Training-Load Change and Running Injury
Very limited evidence, and no support for a "10% rule"
Key Findings
Only 4 articles met the inclusion criteria; 3 of them found an association between an increase in training load and a higher risk of running-related injury:
1. runners who had recently made one or more changes in velocity and/or distance and/or frequency were more often injured than non-injured runners (p = 0.037)
2. increasing average weekly distance by more than 30% vs an increase of less than 10%: hazard ratio 1.59 (95% CI 0.96-2.66) — the CI crosses 1
3. injured runners had increased total distance significantly more in the week before injury than in other weeks (mean difference 86%, 95% CI 12-159%, p = 0.026)
BUT: no difference between a 10% and a 24% average weekly increase (HR 0.8, 95% CI 0.6-1.3) — a "10%" threshold is not supported
Authors' conclusion: there is only very limited evidence linking abrupt changes in training load to higher running-related injury risk (level of evidence 2)
Practical Application
Systematic review, 4 studies: sharp increases in running load are associated with injury in 3 of 4 studies (e.g. >30% vs <10% weekly increase, HR 1.59, CI crossing 1; +86% in the week before injury), but a 10% and a 24% weekly increase carried the same risk — the evidence is "very limited" and does not support a specific "10% rule".
Progress gradually and individually; treat any fixed percentage as a heuristic, not a rule, and the athlete's symptoms — pain, stiffness, a site that stays sore — as the actual gate (not from this paper; the base's own practice).
The abstract-level signal is about the size of the jump (large single-week spikes), not about a magic number: the rule of thumb the base uses is "avoid jumps of a third or more in a week" — inference from the >30% finding, not tested as a threshold.
After a break, restart conservatively (detraining §31); a spike is measured against what the tissues did recently, not what they once tolerated — not from this paper; the base's own practice.
ACWR and similar ratios are contested as metrics (Limitations); use them as a spike detector, not a target — not from this paper; the base's own practice.
Limitations
Only 4 eligible studies; heterogeneity of load/injury definitions; the strongest hazard ratio (1.59) has a confidence interval crossing 1
ACWR (acute:chronic workload ratio) as a metric is methodologically contested (not addressed in the abstract)
Does not provide a precise safe threshold — the conclusion is "very limited evidence", not "there is a rule"
How this entry was checked and changed
Content correction 2026-09-27
Entry said injury type depends on what increases (volume → Achilles/PFP vs intensity, "Run Clever RCT") -> not in this abstract (that is a different study); removed from findings and lead.
Added the review's actual numbers: 4 articles, 3 with an association; >30% vs <10% HR 1.59 (0.96-2.66); +86% the week before injury; no difference 10% vs 24% (HR 0.8); conclusion "very limited evidence".
House advice (symptom gate, restart after break, ACWR use) labelled as the base's own practice.