Marc Vandemeulebroecke (1), Johanna Mielke (1,2), Tillmann Krahnke (1), Peter Quarg (1), Andreas U. Monsch (3)
(1) Novartis Pharma AG, Novartis Campus, 4002 Basel, Switzerland, (2) Dept. of Statistics, TU Dortmund University, Vogelpothsweg 87, 44227 Dortmund, Germany, (3) Memory Clinic, University Center for Medicine of Aging, Felix Platter-Hospital, Schanzenstrasse 55, 4031 Basel, Switzerland
Objectives: The goal was to investigate which cognitive domains carry most information in the earliest stages of cognitive decline. This may be helpful in the development of disease-modifying drugs against Alzheimer’s disease, which must intervene before the onset of clinical symptoms.
Methods: Following the idea of Ueckert et al. [1], Item Response Theory was applied to 14 variables of the CERAD-NAB, its additions (phonemic fluency; Trail Making Test, parts A and B), and the CVLT, based on data from 1,101 cognitively healthy elderly subjects and two subjects with probable AD (39% women; mean age ± sd = 69.9 ± 8.28 years) of the BASEL study. The model was implemented in a Bayesian framework with non-informative priors. The estimated discrimination parameters and Fisher information were used to assess the information content of the different component tests.
Results: ‘CERAD-Word List Learning’ and ‘CVLT-Word List Learning’ as well as ‘CERAD-Word List Delayed Recall’ and ‘CVLT-Word List Long Delay Free Recall’ carried most information in the BASEL sample (17.4%, 16.2%, 15.1% and 11.5%, respectively, of the total amount of information). CVLT variables did not carry more information than corresponding variables of the CERAD-NAB. ‘Trail Making Test (Part B)’, ‘CVLT-Word List Recognition Discriminability’, ‘CERAD-Word List Recognition Discriminability’, ‘Mini Mental Status Examination (MMSE)’ and ‘Semantic Fluency’ carried little information overall, with the word list recognition tasks and the MMSE being informative in the low ability range (only). ‘Phonemic Fluency’, ‘CERAD-Figures-Delayed Recall’, Trail Making Test (Part A), ‘Boston Naming Test’, and ‘CERAD-Figures-Copy’ carried almost no information.
Conclusions: By application of Item Response Theory in the Basel Study on the Elderly, it was possible to quantify the information contribution of individual test items to the CERAD-NAB and CVLT neuropsychological test batteries in this study. Verbal episodic memory as assessed by word list learning and recall tasks carried most information. Hence, this may be a promising domain for the detection of earliest signs of cognitive decline in initially healthy subjects.
References:
[1] Ueckert S, Plan EL, Ito K, Karlsson MO, Corrigan B, Hooker AC (2013). Benefits of an Item Response Theory Based Analysis of ADAS-Cog Assessments. American Conference on Pharmacometrics.
Reference: PAGE 23 (2014) Abstr 3030 [www.page-meeting.org/?abstract=3030]
Poster: Drug/Disease modeling - CNS