Reference values for serum S-100B protein depend on the race of individuals.

Fig. 1B shows the opposite, with a mean bias of ϩ14.5% and a scatter of ϩ10 to ϩ 100% higher values in seven samples having a cTnI value of ϳ1.0 ␮g/L, the clinically most important concentrations. Perhaps the authors inverted the factors (plasma Ϫ serum values in the text and serum Ϫ plasma values in the figure), but they did so without any reason, and this fact markedly clouds the data presentation. Finally, the same scale should be used for the x axis in all of the graphs to permit a direct comparison of the results obtained with the two assay generations. More importantly, the authors make improper use of the recommendations reported in the quoted letter by Dewitte et al. (3) and in the corresponding response by Altman and Bland (4). The reference to the method comparison issue is questionable , the anticoagulant problem being indeed an interference (and preanalytical) issue (5). Nevertheless , the original suggestion by the previous authors was to take logarithms of the original data or, alternatively , the percentage bias, both of which give similar information (4). Accordingly, reporting the data as the percentage difference on the y axis and serum cTnI concentrations on the x axis is the correct and more transparent approach to evaluate possible anticoagulant effects on cTnI assays and the variability among samples (6). Another limitation of the study is that at least two samples used in the second experiments had cTnI concentrations Ͻ0.02 ␮g/L, the detection limit of the assay (7). The use of samples having undetectable cTnI concentrations makes calculation of the sample difference impossible. This could have significantly biased the results obtained with the second-generation assay. Finally, Dorizzi et al. (2) do not provide any information on the time of sampling after hospital admission (or onset of chest pain) of the patients studied, the assay cutoff value(s), or the analytical imprecision at the cutoff value(s) (8). Time-dependent losses of troponin immunoreactivity have clearly been shown in heparin plasma by previous authors (9). In conclusion, the study by Dorizzi et al. (2) shows that the data obtained with the second-generation Dimension assay are significantly different from those obtained previously with the first-generation assay; the scatter was very different (although they do not indicate standard errors of estimates), and the statistical test (and the 95% confidence interval of the mean bias) evaluating the serum/ plasma difference reached significance for …