Before you order that lab test … part 2 - Editorial

Townsend Letter for Doctors and Patients,  Jan, 2004  by Alan R. Gaby

• E-mail
• Print
• Link

In the last issue of the Townsend Letter, I argued that, for many patients, an appropriate diagnosis and effective treatment plan can be developed solely from a careful history and physical exam, supported by a few basic lab tests. More-detailed laboratory testing, including some so-called "alternative" lab tests, may be useful in difficult cases.

Before ordering any test, however, one should ask the following questions: 1) What is the likelihood that the results will influence the management of the patient? 2) What quality-control measures does the laboratory take, in order to assure that their results are accurate and reproducible? 3) How are the reference ranges determined for a particular test (i.e., where does the "normal" population come from; how many standard deviations from the mean does the reference range include)? 4) What is the scientific basis for the interpretation or suggested diagnoses offered by the laboratory?

With regard to question number 1, one does not need to perform a glucose-insulin tolerance test, for example, to be fairly certain that an abdominally obese, Pepsi-guzzling, TV-watching, hypertensive, middle-aged male has the insulin-resistance syndrome. Nor does one need an abnormal test result to prescribe vigorously and confidently an appropriate diet, supplement program, and exercise regimen. While serial glucose-insulin tolerance testing could be used to monitor a patient's progress, so could body weight, abdominal girth, blood pressure, and clinical symptoms. With regard to a patient with chronic gastrointestinal symptoms that began after a course of antibiotics, one does not need a stool analysis to consider a therapeutic trial with probiotics or antifungal medication. Of course, a stool analysis could be considered if the patient fails to respond to these straightforward treatment measures.

Concerning quality control, most laboratories have acceptable procedures in place, including in-house reproducibility checks, and periodic reproducibility checks on samples submitted to outside testing agencies. A few concerns have surfaced, however, in relation to specific laboratories or tests. As I discuss in this month's Literature Review, a study of the hair analysis industry revealed widely discrepant results when split samples taken from the same person were analyzed by different labs. That finding suggests that at least some companies are not providing accurate results for hair analysis.

To assess the reproducibility of food allergy testing, a researcher at Bastyr University obtained 18 tubes of blood from a single healthy individual, drawn on a single occasion. Six tubes were sent to each of three commercial labs, with a different name on each of the six samples. (2) The average difference between the highest and lowest reported values was 73% at one lab and 49% at another lab. That degree of variation is greater than what is considered acceptable by most laboratory standards. For one of these labs, the recommendation to eat or not to eat a particular food was contradicted in 59% of the foods tested in at least two of the six samples. These findings indicate that the results of food allergy testing, at least at some laboratories, are not reproducible and, therefore, cannot be considered reliable.

I also have some concern about the reliability of tests for dehydroepiandrosterone (DHEA), particularly in saliva, but also in serum and urine. The concentration of circulating DHEA-sulfate (DHEA-S) is approximately 1,000 times higher than that of unconjugated (free) DHEA. While DHEA-S levels can be measured accurately, the same may not be true of DHEA, which is present only in nanogram quantities in serum. One laboratory director told me his lab does not offer DHEA testing, because the results are too prone to error. Over the years, in reviewing salivary hormone tests, it has been my impression that an unusually high percentage of the DHEA results are either below normal or at the bottom of the normal range. Serum DHEA levels reported by many labs also seem not to correlate with serum DHEA-S levels. Although some evidence indicates that free DHEA is the active form of the hormone, my clinical experience suggests that measuring DHEA-S provides a more reliable guideline with respect to DHEA therapy.

With regard to reference ranges, for certain tests some laboratories define "normal" within an unusually narrow range, such as one standard deviation from the mean. Defining the normal range that way leads to more than 30% of the test results being classified "abnormal," whereas only about 5% of the results of most conventional laboratory tests are abnormal. It has been argued that "normal" is not necessarily the same as "healthy," and that using the conventional two-standard-deviations-from-the-mean reference range erroneously classifies a lot of sick people as normal. While it is true that more than 5% of the population is unhealthy, it does not necessarily follow that it is proper to narrow the normal range for a lab test. It is an axiom of laboratory testing that narrowing the normal range increases the sensitivity of the test, but decreases its specificity, potentially leading to the conundrum in which, as Bob Dylan said, "you know something is happening, but you don't know what it is, do you ... ?" Of course, in most cases you already knew "something is happening" after taking the medical history.