Prospective Cohort Study of Villarreal CF Youth Players

Quality Assurance Plan: The quality of the data and project processes is guaranteed through a centralized technological infrastructure and a strict biological sample traceability protocol. The pseudonymized database is hosted within the institutional cloud-based data infrastructure of the CEU Cardenal Herrera University (UCH-CEU). The system implements restricted access control through personal, non-transferable credentials, institutional authentication, and automated operation logging (activity logs). The laboratories responsible for biological processing (saliva and stool) operate under international quality certifications ISO 9001, ISO 15189, and ISO 27001. Additionally, the genetic bioinformatic analysis integrates an automated quality control system (QC-System) that filters and validates genetic variants based on international standards (call rate >98%, Hardy-Weinberg equilibrium, and minor allele frequency/MAF).

Data Checks:Automated and manual consistency checks are applied across the different analytical dimensions. In the bioinformatic layer, the AIG (Automated Intelligence Genetics) platform executes automatic duplication checks and filters artifacts during DNA sequencing. In the microbiome dimension, the bioinformatic workflow utilizes QIIME2® and DADA2 tools alongside positive controls (standard microbial communities) and negative controls (blank extractions) to identify and correct methodological biases or batch effects. For the biomechanical (LESS test) and clinical assessments (Lachmeter®), three valid trials are performed per participant, and the measurements are averaged, verifying internal consistency via the intraclass correlation coefficient (ICC), standard error of measurement (SEM), and smallest detectable change (SDC).

Source Data Verification (SDV):The study design requires cross-referencing records with external platforms to ensure the accuracy of longitudinal information. Daily physiological data (resting heart rate and nocturnal heart rate variability) are automatically synchronized from the WHOOP Strap 4.0 wearable devices to the internal monitoring platform of Villarreal CF. Furthermore, the primary variable-injury incidence-is verified through direct clinical confirmation by the Villarreal CF medical team following the club's standardized diagnostic criteria, which are prospectively contrasted against weekly exposure records filled out under technical staff supervision.

Data Dictionary:The protocol defines a set of multiscale variables coded under a unique alphanumeric code for each participant, omitting any directly identifiable information.

The variable dictionary explicitly includes:

• Biomechanical Biomarkers: Peak vertical force, vertical loading rate (first 50-100 ms), mediolateral/anteroposterior forces, contact time, jump height, and Reactive Strength Index (RSI) using ForceDecks platforms, evaluating interlimb asymmetries with a clinical relevance threshold set at >10-15%. Normalized muscle activation (%) and median frequency (Hz) obtained via BTS FreeEMG® surface electromyography. Anterior tibial displacement in millimeters measured by the Lachmeter® device (considering a side-to-side difference >3 mm as clinically relevant).
• Physiological Biomarkers: Nocturnal heart rate variability (HRV) and resting heart rate (RHR) measured via optical photoplethysmography (PPG).
• Genetic Biomarkers: Single nucleotide variants (SNVs) and insertions/deletions within targeted gene categories (ACTN3 rs1815739, ACE rs4343, COL5A1 rs12722, COL1A1 rs1800012, IL6 rs1800795, SOD2 rs4880, PPARGC1A rs8192678, MCT1 rs1049434, CYP1A2 rs762551, FTO rs9939609, MTHFR rs1801133, VDR rs2228570, ADRB2 rs1042713).
• Microbiome Biomarkers: Alpha and beta diversity indices, and taxonomic profiles organized into 14 characteristic profiles (clusters) using FISABIO's CLOM platform.
• Exposure and Injury Variables: Injury type, location, mechanism, time-loss duration, and number of exposure hours to training sessions and competitions (questionnaire adapted from the UEFA Champions League injury study).

Standard Operating Procedures (SOPs):The study operations are governed by sequential, standardized processes:

• Recruitment: Coordinated with the club in consecutive phases including a verbal/written information session, signing of informed consent (or written assent for minors aged ≥12 along with parental/legal guardian consent), and immediate assignment of a pseudonymization code.
• Data Collection: Genetics are collected only once at baseline via buccal swab (using the Overgenes® kit). Gut microbiome, biomechanics (LESS and sEMG), and clinical laxity are systematically measured twice per season (beginning and end). Physiology is recorded daily. Exposure and injury data are logged weekly.
• Management of Individual Findings / Incidental Findings: If high-risk injury profiles are identified, a confidential individual communication protocol is triggered through Dr. Simón Bueno López (Club Medical Service), respecting the participant's signed "right not to know". Disclosing individual data to coaching or technical staff, or using it for contractual or sports selection purposes, is strictly prohibited.
• Management of Changes: A progressive transition from 16S rRNA sequencing to shotgun metagenomics is planned, requiring a technical equivalence control report to be submitted to the Ethics Committee as a minor amendment prior to operational implementation.

Sample Size Calculation:The calculation is based on the documented incidence of ACL injuries in Spanish First Division professional football (approximately 0.0364 injuries per 1,000 total exposure hours) and the significantly higher relative risk (2- to 8-fold higher) observed in female players. Under the assumptions of a 95% confidence level and 80% statistical power, it was estimated that at least 16 to 20 incident ACL cases are required to achieve adequate power for sex-based comparisons. Given that a professional team accumulates roughly 30,000 exposure hours per season, it was determined that this number of events can be achieved through the longitudinal follow-up of a cohort of approximately 200 to 250 athletes monitored continuously over a minimum of two to three consecutive seasons.

Statistical Analysis Plan:The integration of data into a learning matrix is structured across six successive phases combining classical statistics and machine learning:

• Phase 1: Calculation of means, standard deviations, medians, and confidence intervals to characterize the cohort and identify potential confounding variables.
• Phase 2: Group differences evaluated using parametric tests (t-test, ANOVA) or non-parametric tests (Mann-Whitney U, Kruskal-Wallis) depending on data distribution, applying the Benjamini-Hochberg procedure (FDR) for multiple comparison adjustments.
• Phase 3: Logistic regression models to estimate injury risk adjusted for predictors (sex, age, load) and Cox proportional hazards models to analyze time-to-injury events.
• Phase 4: Development of predictive models optimized through the AIG framework using random forests, gradient boosting, penalized regression (LASSO), and multilayer neural networks, utilizing k-fold cross-validation (k=5 or 10) and bootstrap resampling to prevent overfitting.
• Phase 5: Analysis of ROC curves, calculation of the Area Under the Curve (AUC), Hosmer-Lemeshow calibration tests, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). - Phase 6 (Score Development): Integration of variables with the highest predictive weight into a multiscale quantitative index (score) to assign each player an individualized risk level (low, moderate, high), internally validated via cross-validation or temporally by season.