Cognitive-Communicative Predictors of Mild Cognitive Impairment: A Systematic Review and Meta-Analysis

Article information

Audiol Speech Res. 2020;16(4):276-294
Publication date (electronic) : 2020 October 31
doi : https://doi.org/10.21848/asr.200056
1Department of Audiology & Speech-Language Pathology, Hallym University of Graduate Studies, Seoul, Korea
2HUGS Center for Hearing and Speech Research, Hallym University of Graduate Studies, Seoul, Korea
Correspondence: Mi Sook Lee, PhD Department of Audiology & SpeechLanguage Pathology, Hallym University of Graduate Studies, 427 Yeoksam-ro, Gangnam-gu, Seoul 06197, Korea Tel: +82-70-8680-6961 Fax: +82-2-3453-6618 E-mail: mslee2018@hallym.ac.kr
Received 2020 July 28; Revised 2020 September 29; Accepted 2020 October 04.

Abstract

Mild cognitive impairment (MCI) is the preclinical stage and sign of dementia. It is also important for guidance in the prevention and intervention of neurological disease. The purpose of this study was to review literatures on cognitive/communicative and other predictors of MCI patients systematically, and to propose the evidence-based data including effect sizes of them using a meta-analysis method. Fifty-seven researches published since 2010, meeting the inclusion and exclusion criteria, were entered into the analysis. They were analyzed in a methodological and content level, and the effect sizes were calculated by 3 predictors. Predictive values were pooled from cognitive (10 domains), communicative (9 domains), and other (3 domains). The main findings were as follows. Firstly, the general target population for studies was older adults over the age of 55, and most studies included at least 2 types of predictors. Secondly, average effect sizes of 3 predictors in MCI were all significant. Thirdly, cognitive predictors like memory and general cognition had significant and high-level effect sizes. Fourthly, communicative predictors including comprehension and word fluency had moderate-level effect sizes significantly. Lastly, all demographic and neuropsychological (age, education, depression) predictors had significant and moderate-level effect sizes. Our results provide the evidence-based information to predict MCI. Especially, specific cognitive and communicative predictors may contribute to increase the diagnostic and prognostic accuracy in MCI. This study is also expected to present clinically available data and increase the effect in intervention for MCI.

INTRODUCTION

전 세계적인 치매 인구는 약 5,000만 명에 이른다(Alzheimer’s Disease International, 2018). 2050년까지 3초당 1명씩 환자가 추가되어 1억 5,200만 명으로 증가할 것이라는 전망도 있다(Wang et al., 2019). 이 같은 추세를 감안할 때 치매의 전조 증상이자 임상 전(preclinical) 단계에 해당하는 경도인지장애(mild cognitive impairment, MCI)는 임상적으로 매우 중요한 함의를 갖는다.

MCI는 일상 활동에 큰 영향을 주지 않으나 연령 및 교육수준에 비해 인지 저하가 두드러지는 상태로 임상적 진단 기준이 다양하다(Etgen et al., 2010). Morris et al.(2001)은 1) 기억장애의 발현, 2) 전반적 인지 및 기능의 보존, 3) 치매로 진단되지 않은 상태 등을 MCI의 진단 기준으로 제시한 바 있다. National Institute on Aging―Alzheimer’s Association의 보고서(Jack et al., 2018)에 따르면, 1) 객관적 인지 평가상 손상 또는 비정상적 범주의 수행력, 2) 인지의 기초선보다 저하된 인지 능력, 3) 일상 활동을 독립적으로 수행하나 보다 복잡한 활동에서 인지적 어려움이 감지되며 기능이 경미하게 저하된 상태 등을 MCI로 진단한다. 이밖에, Diagnostic and Statistical Manual of Mental Disorders, National Alzheimer’s Coordinating Center 등이 제시한 기준이나 임상적 증상, 생물학적 구조 등에 근거해 MCI를 진단한다.

이처럼 MCI의 진단 기준이 다양하다는 것은 임상적 판단이 매우 중요함을 시사한다. 즉 광범위한 임상적 평가, 임상적 관찰, 뇌영상 촬영, 생물학적 및 신경심리학적 검사, 일상생활 기능에 대한 보고 등을 통해 종합적으로 진단할 필요가 있다. 이는 MCI에 대한 중재를 구상하고 효과를 예측하기 위한 근거 자료가 된다는 점에서도 의의가 있다. 따라서 MCI의 임상적 양상을 영역별로 예측하면 진단적 준거와 중재의 지침으로서 활용할 수 있다.

MCI의 인지적 예측인자는 하위 영역별로 다양한데, 주로 기억력에 대한 논의가 많다. MCI의 주요 진단 근거로서 기억력 손상이 포함되기도 하며(Petersen, 2004), 알츠하이머병(Alzheimer’s disease, AD)으로 진행하지 않고 장기간 일정 수준의 수행력을 유지하는 MCI 환자는 기억력이 상대적으로 높다(Alves et al., 2018). 특히 작업기억 용량이 축소되면 인지-언어 기능을 저하시킨다. 이는 새로운 상황에 효율적으로 적응하기 위한 유동성 지능에도 영향을 주며, 정보 처리 속도, 학습, 구문 이해, 유창성, 담화 산출 등을 방해한다. 주의력은 추론력, 집행기능, 언어 과제 등을 수행하는 데 전제되므로 신경학적 변화에 민감하다(Facal et al., 2014; Park et al., 2017). 즉 MCI로 인해 전전두피질의 크기와 두께가 축소되면 이러한 기제에도 부정적인 영향을 미친다. 예컨대, 인지-언어 과제를 수행할 때 교차감각적(cross-modal) 방해물이나 불필요한 자극을 억제하지 못해 지각 및 인지 처리가 크게 제한된다(Alves et al., 2018; Facal et al., 2014). 또 다른 예측인자인 집행기능은 주의력, 작업기억 등 다른 인지 영역과 함께 복합적인 기능망을 형성하기 때문에, MCI로 인해 이러한 기능이 떨어지면 계획과 의사결정, 반응 조율, 행동 억제, 정보 업데이트 등이 어려울 수 있다. 신경 심리 선별 검사나 지능 검사, 다양한 인지 영역이 포함된 치매 검사 등을 통한 전반적 인지 수행력을 MCI의 예측인자로 삼기도 한다. 인지적 수행이 하위 영역 간의 연계성에 기초하므로 전반적 인지 기능은 정상군으로부터 MCI를 변별하는 데 유용한 요인이다.

MCI를 예측하는 의사소통적 요인으로 단어유창성과 이름 대기 등 어휘-의미적 측면을 다루는 경우가 많다. 예컨대, 노인과 MCI 집단은 어휘-의미적 접근과 반응시간 간에 상관성이 크다(Facal et al., 2016). 이는 처리 자원이 감소할수록 어휘-의미적 접근의 어려움이 가중되기 때문이다. 이름대기와 어휘 학습 능력은 MCI와 AD를 변별하는 진단적 요소라는 보고도 있다(Hu¨bner et al., 2018). MCI의 의사소통적 예측인자로서 이해력을 꼽기도 하는데, 이에 활용되는 과제는 연구마다 상이하다. 즉 복잡한 문자적 및 비문자적 언어, 지시 따르기, 청각적 이해력, 토큰 검사 등을 통해 언어적 해석과 처리를 평가한 후 정상 노인이나 AD 집단과 변별한다. 이밖에 읽기나 쓰기, 비유 언어, 전반적 언어 등이 MCI의 의사소통적 예측 요인으로 활용되기도 한다.

MCI를 예측하는 기타 요인으로 인구통계학적 및 신경심리적 측면이 고려될 수 있다. 예를 들어, 60세 이상 인구의 3%, 75세 이상 노인의 15%가 MCI로 진단된다(Wang et al., 2019). 65세 이상 인구 중 MCI의 발생률이 인구 1,000명당 25.9명에 이른다는 보고도 있다(Panza et al., 2005). 교육수준도 MCI의 보편적인 예측인자에 해당한다. 정상적인 인지 수준으로 회복한 MCI 집단은 상대적으로 연령이 낮고 교육수준은 높게 나타났다(Xue et al., 2019). Sattler et al.(2012)은 MCI와 AD의 발병률을 감소시키는 주요 요인으로서 높은 교육수준과 사회경제적 지위(socioeconomic status, SES)를 꼽았다. 우울증도 MCI와 상관성이 높다. 실제로 우울증이 있는 정상군과 MCI 환자군은 각각 MCI와 치매로 진행할 확률이 높다고 보고된다(Panza et al., 2010). 특히 시설에 거주하는 MCI 환자 중 약 44%가 우울증을 보이는 것으로 나타났다. MCI를 유발하는 신경병리적 속성이 우울증에 관여할 수 있고, 이는 MCI에서 AD로의 진행을 예측하는 인자로서 작용하기도 한다(Houde et al., 2008; Panza et al., 2010).

요컨대, MCI는 치매의 전조 증상으로서 진단과 예측의 차원에서 임상적으로 매우 중요하다. 질환의 속성상 진단 기준이 다양하고 방법론적 범주가 광범위하기 때문에 임상적 판단에 크게 좌우될 수 있다. 따라서 MCI의 변별에 활용할 수 있는 예측 인자를 증거 기반적이고 체계적인 관점에서 제시해야 한다. 이에 본 연구에서는 2010년 이후의 최근 연구들을 중심으로 MCI의 주요 영역별 예측인자를 체계적으로 고찰하고, 메타분석을 통해 효과크기(effect size) 등 증거 기반적 자료를 제시하고자 한다. 구체적인 연구 문제는 다음과 같다.

첫째, MCI의 주요 예측인자에 관한 연구들을 방법 및 내용의 질적 측면에서 살펴보고자 한다.

둘째, 예측인자의 각 영역(인지적/의사소통적/기타)에서 평균 효과크기 및 하위 영역별 효과크기를 산출하고 유의미한 예측인자가 무엇인지를 제시하고자 한다.

MATERIALS AND METHODS

문헌 수집

MCI의 예측인자를 분석하기 위해 다양한 데이터베이스와 학술지 검색을 통해 문헌을 수집하였다. PudMed, CINAHL Plus with Full Text, MEDLINE, Web of Science, EBSCO, OVID, Scopus 등의 데이터베이스를 활용하였다. 국외 학술지에 게재된 국내 연구도 분석 대상에 포함하였다. 2020년 6월 16일자를 기준으로 2010년 이후 게재된 연구에 한해 검색하였다. 구체적인 주제어는 다음과 같다: communication, conversion, detect, diagnose, discriminate, evaluate, language, marker, mild cognitive impairment (MCI), predict (or), prognostic, risk factor.

대상 연구의 선정

본 연구를 위한 논문의 선정 기준은 다음과 같다.

첫째, 연구 설계 방법으로서 확률화 배정 연구나 집단 간 비교 연구로 제한하였다. 동일 집단을 대상으로 한 반복 측정 연구, 단일사례 연구는 제외하였다. 둘째, 실험 집단의 연구 대상으로 MCI는 포함하였고, 파킨슨병 등 동반 장애가 있는 경우, 주관적 인지-언어장애, 기타 뇌 손상으로 인한 인지장애 등은 제외하였다. MCI의 하위 유형으로 기억력의 손상 유무에 따른 기억상실형 MCI (amnestic MCI, aMCI) 및 비기억상실형 MCI (nonamnestic MCI, non-aMCI), 손상 영역의 단일성을 기준으로 단일영역형(single-domain) 및 다영역형(multiple-domain) MCI를 모두 포함하였다. 통제 집단은 인지-언어 수행력이 정상 범주인 경우로 제한하였다. 셋째, 인지적 예측인자로서 주의력, 기억력, 고차원적 인지 등의 하위 영역이나 전반적 인지 능력을 제시한 연구를 포함하였고, 주관적 인지 호소, 신체적 및 심리-행동적 수행력에 대한 결과는 배제하였다. 넷째, 의사소통적 예측인자로서 이해, 이름대기, 읽기 등의 하위 영역이나 전반적 언어 능력을 포함하였고, 주관적 언어 능력은 제외하였다. 다섯째, 기타 예측인자로서 연령, 교육 연수, 우울증 등 인구통계학적 및 신경심리적 측면이 포함되었다. 여섯째, 각 영역별 수행력의 결과치로서 평균 및 표준편차를 제시한 연구만 선정하였다.

메타분석을 위해 1차적으로 총 761개의 논문을 선정하였고, 본 연구의 선정 기준에 따라 704개의 논문이 제외되어 최종적으로 총 57개의 논문을 분석하였다. 이 중 국내 연구는 4개, 국외 연구는 53개였다. 논문 선정 과정은 Figure 1에 순서도로 제시하였다.

Figure 1.

Flow of studies included in the meta-analysis. MCI: mild cognitive impairment.

자료의 코딩

분석 대상에 포함된 연구들은 연구자, 출판 연도, 대상자의 특성, 인지적/의사소통적/기타 예측인자의 하위 영역에 따라 코딩되었다. 예측인자는 제시된 통계값(평균/표준편차)으로 분류하여 코딩하였다. 독립변인은 MCI의 유무이고, 종속변인은 MCI의 인지적/의사소통적/기타 요인에 대한 수행력이었다. 코딩된 자료의 수는 총 402개였는데, 이는 1개 연구에서 2개 이상의 항목에 대한 결과치를 제시한 연구들이 대부분이었기 때 문이다. 예컨대, Torress et al.(2019)은 언어 이해에 대해 3개의 상이한 항목별로 결과치를 제시하였고, Zhou et al.(2016)의 연구에서는 기억력과 쓰기에 대해 다수의 결과치를 산정하였다.

코딩 작업은 언어병리학 교수 1인과 연구 보조원 1인에 의해 실시되었다. 이들은 1급 및 2급 언어재활사 국가자격증을 소지한 자들로 관련 임상 경력이 7년 이상이었으며, 신경언어장애에 대한 평가 및 치료 경험이 있었다. 최종 검토 시 이상치나 확인이 필요한 부분에 대해서는 상호 논의한 후 수정하였다.

연구의 질 및 신뢰도 평가

연구의 질을 평가하기 위해 Gersten et al.(2005)의 필수적인 질 지표(essential quality indicators)를 사용하여 3점 척도(1점: 부적절, 2점: 불명확, 3점: 적절)로 측정하였다. 주요 평가 항목으로는, 연구 대상자의 정보, 집단별 할당 방법, 수행력 평가의 절차 및 내용, 수행력에 대한 결과 제시 방법 등이었다. 57개 논문 중 55개는 평균 3점, 2개는 2.85점으로 평가되어 논문의 질적 수준이 적절한 것으로 확인되었다.

신뢰도를 평가하기 위해 전체 연구의 10%에 해당하는 논문을 우선적으로 선택한 후 2인의 평가자가 각각 코딩하고 효과 크기를 산출하였다. 그 결과, 평가자 간 신뢰도는 100%로 산정되었다.

메타분석

효과크기의 산출 및 해석

메타분석을 위한 인지적 예측인자에는 주의력, 계산력, 시지각력, 지남력, 기억력, 작업기억, 조직화 능력, 추론력, 집행기능, 전반적 인지 등 10개 하위 영역을 포함하였다. 기억력에는 학습과 회상(즉각/지연/자유) 능력이 포함되었다. 작업기억은 기억력의 하위 범주로 분류되기도 하나 본 연구에서는 집행기능 등 다른 영역과의 연계성을 고려하여 개별화하였다. 의사소통적 예측인자는 이해, 구문, 단어유창성, 이름대기, 단어정의, 읽기, 쓰기, 비유언어, 전반적 언어 등 9개 영역을 분석하였다. 단어유창성은 생성이름대기, 이름대기는 대면이름대기 과제의 수행력으로 각각 구분하였다. 기타 예측인자로서 연령, 교육 연수, 우울증 등 인구통계학적 및 신경심리적 측면의 3개 영역을 분석하였다. 이는 대다수의 선행 연구에서 공통적으로 검증된 요인들로, 사회경제적 지위 등 기타 요소는 수입, 재산, 생활양식, 지역사회 활동, 정치적 참여, 타인의 평가 등 지표가 다양해 본 연구의 분석 대상에서 제외하였다.

각 영역별로 수행력을 측정한 결과값을 분석 자료로 사용하였고, 메타분석용 통계 프로그램인 Comprehensive MetaAnalysis version 3 (Biostat Inc., Englewood, NJ, USA)으로 평균 및 영역별 효과크기를 분석하였다. 집단의 평균 및 표준편차에 기반하여 효과크기인 Hedges’s g값을 산정하였다. 효과크기는 연구 간의 표본 크기를 고려한 가중평균 효과크기(weight effect size)를 적용하였고, 95% 신뢰구간을 기준으로 효과크기의 유의성을 평가하였다.

효과크기를 해석하는 기준으로는 0.20 이하 시 ‘작은(small) 효과’, 0.50 수준 시 ‘중간(medium) 효과’, 0.80 이상 시 ‘큰(large) 효과’로 간주하였다(Lipsey & Wilson, 2001).

동질성 검증

개별 연구에서 도출된 효과크기의 통계적 이질성 유무를 확인하기 위해 동질성 검정을 시행하였다. 그 결과, Table 1과 같이 Q-df > 0, I2 ≥ 75%로 나타나 연구 간의 분산이 실제로 존재할 뿐 아니라 각 효과크기의 이질성이 상당함을 입증하였다(Higgins & Green, 2011). 따라서 본 메타분석을 위해 무선효과 모형(random effect model)을 적용하였다(Borenstein et al., 2009).

Verification of homogeneity

출판편의 검증

메타분석 결과의 타당성을 확보하기 위해 출판편의(publication bias)에 대한 검증을 실시하였다(Rosenthal, 1979). 오류의 존재 유무를 알아보기 위해 Figure 2와 같이 funnel plot을 확인한 결과 시각적 비대칭성이 관찰되었다. 비대칭의 보정을 위해 Duval & Tweedie(2000)의 trim-and-fill 기법을 적용하였는데, 보정 전의 risk ratio가 43%인 반면 보정 후에는 8%로 감소함을 알 수 있었다(Table 2).

Figure 2.

Funnel plot for checking the bias.

Statistical adjustment of asymmetry by trim-and-fill

통계적 대칭성을 분석하기 위해 Egger의 회귀분석(Egger’s regression test) 및 Kendall’s tau 검사를 시행하였다. Egger의 회귀분석에서 회귀식 초기값(intercept)은 -2.76으로 통계적 유의성이 있었다(p < 0.001, standard error = 0.45, df = 403.00). Kendall’s tau 검사 결과에서도 연속성 상관 유무 조건에서 모든 tau 값이 -0.06으로 나타나 통계적 유의성을 보였다(z for tau = 1.77, p ≤ 0.05). 따라서 본 연구를 위해 선정된 논문들이 출판편의상 오류가 있음을 알 수 있었다.

오류의 정도를 살펴보기 위해 Orwin(1983)의 안전계수(failsafe N) 공식을 활용하여 분석한 결과, 안전계수가 609.00으로 대상 연구 수에 비해 누락된 연구의 수가 충분히 큰 것으로 나타났다. 궁극적으로 본 연구의 출판편의는 매우 미미한 수준임을 알 수 있었다.

RESULTS

연구의 질적 분석

설계 방법의 측면에서 모든 연구들은 실험 및 통제 집단의 수행력을 비교해 제시하였다. 연구에 포함된 대상자 수는 6명부터 24,649명까지 매우 다양하였다(Appendix 1). 특히 100명 이하의 대상자로 구성된 연구가 총 40개로 대다수를 차지하였다.

연구 대상 중 실험 집단의 평균 연령은 56.9~83.4세, 통제 집단은 57.3~80.8세로 55세 이상 노년층이 주요 대상자였다. 실험 및 통제 집단의 평균 교육 연수는 각각 4.6~17.1년, 5.1~17.0년으로 분석되었다. 실험 집단의 MCI 유형은 aMCI, non-aMCI, 단일영역형 및 다영역형 MCI로 명시된 연구도 있으나, 대다수는 구체적인 하위 유형을 언급하지 않았다.

각 연구별로 실험 및 통제 집단의 예측인자를 제시한 결과는 Appendix 1에 제시하였다. 하위 요소는 인지적/의사소통적/기타 예측인자의 순으로 정리하였다. 한 분야의 예측인자만 포함된 3개 연구를 제외한 나머지는 최소 2개 분야 이상의 예측인자를 제시하였다.

예측인자의 수행력을 파악하는 데 활용된 평가도구는 연구마다 상이하였다. 인지적 예측인자 중 기억력은 Auditory Verbal Learning Test, Rey-Osterrieth Complex Figure의 recall 등, 전반적 인지는 단어-그림 짝짓기, 명령 이행, 문장 처리 등으로 평가되었다. 의사소통적 예측인자의 이해는 matching wordpicture, commands, sentence processing 등, 단어유창성과 이름대기는 Verbal Fluency Test (phonological/semantic), Boston Naming Test (BNT) 등이 제시되었다. 기타 예측인자(인구통계학적 및 신경심리적) 중 우울증은 Geriatric Depression Scale과 Beck’s Depression Inventory II가 주로 활용되었다.

효과크기의 분석

평균 효과크기

MCI 환자군과 통제군 간의 3개 예측인자를 비교한 결과(Table 3), 모든 영역의 평균 효과크기가 유의하게 나타났다. 인지적 예측인자의 경우 g = -0.81 (p < 0.001), 95% CI = [-0.97, -0.66]으로 ‘큰’ 수준이며, 의사소통적 예측인자는 g = -0.55 (p < 0.001), 95% CI = [0.10, 0.34]로 ‘중간’ 수준이었다. 기타 예측인자(인구통계학적 및 신경심리적)의 평균 효과크기는 g = 0.10 (p < 0.001), 95% CI = [-0.00, 0.19]로 ‘작은’ 수준에 해당하였다.

Mean effect sizes in three factors

하위 영역별 효과크기

3개 예측인자에서 각 하위 영역별로 효과크기를 비교한 결과는 Table 4~6에 제시하였다.

Effect sizes in cognitive factors

인지적 예측변인 중 계산력(g = -6.48, p < 0.001), 시지각력(g = -0.89, p < 0.001), 기억력(g = -1.09, p < 0.001), 전반적 인지(g = -1.04, p < 0.001)는 ‘큰’ 효과크기, 주의력(g = -0.51, p < 0.001)과 지남력(g = 0.58, p < 0.001)은 ‘중간’ 수준의 효과크기, 작업기억(g = -0.13, p < 0.001)과 집행기능(g = 0.00, p < 0.001)은 ‘작은’ 효과크기로서 유의하게 분석되었다. 조직화 능력(g = -0.58, p = 0.359)과 추론력(g = -0.43, p = 0.831)의 효과크기는 유의미하지 않았다. 효과크기가 비교적 높은 ‘기억력’의 forest plot은 Figure 3에 제시하였다.

Figure 3.

Forest plot of ‘memory’ factor. CI: confidence interval.

의사소통적 예측인자는 모두 유의한 효과크기를 보였다. ‘중간’ 수준의 효과크기는 이해(g = -0.75, p < 0.001), 단어유창성(g = -0.56, p < 0.001), 이름대기(g = -0.39, p < 0.001), 단어 정의(g = 0.47, p < 0.001), 쓰기(g = -0.71, p < 0.05), 비유언어(g = -0.31, p < 0.001), 전반적 언어(g = -0.64, p < 0.001) 등 7개 요인에서 분석되었다. 구문(g = -0.11, p < 0.05)과 읽기(g = -0.19, p < 0.001)는 ‘작은’ 효과크기를 나타내었다. 가장 높은 효과크기를 보인 ‘이해’ 요인의 forest plot은 Figure 4에 제시하였다.

Figure 4.

Forest plot of ‘comprehension’ factor. CI: confidence interval.

기타 예측인자(인구통계학적 및 신경심리적)는 연령(g = 0.58, p < 0.001), 교육 연수(g = -0.29, p < 0.001), 우울증(g = 0.22, p < 0.001)의 순으로 효과크기가 높게 나타났다. 세 요인은 모두 ‘중간’ 수준의 유의한 효과크기를 보였다. 효과크기가 상대적으로 높은 ‘연령’ 요인의 forest plot은 Figure 5에 제시하였다. Figures 3~5에서 사용된 문헌은 Appendix 2에 목록화하였다.

Figure 5.

Forest plot of ‘age’ factor. CI: confidence interval.

DISCUSSIONS

본 연구는 체계적 고찰과 메타분석을 통해 MCI에 대한 인지적 및 의사소통적, 기타 예측인자를 제시하고자 하였다. 2010년 이후의 57개 국내외 연구들을 대상으로 방법 및 내용의 질적 특성을 살펴보고, 각 예측인자의 효과크기를 분석하였다.

연구의 질적 분석

연구의 방법적 측면에서 MCI 환자군과 통제군은 주로 55세 이상의 노년층으로 구성되었다. 평균 교육 연수는 각각 4.6~17.1년, 5.1~17.0년으로 교육수준의 범주가 매우 광범위한 것으로 분석되었다. 대상자 수는 10명 이하의 소집단부터 2만 명 이상의 대 규모 연구까지 광범위했으나 전체 연구의 약 70%는 10~100명 수준에 해당하였다.

일부 연구에서는 MCI의 유형을 aMCI, non-aMCI, 단일영역형 및 다영역형 MCI로 세분화하였으나(Mohammadi et al., 2018), 하위 유형을 구체화하지 않은 연구가 대다수를 차지하였다. MCI는 기억력의 손상 및 단일성 여부를 기준으로 aMCI와 non-aMCI, 단일영역형 MCI와 다영역형 MCI로 세분화된다(Fischer et al., 2007). 하위 유형에 따라 인지-언어적 양상이나 신경병리학적 진행 속도가 다르다. 예를 들어, 다영역형 aMCI는 AD로 진행될 가능성이 높은 반면, non-aMCI는 루이소체 치매(dementia with Lewy bodies), 전두측두엽변성(frontotemporal degeneration) 등 비알츠하이머형 치매로 진행되는 경우가 많다(Folenza et al., 2009). 하위 유형을 세분화하지 않은 연구들이 대다수이고, 이에 근거해 본 연구가 진행되었으므로 유형별 분류는 분석 기준에 포함되지 않았다. 따라서 향후 연구에서는 MCI의 하위 유형별로 연구 대상이 세분화될 것으로 보인다.

연구의 내용적 분석 결과, 각 연구에서 도출한 예측인자의 수나 유형은 매우 상이하였다. 인지적 및 의사소통적, 기타 예측인자 중 일부 혹은 전체가 포함되기도 하고, 생물학적/신체적/기능적 측면의 인자를 추가적으로 비교한 연구도 있었다. 본 연구에서는 수행력의 통계값으로 평균과 표준편차를 분석 대상으로 삼았으나, Z-score나 예측인자 간 상관계수를 함께 제시한 연구도 있었다. MCI의 변별적 진단이 요구되고 이에 따른 임상적 판단의 중요성이 증대되면서 예측인자의 유형이 보다 세분화될 필요가 있다. 실제로 최근 연구들에서 MCI의 사회경제적 지표, 심리정서적 요소, 뇌영상 연구에 근거한 생물학적 지표 등이 제시되는 추세이다(Burton et al., 2018; Chandra et al., 2019; Dodge et al., 2015; Houde et al., 2008).

수행력을 파악하기 위해 활용된 과제의 다양성은 영역별로 분석된 자료 수와 비례하였는데, 이 같은 특성은 인지적 예측인자에서 가장 두드러졌다. 즉 인지적 예측인자의 하위 유형에 따라 자료 수가 매우 다양하고 수행력을 평가하는 데 활용된 과제가 광범위하였다. 예를 들어, 분석된 자료 수가 많은 기억력은 어휘 학습, 단어 회상(즉각/지연) 및 재인, 기억하여 그리기, 논리적 기억 및 이야기 회상 등의 과제로 평가되었다(Mohammadi et al., 2018; Park et al., 2017). 또 집행기능은 스트룹 검사(Stroop Test), 탑 검사(Tower test), 전반적 집행기능 체계 검사, 기호잇기 검사(Trail making test) 등이 다양하게 활용되었다(Arsenault-Lapierre et al., 2012; Manouilidoua et al., 2016). 분석된 자료 수가 적은 지남력, 조직화 능력, 추론력, 계산력 등은 과제의 유형이 비교적 단일하였다.

이에 반해 의사소통적 예측인자는 분석된 자료 수와 과제 유형 간의 상관성이 비교적 미미하였다. 예컨대, 자료 수가 가장 많은 ‘이해’ 요인은 주로 단어 처리나 문장 이해(지시 따르기) 과제를 통해 평가되었다(Manouilidoua et al., 2016; Payne & Stine-Morrow, 2016). 대다수의 연구에서 단어유창성은 특정 범주(예: animal)나 음운(예: F/A/S)으로, 이름대기는 특정 평가도구(예: BNT)로 수행력을 파악하였다. 이러한 경향성은, 다양화된 신경심리 검사와 영역별 심화검사를 통해 인지 능력을 파악하는 것과 대조적으로 언어 및 의사소통에 대한 세분화된 과제나 도구가 부족한 현실을 반영한다. 반면 대부분의 연구에서 공통적으로 활용되는 평가의 경우 신뢰도와 타당도가 이미 검증되었고 임상적 활용도가 높다. 이는 여러 연구를 통해 도출된 결과치가 일관적임을 시사하므로 본 분석 결과의 정확도를 신뢰할 만하다고 간주할 수 있다.

효과크기의 분석

본 연구에서 3개 예측인자의 평균 효과크기는 모두 유의하였고, 인지적 및 의사소통적, 기타(인구통계학적·신경심리적) 예측인자 순으로 높게 나타났다. 즉 MCI를 변별하는 데 있어 3개 예측인자가 모두 유효하며, 특히 인지적 예측인자는 정상군으로부터 MCI를 변별하는 가장 강력한 예측인자였다. 이는 여러 선행 연구를 통해 규명된 바 있다. 예를 들어, 전반적 인지를 포함한 인지 기능은 MCI를 예측하는 주요 변인으로 작용한다(Rajtar-Zembaty et al., 2019). 또 초기 단계의 MCI 집단에서 인지 능력은 질환의 진전을 예측하는 가장 민감한 변인으로 제시된 바 있는데, 이는 생물학적 지표와도 상관성이 크다(Egli et al., 2015). 특히 단어 학습 등의 기억력 과제는 AD로의 진행을 예측할 수 있다고 보고된다.

예측인자의 영역별로 효과크기를 분석한 결과, 인지적 예측변인 중 계산력, 시지각력, 기억력, 전반적 인지는 ‘큰’ 수준, 주의력과 지남력은 ‘중간’ 수준, 작업기억과 집행기능은 ‘작은’ 수준의 유의미한 효과크기를 나타내었다. 조직화 능력과 추론력은 효과크기가 유의미하지 않았는데, 이는 분석 대상에 해당하는 자료 수가 상대적으로 적은 데 따른 것으로 추측된다.

Rajtar-Zembaty et al.(2019)은 기억력과 전반적 인지가 MCI와 정상 노인을 변별하는 가장 강력한 독립적 요인으로 제시한 바 있는데, 이는 본 연구 결과와도 일치한다. MCI의 진단 기준에 주관적 기억 호소와 객관적 기억 결함이 포함되기도 하며, 기억력장애가 두드러지는 aMCI는 AD로 진행될 확률이 높아 임상적으로 주목되는 유형이다(Petersen, 2004). 특히 본 연구의 기억력에는 단어에 대한 학습이 대다수 포함되었고, 도형, 단어 및 문장, 이야기에 대한 즉각 및 지연, 자유, 단서 회상 등 다양한 회상 과제가 활용되었다. 이는 예측인자뿐 아니라 중재를 계획할 때 활용할 수 있는 증거 기반적 자료로서 유용하다.

본 연구에서 기억력 및 전반적 인지와 유사한 수준의 높은 효과크기를 보인 시지각력은 시공간, 시각구성 등의 과제를 통해 평가되는데, MCI에서 결함이 나타나는 보편적 영역 중 하나이다. 즉 시각 자극의 변별, 새로운 자극의 분석, 친숙한 자극의 재인, 시각 자극의 이해 등에 관여하는 시지각뿐 아니라 운동 반응과 결합된 시각구성력이 저하된다. 이로 인해 MCI 환자는 Clock Drawing Test와 Raven’s Progressive Matrices 등 시공간 과제의 수행력이 낮다. Alves et al.(2018)은 3년 후 AD로의 진행을 예측하는 데 유효한 지표로서 시공간력을 꼽았다.

작업기억과 집행기능은 전전두피질(prefrontal cortex)과 흑질선상체(nigrostriatal)의 도파민 신경전달 체계가 관장하는 영역으로, 노화나 신경학적 질환 등 신경학적 변화에 민감하다. MCI나 AD의 초기부터 두 기능이 손상되므로 일상생활의 과제를 수행하는 데 복합적인 어려움을 초래한다(Lee & Kim, 2020a, 2020b). 따라서 집행기능과 연계된 작업기억의 몇몇 하위 요소는 MCI에서 AD로의 진전을 예측하는 주요 인자에 해당한다(Alves et al., 2018; Facal et al., 2014; Park et al., 2017).

본 연구의 메타분석 결과 의사소통적 예측인자의 효과크기는 모두 유의하였고, 이해, 단어유창성, 이름대기, 단어정의, 쓰기, 비유언어, 전반적 언어 등이 상대적으로 큰 효과크기를 보였다. 특히 이해 및 쓰기, 전반적 언어는 의사소통 영역 중 가장 큰 효과크기를 갖는 것으로 분석되었다. 언어 이해는 기억력, 주의력 할당 등과도 연계되는데, 이 같은 인지적 요구가 증가할수록 제시된 문장을 이해하거나 읽은 내용을 해석하는 데 어려움을 보인다(Peach & Shapiro, 2020). MCI의 경우 하위 유형에 따라 다른 양상을 보이기도 한다. 즉 aMCI는 non-aMCI에 비해 언어 이해와 관련된 인지적 처리에 더 큰 결함이 있어 이해 과제의 수행력이 더 낮다(Payne & Stine-Morrow, 2016; Zhou et al. 2016). 단어 수준에서는 형태론적 규칙에 대한 지식은 보존되나 복잡한 비단어를 이해하는 데 어려움을 보인다. 이는 비문법적 형태의 재인을 지연시키는 집행기능의 저하와도 연관된다(Manouilidoua et al., 2016). 따라서 MCI의 예측인자로서 ‘이해’ 요인의 유효성은 인지적 예측인자, 특히 기억력과 주의력, 집행기능의 영향도 크게 작용한다. 쓰기의 경우 단어나 과제의 유형과 같은 다양한 변인이 작용하나, MCI로 인해 구문적으로 단순하고 정보 단위가 적은 문장을 산출하는 경향이 있다. 글자상 오류도 정상군과의 변별에 활용될 수 있다.

어휘-의미적 측면인 단어유창성과 이름대기도 신경학적 변화에 민감한 영역이다. 이들은 의미기억과 상관성이 높고, 의미적·음운적 및 글자 체계로 조직화되어 상호 연계된 네트워크를 형성하기 때문이다(Hu¨bner et al., 2018; Lee & Kim, 2020b). 따라서 단어유창성과 이름대기의 효과크기는 기억력의 효과크기가 유효한 것과도 크게 관련된다. 예를 들어, 대면이름 대기를 수행할 때 먼저 그림 자극이 의미 영역을 활성화한 후 단어를 산출하기 위해 음운 표상을 자극한다(Peach & Shapiro, 2020). 이때 어휘의 빈도, 어휘 사용의 최신성, 신경학적 변화 등이 영향을 미친다. MCI와 AD는 의미 체계 자체의 결함이나 접근 능력의 저하로 의미적 오류가 빈번해진다(Hu¨bner et al., 2018; Zhou et al., 2016). 단어정의 과제에서 MCI의 수행력이 낮은 것도 이와 동일한 맥락으로 해석될 수 있다(Lee & Kim, 2020b).

메타분석 결과에서 비유언어의 효과크기는 다른 의사소통 영역에 비해 상대적으로 낮으나 유의미한 것으로 분석되었다. 이는 집행기능의 역할과 연계된다. 비유언어는 문자 그대로가 아닌 화용적 추론에 기반하는데, MCI로 인해 집행기능이 저하되면 의미적 부호화를 반복함으로써 억제 기능이 제한된다(Lee & Kim, 2020b). 이로 인해 문자적 해석과 비문자적 해석 간의 전환이 느리고, 드러나지 않은 의미에 대한 추론적 실행이 어려워진다(Alves et al., 2018).

본 연구에서 인구통계학적 및 신경심리적 측면의 기타 예측인자는 연령, 교육 연수, 우울증 순으로 높았고, 모두 ‘중간’ 수준의 유의미한 효과크기를 보였다. MCI 이후 정상적인 인지의사소통 수준으로 회복하는 데 초점을 둔 연구들은 예측인자의 규명이 가역성 비율(reversion rate)을 증진한다는 점에서 중요하다고 강조하였다(Xue et al., 2019). 이는 AD로의 진행을 예방하는 차원에서도 매우 유의미하다. 특히 더 적은 연령과 높은 교육수준은 MCI의 회복에 결정적으로 기여한다(Sattler et al., 2012; Xue et al., 2019). 반면 연령이 높거나 교육수준이 낮을수록 AD로 진행할 위험이 증가한다. 높은 교육수준은 인지 보존능력(cognitive reserve)의 주요 하위 영역으로서 인지의사소통 기능을 보존하는 데 기여한다고 알려져 있다(Lee, 2015). 또 문해 능력을 보존함으로써 건강 관련 지식을 이해하고 활용하는 데 유리할 뿐 아니라 건강한 생활양식과 SES를 유지할 가능성이 높다(Houde et al., 2008).

우울증도 신경학적 질환의 위험 요인 중 하나일 수 있다. 신경학적 질환과 우울증의 동반은 생물학적, 감정적, 인지적 증상 간의 복잡한 상호작용을 반영한다. 우울증 자체가 MCI에서 AD로의 진행을 예측하지는 않으나 2~3년간 지속되는 우울장애는 AD로의 진행 가능성을 증가시킬 수 있다(Houde et al., 2008; Park et al., 2017). 특히 aMCI에 동반된 지속적인 우울 증상은 인지 저하를 더욱 심화시킨다. 우울감(melancholy) 등의 하위 증상은 특정 신경전달물질 체계나 해부학적 네트워크와 연관되어 AD로의 진행을 가중시킨다는 연구 결과도 있다(Houde et al., 2008).

요컨대, 본 연구는 메타분석을 활용하여 MCI를 예측하는 인지적 및 의사소통적 요인과 기타 요인을 도출한 후 질적 및 양적 분석 결과를 체계적으로 논의하였다. 이는 MCI를 변별적으로 진단하기 위한 증거 기반적 지침으로서 활용될 수 있을 것이다. 또 인지-의사소통 양상의 심화나 AD로의 진행을 예방하는데 기여할 것이다. 궁극적으로, MCI 전후의 중재 과정에서 인지-의사소통적 예측인자와의 연계성을 고려한 과제를 구상하고 중재의 효과성을 제고하기 위한 기초로 삼을 수 있을 것이다.

본 연구의 제한점은 다음과 같다. 첫째, 본 메타분석은 데이터베이스 및 학술지에 게재된 연구만을 대상으로 하였다. 따라서 출판편의에 대한 검증을 거쳤음에도 ‘회색 문헌(grey publications)’이나 ‘비관행적 문헌(non-conventional publication)’을 포함하지 않은 데 따른 편향성을 완전히 배제할 수 없을 것이다. 둘째, 본 연구에서 지남력, 조직화 능력 등의 인지적 요인과 구문, 비유언어 등의 의사소통적 요인은 분석된 자료 수가 5개 미만으로 비교적 적은 편이었다. 이는 각 효과크기의 신뢰도와 타당도에 영향을 미칠 수 있을 것이다. 셋째, MCI는 하위 유형에 따라 인지-의사소통 양상이 다르게 나타난다. 후속 연구를 통해 aMCI 및 non-aMCI, 단일 및 다영역형 MCI 등 하위 유형별로 예측인자를 제시하거나 비교할 필요가 있을 것이다. 넷째, 본 연구는 국외 연구를 주요 분석 대상으로 삼았고 국내의 경우 해외 학술지에 게재된 연구로 제한하였다. 이는 본 연구의 선정 기준에 부합하는 국내 연구가 양적으로 매우 부족한 현실을 감안한 것이다. 향후에는 선정 기준의 확대, 방법론적 전환, 문헌의 다양화 등을 통해 국내 연구에 중점을 둔 분석을 시도할 필요가 있다.

Effect sizes in communicative factors

Effect sizes in other factors

Acknowledgements

The authors thank to the participants.

Notes

Ethical Statement

We explained exactly the purpose and process of this study to all participants in advance. And they agreed to participate in it.

Declaration of Conflicting Interests

There are no conflict of interests.

Funding

This work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (NRF-2020S1A5A8040953).

References

1. Alves L., Cardoso S., Maroco J., de Mendonça A., Guerreiro M., Silva D.. 2018;Neuropsychological predictors of long-term (10 Years) mild cognitive impairment stability. Journal of Alzheimer’s Disease: JAD 62(4):1703–1711.
2. Alzheimer’s Disease International.. 2018. World Alzheimer Report 2018 London: Alzheimer’s Disease International.
3. Arsenault-Lapierre G., Bergman H., Chertkow H.. 2012;Word reading threshold and mild cognitive impairment: a validation study. BMC Geriatrics 12:38.
4. Borenstein M., Hedges L. V., Higgins J. P., Rothstein H. R.. 2009. Introduction to Meta-Analysis Chichester: John Wiley & Sons.
5. Burton R. L., O’Connell M. E., Morgan D. G.. 2018;Cognitive and neuropsychiatric correlates of functional impairment across the continuum of no cognitive impairment to dementia. Archives of Clinical Neuropsychology 33(7):795–807.
6. Chandra A., Valkimadi P., Pagano G., Cousins O., Dervenoulas G., Marios P. for the Alzheimer’s Disease Neuroimaging Initiative.. 2019;Applications of amyloid, tau, and neuroinflammation PET imaging to Alzheimer’s disease and mild cognitive impairment. Human Brain Mapping 40(18):5424–5442.
7. Dodge H. H., Mattek N., Gregor M., Bowman M., Seelye A., Ybarra O., et al. 2015;Social markers of mild cognitive impairment: Proportion of word counts in free conversational speech. Current Alzheimer Research 12(6):513–519.
8. Duval S., Tweedie R.. 2000;A nonparametric “trim and fill” method of accounting for publication bias in meta-analysis. Journal of the American Statistical Association 95(449):89–98.
9. Egli S. C., Hirni D. I., Taylor K. I., Berres M., Regeniter A., Gass A., et al. 2015;Varying strength of cognitive markers and biomarkers to predict conversion and cognitive decline in an early-stage-enriched mild cognitive impairment sample. Journal of Alzheimer’s Disease 44(2):625–633.
10. Etgen T., Sander D., Bickel H., Sander K., Förstl H.. 2010;Cognitive decline: The relevance of diabetes, hyperlipidaemia and hypertension. The British Journal of Diabetes and Vascular Disease 10(3):115–122.
11. Facal D., Juncos-Rabadán O., Guardia-Olmos J., Pereiro A. X.. 2016;Temporal changes in semantic and lexical access related to mild cognitive impairment. Aging Clinical and Experimental Research 28:497–504.
12. Facal D., Juncos-Rabadán O., Pereiro A. X., Lojo-Seoane C.. 2014;Working memory span in mild cognitive impairment. Influence of processing speed and cognitive reserve. International Psychogeriatrics 26(4):615–625.
13. Fischer P., Jungwirth S., Zehetmayer S., Weissgram S., Hoenigschnabl S., Gelpi E., et al. 2007;Conversion from subtypes of mild cognitive impairment to Alzheimer dementia. Neurology 68(4):288–291.
14. Forlenza O. V., Diniz B. S., Nunes P. V., Memória C. M., Yassuda M. S., Gattaz W. F.. 2009;Diagnostic transitions in mild cognitive impairment subtypes. International Psychogeriatrics 21(6):1088–1095.
15. Gersten R., Fuchs L. S., Compton D., Coyne M., Greenwood C., Innocenti M. S.. 2005;Quality indicators for group experimental and quasi-experimental research in special education. Exceptional Children 71(2):149–164.
16. HigginsMJ. P. T. GreenS. 2011. March. Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0. The Cochrane Collaboration. Retrieved from https://handbook-5-1.cochrane.org/.
17. Houde M., Bergman H., Whitehead V., Chertkow H.. 2008;A predictive depression pattern in mild cognitive impairment. International Journal of Geriatric Psychiatry 23(10):1028–1033.
18. Hübner L. C., Loureiro F., Tessaro B., Siqueira E. C. G., Jerônimo G. M., Gomes I., et al. 2018;Naming and verbal learning in adults with Alzheimer’s disease, mild cognitive impairment and in healthy aging, with low educational levels. Arquivos de Neuro-Psiquiatria 76(2):93–99.
19. Jack C. R., Jr, Bennett D. A., Blennow K., Carrillo M. C., Dunn B., Haeberlein S. B., et al. 2018;NIA-AA Research Framework: toward a biological definition of Alzheimer’s disease. Alzheimer’s and Dementia: The Journal of the Alzheimer’s Association 14(4):535–562.
20. Lee M. S.. 2015;Meta-analysis of correlation between cognitive-linguistic ability and cognitive reserve in normal aging. The Journal of the Korea Contents Association 15(11):359–373.
21. Lee, M. S. & Kim, B. S. (2020). Effects of working memory intervention on language production by individuals with dementia. Neuropsychological Rehabilitation. [Epub]. https://doi.org/10.1080/09602011.2020.1789479.
22. Lee M. S., Kim B. S.. 2020;Higher order cognition and communication: A preliminary study for the development of brief test of cognitive-communication disorders. Audiology and Speech Research 16(3):236–244.
23. Lipsey M. W., Wilson D. B.. 2001. Practical Meta-Analysis Thousand Oaks, CA: SAGE Publications.
24. Manouilidou C., Dolenc B., Marvin T., Pirtošek Z.. 2016;Processing complex pseudo-words in mild cognitive impairment: The interaction of preserved morphological rule knowledge with compromised cognitive ability. Clinical Linguistics and Phonetics 30(1):49–67.
25. Mohammadi A., Kargar M., Hesami E.. 2018;Using virtual reality to distinguish subjects with multiple- but not single-domain amnestic mild cognitive impairment from normal elderly subjects. Psychogeriatrics 18(2):132–142.
26. Morris J. C., Storandt M., Miller J. P., McKeel D. W., Price J. L., Rubin E. H., et al. 2001;Mild cognitive impairment represents early-stage Alzheimer disease. Archives of Neurology 58(3):397–405.
27. Orwin R. G.. 1983;A fail-safe N for effect size in meta-analysis. Journal of Educational Statistics 8(2):157–159.
28. Panza F., D’Introno A., Colacicco A. M., Capurso C., Del Parigi A., Caselli R. J., et al. 2005;Current epidemiology of mild cognitive impairment and other predementia syndromes. The American Journal of Geriatric Psychiatry 13(8):633–644.
29. Panza F., Frisardi V., Capurso C., D’Introno A., Colacicco A. M., Imbimbo B. P., et al. 2010;Late-life depression, mild cognitive impairment, and dementia: Possible continuum? The American Journal of Geriatric Psychiatry 18(2):98–116.
30. Park J. H., Park H., Sohn S. W., Kim S., Park K. W.. 2017;Memory performance on the story recall test and prediction of cognitive dysfunction progression in mild cognitive impairment and Alzheimer’s dementia. Geriatrics and Gerontology International 17(10):1603–1609.
31. Payne B. R., Stine-Morrow E. A.. 2016;Risk for mild cognitive impairment is associated with semantic integration deficits in sentence processing and memory. The Journals of Gerontology: Series B 71(2):243–253.
32. Peach R. K., Shapiro L. P.. 2020. Cognition and Acquired Language Disorders: An Information Processing Approach (Lee, M. S. & Kim, S. R., Trans.). (1st ed.) Seoul: Hakjisa.
33. Petersen R. C.. 2004;Mild cognitive impairment as a diagnostic entity. Journal of Internal Medicine 256(3):183–194.
34. Rajtar-Zembaty A., Rajtar-Zembaty J., Sałkowski A., Starowicz-Filip A., Skalska A.. 2019;Global cognitive functioning and physical mobility in older adults with and without mild cognitive impairment: Evidence and implications. Folia medica Cracoviensia 59(1):75–88.
35. Rosenthal R.. 1979;The “file drawer problem” and tolerance for null results. Psychological Bulletin 86(3):638–641.
36. Sattler C., Toro P., Schönknecht P., Schröder J.. 2012;Cognitive activity, education and socioeconomic status as preventive factors for mild cognitive impairment and Alzheimer’s disease. Psychiatry Research 196(1):90–95.
37. Torres V. L., Rosselli M., Loewenstein D. A., Curiel R. E., Vélez Uribe I., Lang M., et al. 2019;Types of errors on a semantic interference task in mild cognitive impairment and dementia. Neuropsychology 33(5):670–684.
38. Wang J., Wang L., Zhou X., Wen X., Zhen X.. 2019;Risk factors for predicting progression from normal cognition to mild cognitive impairment: Protocol for a systematic review and meta-analysis of cohort studies. BMJ Open 9(6):e027313.
39. Xue H., Hou P., Li Y., Mao X., Wu L., Liu Y.. 2019;Factors for predicting reversion from mild cognitive impairment to normal cognition: A meta-analysis. International Journal of Geriatric Psychiatry 34(10):1361–1368.
40. Zhou J., Jiang B., Huang X. H., Kong L. L., Li H. L.. 2016;Characteristics of agraphia in Chinese patients with Alzheimer’s disease and amnestic mild cognitive impairment. Chinese Medical Journal 129(13):1553–1557.

Appendices

Appendix 1. Characteristics of Studies Included in the Meta-Analysis

Appendix 2. Lists of Studies Included in the Meta-Analysis

Article information Continued

Figure 1.

Flow of studies included in the meta-analysis. MCI: mild cognitive impairment.

Figure 2.

Funnel plot for checking the bias.

Figure 3.

Forest plot of ‘memory’ factor. CI: confidence interval.

Figure 4.

Forest plot of ‘comprehension’ factor. CI: confidence interval.

Figure 5.

Forest plot of ‘age’ factor. CI: confidence interval.

Table 1.

Verification of homogeneity

Q df p I2 T2
1310.58 51 <0.001 96.11 0.32

Q: heterogeneity between studies within domain, I2: percentage of heterogeneity caused by study differences, T2: between-studies variance

Table 2.

Statistical adjustment of asymmetry by trim-and-fill

Random effect
Risk ratio Lower limit Upper limit
Pre-adjustment 0.43 0.51 0.36
Post-adjustment 0.08 0.01 0.16

Table 3.

Mean effect sizes in three factors

Factor g 95% CI K Q p (Q) I2 (%) FSN
Cognitive -0.81 [-0.97, -0.66] 183 7161.86 < 0.001*** 97.46 725
Communicative -0.55 [0.10, 0.34] 103 1840.80 < 0.001*** 94.46 443
Other 0.10 [-0.00, 0.19] 116 5650.18 < 0.001*** 97.96 1,899
*

p < 0.05,

**

p < 0.01,

***

p < 0.001.

g: Hedge’s g, CI: confidence interval, K: number of the effect size, Q: heterogeneity between studies within domain, p (Q): p-value for heterogeneity, I2: percentage of heterogeneity caused by study differences, FSN: fail safe N

Table 4.

Effect sizes in cognitive factors

Subdomain g 95% CI K Q p (Q) I2 (%) FSN
Attention -0.51 [-1.08, 0.07] 13 519.51 < 0.001*** 97.69 508
Calculation -6.48 [-9.40, -3.56] 4 279.67 < 0.001*** 98.93 544
Visuoperception -0.89 [-1.61, -0.17] 13 383.48 < 0.001*** 96.87 875
Orientation 0.58 [-1.11, 2.28] 2 20.26 < 0.001*** 95.06 NA
Memory -1.09 [-1.38, -0.80] 56 1839.54 < 0.001*** 97.01 413
Working memory -0.13 [-0.34, 0.08] 17 107.40 < 0.001*** 85.10 350
Organization -0.58 [-0.95, -0.21] 2 0.84 0.359 0.00 NA
Reasoning -0.43 [-0.81, -0.04] 2 0.05 0.831 0.00 NA
Executive function 0.00 [-0.32, 0.33] 17 216.98 < 0.001*** 92.63 526
General cognition -1.04 [-1.35, -0.74] 57 2567.79 < 0.001*** 97.82 407
*

p < 0.05,

**

p < 0.01,

***

p < 0.001.

g: Hedge’s g, CI: confidence interval, K: number of the effect size, Q: heterogeneity between studies within domain, p (Q): p-value for heterogeneity, I2: percentage of heterogeneity caused by study differences, FSN: fail safe N, NA: not applicable

Table 5.

Effect sizes in communicative factors

Subdomain g 95% CI K Q p (Q) I2 (%) FSN
Comprehension -0.75 [-1.20, -0.29] 28 795.45 < 0.001*** 96.61 163
Syntax -0.11 [-0.85, 0.64] 2 4.26 < 0.05* 76.52 NA
Word fluency -0.56 [-0.75, -0.36] 37 507.75 < 0.001*** 92.91 177
Naming -0.39 [-0.71, -0.08] 12 120.05 < 0.001*** 90.84 512
Word definition 0.47 [-0.11, 1.05] 2 6.76 < 0.001*** 85.20 NA
Reading -0.19 [-0.70, 0.32] 6 51.88 < 0.001*** 90.36 364
Writing -0.71 [-1.04, -0.37] 7 15.04 < 0.05* 60.10 478
Figurative language -0.31 [-1.70, 1.07] 3 47.14 < 0.001*** 95.76 129
General language -0.64 [-1.24, 0.04] 6 54.26 < 0.001*** 90.79 649
*

p < 0.05,

**

p < 0.01,

***

p < 0.001.

g: Hedge’s g, CI: confidence interval, K: number of the effect size, Q: heterogeneity between studies within domain, p (Q): p-value for heterogeneity, I2: percentage of heterogeneity caused by study differences, FSN: fail safe N, NA: not applicable

Table 6.

Effect sizes in other factors

Subdomain g 95% CI K Q p (Q) I2 (%) FSN
Age 0.58 [0.26, 0.59] 52 1,313.78 < 0.001*** 96.12 248
Education -0.29 [-0.40, -0.17] 48 546.35 < 0.001*** 91.40 95
Depression 0.22 [0.10, 0.34] 16 52.36 < 0.001*** 71.35 349
*

p < 0.05,

**

p < 0.01,

***

p < 0.001.

g: Hedge’s g, CI: confidence interval, K: number of the effect size, Q: heterogeneity between studies within domain, p (Q): p-value for heterogeneity, I2: percentage of heterogeneity caused by study differences, FSN: fail safe N

Study MCI group
Control group
Predictor
n Age Education n Age Education
Aguilar-Navarro et al. (2017) 49 - 10.5 (6.1) 76 - 8.0 (5.7) VP/memory/GC, word fluency, education
Tokuchi et al. (2014) 39 75.8 (12.5) 12.1 (-) 29 77.2 (9.8) 11.3 (-) GC, Age/education
Sachdev et al. (2013) 157 78.5 (4.5) 11.2 (3.5) 66 78.6 (4.5) 12.3 (3.9) Memory/GC, age/education/depression
Kang et al. (2014) 98 74.7 (5.4) 5.4 (4.9) 87 71.6 (5.0) 8.4 (4.8) GC, age/education/depression
Egli et al. (2015) 23 66.7 (9.1) 13.2 (2.8) 13 74.7 (6.0) 14.3 (4.1) GC, age/education
Sheardova et al. (2014) 30 71.9 (9.2) 14.8 (3.4) 42 71.6 (5.0) 15.8 (2.6) Attention/memory/WM/EF, word fluency/naming, age/education
Pandya et al. (2017) 421 75.2 (8.7) 15.5 (3.1) 175 71.5 (9.1) 15.8 (3.1) GC, age/education
Dodge et al. (2015) 14 83.4 (8.8) 14.0 (2.6) 27 78.9 (5.5) 16.6 (2.4) Attention/memory/GC/EF, word fluency, age/education
Albert et al. (2018) 224 56.9 (8.4) 17.1 (2.3) 349 57.3 (10.4) 17.0 (2.4) GC, reading, age/education
Cui et al. (2012) 33 77.9 (4.2) 12.7 (4.0) 153 77.1 (4.4) 12.0 (3.5) GC, age/education
Ellendt et al. (2017) 90 70.7 (7.6) 9.8 (2.0) 95 67.1 (7.2) 10.2 (1.9) Age/education/depression
Stormoen et al. (2014) 21 69.1 (8.8) 12.3 (2.9) 33 69.2 (6.5) 12.4 (3.3) Attention/VP/memory/WM/reasoning/EF/GC, syntax/word fluency/naming/reading, age/education
Thomas et al. (2018) 220 73.6 (7.3) 16.2 (2.8) 305 72.3 (6.7) 16.5 (2.6) Attention/memory/EF, word fluency/naming, age/education
Pereira et al. (2018) 311 72.1 (7.3) 16.1 (2.8) 122 74.8 (7.6) 16 (2.6) Age/education
Alves et al., (2018) 22 65.5 (8.3) 9.5 (5.0) 44 66.4 (7.7) 9.1 (4.8) Attention/calculation/VP/orientation/memory/WM, word fluency/FL, age/education
Shimada et al. (2019) 24,649 73.0 (5.6) 11.0 (2.5) 12,897 69.6 (3.6) 11.7 (2.4) GC, age/education/depression
Kim et al. (2019) 142 70.0 (7.00) 11.5 (5.0) 31 74.5 (6.2) 11.9 (4.6) Calculation/VP/memory/WM/EF/GC, naming, word fluency, age/education/depression
Park et al. (2017) 44 68.3 (8.0) 7.1 (4.7) 23 64.7 (7.8) 7.4 (4.4) Calculation/VP/orientation/memory/WM/E/GC, age/education/depression
Reppermund et al. (2013) 227 78.6 (4.4) 11.6 (3.6) 375 77.9 (4.6) 11.8 (3.5) GC, age/education/depression
Rajtar-Zembaty et al. (2019) 147 71.3 (6.0) 12.1 (3.2) 653 68.9 (4.9) 15.1 (3.1) Attention/VP/memory/GC, word fluency/GL, age/education/depression
Mohammadi et al. (2018) 30 70.1 (1.6) 12.0 (2.3) 30 69.9 (1.4) 13.1 (2.7) Memory/GC, age/education/depression
Juncos-Rabad´an et al. (2013) 32 71.2 (8.5) 10.6 (4.7) 106 63.9 (8.5) 10.3 (4.7) Attention/calculation/memory/EF/GC, word fluency, age/education/depression
Vaughan et al. (2018) 308 71.4 (8.0) - 302 71.7 (8.1) - GC, word fluency, age
Gifford et al. (2015) 1,989 74.0 (8.0) 15.1 (3.5) 467 73.0 (8.0) 15.1 (3.4) Attention/memory/WM/EF/GC, word fluency/naming, age/education
Brodaty et al. (2014) 133 76.0 (7.0) - 52 75.3 (6.7) - GC, age/depression,
Ansai et al. (2017) 40 75.8 (6.3) 4.9 (4.0) 40 73.5 (6.2) 6.3 (4.3) GC, age/education/depression
Xie et al. (2011) 187 80.0 (6.0) - 124 80.0 (7.2) - GC, age
O’caoimh et al. (2012) 154 73.6 (-) 12.2 (-) 630 67.4 (-) 13.8 (-) GC, age/education
Ng et al. (2015) 62 70.8 (6.8) 4.6 (4.4) 176 69.5 (7.4) 5.1 (3.8) GC, age/education
Puente et al. (2014) 20 - - 31 - - Attention/VP/memory/EF/GC, GL
Yokoi et al. (2019) 133 72.6 (6.1) 13.3 (2.7) 101 73.1 (5.6) 12.6 (2.9) GC, age/education/depression
Zhou et al. (2016) 14 69.1 (10.6) - 16 64.1 (9.1) - VP/orientation/memory/GC, comprehension/naming/writing/GL, age
Johns et al. (2018) 13 77.7 (6.8) 15.2 (3.4) 13 75.8 (6.4) 14.3 (3.1) GC, age/education
Fausto et al. (2018) 51 74.5 (6.6) 15.5 (2.8) 46 71.4 (6.1) 15.3 (2.3) Comprehension, age/education
Ding et al. (2019) 372 74.8 (7.3) 15.6 (3.0) 219 75.9 (5.1) 16.1 (2.8) GC, age/education
Zhang et al. (2018) 26 81.2 (5.5) - 27 80.8 (6.8) - GC, age
Andrejeva et al. (2016) 288 70.0 (7.9) - 65 66.6 (8.6) - GC, age
Cardoso et al. (2014) 52 71.8 (8.2) 10.4 (3.6) 31 70.1 (7.4) 11.4 (2.4) Memory/ WM/GC, comprehension/writing/FL, age/education
Kirchberg et al. (2012) 25 76.0 (6.1) 14.8 (3.6) 70 75.3 (8.7) 15.1 (3.3) GC, age/education
Facal et al. (2014) 44 70.3 (9.5) 9.5 (3.8) 58 67.4 (9.1) 10.2 (5.1) Attention/memory/GC, reading/GL, age/education
T´oth et al. (2018) 48 73.1 (8.0) 11.8 (3.3) 36 64.1 (7.1) 12.5 (3.2) VP/GC, age/education
Kim et al. (2015) 68 73.2 (8.9) 12.6 (4.3) 93 72.8 (8.5) 11.5 (3.4) Word definition, age/education
Weakley et al. (2013) 49 72.1 (8.8) 15.0 (3.0) 90 71.4 (9.4) 15.7 (2.7) GC, word fluency/naming, age/education
Fors et al. (2018) 31 70.1 (5.6) 14.1 (3.6) 36 67.9 (7.2) 13.2 (3.4) GC, syntax, age/education
Peter et al. (2016) 20 71.8 (4.9) 14.7 (2.9) 30 69.8 (4.9) 15.1 (3.4) GC, naming, age/education/depression
Tsantali et al. (2013) 28 - - 38 - - Comprehension/naming/reading/writing
Polcher et al. (2017) 29 68.2 (9.0) 14.1 (2.9) 28 67.5 (8.1) 14.5 (3.1) Attention/VP/organization/memory/GC, GL, age/education/depression
Hu¨bner et al. (2018) 9 70.6 (5.9) - 34 71.0 (6.3) - Memory/GC, naming, age
Kowoll et al. (2015) 14 77.5 (12.4) - 6 70.2 (8.2) - VP/GC, age/education/depression
Manouilidou et al. (2016) 23 68.6 (-) 12.4 (3.1) 21 67.8 (-) 14.1 (2.2) WM/EF/GC, comprehension/word fluency, age/education
Payne & Stine-Morrow (2016) 64 77.6 (2.0) 4.9 (0.7) 281 71.0 (0.8) 16.1 (0.3) Memory/GC, comprehension, age
Price et al. (2012) 33 77.6 (7.2) 12.2 (3.1) 33 75.5 (6.2) 13.5 (2.6) Memory/GC, word fluency/naming/reading, age/education
Facal et al. (2016) 15 72.5 (8.5) 9.9 (3.8) 41 65.4 (9.3) 9.8 (4.7) Attention/memory/EF/GC, age/education
Torres et al. (2019) 81 72.8 (7.6) 14.6 (3.0) 50 70.5 (6.0) 15.8 (3.1) GC, comprehension, age/education
Vita et al. (2014) 60 71.3 (6.4) 10.3 (5.0) 20 69.5 (3.7) 10.0 (3.9) GC, word fluency, age/education
Rinehardt et al. (2014) 34 74.7 (6.0) 13.8 (2.2) 34 75.2 (5.3) 13.7 (1.8) GC, word fluency, age/education
Arsenault-Lapierre et al. (2012) 60 75.0 (7.0) 11.2 (3.1) 33 74.7 (6.0) 13.8 (2.9) Attention/memory/EF/GC, word fluency/naming/reading, age/education/depression

VP: visuoperception, WM: working memory, EF: executive function, GC: general cognition, FL: figurative language, GL: general language

Study
Aguilar-Navarro et al.(2017) Clinical and demographic predictors of conversion to dementia in Mexican elderly with mild cognitive impairment. Revista de Investigacion Clinica, 69(1), 33-39.
Tokuchi et al.(2014) Clinical and demographic predictors of mild cognitive impairment for converting to Alzheimer’s disease and reverting to normal cognition. Journal of the Neurological Sciences, 346(1-2), 288–292.
Sachdev et al.(2013) Factors predicting reversion from mild cognitive impairment to normal cognitive functioning: A population-based study. PloS one, 8(3), e59649.
Kang et al.(2014) Factors associated with cognition recovery among elders with mild cognitive impairment in Korea. International Nursing Review, 61(3), 318-326.
Egli et al.(2015) Varying strength of cognitive markers and biomarkers to predict conversion and cognitive decline in an early-stage-enriched mild cognitive impairment sample. Journal of Alzheimer’s Disease, 44(2), 625–633.
Sheardova et al.(2014) Famous landmark identification in amnestic mild cognitive impairment and Alzheimer’s disease. PloS one, 9(8), e105623.
Pandya et al.(2017) Predictors of reversion from mild cognitive impairment to normal cognition. Dementia and Geriatric Cognitive Disorders, 43(3-4), 204-214.
Dodge et al.(2015) Social markers of mild cognitive impairment: Proportion of word counts in free conversational speech. Current Alzheimer Research, 12(6), 513-519.
Albert et al.(2018) Predicting progression from normal cognition to mild cognitive impairment for individuals at 5 years. Brain, 141(3), 877–887.
Cui et al.(2012) Predicting the development of mild cognitive impairment: A new use of pattern recognition. NeuroImage, 60(2), 894–901.
Ellendt et al.(2017) Predicting stability of mild cognitive impairment (MCI): Findings of a community based sample. Current Alzheimer Research, 14(6), 608-619.
Stormoen et al.(2014) Cognitive predictors of medical decision-making capacity in mild cognitive impairment and Alzheimer’s disease. International Journal of Geriatric Psychiatry, 29(12), 1304-1311.
Thomas et al.(2018) Word-list intrusion errors predict progression to mild cognitive impairment. Neuropsychology, 32(2), 235–245.
Pereira et al.(2018) Neuropsychological predictors of conversion from mild cognitive impairment to Alzheimer’s disease: A feature selection ensemble combining stability and predictability. BMC Medical Informatics and Decision Making, 18(1), 137.
Alves et al.,(2018) Neuropsychological predictors of long-term (10 years) mild cognitive impairment stability. Journal of Alzheimer’s Disease, 62(4), 1703–1711.
Shimada et al.(2019) Reversible predictors of reversion from mild cognitive impairment to normal cognition: A 4-year longitudinal study. Alzheimer’s Research and Therapy, 11(1), 24.
Kim et al.(2019) Data-driven prognostic features of cognitive trajectories in patients with amnestic mild cognitive impairments. Alzheimer’s Research and Therapy, 11(1), 10.
Park et al.(2017) Memory performance on the story recall test and prediction of cognitive dysfunction progression in mild cognitive impairment and Alzheimer’s dementia. Geriatrics and Gerontology International, 17(10), 1603-1609.
Reppermund et al.(2013) Impairment in instrumental activities of daily living with high cognitive demand is an early marker of mild cognitive impairment: The Sydney Memory and Ageing Study. Psychological Medicine, 43(11), 2437–2445.
Rajtar-Zembaty et al.(2019) Global cognitive functioning and physical mobility in older adults with and without mild cognitive impairment: evidence and implications. Folia Medica Cracoviensia, 59(1), 75–88.
Mohammadi et al.(2018) Using virtual reality to distinguish subjects with multiple- but not single-domain amnestic mild cognitive impairment from normal elderly subjects. Psychogeriatrics, 18(2), 132-142.
Juncos-Rabad´an et al. (2013) Does tip-of-the-tongue for proper names discriminate amnestic mild cognitive impairment? International Psychogeriatrics, 25(4), 627-634.
Vaughan et al.(2018) Semantic and phonemic verbal fluency discrepancy in mild cognitive impairment: Potential predictor of progression to Alzheimer’s disease. Journal of the American Geriatrics Society, 66(4), 755-759.
Gifford et al.(2015) Inclusion of an informant yields strong associations between cognitive complaint and longitudinal cognitive outcomes in non-demented elders. Journal of Alzheimer’s Disease, 43(1), 121–132.
Brodaty et al.(2014) Progression from mild cognitive impairment to dementia: A 3-year longitudinal study. Australian and New Zealand Journal of Psychiatry, 48(12), 1137–1142.
Ansai et al.(2017) Cognitive correlates of timed up and go subtasks in older people with preserved cognition, mild cognitive impairment, and Alzheimer’s disease. American Journal of Physical Medicine and Rehabilitation, 96(10), 700-705.
Xie et al.(2011) Identifying and characterizing trajectories of cognitive change in older persons with mild cognitive impairment. Dementia and Geriatric Cognitive Disorders, 31(2), 165–172.
O’caoimh et al.(2012) Comparison of the quick mild cognitive impairment (Qmci) screen and the SMMSE in screening for mild cognitive impairment. Age and Ageing, 41(5), 624–629.
Ng et al.(2015) Montreal Cognitive Assessment for screening mild cognitive impairment: Variations in test performance and scores by education in Singapore. Dementia and Geriatric Cognitive Disorders, 39(3-4), 176-185.
Puente et al.(2014) Functional impairment in mild cognitive impairment evidenced using performance-based measurement. Journal of Geriatric Psychiatry and Neurology, 27(4), 253-258.
Yokoi et al.(2019) Discrete effect of each mild behavioural impairment category on dementia conversion or cognitive decline in patients with mild cognitive impairment. Psychogeriatrics, 19(6), 591-600.
Zhou et al. (2016) Characteristics of agraphia in Chinese patients with Alzheimer’s disease and amnestic mild cognitive impairment. Chinese Medical Journal, 129(13), 1553–1557.
Johns et al.(2018) Cognitive modeling as an interface between brain and behavior: Measuring the semantic decline in mild cognitive impairment. Canadian Journal of Experimental Psychology/Revue Canadienne de Psychologie Exp´erimentale, 72(2), 117–126.
Fausto et al.(2018) Comparison of subjective and objective measures of hearing, auditory processing, and cognition among older adults with and without mild cognitive impairment. Journal of Speech, Language, and Hearing Research, 61(4), 945-956.
Ding et al.(2019) Evaluating trajectories of episodic memory in normal cognition and mild cognitive impairment: Results from ADNI. PLoS One, 14(2), e0212435.
Zhang et al.(2018) Evaluation of MESSAGE communication strategy combined with group reminiscence therapy on elders with mild cognitive impairment in long-term care facilities. International Journal of Geriatric Psychiatry, 33(4), 613-622.
Andrejeva et al.(2016) Neurocognitive deficits and effects of cognitive reserve in mild cognitive impairment. Dementia and Geriatric Cognitive Disorders, 41(3-4), 199-209.
Cardoso et al.(2014) Non-literal language deficits in mild cognitive impairment. Psychogeriatrics, 14(4), 222-228.
Kirchberg et al.(2012) Semantic distance abnormalities in mild cognitive impairment: Their nature and relationship to function. The American Journal of Psychiatry, 169(12), 1275-1283.
Facal et al.(2014) Working memory span in mild cognitive impairment. Influence of processing speed and cognitive reserve. International Psychogeriatrics, 26(4), 615-625.
T´oth et al.(2018) A speech recognition-based solution for the automatic detection of mild cognitive impairment from spontaneous speech. Current Alzheimer Research, 15(2), 130-138.
Kim et al.(2015) Abstract word definition in patients with amnestic mild cognitive impairment. Behavioural Neurology, 2015, 580246.
Weakley et al.(2013) Analysis of verbal fluency ability in amnestic and non-amnestic mild cognitive impairment. Archives of Clinical Neuropsychology, 28(7), 721–731.
Fors et al.(2018) Automated syntactic analysis of language abilities in persons with mild and subjective cognitive impairment. Studies in Health Technology and Informatics, 247, 705-709.
Peter et al.(2016) Category and design fluency in mild cognitive impairment: Performance, strategy use, and neural correlates. Neuropsychologia, 93(pt A), 21-29.
Tsantali et al.(2013) Could language deficits really differentiate mild cognitive impairment (MCI) from mild Alzheimer’s disease? Archives of Gerontology and Geriatrics, 57(3), 263-270.
Polcher et al.(2017) Face-name associative recognition deficits in subjective cognitive decline and mild cognitive impairment. Journal of Alzheimer’s Disease, 56(3), 1185-1196.
Hu¨bner et al.(2018) Naming and verbal learning in adults with Alzheimer’s disease, mild cognitive impairment and in healthy aging, with low educational levels. Arquivos de Neuro-Psiquiatria, 76(2), 93-99.
Kowoll et al.(2015) Neuropsychological profiles and verbal abilities in lifelong bilinguals with mild cognitive impairment and Alzheimer’s disease. Journal of Alzheimer’s Disease, 45(4), 1257-1268.
Manouilidou et al.(2016) Processing complex pseudo-words in mild cognitive impairment: The interaction of preserved morphological rule knowledge with compromised cognitive ability. Clinical Linguistics and Phonetics, 30(1), 49–67.
Payne & Stine-Morrow (2016) Risk for mild cognitive impairment is associated with semantic integration deficits in sentence processing and memory. The journals of gerontology. Series B, Psychological Sciences and Social Sciences, 71(2), 243–253.
Price et al.(2012) Semantic verbal fluency strategies in amnestic mild cognitive impairment. Neuropsychology, 26(4), 490–497.
Facal et al.(2016) Temporal changes in semantic and lexical access related to mild cognitive impairment. Aging Clinical and Experimental Research, 28(3), 497–504.
Torres et al.(2019) Types of errors on a semantic interference task in mild cognitive impairment and dementia. Neuropsychology, 33(5), 670–684.
Vita et al.(2014) Typicality of words produced on a semantic fluency task in amnesic mild cognitive impairment: Linguistic analysis and risk of conversion to dementia. Journal of Alzheimer’s Disease, 42(4), 1171-1178.
Rinehardt et al.(2014) Verbal fluency patterns in mild cognitive impairment and Alzheimer’s disease. Dementia and Geriatric Cognitive Disorders, 38(1-2), 1-9.
Arsenault-Lapierre et al. (2012) Word reading threshold and mild cognitive impairment: A validation study. BMC Geriatrics, 12, 38.

Appendix 2 문헌의 순서는 Appendix 1 목록의 순서대로 기재되었다.