Genotype-Based Stratification of Skeletal Muscle and Fat-Free Mass in Elite Athletes: A Pilot Classification and Regression Tree Analysis

Background/Objectives: Body composition is a key determinant of athletic performance and is influenced by both training and genetic factors, yet the combined contribution of several genetic polymorphisms remains insufficiently understood in elite athletes from different sports. This exploratory pilot study aimed to apply classification and regression tree (CRT) analysis to stratify elite ice hockey players and combat sport athletes by genotype with respect to skeletal muscle mass (SMM) and fat-free mass (FFM), and to assess, by cross-validation, whether the resulting structure holds beyond the sample in which it was derived. Methods: In an exploratory secondary analysis of an existing cohort, combat sport athletes (n = 44) and ice hockey players (n = 48) were examined. Body composition was assessed by bioimpedance analysis and eight single-nucleotide polymorphisms were genotyped by polymerase chain reaction. Separate CRT models were constructed for SMM and FFM in each group with the polymorphisms as the only independent variables; model performance was assessed by 10-fold cross-validation. Results: In combat sport athletes the trees included IL6 rs1800795, UCP3 rs1800849 and KCNJ11 rs5219 and accounted for 20.7% of the variance in SMM in the training data; in ice hockey players they included ADRB2 rs1042713, KCNJ11 rs5219 and IL6 rs1800795 and accounted for 17.4%. In every model the cross-validated risk estimate exceeded the total variance of the outcome (cross-validated R2 = −0.27 and −0.32 for SMM in combat sport athletes and ice hockey players, respectively), and standard deviations within terminal nodes remained close to those of the whole groups. ADRB2 rs1042713 produced the most consistent signal: in ice hockey players it formed the first split of both trees, SMM decreased with the number of A alleles (β = −1.47 kg per allele, p = 0.085), and the A/A genotype was more frequent than in combat sport athletes (25.0% vs. 4.5%); compared with reference data for the general Kazakh population (17% A/A), this difference reflected a depletion of A/A among combat sport athletes rather than an excess among ice hockey players. Conclusions: CRT analysis identified subgroups of athletes differing in SMM and FFM, but the models had no predictive value beyond the sample in which they were built. ADRB2 rs1042713 emerged as the single candidate warranting replication, and its depletion among combat sport athletes relative to population reference data requires confirmation in studies with matched controls. These findings are preliminary and hypothesis-generating; they do not establish causal relationships and require validation in larger, independent cohorts.