Oral Contraceptives, Concussion History, and Contact-Sport Exposure: Multimodal MRI and Brain Injury Biomarkers in Female Athletes

Abstract

Neuroimaging and blood-based biomarkers are increasingly utilized to identify long-term effects of concussion and contact-sport exposure (CSE), but female athletes are underrepresented in this work, despite potential sex-related differences. Early research indicates that widely used oral contraceptives (OCs) influence brain structure and function and concussion-related biomarkers. This study investigated whether current oral contraceptive (OC) use moderates associations of concussion history and CSE with magnetic resonance imaging (MRI) measures and brain injury biomarkers in female athletes. Part of a larger cross-sectional study, 68 collegiate-aged female athletes (30 OC users, 38 with no hormonal contraceptive use [NoHC]), completed interviews to retrospectively determine the number of prior concussions (last injury at least 6 months before study enrollment) and years of CSE. Serum glial fibrillary acidic protein, neurofilament light, and total tau levels were measured, along with hippocampal volume, cortical thickness, and Neurite Orientation Dispersion and Density Imaging metrics. Linear regression and Bayesian multilevel modeling tested interactions between OC use and concussion history or CSE on outcomes. Number of prior concussions was negatively associated with hippocampal volumes and cortical thickness. OC use moderated relationships between CSE and cortical thickness, isotropic volume fraction (ISO), and intracellular volume fraction, and between concussion history and orientation dispersion index (ODI). OC users had lower tau levels than the NoHC group. Exploratory models suggested OC pill formulation may moderate associations of concussion history and/or CSE with outcomes and directly affect cortical thickness, ODI, and ISO. These results suggest OC use deserves consideration as an influencing factor in future research and personalized treatment for female athletes.

Introduction

Concussion, a subset of mild traumatic brain injury (mTBI), and repetitive head impacts from contact-sport exposure (CSE) are major public health concerns with potential long-term implications.^1,2 Fluid-based biomarkers and advanced neuroimaging are increasingly used to study the acute effects of concussion and CSE, whereas research on cumulative effects remains more limited.^1,3,4

Emerging evidence demonstrates that neuroimaging measures and brain injury markers are associated with both history of concussion and cumulative CSE. Most consistently, concussion history and CSE are associated with smaller hippocampal volumes,^5–16 and with thinner cortex (e.g., in frontotemporal cortices).^5,17–20 Diffusion MRI studies also demonstrate differences (e.g., higher fractional anisotropy) associated with concussion history and CSE in white matter tracts, including the corpus callosum, cingulum, and uncinate fasciculus.^10,21–26 In addition, increased glial fibrillary acidic protein (GFAP),^27–29 neurofilament light (NfL),^28,29 and tau^30,31 have been linked to CSE.

While this prior work has provided important information regarding potential cumulative and chronic effects of repeat concussion and CSE, the underrepresentation of female athletes is pervasive throughout this literature.^1,3,4 This is particularly problematic, as female athletes incur concussions at higher rates than male athletes in comparable sports.^32,33 Additionally, there is evidence that sex influences associations of concussion history and/or CSE with physiological outcomes, as well as evidence for underlying sex differences in several physiological markers that are thought to be sensitive to neurotrauma (e.g., GFAP, tau, NfL, diffusion MRI).^25,34–41 Thus, studies characterizing the effects of repeat concussion and CSE in female athletes are needed.

Furthermore, evidence that concussion outcomes may differ based on hormonal contraceptive use at the time of injury^42,43 warrants the inclusion of hormonal contraception information when studying concussion in female athletes. Hormonal contraceptives are widely used, with almost 50% of female collegiate athletes and 80% of sexually experienced women in the United States reporting oral contraceptive (OC) use.^44,45 By design, the exogenous ovarian hormones in OCs alter endogenous hormone levels. ⁴⁶ Moreover, both hormonal contraceptive use and hormones have been associated with changes in brain regions with gonadal hormone receptors, impacting hippocampal volume,^47–49 cortical thickness,^50–53 diffusion MRI metrics,^49,54 and GFAP, NfL, and tau.^{29,41,55–58} Additional evidence suggests that these brain changes may be dependent on OC pill formulation,^50,59–61 which can vary in ethinyl estradiol dose and synthetic progestin dose and type. Despite the overlap in the influence of concussion history, CSE, and hormonal contraceptive use on brain injury markers and structural neuroimaging, there is very little research on concussion history/CSE-related effects on these outcomes that includes information on hormonal contraceptive use. To fully understand the chronic and evolving effects of concussion history and CSE, we must understand how athlete-specific factors, such as hormonal contraception, may exacerbate or confound these potential associations with brain health.

The current study begins to address the knowledge gap of the impact of OC use on associations of concussion history and CSE with brain structure and brain injury markers for female athletes. Our primary hypothesis was that OC use moderates associations of concussion history and CSE with brain injury markers (GFAP, NfL, tau), hippocampal volume, cortical thickness, and advanced diffusion MRI (Neurite Orientation Dispersion and Density Imaging—NODDI) measures in female athletes. Exploratory work investigated the role of OC pill formulation in these relationships.

Methods

Collegiate-aged athletes recruited from the community were enrolled as part of a larger study examining the effects of prior concussion and CSE (Project ChronEx), and eligibility criteria are described again here. ¹⁰ Eligible individuals were 18–26 years old and participated in a collegiate sport within the last 12 months. The following enrollment exclusions applied: (1) self-reported psychiatric, neurodevelopmental, neuroendocrine, or severe neurological or autoimmune disease diagnosis; (2) prior moderate/severe TBI; (3) excessive alcohol/substance use as determined by a score of two or greater on the cut down, annoyed, guilty, eye opener Questions Adapted to Include Drug Use screener; (4) neuroactive medication use; (5) concussion within 6 months of enrollment or incomplete recovery from prior concussion; (6) contraindication to study procedures; (7) injury/illness precluding participation in study protocol. All aspects of this study were approved by the Medical College of Wisconsin Institutional Review Board, and participants provided written informed consent. The study aimed to enroll contact- or combat-sport athletes during their off-season. Only athletes self-identifying as female (given the choices female, male, unspecified/transgender, unknown) were included in this portion of the study.

Assessment of prior concussion and contact-sport exposure

Semi-structured interviews collected details for any head or neck injuries that participants recalled throughout their lifespan to retrospectively determine the number of prior concussions, as previously reported. ⁹ Injuries designated as concussions met criteria consistent with the 1993 American Congress of Rehabilitation Medicine (ACRM) definition of mild traumatic brain injury and World Health Organization recommendations (loss of consciousness [≤30 min], post-traumatic amnesia [≤24 h], retrograde amnesia, and/or alteration in mental status [i.e., being dazed or confused]).^62,63

Information on current and prior participation in soccer, rugby, and combat sports with routine head impacts was collected in a similarly structured interview to estimate cumulative years of CSE across the lifespan. ⁶⁴

Contraceptive use

As described in our previous work, participants provided the dosage and brand name of any birth control or contraceptive used (OC, transdermal patch, vaginal ring, shot/injection, intrauterine device or IUD, other implanted device with hormones [e.g., arm implant], abstinence, other) for any reason, including irregular periods. ⁶⁵ The current study focused on participants reporting OC use and those reporting no hormonal contraceptive use (NoHC). Subjects from the parent study who reported use of alternative hormonal contraceptives or Cu IUDs were excluded. Additionally, since OC pill formulation may be important when considering the impact of OC use,^50,59–61 exploratory analyses based on the hormone activity levels (i.e., estrogenic, progestational, and androgenic) of OC pill formulations according to determinations in Dickey ⁶⁶ were conducted to address the heterogeneity of OC use and are detailed in the Supplemental Material.

Blood sample collection and analysis

Blood samples were collected according to processes previously described for Project ChronEx. ¹⁰ Morning, fasting venous blood was collected using Red Top BD Vacutainer tubes. Blood was left to clot at room temperature for 30 min, then centrifuged at 1,500 relative centrifugal force for 15 min; the serum was aliquoted and stored at −80°C. Concentrations of GFAP, NfL, and tau were measured in duplicate using Meso Scale Discovery S-PLEX assays following manufacturer instructions. The average coefficient of variation was 3.65% for GFAP, 4.30% for tau, and 5.20% for NfL. All samples were detectable. Markers with a coefficient of variation greater than 25% were removed (n = 1 GFAP, n = 2 NfL). The ComBat harmonization method was applied to reduce plate effects for GFAP, NfL, and tau markers, implemented in the R “sva” library from the Bioconductor package (R version 4.5.0).^67,68 All markers were natural-log-transformed prior to ComBat adjustment to reduce distributional skewness. Separate ComBat harmonizations were run for each statistical model, with the applicable covariates, as recommended. ⁶⁸

MRI data collection

As previously described,^9,10,69 MRI data were collected on a GE HealthCare (Waukesha, WI) SIGNA Premier 3T scanner using a brain-dedicated receive-only 32-element coil array (Nova Medical, Inc., Wilmington, MA). A three-dimensional (3D) magnetization-prepared rapid gradient echo (MPRAGE) pulse sequence was used to collect the T1-weighted image set with field of view (FOV) = 240 mm, image matrix = 240 × 240, 188 axial slices, 1-mm slice thickness, voxel size 1 × 1 × 1 mm³, repetition time (TR) = 2,279–2,356 ms, echo time (TE) = 2.84–2.97 ms, inversion time = 900 ms, and flip angle = 8°. The T2-weighted image set was acquired via a 3D CUBE pulse sequence with the same FOV, image matrix, number of axial slices, slice thickness, and voxel size, with TR = 3,202 ms, TE = 86.34–93.14 ms, and flip angle = 90°. A single-shot spin-echo EPI was used to collect a multi-shell diffusion-weighted image series with TR = 7,500 ms, TE = 74 ms, FOV = 220 mm, image matrix = 110 × 110, 62 axial slices, 2-mm slice thickness, voxel size 2 × 2 × 2 mm³, flip angle = 90°, and in-plane ASSET acceleration of 2. Ninety diffusion directions, sampled evenly across three b-values (700, 1,000, and 2,800 s/mm²), with six nondiffusion-weighted reference volumes, were collected.

Hippocampal volumes and cortical thickness

Cortical reconstruction, tissue segmentation, and hippocampal segmentation of the T1-weighted images were performed using FreeSurfer 7.2 (https://surfer.nmr.mgh.harvard.edu/), with the T2-weighted images included to improve pial surface estimation and segmentations, ⁷⁰ as previously described. ¹⁰ Bilateral whole hippocampal volume was the primary structure of interest. Intracranial volume was quantified from the T1- and T2-weighted images using Samseg in FreeSurfer 7.2. ⁷¹ Mean cortical thickness was extracted for the following regions of interest (ROIs) based on the Desikan–Killiany Atlas parcellation, ⁷² chosen based on associations with concussion history, CSE, and OC use^52,53,69: bilateral lateral and medial orbitofrontal cortex, pars triangularis, insula, superior and middle temporal gyri, temporal pole, entorhinal cortex, parahippocampal gyrus, rostral and caudal anterior cingulate cortex, and posterior cingulate cortex.

NODDI measures

As in previous Project ChronEx work, ¹⁰ multi-shell diffusion data were corrected at the scanner for susceptibility-induced distortions using the GE PROGRES feature. FMRIB Software Library’s (FSL; version 5.0.11) eddy_openmp corrected for eddy currents and head motion using the replace-outlier option.^73–75 A brain mask was extracted from a distortion-corrected nondiffusion-weighted image volume (b0) by FSL’s brain extraction tool. ⁷⁵ NODDI model metrics (i.e., orientation dispersion index [ODI], isotropic volume fraction [ISO], and intracellular volume fraction [ICVF]) were calculated using the Watson distribution in the CUDA Diffusion Modelling Toolbox (cuDIMOT; FSL version 6.0.4).^76,77 Diffusion metrics were transformed to standard space using Analysis of Functional NeuroImages (AFNI; version 18.3.08) programs. ⁷⁸ Nonlinear warping of NODDI metrics to Montreal Neurological Institute (MNI) space was accomplished in one step by applying a combination of distortion-corrected b0-to-T1 alignment performed in native space and nonlinear T1-to-MNI warping. Pandora Tracula probabilistic maps ⁷⁹ were used to derive a priori white matter ROIs, applying a threshold of 0.95. ROIs included the forceps major (Fmajor), forceps minor (Fminor), and bilateral cingulum angular bundle (CingAng), cingulum cingulate gyrus (CingGyr), and uncinate fasciculus (UF). Average NODDI metrics were extracted for each participant from voxels in each ROI in which at least 90% of participants had data.

Statistical analysis

Demographic variables of the OC and NoHC groups were compared using independent-samples t-tests, Wilcoxon rank-sum tests, Wilcoxon rank-sum exact tests, Fisher exact tests, and Moses test of extreme reactions as appropriate using the gtsummary ⁸⁰ and DescTools ⁸¹ packages in R (version 4.5.0), ⁸² depending on the nature and distribution of the data.

Linear regression models in R were fit for outcomes of hippocampal volumes (left and right), GFAP, NfL, and tau. Independent variables in primary models included the number of prior concussions and OC use (yes vs. no hormonal contraception) and their interaction, as well as years of CSE, OC use, and their interaction. Nonsignificant interaction terms were removed, and models were rerun. To confirm the robustness of results and account for qualitative differences in the hormonal milieu between OC users and nonusers (e.g., ovarian function and endogenous/exogenous hormone exposure), models were run separately in OC and NoHC groups, with prior concussion and CSE as predictors. Such examination of group-stratified processes is recommended in sex difference-related research. ⁸³

Each model was evaluated for linearity, homogeneity of variance, influential observations, collinearity, normality of residuals, and model fit using the easystatsR packages. ⁸⁴ Influential observations (falling outside Cook’s distance) were removed, and models were rerun when applicable. Robust standard errors were calculated for models with heteroscedasticity (identified by the Breusch–Pagan test) using the sandwich R package with Heteroscedasticity-consistent (robust standard error) estimator, type 3 robust estimation. ⁸⁵

Cortical thickness and NODDI metrics (ODI, ISO, ICVF) were analyzed using Bayesian multilevel modeling (BML) via the Region-Based Analysis (RBA) program in AFNI. BML’s unified model estimates the probability that effects across all ROIs are simultaneously nonzero.^86,87 Models included the interactions of OC use with the number of prior concussions and with years of CSE, and with the main effects of the number of prior concussions, years of CSE, and OC use. Subgroup-stratified analyses were conducted for the OC and NoHC groups.

For all bayesian multilevel modeling models, continuous variables were standardized per program recommendations, and the response variables were modeled using a Student’s t-distribution. Models were fit using four Markov chains with 5,000 iterations per chain and 2,500 warmup samples and evaluated using R-hat statistics, posterior predictive checks, and probability integral transformation checks. Evidence for effects was quantified using the posterior probability that an effect was positive (P+). A posterior probability that an effect was positive (P+) value of 0.5 indicates no evidence of an effect, values near 1 indicate strong evidence for positive effects, and values near 0 indicate strong evidence for negative effects. For clarity, only associations with P+ ≥ 0.90 or ≤ 0.10 are described in the main text, but all results, regardless of P+ values, are presented in the Supplemental Tables.

Additionally, sensitivity analyses were conducted by removing three participants in the NoHC group who indicated previous OC use. No other prior hormonal contraceptive use was indicated in the NoHC group.

Results

Demographics

This study included 68 female participants, 30 OC users (13 with prior concussion), and 38 not using any hormonal contraceptive (NoHC; 18 with prior concussion). The OC group did not differ from the NoHC group in CSE, days since injury for those with prior concussion, age, race, ethnicity, waist-to-hip ratio, or socioeconomic status (ps > 0.05). While the Wilcoxon rank-sum test showed the number of prior concussions was not significantly different between groups (p = 0.29), when evaluated with the Moses test of extreme reactions, the NoHC group had more participants with a high number of concussions (p < 0.001). Demographics are listed in Table 1. Comparisons of all outcomes of interest between the OC and NoHC groups are provided in Supplementary Table S1. The OC group had lower tau and right insula thickness and higher left UF ISO.

Table 1.

Sample Characteristics

Variable	N	Overall N = 68 a	Group		Test statistic	p-Value
			NoHC N = 38 a	OC N = 30 a
Concussion history	68				0.11	0.74 b
No prior concussions		37/68 (54%)	20/38 (53%)	17/30 (57%)
Prior concussion(s)		31/68 (46%)	18/38 (47%)	13/30 (43%)
Number of prior concussions	68				649 49	0.29 c <0.001 d
Mean (SD)		0.90 (1.35)	1.16 (1.64)	0.57 (0.77)
[Min, Max]		[0, 6]	[0, 6]	[0, 3]
Years of contact-sport exposure	68				641	0.37 c
Mean (SD)		6.22 (6.26)	6.68 (6.27)	5.63 (6.29)
[Min, Max]		[0, 17]	[0, 16]	[0, 17]
Age	68				0.13	0.90 e
Mean (SD)		20.34 (1.31)	20.36 (1.38)	20.32 (1.24)
[Min, Max]		[18.26, 23.11]	[18.26, 23.11]	[18.26, 22.93]
Days since injury	31				116	0.98 f
Mean (SD)		1,116 (1,149)	1,193 (1,353)	1,010 (827)
[Min, Max]		[166, 5,196]	[190, 5,196]	[166, 2,278]
Waist to hip ratio	68				551	0.82 f
Mean (SD)		0.83 (0.06)	0.83 (0.06)	0.84 (0.07)
[Min, Max]		[0.70, 0.97]	[0.70, 0.94]	[0.71, 0.97]
Socioeconomic status g	68				675	0.20 c
Mean (SD)		50.22 (8.33)	51.12 (8.73)	49.07 (7.80)
[Min, Max]		[20.0, 63.5]	[20.0, 63.5]	[33.0, 61.0]
Race	68					0.19 8
Black or African American		2/68 (2.9%)	0/38 (0%)	2/30 (6.7%)
White		66/68 (97%)	38/38 (100%)	28/30 (93%)
Ethnicity	68					>0.99 h
Hispanic or Latino		0/68 (0%)	0/38 (0%)	0/30 (0%)
Not Hispanic or Latino		68/68 (100%)	38/38 (100%)	30/30 (100%)

^a n/N (%).

^b Pearson’s chi-squared test.

^c Wilcoxon rank-sum test.

^d Moses test of extreme reactions.

^e Welch two sample t-test.

^f Wilcoxon rank-sum exact test.

^g Hollingshead four-factor index of socioeconomic status.

^h Fisher’s exact test.

SD, standard deviation.

OC information

On average, OC users reported taking their current OC for 2.55 years (SD = 1.69, range: 0.08–6.0 years). Within this group, 5 of the 17 concussions occurred during current OC use. Except for one athlete, OC users with CSE started playing contact sports prior to OC use. Of the 30 OC users, 25 provided enough information to determine OC brand, progestin generation, and whether the OC was mono- or triphasic. Nineteen brands representing 11 different formulations were used. Sixteen were monophasic, and nine were triphasic. First-generation progestins were used by 7 OC users, second-generation by 4, third-generation by 13, and fourth-generation by 1. Details on OC formulation are provided in Supplementary Table S2. Information on dose timing and pill phase for OC users was not available.

Hippocampal volumes

No significant interactions were seen between OC use and concussion history or CSE. Left and right hippocampal volumes decreased with an increase in the number of prior concussions. No significant associations were seen in the subgroup-stratified analyses. Results for hippocampal models are in Table 2.

Table 2.

Hippocampal Volume Associations with Prior Concussion, CSE, and OC Use

Outcome measure	Variable	B (SE)	ß [95% CI]	p-Value
Hippocampal volume whole group (N = 68)
Left&	#Cx	−47.5(19.7)	−0.19 [−0.35, −0.03]	0.019
	CSE	4.95(4.68)	0.09 [−0.08, 0.26]	0.29
	OC Use	32.3(60.6)	0.10 [−0.26, 0.45]	0.60
Right&	#Cx	−42.0(20.0)	−0.15 [−0.29, −0.01]	0.039
	CSE	8.50(4.66)	0.14 [−0.01, 0.29]	0.07
	OC Use	−54.4 (62.7)	−0.14 [−0.47, 0.18]	0.39
Hippocampal volume OC group only (N = 30)
Left	#Cx	−97.8(58.8)	−0.26 [−0.57, 0.06]	0.11
	CSE	6.48(7.75)	0.14 [−0.20, 0.48]	0.41
Right	#Cx	−68.7(60.7)	−0.17 [−0.49, 0.14]	0.27
	CSE	5.76(8.01)	0.12 [−0.22, 0.46]	0.48
Hippocampal volume NoHC group only (N = 38)
Left	#Cx	−40.7(23.8)	−0.18 [−0.39, 0.03]	0.10
	CSE	4.39(6.28)	0.07 [−0.14, 0.29]	0.49
Right&	#Cx	−38.5(21.5)	−0.14 [−0.31, 0.02]	0.08
	CSE	9.60(6.55)	0.14 [−0.05, 0.33]	0.15

Significant associations (p < 0.05) are in bold font.

B, unstandardized coefficient; SE, standard error; ß, standardized coefficient; CI, confidence interval; #Cx, number of prior concussions; CSE, years of contact-sport exposure; OC, oral contraceptive; NoHC, no hormonal contraceptive; &, robust standard error with corresponding CI adjustment due to heteroscedasticity in model.

Blood-based brain injury biomarkers

After dropping nonsignificant interaction terms, the models showed lower tau for OC users than the NoHC group, with no other significant associations. Subgroup-stratified models did not show any significant main effects. Brain injury marker model results are in Table 3.

Table 3.

Brain Injury Marker Associations with Prior Concussion, CSE, and OC Use

Outcome measure	Variable	B (SE)	ß [95% CI]	p-Value
Biomarker whole group
GFAP (N = 67)	#Cx	−0.01 (0.03)	−0.06 [−0.31, 0.20]	0.65
	CSE	0.01 (0.01)	0.19 [−0.06, 0.43]	0.14
	OC Use	−0.08 (0.08)	−0.25 [−0.75, 0.25]	0.33
NfL (N = 66)	#Cx	−0.03 (0.03)	−0.13 [−0.39, 0.12]	0.31
	CSE	0.01 (0.01)	0.15 [−0.10, 0.40]	0.23
	OC Use	0.02 (0.08)	0.05 [−0.46, 0.56]	0.85
Tau (N = 68)	#Cx	−0.06 (0.07)	−0.11 [−0.35, 0.14]	0.38
	CSE	0.01 (0.01)	0.12 [−0.12, 0.36]	0.32
	OC Use	−0.45 (0.18)	−0.59 [−1.08, −0.11]	0.017
Biomarker OC group only
GFAP (N = 30)	# Cx	−0.10 (0.09)	−0.21 [−0.61, 0.20]	0.31
	CSE	0.01 (0.01)	0.21 [−0.20, 0.62]	0.31
NfL (N = 30)	# Cx	−0.01 (0.09)	−0.02 [−0.44, 0.40]	0.92
	CSE	0.01 (0.01)	0.14 [−0.28, 0.56]	0.50
Tau (N = 30)	# Cx	−0.20 (0.20)	−0.20 [−0.60, 0.21]	0.33
	CSE	0.03 (0.02)	0.21 [−0.20, 0.62]	0.30
Biomarker NoHC group only
GFAP (N = 37)	#Cx	−0.01 (0.03)	−0.03 [−0.37, 0.31]	0.86
	CSE	0.01 (0.01)	0.21 [−0.13, 0.55]	0.22
NfL (N = 36)	#Cx	−0.04 (0.03)	−0.23 [−0.57, 0.11]	0.18
	CSE	0.01 (0.01)	0.12 [−0.22, 0.46]	0.48
Tau (N = 38)	#Cx	−0.04 (0.07)	−0.10 [−0.44, 0.24]	0.54
	CSE	0.01 (0.02)	0.08 [−0.26, 0.42]	0.64

Significant associations (p < 0.05) are in bold font.

B, unstandardized coefficient; SE, standard error; ß, standardized coefficient; CI, confidence interval; GFAP, glial fibrillary acidic protein; NfL, neurofilament light; Tau, total tau; #Cx, number of prior concussions; CSE, years of contact-sport exposure; OC, oral contraceptive; NoHC, no hormonal contraceptive.

Cortical thickness

Strong evidence existed for a CSE by OC use interaction. Subgroup-stratified analysis showed more years of CSE corresponded with thicker cortex in all ROIs except the right caudal anterior cingulate cortex for the OC group (mean effects 0.028–0.045 mm expected change in cortical thickness associated with a 1-standard-deviation increase in years of CSE), with the largest effects in the bilateral temporal poles. No strong associations with CSE were seen for the NoHC group. For both groups, more prior concussions were associated with thinner cortex, mainly in temporal regions. These associations were more consistent across all regions in the OC group (mean effects −0.028 to −0.020), but there were larger mean effects in specific regions in the NoHC group (mean effects −0.065 to −0.019). Results are shown in Figure 1 with full model results in Supplementary Tables S3 and S4.

FIG. 1.

Associations of prior concussion and contact-sport exposure with cortical thickness split by group. Boxplot colors show the probability that the effect is positive given the model and data (P+), with values closer to 0 indicating evidence for negative relationships and values closer to 1 indicating evidence for positive relationships. Mean effects are expected change in cortical thickness (mm) associated with a 1-standard-deviation increase in predictor. Boxplots show 2.5%, 5%, 50%, 95%, and 97.5% quantile intervals for each effect. *, P+ ≥0.9 or ≤0.1; #Cx, number of prior concussions; CSE, contact-sport exposure; NoHC, no hormonal contraceptive use group; OC, oral contraceptive use group; ROI, region of interest.

Orientation dispersion index

There was strong evidence for an interaction of prior number of concussions by OC use for ODI. The OC group had strong positive relationships between prior concussions and orientation dispersion index (ODI) in all ROIs except the right cingulum angular bundle (CingAng) (mean effects 0.0039–0.0074). The NoHC group showed smaller, more localized positive relationships between prior concussion and ODI in the right cingulum cingulate gyrus (CingGyr) and bilateral UF (mean effects 0.0028–0.0038). Increased ODI with more CSE was also observed bilaterally in the OC group in the UF and CingAng (mean effects 0.0030–0.0052). Results are shown in Figure 2 and detailed in Supplementary Tables S5 and S6.

FIG. 2.

Associations of prior concussion and contact-sport exposure with ODI split by group. Boxplot colors show the probability that the effect is positive given the model and data (P+), with values closer to 0 indicating evidence for negative relationships and values closer to 1 indicating evidence for positive relationships. Mean effects are expected change in ODI associated with a 1-standard-deviation increase in predictor. Boxplots show 2.5%, 5%, 50%, 95%, and 97.5% quantile intervals for each effect. *, P+ ≥0.9 or ≤0.1; #Cx, number of prior concussions; CingAng, cingulum angular bundle; CingGyr, cingulum cingulate gyrus; CSE, contact-sport exposure; Fmajor, forceps major; Fminor, forceps minor; NoHC, no hormonal contraceptive use group; OC, oral contraceptive use group; ODI, orientation dispersion index; ROI, region of interest; UF, uncinate fasciculus.

Isotropic volume fraction

Evidence was strong for a CSE by OC use interaction with ISO. CSE was positively related to ISO only for OC users, bilaterally in the CingAng and UF, and in the right CingGyr (mean effects 0.0022–0.0046), as shown in Figure 3 with additional detail in Supplementary Tables S7 and S8.

FIG. 3.

Associations of prior concussion and contact-sport exposure with ISO split by group. Boxplot colors show the probability that the effect is positive given the model and data (P+), with values closer to 0 indicating evidence for negative relationships and values closer to 1 indicating evidence for positive relationships. Mean effects are expected change in ISO associated with a 1-standard-deviation increase in predictor. Boxplots show 2.5%, 5%, 50%, 95%, and 97.5% quantile intervals for each effect. *, P+ ≥0.9 or ≤0.1; #Cx, number of prior concussions; CingAng, cingulum angular bundle; CingGyr, cingulum cingulate gyrus; CSE, contact-sport exposure; Fmajor, forceps major; Fminor, forceps minor; ISO, isotropic volume fraction; NoHC, no hormonal contraceptive use group; OC, oral contraceptive use group; ROI, region of interest; UF, uncinate fasciculus.

Intracellular volume fraction

Subgroup-stratified analyses to explore the significant interaction between CSE and OC use revealed divergent associations (see Fig. 4). The OC use group had higher ICVF with more CSE in the bilateral CingGyr and Fminor (mean effects 0.0060–0.0079) while the NoHC group had lower ICVF with more CSE in the Fmajor, Fminor, and bilateral CingGyr (mean effects −0.0056 to −0.0050). Additionally, the number of prior concussions and ICVF were negatively associated in the Fmajor and left CingGyr for the OC group only (mean effects −0.0062 to −0.0057), though weaker negative effects were observed across most ROIs for the NoHC group. Detailed results for these models are in Supplementary Tables S9 and S10.

FIG. 4.

Associations of prior concussion and contact-sport exposure with ICVF split by group. Boxplot colors show the probability that the effect is positive given the model and data (P+), with values closer to 0 indicating evidence for negative relationships and values closer to 1 indicating evidence for positive relationships. Mean effects are expected change in ICVF associated with a 1-standard-deviation increase in predictor. Boxplots show 2.5%, 5%, 50%, 95%, and 97.5% quantile intervals for each effect. *, P+ ≥0.9 or ≤0.1; #Cx, number of prior concussions; CingAng, cingulum angular bundle; CingGyr, cingulum cingulate gyrus; CSE, contact-sport exposure; Fmajor, forceps major; Fminor, forceps minor; ICVF, intracellular volume fraction; NoHC, no hormonal contraceptive use group; OC, oral contraceptive use group; ROI, region of interest; UF, uncinate fasciculus.

Exploratory hormone activity level analyses

Twenty-five OC pill users provided sufficient details to determine their OC formulation and its associated hormone activity levels. Methods for this exploratory work are included in the Supplemental Data with information on hormone activity levels are detailed in Supplementary Table S11. This exploratory work suggests that progestational activity levels may be directly related to ODI and ISO, and that estrogenic and androgenic activity levels may be related to cortical thickness. Results also showed potential moderation of associations with prior concussion by hormonal activity levels for right hippocampal volume, GFAP, and ISO, and of associations with CSE for GFAP and tau. Full results are provided in Supplementary Tables S12, S13, S14, S15, S16, S17, S18, S19, S20, S21, S22, S23, S24, S25, S26, and S27 and Supplementary Figures S1, S2, S3, S4, S5, and S6.

Sensitivity analyses removing NoHC group participants with prior OC use

Sensitivity analyses excluding the three NoHC participants with prior OC use did not meaningfully change the hippocampal volume or brain injury marker models. In the cortical thickness, ODI, and ICVF models, a few results shifted across our significance cutoffs. In the ISO model, the CSE by OC use interaction was no longer significant enough to warrant a split analysis, though the NoHC-only results were unchanged. Full results are provided in Supplementary Tables S28, S29, S30, S31, S32, S33, S34, S35, S36, and S37.

Discussion

Effects of cumulative concussion and CSE are of increasing concern, with neuroimaging and blood-based biomarkers emerging as primary physiological outcomes of interest. This study of female athletes replicated associations of concussion history with smaller hippocampal volumes and thinner cortex previously seen in mixed-sex and male-only studies. We found a positive relationship between concussion history and ODI, slightly influenced by OC use. Furthermore, OC use strongly moderated relationships between CSE and cortical thickness, ISO, and ICVF. In contrast, no associations with concussion history or CSE were observed for blood-based biomarkers of brain injury across the whole cohort, though OC users had lower tau concentrations than the NoHC group. Finally, exploratory models indicated OC pill formulation may affect relationships of concussion history and CSE with structural neuroimaging measures and blood-based biomarkers, and may directly impact cortical thickness, ODI, and ISO. These preliminary results demonstrate that OC use should be considered in future research and when personalizing treatment for female athletes with multiple concussions or CSE.

Associations with concussion history

This study adds to the growing literature linking repeat concussion to subtle differences in gray matter structure, notably expanding upon prior work by focusing specifically on female athletes. The negative relationship between hippocampal volume and the number of prior concussions is consistent with most prior work^8,9,13–16 and with results in the mixed-sex parent study. ¹⁰ Similarly, thinner cortex with more prior concussions has been reported elsewhere^20,88 and in the mixed-sex parent study, ⁶⁹ although some prior studies had null findings.^6,89–91 The temporal region concentration of these findings is expected, given their proximity to the bony structure of the skull and the movement of the brain during impacts. ⁹² Previous explanations for smaller hippocampal volumes and thinner cortex that may also apply here include inflammation, stress-related hormones, and altered glutamatergic activity.^9,93–99 Our study suggests that current OC use may not substantially impact these gray matter associations.

White matter associations with concussion history were more sensitive to OC use in the current cohort. Concussion history was positively related to ODI, with stronger and more widespread effects in the OC group, corresponding to more diffusely oriented neurites rather than the typical highly aligned white matter fibers. Lower ICVF with more prior concussions in the Fmajor and left CingGyr of the OC group indicates lower neurite density in these regions. These findings are consistent with proposed TBI pathologies, including white matter injury and increased perivascular spaces.^10,100–102 Increased ODI is also consistent with decreased fractional anisotropy (FA) associated with concussion history in other studies.^36,103,104

The mechanism underlying the OC group’s increased sensitivity to white matter microstructural abnormalities associated with concussion is not yet clear. OC use has been associated with increased inflammation, as evidenced by elevated oxidative stress and C-reactive protein levels, including in this cohort.^65,105–108 With changes to endogenous and exogenous hormones resulting from OC use, the involvement of progesterone, progestins, and estradiol (including brain-derived estradiol) in gliosis, synaptogenesis, myelination, and regulation of neurotransmitters^61,109 is likely altered. Concussion is known to impact these same processes acutely and potentially chronically,^94,100 and it is possible that OC use heightens or otherwise changes these effects. Alternatively, we cannot rule out that differences in the association between prior concussion and white matter microstructure could also reflect some other unmeasured pre-existing difference between groups or could be sensitive to the exact timing of each prior concussion relative to the initiation of OC use.

Blood-based brain injury markers showed no significant associations with concussion (or CSE) in the full cohort. These results suggest that MRI markers are more sensitive than peripheral markers to subtle, localized effects in younger athletes, in agreement with most of the existing literature, which also reports null results.^{27,30,31,110,111} Study design may be important, as prior work in which GFAP, NfL, and/or tau increased with CSE was primarily conducted over the course of a season or preseason^27–29,31 (with the exception of an increase in tau with years of CSE only in males seen by Di Battista et al. ³⁰ ). Additionally, exploratory analyses limited to the OC group did find evidence that hormone activity level (derived from pill formulation) may moderate associations of GFAP and tau with concussion and/or CSE, highlighting pill formulation as an important factor for consideration in future studies on this topic.

Associations with CSE

OC use more strongly moderated the relationship between CSE and brain structure than it did for concussion history. The positive relationship between CSE and cortical thickness observed in the OC group contrasts with most of the existing literature, which either found thinner cortex or no association with CSE,^{5,6,17–19,90,91,112} although we first identified this positive association in female athletes in a split-by-sex analysis in the parent study. ⁶⁹ While we are not aware of any evidence outside this cohort linking thicker cortex to CSE, there are a few instances of region-dependent cortical thickening in the chronic phase of concussion.^113,114 As proposed by Dall’Acqua et al., ¹¹³ neuroinflammation, a stunting of neurotypical cortical thinning, or neuroplasticity-related processes (synaptogenesis, dendritic growth, or gliogenesis) might plausibly explain thicker cortex. The concentration of these effects in OC users may potentially reflect the complex impact of hormonal milieu changes on physiological processes sensitive to repetitive neurotrauma.

Associations of diffusion metrics with CSE also diverged based on OC use. In the OC group, the positive relationships between CSE and ODI, ISO, and ICVF are consistent with prior studies showing less organized white matter and more free diffusion associated with CSE.^115–117 These findings also suggest potential inflammation and/or cytotoxic edema and may align with a low-level chronic inflammatory state associated with OC use as suggested above. In contrast, the negative relationship between CSE and ICVF for the NoHC group may potentially reflect axonal degeneration or demyelination. These results raise the hypothesis that OC use impacts the mechanisms underlying subtle differences in brain microstructure associated with CSE, though additional studies are needed to pursue this hypothesis.

Associations with OC use or formulation

Our main investigation and exploratory work revealed various potential relationships with OC use or pill formulation and our outcomes of interest. The lower tau for OC users in our study appears to be novel, as we are not aware of any other studies that have directly examined this relationship. Tabor et al. ⁴¹ found that tau is negatively associated with plasma progesterone levels. Since OC users would likely have lower endogenous progesterone, our result may appear to contradict Tabor et al., ⁴¹ although it is unknown whether exogenous progestins would show a similar relationship. In contrast to Huibregtse et al. ²⁹ reporting lower GFAP and NfL in hormonal contraceptive users, our study did not find associations of these biomarkers with OCs. This difference may be due to Huibregtse et al. ²⁹ including IUD users or saliva hormone levels in their models.

Exploratory analyses also revealed strong associations between OC hormone activity levels (derived from pill formulations) and cortical thickness and white matter microstructure (ODI and ISO), as well as evidence that OC hormone activity levels moderate associations with concussion and/or CSE across all studied outcomes. These results support our findings of multiple associations in the OC group but not the NoHC group and further highlight the need for additional studies specifically focused on these factors in female athletes.

Study strengths and limitations

Study strengths include strict exclusion criteria to reduce the impact of potentially confounding factors, in-depth interviews used to determine prior concussions and CSE, and inclusion of OC use information, including pill formulation, for exploratory work.

Key limitations include the absence of menstrual cycle phase, blood hormone levels, and OC pill regimen data (dose timing and pill phase), all of which may influence outcomes.^{41,50,54,56,57,59} Future studies should include these factors and investigate with larger sample sizes, enabling examination of hormone activity level effects in combination. Most of the OC group’s concussions and the bulk of the years of CSE did not occur while using OCs, so these results reflect effects of OC use at the time of study participation but not injury. Timing of Cx and of initiation of OC use were also not addressed due to complexity of models given the sample size. Future studies should be designed to more directly address the role of contraceptive use at the time of injury/CSE on chronic outcomes. Generalizability of results may be reduced by the focus on collegiate-aged athletes and the limited racial and ethnic diversity of the sample. Finally, longitudinal studies are needed, since one cannot infer directionality or determine the stability of observed abnormalities from this cross-sectional study.

Conclusion

Results suggest that current OC use may fundamentally change associations between CSE (and to a lesser extent, cumulative concussion) and brain structure in collegiate-aged female athletes and that OC pill formulation may influence these relationships. Larger studies specifically designed to address these questions are desperately needed as the field moves forward to better describe the physiological effects of concussion and CSE in female athletes.

Transparency, Rigor, and Reproducibility Summary

The study and analysis plan were not formally preregistered. Participant recruitment for the larger cross-sectional study occurred from December 2018 to January 2023. The sample size was limited to available participant data, which ultimately resulted in the inclusion of 68 participants in the present analyses. A flowchart of the research process from participant recruitment to the final sample size for analysis is provided in the Supplement Figure S7. All aspects of the study were approved by the Medical College of Wisconsin Institutional Review Board, with written informed consent provided by all participants. Human participants were not informed of their individual imaging or biomarker results during the study. Handling of biofluid samples and measurements, imaging quality control decisions, and analyses were performed by team members blinded to participant characteristics. The biomarker assays were performed by investigators blinded to participant characteristics. Imaging data and biospecimens were labeled using participant ID numbers that were not linked to identifying information. Detailed parameters for imaging and fluid biomarkers and for statistical analyses are included in the text. All equipment, software, and analytical reagents used to perform imaging, preprocessing, and measurements on the fluid biomarkers are widely available from commercial sources, or their sources are described in the text. No replication or external validation studies have been performed or are planned/ongoing at this time. De-identified data and analytic code used to conduct the analyses are not available in a public repository but may be made available (as allowable according to institutional IRB standards) upon request by emailing the corresponding author. The authors have agreed to publish the article using the Mary Ann Liebert Inc. “Open Access” option under appropriate license.

Authors’ Contributions

B.D.G.: Conceptualization, data curation, formal analysis, methodology, software, validation, visualization, writing original draft, writing—review and editing. A.M.B.: Conceptualization, methodology, writing—review and editing. T.K.T.: Investigation, resources, writing—review and editing. J.S.: Conceptualization, writing—review and editing. A.B.: Methodology, writing—review and editing. R.M.: Conceptualization, methodology, writing—review and editing. B.L.B.: Conceptualization, methodology, writing—review and editing. T.B.M.: Conceptualization, funding acquisition, methodology, supervision, writing—original draft, writing—review and editing.

Ethical Considerations

All aspects of the study were approved by the Medical College of Wisconsin Institutional Review Board.

Consent to Participate

Written informed consent was provided by all participants.

Data Availability

The data supporting the findings of this study are available from the corresponding author upon reasonable request, with the execution of necessary data-use agreements.

Abbreviations Used

3D

Three-dimensional

ACRM

American Congress of Rehabilitation Medicine

AFNI

Analysis of Functional NeuroImages

b0

Non-diffusion-weighted image volume

BET

Brain Extraction Tool

BML

Bayesian multilevel modeling

CAGE

Cut down, Annoyed, Guilty, Eye-opener

CingAng

Cingulum angular bundle

CingGyr

Cingulum cingulate gyrus

CSE

Contact-sport exposure

cuDIMOT

CUDA Diffusion Modelling Toolbox

FA

Fractional anisotropy

Fmajor

Forceps major

Fminor

Forceps minor

FOV

Field of view

FSL

FMRIB Software Library

GFAP

Glial fibrillary acidic protein

HC3

Heteroscedasticity-consistent (robust standard error) estimator, type 3

ICVF

Intracellular volume fraction

ISO

Isotropic volume fraction

IUD

Intrauterine device

MNI

Montreal Neurological Institute

MPRAGE

Magnetization-prepared rapid gradient echo

MRI

Magnetic resonance imaging

mTBI

Mild traumatic brain injury

NfL

Neurofilament light

NODDI

Neurite Orientation Dispersion and Density Imaging

NoHC

No hormonal contraceptive (use)

OC

Oral contraceptive

ODI

Orientation dispersion index

P+

Posterior probability that an effect was positive

RBA

Region-Based-Analysis

RCF

Relative centrifugal force

ROI(s)

Region(s) of interest

SD

Standard deviation

TBI

Traumatic brain injury

TE

Echo time

TR

Repetition time

UF

Uncinate fasciculus

WHO

World Health Organization

Cite this article as: Goeckner BD, Beltz AM, Teague T K, et al. Oral contraceptives, concussion history, and contact-sport exposure: Multimodal MRI and brain injury biomarkers in female athletes, Neurotrauma Reports, 2026, 7, 2689288X261473764, doi: 10.1177/2689288X261473764.