This prospective study, approved by the institutional ethics committee in accordance with the Declaration of Helsinki (Protocol code PV4764-3108-BO-ff, date of approval July 29, 2014), recruited competitive amateur cyclists and healthy controls between March 1, 2018, and October 31, 2019. Informed consent was obtained. Cyclists were identified through advertisements at local, registered cycling clubs.

Participants were included if they met the following criteria, based on the standardized self-reported questionnaire: (a) regularly participation in cycling competitions across various distances, and (b) cycling at least 9,000 km in the past year and/or training on average of at minimum 10 h per week. Study exclusion criteria included contraindications for cardiac MRI, any cardiovascular or systemic diseases, and use of illicit drugs.

Cyclists were asked to report their lifetime competition history, years of competitive cycling, weekly exercise volume (in hours), and number of cycling competitions per year using a standardized questionnaire. Cyclists’ intrinsic motivation to compete was assessed using a study-specific questionnaire (Supplementary Material, Table S1) that evaluated individual’s motivation to compete and their competitive drive (minimum-maximum points: 10–70; higher score- more win-oriented, strong competitive drive).

Control participants were included if healthy and if they averaged less than 3 h of exercise per week. Controls were subject to the same exclusion criteria as the cyclists, and ineligible if they engaged in regular structured athletic training.

Prior to cardiac MRI exams, all subjects were instructed to refrain from exercise for 72 h. Prior to cardiac MRI blood samples were drawn from an antecubital vein to obtain hematocrit and markers of cardiac injury, including high-sensitivity troponin T (hs-TnT), creatine kinase (CK) and N-terminal pro–B-type natriuretic peptide (NT-proBNP). Body surface area (BSA) and body mass index (BMI) were calculated from the study participant’s body weight and height.

We note that this cyclist collective was included in a previous report focusing on prognostic outcomes in a mixed-athlete population [8]. The current manuscript represents the primary, detailed investigation of this specific cohort, providing cyclist-specific data and analysis not addressed in the prior aggregate study.

Cardiac MRI exams were performed using a 3.0T clinical MRI system (Achieva, Philips Healthcare, Netherlands) with 32-channel cardiac phased array receiver coil. The protocol (Supplementary Material, Table S2) included gated CINE imaging using balanced steady-state free-precession (bSSFP) in the short axis for measuring LV and right ventricle (RV) volumes and LV mass.

T1 mapping was performed using a Modified Look Locker Inversion Recovery (MOLLI) sequence on 3 double-oblique left ventricular short-axes (SA) slices (apical, mid, and basal) before and 15 min after administration of 0.2mmol/kg gadoterate meglumine (Dotarem, Guerbet, Sulzbach, Germany; rate 2.5 ml/s). T2 mapping was performed using a black-blood, multi-echo GRASE-based sequence on 3 double-oblique left ventricular SA slices (apical, mid, and basal) prior to contrast administration. T1- and T2 maps were generated using a third-party cardiovascular software (cvi42, Circle Cardiovascular Imaging Inc., Calgary, Alberta, Canada).

For LGE imaging, end-diastolic images were acquired 10 min after contrast administration using phase-sensitive inversion recovery (PSIR) sequence in SA orientation covering the entire heart and in standard long-axis views (2-,3- and 4-chamber).

Two observers, a physician (FA) under the supervision of radiologist with 13 years cardiac MRI experience (JS), and a cardiologist with 7 years cardiac MRI experience (TT) both blinded to clinical data, performed image analysis using cvi42 software (Circle Cardiovascular Imaging Inc, Calgary, Alberta, Canada). For volumetric and mass analysis, segmentation of the LV and RV endocardial and epicardial borders was performed using semi-automated contouring, with manual adjustments in standard fashion on SA images [22]. As part of the segmentation, papillary muscles and trabeculae were excluded from the LV mass but included in the LV volume in the primary analysis. All measurements were indexed to BSA. A secondary analysis was performed by a single experienced radiologist (JS) using semi-automated contouring with manual adjustments to reclassify papillary muscles and prominent trabeculations as part of the myocardial mass to allow comparison with prior athletic cohort studies using this type of segmentation.

Native T1, T2, and post-contrast T1 times were measured using a single mid-septal region of interest (ROI) on midventricular short-axis images, excluding areas of focal LGE if present. ECV was calculated using the standard formula [9]. Results are given as the mean of two observers’ measurements.

Focal myocardial fibrosis was defined as an area of LGE present in at least 3 consecutive SA images, or confirmed on perpendicular long-axis images, ensuring the findings represents a significant tissue volume rather than minor transient artifacts at the RVIP of LV or elsewhere. To maximize diagnostic accuracy and eliminate inter-observer variability all images were reviewed in a side-by-side consensus agreement by two observers with 13- and 20-years cardiac MRI experience (JS and GKL), respectively. Athletes with LGE present exclusively at the RV insertion point (RVIP) were classified as RVIP LGE cyclists. Those with LGE in other areas, or in both the RVIP and other locations, were classified as non-RVIP LGE cyclists. Accordingly, non-RVIP LGE was considered the primary phenotype of interest for pathological remodeling analyses, whereas RVIP LGE was analyzed descriptively as a comparator pattern. Quantitative analysis of LGE was performed on SA LGE images by one observer (JS) using a threshold method, with a cutoff of more than 5 standard deviations (SD) above remote, healthy myocardium.

A cardiopulmonary exercise test was performed using a ramp-incremental protocol on an eddy current-braked, high-performance, sport-specific cycle ergometer (Lode Excalibur Sport 911900, Lode BV, Groningen, Netherlands). Subjects wore an appropriately sized silicone oronasal face mask that was securely fitted to their heads using polyurethane headgear (7450 SeriesV2 MaskTM, Hans Rudolph, Inc., Kansas, USA).

The ramp-incremental protocol differed for males and females to ensure all subjects reached maximum exhaustion within 10–12 min. The starting load was 50 W, with a continuous increase of 40 W/min for males and 30 W/min for females. Total exhaustion was defined as a cycling cadence falling below 50 rpm. Breath-by-breath gas exchange analysis was performed throughout the exercise test, to determine maximal oxygen uptake (VO2max), a measure of the participant’s cardiorespiratory capacity and endurance, along with the maximal power output (W) at exhaustion [23].

A 12-lead electrocardiogram (Custo Cardio 100 and 300, Custo Med GmbH, Ottobrunn, Germany) was continuously recorded at rest, during the ramp-incremental protocol, and for 5 min post-exercise. Blood pressure was measured manually every two minutes. Resting and peak systolic blood pressure during exercise, along with heart rate, were recorded.

In addition, body fat was measured using a Harpenden Skinfold Caliper (Baty International, Burgess Hill, England). Skinfold thickness at the ten sites required for the Parizkova method was measured to the nearest 0.1 mm, and body fat was estimated using sex-specific Parizkova’s Eqs [24, 25].

Statistical analysis was performed using the packages table1 (v1.4.3) and tidyverse (v2.0.0) in R (version 4.1.1). Continuous data are presented as mean ± SD, and categorical data are presented as absolute numbers and percentages. Continuous data were compared using two-sided Student’s t tests, and categorical variables were compared using chi-square or Fisher exact test, as appropriate. Given the exploratory nature of the subgroup analyses and the limited sample sizes in specific cohorts, p-values were not adjusted for multiple comparisons to minimize the risk of Type II errors (false negatives) and avoid masking potentially relevant physiological signals [26]. However, Cohen’s d effect sizes are provided to contextualize the magnitude of observed differences.