The 4 Mountains Test: A Short Test of Spatial Memory with High Sensitivity for the Diagnosis of Pre-dementia Alzheimer’s Disease

The study was performed in accordance with the Declaration of Helsinki. All participants gave written, informed consent. Ethics approval was obtained from the UK Research Ethics Committee South East Coast and from Brighton and Sussex University Hospitals NHS Trust (references 10/H1107/23 and 13/LO/0277, respectively).

21 patients with MCI were recruited form the Cognitive Disorders Clinic at Hurstwood Park Neurological Centre, Haywards Heath, West Sussex, UK. MCI was diagnosed according to internationally recognized criteria¹, which specify I) a subjective report of cognitive decline, corroborated by an informant II) objective evidence of cognitive impairment on formal testing III) absence of dementia and IV) preserved activities of daily living and functional independence.

Objective cognitive testing was undertaken using either the Addenbrooke's Cognitive Examination-Revised²⁴ or the Queen Square Screening Test for Cognitive Deficits (EK Warrington 2003) in combination with the Mini Mental State Examination (MMSE)⁷. As part of the clinical diagnostic workup, patients underwent clinical and laboratory assessments to exclude potentially treatable causes of cognitive decline, such as vitamin B12 deficiency or thyroid dysfunction. The presence of significant cerebrovascular disease was a core exclusion criterion, as evidenced by significant vascular lesion load on imaging (the presence of cortical infarcts, extensive and/or confluent White Matter Hyperintensities (WMH) and WMH >10mm diameter), and/or a Hachinski Ischaemic Score >4²⁵. Patient data were compared with that from age-matched healthy controls (HC) without a history of cognitive impairment and with 11 patients with mild AD-related dementia, diagnosed according to the McKhann criteria²⁶.

The MCI patient group was split into MCI biomarker positive (MCI+ve) and MCI biomarker negative (MCI-ve) subgroups on the basis of testing for CSF biomarker evidence of underlying AD pathology, i.e., CSF β-amyloid_1-42 and tau levels. Biomarker positive/negative status was determined using updated cut-off scores²⁷. The detection of positive CSF biomarkers in MCI patients (i.e., the MCI+ve subgroup) would fulfil diagnostic criteria for predementia AD, termed variously as prodromal AD² or MCI due to AD³. Two MCI patients did not undergo CSF testing.

All subjects were tested on a battery of neuropsychological tests, which included testing of the following cognitive domains: premorbid IQ (National Adult Reading Test, Nelson and Willison 1984)²⁸, episodic memory (Rey Auditory Verbal Learning Test, RAVLT, Rey 1941)²⁹, attention and executive function (Trail Making Test A and B, Reitan 1958)³⁰, executive function (lexical and semantic fluency, Benton et al. 1994)³¹, working memory (Digit Span, Blackburn and Benton 1957)³² and higher visual processing (Object Decision Test from the Visual Object and Space Perception test)³³

MRI scanning was undertaken on a 1.5T scanner based at the Clinical Imaging Sciences Centre, Brighton and Sussex Medical School, UK. T1-weighted 3D volumetric MRI data were acquired using a magnetization-prepared rapid-acquisition gradient-echo sequence, with 1 x 1 x 1mm³ voxel size, TI = 600 msec, TE = 4 msec, TR = 1160 msec. 2 AD patients and 4 MCI patients were unable to undergo MRI scanning. Structural correlations were reported for the remaining participants.

Cortical thickness was measured using the open source FreeSurfer package (Massachusetts General Hospital, Harvard University, Boston MA, USA), which, as detailed elsewhere³⁴, involves iterative reconstruction of the white-gray matter interface and pial surface, and subsequent labelling with non-linear morphing to a probabilistic brain atlas. The Desikan probabilistic brain atlas was used³⁵, with the posterior cingulate gyrus and precuneus selected as regions of interest (ROIs), reflecting their putative role in spatial cognition and their early involvement in AD^{36, 37}.

Total hippocampal volumes were measured using the FSL/FIRST tool (FMRIB, Oxford Centre for Functional Magnetic Resonance Imaging of the Brain, Oxford, UK)³⁸. Correlations were not determined for other brain regions, reflecting the study hypothesis. In particular, correlations with frontal brain regions were not calculated since 4MT performance is not impaired in patients with frontotemporal dementia^{18, 19}.

All study groups (MCI, AD, HC), and within the MCI biomarker subgroups, were matched in terms of demographics (age, gender, years of education) (Table 1).

Table 1. Demographics of Participants. Data presented as mean (standard deviation) for A) all participants grouped according to cognitive status (HC = Healthy Control, MCI = Mild Cognitive Impairment, AD = Alzheimer's Disease B) MCI patients grouped according to CSF AD biomarker status. Reproduced, with permission, from Moodley et al. (2015)²⁰.

General neuropsychometric assessment
MCI patients were impaired on tests of episodic memory (RAVLT; delayed recall and recognition memory) and executive function (Trail Making Test A and B). By comparison, and consistent with their diagnostic classification, patients with AD-related dementia were impaired in all cognitive domains (Table 2).

Table 2. Neuropsychometric Data. Neuropsychometric data for all participants, presented as raw scores, in keeping with the UK clinical practice for reporting of neuropsychometric data, described as mean (standard deviation). NART = National Adult Reading Test. MMSE = Mini Mental State Examination (not performed in control subjects). VOSP-OD = Visual Object and Space Perception battery. RAVLT-DR = Rey Auditory Verbal Learning Test-Delayed Recall (List A). RAVLT-RP = Rey Auditory Verbal Learning Test-Recognition Performance (List A). Reproduced, with permission, from Moodley et al. (2015)²⁰.

A direct comparison of the MCI subgroups did not reveal any significant differences in the test scores obtained by the MCI-ve and MCI+ve patients, with the exception of the Trail Making Test "B" (Table 3). There was no significant difference in episodic memory between the 2 MCI groups (RAVLT; delayed recall and recognition memory).

Table 3. Neuropsychometric Results for MCI Patients. Neuropsychometric data for MCI patients, grouped according to CSF AD biomarker status (alpha = 0.004, adjusted for multiple comparisons) presented as raw scores, in keeping with the UK clinical practice for reporting of neuropsychometric data, described as mean (standard deviation). Reproduced, with permission, from Moodley et al. (2015)²⁰.

4MT performance
There were significant differences between study groups in terms of performance on the 4MT test (p <0.001, Table 4). After correction for multiple comparisons, pairwise group comparisons revealed significant differences between healthy controls (HC) and MCI+ve groups (p <0.001), HC and AD (p <0.001), MCI-ve and AD (p <0.001) and, crucially, between MCI-ve vs MCI+ve groups (p = 0.002). No significant difference in PM test scores was observed between HC and MCI-ve (p = 0.3) or between MCI+ve and AD groups (p = 0.6). Figure 2 shows individual 4MT scores and the differences in score between study groups.

Table 4. 4MT Scores. 4MT scores (scored out of 15) for all participants (top) and for MCI patients grouped according to CSF AD biomarker status (middle), with pairwise comparisons (bottom). HC = Healthy Controls; MCI = Mild Cognitive Impairment; AD = Alzheimer's Disease. Reproduced, with permission, from Moodley et al. (2015)²⁰.

Figure 2. 4MT Scores for MCI Patients. 4MT scores (scored out of 15) for MCI patients grouped by CSF AD biomarker status. Reproduced, with permission, from Moodley et al. (2015)²⁰.

The ability of 4MT to differentiate between MCI patients with AD pathology (i.e., MCI-ve and MCI+ve is illustrated by the area under the Receiver Operating Characteristics curve (AUC ROC) (Figure 3). Test performance was associated with an AUC of 0.93; PM scores of 8 or below were associated with 100% sensitivity and 78% specificity for differentiating MCI+ve from MCI-ve individuals.

Figure 3. ROC Curve. ROC curve showing discrimination of MCI patients with and without biomarker evidence of AD. Area under ROC curve 0.93.

Correlations between 4MT and quantitative MRI data
Partial correlations were undertaken for patients with MCI and AD-related dementia, corrected for age and total intracranial volume. After averaging between left and right hemispheres, significant associations were found between PM score and hippocampal volume (r = 0.42, p = 0.03, not surviving the corrected alpha threshold of 0.02), and between PM score and cortical thickness of the precuneus (r = 0.55, p = 0.003). No significant correlation between PM score and the cortical thickness of the posterior cingulate gyrus was observed (r = 0.19, p = 0.4). Scatterplots of these correlations are provided in Figure 4.

Figure 4. Scatterplots Demonstrating Correlation With Structural MRI Data. Scatterplots demonstrating correlation between 4MT score and hippocampal volume (top), cortical thickness of the precuneus (middle) and cortical thickness of the posterior cingulate gyrus (bottom) for all MCI and AD patients. Reproduced, with permission, from Moodley et al. (2015)²⁰.

Testing of 4MT stability and reliability
Psychometric properties of the 4MT were evaluated in a separate cohort of 41 healthy controls without symptoms of cognitive impairment. Participants were retested 7 and 28 d after initial testing. The effect size between the mean score at baseline and at 7 and 28 d was assessed using the Cohen's d statistic. The modest practice effect observed at 7 d (d = 0.35) was eliminated by 28 d (d = 0), indicating that there was no demonstrable practice effect by the latter interval.

A high degree of reliability was found between 4MT performance at baseline and retest. The average measure intraclass coefficient was 0.808 (95% CI 0.54 -0.918, F23, 23 = 5.96, p <0.01) and 0.641 (95% CI -0.115 – 0.862, F16, 16 = 2.49, p <0.05) at 7 and 28 d, respectively. The mean difference in test score was 0.71 ± 1.52 and 0 ± 2.24 at 7 and 28 d, respectively.

The stability and reliability of the 4MT in patient participants will be assessed in upcoming, larger scale studies.