HJAR Sep/Oct 2023

CTE 30 SEP / OCT 2023 I  HEALTHCARE JOURNAL OF ARKANSAS mean squared error (MSE) for CHII-R and CHII-G, followed by duration of play and then CHII. CHII-G and CHII-R better at classifying CTE severity than duration of play or CHII When restricted to individuals with CTE, the BIC provided very strong evidence for improved model fit for models estimating the relationship between CTE severity and either duration of play, CHII-G, or CHII- R compared with the model using CHII. Additionally, there was strong evidence for improved model fit for models using either CHII-G and CHII-R over duration of play alone. There was no evidence of differences in model fit for the model incorporating CHII-G compared to CHII-R. ROC analyses (Fig. 2B) found significantly improved performance in classifying CTE severity for models using CHII-G or CHII- R, but not CHII compared to models using duration of play alone. Similarly, there were significant improvements in classification using CHII-G and CHII-R compared to CHII. There were no significant differences between CHII-G and CHII-R in classifying CTE status. Similarly, 10-fold cross valida- tion analyses found lowest mean error for CHII-R and CHII-G, followed by duration of play and then CHII. Duration of play, CHII-G, or CHII-R preferred over CHII for modeling NFT burden In models adjusted for age at death, there was also a significant relationship between increasing NFT burden and increasing years of football played, in addition to all cumula- tive exposure measures (Table 6). The mod- els incorporating duration of play, CHII-G, and CHII-R, had higher R 2 s and showed very strong evidence for improved model fit (Table 6) compared to models incorpo- rating CHII. Sensitivity analyses: 1) after removing ath- letes who played multiple contact sports, 2) after removing athletes who had other neu- rodegenerative pathologies, and 3) without adjustment for age at death. While all athletes included in the above analyses played football, many had other exposures to RHI besides football; 26.9% of athletes participated in contact sports besides football and 23.4% served in the military (Supplementary Fig. 1). Similar rela- tionships as reported above were observed between exposure measures and CTE out- comes when restricted to athletes with no military history or other contact sport expo- sure besides football (Supplementary Tables 5 and 6). Additionally, some athletes had other neurodegenerative diseases besides CTE; 15.7% of athletes had Alzheimer’s disease, 7.6% had frontotemporal lobar degenera- tion, 16.1% had Lewy body disease, and 3.2% had motor neuron disease. Similar relation- ships as reported above were also observed when excluding athletes with any other neu- rodegenerative process (Supplementary Tables 5 and 6). Adjusting for age at death may remove some of the variance in the neuropathologi- cal outcomes that the RHI exposures mea- sures may explain. To investigate if this was the case, we re-ran our analyses without age in the models. Point estimates for the RHI exposure measures did not meaningfully change, but overall model fit and predictive power were mildly reduced (Supplementary Tables 1–3). DISCUSSION We investigated which components of RHI exposure from football play may be implicated in CTE pathology. The PEM reported here adapts the use of exposure matrices from other disciplines, and builds upon the previous use of CHII to quantify exposure to RHI. To our knowledge, this is the first study to use a PEM to estimate the relationship between different forms of RHI exposure and CTE pathology. This PEM is a tool for quantifying a foot- ball player’s lifetime exposure to RHI, by estimating the cumulative number of head impacts as well as the cumulative linear and rotational accelerations associated with those impacts. We demonstrated an asso- ciation between cumulative RHI exposure and CTE status, CTE severity, and NFT bur- den in football players. In general, model performance and fit using cumulative linear and rotational accelerations were improved compared to models with cumulative num- ber of head impacts and years of play. Addi- tionally, we found no relationship between informant-reported concussion number and CTE status, CTE severity, or NFT bur- den. Position played at the highest level was not associated with CTE status, CTE sever- ity, or NFT burden, with the exception of a