HJBR Sep/Oct 2023

HEALTHCARE JOURNAL OF BATON ROUGE I  SEP / OCT 2023 29 lifetime RHI exposure. Cumulative expo- sure measures appear to better approxi- mate RHI exposure compared to single positions, likely because the former better accounts for the fact that football players change positions throughout their careers and often play multiple positions. Future measures should incorporate special teams participation (e.g. kickoff and punt), as this can vary by week and RHI from these posi- tions may not be otherwise incorporated. This study has several limitations. The study consists of a convenience sample of football playing brain donors, who tend to have greater exposure to RHI than the general population of football players. Even the athletes with lower years of expo- sure to football often had exposure to RHI from other sources, including other contact sports or military service, which were not characterized in this study. Along these lines, 82% of the subjects in this study played at the college level or above; predictions and thresholds from the present study would likely be most applicable to athletes with similar high levels of exposure. However, a substantial number of subjects had lower exposures ( n = 17 with only youth participa- tion and n = 95 only through high school), so we are not extrapolating to exposure ranges for which we have no data. There are several sources of poten- tial measurement error related to the cumulative RHI exposure measures: These measures were not observed directly while the donors were alive, but instead extrapo- lated from helmet accelerometer studies of other recent football players. Additionally, play style has changed over decades 57 , and many of the donors played decades ago, but the extrapolated data do not reflect these changes. Some studies reported means while others reported medians and we considered them equivalently. Also, the helmet accelerometer technology and the minimum thresholds for measured hits (e.g., 10 g vs 15 g) differed across studies. These studies also reported averages across all athletes at a given level, but donors in this cohort, particularly those that ultimately played professionally, likely had different RHI exposures than an average athlete. Additionally, accelerometers in helmets may move independently from the head and may not accurately indicate accelera- tion experienced by the brain 58 . Because the professional football leagues have not made their helmet sensor data available, we extrapolated collegiate athlete exposure to professional athletes. Furthermore, athletes play multiple positions even within a given season; these analyses averaged up to two positions per season but may not accu- rately reflect uneven distributions of play- ing time for a given athlete. Special teams participation could introduce measurement error, as impacts during kickoffs and punts, which tend to be high magnitude, were not incorporated into the PEM-derived cumu- lative exposure measures. Even with sev- eral sources of potential measurement error, measurement error can be overcome with a sufficiently large sample size, provided there is not bias in the estimation. We do not have reason to suspect that measurement error differed by neuropathological status to introduce bias. Given that we were able to find a robust relationship between the esti- mated cumulative measures of RHI expo- sure and CTE pathology in the absence of a reasonable explanation for bias, it seems this sample size was sufficiently large to overcome potential measurement error. There are additional limitations. CTE diag- noses and staging may have been obscured by comorbid pathology. However, we would not expect this to bias the reported results as the pathologists were blinded to the ath- letic history, upon which the RHI exposure measures were based. Concussion num- ber was also obtained retrospectively from informant report and validated with medi- cal records where available; given changes in concussion diagnoses over decades 57 , as well as recall bias and use of informant report, these reported concussion numbers may not accurately reflect concussion expo- sure. As a result, it is possible that a true rela- tionship between number of symptomatic