Background. Recent statements from the American Academy of Pediatrics and Centers for Disease Control and Prevention recommend diagnostic venous blood lead testing within 90 days of a marginally elevated screening test (10–14 μg/dL).
Objective. To evaluate the ability of a marginally elevated capillary (CScr) or venous (VScr) blood lead screening test to predict venous diagnostic (VPb) blood lead (taken within 90 days of the screening test) that would prompt environmental evaluation (≥20 μg/dL).
Design. Population-based follow-up study comparing CScr and VScr with VPb drawn within 90 days of the screening sample. This study population was drawn from all children aged 0 to 4 years who were screened in Worcester County, Massachusetts, and Providence County, Rhode Island, with CScr and VScr during calendar year 1994.
Outcome Measures. To evaluate predictive validity, CScr and VScr were correlated with VPb. CScr, VScr, and VPb results were then separated into the following categories: <10, 10 to 14, 15 to 19, and ≥20 μg/dL. CScr and VScr categories were cross-tabulated against VPb categories, and logistic regression analysis was used to evaluate categorical elevations of CScr and VScr as predictors of VPb ≥20 μg/dL.
Results. Of 31 904 children screened with CScr, 5450 (17.1%) were elevated and 1278 were followed up with VPb within 90 days. Of 14 623 children screened with VScr, 2979 (20.4%) were elevated and 614 were followed up with VPb within 90 days. CScr was only weakly correlated with VPb (r = 0.39), whereas VScr was more strongly correlated with VPb (r = 0.73). Compared with CScr <10 μg/dL, CScr in the 10 to 14 μg/dL range did not identify a higher percentage of children with VPb elevation in any category, and falsely misclassified as lead poisoned some 77% of children. Compared with VScr <10 μg/dL, VScr in the 10 to 14 μg/dL range identified higher percentages of children with VPb in the 10 to 19 μg/dL range but not with VPb ≥20 μg/dL, and falsely misclassified as lead poisoned 42% of children. Compared with screening tests <10 μg/dL, the odds of identifying a child with VPb ≥20 were no different from 1 for CScr of 10 to 14 μg/dL (adjusted odds ratio 1.4 [95% confidence interval 0.3, 6.6]), CScr of 15 to 19 μg/dL (3.2 [0.7, 15.7]), or VScr of 10 to 14 μg/dL (0.9 [0.3, 3.0]). CScr and VScr in the 15 to 19 μg/dL range were associated with significantly higher odds of having VPb ≥20 μg/dL when compared with screening tests <10 μg/dL.
Conclusions. These data indicate that special diagnostic testing within 90 days for children with CScr and VScr in the 10 to 14 μg/dL range does not result in greater identification of VPb ≥20. Raising the set point for diagnostic testing to 15 μg/dL in this sample would eliminate the unnecessary follow-up of 5162 children, of whom 3360 were falsely misclassified as having undue lead exposure.
The Centers for Disease Control and Prevention (CDC) statement of 19911 identified a group of children with blood lead levels (BLL) between 10 and 24 μg/dL, for whom little data were available to guide clinical practice. Since that time, there have been four randomized clinical trials evaluating prevention strategies for children with blood lead elevations in this range (one calcium supplementation,2 two household dust control,3,,4 one soil removal5); none of these prevention strategies has resulted in clinically significant declines of BLL in the treatment group. These data on treatment outcomes suggest that screening for BLL elevations in this range violates a critical concept underlying screening: that early diagnosis and treatment should result in a more favorable outcome for the patient than would otherwise have occurred had screening not taken place.6
A second issue complicates screening of children with mild BLL elevation; laboratory error in the BLL analysis is high in relation to the threshold of concern.7,,8 Most clinical laboratories operate at a level of precision for which the 95% confidence intervals for assay results are ±4 μg/dL for samples in the 10 to 19 μg/dL range.9,,10 These laboratory errors have been highlighted in a recent article in which 2 of 18 commercial laboratories misclassified clinical specimens with BLL of 18 μg/dL as below 10 μg/dL.11 Combined with normal fluctuations in BLL throughout time,12 laboratory errors decrease our confidence in the ability to determine blood elevation in the 10 to 19 μg/dL range with accuracy.
The lead screening debate has focused on the appropriateness of universal screening, with little discussion of the screening BLL cut-off point for an abnormal test result. However, BLL is a continuous variable, and the threshold of concern, 10 μg/dL, is based strictly on concerns about effects of lead on development. The determination of the screening cut-off point for an abnormal level of BLL should be amenable to an evidence-based approach,13 which considers other factors, such as the precision and predictive validity of the screening test, the net cost of a false-positive result, and the benefits of treating a true positive. Raising the screening set point would be expected to decrease false positive misclassification, but is also often associated with an increased likelihood of false negative tests.14 We considered this trade-off in a recent study of paired capillary and venous BLL data. Using receiving-operator characteristics curves, we evaluated the capillary result as a predictor of the venous result15 and suggested that a rational clinical threshold for BLL elevation was 15 μg/dL because this set point minimized the chances of falsely misclassifying a child with BLL <10 μg/dL as elevated and minimized the chances of falsely misclassifying a child with BLL ≥20 μg/dL as only mildly elevated, two outcomes that we wished to avoid.16
In this study, we evaluate the generalizability of our suggestion in a large population-based sample. We focus on the ability of a marginally elevated screening test (10–14 μg/dL) to predict a diagnostic venous lead confirmation result (VPb) that would prompt environmental investigation (VPb ≥20 μg/dL) according to current guidelines.17,,18
We examined lead screening data from the Massachusetts and Rhode Island Childhood Lead Poisoning Prevention Programs for calendar year 1994. Both states require universal annual screening of all children under 5 years of age with BLL analysis. Screening data were obtained for 27 419 and 19 037 children aged 0 to 4 years from Worcester, Massachusetts, and Providence, Rhode Island, counties, respectively. Screening took place in a variety of settings including: doctors' offices; community health centers; Women, Infants, and Children programs; and public clinics. More than 80% of specimens from both counties were analyzed by their respective state laboratories by graphite furnace atomic absorption spectroscopy. In addition, both laboratories participate in the CDC Blood Lead Proficiency Program.
Selection criteria are illustrated in Fig 1. Overall, 12.1% (n = 3319) of the children in Worcester County and 13.9% (n= 3002) of the children in Providence County had more than one lead test. We identified all children with a capillary screening test (CScr;n = 4047) or venous screening test (VScr;n = 2274) that was followed with another test during 1994. From this sample, we studied children who had a CScr followed by a venous diagnostic test (n = 1824) and those who had a VScr followed by a venous diagnostic test (n = 1751).
Current Follow-up Recommendations for Marginally Elevated Blood Lead Screening Tests
The most recent CDC statement on lead poisoning17distinguishes between a screening test and a diagnostic test (“the first venous BLL test performed within 6 months on a child with a previously elevated BLL on a screening test. If the diagnostic test is not performed within 6 months, the next test is considered a new screening test”). The CDC recommends that a screening test in 10 to 14 μg/dL range be confirmed with a diagnostic test within 3 months. A recent statement on lead poisoning by the American Academy of Pediatrics18 recommends that a screening BLL in the 10 to 14 μg/dL range be followed with a confirmatory test within 1 month. To evaluate the utility of these currently recommended diagnostic strategies, we restrict our follow-up period to 90 days or less. Of CScr with follow-up in our sample, 1277 were evaluated within 90 days; of VScr with follow-up, 613 were evaluated within 90 days.
We separated CScr, VScr, and VPb into the following categories: <10, 10 to 14, 15 to 19, and ≥20 μg/dL and cross-tabulated the screening with the diagnostic results. We also graphed the screening against the diagnostic results to evaluate correlation and illustrate bias. Finally, we used logistic regression analysis to determine the ability of elevated screening results to predict VPb ≥20 μg/dL. We evaluated for confounding by including age (dummy variables indicating age in years), season (dummy variables indicating season when the screening test was obtained: winter [January–March], spring [April–June], summer [July–September], and fall [October–December]), interval (time between screening test and VPb in days), and interactions between season and interval. Only interval and the interaction terms (season × interval) were significant and are included in final models; their inclusion did not substantially alter the unadjusted results.
Capillary screening results were obtained in 31 904 children during calendar year 1994 in Worcester and Providence counties, of which 5450 were elevated: 3368 (10.6%) 10 to 14 μg/dL, 1136 (3.6%) 15 to 19 μg/dL, and 946 (3.0%) ≥20 μg/dL. Venous screening results were obtained in 14 643 children, of which 2979 were elevated: 1784 (12.2%) 10 to 14 μg/dL, 679 (4.6%) 15 to 19 μg/dL, and 516 (3.5%) ≥20 μg/dL, reflecting a preference for venous screening among urban Providence County providers where many high-risk children reside. Sixty-one percent (n = 5152) of all children who required diagnostic testing according to the 1997 CDC guidelines had blood lead in the 10 to 14 μg/dL range.
Table 1 shows follow-up rates for CScr and VScr according to characteristics of the study population. Follow-up was substantially and significantly more likely in children with VScr, higher screening blood lead category, younger age, those screened earlier in the year, and those living in Rhode Island (for CScr only), and those who resided in census tracts with higher proportions of households receiving public assistance income.
Table 2 shows VPb categories for CScr tests that were followed up within 90 days. Compared with CScr <10 μg/dL, CScr samples in the 10 to 14 μg/dL range did not identify a higher proportion of children within any category of elevated VPb (P = .33). In addition, CScr in the 10 to 14 μg/dL range falsely identified 393 (77.3%) children with normal VPb as elevated, resulting in unnecessary follow-up in these cases. Follow-up of CScr in the 15 to 19 μg/dL and >20 μg/dL categories resulted in the identification of a significantly higher percentage of children with VPb ≥20 μg/dL (P < .0001 in both cases) compared with CScr <10 μg/dL.
Figure 2 shows a plot of VPb against CScr for the 1278 capillary screened children described in Table 1; the line reflects expected values if there were perfect correlation between CScr and VPb. Figure 2 reveals why CScr is a poor predictor of VPb; a striking positive bias is seen, with CScr exceeding VPb in the majority of cases. Correlation between CScr and VPb is low (Pearsonr = 0.29). Three outliers are not included, for which CScr is 106, 138, 240, and for which VPb is 2, 9, and 7 μg/dL, respectively.
Table 3 shows VPb categories for 614 VScr tests that were followed up within 90 days. Compared with VScr <10, VScr samples in the 10 to 14 μg/dL range did not identify a higher proportion of children with VPb ≥20 μg/dL, but were able to identify higher proportions of children with VPb 10 to 14 μg/dL and 15 to 19 μg/dL (P < .0001). As with the CScr strategy, VScr in the 10 to 14 μg/dL range falsely identified as elevated, a substantial proportion of children (42.0%).
Fig 3 shows a plot of CScr against VPb for the 614 children described in Table 2, revealing that VScr is a better predictor of VPb than CScr. The points are more evenly distributed around the line of perfect correlation, indicating little positive bias in VScr. Correlation between VScr and VPb is higher than was seen for CScr and VPb (Pearson r = 0.78).
The CScr and VScr results are evaluated as predictors of VPb ≥20 μg/dL obtained within 90 days in Table 4. Neither CScr nor VScr in the 10 to 14 μg/dL range were predictive of VPb ≥20 μg/dL (adjusted odds ratio 1.4 and 0.9, respectively). In addition, CScr in the 15 to 19 μg/dL range was not predictive of VPb ≥20 μg/dL (adjusted odds ratio 3.2 [95% confidence interval 0.7, 15.7]). Categories that were predictive of VPb ≥20 included CScr ≥20, VScr 15 to 19, and VScr ≥20 μg/dL.
This study examines screening for lead exposure as practiced in the community setting for a large population-based sample. Regardless of whether the specimen was drawn using capillary or venous techniques, screen positive tests in the 10 to 14 μg/dL range were not predictive of diagnostic tests that would prompt environmental action within a 90-day follow-up period. Moreover, compared with CScr <10 μg/dL, CScr samples in the 10 to 14 μg/dL range did not seem to predict any category of VPb elevation. Finally, both VScr and CScr in the 10 to 14 μg/dL range were associated with substantial false positive misclassification errors, amounting to 42% and 77% of screen positive tests, respectively. Both screening modalities have high misclassification error rates because of the analytic error of the blood lead measurement (±4 μg/dL) is high with respect to the threshold of concern, 10 μg/dL. Much of the variation in VScr is attributable to random analytic error in the analysis of blood lead, explaining why VPb may be lower than VScr. Higher misclassification error rates among CScr samples is probably attributable to positive bias in the measurement of CScr resulting from finger skin contamination with lead.12,,15,19
Because there is currently no available intervention known to produce a clinically significant decrease BLL in children with mild exposure (BLL, 10–19 μg/dL), the utility in identifying and following children in this range depends on the ability of the test to predict BLL elevations that should prompt environmental action (≥20 μg/dL). Further study of interventions for children with blood lead in the 10 to 19 μg/dL range are needed.20 In the meantime, our data suggest little benefit to the CDC 1997 recommendation that children with BLL in the 10 to 14 μg/dL range be followed up with diagnostic measurement within 90 days. Dropping the recommendation for diagnostic testing in this sample would have alleviated the need for primary care providers to recall some 5162 children (61% of all screen positive children), 3360 of whom would have been erroneously told that they had significant lead exposure.
In interpreting our conclusion, it is important to remember that our study addresses only the utility of diagnostic testing within 90 days and not the appropriate periodicity of screening, which should occur as often as every 6 months in some high-risk populations. In addition, among low-risk populations, a marginally elevated screening test may be an indication to rescreen that child within 6 months or sooner in the case of young infants. Finally, we concur with the recommendation by the Center for Disease Control that community-wide primary prevention efforts to remove sources of lead be applied if many children in that community have BLL ≥10 μg/dL.21 In high-risk communities it is often deteriorating paint from substandard rental housing that is responsible for the population-based lead exposure. Rather than focusing on how the renters who have children with marginally elevated blood lead can keep old houses cleaner, we should focus our attention on how to structure state housing policy so that paint sources of lead are permanently removed from old housing.
One limitation with this study has to do with work-up bias, which is a problem with many population-based studies reflecting incomplete follow-up, especially among screen-negative children. Incomplete follow-up does not allow us to calculate sensitivity and specificity of the CScr and VScr screening tests. Follow-up tends to be biased in favor of children who are at higher risk by virtue of their age and place of residence. Follow-up rates were also lower among children who were screened toward the end of the study period because these children were not allowed the full 90 days before the study period ended. We take reassurance that the results from this study concur with those from a previous study in which work-up bias was not a factor.16
In summary, we suggest that clinicians consider a set point of 15 μg/dL for diagnostic VPb evaluation because this minimizes unnecessary follow-up of false positives and because this does not result in a substantial increase in false-negative misclassification of children for whom housing inspection is recommended.
This work was supported by Grant 022315 of the Robert Wood Johnson Foundation.
We thank Brad Prenney, Paul Hunter, and Mary Jean Brown of the Massachusetts Childhood Lead Poisoning Prevention Program for supplying screening data from Worcester County. We also thank Susan Feeley, PhD, Robert Vanderslice, William Dundulis, and Peter Simon of the Rhode Island Department of Health, Family Health Services, for the results from Providence County Lead Screening. The analysis and conclusions are our own.
- Received May 18, 1998.
- Accepted October 19, 1998.
Reprint requests to (J.D.S.) Pediatrics and Adolescent Medicine, Dartmouth-Hitchcock Medical Center, One Medical Center Dr, Lebanon, NH 03756.
- CDC =
- Centers for Disease Control and Prevention •
- BLL =
- blood lead level •
- VPb =
- venous lead confirmation results •
- CScr =
- capillary screening test •
- VScr =
- venous screening test
- ↵Centers for Disease Control and Prevention. Preventing Lead Poisoning in Young Children: A Statement by the Centers for Disease Control. Atlanta, GA: Centers for Disease Control and Prevention; 1991
- ↵Sargent JD, Dalton MA, O'Connor GT, Olmstead EM, Klein RZ. Randomized trial of calcium-phosphorus supplemented infant formula to prevent lead exposure. Am J Clin Nutr. In press
- Lanphear BP,
- Winter NL,
- Apetz L,
- Eberly S,
- Weitzman M
- ↵Hennekens CH, Buring JE. Epidemiology in Medicine. Boston, MA: Little Brown and Company; 1987
- ↵National Research Council. Measuring Lead Exposure in Infants, Children, and Other Sensitive Populations. Washington, DC: National Academy Press; 1993
- Harvey B
- ↵Centers for Disease Control and Prevention. Blood Lead Proficiency Testing. Atlanta, GA: US Department of Health and Human Services, Public Health Service; 1994
- Schonfeld DJ,
- Cullen MR,
- Rainey PM,
- et al.
- ↵Sox HC, Blatt MA, Higgins MC, Marton KI. Medical Decision Making. Boston, MA: Butterworth-Heinemann; 1988
- Sargent JD,
- Dalton M,
- Stukel TA,
- Roda S,
- Klein RZ
- ↵Centers for Disease Control and Prevention. Screening Young Children for Lead Poisoning: Guidance for State and Local Public Health Officials. Atlanta, GA: Centers for Disease Control and Prevention; 1997
- American Academy of Pediatrics, Committee on Environmental Health
- Sargent JD,
- Dalton M,
- Stukel TA,
- Roda S,
- Klein RZ
- Lanphear BP
- Copyright © 1999 American Academy of Pediatrics