REF-1491 · Larsen et al., 2020 · Sports · Paper · PubMed · Free full text

Are CrossFit novices hurt more?

In 168 novices over 8 weeks, 9.5 injured people and 10.6 injuries per 1000 h, 2.66 for new injuries only. The comparison with experienced athletes is across studies, not controlled; 38% answered.

Main caveat: Response rate 38.3%. The authors say the results may over- or underestimate the rates; people who quit because of injury may be missing.
limited evidenceMeasured on recreational to trained athletes

Larsen and colleagues followed new members of CrossFit Copenhagen during "The CrossFit Experiment", an eight-week, free membership with two to three recommended classes a week, and asked how many were hurt. Of 439 who opened the email, 303 answered at baseline and 168 at the end. Exposure came from the booking system plus self-reported open gym: 2634.5 h.

Injury was a problem (pain, soreness, stiffness or swelling) that reduced participation for at least seven days, based on the OSTRC template. Twenty-five people had 28 injuries: 9.5 injured and 10.6 injuries per 1000 h; most were reported as provoked by a previous problem, and only seven were called new, which gives 2.66 per 1000 h.

No risk factor was significant, including introduction classes. The authors conclude that novices are at higher risk than experienced participants, whose injury rates they give from other studies as 0.74 to 3.3 per 1000 h. Our reading: this is a useful first-eight-weeks rate, not a controlled comparison.

The experienced figures are mostly retrospective, while prospective cohorts ran about 9.9 per 1000 h in a later pooled estimate (REF-1489, whose abstract does not name its five cohorts, so this one may be among them); so a difference in method could explain part of the gap, and the 2.66 for new injuries sits inside the experienced range.

None of this says the first weeks are safe either: one of the 168 reported medically diagnosed rhabdomyolysis.

Protocol

Design: prospective cohort with online surveys at baseline and eight weeks; injuries recorded only at the end, so the injury data are retrospective, as the authors note.
Setting: CrossFit Copenhagen, from 1 March 2018; 1500 one-hour sessions a week were available, each with a trainer. Voluntary introduction classes daily, and unsupervised open gym.
Who: n = 168 answered both surveys (51 men, 117 women; age 29.2 ± 7.9 y; BMI 24.3 ± 2.9); above 18 years, not current members. Previous primary sport: running 23.8%, fitness 20.2%, team sports 14.9%, cycling 7.1%, strength training 7.1%; 8.9% did not exercise previously.
Exposure: 2362 classes (median 14, IQR 7 to 21.5), 272.5 h open gym. 49 attended all three introduction classes, 72 one or two, and 47 none.
Injury definition: problem in a body region plus reduced participation for at least seven days. Sensitivity: the Weisenthal definition (total removal for more than 1 week, modified training for more than 2 weeks, or a visit to a health professional).
Analysis: rate = events / total exposure × 1000; univariate Poisson regressions for risk factors.
Funding and interests: "no external funding". Two authors (R.T.L. and A.L.H.) are part-time employees of CrossFit Copenhagen, which the authors state gave access to the data but had no part in design, analysis or writing.

Key Results

MeasureResult
Injured participants25 of 168 (14.9%), 22 with 1 injury, 3 with 2
Injuries28
Rate, injured participants9.5 per 1000 h
Rate, all injuries10.6 per 1000 h
Rate, injuries not provoked by an old problem (n = 7 injuries)2.66 per 1000 h
Sensitivity, Weisenthal definition21 injured, 24 injuries: 8.0 and 9.1 per 1000 h
Rhabdomyolysis, medically diagnosed1
SitesLower back 25%, knee 21.4%, elbow/hand 17.9%, other 17.9%, shoulder 7.1%, neck 3.6%, hip 3.6%, ankle 3.6%
Reported as acute75% of injuries
Risk factor (univariate)IRR (95% CI)p
Male vs female0.57 (0.22 to 1.53)0.266
Pain, soreness, stiffness or swelling beforehand0.87 (0.40 to 1.97)0.768
Class attendance, per class0.96 (0.92 to 1.01)0.144
Some introduction classes vs none3.04 (0.88 to 10.60)0.080
All introduction classes vs none2.56 (0.68 to 9.64)0.165
The comparison with "experienced" is not controlled. The range 0.74 to 3.3 per 1000 h comes from earlier, largely retrospective studies with other definitions. The study has no experienced group of its own.
Response: 168 of the 439 who opened the email (38.3%, the paper's figure); 303 answered at baseline, so 168 of 303 is 55.4% of baseline responders (our arithmetic); those lost were on average 2.4 years younger, otherwise similar except for education.

Mechanisms

Measured: injuries and exposure; no mechanism.
Authors' interpretation: novices may be at greater risk, as in two earlier surveys; introduction classes may attract the less active, which may explain why attending them looked no safer (an explorative association: active people were less likely to attend).
Inference: 21 of 28 injuries were put down to a previous problem, in people 52% of whom had pain or a similar problem in the two weeks before starting; for many, the first weeks may bring out an old problem and not create a new one. It was self-attributed.

Practical Application

Are CrossFit novices hurt more? In 168 novices over 8 weeks, 9.5 injured people and 10.6 injuries per 1000 h, 2.66 for new injuries only. The comparison with experienced athletes is across studies, not controlled; 38% answered.
What it gives the guide: a descriptive rate for the first eight weeks of unaccustomed mixed sessions (B, one provider), and a pointer to ask about old problems at the start: three-quarters of injuries were put down to one.
What it does not establish: that novices are more injured than experienced people, that a start-up programme causes injury, or that introduction classes do or do not help (not significant, wide CI). It does not test progression or a ramp.
Not from this paper; the base's own practice: 2-4 weeks at moderate load to learn the lifts (SYNTH-71, SYNTH-75) and, for pain under one station, remove that station's load first (SYNTH-71); slowing the start for an old problem is a suggestion made here, not a guide line.

Limitations

Response rate 38.3%. The authors say the results may over- or underestimate the rates; people who quit because of injury may be missing.
Recall of injuries at eight weeks, self-reported. No clinical assessment. Attendance recall matched the system (0.4 more classes, 95% CI −1.4 to 2.2).
No control or experienced group. The stated conclusion on higher risk needs one.
Underpowered for risk factors. The authors say so.
Numbers that do not add up. The abstract writes "28 injuries (14.9%)", but 14.9% is 25 people of 168; 28 of 168 would be 16.7% (our arithmetic). Table 2 lists n = 47 for one or two introduction classes and n = 72 for all, whereas the text gives 72 and 49, so the labels may be swapped.
Interests. Two authors work part-time for the provider that ran the programme.
How this entry was checked and changed
Why this level

small cohort (n = 168), with exposure registered objectively but injuries self-reported afterwards and a 38.3% response rate. Read from the full text (open access, CC BY)

Full text (checked 2026-10-30)
168 answered both surveys, a 38.3% response rate; 2634.5 h of training in total (full text)
Injury required at least seven days of reduced participation (full text)
25 injured people gave 9.5 per 1000 h, 28 injuries 10.6, and 7 injuries not provoked by old problems 2.66 (full text)
With the Weisenthal definition: 21 injured, 24 injuries, 8.0 and 9.1 per 1000 h (full text)
Previous studies of experienced participants: 0.74 to 3.3 per 1000 h, mostly retrospective (full text)
Introduction classes: IRR 3.04 (0.88 to 10.60), p = 0.080; none of the risk factors significant (full text)
No external funding; two authors part-time employed at CrossFit Copenhagen (full text)