In a recent book "Does Measurement Measure Up? sub-titled "How Numbers Reveal and Conceal the Truth" (1) John M. Henshaw describes how measurements control our lives. He shows the inter-relationship of measurement and knowledge and takes one beyond Lord Kelvin's written words: "To measure is to know"(2). Henshaw writes that a measurement is a number; but that we need to know: how it was obtained; what it means; how it is being applied; and should it be believed or ignored? He describes many uses of measurements resulting in a single number to describe the effects of many differing parameters. How these numbers are used in ratings, in business, in sports, and in computers is described at length; but measurements produced in hospital laboratories, independent laboratories, or police department or crime laboratories are not included.

      A comparison of the measurement practices of these laboratories with those described by Dr. Henshaw reveals very startling information. To begin this comparison, it is important to realize that every measurement has two values: the test value and the true value. The difference between the two values is the uncertainty of the measurement. More than twenty years ago the National Institute of Standards and Technology (the NIST) participated in the development of an inter nationally uniform approach to expressing measurement uncertainty. Their guide to this approach is Technical Note 1297 (3). It ordains that a measurement result is complete only when accompanied by a quantitative statement of its uncertainty. Although Clinical laboratories often report "normal ranges," these are not quantitative statements of the uncertainty of the actual measurements. Test results used in prosecuting Driving Under the Influence (DUI) cases are presented as single numbers without including ranges of test uncertainty. The uncertainty of a measurement is the sum of the uncertainties of the three main components of an analysis: the methodology, the device, and the analyst. For blood or breath analysis for alcohol concentration additional biological variations may be present. Callum Fraser, (4) a recognized worldwide authority on biological variations, writes "Variation can be described as random fluctuation around a homeostatic setting point. The test results of any person may vary over time due to three factors: pre-analytical influences, analytical random error, and inherent biological variation." Short-term fluctuations in the blood or breath alcohol concentrations, also known as "the steeple effect" are examples of inherent biological variations not detectable by the single sample of blood or dual samples of breath as now used in DUI tests. 

        Today, the usual procedure for obtaining a blood sample for DUI is one in which the phlebotomist applies a tourniquet, cleanses an area near the elbow, injects the needle, releases the tourniquet, and collects two tubes of blood. The stressful nature of the process, which affects the result, may be unavoidable without the use of a catheter. Using a catheter, and waiting for five or ten minutes to allow for the stressful situation to return to a state of normalcy, two or more timed samples of blood could be obtained without incurrence of those pre-analytical variables described above. The most common pre-analytical variables for breath alcohol involve variations in the breath samples. The withholding of breath for only a few minutes prior to delivery of the sample will increase the result; whereas a few rapid exhalations prior to the test will give a lower result. Many officers exhort the person being tested to blow harder and longer. This produces higher results. 

         Quantifying the effects of measurement variables has become easily accomplished and revolutionized since the advent of the computer. Computers convert the data into quality control information. Standard deviations, and daily or monthly graphs of data can easily be produced.  Thus, levels of quality control previously set as percentages (parts per hundred) are now being given in parts per million or billion. 

          Six Sigma, as perfected by Motorola, Inc.(5), is a management tool for improving the quality of a product or service. An objective of the program is the reduction of the number of defects or errors (the unacceptable variations from target values) with the aim of eliminating them entirely. The Six Sigma goal is to achieve one error or defect per 3.4 million opportunities, a 99.9997% accuracy. 

          The contrast between the quality control requirements and goals of business and industry and the quality control goals of clinical and crime laboratories is striking.  Clinical Laboratory Act (CLIA) of 1988, set an acceptable proficiency test result as 4 out of 5 tests per year. That is the acceptable accuracy and the goal of 80% for the clinical laboratories.  The acceptable performance of the alcohol determination of blood in DUI cases is somewhat better at 90%.  This accuracy figure derives from the fact that the tests on the standard solutions used in the determinations must agree within ten percent and in the breath test the two test results must likewise agree within ten percent, but the lower of the two results is the accepted value. 

          Carl Garber, Ph.D. FACB, Director of Statistical Applications at Quest Diagnostics, Inc. Lyndhurst, NJ, (6) compared the laboratory accuracies with those of the airline and banking industries as follows.  If the airline industry set its goals for accuracy in airplane landings at 99.9%, they would experience two unsafe landings per day at its largest airport. The banking industry with that same goal would be deducting 22,000 checks from the wrong accounts per hour! Should the driving public and/or the laboratory directors be satisfied with much lower goals for the accuracy of their measurements?

          James O.Westgard, (7), has challenged clinical laboratory professionals to improve the quality of their laboratory test results.  He reports that the analytical quality of clinical laboratory tests is still a serious problem that requires on-going improvements.  All laboratory directors and others interested in the accuracy of laboratory test results are urged to do whatever is necessary to produce results that meet the standards adopted by our National Institute of Standards and Technology so many years ago.

          I ask of everyone interested in accurate alcohol determinations and, especially, all forensic toxicologists, two questions often used by many for other varied reasons. If not us, who?  If not now, when?

References

1.  John M. Henshaw, "Does Measurement Measure Up?" The Johns Hopkins University Press, Baltimore, MD, 2006. 
2. ibid xi. 
3. National Institute of  Standards and Technology, Tech. Note #1297. B. N.Taylor and C.E. Kuyatt, U.S. Gov't. Printing   Office, Washington, D.C. 3001.
4. Callum G. Fraser, "Biological Variation: From Principal to Practice" AACC Press, Washington, D.C.  
5. Six Sigma, http://en.wikipedia.org/wiki/Six_Sigma.
6.  Carl Garber, Clinmical Laboratory News, April 2004, p. 10-14.  
7. J.O. Westgard and S.A. Westgard, "The Quality of Laboratory Testing Today," Am. J.Clin. Path. 20006, 125, pp.343-354.  

Those wishing to respond, question or add relative information are urged to contact me by email.