Fenêtre de détection et critères d’identification suivant un dopage à la DYNEPO

Code: T08B01MS

The main objectives of the project were: - Testing the sensitivity of the classical WADA positivity criteria when applied to Dynepo™-enriched samples

- Computing a new decisional tool able to discriminate between negative urine profiles, Dynepo™-enriched urine profiles and effort urine profiles.

- Determining the detection window of Dynepo™ following multiple injections on healthy subjects

Main findings

The main outcome of this project was to demonstrate factually that the current WADA criteria, as defined in the WADA2009TD, were not applicable to Dynepo™ detection in urine. Indeed, a formal application of these criteria on the IEF patterns resulting from 126 Dynepo™-enriched samples demonstrated a sensitivity of 9% on 3 weeks for multiple injections of a total of 22’500 IU of Dynepo™. The 3rd criterion, defining the acceptable ratio between the second most intense basic band and the most intense endogenous band, was mainly responsible for this poor sensitivity. We therefore proposed a linear multivariate model (PLS-DA) computed for discriminant analysis on the basis of 196 detectable urine patterns, half of them resulting from Dynepo™-enriched samples. Following cross-validation, this model, based on 3 latent variables (LV), yielded a score characteristic of each individual IEF pattern. This score indicated how representative a sample was of the positive or the negative class. Bootstrap resampling allowed the definition of a cut-off score and consequently, the identification of atypical samples. Applying this new criterion resulted in a sensitivity of 52% on the same 126 samples, without any loss of specificity. This model eventually evaluated Dynepo™ detection window as close to 48 hours, which is in par with the short half-life of the molecule in the organism, when compared to those of epoetins-α and -β

The main interest of this open model is that it is potentially refined each time a new data is computed. In addition, it could be easily generalized to other epoetins, notably alpha and beta. Pattern classification methods have been previously developed for classical epoetins, but the interpretation of Dynepo™ profiles has never been considered. Considering the fact that the current WADA criteria are manifestly not applicable to Dynepo™ detection, our model has returned a good sensitivity versus specificity ratio. It remains however very dependent on the analytical protocol, as any departure from the described procedure would require a specific validation. Altogether, this suggests that the use of the proposed model could be included as an additional piece of evidence in the procedure of EPO doping detection.