Literature review > Issue_2 > Review Land et al. 

The licensing of a new diagnostic test requires rigorous comparison of defined samples against the best existing techniques, in highly skilled laboratories. Once a new diagnostic test is licensed, the measurement of its performance in the real world of diverse routine diagnostic laboratories becomes of greater importance than its performance in selected or centralised laboratories. The most important information is that of the ability of the new assay to be robust, to retain sensitivity and specificity in the face of widely different skilled or experienced technicians. Clearly, this ability is best tested through regular quality assurance panels, but the nature of the panel samples is important, in that the closer the panel represents "the real world", the more valid the interpretation of the results to clinical practice.

Sally Land and colleagues have studied the performance of 57 Australian routine diagnostic laboratories against two panels of common samples. One panel consisted of normal human urine spiked with Chlamydial antigen derived from cell culture, and the second clinical urine samples. The choice of spiked urine for the first panel guarantees the comparability of the results against that expected from clinical samples, and the precise quantification of the spike allows a high degree of confidence in measures of sensitivity. Clinical urine samples suffer from problems of available volume, and unreproducibility between panels.

For the first panel (the C trachomatis spiked normal human urine, at neat, 10x and 100x dilution), 40 of 51 participating laboratories used the Roche R-PCR assay, with detection by manual EIA or by the automated COBAS system. Eleven laboratories used the Abbott ligase A-LCR assay. Samples were tested in duplicate, except for the lowest concentration (x1) sample, which was tested in quadruplicate. For all samples, the R-PCR assay was more sensitive than the A-LCR; at the highest dilution, 9% were detected by R-PCR versus none by A-LCR (p<0.05). Interlaboratory variation was absent at the lowest dilution (100x endpoint), and all laboratories correctly identified the negative sample.

For the second panel, 47/57 used R-PCR and of these, 33/47 used the automated COBAS system and 14/47 manual EIA. The method of detection did not affect the sensitivity of the results. The panel was derived from single clinical samples, one of which was spiked with humic acid to simulate the presence of inhibitors of PCR. Duplicate testing by R-PCR of samples with low levels of C. trachomatis revealed discordant results in 16/47 laboratories. The inhibitor-spiked sample was correctly identified only in those laboratories using the internal control; 16 laboratories using R-PCR C trachomatis without the internal control and all the LCR laboratories reported this sample incorrectly negative.

This well-conducted study is of practical use, even though Abbott has recently withdrawn the A-LCR assay. The R-PCR assay is sensitive, but duplicate clinical urine samples containing low levels of antigen have a high likelihood of discordant results; single testing of urine samples from patients with low organism load will therefore have a high likelihood of false negative results. This is particularly important as many positive urine screening samples are likely to have low organism load; duplicate sample testing would greatly improve detection rates by R-PCR in these clinical samples. The urine spiked with inhibitor also showed the value of routinely using the internal amplification control in the R-PCR assay; in those laboratories not using the internal control, these samples were false negatives. These results should inform the regular use of duplicate testing and the internal amplification control for the R-PCR assay.

about SDI | newsletters | grants | publications | literature reviews

WHO Home - WHO Search - TDR Home - SDI Home - SDI Contact us
(c) WHO/OMS 2001