Carvalho, AndreiaVanrenterghem, JosPataky, Todd C.Robinson, Mark A.Veloso, António P.Moniz-Pereira, Vera2025-12-152025-12-152026-01Carvalho A, Vanrenterghem J, Pataky TC, Robinson MA, Veloso AP, Moniz-Pereira V. How many strides are needed for reliable markerless gait analysis? Gait Posture. 2026;123:110020.0966-6362http://hdl.handle.net/10400.21/22319We gratefully acknowledge the contribution of all volunteers who participated in the study. We thank Lina Schelin and Alessia Pini for their availability to discuss the ICC R-code. Thanks to Paula Bruno for her insight into the statistical analysis. This work was supported by Fundação para a Ciência e a Tecnologia: Grant numbers DOI 10.54499/2020.07958.BD (PhD Grant) and DOI: 10.54499/UIDB/00447/2020, attributed to CIPER – Centro Interdisciplinar de Estudo da Performance Humana (unit 447).Background/aim: Markerless motion capture is an emerging tool for gait analysis. In some populations, e.g., older adults, traditional gait analysis poses practical challenges, such as reduced assessment tolerance, and the number of strides collected can be limited. This study aimed to investigate the influence of the number of strides on test-retest reliability and measurement error of markerless gait biomechanics in older adults. Methods: Twenty strides were extracted from 29 healthy older adults for each of two sessions. Lower-limb kinematics and kinetics were computed. Subsequently, non-consecutive random subsets of k = 2-19 strides were averaged within-subjects and within-sessions, including scenarios with unequal k between sessions. Integrated Intraclass Correlation Coefficients (iICCA,k) and Standard Errors of Measurement (SEM) were calculated for trajectory data. ICCA,k [Confidence Intervals] were computed for the range of motion and peaks. Two arbitrary thresholds for the minimally acceptable number of strides were combined: (1) the smallest k that yielded an ICC within 10 % of the maximum ICC across all k, and (2) an absolute ICC threshold of 0.75. SEM≤ 2º was deemed suitable for kinematics, and SEM%≤ 5 % for kinetics. Results: For joint angles and moments, iICC dropped less than 10 % from the highest iICC when using ≥ 7 strides with an equal number of strides across sessions, attaining iICCs≥ 0.75. Reducing the number of strides in one session had less impact than reducing both equally. Lower Confidence intervals were generally ≥ 0.75 for discrete parameters. Kinematic SEM rarely exceeds 2º. Globally, 4 strides are needed to have a kinetics SEM%≤ 5 %. Conclusion: A minimum of 7 strides contributing to the average observation is generally sufficient to achieve reliable markerless kinematics and kinetics in older adults. These results have particular relevance to populations who may experience limited tolerance for lengthy assessments. Allowing flexibility in stride number collected across sessions, while maintaining reliability, contributes to optimizing data collection strategies.engBiomechanicsIntraclass correlation coefficientOlder adultsStandard error of measurementTest-retestGaitjournal article10.1016/j.gaitpost.2025.110020