Background: The utility of insulinlike growth factor (IGF) axis and collagen markers for a growth hormone (GH) doping test in sport depends on their stability and reproducibility. We sought to determine short-term within-subject variability of these markers in a large cohort of healthy individuals.

Methods: We measured IGF-I, IGF binding protein 3 (IGFBP-3), acid labile subunit (ALS), and the collagen markers N-terminal propeptide of type I procollagen (PINP), C-terminal telopeptide of type I collagen (ICTP), and N-terminal propeptide of type III procollagen (PIIINP) in serum samples obtained on multiple occasions (median 3 per participant) over a 2- to 3-week period from 1103 elite athletes (699 men, 404 women) ages 22.2 (5.2) years [mean (SD)]. We estimated between-subject and within-subject variances by mixed–effects ANOVA.

Results: Within-subject variance accounted for 32% to 36% and 4% to 13% of the total variance in IGF markers and collagen markers, respectively. The within-subject CV ranged from 11% to 21% for the IGF axis markers and from 13% to 15% for the collagen markers. The index of individuality for the IGF axis markers was 0.66–0.76, and for the collagen markers, 0.26–0.45. For each marker, individuals with initial extreme measured values tended to regress toward the population mean in subsequent repeated measurements. We developed a Bayesian model to estimate the long-term probable value for each marker.

Conclusions: These results indicate that in healthy individuals the within-subject variability was greater for IGF-I than for the collagen markers, and that where a single measurement is available, it is possible to estimate the long-term probable value of each of the markers by applying the Bayesian approach. Such an application can increase the reliability and decrease the cost of detecting GH doping.