Answers are grounded in this paper and cite their sources.
Nutrition and exercise trials use laboratory markers in blood, muscle or cells to explain how interventions work. Whether a reported change reflects the intervention depends on the measurement error and on the level of inference. A change in one person must exceed about 2.8 times the scatter of repeated measurements; a group effect smaller than the scatter can still be estimated with enough participants. We reviewed method studies in eight selected measurement areas, from mitochondrial function to telomere length, to identify measurement problems, compare alternative methods and derive recommendations. Six problems recurred. Sample handling and within-person biology often contributed more variation than the analysis; for interleukin-6, the within-person coefficient of variation was 47–57%, against 6.9–10.0% for the assay. Test compounds reacted with some assay reagents. Analysis choices changed results from identical raw data. Change thresholds were not retrieved for phospho-immunoblots, autophagic flux in human muscle or satellite cell counts. Evidence adequate for group means was often inadequate for individuals. Some assays measured a related, not identical, quantity. In the areas reviewed, reliability data were dense where clinical decisions depend on the value and sparse for research endpoints. We propose a validation framework and a reporting checklist for testing.
No takes yet. Share an insight, caveat, or question.
Németh et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: