Early Warning of Suspected Doping from Biological Passport Based on Multivariate Trends

António Júlio Nunes, Paulo Paixão, Jorge Proença, Ricardo J.N.Bettencourt da Silva

Research output: Contribution to journalArticlepeer-review

2 Citations (Scopus)


The indirect identification of doping in sports can be performed by assessing athletes' hematological perturbations from the analysis of blood collected on different occasions. Because prosecution for doping based on this information requires expensive and time-consuming interpretation of blood analysis results by various expert hematologists, mathematical data screening is performed to decide which cases should be forwarded to hematologists. The current Bayesian and univariate screening of data does not process the multivariate trends of blood parameters or take the time interval between samplings into account. This work presents a computational tool that overcomes these limitations by calculating a single score, the hematological perturbation index (HPIx), for which a threshold is defined above which hematologists should be asked to assess the athlete's biological passport. The doping detection from this index, normalized for days difference between samplings based on 3, 4 or 5 consecutive samplings, is associated with true positive result rates (TP) not below 98% and false positive result rates (FP) less than 0.9%. Therefore, this tool can be useful as an early warning system of hematological perturbations to decide which athletes should be more closely monitored and which biological passports should be forwarded to hematologists for medical interpretation of data.

Original languageEnglish
Pages (from-to)44-53
Number of pages10
JournalInternational journal of sports medicine
Issue number1
Publication statusPublished - Jan 2020


  • :biological passport
  • early warning system
  • multivariate analysis
  • hematological perturbations
  • identification uncertainty


Dive into the research topics of 'Early Warning of Suspected Doping from Biological Passport Based on Multivariate Trends'. Together they form a unique fingerprint.

Cite this