Peripheral scope (exercise-programming), but the outcome is patient-important: a running-related injury (RRI) ends the training attempt — route-(c) harm, and most consequential for the deconditioned / obese novice whom the fabric’s exercise-programming taper places in the fabric zone (where a wrong load ramp causes the injury that ends the attempt, not a marginal performance loss). The decision this page answers: does ramping running volume too fast cause RRI, and is the “10% rule” — add no more than 10% to weekly distance — evidence-based? The honest verdict is uncomfortable for a rule that reads as settled: the evidence is insufficient to confirm the causal claim, and there is no evidence at all for the 10% figure itself.
Headline conclusion: «Very limited evidence exists supporting that a sudden change in training load is associated with increased risk of running-related injury.» (Damsted et al., 2018)
Evidence-tier note (confidence: low). Single systematic review (Damsted 2018), tier high,
no meta-analysis — the included studies were too heterogeneous to pool: «due to the heterogeneity
observed in the study designs, the runners’ profiles, as well as the methods used for data collec-tion
and analysis of changes in training load, com-parison of the results of the four included studies in
the present systematic review must be performed with caution.»
(Damsted et al., 2018) Only four studies met eligibility, and
all involved «“novice” and “amateur” runners, which limits the applicability of the findings to other
populations of runners.» (Damsted et al., 2018) So the page
extracts the structure of the evidence, not a pooled magnitude.
The evidence structure — four studies, three point the same way, weakly
The SR found four eligible studies; three reported an association between a load increase and RRI, the directions consistent but each estimate individually weak or uncorrected: (Damsted et al., 2018)
| Study finding (as reported in the SR) | Estimate | Read |
|---|---|---|
| Recent change in velocity/distance/frequency, injured vs non-injured | p=0.037 | significant, but a chi-square the review authors ran themselves on raw proportions |
| Weekly distance increase >30% vs <10% | HR 1.59 (95% CI 0.96; 2.66) | directionally raised, CI crosses 1 — not significant |
| Total distance the week before injury vs other weeks | mean diff 86% (95% CI 12%; 159%), p=0.026 | significant within-runner, but susceptible to reverse causation (the pre-injury week) |
| Weekly increase 10% vs 24% | HR 0.8 (95% CI 0.6; 1.3) | null — no difference between the two |
- The direction is consistent but thin: «a tendency toward an increased injury risk following a sudden increase in training load was identified in three out of the four studies included…» (Damsted et al., 2018)
- For the Cantidio study the SR had to supply the statistics: «no result for statistical comparison was provided. Therefore, as all data were available to run a Chi-square test this was performed by the present authors…» (Damsted et al., 2018) — so one of the three supporting findings is a re-analysis, not an original result, which weakens the independent-study count further.
The 10% rule is unsupported — and the one study built to test it falsifies its special status
The SR is unambiguous: «no evidence exists for the use of the so-called “10% rule”, which is commonly used by runners, coaches and clinicians as a guideline for a maximum increase in training load per week.» (Damsted et al., 2018)
The data go further than unproven. The one estimate that directly tested the 10% boundary found 10% and 24% weekly increases indistinguishable — «no difference was found between a 10% and a 24% average increase in weekly volume (HR=0.8, 95% CI: 0.6; 1.3)» (Damsted et al., 2018) — and only the much larger >30% step showed even a (non-significant) signal. The authors’ own gloss: «An average weekly increase in training load of 24% may not be sufficiently large to reflect the mechanism of running too much, too soon in novice runners.» (Damsted et al., 2018) So the 10% cut has no demonstrated privileged status: a novice can apparently more than double it (to 24%) without a measured excess risk, and the hazardous zone — if one exists — sits well above 10%, nearer 30%, not at it.
Note this is within one study — the SR’s highest-quality study (100% methodological-quality score) produced both the 10%-vs-24% null and the >30% signal. That the same methods, same population and same analysis found 10% and 24% equivalent yet >30% elevated makes the refutation of the 10% figure a within-study contrast, not a fragile between-study artifact. (Damsted et al., 2018)
Underivable-optimum discipline ([[The Underivable Optimum]]). The 10% figure is a round-number
heuristic, not a located knee. The SR states the knee is unknown: «a well-defined threshold for
hazardous or sudden changes in training load above which the risk of injury development is significantly
increased is still unknown.» (Damsted et al., 2018) No curve was
fitted; the only datapoints are the dichotomies above (10% vs 24% null; >30% vs <10% HR 1.59, CI 0.96-2.66).
A threshold quoted as 10% is a bare number stripped of a CI and a studied range — exactly the false
precision the fabric forbids.
The relative-vs-absolute trap — why a novice’s small increase is large
A fixed percentage rule interacts badly with a low baseline. «a small increase in training load might result in a large relative sudden increase (e.g. increasing from 2 to 3 km per week represents an absolute difference of 1 km but a relative sudden increase of 50%).» (Damsted et al., 2018) This is the Baseline Risk and the Relative-Absolute Split logic applied to dose: for a true beginner running a handful of kilometres, trivial absolute additions are large relative jumps, so a percentage-based rule is most restrictive precisely where volume is lowest — and the studies that measure percent increase inherit this, making the exposure metric itself ambiguous across runners at different baselines. It cuts the other way for a high-mileage runner, for whom 10% is a large absolute load.
Evidence-state verdict — INSUFFICIENT, not null, with a weak directional signal
Applying the four evidence states (benefit / harm / no meaningful effect / insufficient) to the SR’s findings (inferred from Damsted et al., 2018):
- This is insufficient evidence, not a null. Very limited evidence with four heterogeneous, un-poolable studies, three directionally positive but individually weak, is the insufficient-evidence state — the expectancy test is not satisfied (a properly-powered prospective cohort with objective load measurement could show a gradient we have not yet seen). It is emphatically not a no-effect finding: the SR does not say ramping fast is safe, it says the evidence cannot yet adjudicate.
- The 10% rule specifically is closer to refuted-as-a-number: the one direct test found 10% and 24% equivalent, so the figure has no support. That is a decision-change — it licenses dropping slavish adherence to the 10% cut — without licensing the opposite (fast ramping is not thereby endorsed).
- The conservative default survives on other grounds. Three directionally-consistent studies plus a coherent tissue mechanism (below) make progress load gradually the low-regret choice even under insufficient evidence: over-caution on a hidden plateau merely forgoes a little fitness, while a too-fast ramp risks the injury that ends the attempt. The default is gradual progression, not the 10% number.
Mechanism bridge to tendon conditioning — consistent, but NOT a matched-parameter composite
The weak epidemiological signal here is consistent with the tendon-adaptation mechanism on Tendon Adaptation to Mechanical Loading: tendon adapts slower than muscle, so load ramped faster than the tissue can condition is an overload pathway. But the two sources measure different quantities, so this is a mechanistic bridge, not a type-F composite — the parameter table forbids treating them as the same claim:
| Parameter | Damsted (this page) | Bohm (Tendon Adaptation) | Same quantity? |
|---|---|---|---|
| Exposure | running training-load ramp (weekly % increase in volume) | muscle contraction intensity (~70% MVC/RM) per session | NO — rate of volume progression vs single-session load magnitude |
| Outcome | clinical running-related injury (whole-body, any RRI) | tendon mechanical surrogates (stiffness, modulus, CSA) | NO — patient-important injury vs tissue surrogate |
| Population | novice & amateur runners (mixed sex) | young recreationally-active men (237/264 male) | partial, not matched |
| Axis of the dose | how fast load rises over weeks | how hard a contraction loads the tendon | different axes (rate vs magnitude) |
So Bohm’s rate-mismatch (muscle outruns tendon) gives a plausible why for Damsted’s signal, and Damsted supplies the epidemiological counterpart Bohm lacks (Bohm measured no injury), but neither confirms the other’s number. Kept as convergence, not pooled evidence. (inferred from Damsted et al., 2018)
The separate causal-grade evidence that strength training reduces injury (RR 0.315, overuse almost halved) lives on Exercise Interventions and Sports Injury Prevention — a different lever (build tolerance) from this page’s question (manage the ramp), and a stronger evidence base.
Transportability — who was studied, and who was not
- Novices and amateurs only, so the findings do not transfer cleanly to competitive or high-mileage runners (for whom the relative-absolute trap inverts and 10% is a large absolute load). (Damsted et al., 2018)
- The exposure metric is self-reported training load across studies using «methods used for data collec-tion and analysis of changes in training load» that differed study to study (Damsted et al., 2018) — i.e. the ramp was measured heterogeneously and often by recall, so Measurement Error in Dietary Assessment’s logic applies by analogy: measurement error flattens a real dose-response, so a failed or weak gradient is weak evidence of no gradient. (inferred from Damsted et al., 2018)
- A broader training-error SR and an updated cohort would test whether the signal firms with better measurement: (the earlier training-error review this SR updates) and — both would bear on whether the >30% signal is real or an artifact of the pre-injury week.
Decision relevance
(inferred from Damsted et al., 2018) — the actionable reading of the SR’s findings:
- Do not treat the 10% rule as evidence-based. It is a round-number heuristic with no support; the one study testing it found 10% and 24% weekly increases equivalent in injury risk. Dropping rigid 10% adherence is a licensed decision-change.
- Still ramp gradually — but for mechanism and low-regret reasons, not the 10% number. Three directionally-consistent studies plus the tendon rate-mismatch mechanism make gradual progression the conservative choice; the number is undefined (the hazardous threshold, if any, is unknown and appears to sit nearer 30% than 10%).
- Watch the relative-absolute trap at low volume. For a true beginner, percentage rules are most restrictive where mileage is lowest; think in absolute steps a deconditioned body tolerates, not a fixed percent.
- This is insufficient evidence, not reassurance. The SR cannot say fast ramping is safe — it says the evidence cannot yet decide. For the deconditioned/obese novice (route-(c) harm, high stakes), the asymmetry favours caution.
- The stronger injury lever is building tolerance (Exercise Interventions and Sports Injury Prevention: strength training), a better-evidenced route than load-ramp management alone.
Sibling cell — the bone-stress-injury counterpart (distinction + convergence)
(inferred from Damsted et al., 2018) — the bone overuse injury (stress fracture) has its own risk-factor SR-MA on Stress Fracture Risk Factors in Runners (Wright 2016). It is a distinction, not a tension: this page’s exposure is the rate of change of load -> any RRI, Wright’s is absolute volume + host factors -> bone stress fracture (parameter table on that page). But the two converge on a structural claim neither makes alone: the popular training-load injury heuristics lack an evidence base — here the 10% rule (ramp rate), on Wright’s page the widespread anecdote that higher mileages cause stress fracture and that cutting volume prevents them (Wright’s own verbatim wording lives on that page). Two different load rules, both received wisdom without support — a type-A convergence, not a permission to ramp freely (both are insufficient-evidence verdicts).