Oxidative stress markers have low repeatability: A meta-analysis and simulation study with implications for measuring physiological condition and fitness

Abstract Markers of oxidative stress are widely used as indicators of physiological state, but their utility for ecological and evolutionary inference remains uncertain due to high intraindividual variation masking associations with environmental conditions, fitness, and other physiological indicators such as telomere length. Although numerous longitudinal studies exist, individual repeatability of oxidative stress measurements is rarely reported explicitly. Here, we present the first meta-analysis assessing individual repeatability of oxidative stress markers, comprising 123 repeatability estimates from 22 studies. Overall, oxidative stress markers exhibited low repeatability on average (Intraclass correlation = 0.164), although repeatability estimates were highly heterogenous across biomarkers, study systems, and ecological contexts. Repeatability was generally low across taxa, sexes, study designs, and environments, although some markers, particularly lipid peroxidation quantified with HPLC, exhibited moderate repeatability in specific contexts. To illustrate the statistical consequences of low repeatability, we used heuristic simulations examining associations between oxidative stress and telomere length under different biological scenarios. These simulations demonstrate how low repeatability substantially reduces statistical power, but this is somewhat mitigated by taking repeated measurements. More broadly, our findings highlight the value of greater cross-disciplinary dialogue, as ecological and biomedical studies often use oxidative stress biomarkers in different ways and may benefit from shared perspectives.