Biological variation of inflammatory and iron metabolism markers in high-endurance recreational athletes; are these markers useful for athlete monitoring?

Study population

The study population has been described in detail previously [12, 13, 21]. Briefly, 30 subjects (15 women), median age 36 years (range 19–53) were recruited from three triathlon clubs in Madrid, Spain and fulfilled the following inclusion criteria: presumably healthy, age>18 years, >13 h of training per week, normal standard medical examination and expected results for a stress test supervised by a physician specialist in sports medicine [12]. The study was approved by the host Institutional Research Ethical Committee and all participants signed informed consent.

Sample collections

To assure a confidence interval (CI) width lower than 1/3 CV_I [22], 11 samples per subject were collected monthly (January–November 2017) on Saturday mornings (8–10 am) after 10 h fasting. In each visit, participants completed a questionnaire about fasting, vitamin supplementation and diet changes, health status and use of medication, date of last competition, details on last training (time, type and intensity) and in females, date of last menstrual period.

All aspects of the preanalytical phase were standardized, as previously reported in detail [13] and are detailed in the Supplementary File.

Sample analysis

All serum samples were analyzed in duplicate for hepcidin, sEPO and sTfR (January–August 2022) and IL-6 (January 2022–February 2023). We used the following methodology: LC-MS/MS for hepcidin [23] (limit of detection (LoD) 0.3 nmol/L); Quantikine ELISA immunoassay for IL-6 (R&D Systems Inc, Minneapolis, MN, USA) (LoD=0.7 ng/L), chemiluminescence immunoassay on Atellica analyzer (Siemens Healthineers^®) for sTfR (LoD=0.5 mg/L) and chemiluminescent immunometric assay on Immulite 2000 XPi analyzer for sEPO (LoD=1.5 IU/L). All samples from each subject were measured in the same analytical run and with the same reagent lot. All quality control results were within set limits.

Statistical analysis

Differences in biomarker mean concentrations related to sex (male/female) or the physical activity performed the previous 24 h (Yes/No) were assessed by t-tests, after a Levene’s test to evaluate the data distribution.

Prior to calculating the BV components, results below the LoD were excluded. Trend analysis was performed for the overall population for all measurements. If the 95 % CI of the slope did not include 0, the trend was considered significant. In this case, trend analysis was performed for each individual by plotting each subject’s results against the collection dates. Subjects with a significant trend during the study were excluded and trend analysis for the overall population was repeated. The resulting data set was considered to represent a steady-state situation and two different approaches were applied to derive BV estimates:

Influence of sex and training on measurand concentrations

Training during the previous 24 h did not influence the mean concentration of any of the measurands (data not shown). However, the mean concentrations were significantly different (p<0.005) between men and women, with women having higher results for sTfR and sEPO and lower for hepcidin (Supplementary Figure 2). Mean concentrations in men and women were 0.73 and 1.01 mg/L for sTfR, 7.7 and 11.7 IU/L for sEPO, and 2.7 and 1.0 nmol/L for hepcidin, respectively (Supplementary Figures 1 and 2).

Biological variation estimation

No trends were identified in the overall study population except for sEPO (slope 0.004, CI 95 % 0.007–0.026). Further study indicated a significant trend for sEPO in two subjects. After excluding these two, there was no longer a trend in sEPO and the remaining subjects were considered to be in a steady state.

For hepcidin and sTfR, 16.5 and 17.1 % of results were under the LoD, respectively. Out of the results over the LoD, 18.6 and 13.5 % of the data were identified as outliers for hepcidin and sTfR, respectively (Table 1). Only two results were identified as outliers for sEPO. The BV estimates derived by the classical approach, with mean concentration, included results and the number of results under the LoD and outliers are shown in Table 2. No significant differences between men and women were identified for BV estimates derived by the classical approach. Still, the distribution of the CVp_(I) derived by the Bayesian model was different between men and women for hepcidin and sTfR (Table 2, Figures 2 and 3). No relation between the CVp_(I) and concentrations for the different analytes was observed by visual inspection (Supplementary Figure 3).

Figure 2:

CVp_(I) of the athletes for hepcidin, sTfR, soluble transferrin receptor; EPO, erythropoietin. Circles indicate males and triangles females.

Figure 3:

Distribution of the CVp_(I) of the athletes for hepcidin, sTfR, soluble transferrin receptor; EPO, erythropoietin. Dashed lines represent -females and continuous lines -males.

For hepcidin, sEPO and sTfR, there was an overlap between the 95 % CI/Crl for the BV estimates obtained by Bayesian and ANOVA methods (Tables 1 and 2). Including exercise and health-related factors in the LMM did not change the CV_I for hepcidin, sEPO and sTfR when including sampling interval, perceived exertion, intensity or duration of training in the week before sampling (data not shown).

Distribution of BV data

When applying the classical CV-ANOVA approach, the strategy is to analyze and eliminate outliers to fulfil ANOVA requirements, including normality and homogeneity between replicates and between subjects. This data management “cleans” the data, which runs a risk of results being biased, especially when the number of excluded results is high, as was the case for some of the measurands included in our study. Conversely, the Bayesian method is robust to extreme values, and outlier and homogeneity analyses are not required before analysis. When interpreting the Bayesian results, the HBR is a tool to evaluate the distribution of the individual CVp_(I) (Figures 2 and 3). If the HRB is above a set cut-off dependent on the study population/design ( 100 / 2 · n o .   s a m p l e ) , the CVp_(I) distribution can be considered heterogeneous. In such a case, the derived average, or median, BV estimate is unlikely to represent the studied population.

For hepcidin, an HBR below 24.9 would in our study indicate a homogenous distribution. However, the HBR was 31.9 for all subjects, and for the female subgroup even higher (HBR=58.5; cut-off <26.7) than for men (HBR=30.7; cut-off <23.4) (Figure 4). Thus, a common CVp_(I) based on a central value (mean) would not represent all the included subjects for hepcidin. This also fits very well with the results of the CV-ANOVA analysis where 18.6 % of the results were excluded as outliers (Table 1), indicating that to obtain data homogeneity, required for CV-ANOVA analysis, was difficult. For sTfR, the HBR was also higher (43.1 and 34.4 for women and men, respectively) than the respective cut-offs (Table 2), indicating non-homogenous distribution for both subgroups. For sEPO, on the other hand, the CVp_(I) distribution was homogeneous for both subgroups. The HBR was, nevertheless, higher in women than in men (18 and 12.5, respectively), indicating a wider distribution, as illustrated in Figure 3. The heterogeneity observed for the individual CVp_(I) indicates that it may not be appropriate to use a mean CVp_(I) for hepcidin and sTfR for the different BV applications such as APS or RCV. Instead, using a higher or lower percentile of the CVp_(I) may be more appropriate, depending on the clinical situation.

Soluble transferrin receptor

sTfR is especially useful in the work-up of anemia in athletes to differentiate between iron deficiency and anemia of chronic disease. Inflammatory processes linked to exercise have previously been shown to inhibit iron absorption [18]. The BV estimates for sTfR were significantly higher in athletes (10.3 % (95 % CI; 8.8–12.0)) than those published for a healthy, general population (6.95 % (6.5–7.3)) [30]. These results could be explained by a higher iron demand in endurance athletes [18, 31]. However, these two studies are not directly comparable, as they had different study designs; weekly samplings over 10 weeks [30] and monthly samplings over one year in our study. Therefore, further study may be necessary to clarify differences between general and athlete populations for sTfr. Furthermore, previous studies in healthy individuals applied classical methods and there needs to be knowledge on whether the heterogeneous BV data distribution of sTfR is specific for athlete populations or also a feature also in the general population.

Hepcidin

There is generally little published BV data for hepcidin, and only one previous BV study, based on within-day samplings, performed in a healthy population could be identified [32]. The CV_I of 30.3 % reported by Krookt et al. in this study [32], derived by ANOVA, was lower than the estimate observed in our study (51.9 %). However, the different sampling intervals make direct comparisons difficult. Two publications have reported BV of hepcidin in hemodialysis patients [26, 33]. In these studies the study duration was two and six weeks, respectively [26, 33] and the CV_I estimate reported were much lower (23 % [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28]) than that obtained for the athletes in this study. Previously, we have assessed BV in this cohort for other iron markers by classical methods (ANOVA). However, for serum ferritin, data were so heterogeneously distributed that we did not estimate the BV [13]. When applying a Bayesian model on serum ferritin, the BV in women for ferritin was very heterogeneously distributed, with a HBR of 51.8, greater than that for men (37.1) (Supplementary Figure 4). Similarly, we also found a high predicted CVp_(I) (63.1 % for females and 49.2 % for males) for ferritin suggesting that both markers may not be under homeostatic control or at least highly influenced by inflammatory mechanisms in athletes [15]. Hepcidin, in combination with other iron markers (iron, transferrin, ferritin and erythroferrone) [34], has been suggested used to detect autologous transfusions – used for doping purposes because its concentration clearly peaks up one day after blood transfusion [35]. However, as demonstrated by our study, hepcidin results show great variability. Therefore, serial results for these measurands (ferritin and hepcidin) would be difficult to interpret in endurance athletes.

Erythropoietin

No study reporting BV data for EPO has been published in the general population or athletes. Thus, this is the first study that reports BV estimates for sEPO. EPO is well known to stimulate the production of red blood cells and due to its performance-enhancing effect, the use of recombinant human EPO (rHuEPO) is prohibited by the World Anti-Doping Agency (WADA). Direct and indirect methods have been developed to detect EPO misuse, especially when rHuEPO microdoses are administered [36]. Among the available methods are different methodologies that attempt to measure EPO directly. However, to detect rHuEPO microdoses administration, one of the suggested methods is the Athlete Biological Passport (ABP). The ABP is a tool used by WADA to monitor selected biomarkers (hemoglobin and reticulocytes among others) over time that flags variations not compatible with a normal physiological condition, bringing out a possible drug misuse. Clark et al. reported changes in performance, physiology, and blood markers included in the hematology ABP module, especially after high dosing followed by microdosed of rHuEPO [37]. However, the ABP utility for this purpose still needs to be demonstrated. Previous publications have reported that sEPO is affected by confounding factors [38]. However, our results indicate that the CV_I of sEPO was apparently not affected by physical activity, duration or intensity. Furthermore, the CV_P(i) estimates were homogeneously distributed and the II relatively low (0.7), which suggest that sEPO could be a potential candidate to be included in the ABP strategy to detect rHuEPO misuse. The immunoassay used in this study (Immulite 2000 XPi analyzer) showed an acceptable performance (CV_A=10.2 %), fulfilling the APS for imprecision of 0.5 × CV_I of about 15 %. Measured by chemiluminescence, circulating EPO levels determination is relatively inexpensive compared with other methods currently in use to detect EPO in urine (isoelectric focusing or sodium dodecylsulfate polyacrylamide gel electrophoresis) [39], and the assay procedure is fully automated. Thus, this marker could be a candidate to add to WADA strategies for detecting rHuEPO misuse and may be considered included in the ABP module as a complementary method to urine EPO detection for increasing the sensitivity. However, further study to evaluate its performance for this purpose must be performed prior to implementation.

Interleukin-6

BV data for I-L6 could not be estimated, as the assay used to measure IL-6 was not sensitive enough to quantify the low concentrations of the athletes included in our study, indicating that any inflammatory process related to exercise is not pronounced by the increase of this measurand. We, however, observed that one of the subjects had detectable values in all the collected samples (range 14.8–37.3 ng/L). Further investigation revealed that this subject had been diagnosed with a mild psoriasis, not mentioned in the recruitment questionnaire, which could likely explain the results. Results for this participant for sTfR, hepcidin and EPO were within the reference intervals.

Study limitations

For the Bayesian model, no prior information on BV components could be identified for EPO in the literature. Therefore, we used the results from the CV-ANOVA and ANOVA analysis derived from the same cohort to set the hyperparameters, however, giving them low credibility. The inclusion of a control group into our study, consisting of non-athletes using the same sampling period, would have been useful to better assess differences in BV in iron and inflammatory markers between athletes and the general population.