The effect of data reduction technic on ligand assay proficiency survey results

The author studied the effects of different data reduction technics on proficiency survey results for radioligand assays. CAP Basic Ligand Assay Survey participants were sent raw count data on a serum folate assay and asked to compute the results for five "unknowns" using their usual method. Analysis of 860 responses showed at least fifteen different data processing methods were used, although 61% of these were some form of a logit transformation. Respondents were divided almost equally between users of computer and manual technics. The coefficient of variation (CV) of the data reduction method at a folate level of 4.5 ng/ml varied from 1% for a spline function (8 participants) to 13% for a manual reciprocal plot (9 participants). Computed weighted logit-log methods (313 participants) gave a 10% CV. Analysis of results for a survey sample by kit and data reduction method showed that the standard deviation (SD) of the calculation was a major contributor to the total interlaboratory SD.