Detection of cerebrospinal fluid biomarkers changes of Alzheimer's disease using a cognitive stress test in persons with subjective cognitive decline and mild cognitive impairment

Alzheimer’s disease features beta-amyloid plaques and tau neurofibrillary tangles, with pathophysiological changes beginning years before dementia onset. Early, accurate detection is essential as disease-modifying therapies are more effective in prodromal phases. Traditional neuropsychological tests show mixed ability to detect early disease or predict progression, especially when deficits are subtle. Cognitive challenge tests like the LASSI-L, which probe proactive semantic interference and recovery from interference, have shown promise against imaging biomarkers but have not been validated against CSF markers that may change earlier than neuroimaging metrics. This study tests the hypothesis that LASSI-L performance relates to CSF amyloid and tau measures and can detect early AD-related biomarker abnormalities in individuals with SCD and amnestic MCI.

Prior work indicates that amnestic MCI is heterogeneous and not uniformly amyloid-positive. Cognitive stress tests (e.g., LASSI-L) refine memory assessment by evaluating proactive semantic interference and recovery, offering improved sensitivity over traditional tests. LASSI-L performance has been linked to brain regions affected in early AD, metabolism, and amyloid load in community and clinic samples, and shows cross-cultural generalizability. However, no studies had examined its association with CSF biomarkers (Aβ1-42, Aβ1-40, p-tau181, t-tau), which may capture earlier pathophysiological changes than PET measures. This gap motivated the current investigation.

Design and participants: Observational study of Spanish-speaking patients consulting for memory loss without functional impairment, enrolled January 2019–June 2023. Comprehensive neuropsychological assessment and LASSI-L were administered; LASSI-L was not used diagnostically. Inclusion criteria: memory complaints; Functional Activities Questionnaire <2; clinical features suggesting AD risk (family history, age at symptom onset >60 years, clinical characteristics, or neuropsych profile suggestive of amnestic MCI); no alternative causes (psychiatric/neurological/medical/sensory). Exclusion: LP contraindications; neurological disorders affecting cognition (e.g., stroke, epilepsy); active psychiatric disease; substance abuse; significant sensory impairment. Final sample: 179 after excluding suspected non-AD pathophysiology (elevated t-tau with normal amyloid). Demographics (final sample): mean age 70.52±7.16 years; education 11.16±4.82 years; 51.95% female. Classification: aMCI vs SCD using FCSRT (−1.5 SD impairment on learning/recall indices = aMCI; otherwise SCD); FCSRT unavailable in 4. Resulting groups: 103 MCI, 72 SCD. Neuropsychological battery included ACE-III and a co-normed Spanish battery (NEURONORMA). LASSI-L paradigm: two 15-word lists from three semantic categories (fruits, instruments, clothing) with controlled presentation and cued recalls. Key scores: Cued A2 (maximal storage), Cued B1 (impact of proactive interference), Cued B2 (recovery from interference), Delayed Recall (DR; 20-min delay). Intrusions were quantified, and percentage intrusion errors (PIE) computed for Cued B1 and B2. CSF biomarkers: Aβ1-42, Aβ1-40, p-tau181, t-tau measured on Lumipulse G600 II (Fujirebio) with quarterly external QC. Cutoffs: Aβ1-42/Aβ1-40 ratio 0.068; Aβ1-42 723 pg/mL; p-tau181 59 pg/mL; t-tau 410 pg/mL; Aβ1-42/p-tau181 ratio 11.8. Statistics: Non-normal distributions; Mann–Whitney U for group comparisons; Spearman correlations; effect sizes via rank biserial correlation (small 0.10–0.29, medium 0.30–0.49, large ≥0.50). Logistic regressions (backward stepwise) to predict biomarker abnormalities using raw LASSI-L scores with age and education; bootstrapped 1,000 resamples; Hosmer–Lemeshow for calibration; accuracy and AUC for discrimination. Analyses performed in SPSS 26, JASP 0.18.1, Jamovi 2.3.21.

- Correlations: Cued A2 correlated moderately with Aβ1-42 and Aβ1-42/p-tau181 and weakly with Aβ1-42/Aβ1-40, p-tau181, and t-tau. Cued B2 and DR showed moderate correlations with Aβ1-42, Aβ1-42/Aβ1-40, and Aβ1-42/p-tau181 and weak correlations with tau measures; PIE-B1 and PIE-B2 showed moderate associations with amyloid-related indices and p-tau181. - Whole-sample amyloid status (Aβ1-42/Aβ1-40): Amyloid-positive patients scored worse on LASSI-L (means±SD): Cued A2 8.77±2.49 vs 10.48±2.26; Cued B1 4.15±2.18 vs 5.32±2.79; Cued B2 6.92±2.61 vs 8.90±2.75; DR 8.03±5.76 vs 12.96±6.51; and had higher intrusions: PIE-B1 56.09%±20.43 vs 39.98%±22.68; PIE-B2 32.86%±19.82 vs 20.17%±16.63 (all p≤0.007; effect sizes mostly medium). - aMCI subgroup: 63.1% amyloid-positive by ratio. Amyloid-positive aMCI showed lower Cued A2, Cued B2, DR and higher PIE-B1/PIE-B2 (medium effects). Elevated p-tau181 (53.3%) associated with lower DR and higher PIE-B1/PIE-B2 (medium effects). Abnormal Aβ1-42/p-tau181 ratio (57.28%) associated with lower Cued A2, Cued B2, DR (large effect for DR) and higher intrusions. - SCD subgroup: 43.05% amyloid-positive. Reduced amyloid ratio associated with lower Cued B2 and DR (medium and small effects, respectively). Elevated p-tau181 (38.88%) associated with lower Cued A2, Cued B2, DR and trends to higher PIE-B1/PIE-B2 (medium effects for significant outcomes). Aβ1-42/p-tau181 ratio groups did not differ significantly on LASSI-L. - Agreement between biomarker classifications was high in both aMCI and SCD (generally >90%). - Logistic regression: • Predicting abnormal Aβ1-42/Aβ1-40: DR (OR 0.904 per point; p<0.001) and age (OR 1.142 per year; p<0.001). Model accuracy 0.749; AUC 0.785 (95% CI 0.716–0.855). AUC in MCI 0.742; SCD 0.799. • Predicting elevated p-tau181: DR (OR 0.905; p=0.003), age (OR 1.084; p=0.010), PIE-B2 (OR 1.021 per percentage point; p=0.050). Accuracy 0.687; AUC 0.771 (95% CI 0.696–0.845). AUC in MCI 0.720; SCD 0.791. • Predicting reduced Aβ1-42/p-tau181: DR (OR 0.900; p=0.002), age (OR 1.088; p=0.007), PIE-B2 (OR 1.022; p=0.041). Accuracy 0.714; AUC 0.788 (95% CI 0.715–0.860). AUC in MCI 0.770; SCD 0.714.

The study demonstrates biological validity of LASSI-L in early AD by linking performance, especially recovery from proactive semantic interference (Cued B2) and 20-minute delayed recall (DR), with CSF amyloid and tau indices. DR emerged as the most discriminative measure, likely reflecting cumulative demands on encoding, resilience to interference, and retrieval. The similar predictive performance across amyloid ratio, p-tau181, and Aβ1-42/p-tau181 ratio reflects biomarker interrelations. Importantly, LASSI-L also differentiated biomarker status within SCD, suggesting sensitivity at very early stages when conventional tests may be less informative. These results support using cognitive challenge paradigms to enrich clinical trial cohorts and guide decisions regarding invasive/expensive biomarker testing, and point to potential gains when combined with plasma biomarkers. Overall, sensitive cognitive tools remain crucial alongside fluid and imaging biomarkers to capture the earliest cognitive changes targeted by disease-modifying interventions.

LASSI-L performance, particularly delayed recall and recovery from proactive semantic interference, is associated with CSF amyloid and tau abnormalities in SCD and aMCI. The test achieves acceptable discrimination of biomarker positivity and can aid early detection and triage for CSF/PET testing, as well as contextualize plasma biomarker findings. Future work should directly compare LASSI-L with traditional tests, integrate multimodal markers (genetics, plasma, imaging) to improve predictive accuracy, and use longitudinal designs to model trajectories and biomarker dynamics.

- Statistical models assumed linearity; more complex/longitudinal modeling may better capture biomarker and cognitive dynamics. - Single-center memory clinic cohort limits generalizability to broader or population-based samples. - Neuropsychological examination influenced CSF referral, precluding evaluation of added value of tests other than LASSI-L for biomarker detection. - PET imaging topography and APOE genotype were not assessed; these could refine associations. - The affiliation list suggests some missing details (e.g., one author’s listed footnote 4 affiliation not provided in the excerpt), though this does not affect results.