Clinical Conversion or Reversion of Mild Cognitive Impairment in Community versus Hospital Based Studies: GDEMCIS (Gwangju Dementia and Mild Cognitive Impairment Study) and CREDOS (Clinical Research Center for Dementia of South Korea)

Abstract

Background: In keeping with increasing interest in dementia, few recent studies suggest that clinical course of mild cognitive impairment vary across different studies with hospital-based subjects showing higher rates of conversion than community-based subjects.

Objective: The main objective of the present study was to assess whether the clinical conversion or reversion rates differ according to recruitment source.

Methods: The baseline study subjects comprised of patients who were diagnosed with mild cognitive impairment in community-based GDEMCIS or hospital-based CREDOS. The two studies had nearly the same protocol and were performed over a similar period. We used propensity score matching for baseline comparability. After that, Cox proportional hazards regression analyses were conducted to estimate the hazard ratios and 95% confidence intervals of clinical conversion or reversion.

Results: Based on 89 GDEMCIS subjects, 1 : 4 propensity score matching was conducted and 356 CREDOS subjects were selected. After adjusting for covariates including baseline demographics, comorbidity, depression, disability, and neuropsychological result, Cox proportional hazard regression analysis for time to clinical conversion indicated that recruitment from hospital-based CREDOS exhibited hazard ratio of 2.13 (95% CI, 1.08–4.21), as compared to recruitment from community-based GDEMCIS. Similarly, Cox proportional hazard regression analysis for time to reversion indicated that recruitment from hospital-based CREDOS exhibited hazard ratio of 0.34 (95% CI, 0.20–0.59), as compared to recruitment from community-based GDEMCIS.

Conclusion: The present study demonstrated that even after the matching process and adjustments for baseline covariates, recruitment source greatly affected the course of mild cognitive impairment.

Keywords

Clinical conversion community dementia hospital mild cognitive impairment reversion recruitment source

INTRODUCTION

Mild cognitive impairment (MCI) is defined as a slight but noticeable and measurable decline in one or more cognitive domains [1]. It is an important public health problem attracting worldwide attention, and many researchers are in the process of determining its clinical implication. Long-term studies suggest that 10 to 20% of those aged ≥65 years may have MCI, and annual conversion rate to dementia ranges from 10 to 15% in hospital samples and 3% to 6% in community samples [2 –5]. Initially, MCI was considered to be a transitional stage between normal cognition and dementia [6]. However, follow-up research showed that MCI progresses in a variety of patterns, and remained relatively stable over time in community samples [5 , 8]. In addition, a recent meta-analysis on reversion from MCI to normal cognition showed that the overall reversion rate of MCI patients was as high as 24% with substantial differences between community-based studies and hospital-based studies (14% versus 31%) [9]. With the wide acceptance of the concept of MCI since the turn of the 21st century, many researchers have made active efforts to identify the factors of MCI that modify the clinical course of MCI. Various factors were proposed including neuroimaging factors and biomarkers [10 –14].

Previous research findings indicate differences in clinical conversion rate to dementia or reversion rate to normal cognition according to recruitment sources; however, a direct, well-controlled and comparative approach to such a phenomenon is lacking. A clear conclusion on the association between recruitment source and course of MCI is required because it may lead to the identification of hidden risk factors, hidden protective factors, or epidemiological biases. However, there is insufficient research due to difficulty in comparability. To the best of our knowledge, only one research study has attempted a direct comparative approach [3]. Farias et al. reported that recruitment source was highly associated with the risk of clinical conversion to dementia (hazard ratio = 3.50, 95% confidence interval, 1.31–9.18, p = 0.01). Furthermore, the strong points of their study were the use of same protocol at the same research center and the inclusion of variables such as brain imaging and neuropsychological testing; whereas, the limitations were the relatively small number of subjects and differences in ethnic distribution between individual groups. In addition, there was no correction for depression, disability, and comorbidity that are associated factors with clinical conversion to dementia [10 , 16].

In the present study, we made an attempt to determine the association of recruitment source and risk of clinical conversion to dementia in MCI patients. In addition, the association of recruitment source and rate of reversion to normal cognition was evaluated in MCI patients. Finally, in both hospital and community-based studies, we compared the differences of baseline characteristics between participants with clinical conversion and participants with reversion to normal cognition. To assess these points, we analyzed hospital and community-based longitudinal studies that used nearly the same protocols on ethnically homogeneous Korean elderly. We also utilized a different statistical approach to determine the independent effect of recruitment source on clinical course of MCI, by adjustment for covariates, including depression, disability, and comorbidity, which were not considered in preceding studies.

MATERIALS AND METHODS

Study participants from GDEMCIS and CREDOS

The study population comprised community-based GDEMCIS (Gwangju Dementia and Mild Cognitive Impairment Study) and hospital-based CREDOS (Clinical Research Center for Dementia of South Korea) participants who were diagnosed with MCI at baseline assessment and were followed at least once after the first visit.

The community-based GDEMCIS recruited participants from the community mental health service center who were diagnosed with normal cognition, MCI, or dementia by a neurologist or psychiatrist. We recruited elderly subjects who resided within a well-defined geographic region in a small city with agricultural districts, Gwangju. All participants underwent comprehensive interviews, neurological examinations, and neuropsychological assessment as in CREDOS. Detailed descriptions of GDEMCIS have been reported previously [17]. Of GDEMCIS participants from March 2005 to August 2010, we retrospectively analyzed data from 89 patients who were diagnosed with MCI at baseline assessment and were followed up at least once after the first visit.

The hospital-based CREDOS recruited participants from university-affiliated hospitals who were diagnosed with normal cognition, MCI, or dementia by a neurologist or psychiatrist. We evaluated all participants with comprehensive interviews, neurological examinations, and neuropsychological assessment, as described in the previous study [18]. In brief, main caregivers were interviewed in depth by the clinician and neuropsychologist. Of CREDOS participants from March 2005 to April 2013, we retrospectively analyzed the data from 1,597 patients who were diagnosed with MCI at baseline assessment and who were followed up at least once after the first visit.

For both GDEMCIS and CREDOS, the same neurologist or psychiatrist continuously assessed and rated the patient at the follow up visit based on comprehensive interviews with patient and caregiver, neurological examinations, and neuropsychological assessment. Finally, we excluded those who met the following criteria: (1) history of significant hearing or visual impairment rendering participation in the interview difficult; (2) history of following neurologic disorder (brain tumor, subarachnoid hemorrhage, epilepsy, encephalitis, and metabolic encephalopathy) or other neurologic conditions that could interfere with the study; (3) history of psychiatric disorder including mental retardation, schizophrenia, bipolar disorder, or other psychiatric conditions that could interfere with the study; (4) history of psychoactive substances other than alcohol; (5) history of physical illnesses or disorders including cancer, renal failure, hepatic failure, severe asthma or chronic obstructive pulmonary disease, or other physical conditions that could interfere with the study.

Standard protocol approvals and patient consents

GDEMCIS was approved by the Institutional Review Board of the Severance Mental Health Hospital. CREDOS was approved by the Institutional Review Board of all participating centers. All participants of CREDOS and GDEMCIS provided informed written consent.

MCI and dementia diagnostic criteria

For both GDEMCIS and CREDOS, we used the operational criteria for the diagnosis of MCI by Petersen: (1) complaints of subjective memory or other cognitive impairment by patients or informants, (2) evidence of objective memory or other cognitive function impairment in neuropsychological tests, (3) largely maintained activities of daily living (ADL) and minimally impaired instrumental activity of daily living (IADL), and (4) absence of dementia in the individual [2]. We used the diagnostic criteria for probable Alzheimer’s disease issued by the National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer’s Disease and Related Disorder Association (NINCDS-ADRDA), and the diagnostic criteria for other types of dementia in the Fourth Edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) [19, 20].

Other covariates

Covariates assessed in the study were age, gender, education, diabetes, hypertension, depression, disability, and impairment of each cognitive domain by Seoul Neuropsychological Screening Battery-dementia version (SNSB-D). Physician-diagnosed hypertension or diabetes was regarded as the presence of disease. Depression was measured using the Korean version of the 15-item Geriatric Depression Scale (SGDS-K) [21]. The optimal cut-off point for screening major depressive disorder is reported as SGDS-K score of ≥8 points, with sensitivity and specificity of 93.6% and 76.0%, respectively [22]. Disability was measured using the Seoul-Instrumental Activity of Daily Living (S-IADL). The optimal cut-off point for screening disability is a S-IADL score of ≥8 points with the sensitivity and specificity of the findings of 83.3% and 93.1%, respectively [23]. Cognitive function was assessed by the presence of impairment of each cognitive domain by SNSB-D. SNSB-D is standardized neuropsychological battery and includes tests of language, verbal/nonverbal memory, attention, and frontal lobe function domains [24]. We assessed each cognitive domain using the following tests: the Korean short version of the Boston Naming Test for language, the Rey Complex Figure Test-delayed recall for visuospatial memory, the Seoul Verbal Learning Test-delayed recall for verbal memory, the Digit Span Test-backward for attention and Controlled Oral Word Association Test (COWAT) for frontal/executive function. In this study, scores below 1 standard deviation (16th percentile) of the age, gender, and education specific norm were classified as impaired based on SNSB-D reference standards [24].

Statistical analysis

Since the aim of the study was to determine independent effect of recruitment source as a factor of clinical course of MCI regardless of known confounding factors, and the participants of GDEMCIS and CREDOS were not randomly assigned, we applied the propensity score matching process. The rationale and methods underlying the use of a propensity score for causal exposure variables are described previously [25]. The propensity for GDEMCIS participants was determined without regard to clinical conversion to dementia, using multivariable logistic regression analysis. A parsimonious model was developed with 8 covariates including age, gender, education, depression, disability, language, verbal memory, and frontal/executive function. This model exhibited a c-statistic of 0.76, indicating good ability to differentiate between GDEMCIS and CREDOS participants. A propensity score for GDEMCIS participants was calculated from the logistic equation for each participant. After that, we used the propensity scores to match participants of GDEMCIS and CREDOS by computerized process. We were able to match 89 GDEMCIS participants to 1 : 4 matched 356 CREDOS participants.

After the propensity score matching process, categorical variables were compared using Chi-square tests. Continuous variables were compared using independent Student’s t-tests between the GDEMCIS and CREDOS participants. We used Kaplan-Meier survival curves to plot the survival curve for clinical conversion to dementia by the recruitment source. Cox proportional hazards regression was used to estimate the hazard ratios (HR) and 95% confidence intervals (CI) of clinical conversion to dementia or reversion to normal cognition. Harrell’s c index was used for assessing prediction performance in Cox proportional hazard regression models. We also evaluated time-dependent integrated area under the curve (iAUC) plot for assessing time-dependent stability of the Cox proportional hazard regression model.

We assessed the relationship between recruitment source and clinical course of MCI within five different models to investigate whether other covariates could explain the association. Model 1 made adjustments for age, sex, and education. Model 2 adjusted for diabetes, hypertension, depression, disability, and all factors adjusted in Model 1. Model 3-1 added cognition (by MMSE score) and all factors in Model 2. Model 3-2 adjusted for cognition (by impairment of cognitive domains in neuropsychological test) and all factors in Model 2.

Finally, in both GDEMCIS and CREDOS, we compared the differences of baseline characteristics between participants with clinical conversion and participants with reversion to normal cognition. Categorical variables were compared using Chi-square tests or Fisher’s exact test. Continuous variables were compared using independent Student’s t-tests or Mann-Whitney U test.

Statistical analyses were performed using Statistical Analysis System (SAS) (Version 9.2, SAS Institute Inc., Cary, NC) and R software packages (Version 3.1.2, R Foundation for Statistical Computing, Vienna, Austria). Statistical significance was considered when p value was < 0.05.

RESULTS

Differences of baseline characteristics between the two groups after propensity score matching were summarized in Table 1. Between these two groups, there were no statistically significant differences in age, gender, diabetes, hypertension, SGDS-K, S-IADL, and SNSB-D results. Hospital-based CREDOS subjects were slightly more educated with a slightly higher mini-mental status exam score. In addition, CREDOS subject exhibited slightly shorter follow up duration with slightly more frequent follow up than GDEMCIS subjects (Table 1). Over the follow up, participants who were recruited from hospital-based CREDOS were more likely to convert to dementia (n = 83, 23.3%), as compared to those recruited from community-based GDEMCIS (n = 9, 10.1%) (Fig. 1). In addition, participants who were recruited from hospital-based CREDOS were less likely to revert to normal cognition (n = 39, 11.0%), as compared to those recruited from community-based GDEMCIS (n = 27, 30.3%).

About clinical conversion to dementia, unadjusted Cox proportional hazard regression analysis for time to clinical conversion indicated that recruitment from hospital-based CREDOS exhibited hazard ratio of 2.50 (95% CI, 1.30–4.83), as compared to recruitment from community-based GDEMCIS. Table 2 showed the HR of clinical conversion for participants who were recruited from CREDOS after adjustment for different potential confounders. In adjusted Cox proportional hazard regression analysis, the direction and strength of the association between recruitment source and clinical conversion remained even after adjustment for potential confounders (adjusted HR for clinical conversion = 2.13, 95% CI: 1.08–4.21) (Table 2).

In unadjusted and adjusted Cox proportional hazard regression analysis for time to reversion to normal cognition, recruitment from hospital-based CREDOS exhibited hazard ratio of 0.39 and 0.34, respectively, as compared to recruitment from community-based GDEMCIS (unadjusted HR for reversion = 0.39, 95% CI: 0.24–0.64, adjusted HR for reversion = 0.34, 95% CI: 0.20–0.59). Thus, participants of hospital-based CREDOS were less likely to revert to normal cognition (Supplementary Table 1).

Harrell’s c index was used for assessing prediction performance of Cox proportional hazard regression models and time-dependent integrated area under the curve (iAUC) plot for time-dependent stability of Cox proportional hazard regression model. Thus, the Cox proportional hazard regression models had relatively fare prediction performance and time-dependent stability (Table 2 and Fig. 2).

Finally, we assessed the differences of baseline characteristics between participants with clinical conversion and participants with reversion. In community-based GDEMCIS, participants with clinical conversion were more disabled in activity of daily living at baseline, as compared to participants with reversion. They also exhibited higher impairment rate in language, verbal memory, and frontal lobe function at baseline neuropsychological test. In hospital-based CREDOS, participants with clinical conversion were more aged with lower MMSE score at baseline, as compared to participants with reversion. They also exhibited higher impairment rate in language, visuospatial memory and frontal lobe function at baseline neuropsychological test result (Table 3).

DISCUSSION

The primary purpose of this study was to investigate the association between recruitment source and the risk of clinical conversion to dementia in MCI patients. Additionally, we assessed the association between recruitment source and rate of reversion to normal cognition in MCI patients. Our study utilized the propensity score matching process to reduce the bias from confounding variables at baseline.

After matching and additional adjusting for potential confounders, there was still a significantly higher risk of clinical conversion to dementia in MCI patients recruited from hospital, as compared to MCI patients recruited from community. Several interpretations are possible for the association of recruitment source with risk of clinical conversion to dementia in MCI patients. First, the stage of MCI in the hospital-based sample might be more advanced than in the community-based sample. This possibility had to be considered above all because the definition of the criteria for MCI diagnosis is somewhat broad. However, the result of this research was robust even after sufficient matching of cognitive functioning according to well-validated neuropsychological test on five different cognitive domains and disability according to caregiver’s reports. Moreover, a validated and brief measure for global cognitive functioning, i.e., Korean Mini-Mental State Examination (K-MMSE) score, was higher with the hospital-based CREDOS group than with the community-based GDEMCIS group (22.9 versus 24.8, respectively). Therefore, there was a low possibility that the significant effect of recruitment source was due to more advanced stage of MCI in the hospital-based CREDOS group. Second, there was a possibility that known risk factors, such as education, depression, and comorbidity, were confounding factors [13]. However, depression and comorbidity were taken into consideration in our study. In addition, although there was little difference in educational levels between the two groups even after matching, the hospital-based CREDOS group, which had increased risk of clinical conversion to dementia, had a higher level of education. Therefore, recruitment source was a relatively independent risk factor that was not affected by confounding factors such as education, depression, and comorbidity. Third, cognitive reserve hypothesis could be a possible explanation for higher clinical conversion rate of hospital-based MCI patients. Considering the cognitive reserve hypothesis, higher education delays the onset of cognitive decline but once it begins it is more rapid in persons with more education [26]. Actually, both our study and previous study of Farias et al. exhibited that hospital-based MCI patients had received more formal education than community-based MCI patients. To make an approach to such a concern, the regression analysis in our study considered an adjustment of education and confirmed non-significant effect of education on clinical conversion. But, this can hardly be considered sufficient for assessing effect of education. Future research requires another statistical approach, such as the structural equation model, that enables a more definite assessment of independent effect of education or other variables. Fourth, it was possible that bias occurred during the processes of assessment, evaluation, and diagnosis [27]. In fact, preceding researchers reported a tendency that cognitively normal conditions in community-based samples are misdiagnosed as MCI, especially in ethnic minorities [28]. However, the present research was conducted on ethnically homogeneous subjects. In addition, the results of neuropsychological test, which could be considered the most important means of assessment, were obtained by objective structured examination. The assessment of depression and disability was relatively objective, because it used SGDS-K, which is a depressive symptom scale involving yes or no responses, as well as S-IADL, which is a structured informant-based questionnaire. Furthermore, bias was minimized in this research, because CREDOS and GDEMCIS began at similar times and the clinician and psychologist received nearly the same education program for both studies. Finally, there was a strong possibility that an uncovered or hidden confounding factor made recruitment source seem like a significant risk factor. This interpretation was most favored. In fact, it is unlikely that recruitment source has a direct biological effect on risk of clinical conversion to dementia. Hence, a process was necessary to confirm which recruitment source-related factors caused the evident risk difference, i.e., to identify the absence of specific factors that eliminates the risk difference. We therefore made efforts to eliminate the effect of many variables that were potentially important confounding factors of dementia conversion. However, the efforts failed to cancel the impact of recruitment source. Accordingly, further investigation of other factors that are potentially related to recruitment source is required. These potential recruitment source-related factors include socioeconomic status, family environmental factor, more detailed comorbidity, subjective health complaints, and perceived health.

Our study also exhibited that there was higher rate of reversion to normal cognition in MCI patients recruited from community, as compared to MCI patients recruited from hospital. For years, previous research has suggested relatively high reversion rate of community-based MCI [8 , 29–31]. However, there was no direct or propensity-matched comparative analysis between participants from community and hospital. Although specific mechanism has not been fully identified yet, higher reversion rate and progress in a various patterns of community-based MCI might be an important cause of different courses between community and hospital-based MCI patients.

Additionally, our study presented baseline differences between participants with clinical conversion and participants with reversion in order to determine predicting factors for clinical conversion or reversion. In both community-based GDEMCIS and hospital-based CREDOS, baseline impairment rate of language and frontal lobe function domain was higher in clinical converter than that of reverter. In terms of memory domain, clinical converter of community-based GDEMCIS showed higher impairment rate of verbal memory than that of reverter, while clinical converter of hospital-based CREDOS showed higher impairment rate of visuospatial memory than that of reverter. Overall, neuropsychological test result at baseline provided useful information about the course of MCI, which is of potential use to the clinician. Another interesting result was that baseline disability was more common only in clinical converter of community-based GDEMCIS than that of reverter. In other words, there was no significant difference of disability between clinical converter and reverter of hospital-based CREDOS. Some recent community-based studies indicated the importance of disability or physical frailty as a contributing factor to the course of MCI [32 –34]. Therefore, we cautiously suggest that baseline disability is valuable information for clinicians or public health service officers to estimate future clinical conversion or reversion of MCI in the community. Similarly, we exhibited baseline differences of age and MMSE score between clinical converter and reverter of hospital-based CREDOS that could be useful information to hospital clinicians.

Our study had several strengths, including a longitudinal design with a long follow-up period up to 74.3 months and an ethnically homogeneous sample of patients with MCI. Also, diagnostic classification of our study was carried out directly by a neurologist or a psychiatrist who had abundant clinical experiences. Finally, propensity score matching and additional adjustments were also made to identify the independent effect of recruitment source. This statistical approach enables an even more rigorous adjustment for selection bias and confounding factors than would be possible with standard multivariable analysis [25, 35].

Despite its strengths, our study also had some limitations. First, unidentified selection bias was possible, including factors other than cognitive function, education, disability, and depression, which were the focus of this research. Such factors include socioeconomic status, lifestyle factors, subjective health complaints, and medical care seeking behaviors that could be associated with different enrollment process and environment between two studies. Second, the study is representative and can be generalized to a limited extent, because the number of subjects was insufficient and the subjects of GDEMCIS were recruited from only a single region. Generally, as the community-based subjects with MCI are likely to be more heterogeneous than the hospital-based ones, sufficient number of subject is important to obtain representability and generalizability [36]. Future large-scale study is greatly needed to attain more clear and definite conclusion.

In summary, we demonstrated a significantly higher rate of clinical conversion to dementia in the MCI patients recruited from hospital, as compared to MCI patients recruited from community. Our study results also indicated a higher rate of reversion to normal cognition in MCI patients recruited from the community, as compared to MCI patients recruited from the hospital. In other words, recruitment source was a highly significant factor, despite accounting for primary factors such as baseline cognitive function and demographic factors including education, disability and depression. The important implication of this research is that an uncovered or hidden risk factor is responsible for the big difference in MCI progress between the MCI patients recruited from hospital vs. community. The findings of the current study will facilitate future MCI research on the causes of risk difference between hospital- and community- based patient populations.

Footnotes

ACKNOWLEDGMENTS

This study was supported by a grant from the Korea Health 21 R&D Project, Ministry of Health & Welfare, Republic of Korea (HI10C2020) and National Research Foundation of Korea (NRF) Grant funded by the Korean Government (MSIP) (No.2015R1A5A7037630).

Authors’ disclosures available online ().

Appendix

References

Petersen

, Caracciolo

, Brayne

, Gauthier

, Jelic

, Fratiglioni

(2014) Mild cognitive impairment: A concept in evolution. J Intern Med 275, 214–228.

Petersen

, Smith

, Waring

, Ivnik

, Tangalos

, Kokmen

(1999) Mild cognitive impairment: Clinical characterization and outcome. Arch Neurol 56, 303–308.

Farias

, Mungas

, Reed

, Harvey

, DeCarli

(2009) Progression of mild cognitive impairment to dementia in clinic- vs community-based cohorts. Arch Neurol 66, 1151–1157.

Grober

, Lipton

, Hall

, Crystal

(2000) Memory impairment on free and cued selective reminding predicts dementia. Neurology 54, 827–832.

Brodaty

, Heffernan

, Kochan

, Draper

, Trollor

, Reppermund

, Slavin

, Sachdev

(2013) Mild cognitive impairment in a community sample: The Sydney Memory and Ageing Study. Alzheimers Dement 9, 310–317e311.

Morris

, Storandt

, Miller

, McKeel

, Price

, Rubin

, Berg

(2001) Mild cognitive impairment represents early-stage Alzheimer disease. Arch Neurol 58, 397–405.

Panza

, Capurso

, D’Introno

, Colacicco

, Capurso

, Solfrizzi

(2007) Heterogeneity of mild cognitive impairment and other predementia syndromes in progression to dementia. Neurobiol Aging 28, 1631–1632;discussion 1633-1634.

Ganguli

, Snitz

, Saxton

, Chang

, Lee

, Vander Bilt

, Hughes

, Loewenstein

, Unverzagt

, Petersen

(2011) Outcomes of mild cognitive impairment by definition: A population study. Arch Neurol 68, 761–767.

Malek-Ahmadi

(2016) Reversion from mild cognitive impairment to normal cognition: A meta-analysis. Alzheimer Dis Assoc Disord. doi: 10.1097/WAD.0000000000000145

10.

Cooper

, Sommerlad

, Lyketsos

, Livingston

(2015) Modifiable predictors of dementia in mild cognitive impairment: A systematic review and meta-analysis. Am J Psychiatry 172, 323–334.

11.

Trzepacz

, Yu

, Sun

, Schuh

, Case

, Witte

, Hochstetler

, Hake

, Alzheimer’s Disease Neuroimaging Initiative (2014) Comparison of neuroimaging modalities for the prediction of conversion from mild cognitive impairment to Alzheimer’s dementia. Neurobiol Aging 35, 143–151.

12.

Westman

, Muehlboeck

, Simmons

(2012) Combining MRI and CSF measures for classification of Alzheimer’s disease and prediction of mild cognitive impairment conversion. Neuroimage 62, 229–238.

13.

Lee

, Ritchie

, Yaffe

, Stijacic Cenzer

, Barnes

(2014) A clinical index to predict progression from mild cognitive impairment to dementia due to Alzheimer’s disease. PLoS One 9, e113535.

14.

, Toledo

, Zee

, Wolk

, Xie

, Ou

, Shacklett

, Parmpi

, Shaw

, Trojanowski

, Davatzikos

, Alzheimer’s Neuroimaging Initiative (2014) Integration and relative value of biomarkers for prediction of MCI to AD progression: Spatial patterns of brain atrophy, cognitive scores, APOE genotype and CSF biomarkers. Neuroimage Clin 4, 164–173.

15.

Van der Mussele

, Fransen

, Struyfs

, Luyckx

, Marien

, Saerens

, Somers

, Goeman

, De Deyn

, Engelborghs

(2014) Depression in mild cognitive impairment is associated with progression to Alzheimer’s disease: A longitudinal study. J Alzheimers Dis 42, 1239–1250.

16.

Peres

, Chrysostome

, Fabrigoule

, Orgogozo

, Dartigues

, Barberger-Gateau

(2006) Restriction in complex activities of daily living in MCI: Impact on outcome. Neurology 67, 461–466.

17.

Son

, Lee

, Chung

, Chang

, Roh

, Kim

, Jin

, Back

, Kim

, Lee

, Choi

, Noh

, Lim

, Chung

, Hong

, Oh

(2015) Increased plasma levels of heat shock protein 70 associated with subsequent clinical conversion to mild cognitive impairment in cognitively healthy elderly. PLoS One 10, e0119180.

18.

Hong

, Yoon

, Shim

, Cho

, Shin

, Kim

, Yang

(2011) APOE epsilon4 allele status in korean dementia patients with severe white matter hyperintensities. J Alzheimers Dis 24, 519–524.

19.

American Psychiatry Association (1994) Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, American Psychiatric Association, Washington, DC.

20.

McKhann

, Drachman

, Folstein

, Katzman

, Price

, Stadlan

(1984) Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology 34, 939–944.

21.

Bae

, Cho

(2004) Development of the Korean version of the Geriatric Depression Scale and its short form among elderly psychiatric patients. J Psychosom Res 57, 297–305.

22.

Lee

, Kim

, Chang

, Kim

, Lee

, Bae

, Cho

(2013) The use of the Korean Version of Short Form Geriatric Depression Scale (SGDS-K) in the community dwelling elderly in Korea. J Korean Geriatr Psychiatry 17, 37–43.

23.

, Kim

, Kwon

, Kim

, Lee

, Ko

, Jo

, Kim

(2004) A study on the reliability and validity of Seoul-Instrumental Activities of Daily Living (S-IADL). J Korean Neuropsychiatr Assoc 43, 189–199.

24.

Kang

, Na

(2003) Seoul Neuropsychological Screening Battery (SNSB), Human Brain Research & Consulting Co, Seoul.

25.

Joffe

, Rosenbaum

(1999) Invited commentary: Propensity scores. Am J Epidemiol 150, 327–333.

26.

Hall

, Derby

, LeValley

, Katz

, Verghese

, Lipton

(2007) Education delays accelerated decline on a memory test in persons who develop dementia. Neurology 69, 1657–1664.

27.

Hoyt

(2000) Rater bias in psychological research: When is it a problem and what can we do about it? Psychol Methods 5, 64–86.

28.

Manly

, Jacobs

, Sano

, Bell

, Merchant

, Small

, Stern

(1998) Cognitive test performance among nondemented elderly African Americans and whites. Neurology 50, 1238–1245.

29.

Palmer

, Backman

, Winblad

, Fratiglioni

(2008) Mild cognitive impairment in the general population: Occurrence and progression to Alzheimer disease. Am J Geriatr Psychiatry 16, 603–611.

30.

Mitchell

, Shiri-Feshki

(2009) Rate of progression of mild cognitive impairment to dementia–meta-analysis of 41 robust inception cohort studies. Acta Psychiatr Scand 119, 252–265.

31.

Gao

, Unverzagt

, Hall

, Lane

, Murrell

, Hake

, Smith-Gamble

, Hendrie

(2014) Mild cognitive impairment, incidence, progression, and reversion: Findings from a community-based cohort of elderly African Americans. Am J Geriatr Psychiatry 22, 670–681.

32.

Jekel

, Damian

, Wattmo

, Hausner

, Bullock

, Connelly

, Dubois

, Eriksdotter

, Ewers

, Graessel

, Kramberger

, Law

, Mecocci

, Molinuevo

, Nygard

, Olde-Rikkert

, Orgogozo

, Pasquier

, Peres

, Salmon

, Sikkes

, Sobow

, Spiegel

, Tsolaki

, Winblad

, Frolich

(2015) Mild cognitive impairment and deficits in instrumental activities of daily living: A systematic review. Alzheimers Res Ther 7, 17.

33.

Boyle

, Buchman

, Wilson

, Leurgans

, Bennett

(2010) Physical frailty is associated with incident mild cognitive impairment in community-based older persons. J Am Geriatr Soc 58, 248–255.

34.

Ouchi

, Akanuma

, Meguro

, Kasai

, Ishii

, Meguro

(2012) Impaired instrumental activities of daily living affect conversion from mild cognitive impairment to dementia: The Osaki-Tajiri Project. Psychogeriatrics 12, 34–42.

35.

Gum

, Thamilarasan

, Watanabe

, Blackstone

, Lauer

(2001) Aspirin use and all-cause mortality among patients being evaluated for known or suspected coronary artery disease: A propensity analysis. JAMA 286, 1187–1194.

36.

Ganguli

, Fu

, Snitz

, Hughes

, Chang

(2013) Mild cognitive impairment: Incidence and vascular risk factors in a population-based cohort. Neurology 80, 2112–2120.