A brief digital symbol-coding task outperforms clinical benchmarks for alzheimer’s trial pre-enrichment

Background

Screen failure rates in Alzheimer’s (AD) trials can exceed 40% in mild AD studies (Hampel 2023) and be higher for early disease (prodromal AD >75%; preclinical AD >85%;Langbaum 2023; Aisen 2025) or in rare phenotypes.

Scalable, low-cost tools are needed to identifyindividuals most likely to have early AD pathology or measurable cognitive decline before costly PET or CSF screening. NeuLogiq® is a multimodal digital endpoint platform used in CNS clinical trials. Here we report recent interim data on Symbol Swap, a 2-minute tablet-based implementation of the Symbol Coding Task (Figure 1; Jaeger 2018; Dyer 2025) and compare its pre-enrichment performance to traditional paper and pen tests.

We present concurrent data from one completed at-home study, and breaking interim data from two large in-clinic studies (total n=508 participants).

**METHODS:

Study 1** - CNS-101 study (N=46 dementia patients; N=54 controls, 7 UK sites; funded by Cumulus Pharma Consortium and IUK): data from the first at-home use of Symbol Swap, and compared to ADAS-Cog (~45min trained clinician time). Pathology status used ALZpath phosphorated-tau 217 (Ashton 2024).

Study 2 - Fastball i4i study (target N=1000, in progress, conducted at Southmead hospital memory clinic; funded by NIHR): Symbol Swap was administered once during clinical assessment and compared to MoCA (~20min trained clinician time). Pathology status as Study 1.

Study 3 - BioHermes-2 (target N=1000 including MCI, dementia & controls, in progress; with IUK funding): Conducted across 20 US & UK sites, Symbol Swap was administered once in clinic and compared to MMSE (~10min trained clinician time).

Across all studies the AUC-ROC rank association was calculated between performance on Symbol Swap and phenotypes relevant to clinical trial recruitment.

Results

Table 1: Discriminative power of Symbol Swap digital coding task to detect clinical and pathological phenotypes, compared to clinical benchmarks assessments. Accuracy of rank association is quantified using AUC-ROC (Area Under the Receiver Operating Characteristic Curve).

AUC accuracy per study and cohort contrast		Symbol Swap (digital DSST)	Clinical Benchmark (paper/pen)
Study 1	Dementia (n=46) vs Controls (n=54)	0.91	0.97	ADAS-Cog 13
Study 1	p-tau217+ (n-20) vs p-tau217- (n=24), Controls only	0.77	0.57	ADAS-Cog 13
Study 1	p-tau217+ (n=52) vs p-tau217- (n=34), All Participants	0.81	0.73	ADAS-Cog 13
Study 2	AD (MCI and Dementia, n=60) vs Controls (n=113)	0.92	0.92	MoCA
Study 2	MCI-AD (n=11) vs Controls (n=113)	0.92	0.82	MoCA
Study 2	p-tau217+ (n=73) vs p-tau217- (n=48), All Participants	0.74	0.69	MoCA
Study 3	MCI (n=124) vs Controls (n=111)	0.68	0.74	MMSE

Conclusions

In post-hoc screening of control, MCI and AD-dementia participants across three studies, a short (~2mins) patient-friendly symbol coding task compares well to long burdensome assessments that require trained clinical staff. Additionally, Symbol Swap can detect Alzheimer’s pathology, even in clinically healthy controls. In the Alzheimer’s space, this could precede the use of plasma biomarkers, or be combined in a multimodal composite including further cognitive and speech endpoints. In psychiatry, the same principles may hold where cognition is impaired (e.g. in MDD or Schizophrenia) and where novel treatments may have benefits beyond core mood symptoms.

References

Aisen, Paul, et al. “Advancing the science of recruitment for Alzheimer’s clinical trials: Challenges and opportunities.” The Journal of Prevention of Alzheimer’s Disease 12.7 (2025): 100230. Ashton, Nicholas J., et al. “Diagnostic accuracy of a plasma phosphorylated tau 217 immunoassay for Alzheimer disease pathology.” JAMA neurology 81.3 (2024): 255-263. Dyer, John, et al. “Temporal Dynamics of Subtle Cognitive Change: Validation of a User-Friendly Multidomain Digital Assessment Using an Alcohol Challenge.” Journal of Medical Internet Research 27 (2025): e55469. Hampel, Harald, et al. “Blood-based biomarkers for Alzheimer’s disease: current state and future use in a transformed global healthcare landscape.” Neuron 111.18 (2023): 2781-2799. Jaeger, Judith. “Digit symbol substitution test: the case for sensitivity over specificity in neuropsychological testing.” Journal of clinical psychopharmacology 38.5 (2018): 513-519. Langbaum, Jessica B., et al. “Recommendations to address key recruitment challenges of Alzheimer’s disease clinical trials.” Alzheimer’s & Dementia 19.2 (2023): 696-707.