REF-1227 · Cloosterman et al., 2026 · Journal of Athletic Training · Paper · PubMed · Free full text

Does a jump in my acute:chronic load ratio predict injury?

Barely: in 461 recreational runners a duration-based ratio went with HR 1.33, and nothing held once runners already injured were left out. Not a risk score.

B+moderate evidenceMeasured on recreational to trained athletes

Cloosterman and colleagues used GPS training records shared by adult runners who had registered for 10 km to marathon events in the Netherlands and who took part in a randomized trial on injury prevention.

From each training session they computed the acute:chronic workload ratio (ACWR) by distance, by duration and by speed, and related it to the onset of a running-related injury with Cox regression. Of 461 runners (20,425 sessions), 42.3% reported a new injury during follow-up.

Unadjusted, a higher ACWR went with injury for distance (hazard ratio 1.32, 95% CI 1.00 to 1.74) and duration (1.39, 1.07 to 1.79); adjusted for confounders only the duration-based ratio remained (1.33, 1.02 to 1.74), and with runners injured at baseline (15.8%) excluded no ACWR was significantly associated with a new injury.

The authors conclude that the associations are small and inconsistent, that their clinical relevance is questionable, and that the role of general training-load advice in preventing running injuries is uncertain.

Our reading: a second large running cohort in which a weekly-style load ratio carries little signal; it fits the critique of the ratio in REF-365 and the null for week-to-week change in REF-1102, while REF-1102 found the risk in single runs longer than the recent longest. Abstract only: the ratio windows, the covariates and how injury was defined could not be checked.

Protocol

Design: prospective cohort within a randomized controlled trial on running-injury prevention; consenting runners received 3 follow-up questionnaires and were asked to share GPS data (abstract). Trial arm, follow-up length and how the trial intervention was handled in the analysis are not given in the abstract.
Participants: adult runners registered for running events of 10 to 42.195 km in the Netherlands; n = 461 with GPS data (72.9% men; age 43.8 ± 12.4 y); 20,425 training sessions.
Exposure: ACWR calculated from distance, duration or speed of each training session; the acute and chronic window lengths and whether rolling or exponentially weighted averages were used are not given in the abstract.
Outcome: onset of a running-related injury (RRI), from questionnaires; 42.3% reported a new RRI during follow-up; the injury definition is not given in the abstract.
Analysis: Cox regression, unadjusted and adjusted for confounders (which ones is not given in the abstract); sensitivity analysis excluding runners with an RRI at baseline (15.8%).
Affiliations and funding: Erasmus MC Rotterdam (general practice, orthopaedics and sports medicine, biostatistics), Move-Metrics (Ede) and a physiotherapy practice (record metadata); funding and conflicts not available (abstract only).

Key Results

ACWR based onUnadjusted hazard ratio (95% CI)Adjusted hazard ratio (95% CI)Excluding runners injured at baseline
Distance1.32 (1.00 to 1.74)not significant (value not given in the abstract)no significant association
Duration1.39 (1.07 to 1.79)1.33 (1.02 to 1.74)no significant association
Speedno association reported in the abstract (values not given)not given in the abstractnot given in the abstract
Incidence: 42.3% of the 461 runners reported a new injury during follow-up.
Unit of the hazard ratio: whether 1.33 is per 0.1 or per 1.0 of ACWR, or for a category, is not given in the abstract; the size of the effect per unit of load change therefore cannot be read.
Abstract tension: the Conclusions say "We identified an association between the ACWR and RRI onset in runners based on distance and duration respectively", while the Results say only the duration-based ratio remained after adjustment and none remained after excluding the baseline-injured.

Mechanisms

Measured: small positive associations between a higher ACWR and a new injury; the duration-based one survived adjustment for confounders (HR 1.33), the distance-based one did not, and no ACWR was significantly associated once runners injured at baseline were excluded.
Authors' interpretation: the hazard ratios are small and the outcomes inconsistent, so the clinical relevance is questionable and general training-load advice has an uncertain role in preventing running injuries.
Inference: the association lost significance without the baseline-injured; the abstract gives no hazard ratio for that analysis, so it may have shrunk or only lost power. If it shrank, runners carrying an injury may both train more erratically and report new injuries more, so part of the signal could be the injury driving the load pattern.
Inference: the ratio is an inaccurate, ambiguous metric that adds noise and creates statistical artefacts (REF-365), which would blur any real relation; a session-level spike measure (REF-1102) may be closer to how tissue is overloaded, but this study did not test one.

Practical Application

Does a jump in my acute:chronic load ratio predict injury? Barely: in 461 recreational runners a duration-based ratio went with HR 1.33, and nothing held once runners already injured were left out. Not a risk score.
Treat an ACWR number as a description of how load changed, not a forecast (B+ from the abstract, observational); this matches REF-365.
For runners, the single long run matters more than the weekly ratio: in the larger cohort of REF-1102 (B+, 5205 runners) the risk sat in one run more than about 10% longer than the longest of the previous 30 days, while week-to-week change showed no relation.
Already niggling? Without the runners injured at baseline the association was no longer significant (no hazard ratio given in the abstract); a current complaint is a stronger reason for caution than any ratio.
Cyclists: a running-injury cohort; do not quote it as a tissue-injury rule for riders (transfer).

Limitations

Observational: association, not cause; reverse causation through existing injuries is plausible (the association disappeared without them).
Abstract only: ACWR windows and calculation, covariates, the injury definition, follow-up length, the hazard-ratio unit, the speed-based results and the trial intervention's handling are not given; the numbers above could not be checked against tables.
Selection: runners willing to share GPS data from one prevention trial; mostly men in their forties.
Self-reported injuries by periodic questionnaire.
Running only: the load-injury link is bone-and-tendon mediated; transfer to cycling is weak.
Reporting: the Conclusions name distance and duration associations, the adjusted Results only duration.
Funding and conflicts: not available; one author's affiliation is Move-Metrics, Ede (record metadata); whether it bears on the study is not stated in the abstract.
Written from the abstract and record metadata (full text not read).
How this entry was checked and changed
Why this level

(from the abstract) — large prospective cohort (n = 461; the rubric's 200 or more); observational, injuries self-reported by questionnaire, and the ratio's window lengths, the model's covariates and the injury definition are not given in the abstract. The full text is not open (PMC embargo; the publisher page loads only through a script and was not read)

Content correction 2026-10-27
Correction 2026-10-27: pointers to SYNTH-29 aligned with its 2026-10-27 rewrite (single-run rule for runners; deload on signals first, a planned lighter week acceptable).