To assess whether training-load measures from wearable inertial sensors are associated with musculoskeletal injury in athletes.
Approach:
Design: Systematic review and meta-analysis of longitudinal observational studies identified in six databases through June 2026.
Evidence: Included 11 studies involving 508 athletes across six sports cohorts and one military cohort; studies assessed cumulative load magnitude and acute-to-chronic workload ratio (ACWR).
Key Findings:
Across all 11 studies, higher sensor-derived training loads were associated with injury; the exploratory pooled effect estimate was 2.74, but mixed effect measures limited interpretation.
For cumulative load magnitude, five studies produced a pooled effect estimate of 2.86, with no statistical heterogeneity detected. The authors cautioned that the small evidence base limits interpretation.
For ACWR, five studies produced a pooled effect estimate of 2.64, but heterogeneity was substantial and the association was not statistically significant under the primary conservative inference method.
Results varied by effect measure: three studies reporting relative risks had a pooled estimate of 2.98, while the pooled odds-ratio result from five studies was not statistically significant.
All outcomes were rated very low certainty; none of the studies had low overall risk of bias.
Interpretation:
The authors reported a possible association between inertial-sensor-derived training-load measures and musculoskeletal injury, particularly for cumulative load magnitude. The evidence does not establish predictive performance, causality, injury-prevention effects, or specific decision thresholds. Cumulative load magnitude and ACWR were not directly compared within the same cohorts.
Limitations:
Studies were small (median sample size 45), with six rated as having some risk-of-bias concerns and five as high risk.
Confounding was generally inadequately controlled; some studies may have included overlapping cohorts.
Differences in sensors, exposure and injury definitions, and data processing limited comparability; publication bias could not be excluded.
Combining different effect measures made the overall pooled estimate exploratory.
Conclusion:
The review found a possible association between wearable-sensor training-load measures and musculoskeletal injury, but the evidence was insufficient to establish causality, identify athletes who will be injured, or recommend numerical thresholds.
