Peripheral scope (exercise-programming), but the outcome is patient-important: a running-related injury (RRI) is an overuse injury that stops or modifies training for weeks (route-(c) harm, and the dropout that ends an exercise attempt). The decision this page answers: do baseline biomechanical and musculoskeletal (MSK) measurements — gait kinematics/kinetics, muscle strength, joint range of motion, lower-limb alignment, foot posture — predict who develops RRI, and should a runner be screened on them to prevent injury? The headline is a strong negative. (inferred from Peterson et al., 2022)
Headline — the null
A gold SR+MA of prospective studies only (the design that can support a risk-factor claim): «Thirty studies (3404 runners), testing over 100 discrete biomechanical and musculoskeletal risk factors for RRI, were included. Nineteen studies were pooled in twenty-five separate meta-analyses.» (Peterson et al., 2022).
The verdict: «Overall, results of the meta-analyses in this systematic review do not support the role of biomechanical and musculoskeletal measures as risk factors for the development of RRI in non-elite runners. Despite significant findings in several individual studies, twenty-three of twenty-five pooled analyses detected no relationship» (Peterson et al., 2022).
The decision-change — screening is not justified. «There is currently insufficient evidence for any musculoskeletal or biomechanical assessments as risk factors for RRI to recommend their use for preventative screening, clinically» (Peterson et al., 2022). Gait analysis, strength/ROM batteries and alignment screens, widely sold to runners as injury-prevention tools, are not supported by the prospective evidence as predictors of who gets hurt.
The two significant pooled effects — both clinically trivial
Only 2 of 25 pooled analyses reached significance, and the authors explicitly down-weight both:
| Measure | Effect (95% CI) | Studies | Read |
|---|---|---|---|
| Knee extension strength (lower in injured) | SMD -0.19 (-0.36 to -0.02), p=0.03, I2=0% | 4 | «may not be clinically important»; a «trivial effect size» |
| Hip adduction velocity, stance (lower in injured) | MD -12.80 deg/s (-25.22 to -0.38), p=0.04, I2=0% | 2 | «interpreted very cautiously»; both constituent studies individually null, one small/uncertain quality |
(Peterson et al., 2022) (Peterson et al., 2022)
The source’s own summary: «the associated trivial to small effects sizes suggest these findings should be treated with caution» (Peterson et al., 2022). Neither effect is a bare threshold; both are carried with their CI. The knee-extension CI’s upper bound (-0.02) nearly touches the null, and the hip-adduction pool rests on two studies that were each individually non-significant — a pooled significance that is fragile by construction.
Risk factor, not a modifiable lever — the epistemic boundary
These are predictors measured at baseline in prospective cohorts, not interventions. A null predictor result says baseline knee-extension strength does not identify who will get injured — it does not say strengthening the knee will not help an injured or deconditioned runner. The latter is an intervention question answered by Exercise Interventions and Sports Injury Prevention (where strength training does reduce sports-injury risk in RCT-MA). Do not launder this screening-null into a claim that strength or mobility work is useless; the two questions sit on different evidence designs and different causal objects. (inferred from Peterson et al., 2022)
Why the null — the load-capacity causal shift
Peterson reads the null through a causal framework in which biomechanics is not a standalone cause: «atypical biomechanical function alone does not cause RRI but may interact with training characteristics as an effect measure modifier to contribute to the risk of RRI», and RRI «occurs when a runner possesses multiple risk factors, and then participates in running under particular circumstances to a degree which exceeds a structure’s load capacity» (Peterson et al., 2022). The null for static biomechanics «is given some credence» by exactly this: biomechanical parameters were «considered, independent of participation-related variables» (Peterson et al., 2022). The modifiable signal, on this reading, lives in training load, not in the runner’s static anatomy -> Training Load Progression and Running Injury.
Relation to Ceyssens 2019 — expands, largely subsumes
Peterson cites Ceyssens et al. 2019 (ref [37]) and positions itself as building on it: «The results of the present review also expand on those of Ceyssens et al. [37] who performed a systematic review of biomechanical risk factors for RRI using a narrative synthesis, justifying the omission of meta-analyses based on between-study heterogeneity» (Peterson et al., 2022). Peterson covers the same gait variable families Ceyssens reviewed — running kinematics (9 studies), kinetics (8 studies), spatiotemporal parameters — but reports them in narrative synthesis because between-study heterogeneity precluded pooling, and adds meta-analysis of the poolable static/strength/alignment measures that Ceyssens did not pool. Peterson is newer (2022 vs 2019), higher-design (SR+MA vs qualitative SR), and reaches the same direction. (inferred from Peterson et al., 2022)
Transportability — the non-elite runner is not the obese novice
Peterson’s stratum is non-elite adult runners who are already running; the review excluded participants confirmed as competitive/elite where it could. This null does not transport to the sedentary, obese, or untrained novice: it says baseline biomechanical variation among established runners does not predict injury within that group — it is silent on whether an unconditioned joint or tendon in someone starting from inactivity carries a different risk profile (where the connective-tissue conditioning and impact-grading concerns of the fabric’s deconditioned-novice stratum apply). The support-factor profile differs. Treat the non-elite-runner null as evidence for the running stratum, not for the exercise-prescription decision of a beginner. (inferred from Peterson et al., 2022)
Synthesis — intrinsic screening is low-yield; load is the lever
Two independent prospective-evidence reviews converge on the same conclusion from disjoint author teams
and different injury scopes [E-independent]:
| Peterson 2022 | Wright 2015 | |
|---|---|---|
| Scope | all-cause RRI | stress fracture (an RRI subtype) |
| Factors | biomechanical + MSK measurements | all intrinsic risk factors |
| Verdict on biomechanical/anatomical measures | 23/25 pooled analyses null | a near-total absence of supported risk factors (verbatim on Stress Fracture Risk Factors in Runners) |
| What survives | 2 pooled effects, both clinically trivial | only prior stress fracture (~5x) + female sex (~2.3x) |
| Nature of survivors | trivial-effect measures, not actionable | non-modifiable markers (history, sex), not biomechanical |
Same-quantity check: both reviews test biomechanical/anatomical measures as predictors of running injury and both return null-to-trivial there — the overlapping quantity agrees. Wright’s two survivors (prior fracture, female sex) lie outside Peterson’s measurement scope, so they neither confirm nor contradict it; they are different variables, not a joined tension. The emergent claim is that intrinsic/static-trait screening for running injury is low-yield across injury types — the predictors that survive are either non-modifiable markers or clinically trivial, and the actionable, modifiable signal sits in training load, not in the runner’s anatomy -> Training Load Progression and Running Injury, and the prevention evidence that does work is interventional strength/neuromuscular training, not screening -> Exercise Interventions and Sports Injury Prevention. (inferred from Peterson et al., 2022)
Confidence: low — the direction is corroborated across two independent reviews, but Peterson’s own base is heterogeneity-limited (many variables unpoolable; measurement heterogeneity across overground vs treadmill, isokinetic vs hand-held dynamometry) and the surviving effects are fragile. The null is believed as a screening-is-not-justified decision, not as proof that biomechanics is causally inert.
The open loop
No operation here grades this against realized injury outcomes; the finding is that current prospective association evidence does not support biomechanical screening, which is a statement about the evidence, not a validated prevention protocol. (inferred from Peterson et al., 2022)