Assessing running stability: a systematic review of methods

Running stability is increasingly investigated in biomechanics and motor control research. However, the term is operationalized using diverse methods that may reflect different theoretical constructs, which complicates cross-study comparison and interpretation. This systematic methodological review aimed to identify, evaluate, and synthesize quantitative methods used to assess stability during running and to clarify the stability constructs represented by these methods.

The review was conducted according to PRISMA 2020 and PERSiST guidelines. PubMed, Web of Science, Scopus, and Google Scholar were searched for English-language studies published or available online from 2000 to the final search date of 17 December 2025. Eligible studies were primary experimental studies involving human running tasks, biomechanical data, and a quantitative stability-related outcome. Non-running tasks, studies without original human running data, non-quantitative studies, modelling or simulation studies, reviews, conference papers, animal studies, and robotic studies were excluded. Methods were synthesized qualitatively and categorized according to their conceptual basis. Reporting quality was assessed using a structured quality assessment tool.

After screening and eligibility assessment, 43 studies were included. Stability was assessed using only linear methods in 9 studies, only nonlinear methods in 20 studies, combined linear and nonlinear methods in 7 studies, and coordination-based methods in 7 studies. Linear methods mainly quantified the magnitude of variability or discrete features of movement execution. Nonlinear methods primarily characterized the temporal structure of variability through measures of trajectory divergence, signal regularity, temporal correlations, and return behavior. Coordination-based methods examined the functional variability of elemental variables with respect to task-relevant performance variables. Reporting quality was moderate overall, with limited reporting of reliability, the number of analyzed strides, and method-specific processing details.

The findings show that running stability is not represented by a single universal measure. Instead, different methods capture distinct aspects of movement variability, discrete features of movement execution, perturbation sensitivity, return behavior, or task-related coordination. Future studies should justify method selection according to the stability construct of interest and report key processing parameters to enhance reproducibility, interpretation, and comparability across studies.

This systematic review was registered on the Open Science Framework prior to study screening ( https://osf.io/e73xw ).