It's rare that I disagree with the AAP Grand Rounds editors about an article; this is one such instance where fatal flaws in study design render the conclusions suspect.
Source: Sakata J, Nakamura E, Suzuki T, et al. Efficacy of a prevention program for medial elbow injuries in youth baseball players. Am J Sports Med. 2018;46(2):460-469; doi:10.1177/0363546517738003. See AAP Grand Rounds commentary by Dr. William Hennrikus (subscription required).
The study in question was a nonrandomized controlled trial, a study design that ought to raise eyebrows from the start. For this study about an exercise program to prevent medial elbow injuries in youth baseball players, that's only the start of the problems.
Investigators, for reasons unclear to me, decided to use this nonrandomized design to look at an exercise program consisting of 9 stretching and 9 strengthening exercises to help prevent throwing injuries; previous studies had suggested some benefit for the exercise program. The authors recruited players from 33 different youth baseball teams from 2 independent regional leagues in a single city in Japan, excluding those who had prior elbow or shoulder pain. Without much explanation, the authors state that the participants were "purposefully allocated" to the exercise or control group based on "regional location." One-year follow-up looking at participants' diaries and interval clinical assessments were utilized to look at injury outcomes. The single examiner performing all the outcome measures was said to be blinded to the group assignment, though that can be difficult and open to question. (Example: if the examiner knows that each individual team's players all had the same group assignment, all it takes is 1 player or parent mentioning the exercise program to unblind the examiner to that group's assignment.) Another issue is whether different teams had different injury rates prior to the study's commencement. If "injury-prone" teams somehow ended up disproportionately in the control group assignment, this could be very misleading. I couldn't find anywhere in the study that the authors considered this.
Also remembering that the participants weren't blinded to their group assignment, this study design has significant risk for contamination, i.e. a player in the control group hearing about the exercise program and deciding to adopt the exercises, while still continuing to be "counted" in the control group. The authors state that "no information about the prevention program was available to the control group" because the regional leagues were independent of one another. I don't know much about Japanese youth baseball players, but it's hard for me to fathom that players from 1 league would never talk to players from another, especially over the year that the study was ongoing. The authors didn't make an attempt to determine if contamination occurred.
In spite of these and other study limitations, we can look at the bottom line results: the exercise group experienced medial elbow injury at a rate of 0.8/1000 athlete-exposures (an athlete-exposure is 1 athlete participating in 1 practice or game where risk of elbow or shoulder is possible) compared to a rate of 1.7 in the control group, a statistically significant difference. Lending support to that finding was the fact that measurements of physical function like increased joint flexibility correlated with a lower rate of injury, suggesting a mechanism for the stretching and strengthening exercises to have produced the better outcome. Also, participants in the exercise group who were more compliant with the exercise program had lower injury rates than those who were less compliant.
Sometimes an article's conclusions sound so logical readers might overlook issues with the study design that weaken those conclusions. This study has enough flaws that I wouldn't call it a controlled trial, better to relegate to an observational study design. Reading through the study I couldn't help but recall an infamous 1979 study on Kawasaki Disease, where 5 different treatment approaches were allocated based on which hospital a patient received treatment. That study found the rate of coronary aneurysms to be 20% in children receiving just an antibiotic, 64.7% in a steroid-treated group, and 11% in an aspirin-treated group. This study led to the conclusion that corticosteroid therapy was contraindicated in Kawasaki Disease, significantly delaying trials of steroids in Kawasaki patients. Of course, state of the art in study design in 1979 is a far cry from today's world of evidence-based medicine; I can't really fault those early Kawasaki Disease researchers.
Nonrandomized group allocation is almost always a bad idea. In the real world, it isn't always possible to blind study subjects and researchers as to treatment assignment, but cluster randomization is a better method to limit bias when individual subject randomization isn't feasible.
