ABPS {ABPS}

A function for calculating the Abnormal Blood Profile Score

Description

The ABPS function computes the Abnormal Blood Profile Score from seven haematological markers. Higher values of this composite score are associated with a higher likelihood of blood doping.

Usage

ABPS(haemdata = NULL, HCT = NULL, HGB = NULL, MCH = NULL, MCHC = NULL,
  MCV = NULL, RBC = NULL, RETP = NULL)

Arguments

haemdata

a vector or data frame containing (at least) the 7 haematological variables, either with the same names as the parameters below, or (not recommended) without names but in the same order as the parameters.

HCT

haematocrit level [%]

HGB

haemoglobin level [g/dL]

MCH

mean corpuscular haemoglobin [pg]

MCHC

mean corpuscular haemoglobin concentration [g/dL]

MCV

mean corpuscular volume [fL]

RBC

red blood cell count [10^6/uL]

RETP

reticulocytes percent [%]

Details

The ABPS uses the seven haematological variables (HCT, HGB, MCH, MCHC, MCV, RBC, RETP) in order to obtain a combined score. This score is more sensitive to doping than the individual markers, and allows the detection of several types of blood doping using a single score.

The combined score is based on two classification techniques, a naive Bayesian classifier and an SVM (Support Vector Machine). The two models were trained using a database of 591 blood profiles (including 402 control samples from clean athletes and 189 samples of athletes who abused of an illegal substance); the two scores were then combined using ensemble averaging to obtain the final ABPS score.

The ABPS is part of the Athlete Biological Passport program managed by the World Anti-Doping Agency. While it is not a primary marker of doping, it has been used as corroborative evidence (see e.g. https://jurisprudence.tas-cas.org/Shared%20Documents/2773.pdf)

Value

a vector containing the ABPS score(s). Scores between 0 and 1 indicate a possible suspicion of doping; a score above 1 should only be found in 1 in 1000 male athletes.

Note

The values for the markers can be specified using either a data frame containing (at least) the 7 haematological variables, or using seven named parameters, but not both at the same time.

The calculation of the ABPS depends on two sets of parameters, for the two machine learning techniques (naive Bayesian classifier and Support Vector Machine), which are provided in the package.

Each parameter must be in a prespecified range; parameters outside this range are constrained to the min (respectively max) values.

Note that several versions of the ABPS were developed (including several different combinations of parameters). The version provided in this package provides the same results as the WADA version included in their ADAMS database. However, some values calculated with other versions of the software have also been distributed (see the help page for the blooddoping dataset for an example).

References

Sottas, P.E., N. Robinson, S. Giraud, et al., Statistical classification of abnormal blood profiles in athletes. Int J Biostat, 2006. 2(1): p. 1557-4679.

https://jurisprudence.tas-cas.org/Shared%20Documents/2773.pdf

Examples

ABPS(HCT=43.2, HGB=14.6, MCH=31.1, MCHC=33.8, MCV=92.1, RBC=4.69, RETP=0.48)
ABPS(data.frame(HCT=43.2, HGB=14.6, MCH=31.1, MCHC=33.8, MCV=92.1, RBC=4.69, RETP=0.48))
ABPS(c(43.2, 14.6, 31.1, 33.8, 92.1, 4.69, 0.48))
data(blooddoping); ABPS(blooddoping)
data(bloodcontrol); ABPS(bloodcontrol)