Assessing the Prevalence of Doping Among Elite Athletes: An Analysis of Results Generated by the Single Sample Count Method Versus the Unrelated Question Method

In 2011, a group of researchers investigated the 12-month prevalence of doping at the 13th International Association of Athletics Federations World Championships in Athletics (WCA) in Daegu, South Korea, and also at the 12th Pan-Arab Games (PAG) in Doha, Qatar. The prevalence of doping at each event was estimated using an established randomized response method, the Unrelated Question Model (UQM). The study, published in 2018, found that the prevalence of past-year doping was at least 30% at WCA and 45% at PAG. At both events, separate data sets were collected in addition to the UQM data using a new method, the single sample count (SSC). Recently, Petroczi et al. have reported 12-month doping prevalence estimates for these two events based on the SSC data. These investigators obtained substantially lower prevalence estimates using the SSC and suggested that the 2018 estimates based on the UQM may have been too high. However, in this communication, we point out several possible shortcomings in the methods of Petroczi et al. and show that their SSC data would be equally compatible with a high 12-month doping prevalence comparable to the UQM estimates published in 2018. A prior study of the prevalence of past-year doping and dietary supplement use among elite athletes, conducted at two international sporting events, used two randomized response techniques-the unrelated question method (UQM) and the single sample count (SSC)-to ensure the anonymity of the participants and thus encourage honest responses.The UQM analysis, published in 2018, found the prevalence of past-year doping to be at least 30% and 45% at the two events, respectively, and the prevalence of past-year dietary supplement use at the second event about 70%-a figure congruent with other studies of supplement use among elite athletes. However, the SSC analysis, published in 2022, yielded rates of only 21.2% and 10.6% for doping and only 8.6% for dietary supplements.In a reanalysis of the SSC data, using a slightly different model that considers both cheating and underreporting due to cognitive limitations, we show that the SSC data can yield much higher prevalence estimates, very similar to those of the UQM.