Vascular adaptation in elite female and male rowers across a competitive season

Abstract

The ‘athlete's artery’ phenotype describes exercise-induced vascular adaptation whereby athletes have enlarged conduit arteries resulting from chronic endurance exercise. However, studies remain limited to males, with few evaluating changes in response to training. This study aimed to compare upper and lower limb adaptations across a 21-week competitive season between elite female and male rowers. Twenty-one athletes (females n = 10) were assessed at three timepoints across a 21-week season: early- (ES), mid- (MS), and late-season (LS). High-resolution duplex ultrasonography assessed brachial and femoral artery diameter, flow-mediated dilation (FMD%), and ischaemic hand-grip exercise (vasodilatory capacity, VD%). Data was analysed using Bayesian repeated measures ANOVA (training × sex). The time-course for peak brachial diameter differed for sex with largest diameter for females at LS (4.3 ± 0.3 cm) compared to MS for males (5.1 ± 0.3 cm). Similarly, brachial FMD% differed by sex with training where females had largest FMD% at LS and males at MS. However, shear-normalised brachial FMD% showed no effect of training or sex. No changes in VD% were observed. Femoral artery diameter was larger in males, while no sex or training effects were evidence for femoral FMD%. In the upper limb, brachial diameter increased with training which differed by sex, suggesting a sex-specific response. While brachial FMD% also improved with training in each sex, normalising for shear rate removed all apparent differences, suggesting a shear-mediated response. Lower limb vasculature was less impacted by training, with a moderate effect for sex.

Highlights

• •Baseline and peak brachial artery diameter differ between male and female rowers.
• •Peak diameter and FMD% demonstrate sex-specific responses over a training season.
• •Both sexes completed similar total training volume over the 21-week study period.
• •Males spent more time training at higher intensities than females.
• •Training intensity may explain vascular adaptation variation between the sexes.

1. Introduction

Chronic exercise causes arterial physiological adaptation identified as the ‘athlete's artery’ (Naylor et al., 2021). Longitudinal studies in previously untrained males have shown an increase in arterial diameter with training, specific to exercise modality (Spence et al., 2013), which is also commonly observed in endurance and resistance trained male athletes (Churchill, 2020; Naylor et al., 2021; Rowley et al., 2012). This exercise-induced adaptation is localised to the arteries supplying the exercising limb and the magnitude of adaptation is positively associated with training duration and load (Rowley et al., 2012; Tao et al., 2023). Arterial adaptation is often preceded by endothelial-dependent functional changes within conduit arteries in untrained participants (Tinken et al., 2008). However, research in athletes indicates chronic training could have little to no effect on functional outcomes (Churchill, 2020), evident by a cross-sectional study in males that showed no differences in brachial flow mediated dilation (FMD%) in athletes compared to controls (Naylor et al., 2021). While the mechanism for these divergent findings is still unclear, arterial function is strongly dependent on diameter (Silber et al., 2005), suggesting that the increased diameter in arteries of athletes results in reduced endothelial function, representative of the “athlete paradox” (Green et al., 2012). The overwhelming majority of this research has been conducted in male, untrained participants, or is cross-sectional in nature, highlighting a gap within our knowledge of female athlete arterial adaptation over time.

Despite the Paris Summer Olympics in 2024 reaching gender parity for the first time (IOC, 2024), sport and exercise science research trails behind in female representation (Smith et al., 2022). A recent systematic review evaluating research on the effects of vascular adaptation to exercise found that only 39 % of participants were female (Thompson et al., 2024). Furthermore, only ∼5 % of included research studies were conducted using best practice guidelines (Elliott-Sale et al., 2021; Thompson et al., 2024). Emerging research has suggested that the menstrual cycle not only has a trivial impact on performance (McNulty et al., 2020) but can influence macrovascular endothelial function (Williams et al., 2020). Oestrogen receptors are found in endothelial cells (Gavin et al., 2009), suggesting that oestrogen may play a role in mediating FMD% in premenopausal women (Holder et al., 2019; Moreau et al., 2024). Subsequently, females may have a more marked vascular response to chronic exercise compared to males, independent of modality (Green et al., 2023). More research is needed to understand the interrelationship between female-specific physiology and vascular adaptation to chronic exercise.

Long-term studies in previously untrained participants (Green et al., 2023; Spence et al., 2013) and cross-sectional studies in male (Naylor et al., 2021) and female athletes (Grandys et al., 2023), have provided insight into the effect of training on the arterial system. However, the specific time-course changes occurring in elite female athletes, and the training response between sexes, remains largely unexplored. This observational study aimed to compare arterial responses in the upper and lower limbs of elite female and male rowers across a 21-week competitive season. Specifically, early in the season (early-season), at the end of the general preparatory and overall fitness phase (mid-season) and at the conclusion of the season (late-season), prior to major competition. Based on prior research, we hypothesise arterial parameters will differ by sex, while the magnitude and direction of any training-induced adaptation remains exploratory.

2. Materials and methods

Written, informed consent was given by all participants prior to participation in the study. The Curtin University Human Research Ethics Committee approved this study (HRE2020-0510).

2.1. Participants

Athletes (20 ± 1 y, n = 22; female n = 11) were recruited from Western Australian Institute of Sport and classified according to the Participant Classification Framework (McKay et al., 2022) with n = 17 state-level representatives (Tier 3 Highly Trained; female n = 8) and n = 5 on the national junior team (Tier 4 Elite; female n = 2). A priori sample size calculation was not required as the study included all eligible athletes within the target population. Participants underwent medical screening prior to enrolment to confirm no history of cardiovascular disease and were free from injuries which would prevent them from training. This cohort also underwent cardiac echocardiographic assessments as part of a separate investigation reported elsewhere (Henley-Martin et al., 2025). A female-specific questionnaire was used to determine menstrual status including history (previous 3 months) and current use of hormonal contraception (including type, duration and formulation). Participants were confirmed to have used either the same hormonal contraception or no contraception prior to the study and asked to maintain this throughout, unless directed otherwise by their medical professional. Participants were not excluded based on use, and menstrual status was not controlled for (Stanhewicz and Wong, 2020) but was documented throughout the study using a combination of subjective and objective measures (Elliott-Sale et al., 2021).

2.2. Study overview

Athletes were recruited in September 2020 and observed over the 2020–2021 competitive rowing season (21-weeks) until March 2021. Testing occurred at key time-points within the season; early-season (ES), mid-season (MS), and late-season (LS; before major competition). The time between testing was considered as separate training blocks for Block 1 (ES-MS, general preparation, overall fitness and base strength phase) which was from September to December 2020 (average duration 9 ± 1 weeks) and Block 2 (MS-LS, race-specific and power phase) which was between December 2020 and March 2021 (average duration 12 ± 3 weeks). As this study was conducted in the southern hemisphere, ES coincided with winter-spring transition (∼17 °C average daily temperature), MS was peak summer (∼22 °C average daily temperature) and LS was late summer-early Autumn (∼24 °C average daily temperature). Rowing is considered a summer sport, often major competition is in the European (northern hemisphere) summer, meaning that the LS phase (race work) would closely mirror what would be seen in ‘typical’ elite rowing populations. The testing protocol consisted of peripheral vascular assessment, aerobic capacity test, and body composition (ES and LS only) and anthropometric measures. All measures were completed within the same week for an individual athlete, with all measures in a single assessment captured on the same day e.g. peripheral vascular assessment and venous blood sample (female athletes), while other measures were obtained on different days within the same testing block over a 3–4-week period.

2.3. Peripheral vascular assessment

Vascular assessments were conducted in a temperature-controlled room, in accordance with expert consensus guidelines and recommendations for participant preparation (Thijssen et al., 2019), with the exception that exercise and food were only avoided for ≥4 h prior to assessment due to the athletes highly demanding training schedules. Following 20-min of supine rest, participants blood pressure (systolic, diastolic and mean arterial pressure) was measured using an automated sphygmomanometer (Connex ProBP 3400, Welch Allyn) and resting heart rate was determined using a 12-lead electrocardiogram (ECG; Nassif Associates Inc., NY, USA).

Simultaneous FMD% of the brachial and femoral arteries was measured at rest using high-resolution duplex ultrasonography (Terason t3200, Burlington, MA, USA) by two experienced sonographers. Two rapid-inflation pneumonic occlusion cuffs (DE Hokanson Inc., Bellevue, WA, USA) were placed on the forearm and above the knee. Using a 10 mHz linear-array ultrasound probe, the brachial (∼8 cm from the elbow) and superficial femoral arteries (∼15 cm from the inguinal fold) were imaged and a 1-min baseline was recorded using Camtasia software (Camtasia Studio 8, TechSmith, Okemos, MI, USA). The cuffs were then simultaneously inflated to 220 mmHg for 5 min. Recording started at 4 min and 30-s and continued for 3-min after cuff deflation. Location of the cuffs, probes, and sample volume depth were photographed and matched during the ischemic exercise condition and all subsequent study measurements.

After a 20-min rest period, maximal brachial artery structure (vasodilatory capacity, VD%) was assessed using an ischemic handgrip exercise condition, which has been extensively used in younger (Naylor et al., 2005) and older cohorts (Green et al., 2018) as a valid index of arterial lumen remodelling. A cuff was placed on the forearm and following 1-min of baseline recording, inflated to 220 mmHg for 5-mins. Using a hand gripper, participants performed hand grip exercise during the middle 3-min of cuff inflation, at a rate of approximately one contraction every 2.4 s (25 contractions per minute) externally paced using a metronome set at 50 beats per minute Screen recordings were captured for the final 30-s of exercise and continued for 4-min after the cuff deflation.

The recordings from the vascular assessments were all analysed by the same researcher (SHM) using custom-designed edge detection and wall-tracking software (FMD%/Blood Flow 4.0, LabView 10.0, National Instruments, TX, USA). Measuring frequency was 30 Hz and were obtained by selecting two regions of interest, one on the Doppler blood flow velocity profile and the other on the B-mode image (Thijssen et al., 2019). Both regions were analysed simultaneously to determine the peak diameter following cuff deflation (automated algorithm) and both FMD% and VD% (change from baseline to peak diameter, %), normalised to shear rate area under the curve (SRAUC [FMD%/SRAUC]). Time to peak was determined as the time from cuff deflation to peak diameter. A blinded re-analysis of ten vascular assessments (brachial and femoral) chosen by a random number generator were used to determine intra-rater reliability (SHM). This was assessed using both the intraclass correlation coefficient (ICC) and the coefficient of variation (CoV%). Baseline and peak artery diameters showed excellent reliability across both metrics (ICC ≥0.996; CoV ≤1.8 %). In contrast, relative change measures (FMD%) and time to peak (TTP) showed good-to-excellent ICCs (0.790–0.922), but higher CoV values (15–29 %), reflecting greater absolute measurement variability as observed by others that assessed the femoral artery (Daniele et al., 2024).

2.4. Performance measures

Athletes completed a graded exercise test at each timepoint (on a separate day to the peripheral vascular assessment) on a rowing ergometer (Concept II, Morrisville, NC), in line with the Rowing Australia Protocol (Rice, 2021). This was used to determine peak oxygen consumption ($\dot{\mathrm{V}}{\mathrm{O}}_2\text{peak}$) with expired breath collected and analysed via a metabolic gas analysis system (TrueOne 2400, Parvo Medics, Inc., UT, USA) as part of the athletes typical monitoring. Workload was assigned for each of the seven 4-min stages based on individual 2 km time trial results, with the final stage being maximal. Heart rate and $\dot{\mathrm{V}}{\mathrm{O}}_2$ were recorded continuously, while blood lactate concentration was taken during the 1-min break between stages with peak exercise heart rate determined as the highest heart rate achieved in the final stage.

2.5. Body composition and anthropometry

A whole-body dual-energy X-ray assessment (DXA; Lunar Prodigy, GE Medical Systems, Madison, WI, USA) was used to assess body composition at ES and LS only, on a separate day to the graded exercise test and peripheral vascular assessment as part of routine athlete monitoring. Total regional (arms and legs) mass, total lean body mass and total body fat percentage were assessed. Body mass and standing stretch stature, relaxed bicep girth (measured at the mid-point between the acromion process and head of the radius) and thigh girth (measured at the mid-point between the anterior superior iliac spine and base of the patella) were assessed by the same International Society for the Advancement of Kinathropometry accredited Anthropometrist (SHM) during the same session as the peripheral vascular assessment. Mass and height were used to determine body surface area (BSA; Mosteller equation; (square root ((height/mass) × 3600)).

2.6. Training volume

All training sessions were tracked throughout the season for each athlete including number and type of session (conditioning or resistance). Heart rate was recorded (Garmin Forerunner 735XT, Garmin International, Inc., USA) during all conditioning sessions including on-water, ergometer and cross-training and stored on TrainingPeaks (TrainingPeaks, Louisville, KY). Heart rate was used as a measure for intensity, based on time (minutes) in each heart rate zone, using a 5-zone model (T1: 50 % of $\dot{\mathrm{V}}{\mathrm{O}}_2$ peakto the midway point between 50 % $\dot{\mathrm{V}}{\mathrm{O}}_2$ peak and lactate threshold 1, T2: T1 to lactate threshold 1, T3: lactate threshold 1–95 % of lactate threshold 2, T4: 95–102 % lactate threshold 2, T5: above lactate threshold 2). Lactate thresholds and $\dot{\mathrm{V}}{\mathrm{O}}_2$ peak were based on the most recent multistage maximal testing and determined using the modified D_max method (Watts et al., 2022). Lactate threshold 1 was defined as the intensity preceding a 0.4 mmol L⁻¹ increase in blood lactate above baseline while lactate threshold 2 was the point on blood lactate intensity curve at maximal distance from a line connecting lactate threshold 1 and the finishing intensity (Bourdon, 2013).

2.7. Menstrual status

Female participants underwent a venous blood sample collected from the antecubital vein and analysed offsite (PathWest Laboratory Medicine, Nedlands, Western Australia) to determine serum hormone concentration (oestradiol E2, progesterone P4 and testosterone T) in the same session following the peripheral vascular assessment. Females also completed a bespoke weekly self-reported menstrual cycle diary, starting one week before ES and asked participants to log start/end dates of their menstrual period or withdrawal bleed. Ovulation testing was not completed due to the observational nature of this study. Hormone analysis and cycle tracking were used to determine cycle length, phase, and number of menstrual periods or withdrawal bleeds, as defined within Elliott-Sale et al. (2021). Given that this was an observational study, we did not control for menstrual status, nor exclude individuals using any type of hormonal contraception or those experiencing any menstrual disturbance.

2.8. Statistical analysis

Statistical analyses were conducted using JASP (Version 0.19.1.0 University of Amsterdam, The Netherlands). To assess differences between the groups (male vs. female) for age, training age and height measured at early-season, a Bayesian t-test was used. To assess the effect of training and differences between sex, a Bayesian two-way repeated measures ANOVA was applied (training × sex). The model best supported by the data for each variable is reported with the Bayes factor (BF₁₀) for all models provided in the supplementary material. Bayes factor was interpreted with the following strength of evidence: <1 supporting the null hypothesis, 1-3 marginal/anecdotal (i.e., insufficient evidence to conclude an effect), 3–10 moderate, 10–30 strong, 30–100 very strong and >100 extreme evidence in favour of a meaningful difference/effect (Faulkenberry et al., 2020).

Bayesian statistics differs from traditional frequentists statistics in their interpretation. Bayesian statistics are not dependent on sample size or underpinned by the distribution of data, offering an advantage for the analysis of small sample sizes, such as within this study. Evaluations are made using BF₁₀ and credibility intervals, instead of p values, to determine the degree of support for an alternative model (>1, i.e. sex, training, sex + training or an interaction) over the null hypothesis (<1, i.e. the variables have no effect on the outcome; (Faulkenberry et al., 2020). Credibility intervals (95% CI) are interpreted in the same way as confidence intervals, corresponding to the bound within which the true value is likely to fall with 95 % certainty (Faulkenberry et al., 2020). In the absence of strong expectations about the size and direction of most effects examined here, we opted to use JASP's default non-informative Cauchy priors for all analyses.

3. Results

Twenty-two athletes were initially recruited, but one female participant withdrew from the study after ES due to injury, resulting in a final sample of ten females (19.2 ± 0.9 years, [95% CI: 18.5, 19.8]) and eleven males (20.4 ± 1.7 years, [19.3, 21.5]). No differences in age or training age (Female: 2.0 ± 1.2 years, [1.1, 2.9]; Male: 3.5 ± 2.0 years, [2.2, 4.9]) were evident between male and female participants (both BF₁₀ = 1.7). Strong sex differences in height were present (BF₁₀ = 22.0) with males taller than females (188.3 ± 6.9 cm, [183.7, 193.0] vs. 177.9 ± 5.9 cm, [173.7, 182.1]). Very strong anthropometric sex differences were evident (Fig. 1) including body mass (BF₁₀ = 44.6), BSA (BF₁₀ = 33.4), LBM (BF₁₀ = 44.2), leg mass (BF₁₀ = 45.9) and arm mass (BF₁₀ > 1000), larger in males in all instances (Table 1). Total body fat was greater in females (BF₁₀ = 32.4) and arm girth was larger in males (BF₁₀ = 7.9). Leg girth differences were supported by independent effects for sex and training (BF₁₀ = 21.5) meaning males had larger leg girths compared to females, and leg girth decreased over time in both sexes, with the lowest value measured at LS. Females had substantially lower systolic blood pressure (BF₁₀ > 1000) and mean arterial pressure (BF₁₀ = 10.8) compared to males. There was no effect on diastolic blood pressure (BF₁₀ = 0.4).

Fig. 1.

Participant characteristics including body mass (A), body surface area (B), dual-energy absorptiometry derived lean body mass (C) and body fat percentage (D) of elite female (n = 10) and male (n = 11) rowers across a competitive season (21 weeks), measured at three timepoints: early-, mid- and late-season. Pink represents females and blue represents male, error bars 95 % credible interval. The model best supported by the data and the Bayes factor (BF₁₀) are reported.

Table 1.

Dual-energy absorptiometry, blood pressure, heart rate, and anthropometric measures (descriptive mean ± standard deviation [95% credible interval]) of elite female (n = 10) and male (n = 11) rowers across a competitive season (21 weeks), measured at three timepoints: early-, mid- and late-season. The best model to explain the data from a Bayesian repeated measures of analysis and the Bayes factor (BF₁₀) are reported.

Dependent Variable	Female			Male			Best model	BF10
	Early	Mid	Late	Early	Mid	Late
Total arm mass, kg	8.0 ± 1.3 [7.2, 8.7]	–	7.5 ± 1.0 [6.8, 8.2]	11.2 ± 1.5 [10.1, 12.2]	–	11.1 ± 1.6 [10.0, 12.2]	Sex	>1000
Total leg mass, kg	26.6 ± 3.0 [24.5, 28.7]	–	26.3 ± 2.5 [24.5, 28.1]	33.3 ± 3.6 [30.7, 35.9]	–	33.0 ± 3.8 [30.3, 35.8]	Sex	45.9
Upper arm girth, cm	28.2 ± 2.5 [26.3, 30.2]	28.0 ± 2.6 [26.0, 30.0]	29.4 ± 3.4 [26.8, 32.0]	31.7 ± 1.7 [30.1, 33.3]	32.2 ± 2.4 [29.9, 34.4]	32.2 ± 1.6 [30.7, 33.6]	Sex	7.9
Upper leg girth, cm	56.3 ± 3.3 [53.7, 58.9]	55.6 ± 2.9 [53.4, 57.8]	54.9 ± 2.8 [52.8, 57.1]	59.4 ± 3.4 [56.6, 62.2]	59.8 ± 3.5 [56.9, 62.7]	58.8 ± 2.9 [56.3, 51.2]	Training + sex	21.5
Systolic blood pressure, mmHg	110.1 ± 8.6 [103.5, 116.7]	109.3 ± 5.6 [105.0, 113.6]	106.2 ± 4.1 [103.1, 109.3]	120.9 ± 6.1 [115.8, 126.0]	126.3 ± 7.8 [119.7, 132.8]	128.9 ± 4.0 [125.5, 132.2]	Sex	>1000
Diastolic blood pressure, mmHg	63.9 ± 8.0 [57.7, 70.0]	61.7 ± 3.9 [58.7, 64.6]	60.1 ± 3.1 [57.7, 62.6]	63.1 ± 5.1 [58.9, 67.4]	59.6 ± 7.7 [53.2, 66.1]	66.3 ± 4.7 [62.4, 70.1]	Training	0.4
Mean arterial pressure, mmHg	79.3 ± 8.0 [73.1, 85.5]	77.6 ± 4.0 [74.5, 80.6]	75.5 ± 2.9 [73.2, 77.7]	82.4 ± 5.0 [78.2, 86.5]	81.8 ± 6.2 [76.6, 87.1]	87.1 ± 3.4 [84.3, 90.0]	Sex	10.8
Resting heart rate, BPM	54 ± 8 [49, 60]	52 ± 9 [45, 58]	51 ± 7 [46, 56]	52 ± 7 [46, 57]	47 ± 5 [45, 58]	47 ± 5 [43, 51]	Training	7.9
Peak exercise heart rate, BPM	200 ± 7 [195, 206]	197 ± 6 [192, 202]	198 ± 6 [193, 203]	199 ± 7 [193, 204]	199 ± 6 [194, 203]	199 ± 8 [193, 205]	Training	0.6

Sex indicates a difference between males and females, training indicates a difference across the training season, Training + sex indicates independent effects of sex and training, interaction refers to an interaction between sex + training + sex∗training.

BF₁₀: level of evidence 0–1: no effect, 1–3: anecdotal, 3–10: moderate, 10–30: strong, 30–100: very strong, >100: extreme.

3.1. Peripheral vascular

All vascular data is presented in Table 2. Males had larger baseline brachial diameter compared to females, both when measured before FMD% and before ischemic exercise condition (BF₁₀ = 144.3 and BF₁₀ = 690.2, respectively). An interaction effect was evident for FMD% brachial peak diameter (BF₁₀ = 723.4, Fig. 3) where males had larger absolute diameter, which peaked at MS, whereas diameter in females progressively increased peaking at LS. When indexed to BSA, no differences were present for baseline diameter of the brachial artery (BF₁₀ = 0.8). Peak diameter of the brachial artery following ischemic exercise was overwhelmingly higher in males (BF₁₀ > 1000). Brachial FMD% SRAUC (BF₁₀ = 25.1) showed a training effect, peaking at MS in both sexes, but no differences were found in SRAUC following ischemic exercise (BF₁₀ = 0.9). An interaction was also observed in brachial artery FMD% (BF₁₀ = 10.9), which was greater in females and continued to increase throughout the season, whereas males peaked at MS (Fig. 2). Normalising FMD% to SRAUC removed any effect of training and sex (BF₁₀ = 0.5). Limited evidence was found for VD% to support either training or sex differences (BF₁₀ = 1.8). No effect was observed for time to peak after occlusion (BF₁₀ = 0.9), VD% normalised for SRAUC (BF₁₀ = 0.5), or following ischemic exercise (BF₁₀ = 0.4).

Table 2.

Upper and lower limb vascular ultrasound outcomes (descriptive mean ± standard deviation [95 % credible interval]) in elite female (n = 10) and male (n = 11) rowers across a competitive season (21 weeks), measured at three time-points: early-, mid- and late-season. The best model to explain the data from a Bayesian repeated measures of analysis and the Bayes factor (BF₁₀) are reported.

	Female Athletes (n = 10)			Male Athletes (n = 11)
	Early	Mid	Late	Early	Mid	Late	Best model	BF10
Brachial Artery Flow Mediated Dilation
Baseline diameter, mm	3.9 ± 0.3 [3.7, 4.1]	3.9 ± 0.3 [3.7, 4.2]	4.0 ± 0.3 [3.7, 4.2]	4.7 ± 0.5 [4.3, 5.1]	4.8 ± 0.4 [4.5, 5.1]	4.7 ± 0.4 [4.4, 5.0]	Sex	144.3
Baseline diameter/BSA, mm/m2	2.1 ± 0.3 [1.9, 2.3]	2.1 ± 0.2 [1.9, 2.3]	2.1 ± 0.2 [2.0, 2.3]	2.2 ± 0.3 [2.0, 2.5]	2.3 ± 0.3 [2.1, 2.5]	2.2 ± 0.3 [2.0, 2.4]	Sex	0.8
Peak diameter, mm	4.1 ± 0.3 [3.9, 4.4	4.2 ± 0.3 [4.0, 4.4]	4.3 ± 0.3 [4.1, 4.5]	4.9 ± 0.5 [4.5, 5.2]	5.1 ± 0.3 [4.9, 5.4]	4.9 ± 0.4 [4.6, 5.2]	Interaction	723.4
FMD, %	5.9 ± 2.3 [4.3, 7.6]	7.6 ± 3.2 [5.2, 9.9]	8.0 ± 2.7 [6.1, 9.9]	4.1 ± 1.7 [2.7, 5.4]	6.6 ± 2.7 [4.5, 8.6]	3.9 ± 2.8 [1.8, 6.1]	Interaction	10.9
SRAUC × 103, s−1,	19.1 ± 8.3 [13.2, 25.0]	28.8 ± 9.0 [22.4, 35.2]	23.4, 8.2 [17.5, 29.2]	19.0 ± 9.0 [12.1, 25.9]	23.2 ± 8.3 [16.9, 29.5]	17.4 ± 4.7 [13.8, 21.0]	Training	25.1
FMD%/SRAUC, × 103, au	0.32 ± 0.08 [0.26, 0.39]	0.26 ± 0.06 [0.21, 0.31]	0.35 ± 0.09 [0.29, 0.41]	0.28 ± 0.06 [0.15, 0.41]	0.31 ± 0.17 [0.18, 0.44]	0.21 ± 0.13 [0.12, 0.31]	Sex	0.5
Time to peak, s	37.9 ± 14.7 [27.4, 48.4]	55.6 ± 17.3 [43.3, 68.0]	41.1 ± 5.9 [36.9, 45.3]	65.8 ± 59.7 [20.0, 111.7]	47.8 ± 8.4 [41.3, 54.3]	43.7 ± 13.6 [33.2, 54.1]	Sex	0.9
Femoral Artery Flow Mediated Dilation
Baseline diameter, mm	6.1 ± 0.6 [5.7, 6.5]	6.2 ± 0.4 [5.9, 6.5]	6.2 ± 0.6 [5.8, 6.5]	6.7 ± 0.5 [6.4, 7.1]	6.9 ± 0.6 [6.4, 7.3]	6.8 ± 0.6 [6.3, 7.2]	Sex	5.8
Peak diameter, mm	6.9 ± 0.8 [6.4, 7.5]	6.7 ± 0.4 [6.4, 6.9]	6.6 ± 0.4 [6.3, 6.9]	7.3 ± 0.5 [6.9, 7.7]	7.3 ± 0.7 [6.8, 7.8]	7.2 ± 0.6 [6.7, 7.7]	Sex	3.6
FMD, %	9.7 ± 7.0 [4.7, 14.8]	8.2 ± 3.4 [5.8, 10.6]	6.8 ± 3.4 [4.4, 9.3]	8.1 ± 5.8 [3.7, 12.6]	6.2 ± 3.7 [3.4, 9.1]	6.9 ± 3.1 [4.5, 9.4]	Sex	0.4
SRAUC × 103, s−1,	18.5 ± 9.3 [11.8, 25.1]	19.5 ± 7.0 [14.5, 24.5]	15.2 ± 7.3 [10.0, 20.5]	20.4 ± 9.2 [13.3, 27.5]	18.8 ± 8.7 [12.1, 25.5]	14.2 ± 5.1 [10.2, 18.1]	Training	0.9
FMD%/SRAUC × 103, au	0.59 ± 0.35 [0.34, 0.84]	0.45 ± 0.28 [0.25, 0.65]	0.48 ± 0.21 [0.33, 0.62]	0.37 ± 0.27 [0.16, 0.58]	0.47 ± 0.48 [0.11, 0.84]	0.53 ± 0.27 [0.32, 0.74]	Sex	0.4
Time to peak, s	81.1 ± 47.2 [47.3, 114.9]	112.4 ± 57.5 [71.3, 153.6]	79.2 ± 44.7 [47.2, 111.1]	99.9 ± 57.8 [55.5, 144.4]	91.2 ± 63.7 [42.2, 140.1]	79.3 ± 48.6 [41.9, 116.7]	Sex	0.3
Brachial Artery Ischemic Exercise Condition
Baseline diameter, mm	3.9 ± 0.4 [3.7, 4.2]	4.0 ± 0.4 [3.7, 4.3]	4.0 ± 0.3 [3.7, 4.2]	4.7 ± 0.4 [4.4, 5.0]	4.9 ± 0.3 [4.6, 5.1]	4.7 ± 0.3 [4.5, 5.0]	Sex	690.2
Baseline diameter/BSA, mm/m2	3.3 ± 0.4 [3.0, 3.6]	3.3 ± 0.2 [3.1, 3.5]	3.3 ± 0.3 [3.1, 3.5]	3.2 ± 0.2 [3.0, 3.4]	3.2 ± 0.3 [3.1, 3.4]	3.2 ± 0.2 [3.0, 3.3]	Sex	0.6
Peak diameter, mm	4.4 ± 0.3 [4.2, 4.6]	4.5 ± 0.4 [4.3, 4.8]	4.5 ± 0.2 [4.3, 4.6]	5.2 ± 0.4 [4.9, 5.5]	5.3 ± 0.3 [5.2, 5.5]	5.2 ± 0.3 [4.9, 5.4]	Sex	>1000
VD, %	13.1 ± 4.5 [10.0, 16.3]	14.7 ± 6.0 [10.4, 19.0]	14.0 ± 3.5 [11.5, 16.6]	11.2 ± 4.2 [7.9, 14.5]	9.4 ± 4.6 [5.9, 12.9]	9.7 ± 3.9 [6.7, 12.7]	Sex	1.8
SRAUC × 103, s−1,	55.5 ± 30.8 [33.4, 77.5]	58.6 ± 26.5 [39.7, 77.6]	57.7 ± 22.9 [41.3, 74.1]	45.0 ± 25.8 [25.1, 64.8]	40.8 ± 16.2 [28.4, 53.3]	48.9 ± 19.0 [34.3, 63.5]	Sex	0.9
VD/SRAUC × 103, au	0.28 ± 0.13 [0.18, 0.37]	0.26 ± 0.09 [0.20, 0.32]	0.27 ± 0.11 [0.19, 0.34]	0.28 ± 0.11 [0.19, 0.37]	0.24 ± 0.08 [0.17, 0.30]	0.21 ± 0.08 [0.15, 0.28]	Sex	0.5
Time to peak, s	86.6 ± 37.6 [59.7, 113.5]	82.1 ± 27.8 [62.3, 102.0]	73.4 ± 13.6 [63.7, 83.2]	82.7 ± 24.3 [64.0, 101.3]	85.9 ± 34.6 [59.3, 112.6]	83.7 ± 25.0 [64.5, 103.0]	Sex	0.4

FMD%: flow-mediated dilation, SRAUC: shear rate area under the curve, VD: vasodilatory capacity.

Sex indicates a difference between males and females, Training indicates a difference across the training season, Training + Sex indicates independent effects of sex and training, Interaction refers to an interaction between sex + training + sex ∗ training.

BF₁₀: level of evidence 0–1: no effect, 1–3: anecdotal, 3–10: moderate, 10–30: strong, 30–100: very strong, >100: extreme.

Fig. 3.

Brachial flow mediated dilation (A), brachial peak diameter (B), femoral flow mediated dilation (C) and femoral peak diameter (D) of elite female (n = 10) and male (n = 11) rowers across a competitive season (21 weeks), measured at three timepoints: early-, mid- and late-season. Each dot represents and individual athlete, white is female, and black is male; the dotted line and pink shaded area represents female average and 95 % credible interval, the solid line and blue shaded area represents male average and 95 % credible interval. The model best supported by the data and the Bayes factor (BF₁₀) are reported.

Fig. 2.

Panel A–B: Peak oxygen performance ($\dot{\mathrm{V}}{\mathrm{O}}_2$ peak) absolute (A) and relative to body mass (B) of elite female (n = 10) and male (n = 11) rowers across a competitive season (21 weeks), measured at three timepoints: early-, mid- and late-season. Each dot represents and individual athlete, white is female, and black is male; the dotted line and pink shaded area represents female average and 95 % credible interval, the solid line and blue shaded area represents male average and 95 % credible interval. The model best supported by the data and the Bayes factor (BF₁₀) are reported. Panel C: Training intensity (average minutes spent in each heart rate training zone per week) in elite female (n = 10) and male (n = 11) rowers across a competitive season (21 weeks). Heart rate zones are relative to lactate thresholds 1 and 2 using a 5-zone model as follows; T1: between 50 % of $\dot{\mathrm{V}}{\mathrm{O}}_2$ peak and the midway point between 50 % $\dot{\mathrm{V}}{\mathrm{O}}_2$ peak and lactate threshold 1, T2: between the top of T1 and lactate threshold 1, T3: between lactate threshold 1 and 95 % of lactate threshold 2, T4: between 95 and 102 % lactate threshold 2, T5: above lactate threshold 2.

Femoral baseline diameter and peak diameter (Fig. 3) demonstrated a moderate effect for sex (BF₁₀ = 5.8 and BF₁₀ = 3.6, respectively), where males presented with larger artery diameter compared to females. This effect on baseline diameter was negligible once indexed to BSA (BF₁₀ = 0.6). No meaningful effects were found for femoral FMD% (BF₁₀ = 0.4), shear-normalised FMD% (BF₁₀ = 0.4), time to peak (BF₁₀ = 0.3) or SRAUC (BF₁₀ = 0.9).

3.2. Menstrual status

Aggregate menstrual status data is presented in Table 3. Six participants were using hormonal contraception at ES testing. Triphasic oral contraception was the most frequent (n = 4; n = 2 LEVLEN® ED (levonorgestrel, ethinylestradiol), n = 2 Yasmin (drospirenone, ethinylestradiol)) with long-acting reversable contraception being used by two athletes (n = 1 implantable device (Implanon NXT®), n = 1 intrauterine device (Mirena ®)). Two athletes changed their hormonal contraception use during the study, one ceased hormonal contraceptive use altogether, and one changed from combined oral contraceptive pill to a hormonal intrauterine device between MS and LS (Yasmin to Mirena®). Of those who were initially not using hormonal contraception (n = 4), only one could be classified as naturally menstruating with an average cycle length 26.0 ± 1.7 days and six reported periods during the study. According to Elliott-Sale et al. (2021), two athletes were classed as oligomenorrhea due to cycle length >35 days (41.4 ± 18.4 and 43.0 ± 21.2, respectively) and one athlete was classified as having exercise-induced secondary amenorrhea due to the absence of menstrual bleeds for the duration of the study. These menstrual irregularities were classified based on self-reported information and not diagnosed by a medical professional. A detailed table including cycle length, hormone concentrations and use of contraceptive use is available in supplementary materials (Supplementary Table 4).

Table 3.

Aggregated menstrual status and hormone profiles elite female rowers (n = 10) across a competitive season (21 weeks), measured at three time-points: Early-, Mid- and Late-Season. Individual data is provided in supplementary materials (Supplementary Table 4).

Menstrual Status	n	Mean interval between bleeds days ± pooled SD	Mean number of bleeds ± SD	Mean E2 (pmol/L)	Mean P4h (nmol/L)	Mean T (nmol/L)
Naturally menstruating (NM)	5a	35.5 ± 13.3b	3.6 ± 1.9	Earlyc: 140 Mid: 394 ± 204 Late: 329 ± 283	Earlyc: 10.0 Mid: <1–31 Late: <1–36	Earlyc: <1 Mid: 1.0 ± 0.2 Late: 1.1 ± 0.2
Oral contraceptive pill (OCP)	4a,d	31.6 ± 14.4	4.5 ± 0.6	Earlyc,e: <40 Mide: <40 Latee: <40	Earlyc: <1 Mid: <1–2 Late: <1	Earlyc: 0.7 Mid: 0.8 ± 0.2 Late: 0.7 ± 0.1
Hormonal intrauterine device (IUD)	2d	36.3 ± 16.6f	2.5 ± 2.1	Early: 170 Mid: 1700 Lateg: 169 ± 100	Early: <1 Mid: <1 Late: <1–1.0	Early: 1.0 Mid: 1.2 Late: 1.0 ± 0.2
Implant	1	45.7 ± 35.8	4	Early: 110 Mid: 240 Late: 140	Early: <1 Mid: <1 Late: <1	Early: 0.6 Mid: 0.9 Late: 0.8

E2: Oestradiol, P4: Progesterone, T: Total testosterone.

^an = 1 athlete transitioned from OCP to NM between early- and mid-season.

^bBased on n = 4 athletes as one athlete experienced 1 period over the 21-week study.

^cBased on n = 1 athlete as blood samples unable to be obtained.

^dn = 1 athlete transitioned from OCP to IUD between mid- and late-season.

^eAll E2 values < 40 pmol/L were below the detection limit of the assay and are consistent with expected suppression under hormonal contraception.

^fBased on n = 1 athlete who transitioned from OCP to IUD between mid-and late-season.

^gBased on n = 2 athletes as n = 1 transitioned from OCP to IUD between mid- and late-season.

^hRange provided as several measured values below the detection limit of the P4 assay (<1 nmol/L).

3.3. Performance measures

Independent effects of time and sex were evident for absolute $\dot{\mathrm{V}}{\mathrm{O}}_2$ peak (BF₁₀ > 1000). Males had a greater $\dot{\mathrm{V}}{\mathrm{O}}_2$ peak compared to females, and $\dot{\mathrm{V}}{\mathrm{O}}_2$ peak was lowest for all athletes at ES (Fig. 2). When expressed relative to body mass, an interaction was present for $\dot{\mathrm{V}}{\mathrm{O}}_2$ peak (BF₁₀ > 1000) where males had consistently higher relative $\dot{\mathrm{V}}{\mathrm{O}}_2$ peak compared to females at all timepoints. However, relative $\dot{\mathrm{V}}{\mathrm{O}}_2$ peak increased following Block 1 in both sexes, which continued to increase after Block 2 in females only, reaching a highest value at LS. Males however experienced a decrease in relative $\dot{\mathrm{V}}{\mathrm{O}}_2$ peak after Block 2, with the highest value recorded at MS.

3.4. Training volume and intensity

Training volume did not differ across sex or throughout the season for number of conditioning sessions per week (BF₁₀ = 0.6; Females Block 1: 11.0 ± 3.4 [8.6, 13.5], Block 2: 12.7 ± 2.7 [10.7, 14.6]; Males Block 1: 12.6 ± 3.2 [10.4, 14.7], Block 2: 12.5 ± 4.2 [9.7, 15.3]) or resistance sessions per week (BF₁₀ = 1.2; Females Block 1: 2.9 ± 1.1 [2.1, 3.7], Block 2: 2.3 ± 0.6 [1.9, 2.6]; Males Block 1: 1.7 ± 1.3 [0.8, 2.6], Block 2: 2.1 ± 0.9 [1.5, 2.7]). Time spent at each heart rate zone did not differ for sex or across the season for T1, T2, or T3 (all BF₁₀ < 1.0, Fig. 2). An interaction effect was present for time spend in T4 and T5, with moderate evidence for T4 (BF₁₀ = 4.7) and strong evidence for T5 (BF₁₀ = 32.8) indicating females spent less time in T4 and T5 compared to males, however males spent more time overall in T4/T5 during Block 1 compared to Block 2 (Fig. 2).

4. Discussion

This study assessed sex-specific arterial adaptation in the upper and lower limbs of elite rowers across a 21-week competitive training season. Clear distinctions between sexes were present in baseline and peak brachial artery diameter, however only moderate evidence was found for the differences within the femoral artery. Sex-specific adaptation was evident across the season within the brachial artery, specifically peak diameter and to a lesser extent, FMD%, which continued to increase across the season in females whereas males peaked mid-season.

This study showed strong evidence for distinct differences between sexes in the peripheral arteries of elite rowers, with baseline diameter of the brachial artery and peak diameter larger in males compared to females. Further to this, females had increased FMD% compared to males, likely compensating for the smaller brachial diameter (Green et al., 2013), indicative of the classic ‘athlete artery’ phenotype. Conversely, within the femoral artery only a moderate difference between sexes were evident for baseline diameter, and no difference in FMD%. With already enlarged femoral arteries compared to non-athletic controls from other studies (male: 6.45 ± 0.58 mm (Naylor et al., 2021), female: 5.5 ± 0.4 mm (Daniele et al., 2024), it is unclear if the findings in our study emphasise an existing training effect ceiling or highlights the lack of sex-differences within the femoral artery of trained athletes. Previous literature has consistently shown that individuals with enlarged artery structure can paradoxically present with decreased FMD%, particularly in athletes (Green et al., 2013; Naylor et al., 2006, 2021; Rowley et al., 2012), and also that sex is a key driver of vascular adaptation (Green et al., 2023, 2025; Holder et al., 2019; Moreau et al., 2024). Further, it is hypothesised that functional adaptations precede structural adaptations, which can result in a ‘normalized’ FMD% (Green and Smith, 2018), which may have occurred to some extent across the training season. There are myriad factors that can affect vascular diameter, including but not limited to, oestrogen status (Moreau et al., 2013), body size (Naylor et al., 2021) and exercise modality (Green et al., 2023). Within the present study, sex differences were voided when baseline diameter was indexed to BSA, suggesting arterial size, particularly in athletes, may be strongly associated with body size, particularly in a cohort of athletes with a similar training load. Due to the substantial variability in menstrual status among our female participants, it is difficult to determine the exact role of oestrogen on vascular measures within this population. Nevertheless, a sex difference is evident within the brachial artery of elite level rowers, despite a lack thereof within the femoral artery. However, owing to the observational that normalising FMD% to shear rate negated these differences, highlights the notion that exercise-induced adaptation may be mediated by shear, regardless of sex.

We observed sex-specific training-induced adaptation in the brachial artery, specifically peak diameter and FMD%, but not the femoral artery. Braichal artery peak diameter in females consistently increased throughout the season, with the highest value recorded at LS. In contrast, males peaked at mid-season, with a decline towards the end season. In females, brachial artery FMD% followed a similar pattern by continuing to increase during the season, whereas in males, FMD% peaked at mid-season. In the femoral artery however, neither diameter nor FMD% responded to training in either sex. Together, this suggests continued plasticity of the brachial artery, as evident in a previous study in elite male rowers, detailing an increase in brachial artery diameter upon returning to training after off-season (Naylor et al., 2006). Unlike Naylor et al. (2006), the present study continued to show changes throughout the season. Our findings also reinforce the notion of localised haemodynamic stimuli favouring upper limb vasculature adaptation in rowers (Rowley et al., 2012; Tao et al., 2023). The repetitive nature of regional loading in the upper limb during rowing (e.g. high intensity pulling, isometric gripping), along with the emphasis on concurrent resistance training, may result in greater continuous upper limb shear stimulus compared to the lower limb. Despite the necessity of the lower limbs in power production during the drive phase of the rowing stroke, the more cyclical eccentric-concentric loading pattern may result in a more pulsatile, interrupted flow leading to relatively lower shear exposure, especially when compared to lower-limb dominant endurance exercise like running and cycling (Thijssen et al., 2009). Furthermore, this high-force loading of the legs during rowing may indeed blunt or counteract any endurance-related adaptation signalling in the femoral artery by instead promoting arterial stiffness or reduced endothelial sensitivity (Königstein et al., 2023). However, as we did not directly assess the underlying signalling mechanisms in the present study, this interpretation remains speculative.

Ashor et al. (2015) revealed a positive relationship between chronic aerobic exercise intensity and FMD%. Assessing training across the season in our athletes, the sexes were matched in most areas except time spent in T4 and T5, with females recording less time (∼17 min) at higher intensities, yet a consistent total training duration per block. In comparison, males completed more time (∼11 min) at higher intensities in Block 1 (ES-MS) compared to Block 2. This increased time at a higher intensity in Block 1 could explain the peak in males at MS, adding an increased ‘dose’ of high-intensity minutes (Ashor et al., 2015), yet does little to explain the continual increase in brachial diameter and FMD% in females throughout the season. Furthermore, our study shows synchronized changes in both brachial FMD% and diameter, contrasting previous literature (Tinken et al., 2008). Tinken et al. (2008) assessed vascular function and diameter at 2-week intervals, unlike our study, which conducted measurements at greater intervals and therefore may have resulted in us being unable to detect these small variations. The authors also recruited “recreationally active” males, not athletes hence the cohort of athletes in the present study were already highly-trained, suggesting they may have already experience arterial adaptation as a result of chronic training prior to assessment. This is further evidenced when comparing brachial baseline diameter of male rowers 4.7 ± 0.5 mm vs untrained males' 4.2 ± 0.4 mm (Tinken et al., 2008). Also, we observed that brachial VD% did not change with training or differ by sex, which may reflect a functional ‘ceiling’ of the structurally enlarged conduit arteries of the athlete phenotype (Green et al., 2012). Only a small proportion of studies on exercise-induced vascular adaptation has focussed exclusively on athletes (18 %), with most (88 %) of these being cross-sectional comparing athletes to sedentary controls (Thompson et al., 2024), highlighting a need for more research into the specific time-course of change within competitive athletes.

This study aimed to address the methodological considerations for female participants and included menstrual status data, highlighting the importance of transparency when including female population (McNulty et al., 2020; Smith et al., 2022). Recent research has shown that hormonal phases within naturally cycling and oral contraceptive users has minimal influence on endothelial function (Williams et al., 2020) however this is highly dependent on the type of contraceptive, dose, duration of use and administration route (Stone et al., 2024). Within the homogenous group of female athletes included in this study there were large variation in contraception use, regularity of menstrual cycle and in some cases, menstrual dysfunction. Impairment of FMD% has been linked to menstrual dysfunction, specifically the oestrogen deficiency that occurs in secondary amenorrhea (Tegg et al., 2024), likely in at least one athlete within this study. Menstrual dysfunction could also indicate low-energy availability, particularly during high training loads, such as within competitive season (Mountjoy et al., 2023). This reiterates the need for a multidisciplinary approach to working alongside female athletes to ensure optimum health and performance.

An important limitation to acknowledge within this study is the sample size. Rowing athletes included within this study were classified as a minimum of Tier 3 and our sample captured most of the athletes fitting this criterion within an immediate geographical area. Also, Bayesian statistics used within the analysis compensate for sample size, providing a more robust methodology for analysis with the current sample. The present study was not designed to directly assess underlying physiological mechanisms of arterial remodelling; however, the observed phenotypic adaptations offer relevant insights into athlete responses to prolonged training stimuli and provide a foundation for future mechanistic investigations. We also did not employ a non-athletic control group which limits the ability to make direct comparisons to untrained vascular phenotypes. Also, we did not observe an inverse diameter–FMD% relationship in the brachial artery of our athletes; both increased with training. The stronger evidence for diameter change suggests that structural adaptation may be more prominent than functional change in this cohort, aligning with aspects of the ‘athlete's artery-paradox’ hypothesis. Finally, training data was limited to group averages outlining the number of sessions per week and time spent in respective heart rate zones. We acknowledge that quantifying training load, specifically in rowing athletes, is difficult due to the multimodal nature of training and therefore our summary may be limited (Watts et al., 2024).

5. Conclusions

In summary, we found distinct differences in the brachial artery when comparing female and male elite rowers. Specifically, males had larger brachial baseline and peak diameters while females had increased FMD%, however this appears to be entirely shear-mediated. Further, these differences were not observed in the lower limb peripheral vasculature. Differences in anthropometric variables were apparent between sexes, such as height, weight, BSA, LBM, fat free mass and, leg and arm mass, likely attributing to the vascular discrepancies between sexes. Despite completing the same volume of training, males completed more time at higher heart rate intensities, thereby potentially contributing to a divergent response to training.

CRediT author statement

Henley-Martin: conceptualization, methodology, formal analysis, investigation, writing – original draft, visualization, project administration, Brade: conceptualization, methodology, investigation, writing – review & editing, supervision, Riddell: methodology, formal analysis, writing – review & editing, supervision, Watts: investigation, writing – review & editing, project administration, Maiorana: writing – review & editing, supervision, Naylor: writing – review & editing, supervision, resources, Binnie: investigation, writing – review & editing, project administration, supervision, resources, Spence: conceptualization, methodology, investigation, writing – review & editing, visualization, project administration, resources.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

Data will be made available on request.