Does CrossFit work, by the evidence?
This 2018 review cannot say: 31 articles, 2 at low risk of bias, and 4 small studies pooled with nothing significant. That is missing evidence, not proof of no effect. No study of load management.
Claudino and colleagues searched PubMed, Web of Science, Scopus, Bireme/MedLine and SciELO in November 2016 using the keyword CrossFit, and included 31 articles after one retraction. Topics: body composition, psycho-physiological responses, injury risk, life and health, and psycho-social behaviour.
Only 6% of articles (two trials by Eather, both in adolescents) reached a high level of evidence at low risk of bias. Four short-term interventions gave 43 variables on endurance and body composition; none reached significance.
The authors summarise questionnaire studies as 19-74% and 1.9-3.1 per 1000 h (the 1.9 is shoulder injuries only); they say this may reflect sampling bias or inadequate management of training volume. They found no study of training-load management and say current load control rests on coaches' experience.
Their only positive signal is preliminary: a higher sense of community, satisfaction and motivation. Our reading: this review documents a thin literature at one moment. Its null is not a finding that CrossFit does not change fitness: a few small studies, split by sex, cannot show that.
Its lasting use for the base is the plain statement of the gap, that no study quantified, monitored or regulated load.
Protocol
Key Results
| Item | Result |
|---|---|
| Articles at low risk of bias | 2 of 31 (6%); both level 2 trials in adolescents |
| Pooled analysis | 4 articles, 43 variables (VO2max; body mass, BMI, fat, lean mass, waist); no significant result for any variable |
| Effect sizes | In Fig. 3 only; not available in the text we hold |
| Injury range in questionnaire studies | "between 19 and 74%" in the authors' summary, though the 19% study is described as a share of "all injuries"; 1.9 to 3.1 per 1000 h, the 1.9 from a shoulder-only survey |
| Injury and experience | One survey: 35% in those over 6 months vs 23% under 6 months (p = 0.004), prevalence, not a rate per hour, so longer exposure may explain it; it does not cleanly contradict REF-1491's conclusion (our reading) |
| Workload responses | Heart rate 54 to 98% of HRmax, lactate 6 to 15 mmol/L, RPE 8 to 9 across workouts |
| Training-load management | No study found |
| Psycho-social | Preliminary higher sense of community, satisfaction and motivation; not related to retention in one study |
Mechanisms
Practical Application
Limitations
How this entry was checked and changed
the review form earns more, but what it carries is mostly high risk of bias (mean CONSORT rating 37%), case reports and questionnaires, and the pooled analysis of four small interventions cannot show presence or absence of an effect; the search is also old. Read from the full text (open access, CC BY); the numbers in the forest plot are in an image (Fig. 3), not read