Use of a modified hidden pathway maze test in patients with Alzheimer’s Disease dementia

By Pradeep Nathan | Jul 22, 2019


Alzheimer’s Association International Conference (AAIC) 2019
14 – 18 July 2019
Los Angeles, U.S.
Alzheimer’s Association

Conference Website


Poster Abstracts


Chris J Edgar1, Tim Tasker2, Adrian Schembri1, Babli Millais2, Bharat Ruparelia2, Odile Dewit2, Stephen Jones2, Paul Maruff1, Pradeep J. Nathan2

1.  Cogstate, New Haven, CT, USA
2.  Sosei Heptares, Cambridge, UK

Background: Maze tests have a long history as measures of visuospatial learning and executive function.  The importance of impairments in both visuospatial learning and executive dysfunction has been increasingly recognized in early AD. In contrast to presented mazes, hidden pathway mazes use underlying rule sets to create decision points meaning greater volumes of data are obtained. Computerized versions (e.g. the Groton Maze Learning Test [GMLT]) provide greater analytical scope than manual tests. However, use of extensive rule sets may prove challenging in AD dementia. Here we present basic psychometric analyses for a simplified GMLT in mild-moderate AD dementia.

Methods: The modified GMLT uses a 28-step pathway hidden in a 10x10 tile grid. The pathway is found by selecting one tile at a time revealing a green checkmark which remains visible if correct. A more limited set of instructions is given compared to the usual test. Five trials are given, with the number of errors recorded. A single, reverse recall trial is also given after a short delay. Data were collected in a PIb clinical trial in AD patients with MMSE 12-24 (NCT03456349). Analyses were performed for missingness, presence of floor and ceiling effects, test retest reliability and learning/practice effects.

Results: Data were collected for 59 patients at the baseline assessment (mean age 73.3 [SD 7.58]; 72.9% female). No ceiling effect was present in the population. Whilst there is no theoretical floor (maximum number of errors), several patients made greater numbers of errors (>20 per trial) suggesting a proportion of the trial population found the test challenging. Data will be reported exploring the impact of cognitive impairment on test length, completion and reliability.

Conclusions: Hidden pathway maze learning may be a value probe of short-term memory and executive function in AD. Whilst the removal of complex rule sets may make the test more feasible in the dementia stages of the disease, increasing severity of impairment may limit utility for some patients.

View Poster



Pradeep J. Nathan PhD 1,4,5, Geor Bakker PhD1,2,3, Alex Godwood MSc1, Claudia Vingerhoets PhD2,3, Jan Booij PhD MD2, Oswald Bloemen PhD MD3,6, Matthan W. Caan PhD2, and Therese van Amelsvoort PhD MD3

1.  Sosei Heptares, Cambridge, United Kingdom
2.  Department of Radiology and Nuclear Medicine, Amsterdam UMC, University of Amsterdam, the Netherlands
3.  Department of Psychiatry and Neuropsychology, Maastricht University, the Netherlands
4.  Brain Mapping Unit, Department of Psychiatry, University of Cambridge, Cambridge, United Kingdom
5.  School of Psychological Sciences, Monash University, Melbourne, Australia
6.  GGZ Centraal, Center for Mental Health Care Innova, the Netherlands

Objectives: The basal forebrain cholinergic neurons provide a major source of cholinergic innervation to the cortex and hippocampus and play a critical role in modulating learning and memory, in part through activation of M1 receptors. Smaller basal forebrain nuclei (BFN) volumes have been shown to be associated with faster rates of cognitive deterioration in Alzheimer’s disease (AD). The relationship between BFN volume and M1 receptor sensitivity is unknown and may underlie greater memory impairments. The objective of this study was to assess whether BFN volumes can predict greater cognitive sensitivity to muscarinic M1 receptor antagonism by biperiden.

Methods: BFN volumes were quantified from T1 weighted 3TMRI scans. A total of 59 subjects (aged 18-40) were scanned. Cognition was assessed twice for all subjects using the Cambridge neuropsychological test automated battery, once under placebo, and once after the administration of 4mg of the M1 selective antagonist biperiden.

Results: Regression analysis showed no significant predictive relationship between the nucleus basalis volumes (CH4) and biperiden induced cognitive impairments. Medial-septal nuclei and the vertical and diagonal band of Broca nuclei (CH1,2,3) that innervate the hippocampal formation, however, did significantly predict biperiden induced impairment on delayed recall of verbal memory, with smaller CH1,2,3 volumes predicting greater impairment on delayed word recall (F=4.81, p=0.032).

Conclusion: M1 receptor modulation of memory may depend on the integrity of basal forebrain cholinergic neurons, The CH1,2,3 BFN volume may be a potential biomarker predictive of superior cognitive efficacy of drugs targeting the cholinergic system including M1 receptor agonists.

View Poster