**4. Preliminary statistical considerations**

The use of inulin as GFR marker is justified by physiologic studies. The others markers that will be proposed thereafter will be compared to inulin measurements. Therefore, the use of other markers will be justified not by physiological studies (even if some physiological studies exist for some markers) but by studies comparing these markers with inulin. Unhopefully, most of these studies comparing different GFR tests lack of strong statistical methodology. Actually, most of the authors have only shown a good correlation between the markers, which is expected but not sufficient. Ratio of new markers results on inulin results are also used (the result being considered as good if ratio is near to 1). The use of such ratio may be misleading (for example, if one method overestimates true GFR in low GFR levels but underestimates GFR in high levels, the ratio will be near to 1 although the method is actually not precise enough). To compare the performance of a new GFR measurement compared to inulin, we need to know the bias (mean difference between the two results) and the precision (standard deviation (SD) around the bias) of this new measurement. Bland and Altman analysis is thus required (Bland & Altman, 1986).

Regarding the other GFR markers, we must also stress that GFR can be measured by plasma clearance and using a bolus injection (instead of constant infusion rate) which makes the GFR measurement much more simple. Method to measure GFR by plasma clearances can be very different (number of samples, timing of samples, mathematical model used). We must keep in mind that results of plasma and urinary clearances are not strictly comparable (plasma clearances overestimate urinary clearances even if the overestimation decreases if plasma samples are drawn after 24 hours) and this must be integrated when these GFR methods are compared (Agarwal et al., 2009; Stolz et al., 2010).
