Introduction: As disease-modifying therapies become available for Alzheimer’s disease (AD), detection of AD in early stages of illness (mild cognitive impairment [MCI], early dementia) becomes increasingly important. Biomarkers for AD in low- and middle-income countries (LMICs) are costly and not widely available; hence, it is important to identify cognitive tests that correlate well with AD biomarker status. In this study, we evaluated the memory alteration test (M@T) to detect biomarker-proven AD and quantify its correlation with neurodegeneration and cerebrospinal fluid (CSF) AD biomarkers in a cohort of participants from Lima, Peru. Methods: This is a secondary analysis of a cohort of 185 participants: 63 controls, 53 with amnestic MCI (aMCI), and 69 with dementia due to AD. Participants underwent testing with M@T and a gold standard neuropsychological battery. We measured total tau (t-tau), phosphorylated tau (p-tau), and beta-amyloid (β-amyloid) in CSF, and evaluated neurodegeneration via medial temporal atrophy score in MRI. We used receiver-operator curves to determine the discriminative capacity of the total M@T score and its subdomains. We used the Pearson coefficient to correlate M@T score and CSF biomarkers. Results: The M@T had an area under the curve (AUC) of 0.994 to discriminate between controls and cognitively impaired (aMCI or AD) patients, and an AUC of 0.98 to differentiate between aMCI and AD patients. Free-recall and cued recall had the highest AUCs of all subdomains. Total score was strongly correlated with t-tau (−0.77) and p-tau (−0.72), and moderately correlated with β-amyloid (0.66). The AUC for discrimination of neurodegeneration was 0.87. Conclusion: The M@T had excellent discrimination of aMCI and dementia due to AD. It was strongly correlated with CSF biomarkers and had good discrimination of neurodegeneration. In LMICs, the M@T may be a cost-effective screening tool for aMCI and dementia caused by AD.

As the global population ages, the number of persons living with Alzheimer’s disease (AD) and related neurodegenerative disorders is expected to increase from 57.4 million cases globally in 2019 to 152.8 million cases in 2050 [1]. The rising disease burden is expected to shift from high-income countries to low- and middle-income countries (LMICs), with the latter accounting for 65% of the global economic burden of AD by 2050 [2]. Among LMICs, Latin American countries are especially vulnerable due to the existing gaps in knowledge and lack of clinical diagnostic and treatment tools and infrastructure [3]. As new disease-modifying therapies targeting the pathophysiology of AD become available in the USA and other high-income countries, providing LMICs with the clinical tools needed to improve detection of persons with mild cognitive impairment (MCI) and early dementia caused by AD is an important step toward the goal of addressing global AD inequities surrounding emerging disease-modifying therapies.

Currently, the amyloid/tau/neurodegeneration (ATN) framework allows for the structured classification of persons with AD in research settings [4]. Persons who demonstrate evidence of cerebral amyloid (A+) and tau (T+) are classified as having AD, regardless of severity of cognitive impairment. These criteria require acquisition of AD biomarkers for classification, including positron emission tomography imaging modalities, or measurement of beta-amyloid (β-amyloid), total tau (t-tau), or phosphorylated tau (p-tau) protein in cerebrospinal fluid (CSF). As AD biomarker testing modalities continue to expand and become increasingly available and affordable in LMICs for research and clinical purposes, it is important to continue to identify cognitive screening tools that are highly sensitive and specific to clinical syndromes that correlate with typical AD neuropathological change, especially during early stages of illness (MCI, early dementia), when affected individuals stand to benefit the most from emerging disease-modifying therapies for AD.

The memory alteration test (M@T) is a brief cognitive test (BCT) designed by Spanish researchers for discrimination of amnestic MCI (aMCI) and early dementia due to AD [5]. As the name suggests, it emphasizes memory domains, including verbal and semantic memory. As such, the M@T parallels AD progression in most persons affected by the disease, who typically experience decreased recent episodic memory early in the disease, followed by impairments in other cognitive domains as the neurodegenerative disease process progresses [6].

As AD biomarker testing is still very limited in LMICs, prioritizing testing patients with a higher likelihood of AD would result in better use of resources. Thus, specific BCTs, like the M@T, could be used for improving sample specificity in these populations. Accordingly, in this study, we examine the performance of the M@T at discriminating between different clinical stages of AD in a cohort of persons with AD and controls who underwent clinical and AD biomarker examinations. We hypothesize that the M@T can help identify AD biomarker-positive cases from biomarker-negative controls.

Study Design and Participants

This is a secondary analysis from a previously reported case-control study of healthy and cognitively impaired participants [7]. Briefly, for this study, we included 63 cognitively healthy participants, 53 participants with aMCI, and 69 participants with dementia due to AD. All cases (MCI and dementia groups) were recruited from the Instituto Peruano de Neurociencias (Peruvian Institute of Neurosciences; IPN) between January 2019 and October 2021. We excluded participants who would be unlikely to engage with testing, including those with sequelae of severe traumatic brain injury, active substance use, uncontrolled epilepsy, and uncontrolled severe psychiatric illness.

We classified participants by group as follows. (1) Controls were clinically healthy volunteers per neuropsychological testing and neurological evaluation, that were recruited from local newspapers, radio and social media, and matched by age and sex; (2) aMCI were participants diagnosed as MCI in accordance with the mild neurocognitive disorder criteria of the Diagnostic and Statistical Manual of Mental Disorders – 5 (DSM-5), that presented with scores more than 1.5 SD from mean compared to norms on one of the tasks within the verbal and/or visual episodic memory domain; and (3) participants with AD were diagnosed according to the published criteria from McKhann et al. [8], and who had positive CSF p-tau consistent with AD pathophysiology.

Procedures

Neurocognitive Assessment

All study participants underwent in-depth cognitive testing with BCSTs, including the M@T and the Rowland Universal Dementia Assessment Scale (RUDAS), and gold standard neuropsychological battery of IPN. The RUDAS and M@T have been previously validated in Peru and have been shown to appropriately discriminate between cognitively healthy individuals and those with aMCI and dementia [9, 10].

The gold standard neuropsychological battery includes the following tests: Rey Auditory Verbal Learning Test (RAVLT), Logical Memory Subtest of the revised Wechler Memory Scale (LMS), Trail Making Tests A and B, Rey Complex Figure (RCF), Boston Naming Test, Wisconsin Card Sorting Test, Letter-Number (subtest of the Weschler Adult Intelligent Scale III), Digit Span, Strub-Black Picture Copying, and the WAIS-III Cubes Test, as has previously been described [11]. For aMCI evaluation, we considered two tests for memory evaluation: the RAVLT for verbal episodic memory and the RCF for visual episodic memory.

CSF Biomarkers

All participants underwent lumbar puncture. b-amyloid, p-tau, and t-tau were measured using INNOTEST (Fujirebio) enzyme-linked immunosorbent assay kits (cut-off values: Aβ42 < 514 pg/mL, t-tau >180.9 pg/mL, and p-tau >42.4 pg/mL). Only p-tau was used as an inclusion criterion for AD patients.

Neurodegeneration

All participants were evaluated by brain MRI. Imaging took place at the Diagnostico Por Imagenes Imaging Center in Lima, Peru, using an existing protocol. Brain images were acquired from a 3-Tesla Siemens Skyra MR System. Study participants were scanned with a standardized MRI protocol containing volumetric T1-weighted magnetization-prepared rapid gradient echo and fluid attenuation inversion recovery sequences. Neurodegeneration was determined by a medial temporal atrophy (MTA) score of 1.0 for persons under 65, 1.5 for persons between 66 and 74 years of age, and 2 for those 75 years of age or older, following previously described protocols [12, 13].

Statistical Analysis

We first describe the demographic and cognitive characteristics of the population using measures of central tendency and dispersion for continuous variables and frequencies for categorical ones. We compare these among the control, aMCI, and AD patient groups using the ANOVA and χ2 tests. The total and domain M@T scores among groups are presented as box plots.

We then obtained a receiver-operator curve to determine the discriminative capacity of the total M@T score as well as its subdomains. We used a nonparametric approach for graphs and area under the curve (AUC) estimation. Additionally, we tested the total RUDAS score, a global BCT, and compared it to the memory-centered M@T. We assessed discrimination between cognitively intact (controls) and impaired (aMCI or dementia) participants, between controls and aMCI, non-dementia (aMCI and controls) and dementia, and aMCI and dementia. Likewise, we evaluated discrimination of neurodegeneration by positive MTA on all participants. We report the AUC with its 95% confidence interval. We considered AUC of over 0.95 as excellent, and those from 0.8 to 0.95 as good. We also report the optimal cut-off point estimated by the Youden Index.

Similarly, we obtained the Pearson coefficient and performed a linear regression between total M@T scores and subdomains with CSF biomarkers. Given that cognitively healthy subjects had normal biomarkers, and this may skew the analysis toward a stronger correlation, we performed a sensitivity analysis without control participants. Additionally, we evaluated the correlation between M@T subdomains and CSF biomarkers.

Finally, given that a high proportion of female participants belonged to the cognitively impaired groups, we performed a sensitivity analysis of discrimination and correlation to adjust for sex. We used R 4.2 (Vienna, Austria) for all statistical analysis. We considered a p value of <0.05 to be statistically significant unless otherwise stated.

We included a total of 185 participants: 63 in the control group, 53 with aMCI due to AD, and 69 with dementia due to AD. Median age was 75 (IQR: 70–78) and median years of education were 12 (IQR: 10–14). None of these characteristics were significantly different between groups. The majority were females (57.3%) and a greater proportion, although not statistically significant (p = 0.06), belonged to cognitively impaired groups.

CSF biomarkers were significantly different among all three groups. For the control group, 68% were amyloid positive (A+), while 38.1% were tau positive (T+). All MCI patients were A+ but only 86.8% were T+, while all AD patients were both A+ and T+. Likewise, for the neurodegeneration criteria, a total of 81.1% and 100% of aMCI and dementia participants, respectively, fulfilled neurodegeneration criteria defined by MTA score (Table 1).

Table 1.

Patient demographic and cognitive characteristics by diagnostic group

CharacteristicsControl (N = 63)MCI (N = 53)AD (N = 69)Total (N = 185)p value
Sex (F) 43 (68.3%) 30 (56.6%) 33 (47.8%) 106 (57.3%) 0.06 
Years of education, median (IQR) 12 (10–14) 12 (10–14) 12 (11–12) 12 (10–14) 0.791 
Age, years, median (IQR) 76 (72–79) 75 (71–78) 74 (69–78) 75 (70–78) 0.419 
M@T score, median (IQR) 
 Total score 45 (44–47) 34 (32–35) 18 (17–21) 34 (21–44) <0.001 
 Encoding 9 (9–10) 9 (8–9) 6 (5–8) 8 (7–9) <0.001 
 Temporal orientation 5 (4–5) 4 (4–5) 3 (3–4) 4 (3–5) <0.001 
 Semantic memory 13 (12–13) 12 (12–13) 8 (7–9) 12 (9–13) <0.001 
 Free recall 9 (8–9) 4 (3–4) 0 (0–1) 3 (1–8) <0.001 
 Cued recall 10 (9–10) 5 (4–5) 2 (1–2) 4 (2–9) <0.001 
RUDAS score, median (IQR) 26 (25–28) 22 (21–23) 19 (18–21) 23 (21–26) <0.001 
MTA, median (IQR) 1.5 (0.5–1.5) 2 (1.5–2.5) 2.5 (2.5–3.5) 2 (1.5–2.5) <0.001 
CSF biomarkers, median (IQR) 
 Aβ42 456 (378.5–547.5) 338 (305–389) 273 (236–323) 341 (284–423) <0.001 
 t-tau 75.2 (65.5–86.5) 105.6 (96.5–114.8) 167.4 (143.6–187.8) 105.6 (85.7–148.3) <0.001 
 p-tau 35.5 (28.9–50.6) 65.2 (49.2–74.1) 87.8 (76.3–98.1) 67.2 (43.1–79.3) <0.001 
ATN, n (%) 
 A+ 43 (68.3) 53 (100.0) 69 (100.0) 165 (89.2) <0.001 
 T+ 24 (38.1) 46 (86.8) 69 (100.0) 139 (75.1) <0.001 
 N+ 19 (30.2) 43 (81.1) 69 (100.0) 131 (70.8) <0.001 
CharacteristicsControl (N = 63)MCI (N = 53)AD (N = 69)Total (N = 185)p value
Sex (F) 43 (68.3%) 30 (56.6%) 33 (47.8%) 106 (57.3%) 0.06 
Years of education, median (IQR) 12 (10–14) 12 (10–14) 12 (11–12) 12 (10–14) 0.791 
Age, years, median (IQR) 76 (72–79) 75 (71–78) 74 (69–78) 75 (70–78) 0.419 
M@T score, median (IQR) 
 Total score 45 (44–47) 34 (32–35) 18 (17–21) 34 (21–44) <0.001 
 Encoding 9 (9–10) 9 (8–9) 6 (5–8) 8 (7–9) <0.001 
 Temporal orientation 5 (4–5) 4 (4–5) 3 (3–4) 4 (3–5) <0.001 
 Semantic memory 13 (12–13) 12 (12–13) 8 (7–9) 12 (9–13) <0.001 
 Free recall 9 (8–9) 4 (3–4) 0 (0–1) 3 (1–8) <0.001 
 Cued recall 10 (9–10) 5 (4–5) 2 (1–2) 4 (2–9) <0.001 
RUDAS score, median (IQR) 26 (25–28) 22 (21–23) 19 (18–21) 23 (21–26) <0.001 
MTA, median (IQR) 1.5 (0.5–1.5) 2 (1.5–2.5) 2.5 (2.5–3.5) 2 (1.5–2.5) <0.001 
CSF biomarkers, median (IQR) 
 Aβ42 456 (378.5–547.5) 338 (305–389) 273 (236–323) 341 (284–423) <0.001 
 t-tau 75.2 (65.5–86.5) 105.6 (96.5–114.8) 167.4 (143.6–187.8) 105.6 (85.7–148.3) <0.001 
 p-tau 35.5 (28.9–50.6) 65.2 (49.2–74.1) 87.8 (76.3–98.1) 67.2 (43.1–79.3) <0.001 
ATN, n (%) 
 A+ 43 (68.3) 53 (100.0) 69 (100.0) 165 (89.2) <0.001 
 T+ 24 (38.1) 46 (86.8) 69 (100.0) 139 (75.1) <0.001 
 N+ 19 (30.2) 43 (81.1) 69 (100.0) 131 (70.8) <0.001 

ATN, amyloid/tau/neurodegeneration.

The M@T total score was significantly different among control, aMCI, and dementia participants (p < 0.001), with higher median total scores for control patients (45, IQR: 44–47), followed by aMCI (34, IQR: 32–35), compared to dementia (18, IQR: 17–21). All subdomains were also different between groups (p < 0.001). Graphically, the largest differences were observed for the free-recall and cued recall subdomains between groups (Fig. 1).

Fig. 1.

Boxplot of M@T and domain scores by diagnostic groups. Y-axis corresponds to total M@T scores (left) and domain scores (right). aMCI, amnestic mild cognitive impairment; AD, Alzheimer’s disease.

Fig. 1.

Boxplot of M@T and domain scores by diagnostic groups. Y-axis corresponds to total M@T scores (left) and domain scores (right). aMCI, amnestic mild cognitive impairment; AD, Alzheimer’s disease.

Close modal

Discrimination of Clinical Groups

We assessed the diagnostic capacity of M@T total score by performing a receiver-operator curve (ROC) and obtaining the AUC with 95% CI. We also determined the optimal Youden-Index cut-off for each comparison. To discriminate between cognitively healthy and cognitively impaired (aMCI and dementia) participants, we found an AUC of 0.994 (95% CI: 0.98–1.00), with an optimal cut-off of 37. For dementia AD versus non-dementia (aMCI and controls), AUC was 0.989 (95% CI: 0.98–1.00), with optimal cut-off of 28. Lastly, for aMCI versus dementia, AUC was 0.98 (95% CI: 0.95–1.00), with optimal cut-off at 26 (Fig. 2).

Fig. 2.

Receiver-operator curves (ROC) for M@T total score for discrimination of cognitively intact patients (a), dementia due to AD (b), and dementia due to AD compared to aMCI due to AD (c). MCI, mild cognitive impairment; AD, Alzheimer’s disease.

Fig. 2.

Receiver-operator curves (ROC) for M@T total score for discrimination of cognitively intact patients (a), dementia due to AD (b), and dementia due to AD compared to aMCI due to AD (c). MCI, mild cognitive impairment; AD, Alzheimer’s disease.

Close modal

For comparison purposes, we assessed the performance of RUDAS for discriminating between the same patient groups. We found that RUDAS performed similarly well for discriminating between cognitively healthy and cognitively impaired patients with an AUC of 0.98 (95% CI: 0.97–1.00, p = 0.33). However, for differentiating between AD dementia patients versus non-dementia, AUC was significantly lower at 0.924 (95% CI: 0.89–0.96, p < 0.001). Likewise, when comparing aMCI versus AD patients, AUC was significantly lower for RUDAS at 0.84 (95% CI: 0.77–0.92, p < 0.001).

For M@T subdomains, free-recall and clued recall had excellent AUCs for all comparisons, while semantic memory was excellent for discrimination between dementia and non-dementia (aMCI and controls), or dementia and aMCI. The remaining subdomains presented good AUCs for all comparisons (Table 2).

Table 2.

Total M@T and subdomain discrimination among diagnostic groups

M@T scoresControl versus MCI/ADNon-AD versus ADMCI versus AD
Total score 0.99 (0.98–1.00) 0.99 (0.98–1.00) 0.98 (0.95–1.00) 
Encoding 0.79 (0.72–0.87) 0.88 (0.83–0.93) 0.85 (0.77–0.92) 
Temporal orientation 0.80 (0.73–0.87) 0.87 (0.83–0.92) 0.82 (0.74–0.90) 
Semantic memory 0.80 (0.73–0.88) 0.98 (0.96–1.00) 0.97 (0.94–1.00) 
Free recall 0.99 (0.99–1.00) 0.96 (0.94–0.99) 0.95 (0.91–0.99) 
Cued recall 1.00 (1.00–1.00) 0.97 (0.96–1.00) 0.97 (0.94–1.00) 
M@T scoresControl versus MCI/ADNon-AD versus ADMCI versus AD
Total score 0.99 (0.98–1.00) 0.99 (0.98–1.00) 0.98 (0.95–1.00) 
Encoding 0.79 (0.72–0.87) 0.88 (0.83–0.93) 0.85 (0.77–0.92) 
Temporal orientation 0.80 (0.73–0.87) 0.87 (0.83–0.92) 0.82 (0.74–0.90) 
Semantic memory 0.80 (0.73–0.88) 0.98 (0.96–1.00) 0.97 (0.94–1.00) 
Free recall 0.99 (0.99–1.00) 0.96 (0.94–0.99) 0.95 (0.91–0.99) 
Cued recall 1.00 (1.00–1.00) 0.97 (0.96–1.00) 0.97 (0.94–1.00) 

Correlation with CSF Biomarkers

To determine the correlation between M@T scores and CSF biomarkers, we obtained the Pearson coefficient and performed a linear regression. When including all patients, Pearson’s coefficient was 0.66 (95% CI: 0.57–0.74) for M@T and b-amyloid, −0.77 (95% CI: −0.82 to −0.70) for M@T and t-tau, and −0.72 (95% CI: −0.78 to −0.64) for M@T and p-tau (Fig. 3). In linear regression, M@T was a statistically significant predictor of all CSF biomarkers (p < 0.001) with an R2 of 0.43, 0.59, and 0.51 for b-amyloid, t-tau, and p-tau, respectively.

Fig. 3.

Correlation between M@T total score (x-axis) and CSF biomarkers in mg/dL (y-axis): b-amyloid (a), t-tau protein (b), and p-tau protein (c).

Fig. 3.

Correlation between M@T total score (x-axis) and CSF biomarkers in mg/dL (y-axis): b-amyloid (a), t-tau protein (b), and p-tau protein (c).

Close modal

We then performed a sensitivity analysis excluding cognitively healthy participants. We found that Pearson’s coefficient was 0.43 (95% CI: 0.26–0.56) for b-amyloid, −0.55 (95% CI: −0.66 to −0.41) for t-tau, and −0.47 (−0.6 to −0.32) for p-tau. In linear regression, M@T was still and statistically significant predictor of all CSF biomarkers (p < 0.001), but R2 was lower at 0.183, 0.301, and 0.220 for b-amyloid, t-tau, and p-tau, respectively.

Testing subdomain correlation with CSF biomarkers, we found that free-recall and cued recall had the highest Pearson’s coefficients for all biomarkers. Linear regression identified all subdomains as statistically significant predictors of biomarkers. However, the highest R2 values were reported for the free-recall and cued recall domains (Table 3).

Table 3.

Correlation between CSF biomarkers and M@T total and sub-domain scores

M@T scoreb-amyloidt-tau proteinp-tau protein
Pearson’s coeff. (95% CI)βp valueR2Pearson’s coeff. (95% CI)βp valueR2Pearson’s coeff. (95% CI)βp valueR2
Total 0.66 (0.57–0.74) 6.19 <0.001 0.439  −0.77 (−0.82–−0.7) −3.2 <0.001 0.591  −0.72 (−0.78–−0.64) −1.61 <0.001 0.512  
Immediate recall 0.47 (0.35–0.58) 26.65 <0.001 0.224  −0.55 (−0.65–−0.44) −13.87 <0.001 0.305  −0.48 (−0.58–−0.36) −6.46 <0.001 0.227  
Orientation 0.41 (0.29–0.53) 43.49 <0.001 0.171  −0.53 (−0.63–−0.42) −24.9 <0.001 0.283  −0.51 (−0.61–−0.39) −12.86 <0.001 0.259  
Semantics 0.49 (0.37–0.59) 19.15 <0.001 0.239  −0.69 (−0.75–−0.6) −11.95 <0.001 0.47  −0.57 (−0.66–−0.46) −5.32 <0.001 0.319  
Free recall 0.68 (0.59–0.75) 19.92 <0.001 0.456  −0.73 (−0.79–−0.65) −9.57 <0.001 0.531  −0.74 (−0.8–−0.66) −5.22 <0.001 0.543  
Cued recall 0.7 (0.62–0.77) 21.19 <0.001 0.493  −0.76 (−0.82–−0.7) −10.25 <0.001 0.583  −0.72 (−0.79–−0.65) −5.24 <0.001 0.523  
M@T scoreb-amyloidt-tau proteinp-tau protein
Pearson’s coeff. (95% CI)βp valueR2Pearson’s coeff. (95% CI)βp valueR2Pearson’s coeff. (95% CI)βp valueR2
Total 0.66 (0.57–0.74) 6.19 <0.001 0.439  −0.77 (−0.82–−0.7) −3.2 <0.001 0.591  −0.72 (−0.78–−0.64) −1.61 <0.001 0.512  
Immediate recall 0.47 (0.35–0.58) 26.65 <0.001 0.224  −0.55 (−0.65–−0.44) −13.87 <0.001 0.305  −0.48 (−0.58–−0.36) −6.46 <0.001 0.227  
Orientation 0.41 (0.29–0.53) 43.49 <0.001 0.171  −0.53 (−0.63–−0.42) −24.9 <0.001 0.283  −0.51 (−0.61–−0.39) −12.86 <0.001 0.259  
Semantics 0.49 (0.37–0.59) 19.15 <0.001 0.239  −0.69 (−0.75–−0.6) −11.95 <0.001 0.47  −0.57 (−0.66–−0.46) −5.32 <0.001 0.319  
Free recall 0.68 (0.59–0.75) 19.92 <0.001 0.456  −0.73 (−0.79–−0.65) −9.57 <0.001 0.531  −0.74 (−0.8–−0.66) −5.22 <0.001 0.543  
Cued recall 0.7 (0.62–0.77) 21.19 <0.001 0.493  −0.76 (−0.82–−0.7) −10.25 <0.001 0.583  −0.72 (−0.79–−0.65) −5.24 <0.001 0.523  

Discrimination of Neurodegeneration

We evaluated positive amyloid/tau/neurodegeneration criteria, defined by MTA, in all participants. By diagnostic group, 27.9% of controls fulfilled criteria, 83.3% of aMCI, and 100% of AD. We then tested the ability of M@T to discriminate neurodegeneration among all participants. We obtained an AUC of 0.87 (95% CI: 0.82–0.94). The optimal YI cut-off value was estimated at 39.2 (online suppl. Fig. 1; for all online suppl. material, see https://doi.org/10.1159/000534157).

Sensitivity Analysis

We found no differences when performing a sensitivity analysis in only male participants or only controls (online suppl. Tables 1, 2).

We performed a diagnostic assessment of the M@T in healthy participants and participants with aMCI and dementia due to AD, based on clinical and AD biomarker assessments. We found that the M@T exhibited excellent discrimination between all clinical groups, and it also correlated with biological and imaging markers of AD. We found strong correlation with t-tau protein, and moderate correlation with b-amyloid and p-tau. Likewise, we found good discrimination of neurodegeneration, as defined by MTA score.

Clinically, we found that M@T total score had an excellent diagnostic performance between all three comparisons: control versus aMCI/dementia, non-dementia (controls and aMCI) versus dementia, and aMCI versus dementia. Similar findings describing the excellent discrimination of M@T have been previously described in Portugal, the UK, and Spain [14‒16]. A recent meta-analysis evaluating and comparing four BCTs (M@T, MMSE, MoCA, and ACE-R) for MCI discrimination, found that M@T had the highest sensitivity among all [17]. This is in keeping with our findings that show M@T performing significantly better than RUDAS for AD discrimination. The meta-analysis also reported a pooled AUC of 0.961, like the one reported in our study (0.98).

Pathophysiologically, we observed high correlation between M@T total score and t-tau (−0.77), as well as p-tau (−0.72). There was moderate correlation with b-amyloid (0.66). However, performing a sensitivity analysis without controls yielded significant, but weaker correlations, at 0.43, −0.55, and −0.47, respectively. Moreover, M@T adequately discriminates between patients with positive and negative neurodegeneration (defined by MTA) with an AUC of 0.87. Although to our knowledge, no previous study has evaluated these relationships for the M@T, other BCSTs have been linked to serological and imaging AD biomarkers. Specifically, the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) has been found to be correlated to cerebral amyloid burden and lower hippocampal volumes [18]. Additionally, a recent paper described that one of its domains, the Delayed Memory Index, showed an even improved correlation [19]. The RBANS can take up to 30 min to administer, whereas the M@T takes 5 min in controls and 7 min in cognitively impaired persons [5, 20].

As this was a cross-sectional analysis, evolution from aMCI to dementia due to AD could not be assessed. However, a previous study has compared the capability of BCTs (MMSE and clock drawing test) and CSF biomarkers in predicting AD progression. They found no significant differences between their diagnostic performances [21]. This is in keeping with our findings of BCT-CSF correlation, but a follow-up, prospective study should be done to evaluate this specific hypothesis of AD progression prediction.

We also found that two domains, free-recall, and clued recall, had excellent AUCs for aMCI due to AD and dementia due to AD discrimination, and were highly correlated with CSF biomarkers. In this regard, the delayed recall domain in the Seoul Verbal Learning Test has been associated with increased tau binding in the lateral and medial temporal cortices [22]. Meanwhile, we described that the semantic memory domain had excellent discrimination only for AD, and not aMCI. This category has been evaluated using the Multilingual Naming Test (MINT) in a previous study and was found to be more strongly associated with increased tau binding in the left anterior temporal lobe and bilateral orbitofrontal regions [23]. Given that tau deposition seems to precede cortical atrophy and symptom progression in persons with AD [24] and that this distribution correlates with disease progression in typical AD [25], this could explain the M@T’s excellent discriminative capacity.

The M@T had excellent discrimination of aMCI due to AD and dementia due to AD in Lima, Peru, a LMIC. It was also strongly correlated with CSF biomarkers and had good discrimination of neurodegeneration. The recall subdomain presented the highest diagnostic capacity. Its use as a screening tool could allow LMICs to identify persons with aMCI or early AD who may then undergo biomarker confirmation.

Limitations

This is a secondary analysis of a previously published cohort of biomarker-proven AD. Thus, patient selection was not optimized for this research question, making it prone to selection bias. Patients included in the study attended a specialized neurology clinic, so referral bias may be present. However, these are likely mitigated by the community recruitment strategies used for the first study. Additionally, only patients with aMCI and dementia due to AD were included, so follow-up studies should include patients with different causes of dementia. Finally, given that this is a local sample, and the test was administered in Spanish, these results are not necessarily generalizable to other countries/languages.

The study was conducted in accordance with the latest revision of the Declaration of Helsinki, and ethical approval for this study was obtained from the Committee for Medical and Health Research Ethics, Hospital Nacional Docente Madre-Niño-HONADOMANI “San Bartolomé” (No: 10777-18). Participation in the study was voluntary and anonymous. We used an informed consent which was signed by patients and their legal guardian or next of kin. We did not involve patients or the public in the research study design or implementation.

The authors have no conflicts of interest to declare.

This work was self-funded. NC and RM were supported by the National Institute of Health (R56AG069118-01) and the Multi-Partner Consortium to Expand Dementia Research in Latin America (ReDLat), supported by the National Institutes of Health, the National Institutes of Aging (R01 AG057234). MD was supported by the National Institute of Health (1-K23-MH131466-01), the Alzheimer’s Association (AARGD-22-924896), and the American Academy of Neurology.

N.C. designed the study. N.C., R.M., M.D., and S.L. supervised data collection. R.M., D.C., J.C.C., F.B., and J.C.H. collected the data. M.M., D.C.-M., E.H.-P., and D.L. were responsible for the statistical design of the study. M.M. carried out the statistical analysis. M.M. and N.C. wrote the paper. R.M., E.H.-P., D.L., M.M.D., and S.L. critically reviewed the manuscript. All authors agreed to the final version.

Authors confirm that the data supporting the findings of this study are available upon reasonable request. Data are not publicly available due to ethical reasons. Further inquiries can be directed to the corresponding author.

1.
GBD 2019 Dementia Forecasting Collaborators, Steinmetz JD, Vollset SE, Fukutaki K, Chalek J, Abd-Allah F, et al. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the Global Burden of Disease Study 2019. Lancet Public Health. 2022;7(2):e105–25.
2.
Nandi A, Counts N, Chen S, Seligman B, Tortorice D, Vigo D, et al. Global and regional projections of the economic burden of Alzheimer’s disease and related dementias from 2019 to 2050: a value of statistical life approach. EClinicalMedicine. 2022;51:101580.
3.
Ibanez A, Parra MA, Butler C; Latin America and the Caribbean Consortium on Dementia LAC-CD. The Latin America and the caribbean Consortium on dementia (LAC-CD): from networking to research to implementation science. J Alzheimers Dis. 2021;82(s1):S379–94.
4.
van der Flier WM, Scheltens P. The ATN framework – moving preclinical Alzheimer disease to clinical relevance. JAMA Neurol. 2022;79(10):968–70.
5.
Rami L, Molinuevo JL, Sanchez-Valle R, Bosch B, Villar A. Screening for amnestic mild cognitive impairment and early Alzheimer’s disease with M@T (Memory Alteration Test) in the primary care population. Int J Geriatr Psychiatry. 2007;22(4):294–304.
6.
Zvěřová M. Clinical aspects of Alzheimer’s disease. Clin Biochem. 2019;72:3–6.
7.
Custodio N, Malaga M, Chambergo-Michilot D, Montesinos R, Moron E, Vences MA, et al. Combining visual rating scales to identify prodromal Alzheimer’s disease and Alzheimer’s disease dementia in a population from a low and middle-income country. Front Neurol. 2022;13:962192.
8.
McKhann GM, Knopman DS, Chertkow H, Hyman BT, Jack CR, Kawas CH, et al. The diagnosis of dementia due to Alzheimer’s disease: recommendations from the National Institute on Aging‐Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement. 2011;7(3):263–9.
9.
Custodio N, Montesinos R, Lira D, Herrera-Perez E, Chavez K, Hernandez-Córdova G, et al. Validation of the RUDAS in patients with a middle-level education in Lima, Peru. Am J Alzheimers Dis Other Demen. 2019;34(7–8):513–22.
10.
Custodio N, Herrera-Pérez E, Montesinos R, Lira D, Metcalf T. Brief cognitive tests validated in Peru for detection of cognitive impairment a systematic mapping of the scientific literature. Dement Neuropsychol. 2020;14(2):134–44.
11.
Custodio N, Lira D, Herrera-Perez E, Montesinos R, Castro-Suarez S, Cuenca-Alfaro J, et al. Memory alteration test to detect amnestic mild cognitive impairment and early Alzheimer’s dementia in population with low educational level. Front Aging Neurosci. 2017;9:278.
12.
Rhodius-Meester HFM, Benedictus MR, Wattjes MP, Barkhof F, Scheltens P, Muller M, et al. MRI visual ratings of brain atrophy and white matter hyperintensities across the spectrum of cognitive decline are differently affected by age and diagnosis. Front Aging Neurosci. 2017;9:117.
13.
Ferreira D, Verhagen C, Hernández-Cabrera JA, Cavallin L, Guo C-J, Ekman U, et al. Distinct subtypes of Alzheimer’s disease based on patterns of brain atrophy: longitudinal trajectories and clinical applications. Sci Rep. 2017;7:46263.
14.
Ozer S, Noonan K, Burke M, Young J, Barber S, Forster A, et al. The validity of the memory alteration test and the test your memory test for community‐based identification of amnestic mild cognitive impairment. Alzheimers Dement. 2016;12(9):987–95.
15.
Sousa M, Pereira A, Costa R, Rami L. Initial phase of adaptation of memory alteration test (M@T) in a Portuguese sample. Arch Gerontol Geriatr. 2015;61(1):103–8.
16.
Carnero-Pardo C, Rego-García I, Mené Llorente M, Alonso Ródenas M, Vílchez Carrillo R. Diagnostic performance of brief cognitive tests in cognitive impairment screening. Neurologia. 2022;37(6):441–9.
17.
Breton A, Casey D, Arnaoutoglou NA. Cognitive tests for the detection of mild cognitive impairment (MCI), the prodromal stage of dementia: meta‐analysis of diagnostic accuracy studies. Int J Geriatr Psychiatry. 2019;34(2):233–42.
18.
Ottoy J, Niemantsverdriet E, Verhaeghe J, de Roeck E, Struyfs H, Somers C, et al. Association of short-term cognitive decline and MCI-to-AD dementia conversion with CSF, MRI, amyloid- and 18F-FDG-PET imaging. Neuroimage Clin. 2019;22:101771.
19.
Suhrie KR, Hammers DB, Porter SM, Dixon AM, King JB, Anderson JS, et al. Predicting biomarkers in intact older adults and those with amnestic mild cognitive impairment, and mild Alzheimer’s disease using the repeatable battery for the assessment of neuropsychological Status. J Clin Exp Neuropsychol. 2021;43(9):861–78.
20.
Loughan AR, Braun SE, Lanoye A. Repeatable battery for the assessment of neuropsychological Status (RBANS): preliminary utility in adult neuro-oncology. Neurooncol Pract. 2019;6(4):289–96.
21.
Palmqvist S, Hertze J, Minthon L, Wattmo C, Zetterberg H, Blennow K, et al. Comparison of brief cognitive tests and CSF biomarkers in predicting Alzheimer’s disease in mild cognitive impairment: six-year follow-up study. PLoS One. 2012;7(6):e38639.
22.
Cho H, Choi JY, Hwang MS, Lee JH, Kim YJ, Lee HM, et al. Tau PET in Alzheimer disease and mild cognitive impairment. Neurology. 2016;87(4):375–83.
23.
Digma LA, Madsen JR, Reas ET, Dale AM, Brewer JB, Banks SJ, et al. Tau and atrophy: domain-specific relationships with cognition. Alzheimers Res Ther. 2019;11(1):65.
24.
La Joie R, Visani AV, Baker SL, Brown JA, Bourakova V, Cha J, et al. Prospective longitudinal atrophy in Alzheimer’s disease correlates with the intensity and topography of baseline tau-PET. Sci Transl Med. 2020;12(524):eaau5732.
25.
Younes L, Albert M, Moghekar A, Soldan A, Pettigrew C, Miller MI. Identifying changepoints in biomarkers during the preclinical phase of Alzheimer’s disease. Front Aging Neurosci. 2019;11:74.