Effectiveness of quality control methods for glomerular filtration rate calculation.

Abstract

In this work, we aimed to identify the types of errors encountered in glomerular filtration rate (GFR) measurement and test the effectiveness of all published quality control (QC) methods for detection of clinically significant errors. A total of 412 GFR tests were carried out on adults and children. The three-point slope-intercept glomerular filtration rate (SI-GFR) was compared with the nine-point 'area under curve' calculation as a gold standard to determine the error in SI-GFR. The Durbin-Watson test was used to characterize the nature of the errors. The sensitivity, specificity and positive predictive value (PPV) of QC methods for detecting clinically significant errors were calculated and receiver operating characteristic curves were constructed. The QC methods were also applied to a dataset of 100 four-point GFR tests from different institutions. Model failure is the dominant cause of clinically significant error in this dataset, with individual point measurement errors only giving rise to clinically significant errors in a small number of cases. No QC test had an acceptable combination of sensitivity, PPV and specificity. The correlation coefficient QC test had the largest area under the receiver operating characteristic curve (0.73). No other QC test had an area greater than 0.57. All the QC methods have poor sensitivity and PPV for detecting clinically significant errors and so cannot be relied on to ensure a robust measurement of GFR, underlining the need for careful working practices and a thorough system of measurement checks. We found no evidence for the value of multiple sampling with respect to QC; until such evidence is published, their clinical utility is unproven.