Biochemical and Perceptual Markers of Physiological Stress During Acute Exercise Overload in U20 Elite Basketball Players

Abstract

The allostatic load index (AL_index) measures the cumulative physiological burden on the body due to stress. This prospective cohort study examined the relationships between certain molecular biomarkers, physical variables, and psychometric variables during deload and overload microcycles to contribute to developing an AL_index in professional team-sport athletes. Twelve elite male basketball players (18.3 [0.9] years; 77.2 [5.7] kg; 185 [9.0] cm) were monitored during two microcycles (deload and overload). Blood creatine kinase (CK) and urea levels, countermovement jump (CMJ), session-RPE (RPE × session duration [min], its exponentially weighted moving average [EWMA]), and a cumulative wellness score (sleep, stress, fatigue, muscle soreness, and mood) were assessed at different time points. Bayesian and robust statistics (Cohen’s ξ) were employed. CK rose from 222 U/L (deload) to 439 U/L (overload; +98%, large effect ξ = 0.65), while session-RPE load more than doubled (270 [269] AU to 733 [406] AU, ξ > 0.8). No difference was found in urea and wellness scores (cumulative or other components). CK levels showed moderate positive correlations with both EWMA of session-RPE (ρ = 0.346, p = 0.002) and reduced sleep quality (ρ = 0.25, p = 0.018). Bayesian modeling identified the EWMA of session-RPE as the strongest predictor of jump-defined fatigue (β = 0.012, 95% HDI [0.004, 0.021]), while CK demonstrated a small negative association (β = −0.009, HDI [−0.016, −0.001]). Finally, a principal component analysis (PCA) revealed that CK and the EWMA of session-RPE were robust indicators of physiological stress. A parsimonious index based on PCA loadings ([0.823 × CK] + [0.652 × EWMA of session-RPE]) demonstrated strong discriminative validity between microcycle phases (overload: 515, 95% HDI [442, 587] versus deload: 250, 95% HDI [218, 283], BF₁₀ > 100,000). CK and session-RPE may serve as sensitive biomarkers for inclusion in the AL_index for team sport athletes.

1. Introduction

High-intensity exercise-induced stress triggers systemic responses across the muscular, neuroendocrine, and nervous systems. The application of the allostatic model to physical training facilitates understanding of the systemic changes generated by exercise, given the magnitude of the applied stimulus [1]. This paradigm provides insight into the adaptation process as variational and relational stability [2], which enables anticipation of stress-related requirements in an athlete’s physiology [3]. Given the inherent capacity of organic systems for physiological resilience, monitoring the changes in critical biomarkers may help optimize adaptive training outcomes by allowing for appropriate adjustments in training loads and preventing adverse health conditions in sports, such as overtraining or relative energy deficiency in sport (RED-S) syndromes [1].

Creatine kinase (CK) and urea are widely utilized biomarkers in sports [4,5,6]. Their widespread use underscores their pivotal role in identifying overtraining and muscular damage, as they correlate notably with diminished exercise tolerance [6]. Variations in peak values and time intervals for these biomarkers have been observed. For example, CK elevations typically occur within 24 to 48 h post-exercise, returning to baseline levels between 48 and 120 h after training [4]. These variations depend on internal and external factors, including sport type, athlete position, and microcycle intensity and volume [7,8]. Although CK serves as a robust sports biomarker [9,10], combining several validated biomarkers increases sensitivity, enabling earlier detection of exercise-induced physiological stress when monitoring performance, recovery, and health in athletes [11].

In addition to molecular biomarkers, subjective self-reported scales have been developed as alternative methods for monitoring training load due to their cost-effectiveness and simplicity [12,13,14]. In particular, session rating of perceived exertion (session-RPE, also known as the Foster index) [15] has been shown to be valid for monitoring internal training load for children and adolescent athletes [16]. It is one of the most frequent variables used to determine the amount of fatigue and monitor training load in elite team sports athletes [17,18,19]. Other measurements to assess neuromuscular readiness and sports performance, providing insights into athletes’ recovery from training, include the countermovement jump (CMJ) [20,21]. The increase in internal and external load has been shown to be negatively related to jump height and other variables of the CMJ in various sports modalities [21,22,23,24]. Hence, CMJ’s value for fatigue monitoring in basketball [25].

The allostatic load index (AL_index) is used to measure the cumulative physiological burden on the body due to stress [3,26]. It is calculated from a panel of molecular and physiological biomarkers representing sub-clinical or clinical deviations to assess the overall load on the main physiological systems (i.e., cardiovascular, metabolic, immune, and neuroendocrine) [3]. Usually, these biomarkers are given equal weightage and added to achieve an index from 0 to approximately 9–12 [27]. With over 30 years of research in both general and clinical populations [28], this topic is moving toward sports sciences [1,29,30,31], providing a more comprehensive understanding of the dose/load–response/effect relationship. This study aimed to examine relationships between molecular, physical, and psychometric biomarkers across deload and overload microcycles. Given the versatility of the AL_index in component selection, we expect to identify key elements suitable for inclusion in this measure.

2. Results

2.1. Participants

Twelve professional male athletes from the Antioquia basketball team participated and completed this prospective cohort study (18.3 [0.9] years; 77.15 [5.7] kg; 185.3 [9.0] cm; 22.53 [2.1] kg·m⁻²). We found significant differences between deload and overload microcycles in several physiological and perceptual variables (

Table 1. Characteristics of the study population during microcycles.

					95% Confidence Interval
	Microcycle	N	Mean	SD	Inferior	Superior	BF10	p Value	ξ
CK (U/L)	Deload	36	221.6	105.6	185.85	257.31	608.8	<0.001	0.653
	Overload	54	438.9	284.08	361.41	516.48
Urea (mg/dL)	Deload	36	5.48	1.062	5.12	5.84	0.42	0.479	0.094
	Overload	54	5.81	1.407	5.43	6.2
CMJ post-Blood	Deload	24	33.65	6.854	30.75	36.54	1.12	0.117	0.353
	Overload	44	36.19	4.31	34.88	37.5
CMJ Pre-Workout	Deload	36	38.31	5.136	36.57	40.05	0.70	0.005	0.455
	Overload	43	40.8	8.082	38.31	43.29
CMJ Post-Workout	Deload	36	40.76	5.729	38.82	42.7	0.745	0.029	0.407
	Overload	43	43.52	8.621	40.87	46.17
Session-RPE	Deload	36	269.7	268.68	178.81	360.63	1.23 × 105	<0.001	0.833
	Overload	45	732.7	405.81	610.75	854.58
EWMA of Session-RPE	Deload	36	158.3	79.823	131.31	185.32	1.11 × 108	<0.001	0.841
	Overload	48	418.1	195.32	361.39	474.82
Sleep	Deload	36	3.36	1.222	2.95	3.77	1.29	0.053	0.329
	Overload	54	3.85	1.071	3.56	4.14
Stress	Deload	36	2.72	0.944	2.4	3.04	0.25	0.833	0.050
	Overload	54	2.59	1.221	2.26	2.93
Fatigue	Deload	36	2.58	1.251	2.16	3.01	0.70	0.08	0.265
	Overload	54	3.04	1.331	2.67	3.4
Muscle soreness	Deload	36	2.25	0.967	1.92	2.58	58.92	0.003	0.496
	Overload	54	3.15	1.25	2.81	3.49
Mood	Deload	36	2.69	1.064	2.33	3.05	0.43	0.172	0.233
	Overload	54	2.98	1.09	2.68	3.28
Wellness Score	Deload	36	13.61	3.871	12.3	14.92	0.52	0.118	0.231
	Overload	56	15.05	5.358	13.62	16.49

Data is presented as geometric mean (standard deviation) unless otherwise indicated. CK: creatine kinase; CMJ: countermovement jump; EWMA: exponentially weighted moving average; RPE: rating of perceived exertion. The confidence interval of the mean assumes that sample means follow a t-distribution with N−1 degrees of freedom. Sample sizes varied across variables due to missing data. ξ: Cohen’s ξ; Statistically significant (p < 0.05 of the two-tailed p value) for the Yuen–Dixon test would indicate a difference between deload and overload.

). CK levels were markedly higher in the overload microcycle (438.9 U/L) compared to the deload phase (221.6 U/L, BF₁₀ = 608.8, ξ = 0.653). Similarly, both the session-RPE and its EWMA were significantly elevated in the overload microcycle (p < 0.001, ξ > 0.8). Muscle soreness was also significantly higher (BF₁₀ = 58.92, ξ = 0.496), while no differences were found in urea levels or the wellness score (including several of its components).

2.2. Outcome Data

Spearman correlation analysis revealed significant associations between physiological and perceptual markers. CK levels showed moderate positive correlations with session-RPE (ρ = 0.229, p = 0.043), its EWMA (ρ = 0.346, p = 0.002), and muscle soreness (ρ = 0.386, p < 0.001). CK also showed a significant but small correlation with sleep (ρ = 0.25, p < 0.05). CMJ performance (pre/post) was strongly correlated (ρ = 0.831, p < 0.001) but inversely related to urea (post: ρ = 0.242, p = 0.031). Mood correlated most strongly with cumulative wellness score (ρ = 0.817, p < 0.001). A Bayesian correlation matrix for the study variables is shown in Figure 1. The full correlations are presented in Supplementary File S1.

The Bayesian linear regression for fatigue (CMJ performance) revealed EWMA of the session-RPE as the most influential predictor (β = 0.012, 95% HDI [0.004, 0.021]), while CK demonstrated a small negative association (β = −0.009, HDI [−0.016, −0.001]). The muscle damage model showed the microcycle had the strongest effect on CK levels (β = 169.5 U/L, HDI [73.9, 265.2]), with very strong evidence (BF_inclusion = 5.2). Mood was negatively associated (β = −86.4, HDI [−157.0, −15.7]), while pre-workout CMJ showed a protective effect (β = −7.2, HDI [−13.2, −1.1]). The model suggests high-intensity training increases CK by ~170 U/L, while better mood and neuromuscular readiness may mitigate muscle damage.

Additionally, we developed an index integrating biochemical and perceptual markers through PCA (Figure 2A,B).

The varimax-rotated solution (Bartlett’s χ² = 112, p < 0.001) yielded two interpretable components explaining 67.5% cumulative variance (

Table 2. Loadings and components statistics.

Variable	PC1	PC2	Uniqueness	KMO
CK (U/L)	0.823		0.296	0.557
Cumulative Wellness	0.540		0.693	0.558
Mean CMJ Post-Workout		0.809	0.274	0.545
Session-RPE	0.500	0.722	0.228	0.694
EWMA of session-RPE	0.652	0.662	0.135	0.625

CK: creatine kinase; CMJ: countermovement jump; EWMA: exponentially weighted moving average; KMO: Kaiser–Meyer–Olkin measure; PC: principal component.

). Component 1 (34.3% variance) represented physiological stress (CK: λ = 0.823; EWMA of session-RPE: λ = 0.652), while Component 2 (33.1%) reflected acute neuromuscular fatigue (CMJ Post-Workout: λ = 0.809; session-RPE: λ = 0.722).

Due to cross-loadings in RPE variables (session-RPE and its EWMA) and high uniqueness of CMJ (0.809), we refined the index to focus on the physiological stress component as follows:

$$\mathrm{PCA} \mathrm{physiological} \mathrm{stress} \mathrm{index}=0.823\times \mathrm{C}\mathrm{K}\left({\mathrm{U}·\mathrm{L}}^{-1}\right)+0.652 \times \mathrm{E}\mathrm{W}\mathrm{M}\mathrm{A} \mathrm{s}\mathrm{R}\mathrm{P}\mathrm{E}$$

(1)

This parsimonious index prioritizes biomarkers with robust loadings while maintaining sensitivity to training load dynamics. The PCA-derived physiological stress index demonstrated strong discriminative validity between microcycle phases. Mean values were 2.1 times higher during high-intensity microcycles (mean: 515, 95% HDI [442, 587]) versus low-intensity (mean: 250, 95% HDI [218, 283]), with an absolute difference of 265 units (515 [high]–250 [low]). Non-overlapping credible intervals, the extreme Bayes factor (BF₁₀ > 100,000), and low measurement error (<2%) confirm the index’s sensitivity to training load changes at both group and individual levels (Figure 2C,D). In fact, we observed clear divergence in CK and EWMA of session-RPE patterns between microcycle phases that were associated with the PCA-derived physiological stress index (Figure 3).

3. Discussion

The aim of this study was to examine relationships between molecular biomarkers, perceived effort, a wellness cumulative score, and vertical jump performance in elite male basketball players during two microcycles (deload and overload) in an effort to identify sensitive elements of physiological stress. The main findings include that CK and session-RPE (particularly its EWMA) exhibited a significant association with overload. In fact, we were able to identify a PCA-derived physiological stress index including only those two variables that strongly discriminated between microcycle phases. These findings support our recent proposal of including CK and session-RPE within the integrated biomarker panel of AL_index for elite athletes [1].

Fatigue, characterized by declines in physical and cognitive performance, may originate from central or peripheral mechanisms and is intrinsically related to an athlete’s recovery capacity following training load. Numerous studies have established associations between workload and perceived recovery/fatigue, consistently demonstrating that increased training load correlates with elevated fatigue levels [17,24,32,33,34]. Notably, research indicates that during competitive periods, perceptions of both stress and recovery intensify, highlighting the significant influence of psychological factors on these measurements [17,35,36,37]. The Hooper index has been proposed as an indicator of autonomic system readiness, given its positive associations with heart rate variability [38]. This underscores its value for monitoring athletes’ fatigue and recovery states. While some studies report that changes in Hooper index components (particularly muscle damage, fatigue, and sleep) correlate with increased training load [17,39], our analysis revealed very strong evidence (BF₁₀ ≈ 59) only for differences in the muscle damage component.

We identified positive associations between CK and self-rated sleep quality, indicating that higher CK levels corresponded to poorer perceived sleep. Previous studies in team sport athletes, including basketball, have demonstrated that increased training load negatively impacts sleep quality [40,41,42]. Berriel et al. (2020) found an association between CK levels and the recovery category of the RESTQ-sport questionnaire, which includes sleep quality, revealing that reduced CK levels improved scores in this category [17]. Further research is needed to clarify the connections between sleep and various performance metrics (athletic/match performance, training load, injury risk) in team sports, given the inconsistent findings in existing literature [43]. While urea concentration has been proposed previously as a marker of increased nucleotide cycle turnover and protein breakdown (potentially indicating muscle damage following competition or high-load microcycles [44,45]), our study found no significant differences between deload and overload microcycles. Furthermore, its limited explanatory power at the microcycle level and low variance in the PCA suggest urea may not be suitable for training load monitoring in U20 elite basketball players.

In relation to fatigue effects on CMJ performance pre- and post-workout, our Bayesian model comparison revealed that the combination of CK and EWMA of session-RPE best explained CMJ variance (BF₁₀ = 16.26). However, contrasting findings exist. Heishman et al. (2020) observed no correlation between accelerometer-derived training load and jump height in basketball players during precompetitive phases, despite large load fluctuations (925 vs. 3120 UA) [23]. This aligns with the training load reported for elite European basketball players in precompetitive periods (3340 SD 256) [46], suggesting that the short-term effects of training might yield positive outcomes on neuromuscular characteristics associated with jumping. Notably, soccer studies during preseason align with our results, linking elevated session-RPE (and its EWMA) to improved CMJ outcomes [47,48], supporting the training-injury prevention paradox [49]. While CMJ height alone may lack sensitivity to detect workload variations across seasonal phases [32,50], its force-time derivatives (e.g., ground contact time, reactive strength index) show promise for fatigue monitoring [51,52].

Although there are research advances on external load in basketball, there are still unanswered questions about athletes, particularly U20 players, dealing with condensed schedules (e.g., tournaments with multiple games in short periods). Full-game scenarios have the largest kinematic demands [53], and research on professional and young players demonstrates positional and game-phase differences in external load [54,55]. While Reina et al. (2020) examined variations on external load in youth tournaments, their approach did not assess cumulative fatigue across consecutive games [54]. It is important to note the increase in youth tournaments and the elevated injury risks associated with inadequate recovery in this population [56]. Our findings highlight the usefulness of CK and EWMA of session-RPE as discriminators of overload phases, reinforcing them as key indicators for tracking athletes during these demanding schedules. Future studies should integrate this biomarker-driven approach with multidimensional external load metrics (e.g., PlayerLoad^TM, accelerations, etc.) to optimize load management [57].

Finally, the PCA has demonstrated utility across multiple analytical domains, including dimensionality reduction, pattern identification, exploratory categorization, and classification tasks [58]. We previously used hierarchical clustering on principal components to generate profiles of college students based on physical self-concept [59]. Although we did not implement additional machine learning techniques (e.g, clustering), it should be noted that the PCA has enhanced efficacy in reducing data dimensionality while retaining fatigue-associated patterns, as demonstrated by Miaoulis et al. (2025) [60]. The authors emphasized the need for future benchmarking of this type of PCA-derived fatigue metric against established scales like RPE, verifying its ability to capture both subjective and physiological manifestations of fatigue. In this study, we successfully identified for the first time a PCA-derived physiological stress index comprising only CK and the EWMA of session-RPE to discriminate between deload and overload phases. This definitively contributes to the identification of key elements from a set of biomarkers that might be integrated into an AL_index to assess physiological burden in elite team athletes.

Limitations and Future Directions

Although in a pioneering attempt to assess and integrate molecular, neuromuscular, and psychometric biomarkers for an AL_index in elite basketball players during deload and overload microcycles, we are aware that this study is not without limitations. Firstly, there is the small sample size of elite athletes; however, they represented the complete U20 Antioquia team with extensive competitive experience (+5 years), and a robust Bayesian statistical approach was implemented. It should be noted that sub-analyses by player position were not performed due to sample size limitations. Secondly, only CK and urea concentrations were biochemically assessed after blood sampling. Thirdly, the inclusion of perception-related variables in the AL_index requires further validation in other athletic populations, though our results demonstrate their potential utility, at least in U20 elite basketball players. In this study, we employed a modified version of the previously validated Hooper index [61,62], adding mood to the classical components of sleep quality, stress levels, fatigue levels, and muscle soreness. Other perceptual questionnaires, such as the profile of mood states (POMS), the recovery-stress questionnaire for athletes (RESTQ-Sport), and the multicomponent training distress scale (MTDS), could further elucidate the relationship with physiological variables [25,36]. These tools may help unravel the complexity underlying human health and performance through diverse statistical approaches [63].

Finally, we developed a parsimonious yet practical index combining CK and session-RPE that demonstrated exceptional discriminative power between deload and overload phases, providing coaches with a potential monitoring tool for assessing readiness in competitively scheduled basketball players. Notwithstanding, we technically did not validate the AL_index since this implies measuring several biomarkers of cardiovascular, neuroendocrine, metabolic, and inflammatory domains. In fact, the modest KMO of our PCA-based index indicates limited sampling adequacy for the current variable set. Future studies should explore alternative measures for this construct. This also reinforces the need to include additional variables to have a robust AL_index and suggests that our index may not fully capture the multidimensional nature of physiological stress. We recently proposed that the AL_index in athletes can be derived from a set of biomarkers that indicate sub-clinical or clinical deviations, providing a multi-system approach to the adaptation process and enabling the anticipation of stress-related dysregulations [1]. Beyond CK and session-RPE, future research should validate other biomarkers of cardiovascular (e.g., heart rate variability), neuroendocrine (e.g., free testosterone-to-cortisol ratio, S100 calcium-binding protein B), metabolic (e.g., resting blood lactate, energy availability), and inflammatory (e.g., C-reactive protein, hepcidin) domains to improve the construct validity of physiological burden assessment in athletes through an AL_index [1]. This will contribute to the use of a simple tool that helps lower costs, speed up data collection and analysis, reduce evaluation time, and accurately monitor high-performance athletes. Readers are encouraged to use the 4Rs App (https://dbss.shinyapps.io/4RsApp/, accessed on 22 April 2025) to implement this practical tool for understanding the allostasis-interoception paradigm and for monitoring and decision-making through the 4Rs of sports nutrition to optimize training adaptations in professional settings.

4. Methods

4.1. Study Design

A prospective cohort study with an observational correlational design was conducted following the STROBE guidelines [64]. Elite basketball players followed their usual training program over a pre-planned two-microcycle period. During the first week, three assessment training days took place within a deload microcycle, while in the following days, five assessment training days were conducted during an overload microcycle that included official matches. Blood samples were collected for biomarker monitoring, and CMJ tests were performed before and after each training session (eight non-consecutive days over the two microcycles). Figure 4 summarizes the study protocol.

4.2. Setting

Participants were recruited on 27 May 2022 and were informed about the characteristics of the study, and those who agreed to participate voluntarily were included. In total, 12 athletes provided informed consent on 30 May, prior to the first testing session.

To monitor the initial state of biochemical variables, a deload microcycle took place from Tuesday, 30 May, to Thursday, 2 June 2022. This period was characterized by low training intensities (participants refrained from any additional physical activity) and included an overnight fast starting at 8:00 p.m. the day before blood sampling. Following this, an overload microcycle was implemented, featuring increased volume and intensity due to four matches (

Table 3. Training structure across deload and overload microcycles.

Microcycle	Strength Training	On-Court Training	Competition Load
Deload	Exercises: Loaded jumps, squat, clean, and bench press Volume: 12 total sets Intensity: RPE 4–5 (40–50% 1RM) Reps: 4–6 (of 8–12 possible) Rest: 2–3 min	Exercises: Individual, partner, and group shooting Volume: 45–60 min Intensity: Low Rest: 2–5 min	None
Overload			Four high intensity matches during the week

Both microcycles followed identical 3-day strength and 5-day on-court training protocols. However, the overload microcycle (Week 2) included high intensity matches within a blitz tournament.

). Samples were collected from 13 June to 17 June 2022. All training sessions were carried out at the Iván de Bedout Coliseum at the Atanasio Girardot Sports Complex (Medellín, Colombia).

Biochemical samples and questionnaire data were collected during both the deload and overload microcycles. Blood samples were taken at the Indeportes Laboratory (Sports Medicine Department, Medellín, Colombia). Questionnaire responses were entered into a database on Monday, 20 June, and the biochemical test results—received on Thursday, 23 June—were subsequently transferred to an Excel spreadsheet for analysis.

4.3. Participants

The sample was taken by convenience, using a captive and representative population of elite basketball players (the Antioquia team). Most were preselected players of a U20 2022 cycle (https://altoslogros.deportesant.gov.co/torneo/esquema/4/torneo/164786, accessed on 28 October 2024) who participated in the 2023 Colombian National Games in the 5 vs. 5 men’s basketball competition. To be eligible, the inclusion criteria were: age between 17 and 20 years, at least 5 years of sports competition experience, and theoretical and practical knowledge of tools for load monitoring (wellness questionnaire and RPE). All participants were free of any endocrine-metabolic conditions and voluntarily expressed interest in participating after receiving information about the characteristics and aims of the research.

4.4. Variables

Molecular biomarkers included CK (U/L) and urea (mg/dL). Psychometric variables encompassed session-RPE and a wellness score built with sleep, stress, fatigue, muscle soreness, and mood scales. Physical measures correspond to height at CMJ (cm).

4.5. Data Sources/Measurement

The medical examinations were carried out as follows: The participants had to fast from 8 p.m. the day before the blood tests, and they had to arrive as rested as possible (avoid running or performing medium or high effort activities) to the laboratory between 6:30 a.m. and 9 a.m.

4.5.1. Biochemical Analysis

Blood samples were collected under fasting conditions following a 15 min seated rest period. Capillary blood sampling (0.5 mL ≈ 20 drops) was performed using 2.0 mm lancets to ensure consistent blood flow. Prior to sampling, during resting, the hand was immersed in a 42 °C water bath for 15 min to enhance circulation. The ring finger was selected as the puncture site due to its reduced callosity and pain sensitivity. Following puncture, the finger was positioned downward to facilitate gravity-assisted blood flow, and gentle pressure was applied to promote droplet formation. Blood samples were collected in S-Monovette^® capillary tubes containing clot activator gel (Sarstedt, Newton, NC, USA). Total CK and urea serum concentrations were analyzed using a COBAS c-111 automated analyzer (Roche Diagnostics, Indianapolis, IN, USA). Validation of the COBAS c-111 has shown within-run and between-run coefficients of variation ranging from 0.4 to 0.6% and from 0.0 to 2.82% for CK and urea, respectively [65].

4.5.2. Psychometric Variables

Self-reported measures assessing perceptual or psychological states were collected post-exercise [18,61]. Each player accessed Google Forms on their mobile devices and completed the required questionnaires when requested by the researchers. These forms included the RPE scale (from 1 to 10—not measured on the final study day) and a cumulative wellness score based on the Hooper Index [66], which assessed subjective perceptions of sleep quality, stress levels, fatigue levels, muscle soreness, and mood (on a 1 to 7 Likert scale from “very very low-or-good” [point 1] to “very very high-or-bad” [point 7]) [13]. Session-RPE was computed by considering the duration of the session in minutes, according to Foster et al. [15]. Finally, to quantify accumulated training load, we calculated the exponentially weighted moving average of session-RPE (EWMA of session-RPE). This method prioritizes recent sessions while retaining a decaying memory of past workloads, offering a more sensitive measure of acute fatigue than simple rolling averages [67]. The EWMA of session-RPE was calculated as shown in Equation (1):

$$\mathrm{E}\mathrm{W}\mathrm{M}\mathrm{A} \mathrm{o}\mathrm{f} \mathrm{s}\mathrm{e}\mathrm{s}\mathrm{s}\mathrm{i}\mathrm{o}\mathrm{n} \mathrm{R}\mathrm{P}\mathrm{E}=\mathsf{\lambda }\times {\mathrm{s}\mathrm{R}\mathrm{P}\mathrm{E}}_{\mathrm{c}\mathrm{u}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{t}}+\left(\left(1-\mathsf{\lambda }\right)\times {\mathrm{s}\mathrm{R}\mathrm{P}\mathrm{E}}_{\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{v}\mathrm{i}\mathrm{o}\mathrm{u}\mathrm{s}}\right)$$

(2)

where λ = 2/N + 1 represents the weighting decay factor (using a 7-day window in our study), sRPE_current is the current session-RPE, and sRPE_previous is the prior EWMA value. This method follows established evidence demonstrating that EWMA better captures positional workload variations in team sport athletes [68].

4.5.3. Countermovement Jump (CMJ)

CMJ height was assessed using a dual force plate system (Vald Performance, Brisbane, Australia). Measurements were taken after each blood sample collection, as well as before and after every training session. Participants performed five maximal CMJ trials for each assessment, with the mean value used for analysis. Our research group has previously reported a coefficient of variation of 2.2% and an intraclass correlation coefficient of 0.99 (95% CI: 0.96–1.00) for CMJ measurements [69].

4.6. Bias

Urea and CK, key biomarkers of exercise intensity and recovery, were analyzed under standardized conditions. A single physician collected all samples, while an independent researcher recorded perceptual data. All biochemical analyses were performed in the morning in one laboratory to ensure consistency. CMJ testing used a platform with automated data transmission. A dedicated coordinator compiled daily records (sample timing, donor ID, and perceptual responses) and verified complete data reporting to the principal investigator.

4.7. Statistical Analysis

The descriptive statistics are expressed as mean and standard deviation (SD). Based on current recommendations [70] and analytical procedures of the DBSS Research Division [71,72], we used Yuen’s test [73] with 20% trimmed means (μ_t) and 20% winsorized standard deviations (σ_w) as a robust statistical method for small simple sizes. To determine statistical significance in the analysis between groups, we examined the 95% confidence intervals (CIs) for the difference (Δ) between the deload and overload microcycles. If the 95% CI excludes zero, the difference will attain significance at the p < 0.05 level. We used an explanatory measure of effect size (Cohen’s ξ) for robust independent comparisons with 0.15, 0.35, and 0.5 for small, medium, and large effects, respectively [74]. To determine the association between all variables, a Spearman rank correlation was used (rho). For the correlation analysis, we also estimated the likelihood ratio (Bayes Factor [BF]), which is the most widely accepted measure to quantify how much evidence a data set provides for a hypothesis [75]. We standardized all continuous predictor variables using z-score normalization prior to analysis. To assess neuromuscular fatigue and muscle damage, we also performed Bayesian linear regression analyses. The coefficients used a Jeffreys–Zellner–Siow prior with a scale of 0.354 as a robust approach for small effect sizes in physiological studies. Model selection employed a uniform prior to avoid comparison bias, which might be appropriate for exploratory research. We reported BF₁₀ to quantify evidence strength for the alternative hypothesis (H₁: effect exists) versus the null (H₀: no effect), along with highest density intervals (HDI). Finally, we conducted a principal component analysis (PCA) with varimax rotation on all 15 perceptual, physiological, and performance variables (sleep, stress, fatigue, DOMS, mood, cumulative wellness, RPE, time, session RPE, EWMA session RPE, urea, CK, CMJ post-blood sample, CMJ pre-workout, and CMJ post-workout). The analysis was validated through Bartlett’s test of sphericity (confirming sufficient intercorrelations) and the Kaiser–Meyer–Olkin (KMO) measure for sampling adequacy (both overall and for individual variables). Statistical analyses were performed using the ESCI, WRS2, and BMRS packages in the Rj module of Jamovi v2.3.26 [76].

5. Conclusions

This study provides novel insight into perception and physiological response to training load in U20 elite basketball players. Among biomarkers and psychometric variables examined, CK and the EWMA of session-RPE were found to be robust indicators of physiological stress, capable of discriminating between deload and overload microcycles within a PCA-derived index. The significant correlations of these variables with mood, sleep quality, and CMJ performance underscore their relevance in monitoring training and cumulative stress. These findings support the integration of CK and session-RPE for the development of a team sport-specific AL_index to guide individualized training and recovery strategies.