Interestingly, the uncertainty of the ”reference method” e.g. iohexol or iothalamate estimations is never included. The major sources of uncertainty there is the estimation of the distribution volume. For this usually another algorithm is applied with unknown uncertainty.

Further, the factors and exponents applied to the eGFR algorithms, whether based on Cystatin C or creatinine, are not usually considered.

Finally, the algorithms are evaluated using population data and comparing epidemiological parameters. However, take a look at the scatter plots being the basis for the regression analysis and you will find a most remarkable spread around the regression line. That is where the patients we treat will be found. Algorithms may be useful to treat populations but that is rarely our task.

Most likely, this is what David James hints at. It is never wise – or even acceptable – to move from information regarding the central tendency of a data set to individual values, whereas the opposite is the recognized, legitimate mode of operation. Consequently, the appraisal of the uncertainty in using the algorithms is most likely underestimated. However, nephrologists tend to accept an A-zone around a regression of +/- 30 %!?

 

Anders Kallner

 

 

Anders Kallner, MD, PhD

Assoc Professor (R)

Dept Clin Chem

Karolinska University Hospital

171 76 Stockholm

Phone +46 851774943