After reading a copy of the study titled “Injury Rate and Patterns Among CrossFit Athletes,” I wrote this post, which included a copy of a letter to the authors. A week later, Dr. Brian Giordano sent his response:
In our recent paper titled, “Injury Rate and Patterns Among Crossfit Athletes,” the use of the term “injury rate” was questioned. Specifically, the reader queried whether our use of the term “injury rate” referred to incidence or prevalence. We appreciate the reader’s thoughtful comments and agree that this point deserves further clarification.
In our study, we described injury rate as a measure of prevalence, not incidence, since we did not report exposure hours. The ratio we report is that of subjects with a new injury divided by the total number of subjects over a specified period of calendar time (6 months). As pointed out, this is not an incidence rate, but rather a prevalence rate. To be considered an incidence rate, the numerator would have been the same, but the denominator would be total exposure time of Crossfit.
We decided not report injuries in terms of exposure hours for several reasons. Because our survey collected data at a single point in time, we were unclear whether the participant was reporting time before or after the injury, which could lead to either an underestimate or overestimate of training time. Secondly, the variables we collected to calculate exposure were recorded in categories (time per session and number of sessions per week) and therefore are naturally imprecise. Lastly, these represent averages over time and are therefore subject to recall bias. While recalling whether the participant sustained an injury is also subject to recall bias, studies have shown that athletes are quite accurate when recalling whether they sustained an injury over a 12-month time period 1. Therefore, if we were to report our injury statistics in terms of injuries per exposure it would be subject to greater recall bias and therefore we would be less certain of its accuracy. Nonetheless, the purpose of the study was to ascertain whether injuries were incurred during Crossfit training and whether there were any identifiable risk factors which shared a relationship with specific injury patterns. Any associations could then be further scrutinized and injury prevention strategies could be implemented, and subsequently studied.
Of note, we did reference incidence at one point in our publication (Page 3, line 15), which states, “This was a single survey that recorded demographic and participant characteristics (ie, sex, age, experience, maximum lifts, and times) and the incidence and characteristics of injuries incurred over the 6 months prior to completing the survey.” This, however, was referring to the survey itself, as participants were required to record any new injuries they incurred during that period of time. The injury rates we report are a measure of prevalence, not incidence, as the new-onset injuries reported in the survey were not reported in terms of exposure hours. While our reported injury prevalence rate can be compared directly with some epidemiological sport injury study formats, other studies cited in our paper report injuries using exposure time. 2 Therefore, any direct comparison of our reported injury rate with other studies should be made judiciously due to these variations in methodology. This, however, should not undermine our conclusion that Crossfit athletes incur a variety of injuries during their training. In addition, how we reported the injury rate did not affect our secondary conclusions of trends and associations between injury rate and demographic categories, gym characteristics, and athletic abilities among Crossfit participants.
Again, we would like to thank the reader for their inquiry and allowing us to clarify this issue. We agree that this is a measure of injury prevalence rate, not incidence rates, and that the reader should be aware of this difference, especially when comparing our reported rate to that of other sports.
Brian Giordano, MD
Benjamin Weisenthal, BA
1. Gabbe BJ, Finch CF, Bennell KL, et al: How valid is a self reported 12 month sports injury history? Br J Sports Med. 2003;37(6):545-547.
2. Jacobsson J, Timpka T, Kowalski J, et al: Prevalence of musculoskeletal injuries in Swedish elite track and field athletes. Am J Sports Med. 2012;40(1):163-169.
I believe this sufficiently answers my first two questions, but it doesn’t answer my third question:
I can find no basis in scholarly literature for defining “injury rate” as the percentage of solicited survey respondents who report injuries. The most authoritative documents on the subject repeatedly define injury rate as a ratio of the number of injuries per unit of exposure time.
Given this information, how do you justify referring to your calculation as an “injury rate?”
It seems to me that referring to a measure of injury prevalence as an “injury rate” is simply a misuse of the term. I requested clarification on this third point and will post any further correspondence here.
Why does this matter? A percentage of injury is nearly meaningless without context. You need to know how much training the subjects did and for how long. Did they train for 6 weeks, 6 months, or 6 years?
For example, this study found a 19.4% rate of injury. Is that high or low? The logical conclusion would be to compare it to other studies. And Giordano et. al tried to do just that:
“Although the mechanism of injury would not be the same between running and CrossFit, running would be a reasonable logical alternative for exercise in the recreational athlete. As a reference, epidemiological studies on runners have reported injury incidence rates ranging from 19.4% to 79.3%. Elite track and field athletes, while a very different population but with similar high-intensity ballistic mechanisms, demonstrated a 42.8% 1-year retrospective injury prevalence, with gradual onset inflammation and pain being the most common type of injury.”
The casual reader might compare Giordano’s 19.4% rate with the rates found for runners above. But you can’t really do that. The track and field study found a 42.8% prevalence over 1-year of running. How long did the subjects in Giordano’s study do CrossFit for?
Well, Giordano refers to a 6-month period frequently, but in fact some of the subjects may have trained CrossFit for even less time than that: “The majority had been involved with CrossFit for 0 to 6 months.”
So, without a real injury rate, or even a solid time frame in which to express the injury prevalence, Giordano’s 19.4% figure isn’t useful.