Page 129 - 2022 Spring JSOM

TABLE 4 Meta-Analyses of Anatomical Locations
Publication Bias Statistics
Effect Size for
Proportion of Duval & Tweedie Trim and Fill Begg and Mazumdar
Injured Locations, % Q-Statistic I 2 Trimmed & Filled Adjusted Effect Size Rank Correlation
Anatomic Locations (95% CI) p-value (%) Studies (n) (%) p-value
Shoulder 26 (23–29) <.01 62 2 25 (24–27) .79
Back/spine 26 (21–31) <.01 85 5 27 (25–28) .62
Knee 14 (11–17) <.01 69 0 — .53
Wrist/hand/fingers 12 (11–14) <.01 64 5 14 (12–15) .62
Arm/elbow 10 (9–11) <.01 88 7 14 (12–16) .92
Ankle/foot 6 (5–9) <.01 61 3 7 (6–8) .52

considerable variability for both types of studies; publication wrist/hands/fingers (12%), arm/elbow (10%), and ankle/
bias was present for both prospective cohort and retrospec- foot (6%). However, it is very important to note that in all
tive cross-sectional investigations. For prospective cohort in- meta-analyses conducted here there was considerable variabil-
2
vestigations, the Trim and Fill Procedure suggested one study ity as indicated by the Q-statistic and I . I was 95% for over-
2
was missing and imputing this one considerably increased the all injury prevalence and 100% for overall injury rate. Injury
injury rate effect size and 95% CI. Studies involving similar prevalence ranged from 12% to 74% and injury rates from
injury definitions had an injury rate effect size lower than that 0.04 to 18.9 injuries/1000 h of training. It was expected that
of the analysis with all studies (Figure 3), with a wider 95% the wide range of reporting periods and differences in injury
CI, considerable variability, and indication of publication bias. definitions might at least partly account for this, but when
CrossFit investigations had a lower overall injury rate effect studies with these factors were considered separately the large
size compared to the analysis with all studies included (Figure amount of variability remained. The majority of studies (71%)
3), considerable variability, and indication of publication bias; had very low methodological quality primarily due to risk of
the Trim and Fill Procedure suggested 4 studies were missing recall bias, lack of sample size justifications, and insufficiency
and imputing these considerably increased the overall injury in accounting for potential confounders. Most interesting was
rate effect size and 95% CI. Participants training in Cross- the fact that the generally higher-quality prospective cohort in-
Fit-affiliated gyms had a higher injury rate effect size than that vestigations had an injury rate that was more than double that
of the analysis with all studies (Figure 3) with considerable of the generally lower-quality retrospective cross-sectional in-
variability and an indication of publication bias; the Trim and vestigations (discussed more below).
Fill Procedure suggested three studies were missing and im- HIFT involves wide variety of physical activities, but running,
puting these resulted in a considerably higher estimated injury resistance training, and gymnastic movements largely predomi-
rate effect size and 95% CI. HIFT programs that were not nate. 50,51 Resistance training activities incorporate a wide range
CrossFit also had a higher injury rate effect size compared to of exercises including those used in powerlifting (squat, dead-
the analysis with all studies (Figure 3), with considerable vari- lift, and bench press), weightlifting (snatch and clean and jerk),
ability and little indication of publication bias.
and other types (e.g., overhead press, push press, barbell row,
Table 3 shows the 18 studies that provided the number of inju- etc.). Gymnastic exercises use rings (e.g., ring muscle ups), bars
ries by anatomical locations. There were six major anatomical (e.g., bar muscle ups), and the floor (handstand walk). It is
51
locations that could be analyzed as shown in Table 3. These useful to compare the HIFT injury data in the current review
six locations accounted for an average ± SD of 83 ± 13% of with that of other similar exercise activities to see if injury prev-
all injuries. It should be noted that some studies only reported alence and rates are similar. Injury incidence in running has
one injury for each person while other studies allowed more varied widely, with a systematic review indicating incidences
than one injury per person. Table 4 shows the meta-analysis ranged from 26% to 92% when all anatomical locations were
on the effect sizes for injury proportions (%) at each anatom- included. A more recent review that included 31 individual
52
ical location. The shoulder and back/spine both accounted for studies found a mean ± SD injury prevalence of 45 ± 13%. A
53
about 26% of injuries, followed by (in order of decreasing in- meta-analysis of injuries in runners found overall injury rates
jury proportion) the knee, wrist/hands/fingers, arm/elbow, and of 17.8 and 7.7 injuries/1000 hours of training for novice and
ankle/foot. There was considerable variability in injury pro- recreational runners, respectively. For weightlifting activities,
54
portions for all locations. The Begg and Mazumdar statistic injury prevalences have varied from 16% to 90% and rates
indicated little publication bias at any location, but the Trim have varied from 2.4 to 3.3 injuries/1000 h. For powerlift-
and Fill Procedure suggested some studies were missing. How- ing, an injury prevalence of 43% has been reported (only one
ever, after imputing these missing studies the adjusted values study) with injury rates varying from 1.0 to 5.8 injuries/1000
were similar to the unadjusted ones at all anatomical sites. h. 55,56 Secondary meta-analyses of the weightlifting data in
one review article produced an overall injury prevalence of
55
42% (95% CI = 15–75%, I = 98%, Begg and Mazumdar
2
Discussion
p-value = .31) and injury rate of 2.86 injuries/1000 h (95% CI
This review and meta-analyses suggested that the overall in- = 2.35–3.38 injuries/1000 h, I = 98%, Begg and Mazumdar
2
jury prevalence during HIFT was 36% (95% CI = 32–41%) p-value = .06). For powerlifting, secondary meta-analysis on
with an injury rate of 4.29 injuries/1000 h of training (95% data in this same review produced an overall injury rate of
55
CI = 3.35–5.23 injuries/1000 h). The most commonly injured 3.58 injuries/1000 h (95% CI = 2.70–4.47 injuries/1000 h, I =
2
anatomical locations (with % of total injuries in parentheses) 100%, Begg and Mazumdar p-value <.01). Systematic reviews
were the shoulder (26%), back/spine (26%), knee (14%), on gymnastic injuries report weighted average injury rates of
High-Intensity Functional Training Injuries | 127