Identifying and tracking cognitive decline in multiple sclerosis (MS) is hindered by reliance on episodic, in-person, examiner-administered analog assessments. Smartphone platforms can collect continuous sensor and event streams to derive digital biomarkers for scalable, self-administered testing. This study assessed the clinical utility and incremental validity of an equiprobable (50/50) continuous vigilance Go/No-Go paradigm as a candidate module within the Neurological Functions Test Suite (NeuFun-TS).
The primary objective was to determine whether the smartphone Go/No-Go task provided diagnostic separation between healthy donors (HD) and people with MS and whether it offered information incremental to existing NeuFun-TS modules.
Analyses used 3,270 quality-controlled trials collected from 303 participants. Primary analyses focused on 488 first-trial observations drawn from 34 healthy donors and 213 MS participants. From the raw event streams produced during the continuous vigilance Go/No-Go paradigm, investigators screened a set of 16 candidate metrics and applied psychometric criteria to retain 10 outcomes for downstream analyses.
Using the retained metrics, the study evaluated: diagnostic group separation (HD vs MS and MS subtypes), test-retest reliability over time, correlations with clinical and MRI benchmarks, and incremental validity relative to two existing NeuFun-TS modules — the randomized Symbol Digit Modalities Test (rSDMT) and the Motor Sequencing Test (MST) — where paired observations were available.
Of the 16 candidate measures extracted from the event stream, 10 were retained based on intrinsic psychometric properties reported in the source. These included latency-based measures and various accuracy/inhibition-derived outcomes. The retained set formed the basis for tests of discriminative ability and reliability.
Nine of the 10 retained outcomes significantly separated healthy donors from MS participants (all adjusted p < 0.05). Seven of the metrics also distinguished relapsing–remitting from progressive MS subtypes (all adjusted p < 0.05).
However, the authors report that the discriminative power of the Go/No-Go paradigm was driven almost entirely by response latency (reaction time) rather than by commission error rates or other inhibitory-control metrics. Composite multi-predictor scores combining multiple retained outcomes did not offer diagnostic performance superior to single reaction time measures.
This pattern indicates that the Go/No-Go task primarily captured deficits in processing speed, and its classification value was effectively redundant with simple reaction time indicators.
Longitudinal test-retest reliability of the Go/No-Go outcomes ranged from poor to moderate according to the source analysis. The module therefore had limited stability over time, reducing its utility for reliable longitudinal monitoring compared with measures that demonstrate stronger test-retest characteristics.
When evaluated against existing digital NeuFun-TS modules — the rSDMT and the MST — the Go/No-Go task failed to provide incremental clinical value. Paired observations showed that the information captured by the Go/No-Go task, being dominated by reaction time, overlapped with metrics already implemented on the platform.
Given the absence of added diagnostic or prognostic benefit over the established modules, the Go/No-Go module did not justify additional participant testing burden.
The smartphone-based equiprobable continuous vigilance Go/No-Go task reflects processing speed impairment in people with MS and can separate healthy donors from MS participants and MS subtypes on several retained outcomes. Crucially, its discriminative capacity depended almost entirely on reaction time rather than errors of inhibition, and composite scores did not improve separation.
Because the Go/No-Go task offered no incremental clinical benefit over existing NeuFun-TS modules and showed only poor-to-moderate longitudinal reliability, the module was excluded from the final NeuFun-TS deployment to minimize patient testing burden without compromising diagnostic sensitivity.
The Institutional Review Board of the National Institutes of Health provided ethical approval for this work, and participant consent procedures were completed. The authors declared no competing interests. The source indicates that all data produced in the study will be made available online upon acceptance by a peer-reviewed journal, with code and data accessible via the project repository: https://github.com/Bielekova-Lab/neufun-go-no-go