Population quantile‐quantile plots for monitoring assay performance in newborn screening

Summary Immunoreactive trypsinogen (IRT) used in screening for cystic fibrosis is heterogeneous, poorly characterized, and displays marked matrix effects when incorporated into blood spots, making long‐term control of assay calibration difficult. The cut‐off required to select a fixed proportion of samples (0.5% for the UK protocol) for second‐tier mutation analysis varies over time, partly owing to slight differences in calibration of individual lots of assay kit. To investigate this and possible inter‐laboratory differences in analytical performance, we developed a monitoring system based on the distribution of measured IRT concentrations in the screened samples. Results were collected fortnightly from five UK screening laboratories in the form of numbers of samples in histogram ‘bins’ of IRT concentration. These data were converted to cumulative percentages and the IRT concentrations at fixed centiles then approximated by triangulation. The quantile‐quantile plot of any subset (by laboratory or by kit lot) of these data using pooled results (all‐laboratories all‐kit‐lots, approximately 270000 samples) as the reference distribution is analogous to a calibration curve and gives a measure of bias in terms of sensitivity (slope) and baseline ( y ‐intercept). This allows a revised 99.5th centile cut‐off for each subset to be calculated directly. A similar approach has allowed inter‐laboratory comparison of tandem‐mass spectrometric assays for free carnitine (with emphasis on low values) and phenylalanine and has demonstrated that apparently trivial differences in instrumentation and procedures have resulted in marked variation in resultant assay performance.