REF-1490 · Mehrab et al., 2023 · International Journal of Sports Medicine · Paper · PubMed · Free full text
Is switching between prescribed and scaled loads risky?
One cohort of 406: people who alternated had higher injury odds than those always scaled (OR 3.5, CI 1.7-7.3). Limited evidence, an association, not proof that it causes injury.
Main caveat: One cohort carries both named risk factors. Szeles 2020, read here through this review, not directly.
Mehrab and colleagues searched seven databases to June 2021 for risk factors of CrossFit injury. From 9,452 records they included three prospective cohorts (Larsen 2020, Moran 2017, Szeles 2020) in a best-evidence synthesis and 12 cross-sectional studies as a source of new hypotheses; only two cohorts were at low risk of bias, so they did not pool.
They report strong evidence that weekly training days and participation in other sports do not affect injury risk, limited evidence that switching between prescribed and scaled loads is associated with more of it (one low-risk cohort) and that longer participation protects (the same cohort), and conflicting evidence on age, sex and previous injury.
The authors add that novices and those increasing load could use closer coaching, and say they cannot be sure this would reduce injuries. Our reading: the alternating finding is the one a progression guide would reach for, and it is also the weakest to act on.
Its comparison group is people who always scale, not people who always do the prescribed version; a person alternating may be moving up, backing off because something already hurts, or uneven across movements. It does not show that changing a load causes injury, and it says nothing of how much.
Protocol
Search: Embase, Medline, Web of Science, Cochrane, CINAHL, Google Scholar, SportDiscus, up to June 2021; 9,452 records, 5,129 after duplicates, 57 read in full.
Eligibility: CrossFit participants of all ages; at least one variable linked with injury; an injury definition including medical attention or time loss; prospective cohort or case-control with 20 or more participants. Cross-sectional studies were reviewed only to find new candidates.
Studies and participants: three prospective cohorts, n = 691 (172 sustained an injury), 315 men and 376 women, mean age 32 ± 8 y: Larsen 2020 (n = 168 novices, 25 injured, 8 weeks, Copenhagen, low risk of bias), Moran 2017 (n = 117, 14 injuries, 12 weeks, UK, high risk), Szeles 2020 (n = 406, 133 injured, 12 weeks, 13 gyms in one Brazilian city, low risk). 12 cross-sectional studies, 6,062 participants, 2,050 injured.
Quality: Quality in Prognostic Studies (QUIPS) for cohorts; cross-sectional studies predetermined at high risk. Weakest domains: participation (57% of points) and attrition (40%); no cohort gave a sample size calculation.
Synthesis: van Tulder best-evidence scale. Strong means two or more low-risk studies, generally consistent; limited means one low-risk study, or consistent findings in one or more high-risk studies. Pooling required at least three low-risk cohorts on an outcome, which was not met.
What "scaled" means there: Rx is the prescribed training load, and scaled is the alternative to it; the review defines it no further, and does not say how alternating was asked or how often it happened. In CrossFit usage, scaling means adapting a prescription to the athlete's ability (REF-1493).In its discussion it treats progressive scaling as adjusting a workout's difficulty on the go so a tired athlete can keep moving, which is not the same behaviour as switching between sessions.
Funding and interests: no conflicts declared; no funding statement printed. The first author wrote one of the included cross-sectional studies (Mehrab 2017).
Key Results
| Risk factor | Cohort(s) | Estimate (95% CI) | Evidence by the review's scale |
|---|---|---|---|
| Alternating Rx and scaled vs solely scaled | Szeles, n = 406 | Crude OR 3.6 (1.8-7.4); adjusted OR 3.5 (1.7-7.3), for CrossFit experience and previous injury | Limited |
| Rx vs scaled | Szeles, n = 406 | No difference | Reported in Table 4d, no estimate |
| Duration of participation | Szeles, n = 406 | Crude OR 0.8 (0.6-1.1), not significant; adjusted OR 0.7 (0.5-1.0), p<0.05 | Limited |
| Weekly training days, exposure | Larsen; Szeles | Szeles OR 0.9 (0.8-1.0), not significant | Strong: no effect |
| Other sports | Larsen; Szeles | No effect | Strong: no effect |
| Previous injury | All three | Szeles OR 3.0 (1.3-6.9), adjusted 3.2 (1.4-7.7); others no effect | Conflicting |
| Male sex | All three | Moran RR 4.62 (90% CI 1.32-16.10); others no effect | Conflicting |
| Age | All three | Szeles OR 0.998 (0.996-0.999); others no effect | Conflicting |
Cross-sectional direction on duration, opposite to the cohort: six studies linked longer participation with more injury and one with less. A cross-sectional survey counts injuries collected over longer exposure, so this does not cancel the cohort; it also does not support reading duration as protective with any confidence (our reading).
Not significant in any cohort: bodyweight, BMI, Functional Movement Screen, coaching involvement, stretching, preventive exercises, protective equipment; each "limited evidence" of no effect on the review's scale.
Mechanisms
Measured: self-reported associations over 8-12 weeks; no mechanism.
Authors' interpretation: training load and experience may matter in injury aetiology; novices and those increasing their load could benefit from closer supervision, and progressive scaling could be avoided in novices. They state that no prevention study tests this.
Inference: alternating may be a marker, not a cause: the person may be in transition, returning from a niggle, or strong in some movements and not in others. The review did not separate these, and the adjustment covers only experience and previous injury.
Practical Application
Is switching between prescribed and scaled loads risky? One cohort of 406: people who alternated had higher injury odds than those always scaled (OR 3.5, CI 1.7-7.3). Limited evidence, an association, not proof that it causes injury.
What it can be used for: a reason to look at why an athlete's load keeps moving up and down, not a rule against changing loads. Moving load is what progression is.
What the review does establish, on its own "strong" label: no link between days a week and injury in two cohorts of 8-12 weeks; that is short follow-up, not a licence for more days.
Does not test: a progression rule, a load jump, a rep scheme, stations or carries. No preventive intervention was studied. Injury definitions differed between studies (Appendix 4 of the review).
Not from this paper; the base's own practice: when the log shows a load going up and down without a plan, name the reason and hold one load for a few sessions.
Limitations
One cohort carries both named risk factors. Szeles 2020, read here through this review, not directly.
Short follow-up. 8-12 weeks. The authors want longer.
Heterogeneous injury definitions and methods. Between-study comparison is difficult, in the authors' words. No consensus definition of a CrossFit injury exists.
Observational, self-reported. Attrition and participation are the weakest domains.
Rounding of the protective finding. The adjusted CI ends at 1.0 and the crude estimate was not significant.
Reporting does not fully add up. Table 4a marks Szeles's age effect "=" beside OR 0.998 with p<0.05 while the text reports lower age as a risk; Table 4c marks demonstration of proper form "=" beside OR 0.30 (0.07-0.97), p<0.05. The discussion mentions relatively small effect sizes beside an OR of 3.5. The 172 total adds participants injured in two cohorts to 14 injuries in Moran.
English only. Selection bias named by the authors.
Interests. None declared; one author's own survey is among the cross-sectional studies.
How this entry was checked and changed
Why this level
a systematic review, but of observational cohorts, without pooling, and each key risk factor rests on a single cohort of 406; the two named findings are "limited evidence" by the review's own scale. Read from the full text (open access, CC BY-NC-ND 4.0)
Full text (checked 2026-10-30)
Three prospective cohorts, 691 participants, 172 with an injury, follow-up 8 to 12 weeks (full text)
Alternating between Rx and scaled vs solely scaled: OR 3.6 (1.8–7.4), adjusted OR 3.5 (1.7–7.3) for CrossFit experience and previous injury; n = 406 (full text)
Duration of participation: OR 0.8 (0.6–1.1) not significant; adjusted OR 0.7 (0.5–1.0), p<0.05 (full text)
Twelve cross-sectional studies, 6,062 participants, 2,050 injured; longer participation linked with more injury in six, less in one (full text)
Pooling required at least 3 low-risk cohorts; only two were at low risk, so no meta-analysis (full text)
Authors state there are no injury prevention studies on these factors, so they cannot be sure closer supervision would reduce injuries (full text)