Premium
Optimizing the memory binding test for detection of aMCI and dementia
Author(s) -
Mowrey Wenzhu,
Lipton Richard B.,
Katz Mindy J.,
Ramratan Wendy S.,
Loewenstein David A.,
Buschke Herman
Publication year - 2020
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1002/alz.044596
Subject(s) - dementia , receiver operating characteristic , cognitive impairment , audiology , psychology , discriminative model , test (biology) , recall , cognition , medicine , cognitive psychology , psychiatry , artificial intelligence , disease , computer science , paleontology , biology
Background The Memory Binding Test (MBT) Version 1 demonstrated good discriminative validity for distinguishing persons with dementia and amnestic mild cognitive impairment (aMCI) from cognitively normal elder controls (CN). Version 2 of the MBT test was developed to improve brevity, which is essential for screening. We aimed to compare these two versions in terms of cross‐sectional discriminative validity to distinguish A. aMCI vs. CN, B. aMCI and dementia vs. CN, and C. dementia vs. CN and aMCI. Method Version 1 and Version 2 (for differences see Table 1) of the MBT were administered to independent and systematically recruited samples between May 2003 and December 2007 (Age: 70+; Version 1: 20 dementia cases, 31 aMCI, 246 CN; Version 2: 13 dementia cases, 29 aMCI, 236 CN). Scores on the MBT indices were compared between the versions. The partial area under the receiver operating characteristic curve (ROC AUC) for specificities ≥ 0.70 was compared between indices and versions (Table 1). Specificities were compared between versions when the sensitivity values were comparable. Result The MBT indices were not significantly different between versions for the diagnosis groups of aMCI and dementia. In CN, Version 2 yielded higher scores on number of items Cued Recall for List 1 (CR‐L1), number of Pairs In the Paired condition (PIP), and number of Total Items recalled in the Paired condition (TIP) (14.8±1.5 vs. 14.4±1.7, 11.0±3.2 vs. 9.9±3.8, 26.3±4.0 vs. 24.9±5.0, respectively, p=0.005) and PIP and TIP remained higher (p=0.02) when adjusting for potential confounders of age, gender, education and global cognitive function. Partial AUC comparison shows that TIP was the optimal index for Version 1 and TIP partial AUCs were not significantly different between two versions (Figure 1, Table 2). The specificities were higher in Version 1 vs. Version 2 in Comparisons A and B, while the specificities were comparable in Comparison C (Table 3). Conclusion MBT Version 1 is better at distinguishing aMCI, or aMCI and dementia combined, from cognitively normal elderly. To distinguish dementia vs. cognitively normal elder controls and aMCI, Version 1 and Version 2 are comparable. We recommend using Version 1 for future studies.