Cognitive Endpoints in Mitochondrial Disease
Digital cognitive tests may reduce trial burden, but their scores become interpretable only when they are anchored to domain-specific assessment and the noncognitive factors that shape timed performance.
Evidence
A newly published study and two independent full-text cohort studies examine cognitive measurement in primary mitochondrial disease from different angles. The new work asks whether shorter digital tests could serve as efficient research endpoints. The earlier studies show what broad batteries can reveal—and what a single score can hide—when disease phenotype, motor speed, baseline intellectual ability, and repeated testing influence performance.
The July 31 study compared digital and conventional assessments in two cohorts. The Newcastle cohort included 45 clinic patients; the KHENERGYZE clinical-trial cohort included 27 patients recruited across four European countries. Newcastle participants completed the Addenbrooke's Cognitive Examination and Montreal Cognitive Assessment alongside Cogstate and the Test of Attentional Performance. The ingested abstract reports strong correlations between conventional and digital measures. It also reports a high prevalence of moderate-to-severe perceived fatigue in both cohorts, reduced reaction times in more than half of participants, and associations between performance and factors including education and employment. Exact correlation coefficients and longitudinal responsiveness were not available in the abstract.
A 2019 prospective case-control study used a much deeper battery in 49 adults with m.3243A>G or m.8344A>G mitochondrial DNA variants and 32 controls. Participants were assessed at baseline, six months, and 18 months. The tests covered general cognition, verbal comprehension, perceptual reasoning, working memory, processing speed, executive function, and memory. Against population norms, 45% of patients scored at least one standard deviation below the mean on current full-scale IQ and 22% scored at least two standard deviations below it.
That comparison changed after baseline ability was addressed. Controls were matched within five years of age and within ten points on an estimate of premorbid full-scale IQ. Patients still performed worse on current general cognition and several domains, but they did not differ from matched controls on the memory tests. Motor speed also modified the processing-speed result: patients with faster motor speed did not perform worse than controls, whereas those with slower motor speed did. The estimated difference between premorbid and current cognition averaged 8.37 IQ points, yet patients and controls showed similar practice-related improvement over 18 months and no significant group-by-time interactions.
A separate 2020 study narrowed the phenotype to chronic progressive external ophthalmoplegia, or CPEO. It compared 28 patients at a Beijing center with 38 age- and education-matched controls. The battery covered global cognition plus executive, language, working-memory, memory, and visuospatial domains. Patients had lower Montreal Cognitive Assessment scores and lower Boston Naming Test scores. They were slower on Trail Making Test Part B, but did not make more errors under the study's corrected significance threshold. Working-memory, memory, and visuospatial measures did not differ between groups.
Within the CPEO group, Trail Making completion time was associated with non-ophthalmoplegia disease severity and disease duration, not with the ophthalmoplegia score. The regression model explained 57% of the observed variance. This finding argues against eye-movement limitation as the sole explanation for slower performance, but it does not isolate a purely cognitive cause. The participants also differed from controls in body mass index and a sleep-behavior questionnaire, and the study was cross-sectional.
Across the three studies, broad agreement exists at the level of measurement: cognitive difficulty is detectable even outside catastrophic central-nervous-system events, and speed-sensitive tasks repeatedly identify differences. The results are not direct replications. The studies sampled different genetic and clinical groups, used different instruments and comparators, and asked different questions about correlation, cognitive profile, and change over time.
Analysis — Speed Is a Composite Endpoint
The cross-study inference is that reaction time may be a useful mitochondrial-disease endpoint precisely because it is sensitive, but that sensitivity is also its main interpretive weakness. This is analysis rather than a conclusion jointly tested by the papers. The new digital study connects computerized measures with conventional assessments while flagging fatigue, education, and employment as influences. The 2019 study shows that motor speed can alter an apparent processing-speed deficit and that premorbid ability changes the interpretation of memory results. The CPEO study finds slow task completion without excess errors and links that speed to broader disease burden and duration rather than to ophthalmoplegia alone. Together, these patterns suggest that a faster trial endpoint is not automatically a more specific endpoint. A rigorous validation study would predefine the intended cognitive domain, measure motor function and fatigue on the same day, estimate practice effects, and test whether within-person digital change tracks an independently measured change over time. That design could separate a useful composite signal from a claim of isolated cognitive decline.
Limitations
The newest source was available only as a PubMed abstract, so its detailed eligibility criteria, genotype mix, missing-data handling, test-level estimates, and overlap between the Newcastle and trial cohorts could not be evaluated. Its correlations establish agreement between measures at observed time points, not responsiveness to disease progression or treatment.
The 2019 study was modest in size, restricted to two mitochondrial DNA variants, and lost 15% of patients by 18 months because of health deterioration. Its premorbid estimate was based on a reading test rather than a true pre-disease measurement, and repeated testing produced practice effects. The 2020 CPEO study was single-center and cross-sectional, excluded severe visual or hearing deficits that could affect testing, and did not establish longitudinal change. Its patients differed from controls on body mass index and sleep-behavior scores. None of the studies tested one locked digital endpoint prospectively across the same diverse population, and their numerical results cannot be pooled.