Recently, I was contacted by an attorney seeking an expert witness for assay of blood ethanol by headspace sampling using gas chromatography with a flame ionization detector and data station. I did not qualify as an expert, but was able to refer the attorney to several colleagues with more relevant experience. However, I was disturbed by the reported practices and lack of internal controls for the assay as described by the attorney.
A quick web search shows that all of the major GC vendors have reported excellent results with their autosamplers feeding samples to their instruments. It seems that the instrumentation and columns can deliver precise results (RSDs about 2%). Of course, the instrument is only part of the assay. The generally accepted description is that the estimated uncertainty of an assay is the square root of the sum of the variances; see Eq. (1):
(σAssay) = (σAnalytics2 + σSample Prep2 + σSample Collection2 +σSample Other2)0.5 (1)
This means that the variation of the largest component of assay protocol dominates the uncertainty of the assay results. For example, if the RSD of Sample Prep is 10 times larger than the other elements of the assay protocol, then (σAnalytics= σSample Collection + σSample Other = 1 + 1 + 1 = 3. σSample Prep2 = 102. The total predicted variance for the assay is (100 + 3) = 103. The predicted uncertainty (σAssay) is (103)0.5 = 10.14889. Thus, the predicted uncertainty is about 1.5% larger than the 10% arising from sample prep. The other factors are much less significant than σSample Prep2. Thus, the other elements in the assay are relevant only if one can reduce the dominant factor (Sample Prep).
However, upon consulting several sources, I became concerned about the lack of controls in sample preparation. For example, the area counts for the internal standard showed a %RSD >30% with n = 80. This is large compared to the precision for the GC. Was this sloppy pipetting, a leaky sample tube, or poor equilibration? For the accused, this can be crucial in estimating the validity of borderline results.
There were no obvious controls on pipetting. Artel has shown that pipetting is a technique-dependent (i.e., operator-dependent) technology. Anecdotal reports show that %CV obtained by untrained operators using air-driven pipets frequently exceeds 30%. Are pipetting technique and equipment an uncontrolled variable in the assay of blood alcohol?
The settings for integration of the chromatogram were very far from the nominal settings recommended by the vendor. This appeared to suppress other peaks in the chromatogram.
So what is the state-of-the-art in blood alcohol assays today?
Is headspace GC (HSGC) the method of choice?
What is good laboratory practice in this important assay?
Do assay variances correlate with individual analysts?
Your views, please.
Robert L. Stevenson, Ph.D., is a Consultant and Editor of Separation Science for American Laboratory/Labcompare; e-mail: firstname.lastname@example.org.