Bad science: the instrumental data in the Floyd Landis case

In 2006 Floyd Landis won the world's most prestigious bicycle race, the Tour de France. However, not many days after the race's conclusion it was released to the press that the Laboratoire National de Dépistage du Dopage (LNDD) had found Landis' urine after stage 17 positive for synthetic testosterone.

This review examines the instrumental data and methodology used by LNDD in the Landis case. The conclusions reached by LNDD were based on results of 2 separate instrumental methods. Subsequent to urine extraction and possibly derivatization, samples were initially screened via gas chromatography/mass spectrometry (GC/MS) using selected ion monitoring (SIM) to measure the ratio of testosterone to epitestosterone (T/E). Final confirmation of exogenous testosterone was determined by measuring the (13)C/(12)C stable isotope ratios in 4 metabolites of testosterone via gas chromatography combustion stable isotope ratio mass spectrometry (GC-C-IRMS).

T/E ratios determined by LNDD in Landis' stage 17 urine were unreliable due to the combined factors of an unsatisfactory extraction, high GC background, failure to obtain baseline peak separation for epitestosterone, unreliable quantization of the epitestosterone peak due to both peak overlap and because it was barely above background noise, and because only a single ion mass (432) rather than a minimum of 3 was used for SIM (in violation of both LNDD's SOP and WADA procedures). GC-C-IRMS methodology is less well known to the analytical chemistry community, but here too the results obtained by LNDD were unreliable. GC-C-IRMS errors can be briefly summarized as uncertain peak identification, unsuitable standards, and unreliable (and possibly biased) calculation of (13)C/(12)C ratios due to peak overlap as well as LNDD's usage of manual peak integration rather than use of the instrument system software.