LAUSR

We read the recent meta-analysis [1] on the frequency and magnitude of in vivo head impacts in contact sports with great interest, as there is a lack of multi-sport head impact exposure comparisons in the literature. This is in large part due to the wide variance in methodologies used to collect head kinematics data and differentiate true head impacts from false impacts. These inconsistencies greatly limit the comparisons that can be made across sports and even across studies within the same sport. In their review [1], the authors compared the impact exposure for contact sports athletes in 21 different studies of various sports. The cited studies used a variety of sensor systems for measuring head kinematics. Sensor devices were installed within the helmet, on skinmounted patches, or in players’ mouthguard with a simple 10–15 g threshold used to record an acceleration event (AE). However, methods for distinguishing real head impacts from false positive AEs varied widely between studies; for example, only a third of the studies used video verification of impacts, while many considered all AEs over the 10–15 g threshold to be true impacts. This major limitation calls into question the key findings of this review and sheds light on a major issue plaguing this field of research. Although simple linear acceleration thresholds are commonly used to distinguish a head impact, this is likely inadequate as sensors can record a large number of false positives from a number of different sources such as sensor handling. This issue is amplified when using light-weight wearable head impact sensors that are easily manipulated. For example, our group has extensively developed and used an instrumented mouthguard sensor with a 10 g linear acceleration threshold. Using comprehensive video verification of each impact and non-impact from five different camera angles, we have found that the vast majority of AEs are false positives [2]. Our data suggest that an offensive lineman starter with an instrumented mouthguard may have over 1000 AEs recorded from a single game, with only 50 of these AEs being true head impacts, with other events triggered by mouthguard handling (before, during, and after the game) or other sources such as spitting or chewing. The number of false head impacts due to handling may be reduced for sensors integrated in more massive objects (e.g. helmet sensors) as these may be less sensitive to small perturbations. However, impact detection still remains a challenge for these devices. Idealized laboratory testing of commercial helmetmounted sensors reveals issues with false positives and false negatives, which may be compounded in the relatively complex and unknown impact conditions on the field [3, 4]. As such, depending on the sensor package and mounting, the number of AEs recorded over 10/15 g acceleration thresholds could contain varying numbers of false positives or negatives. In this case, the variations across studies and across sports may not only be from the exposure differences, but also from the variations in sensor impact detection performance. Given these shortcomings, it is difficult to interpret or compare exposure statistics reported without head impact verification, and we strongly recommend that any future head impact study which publishes exposure data uses video verification or a head impact detection algorithm validated This reply refers to the article available at https ://doi.org/10.1007/ s4027 9-019-01135 -4