Multivariate brain-cognition covariance supports the criterion validity of cognitive screening performance

This study validates the Riga Cognitive Screening Task as an effective tool for identifying cognitive impairment in older adults by demonstrating significant multivariate covariance between its performance scores and cortical thickness in temporal and parietal brain regions.

The Gist

The Big Picture: Checking the "Thermometer"

Imagine you are a doctor trying to figure out if a patient has a fever. You have a new thermometer called the Riga Cognitive Screening Task (RiTa). Before you trust this new thermometer to diagnose patients, you need to prove it actually measures temperature and isn't just measuring how hot the patient's hand feels or how sweaty they are.

In the world of brain health, "temperature" is cognitive decline (like memory loss or confusion), and the "fever" is Alzheimer's or dementia.

This study asked a simple question: Does this new brain test (RiTa) actually match what is happening inside the brain physically?

The Cast of Characters

• The Test (RiTa): A new quiz designed to check how well older adults think, remember, and solve problems.
• The "Real" Proof (The MRI): Instead of just trusting the quiz score, the researchers looked inside the participants' brains using a giant camera (an MRI scanner). They specifically looked at the thickness of the brain's outer layer (cortical thickness).
  • Analogy: Think of the brain's outer layer like the rind of an orange. In a healthy orange, the rind is thick and firm. As the orange gets old or starts to rot (dementia), the rind gets thin and shriveled. The researchers were checking if people who scored low on the quiz also had "thinner rinds" in specific parts of their brains.
• The Participants: 106 older adults (average age 70) with varying levels of memory and thinking skills.

The Detective Work: Finding the Hidden Pattern

The researchers didn't just look at one thing at a time (like "Does memory score match the hippocampus?"). That would be like checking if the engine temperature matches the tire pressure separately.

Instead, they used a powerful statistical tool called Partial Least Squares Correlation (PLS-C).

• Analogy: Imagine you have a huge orchestra. You want to know if the music (the brain test scores) matches the instruments (the brain structure). Instead of asking, "Did the violinist play loud?" and "Did the drum get hit hard?", PLS-C listens to the whole symphony at once. It looks for a hidden "conductor" (a latent variable) that explains why the whole group is playing together.

What They Found

1. The Match Was Real: The study found a strong "symphony." When people scored lower on the RiTa test (meaning they had more trouble with memory, reasoning, and language), their brains showed thinning in specific areas known to be the "first responders" in Alzheimer's disease (like the temporal and parietal lobes).
2. The "Control" Check: To make sure they weren't just seeing random noise, they checked a part of the brain that shouldn't be affected by this specific type of memory loss (the anterior cingulate cortex). It showed no connection to the test scores. This proved the pattern was specific and real, not a fluke.
3. Age and Education: Even after accounting for how old the people were and how much school they finished, the connection between the test scores and the brain thinning remained strong. This means the test is measuring the brain's health, not just how "smart" someone was in school or how old they are.

Why This Matters

• No More Guessing: Many existing tests can be tricky for people with different education levels or cultural backgrounds. This study suggests RiTa is a robust tool that aligns with the actual physical changes in the brain.
• Early Detection: Because the test matches the physical thinning of the brain, it could help doctors catch cognitive issues earlier, before they become severe.
• A New Way to Test: The study showed that looking at the whole pattern of brain and behavior (the orchestra) is better than looking at single parts in isolation.

The Bottom Line

The researchers successfully proved that the Riga Cognitive Screening Task (RiTa) is a valid tool. It's not just a paper quiz; it is a window that accurately reflects the physical state of the aging brain. If the test says a patient is struggling, the MRI confirms that the "rind" of their brain is indeed thinning in the right places.

In short: The new test works, and it's backed up by hard evidence from inside the brain itself.

Technical Summary

1. Problem Statement

The study addresses the critical issue of underdiagnosis and misdiagnosis of dementia and mild cognitive impairment (MCI) in older adults. Current prevalence estimates suggest a significant gap between actual dementia rates and diagnosed cases in Europe. Existing screening tools (e.g., MMSE) often suffer from:

• Cultural and educational bias.
• Insensitivity to early deficits in reasoning, semantic processing, and executive control.
• Lack of rigorous psychometric validation, particularly regarding criterion validity (the extent to which test scores correlate with an external, objective criterion).

The authors aim to validate the Riga Cognitive Screening Task (RiTa), a new integrative tool designed to assess cognitive performance while accounting for cognitive reserve (lifestyle and life experiences). Specifically, they seek to establish its concurrent criterion validity by correlating RiTa performance with objective neuroanatomical markers of pathological aging.

2. Methodology

Participants

• Sample: 106 older adults ($M_{age} = 70.49$, $SD = 8.08$; 35.8% male).
• Inclusion Criteria: Native Latvian speakers, no active oncological/psychiatric/neurological treatment (except for MCI/AD-related cognitive impairment).
• Cognitive Range: Participants had varying levels of functioning, with Montreal Cognitive Assessment (MoCA) scores ranging from 10 to 29 ($M=22.27$), ensuring a spectrum from normal aging to significant impairment.
• Exclusion: No specific exclusion based on MoCA score to preserve natural variance in cognitive ability.

Data Collection

1. Cognitive Assessment: Participants completed the RiTa, which includes 11 tasks covering:
   • Episodic/semantic verbal memory, working memory, orientation, fluency (phonemic/semantic), visuospatial abilities, attention, reasoning, inhibition, and motor abilities.
2. Neuroimaging: All participants underwent 3T MRI scanning.
   • Sequence: 3D T1 SPGR was used for structural analysis.
   • Processing: Used the FreeSurfer pipeline (v8.1.0) with SynthSeg (deep learning-based segmentation) to generate cortical thickness estimates.
   • Regions of Interest (ROIs): 18 specific regions selected based on vulnerability in Alzheimer's Disease (AD), including medial temporal (entorhinal, parahippocampal), lateral temporal (fusiform, middle/inferior temporal), and parietal (precuneus, inferior parietal lobule, supramarginal gyrus) cortices.

Statistical Analysis

• Primary Method: Behavioral Partial Least Squares Correlation (PLS-C).
  • Unlike univariate correlations, PLS-C is a multivariate technique that identifies latent variables (LVs) maximizing the shared covariance between two data matrices:
    • Matrix X: Cortical thickness of 18 ROIs.
    • Matrix Y: Performance scores on 12 RiTa cognitive tasks.
  • Significance Testing: Permutation testing (1,000 permutations) to determine if the observed covariance is greater than chance.
  • Controls: Age and education were regressed out to test the robustness of the brain-cognition link. A negative control region (rostral anterior cingulate cortex) was included to ensure anatomical specificity.

3. Key Results

Latent Variable Extraction

Three latent variables were extracted, but only LV1 was statistically significant ($p < .001$) and interpretable:

• Correlation Strength: LV1 showed a strong brain-cognition correlation ($r = .605$).
• Post-Control Analysis: After controlling for age and education, the correlation remained significant ($r = .469, p = .008$), though slightly attenuated.

Brain Pattern (LV1 Saliences)

The significant latent variable linked cognitive performance to a distributed pattern of cortical thinning in regions known to be vulnerable in AD:

• Core Contributors: Bilateral precuneus, inferior parietal lobule, supramarginal gyrus, and middle temporal cortex.
• Hemispheric Specificity:
  • Left: Entorhinal cortex, parahippocampal cortex.
  • Right: Supramarginal and middle temporal cortices.
• Specificity: The rostral anterior cingulate cortex (negative control) showed negligible salience, confirming the pattern was specific to AD-vulnerable networks.

Cognitive Pattern (LV1 Saliences)

The cognitive profile associated with the brain pattern included:

• Core Domains: Inhibition, reasoning, phonemic/semantic fluency, immediate/delayed semantic verbal memory, orientation, and working memory.
• Supporting Domains: Visuospatial abilities, episodic memory, and processing speed.
• Weak Contribution: Motor functions.

Stability

• The pattern of brain saliences remained highly stable after controlling for age and education (Salience similarity $r = .89$; Jaccard index $J = .50$), indicating the multivariate brain-cognition relationship is robust and not merely an artifact of demographic variables.

4. Key Contributions

1. Validation of RiTa: The study provides empirical evidence that the Riga Cognitive Screening Task possesses high criterion validity. Performance on RiTa significantly covaries with structural brain integrity in regions associated with neurodegeneration.
2. Methodological Innovation: Demonstrates the utility of Multivariate PLS-C over traditional univariate correlations for validating cognitive screening tools. This approach captures system-level reorganization and distributed brain-behavior relationships rather than isolated structure-function pairs.
3. Continuous Spectrum Approach: By including participants across a wide MoCA spectrum (10–29) rather than strict clinical vs. control groups, the study avoids "criterion contamination" and better reflects the continuous nature of cognitive decline.
4. Neuroanatomical Specificity: Confirms that RiTa performance is specifically linked to the medial temporal, lateral temporal, and parietal networks, which are the hallmark sites of early Alzheimer's pathology.

5. Significance and Limitations

Significance:

• Clinical Utility: The findings suggest RiTa is a reliable tool for early detection of cognitive dysfunction, particularly memory-associated impairments, as it aligns with biological markers of disease.
• Psychometric Rigor: Establishes a framework for validating future screening tools using neuroimaging criteria, moving beyond self-report or simple correlation with other behavioral tests.
• Understanding Aging: Highlights that cognitive decline in older adults is a distributed process involving multiple cortical domains, which can be effectively captured by multivariate modeling.

Limitations:

• Cross-Sectional Design: The study cannot establish predictive validity (i.e., whether RiTa scores predict future decline) or causality.
• Single Time Point: Cognitive performance can be influenced by transient factors (fatigue, sleep), and cortical thickness is a static measure at one point in time.
• Criterion Scope: Using only cortical thickness limits the understanding of other neural mechanisms (e.g., functional connectivity, white matter integrity).
• Demographics: The sample was predominantly highly educated (mean 14 years), which may limit generalizability to populations with lower educational attainment, though education was statistically controlled.

Conclusion:
The study successfully validates the Riga Cognitive Screening Task as a psychometrically sound instrument for assessing cognitive impairment, supported by a robust, multivariate correlation with cortical atrophy patterns characteristic of pathological aging.