The Professional Athletes Brain Health study (PABHS) is composed of active and retired professional fighters (boxers and mixed martial artists), along with controls. Active fighters were required to have at least 1 professional fight within 2 years of enrollment and be training with the intent to compete. Retired fighters were included if they had been boxers or mixed martial artists, had a minimum of 10 professional fights, had no sanctioned fights for at least 2 years, and did not intend to return to competition. Control subjects were recruited from outreach efforts in the community and could not have any prior history of neurological disorders, head trauma, military service, or participation at a high school level or higher in a combat sport or a sport in which head trauma can be anticipated to occur, such as football, wrestling, hockey, rugby, soccer, or rodeo. Enrollment in the PABHS began in 2011 and has been continuous since then. Each participant is seen on an annual basis and, for active fighters, not sooner than 45 days from a sanctioned fight to reduce the potential acute effects of head impacts sustained in competition. Because of a variety of reasons (training and competition schedule, travel issues, other obligations), participants who missed a study visit were allowed to remain in the study with the next study visit conducted as soon as they were available. We consider the “baseline” blood levels as the ones that were drawn at the first study visit. Data for this study were collected between 2011 and 2018.

At each visit, blood sampling is obtained, along with a battery of other tests including MRI brain imaging, computerized cognitive testing, and exposure information. The PABHS was approved by the Cleveland Clinic Institutional Review Board, and written informed consent was obtained from all participants. Methods of recruitment and study procedures have been described previously [12].

Cognitive function was assessed by a computer-based battery consisting of four subtests of the CNS Vital Signs (CNS Vital Signs, North Carolina) including verbal memory, symbol digit coding, Stoop, and a finger tapping test. CNS Vital Signs offers robust and reliable measurements of cognition which are computerized but are supervised by a technician [13]. Results from these tests are used to make up scores in various clinical domains: verbal memory, processing speed, psychomotor speed, and reaction time. Raw scores were used in the analyses.

A high-resolution T1-weighted anatomical MRI was obtained on a 3 T MRI scanner (Siemens Verio from April 2011 through October 2015 and Siemens Skyra from December 2016 to the present) with a 32-channel head coil to acquire structural three-dimensional T1-weighted magnetization prepared rapid acquisition gradient echo images (repetition time msec/echo time msec, 2300/2.98; resolution, 1 X 1 X1.2 mm3. Volumes of the hippocampus, amygdala, superior temporal, various frontal regions, anterior cingulate, and total gray matter and subcortical grey matter including thalamus, caudate, and putamen, along with corpus callosum and total white matter volume, were calculated using the automated full brain segmentation process in the Freesurfer 6.0 software. These regions have been shown in pathological series and our prior work to be affected in those with extensive RHI [14, 15]. The volumes of each structure were measured in both hemispheres separately and an average volume calculated for structures that have bilateral representation. The regional volumes were adjusted for total intracranial volume (TIV) by adding TIV as a covariate. A quality control step was performed using the FreeSurfer’s quality analysis tools (https://surfer.nmr.mgh.harvard.edu/fswiki/QATools) to guarantee only data with high-quality cortical reconstruction from FreeSurfer were included in the analyses.

The blood samples were collected in EDTA tubes and centrifuged at 3200 rpm for 10 min to separate plasma from blood cells. The supernatant was aliquoted in 2 ml portions that were immediately frozen and stored at – 80° pending analysis. For all measured biomarkers (commercially available or in-house developed), plasma samples were allowed to thaw at room temperature for 45 min, after which they were vortexed (at 400 rpms for 30 s) and centrifuged (4000 g for 10 min). Internal quality controls (iQC) were included on each plate before and after the analyzed samples to determine inter- and intra-assay variability (intra- and inter-assay variation was < 15% for all biomarkers). All blood biomarkers were measured using a Simoa HD-X platform (Quanterix, Billerica, MA, USA) at the Clinical Neurochemistry Laboratory, Sahlgrenska University Hospital, Mölndal, Sweden. NfL and GFAP were measured using commercially available Simoa kits (NF-light™ #103,186 and GFAP #102,336, Quanterix, Billerica, MA, USA) following manufacturer specifications. In-house developed plasma p-tau231 and NTA tau are measured following published protocols [16, 17]. In brief, plasma p-tau231 assay is comprised by a mouse monoclonal antibody selective against phosphorylated tau at threonine 231 (ADx253, ADx Neuroscience) and a biotinylated mouse monoclonal antibody with epitope at aa 6–18 (Tau12, #806,501, Biolegend). Recombinant full-length Tau-441 phosphorylated in vitro by glycogen synthase kinase 3β (#T08-50FN, SignalChem) was used as the calibrator. The plasma NTA assay targets N-terminal tau fragments using two mouse monoclonal antibodies, with epitopes at aa 6–18 (Tau12, #806,501, Biolegend) and 194–198 (HT7, #MN1000, Thermo Scientific). Recombinant non-phosphorylated full-length Tau-441 (#T08-54N, SignalChem) was used as the calibrator.

Genotyping of apolipoprotein E (APOE) alleles was performed using real-time PCR restriction fragment length polymorphism analysis. Briefly, genomic DNA was collected from blood DNA extracted using Qiamp DNA blood maxi kit (Qiagen), and APOE genotyping was performed using Applied Biosystems TaqMan SNP Genotyping Assay.

STROBE reporting guideline was adhered to in preparing this manuscript.

The cohort was divided into four groups for analyses: active boxers, active MMA fighters, retired boxers, and controls. We chose to divide the active fighters by their fighting discipline because of prior findings from the PABHS that indicate active boxers show lower regional gray matter volumes and lower scores on cognitive tests after adjusting for number of professional fights and other factors compared to the active MMA fighters [15, 18]. We examined only retired boxers because we did not have enough retired MMA fighters in the cohort to analyze separately.

For the comparison of demographic data between the four groups in this study cohort, Kruskal–Wallis test was used for continuous outcomes (e.g., age, number of fights), and the chi-squared test was used for categorical outcomes (e.g., race). For continuous outcomes, we reported the interquartile range (IQR) values with the mean value. For APOE ε4 positivity, Fisher’s exact test was used for comparing the four groups [19].

We used linear regression models to compare the baseline blood biomarkers (GFAP, NfL, p-tau231, NTA) between the four groups after controlling for age, gender, race, education years, and number of fights (Fig. 1). Race was included as a covariate because of recent reports from the Alzheimer’s disease literature that suggest some fluid biomarkers may vary by race [20, 21]. In the PABHS cohort, race was determined by self-report and included White, African American, Asian, Pacific Islander, and American Indian/Native Alaskan. Those who did not designate a race were placed into the category of other. We chose to adjust for number of fights as previous findings from the PABHS indicate that this exposure measure itself predicts both cognitive and MRI volumetric outcomes [22]. In Fig. 1, the reported p-values are the ones after the multiple testing correction by using the Tukey’s approach.

Baseline levels of GFAP, NfL, p-Tau 231, and NTA Tau in control subjects, active MMA fighters, active boxers, and retired boxers (measured in pg/mL)

For the association between blood biomarkers at baseline and cognitive performance or MRI regional volume at baseline, linear regression models were performed with the covariates: age, gender, race, education years, and number of fights. Though we would have preferred to compare our fighter groups to the controls, we felt we did not have enough controls to match either the younger active fighter groups or the older retired boxer group. Furthermore, plasma levels of all the biomarkers we studied are known to be influenced by age. Thus, we performed within group analysis. When MRI regional volumes are the outcome of interest, another two covariates were added to the statistical model: scanner type and the total intracranial volume (TIV). These two covariates were also added in the following statistical models for repeated MRI regional volumes.

Participants were included in longitudinal analyses if they had two or more visits including blood biomarker, MRI, and cognitive data. Linear mixed effect models were used to assess the relationship between the longitudinal cognitive performance or MRI regional volume and each longitudinal blood biomarker data. As above, each group was analyzed separately. The outcomes are MRI regional volumes or cognitive measures. The fixed effects are blood biomarker, group, and their interaction, age, gender, race, education years, and number of fights. The fixed effect group is a categorical variable with 4 separate arms. The other two covariates, scanner type and the total intracranial volume, were added in the models for repeated MRI regional volumes. The correlation for outcome from the same participant is assumed to be the compound symmetry structure. We also ran linear mixed models to evaluate the relationship between baseline blood biomarker level and the longitudinal cognitive performance or MRI regional volume. The same fixed effects were included in the model.

We checked the statistical model assumptions by virtually inspecting the following plots: residual VS fitted value plot, Q-Q plot. The statistical software SAS was used in the analyses, and software R was used for some plots. All the tests are two-sided with the significance level of 0.05.