Page 125 - 2022 Spring JSOM

Results FIGURE 2 Forest plot of studies examining injury prevalence
during high-intensity functional training (95% CI = 95% confidence
Figure 1 shows the number of publications included and ex- interval; std = standardized).
cluded at each stage of the literature review. There were 112
full-text articles examined; after reviewing these, 29 articles
were found to meet the inclusion criteria. One article exam-
30
ined a subgroup from a prior study; so only the study with
31
the larger sample size was used in the analysis.
31
Table 1 provides a summary of the methodology and data
collected on each of the 28 selected studies. The number of
17
participants in each study ranged from 6 to 3,049 with a
30
total of 11,098 participants in all studies combined. Twen-
ty-four studies 17,22–30,32–45 examined injuries during CrossFit
Training and only four 46–49 examined injuries during other
types of HIFT. There were 21 retrospective cross-sectional
studies 22–30,32,33,35–37,40,41,43–45,47,49 and 7 prospective cohort stud-
ies. 17,34,38,39,42,46,48 All but two studies 34,46 obtained injuries by
self-report. The two prospective cohort studies that did not
involve self-reports obtained injuries from medical records FIGURE 3 Forest plot of studies examining injury rates during high-
screening or injuries were recorded by coaches and research- intensity functional training (95% CI = 95% confidence interval; std
46
= standardized).
34
ers. Injury definitions varied widely, but eight studies 22–29
used a very similar injury definition, as mentioned earlier.
37
Injury reporting periods ranged from 6 weeks to 2 years,
48
with 11 studies 23,26,27,29,32,33,36,40,43,45,47 asking participants about
injuries at any time during HIFT. Average ages of participants
ranged from 19 to 40 years. The average ± standard devia-
34
48
tion proportion of men in all studies combined was 54 ± 13%.
32
Injury prevalence varied widely from a high of 73.5% to a
34
low of 12.0%. Injury rates also varied widely from a high
39
42
of 18.9 injuries/1000 h to a low of 0.04 injuries/1000 h.
The methodological quality assessment resulted in 20 studies
(71%) rated as “poor” and 8 (29%) rated “moderate” meth-
odological quality. Studies rated as poor generally had signifi-
cant risk of recall bias, lack sample size justifications, did not
assess potential confounders, and did not assess injuries prior prevalence effect size slightly enlarging it with a wider 95%
to initiation of HIFT. CI. Studies with similar injury definitions had a similar injury
prevalence effect size compared to the analyses with all studies
Figure 2 provides a forest plot and summary of the meta-anal-
ysis for all studies reporting on injury prevalence. The overall included (Figure 2); 95% CIs were wider, there was consid-
injury prevalence was 36.3% with 95% CI = 32.8–41.1%. The erable variability, and there was no indication of publication
Q-statistic and I indicated considerable variability in the prev- bias. For studies involving CrossFit training, injury prevalence
2
alence estimates. The Begg and Mazumdar rank order correla- effect sizes were similar to that of the analysis considering all
tion was not significant suggesting little publication bias and studies (Figure 2), with similar 95% CIs, considerable variabil-
the Trim and Fill Procedure suggested no studies were missing. ity, and no indication of publication bias. Participants training
in CrossFit-affiliated gyms had an injury prevalence effect size
Figure 3 provides a forest plot and summary of the meta-anal- similar to the analysis with all studies combined (Figure 2),
ysis for all studies reporting on injury rates. The overall injury considerable variability, and some indication of publication
rate was 4.29 injuries/1000 h of training with 95% CI = 3.35– bias. HIFT programs other than CrossFit also had similar in-
2
5.23 injuries/1000 h of training. The Q-statistic and I indi- jury prevalence effect size compared to the analysis with all
cated considerable heterogeneity in the injury rate estimates studies (Figure 2), considerable variability, and little indication
and the Begg and Mazumdar rank order correlation was sig- of publication bias. As the injury reporting period increased
nificant suggesting publication bias. The Trim and Fill Proce- from 6 months, to 1 year, to any time during CrossFit training,
dure suggested four studies were missing and when these were the injury prevalence effect sizes increased modestly with con-
imputed, the adjusted injury rate was 5.47 injuries/1000 h of siderable variability at all reporting periods. Among studies
training with 95% CI = 2.80–8.15 injuries/1000 h of training. with reporting period of 6 months there was some indication
of publication bias. The Trim and Fill Procedure suggested two
The top part of Table 2 shows the meta-analyses on specific studies were missing among investigations asking about inju-
moderator variables for injury prevalence. Prospective cohort ries any time during HIFT and when these were imputed, the
studies had a lower injury prevalence effect size than retro- adjusted injury prevalence was slightly higher with a wider
spective cross- sectional studies with considerable variability 95% CI.
for both types of study designs. The Begg and Mazzumdar Test
suggested little publication bias for either study design. The The lower part of Table 2 shows meta-analyses on specific
Trim and Fill Procedure suggested that three studies were miss- covariates for injury rates. Injury rate effect sizes were con-
ing for retrospective cross-sectional studies; imputing these siderably higher for prospective cohort studies compared to
missing studies only had a minor influence with the adjusted retrospective cross-sectional investigations, but there was

High-Intensity Functional Training Injuries | 123