To the Editors:
We read with interest the article by Rosias et al. 1 in a recent issue of the European Respiratory Journal. The authors attempted to validate a number of assays using exhaled breath condensate. The values described by Rosias et al. 1 were very similar to those reported by many laboratories, but our interpretation of the data is different. Our experience in validating similar assays in this medium suggests that measuring mediators at these low concentrations is problematic and, as such, we have the following concerns with the article which other researchers need to be aware of.
First, Rosias et al. 1 based coefficient of variation (CV) on a mean of spiked samples, where the spike was far in excess of the measured values. The CVs given were based upon the sample plus spike, rather than the sample alone. In the majority of assays, the working range produces CVs which are <15% 2. However, in our experience, at low concentrations, the CVs become much higher due to the nonlinearity of the assay 3. Providing the CVs of spiked samples artificially lowers the CV, and does not give a true representation of the inherent variability of measuring mediators near the lower levels of detection of the assay.
Secondly, the authors stated that “The use of CV is not always the ideal way to express variability: when mean values are low, CV values can be abnormally high” 1. In our experience, increased CVs at low levels of mediators are a genuine reflection of the increased variability of the assay (signal to noise ratio). At low concentrations, assays often become nonlinear and, since all results are based on interpolation, the CV will, by nature, be inherently high 3. Figure 1⇓ demonstrates a typical ELISA optical density versus concentration curve for secretory leukoprotease inhibitor plotted using the CVs observed for the assay at varying concentrations of protein. The figure clearly demonstrates that where the assay becomes nonlinear, the CVs increase from 10% to up to 70%. This truly reflects the variability of the assay in these circumstances.
A typical ELISA optical density (○) versus concentration curve for secretory leukoprotease inhibitor (SLPI). The typical sigmoid curve is demonstrated (––––) together with the coefficient of variation (•) of sample measurements at different SLPI concentrations. A major increase in the variability of the repeated measures is seen for the plateau of the curve (- - - -).
Thirdly, Rosias et al. 1 suggest that a matrix effect is seen when samples are spiked with 10 pg·mL−1 of mediator, but not at 100 pg·mL−1. This suggests that at low levels of spiking, the components of the sample might interfere with the assay or mediator, which is a recognised phenomenon. However, spiking with 100 pg·mL−1, or even 10 pg·mL−1, is likely to overwhelm and disguise this matrix effect, considering that the actual sample values are between 1–6 pg·mL−1 and so any compounds that interfere with the assay are likely to be low concentrations themselves. If there is a matrix effect at low concentrations, this will have a significant effect on the reliability of the readings produced and should be taken into consideration when using the assay.
Fourthly, the authors speculate that variations in biomarker measurements can be explained by differences in dilution, quality of condensate, different collection techniques, sample processing, storage conditions and analytical techniques. Although all of this may be true, using the low concentrations cited by Rosias et al. 1 we have described wide variability in a single sample, principally due to the variability within the assay 3. We do not feel that it is prudent to comment on other sources of variability when the variability of the assay itself is so high.
Finally, the authors do not define the lower limit of quantification for any assay. The lower limit of detection defines the difference from zero but is not able to predict where accurate measurements begin. Lower limits of quantification defines the concentration where the assay is linear, and where CVs and spike returns are known and acceptable for the given assay 4.
We feel that the lower limit of quantification is central to the measurement of mediators at these concentrations within this matrix and using current assay methodology suggests that exhaled breathe condensate is not an ideal collection method for protein measurements 3. Whether the reliability of more sensitive proteomics methodologies overcomes these issues remains to be seen.
Statement of interest
None declared.
- © ERS Journals Ltd
