Rethinking the threshold for an abnormal capillary blood lead screening test.

Abstract

To examine the test characteristics of the capillary blood lead screening test as a predictor of elevated venous blood lead levels, using receiver operating characteristic (ROC) curves. To consider a rational capillary blood lead cutoff value in the context of what has been learned about the screening test and what is understood about the clinical course of children with elevated blood lead levels in the mild range (0.48-0.92 mumol/L [10-19 micrograms/dL]). In a clinical trial, 513 urban children aged 6 years and younger were screened for lead exposure. Paired samples of venous blood were drawn from all children. For these children we examine the ROC curves for capillary blood lead levels as a predictor of elevated venous blood lead levels above 2 thresholds, 0.48 and 0.97 mumol/L (10 and 20 micrograms/dL). Contaminated capillary specimens were defined as those in which the capillary result exceeded the venous result by 0.12 mumol/L (2.5 micrograms/dL) or more (n = 49). Test sensitivity and false-positive rate (equal to 1-specificity) as a function of the capillary screening cutoff value. Area under the ROC curve as a measure of screening test performance. Venous blood lead levels were 0.48 mumol/L (10 micrograms/dL) or more in 20.5% and 0.97 mumol/l (20 micrograms/dL) or more in 2.3% of children. Measurement of capillary blood lead levels performed very well as a screening test with an area under the ROC curve of 0.97 at the 0.48 mumol/L (10-micrograms/dL) threshold and 0.99 at the 0.97-mumol/L (20-micrograms/dL) threshold. For a capillary cutoff value of 0.39 mumol/L (8 micrograms/dL) and an elevated blood lead level threshold of 0.48 mumol/L (10 micrograms/dL), test sensitivity is 100% and the false-positive rate is 23%. Test sensitivity drops to 91%, 63%, and 45% at capillary cutoff values of 0.48, 0.58, and 0.68 mumol/L (10, 12, and 14 micrograms/dL), respectively. The false-positive rate drops to 8%, 2%, and 1% at capillary cutoff values of 0.48, 0.58, and 0.68 mumol/L (10, 12, and 14 micrograms/dL), respectively. Changing the contamination rate by appending or deleting contaminated capillary specimens from the data set had little effect on the area under the ROC curve at either threshold. In this sample of children, capillary blood lead measurement performed well as a screening test for elevated venous blood lead levels. Altering the capillary specimen contamination rate has little effect on the rest characteristics because much of the misclassification error resulted from random analytic error in the analysis of blood lead levels, which is high compared with the threshold of concern (0.48 mumol/L [10 micrograms/dL]). Because of lack of data on clinical outcomes for children with elevated blood lead levels in the 0.48- to 0.92-mumol/L (10- to 19-micrograms/dL) range, we suggest that the greatest utility be placed on avoiding false-positive misclassification. A clinical capillary screening cutoff value of 0.72 mumol/L (15 micrograms/dL) would avoid most false-positive results and would permit 100% sensitivity in detecting children with blood lead levels of 0.97 mumol/L (20 micrograms/dL) or higher.