis results of variable combinations 1) to 8) for 162 participants are detailed in Table 3. ROC curves for the prediction of dementia using the STMT score, age, sex, and elapsed years as independent variables for 162 people are shown in Fig. 1A. To determine which combination of independent variables was a good fit as a model for predicting dementia at the second examination, we 1) assessed which combination of independent variables had the best AUC, and if there were similar combinations, and 2) assessed which combination had the best OMQ. If comparable combinations still existed, then 3) the combination with the best sensitivity and specificity was identified to determine the combination of independent variables that resulted in a well-fitted model. The combination of STMT score, age, sex, and elapsed years as independent variables demonstrated the best fit.

Details of the analysis using the combination of variables 9) to 15) for 96 participants are presented in Table 3. The ROC curves for the prediction of the onset of dementia using the STMT score, treatment status for the threee diseases, age, sex, and elapsed years in 96 cases is shown in Fig. 1B. The ROC analysis for the 96 cases showed that the combination employing the STMT score, treatment status for the 3 diseases, age, sex and elapsed years produced the best fit.

Using the independent variables of the best predictive model of dementia (STMT score, age, sex, and elapsed years [n=162]; STMT score, age, sex, elapsed years, and treatment status for the three diseases [n=96]), binary logistic regression analysis (forced entry method) was performed to calculate the probability of developing dementia (P). The dependent variable was whether the individual actually developed dementia.

The results for the two cohorts, comprising 162 patients and 96 patients, are shown below.

For each of the 162 and 96 cohorts, P was calculated for each case by entering the actual values for the independent variables of STMT score, sex, age, elapsed years, and treatment status for the three diseases.

In the case of the cohort comprising all 162 patients, a P greater than 75.1 was predictive of dementia at the time of the second examination, while in the case of the cohort comprising 96 patients undergoing treatment for the three diseases, the cut-off value for the P was 74.7.

Discriminant analysis was then performed with the calculated P as the independent variable and the occurrence of dementia at the second examination as the dependent variable. Discriminative predictive values, sensitivity and specificity were calculated. Table 4 shows the numbers and percentages of patients for whom the discriminant analysis predicted or not predicted dementia among the patients who are actually diagnosed or not diagnosed dementia in the second examination in the cohorts comprising 162 and 96 patients.

For n=162, the discrimination accuracy was {(43+83)/162}×100=77.8, sensitivity was (83/104)×100=79.8, and specificity was (43/58)×100=74.1. For n=96, the discrimination accuracy was {(19+53)/162}×100=75.0, sensitivity was (53/65)×100=81.5, and specificity was (19/31)×100=61.3.

One aim of the study was to investigate whether the STMT could be used to predict the onset of dementia. In 162 patients, when the STMT score alone was used as the independent variable to predict dementia onset by ROC analysis, the AUC and OMQ values were 0.75 and 0.67, respectively. AUC values < 0.5, 0.5–0.7, and > 0.7 suggest low, moderate, and high reliability of the model; respectively. Similarly, OMQ values < 0.60, 0.60–0.67, and > 0.67 indicate that the model is moderately reliable, moderately reliable but with room for improvement, and highly reliable, respectively. Accordingly, our model for predicting dementia using only the STMT score is moderately reliable. Therefore, predicting the onset of dementia may be possible using only the STMT score.

Another aim of the present study was to examine which independent variables could be used in addition to the STMT score to improve its performance in predicting the onset of dementia. Therefore, the effects of the addition of age, sex, and elapsed years as independent variables were examined in terms of changes in AUC and OMQ values. In the cohort comprising all 162 patients, the effects of the addition of age, sex, and elapsed years one by one were not noticeable regarding AUC and OMQ values. When age, sex and elapsed years were all added simultaneously with STMT score, both AUC and OMQ values were the largest and the reliability of the model improved. Cognitive function tends to decline with age [13]. There are sex differences in dementia, with women developing Alzheimer’s disease at twice the rate of men [8, 12], which has been attributed to a decline in female hormonal function [9]. Additionally, elapsed years increase with aging. Accordingly, it seems reasonable to conclude that patients with lower STMT scores, older age, and female sex are more likely to develop dementia.

The study additionally examined which combination of the variables make the best-fitted predictive model for the development of dementia when the treatment statuses of the three diseases were considered as independent variables. The results showed that when the treatment statuses of hypertension, hyperlipidemia and diabetes, in addition to STMT score, age, sex and elapsed year, were all considered as independent variables, the AUC value was 0.79 and the OMQ value was 0.70, indicating that the prediction model was highly reliable.

Studies have reported the association between hypertension and risk of cognitive impairment, with untreated long-term hypertension leading to structural changes in the arterial wall that cause chronic cerebral hypoperfusion [14, 15]. Factors such as age at the onset of hypertension, chronicity, and antihypertensive medication play an important role in determining the risk of cognitive impairment [16, 17]. Meawhile, hypotension is also associated with the risk of developing dementia. The Swedish Kungsholmen Project reported that a diastolic blood pressure < 65 mmHg was associated with a relative risk of dementia of 1.5. Furthermore, in older people with a systolic blood pressure < 160 mmHg, a fall in systolic blood pressure of ≥ 15 mmHg is associated with a 3.1-fold increased risk of developing dementia [18].

Hyperlipidemia is a cardiovascular risk factor and strongly associated with atherosclerotic lesions and ischemic stroke [19]. Furthermore, type 2 diabetes is a known risk factor for atherosclerotic lesions and ischaemic stroke, with probabilities of 21% and 18–22% respectively [20]. In general, 20% of individuals are expected to develop dementia after a stroke, with the risk of dementia reported to increase by 3% annually [21]. A systematic review of Mendelian randomized studies reported that higher total cholesterol and low-density lipoprotein cholesterol levels were associated with a higher risk of Alzheimer’s disease. In contrast, higher high-density lipoprotein cholesterol levels have been reported to lead to a lower incidence of Alzheimer’s disease [22].

Several studies have established that hyperlipidemia increases the risk of cognitive impairment in patients with type 2 diabetes, and hyperglycemia is involved in the development of vascular lesions, leading to vascular dementia and Alzheimer’s disease [23]. Complications associated with diabetes, e.g. hypoglycemia, are risk factors for dementia. Severe hypoglycemic episodes in streptozotocin-induced diabetic rats were significantly associated with neuronal damage in both cortex and hippocampus [24]. Additionally, patients with hypoglycemia have a twofold higher risk of developing dementia compared with those without [14, 19, 25–27]. Increased frequency of hypoglycemic episodes has been reported to increase the risk of dementia progression [26].

A longitudinal study examining concurrent hypertension and diabetes reported that hypertension is a risk factor for cognitive decline only when it is accompanied with diabetes [28]. However, contrasting findings have been obtained from other studies that did not observe a similar relationship between hypertension and diabetes [29, 30]. Furthermore, some reports have examined the incidence of dementia in patients with both hypertension and diabetes and observed no increasing trend.

Despite some contradicting reports, the majority of literature supports the notion that the presence of the three diseases contributes to the development of dementia. This corroborates the findings of this study that people with lower STMT scores, older age, female sex, and the presence of the three diseases are more likely to develop dementia.

Early detection of MCI is important to prevent the onset of dementia. Regarding this, our study confirms that the STMT, a tool for the early detection of memory impairment, can be used as a tool for the prediction of dementia. Since the STMT is a simple tool that does not require specialized instruments, it may be used in daily clinical practice. MoCA is recognized for its high accuracy in identifying patients with MCI and is used in clinical practice [6]. However, MoCA-j is difficult to incorporate widely into routine clinical practice due to its high time burden. A simple test for predicting dementia using shape recognition has been proposed [31], which assesses more cognitive functions, such as constructive skills, in addition to memory functions. Thus, it may be difficult to obtain stable evaluation results. In addition, a computerized touchscreen-based test for the early diagnosis of dementia has been developed by Tsuboi et al. [32], which requires specialized tools such as touchscreens. In comparison, the STMT needs no special tools and only considers memory functions, which may lead to fewer discrepancies in the assessment results. In the STMT, the probability of developing dementia (P) is calculated for each individual by substituting the necessary values (the STMT score, sex, age, and treatment statuses of the three diseases) into the discriminant formula given in the Results section. A P > 74.7 predicts the development of dementia.

This study was based on data obtained from a single medical institution, which could introduce bias in interpreting the results. Future studies should include data from multiple institutions. Additionally, this study did not examine the actual values of LDL cholesterol, HbA1c, and other indicators for patients treated for hypertension, hyperlipidemia, and diabetes, nor did it assess the extent of disease control. Future research should investigate in detail how treatment of these three diseases affects the results. Moreover, other factors influencing the development of dementia, such as the subject's living environment, should also be considered.

This study aimed to assess the predictive ability of the STMT for the development of dementia and further evaluate the impact of the three diseases on the prediction. The prediction results based on the STMT score alone were statistically significant but the accuracy was further improved when age, sex, elapsed years, and the three diseases were added as predictor variables. Although the STMT is a simple method that does not require specialized tools, the results of this study show that the it is a reliable predictor of dementia onset. In conclusion, the STMT is a useful tool for predicting dementia onset in clinical practice.

Participants

Herein, patients presenting to the Forgetfulness Outpatient Clinic at Azuma Street Clinic were included. At the first visit, a dementia specialist (N.K.) conducted a comprehensive interview along with a physical and neurological examination. Concurrently, non-participating staff administered the Mini Mental State Examination (MMSE) and the STMT. The same tests were performed during a second consultation at specific intervals. The diagnosis of dementia was made according to the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision. The decision to see the patient for a second visit was left to the patient.

Of the 2,913 patients aged ≥ 40 years who first visited the Forgetfulness Outpatient Clinic at the Azuma Street Clinic between 1 April 2011 to 31 March 2021, 162 (47 men and 115 women, mean age 79.5 years) returned for a second visit (Table 1). As shown in Table 2, the time between the first and second visit ranged from 1 to 7 years, with a mean duration of 2.20 years. At the time of the first and second visits, these patients were assessed and checked whether they were receiving treatment (including medication and diet) for the three diseases, which revealed that 96 patients (25 men and 71 women) were undergoing treatment. Table 2 presents detailed characteristics of the 96 patients.

Data analysis

All data were analyzed using SPSS ver. 27. First, the accuracy of the STMT in predicting the development of dementia was assessed in 162 out of 2,913 patients for whom data from the second-visit were available.

We used all comibinations of four independent variables, namely STMT score, age, sex, and years to diagnosis at the second examination (hereafter, referred to as ‘elapsed years’) as independent variables in the following analyses. The eight combinations were as follows: 1) STMT; 2) STMT and age; 3) STMT and sex; 4) STMT and elapsed years; 5) STMT, age, and sex; 6) STMT, age, and elapsed years; 7) STMT, sex, and elapsed years; 8) STMT, age, sex, and elapsed years.
Binary logistic regression analysis was performed to determine predictive probabilities, with each combination of 1) to 24) as independent variables and the presence of dementia at the second examination as the dependent variable. When the independent variable was 1), the predictive probability was defined as probability I; when the variable was 2), the probability was probability II, and so on.
A receiver operating characteristic curve (ROC) analysis was performed with the calculated prediction probabilities I-XXIV as independent variables and the presence of dementia at the second examination as the dependent variable. The effect of treatment for the three diseases on the prediction of dementia onset was then examined in 96 patients (25 men and 71 women) out of 162 patients who were being treated for the three diseases.
For the 96 patients, in addition to STMT score, age, sex, and elapsed years, the presence or absence of treatment (hereafter, referred to as the treatment status) for the three diseases were used as an independent variable to predict dementia onset. However, repeating calculations for all 56 possible combinations was unrealistic. Therefore, we restricted the analyses to the following seven combinations of independent variables when assessing probabilities: 9) STMT, age, sex, elapsed years, and hypertension; 10) STMT, age, sex, elapsed years, and hyperlipidaemia; 11) STMT, age, sex, elapsed years, and diabetes; 12) STMT, age, sex, elapsed years, hypertension, and hyperlipidaemia; 13) STMT, age, sex, elapsed years, hypertension, and diabetes; 14) STMT, age, sex, elapsed years, hyperlipidaemia, and diabetes; 15) STMT, age, sex, elapsed years, hypertension, hyperlipidaemia, and diabetes.
Binary logistic regression analysis was performed using all eight combinations as independent variables and the presence of dementia at the second examination as the dependent variable. When the independent variable was 9), the predictive probability was defined as probability IX; when the variable was 10), the probability was probability X, and so on.
ROC analysis was performed with the calculated predictive probabilities IX-XV as independent variables and the presence of dementia at the second examination as the dependent variable.
After ROC analysis for 162 or 96 patients, the best combination of independent variables with the highest accuracy was selected for each patient cohort. A binary logistic analysis was then performed with the presence of dementia at the second examination as the dependent variable to create a dementia prediction equation. For 162 individuals, STMT score, age, sex, and elapsed years were included in the prediction equation; for 96 individuals, STMT score, treatment status of the three diseases, age, sex, and elapsed years were included as independent variables. The predicted probability of dementia (P) was then determined.
Discriminant analysis was performed using the obtained P as the independent variable and whether the patient was diagnosed with dementia at the second examination as the dependent variable, and the discriminant predictive value, sensitivity and specificity were calculated.

Sensitivity was defined as “the probability that a patient who should be diagnosed with dementia was correctly classified as having dementia” and calculated as (B/A) x 100, where B is the number of patients predicted to develop dementia and A is the number of patients who actually developed dementia.

Specificity was defined as “the probability that a patient who should be diagnosed as not having dementia was correctly classified as not having dementia” and was calculated as (D/C) x 100, where D is defined as the number of patients predicted not to develop dementia and C as the number of patients who did not actually develop dementia. The discriminant accuracy rate, which represents the percentage of correct discriminant results, was calculated as [(B+D)/(A+C)] × 100; an accuracy rate of 75 or more is considered good [12].

ACKNOWLEDGEMENTS:This work was supported by the JST Moonshot R&D (Grant Number JPMJMS2237-5).

AUTHOR CONTRIBUTIONS: Chiaki Oshiyama was responsible for writing the entire paper and organizing and analyzing the acquired data.Naoto Kobayashi collected and managed the data used in this study and supervised this paper.Takuichi Nishimura supervised the entire project and checked and proofread the English text. Shin-ichi Niwa wrote the entire paper and supervised its overall review.

DATA AVAILABILITY STATEMENT : The data used in this study will not be made public, but if you would like to use it, please contact Oshiyama.

ADDITIONAL INFORMATIONS： There are no conflicts of interest regarding this study.

（Ethical considerations）

This study complies with the "Declaration of Helsinki: Ethical Principles for Medical Research Involving Human Subjects (revised in 2013 by the World Medical Association)" and the "Ethical Guidelines for Life Sciences and Medical Research Involving Human Subjects (partially revised in 2022 by the Japanese Ministry of Education, Culture, Sports, Science and Technology, the Ministry of Health, Labor and Welfare, and the Ministry of Economy, Trade and Industry)" and has been approved and conducted by the Ethics Committee of the Medical Corporation Kozanso (approval number: 2017-3). Azumadori Clinic is an affiliate of the Medical Corporation Kozanso.

Each participating subject was given a detailed explanation of the study, including the potential burden, expected risks, measures to minimise them and data handling procedures. Only subjects who provided written informed consent were included in the study. If a subject who has consented wishes to discontinue participation after the start of the study, their wishes were respected. Participation or non-participation in this study is at the subject's own voluntary discretion, and subjects will not be disadvantaged if they do not agree to participate or discontinue participation.

Personal information was protected by linkable anonymisation. The measurement data in this study were assigned will be assigned only a code number that identifies the individual, and were anonymized and processed so that individuals cannot be identified from the results. The code number was used as a personal ID. The code number and individual name linking sheet was securely managed by a non-participating staff of Azumadori Clinic. The consent form, and response form were securely stored in a locked place at the clinic. When the research results are published, the subjects’ names are not published under any circumstances. Personal information will not be handled beyond the scope necessary to achieve the purpose of use stated in the consent form without the subject's consent. Personal information obtained in this study will be strictly managed to prevent it from leaking to the outside, and after a certain period of time (10 years after the research publication), all materials will be rendered physically unreadable and appropriately destroyed.

Petersen RC (2004) Mild cognitive impairment as a diagnostic entity. J Intern Med 256:183-194. https://doi.org/10.1111/j.1365-2796.2004.01388.x
Smirnov DS, Ashton NJ, Blennow K, Zetterberg H, Simrén J, Lantero-Rodriguez J, Karikari TK, Hiniker A, Rissman RA, Salmon DP, Galasko D (2022) Plasma biomarkers for Alzheimer's Disease in relation to neuropathology and cognitive change. Acta Neuropathol 143:487-503. https://doi.org/10.1007/s00401-022-02408-5
Ministry of Health, Labour and Welfare (Japan) HP (in Japanese)　https://www.mhlw.go.jp/stf/seisakunitsuite/bunya/0000089508_00007.html
Nobuyuki Okamura. Kazuhiko Yanai. Katsutoshi Furukawa. Hiroyuki Arai. Yukishita Kudo. Biomarkers of Neurological Disease (In Japanease). Jpn J Clin Pharmacal Ther 41(3)2010.
Petersen RC, Doody R, Kurz A, Mohs RC, Morris JC, Rabins PV, Ritchie K, Rossor M, Thal L, Winblad B (2001) Current concepts in mild cognitive impairment. Arch Neurol 58:1985-1992. https://doi.org/10.1001/archneur.58.12.1985
Fujiwara Y, Suzuki H, Yasunaga M, Sugiyama M, Ijuin M, Sakuma N, Inagaki H, Iwasa H, Ura C, Yatomi N, Ishii K, Tokumaru AM, Homma A, Nasreddine Z, Shinkai S (2010) Brief screening tool for mild cognitive impairment in older Japanese: validation of the Japanese version of the Montreal Cognitive Assessment. Geriatr Gerontol Int 10:225-232. https://doi.org/10.1111/j.1447-0594.2010.00585.x
Kobayashi N, Nakano K, Tago H, Niwa S (2010) Development of a simplified Short-Term Memory recall Test (STMT) and its clinical evaluation. Aging Clin Exp Res 22:157-163. https://doi.org/10.1007/BF03324790
Mauvais-Jarvis F, Bairey Merz N, Barnes PJ, Brinton RD, Carrero JJ, DeMeo DL, De Vries GJ, Epperson CN, Govindan R, Klein SL, Lonardo A, Maki PM, McCullough LD, Regitz-Zagrosek V, Regensteiner JG, Rubin JB, Sandberg K, Suzuki A (2020) Sex and gender: modifiers of health, disease, and medicine. Lancet 396:565-582. https://doi.org/10.1016/S0140-6736(20)31561-0
Vegeto E, Villa A, Della Torre S, Crippa V, Rusmini P, Cristofani R, Galbiati M, Maggi A, Poletti A (2020) The role of sex and sex hormones in neurodegenerative diseases. Endocr Rev 41:273-319. https://doi.org/10.1210/endrev/bnz005
Deschaintre Y, Richard F, Leys D, Pasquier F (2009) Treatment of vascular risk factors is associated with slower decline in Alzheimer disease. Neurology 73:674-680. https://doi.org/10.1212/WNL.0b013e3181b59bf3
Kivipelto M, Helkala EL, Laakso MP, Hänninen T, Hallikainen M, Alhainen K, Soininen H, Tuomilehto J, Nissinen A (2001) Midlife vascular risk factors and Alzheimer's disease in later life: longitudinal, population based study. BMJ 322:1447-1451. https://doi.org/10.1136/bmj.322.7300.1447
Kakio Kurita. Akinori Hidaka. Series on Stochastic Models in Informatics and Data Science1- Statistical Pattern Recognition and Discriminant Analysis (in Japanese)147-155(2015.1). https://www.coronasha.co.jp/np/data/tachiyomi/978-4-339-02831-7.pdf
Livingston G, Huntley J, Sommerlad A, Ames D, Ballard C, Banerjee S, Brayne C, Burns A, Cohen-Mansfield J, Cooper C, Costafreda SG, Dias A, Fox N, Gitlin LN, Howard R, Kales HC, Kivimäki M, Larson EB, Ogunniyi A, Orgeta V, Ritchie K, Rockwood K, Sampson EL, Samus Q, Schneider LS, Selbæk G, Teri L, Mukadam N (2020) Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet. 2020;396:413-446. https://doi.org/10.1016/S0140-6736(20)30367-6. Erratum in: Lancet. 2023;402:1132. https://doi.org/10.1016/S0140-6736(23)02043-3
Borshchev YY, Uspensky YP, Galagudza MM (2019) Pathogenetic pathways of cognitive dysfunction and dementia in metabolic syndrome. Life Sci 237:116932. https://doi.org/10.1016/j.lfs.2019.116932
Forette F, Boller F (1991) Hypertension and the risk of dementia in the elderly. Am J Med 90:14S-19S. https://doi.org/10.1016/0002-9343(91)90430-6
Walker KA, Power MC, Gottesman RF (2017) Defining the relationship between hypertension, cognitive decline, and dementia: a review. Curr Hypertens Rep 19:24. https://doi.org/10.1007/s11906-017-0724-3
Gilsanz P, Mayeda ER, Glymour MM, Quesenberry CP, Mungas DM, DeCarli C, Dean A, Whitmer RA (2017) Female sex, early-onset hypertension, and risk of dementia. Neurology 89:1886-1893. https://doi.org/10.1212/WNL.0000000000004602
Qiu C, von Strauss E, Winblad B, Fratiglioni L (2004) Decline in blood pressure over time and risk of dementia: a longitudinal study from the Kungsholmen project. Stroke 35:1810-1815. https://doi.org/10.1161/01.STR.0000133128.42462.ef
Klop B, Elte JW, Cabezas MC (2013) Dyslipidemia in obesity: mechanisms and potential targets. Nutrients 5:1218-1240. https://doi.org/10.3390/nu5041218
Mankovsky BN, Ziegler D (2004) Stroke in patients with diabetes mellitus. Diabetes Metab Res Rev 20:268-287. https://doi.org/10.1002/dmrr.490
Pendlebury ST, Rothwell PM (2009) Prevalence, incidence, and factors associated with pre-stroke and post-stroke dementia: a systematic review and meta-analysis. Lancet Neurol 8:1006-1018. https://doi.org/10.1016/S1474-4422(09)70236-4
Kuźma E, Hannon E, Zhou A, Lourida I, Bethel A, Levine DA, Lunnon K, Thompson-Coon J, Hyppönen E, Llewellyn DJ (2018) Which risk factors causally influence dementia? a systematic review of mendelian randomization studies. J Alzheimers Dis 64:181-193. https://doi.org/10.3233/JAD-180013
Xu W, Qiu C, Winblad B, Fratiglioni L (2007) The effect of borderline diabetes on the risk of dementia and Alzheimer's disease. Diabetes 56:211-216. https://doi.org/10.2337/db06-0879
Bree AJ, Puente EC, Daphna-Iken D, Fisher SJ (2009) Diabetes increases brain damage caused by severe hypoglycemia. Am J Physiol Endocrinol Metab 297:E194-201. https://doi.org/10.1152/ajpendo.91041.2008
Lin CH, Sheu WH (2013) Hypoglycaemic episodes and risk of dementia in diabetes mellitus: 7-year follow-up study. J Intern Med 273:102-110. https://doi.org/10.1111/joim.12000
Mehta HB, Mehta V, Goodwin JS (2017) Association of hypoglycemia with subsequent dementia in older patients with type 2 diabetes mellitus. J Gerontol A Biol Sci Med Sci 72:1110-1116. https://doi.org/10.1093/gerona/glw217
Cox SR, Lyall DM, Ritchie SJ, Bastin ME, Harris MA, Buchanan CR, Fawns-Ritchie C, Barbu MC, de Nooij L, Reus LM, Alloza C, Shen X, Neilson E, Alderson HL, Hunter S, Liewald DC, Whalley HC, McIntosh AM, Lawrie SM, Pell JP, Tucker-Drob EM, Wardlaw JM, Gale CR, Deary IJ (2019) Associations between vascular risk factors and brain MRI indices in UK Biobank. Eur Heart J 40:2290-2300. https://doi.org/10.1093/eurheartj/ehz100
Hassing LB, Hofer SM, Nilsson SE, Berg S, Pedersen NL, McClearn G, Johansson B (2004) Comorbid type 2 diabetes mellitus and hypertension exacerbates cognitive decline: evidence from a longitudinal study. Age Ageing 33:355-361. https://doi.org/10.1093/ageing/afh100
Nagai M, Hoshide S, Kario K (2010) Hypertension and dementia. Am J Hypertens 23:116-124. https://doi.org/10.1038/ajh.2009.212
van den Berg E, Kloppenborg RP, Kessels RP, Kappelle LJ, Biessels GJ (2009) Type 2 diabetes mellitus, hypertension, dyslipidemia and obesity: a systematic comparison of their impact on cognition. Biochim Biophys Acta 1792:470-481. https://doi.org/10.1016/j.bbadis.2008.09.004
Kørner EA, Lauritzen L, Nilsson FM, Lolk A, Christensen P (2012) Simple scoring of the Clock-Drawing test for dementia screening. Dan Med J 59:A4365.
Tsuboi K, Harada T, Ishii T, Morishita H, Ohtani H, Ishizaki F (2009) Evaluation of the usefulness of a simple touch-panel method for the screening of dementia. Hiroshima J Med Sci 58:49-53.

Table 1. Patients’ demographic data

The breakdown of the patients(Second time)n=162
classification		40-49	50-59	60-69	70-79	80-89	90-99	total

	all	1	3	6	46	89	17	162
	male	0	2	3	12	24	6	47
	female	1	1	3	34	65	11	115

dementia	all	0	1	3	23	63	14	72
non-dementia		1	2	3	23	26	3	90
not determined		0	0	0	0	0	0	0

dementia	male	ー	0	2	5	17	4	21
non-dementia		ー	2	1	7	7	2	26
not determined		ー	0	0	0	0	0	0

dementia	female	0	1	1	18	46	10	51
non-dementia		1	0	2	16	19	1	64
not determined		0	0	0	0	0	0	0

hypertension	all	0	0	1	18	35	5	60
non-hypertension		1	1	0	4	23	8	36
not determined		0	2	5	24	31	4	66

hypertension	male	ー	0	0	5	10	1	15
non-hypertension		ー	0	0	0	9	4	14
not determined		ー	2	3	7	5	1	18

hypertension	female	0	0	1	13	25	4	45
non-hypertension		1	1	0	4	14	4	22
not determined		0	0	2	17	26	3	48

hyperlipidemia	all	0	0	0	7	15	2	30
non-hyperlipidemia		1	1	1	15	43	11	66
not determined		0	2	5	24	31	4	66

hyperlipidemia	male	ー	0	0	1	2	0	7
non-hyperlipidemia		ー	0	0	4	17	5	22
not determined		ー	2	3	7	5	1	18

hyperlipidemia	female	0	0	0	6	13	2	23
non-hyperlipidemia		1	1	1	11	26	6	44
not determined		0	0	2	17	26	3	48

diabetes	all	0	0	0	2	12	1	13
non-diabetes		1	1	1	20	46	12	83
not determined		0	2	5	24	31	4	66

diabetes	male	ー	0	0	2	3	0	4
non-diabetes		ー	0	0	3	16	5	25
not detamined		ー	2	3	7	5	1	18

diabetes	female	0	0	0	0	9	1	9
non-diabetes		1	1	1	17	30	7	58
not determined		0	0	2	17	26	3	48

Table 2: Descriptive statistics of 96 patients undergoing treatment for hypertension, hyperlipidemia, and diabetes.

dependent variable			mean	SD	prevalence
Second diagnosis is dementia or not	all (n=96)	age	79.0	8.6	-
		Years to second diagnosis	2.2	1.5	-
		STMT	3.2	2.0	-
		hypertension			60%
		hyperlipidemia			31%
		diabetes			14%
	male (n=25)	age	79.2	10.9	-
		Years to second diagnosis	1.9	1.3	-
		STMT	2.7	1.8	-
		hypertension	-	-	68%
		hyperlipidemia	-	-	40%
		diabetes	-	-	20%
	female (n=71)	age	78.9	7.7	-
		Years to second diagnosis	2.2	1.6	-
		STMT	3.4	2.0	-
		hypertension	-	-	58%
		hyperlipidemia	-	-	28%
		diabetes	-	-	11%

STMT, Short-Term Memory Recall Test.

Table 3. Results of prediction of dementia onset using the STMT score, age, sex, and elapsed years in the cohort of 162 patients and using the previous factors combined with the presence or absence of treatment for hypertension, hyperlipidemia, and diabetes in the cohort of 96 patients.

dependent variable	independent variable	sensitivity	specificity	accuracy rate	AUC	OMQ
the second time dementia or not (n=162)	STMT1	67.30	63.80	66.00	0.75	0.67
	STMT1, AGE	74.04	74.14	74.07	0.76	0.68
	STMT1, SEX	60.58	74.14	65.43	0.79	0.72
	STMT1, ELAPSED YEARS	64.42	68.97	66.05	0.76	0.68
	STMT1, AGE, SEX	71.15	75.86	72.84	0.79	0.72
	STMT1, AGE, ELAPSED YEARS	73.08	77.59	74.69	0.77	0.69
	STMT1, SEX, ELAPSED YEARS	68.27	67.24	67.90	0.79	0.72
	STMT1, AGE, SEX, ELAPSED YEARS	72.10	74.10	72.80	0.80	0.72
the second time dementia or not (n=96)	STMT, AGE, SEX, ELAPSED YEARS HYPERTENSION	72.10	71.40	71.90	0.77	0.67
	STMT, AGE, SEX, ELAPSED YEARS, HYPERLIPIDAEMIA	70.50	62.80	67.70	0.76	0.67
	STMT, AGE, SEX, ELAPSED YEARS, DIABETES	72.10	65.70	69.80	0.78	0.68
	STMT, AGE, SEX, ELAPSED YEARS, HYPERTENSION, HYPERLIPIDAEMIA	68.90	68.60	68.80	0.77	0.67
	STMT, AGE, SEX, ELAPSED YEARS, HYPERTENSION, DIABETES	70.50	62.90	67.70	0.79	0.69
	STMT, AGE, SEX, ELAPSED YEARS, HYPERLIPIDAEMIA, DIABETES	70.50	62.90	67.80	0.79	0.70
	STMT, AGE, SEX, ELAPSED YEARS, HYPERYENSION, HYPERLIPIDAEMIA, DIABETES	70.50	68.60	69.80	0.79	0.70

The sensitivity, specificity, area under the receiver operating characteristic curve (AUC), and overall model quality (OMQ) values are shown.

Table 4. Analysis of sensitivity and specificity.

		Actual Results	（prediction） not dementia occurred at the second assessment	（prediction） dementia occurred at the second assessment	Total
n=162	frequency (N)	no dementia occurred at the second assessment	43	15	58
	frequency (N)	dementia occurred at the second assessment	21	83	104
	%	no dementia occurred at the second assessment	74.1	25.9	100
	%	dementia occurred at the second assessment	20.2	79.8	100
n=96	frequency (N)	no dementia occurred at the second assessment	19	12	31
	frequency (N)	dementia occurred at the second assessment	12	53	65
	%	no dementia occurred at the second assessment	61.3	38.7	100
	%	dementia occurred at the second assessment	18.5	81.5	100

Following discriminant analysis, the classification results are based on the prediction of whether dementia occurred at the second examination and comparing them with the actual presence or absence of dementia at that time

No competing interests reported.

Prediction of Dementia Onset using the Short-Term Memory Recall Test

Status:

Version 1

Abstract

Figures

INTRODUCTION

RESULTS

DISCUSSION

LIMITATIONS

CONCLUSIONS

METHODS

Declarations

References

Tables

Additional Declarations

Status:

Version 1