Accuracy of on-the-spot analysis for vholesterol
ACP J Club. 1991 Sept-Oct;115:56. doi:10.7326/ACPJC-1991-115-2-056
Bachorik PS, Cloey TA, Finney CA, Lowry DR, Becker DM. Lipoprotein-cholesterol analysis during screening: accuracy and reliability. Ann Intern Med. 1991 May 1;114:741-7.
To determine the accuracy and reliability of lipoprotein-cholesterol measurements made during screening.
Comparison of the sensitivity and specificity of measurements made in the field with laboratory analysis.
Satellite, nonlaboratory site of the Johns Hopkins Heart Disease Prevention Programs.
154 adults; 80 (52%) had fasted for ≥ 12 hours and 74 had eaten ≤ 5 hours before screening.
Description of test and criterion standard
Finger-prick (capillary) and venous blood samples were taken and immediately analyzed in desktop centrifugal analyzers (Abbott Vision) by trained operators. Both samples were analyzed for total cholesterol (TC). Split venous samples were also analyzed for triglyceride levels in whole blood and for high-density lipoprotein-cholesterol (HDLC) level in plasma. Low-density lipoprotein-cholesterol (LDLC) was calculated using the Friedewald equation.
Cooled venous samples were transported to the laboratory, and the same measurements were repeated within 2 days. Laboratory measurements were calibrated according to an external standardization program.
Main outcome measures
Comparison of field and laboratory lipoprotein-cholesterol measurements.
All laboratory measurements had a negative bias (mean, 1% for TC; 4%, triglyceride; and 0.6%, HDLC) compared with the external standards; differences between screening values and standards were thus underestimated.
Capillary TC measurements were not significantly different from laboratory values: sensitivity and specificity were 96% and 92%, respectively, for fasting samples; and 98% and 100%, respectively, for nonfasting samples, using a cut point of 5.17 mmol/L (200 mg/dL). Mean venous TC measurements were 5.4% and 3.8% lower than laboratory values in samples from fasting and non-fasting participants, respectively (P < 0.001). The sensitivity and specificity of screening venous samples for TC were 88% and 100%, respectively, for both fasting and nonfasting samples. Mean screening values for triglycerides and HDLC were lower than laboratory measurements (9.8% and 11.2%, respectively). Sensitivity and specificity of screening venous samples for elevated LDLC were 93% and 85%, respectively, using a cut point of 3.36 mmol/L (130 mg/dL). HDLC screening values differed by as much as 40% from laboratory values in 95% of the samples.
Participants with high LDLC were identified as accurately by measuring TC as by calculating LDLC in the field. HDLC field values were not accurate.
Sources of funding: The Johns Hopkins Hospital and Abbott Laboratories.
Address for article reprint: Dr. P.S. Bachorik, The Johns Hopkins Hospital, CMSC 604, 600 North Wolfe Street, Baltimore, MD 21205.
Should HDL measurement be the first step in strategies for detecting and managing lipid disorders? The National Cholesterol Education Program (NCEP) Adult Treatment Panel guidelines treat a low HDL level as an independent risk factor for coronary disease and use the (calculated) LDLC level to determine treatment cutoff values, but lipoprotein fractionation is only recommended for individuals who are first found to have a high-risk TC level. Some experts claim that this strategy is misguided because HDLC and LDLC levels are stronger predictors of coronary risk than the TC level. Further, many persons with a low HDLC or elevated LDLC level have a "desirable" TC level. On the other hand, fractionation is costlier and often less accurate than TC measurement. Studies that demonstrated that HDLC and LDLC levels have a strong association with coronary risk used highly accurate, well-standardized assays; routine test results from clinical laboratories may not predict risk nearly as well. As the NCEP and others have argued, "accurate and standarized methodology (for HDL measurement) is not widely available in clinical laboratories in the United States...more harm than good could come from inaccuracies in reports of HDL-cholesterol levels" (1). The same concerns apply to the calculated LDLC level.
These two studies add fuel to the controversy over the role of TC in screening for lipoprotein disorders. Using meticulously standarized laboratory data from the Lipid Research Clinics program, Bush and Riedel conclude that the strategy of measuring TC first is too insensitive. They found that nearly half of adults at high risk for coronary heart disease on the basis of their LDLC or HDLC levels and other risk factors would have cholesterol levels too low to trigger fractionation when using the NCEP protocol. The study by Bachorik and colleagues, however, may have more relevance to what occurs in practice. Most Americans will be tested in a doctor's office, hospital, or at a field-testing site, where their lipoprotein levels are likely to be measured with a desktop analyzer or by a laboratory procedure that is far less accurate than the procedures used in the Lipid Research Clinics. Bachorik and colleagues show why it may be premature to use relatively inaccurate procedures for HDLC and LDLC measurement as the initial screening tests for lipid disorders.
Their study evaluates the relationship between the "screening" lipoprotein levels measured on a popular desktop analyzer and "gold standard" levels as measured by their CDC-standardized laboratory procedure. They found large disparities between screening values and laboratory values for both the HDLC and LDLC levels. The screening cholesterol measurements, whether obtained from capillary or venous blood, corresponded more closely to the gold standard values. The accuracy of the screening cholesterol was so much greater that it was as accurate as the screening LDLC level in predicting the "true" LDLC level. Their findings also suggest that the inaccurate LDLC levels would be no better than TC levels as an initial screening test for identifying individuals who should receive treatment.
Although Bachorik and colleagues studied a particular desktop analyzer, annual proficiency surveys of the College of American Pathologists confirm that most laboratories measure TC more accurately than HDLC or triglyceride levels. Only when measurements of HDLC and LDLC levels are both accurate and convenient will they be appropriate initial tests for the detection of lipoprotein disorders.
Alan M. Garber, MD, PhD
Stanford University School of Medicine Stanford, California