CLIN. CHEM. 40/9, 1730-1734 (1994) #{149}Automation and Analytical Techniques

1730 CLINICALCHEMISTRY, Vol. 40, No. 9, 1994

Graphite Furnace Atomic Absorption Spectroscopic Measurement of Blood Lead inMatrix-Matched StandardsDesmond I. Bannon,”3 Catherine Murashchik,’ Caroline R. Zapf,’ Mark R. Farfel,”2and J. Julian Chisoim, Jr.1

Now that the level of concern for a toxic blood lead con-centration is 0.482 moI/L (10 gIdL), laboratories mustmeet new requirements to shorten analysis times andincrease accuracy and precision of blood lead determina-tions. We used a matrix-matching method to estimate thelead concentration in blood by graphite furnace atomicabsorption spectroscopy (GFAAS). For CDC proficiencysamples and the NIST-Certified Blood Reference stan-dard, the performance of this method compared favorablywith that of previously published GFMS methods and ofthe anodic stripping voltammetric method routinely usedin our laboratory. At lead concentrations of 0.242 .tmol/L(5.01 gIdL) and 1.478 Lmol/L (30.63 1.tg/dL),within-runCVs were 2.78% and 0.68%, respectively; between-runCVs were 4.9% and 1.35%. In 52 study sampleswith leadcontent ranging from 0.097 to 3.812 &mol/L (2 to 79gIdL), 87% of results by the matrix-modified methodwere within 0.048 moI/L (1 pgIdL) of consensus values.

IndexingTerms:toxicology/callbration/anodic stripping voltamme-by compared

The adverse health effects of exposure to lead in sen-sitive populations has been well documented (1, 2). Inparticular, the spectrum of chronic effects in childrenfrom low-dose exposure is a serious public health con-cern. The recent lowering by the US Centers for DiseaseControl and Prevention (CDC) of the definition of anabove-normal blood lead concentration-to 0.483moJJL (10 pg/dL)-refiects this concern.4 Precise, accu-rate measurements of blood lead �0.483 tmolfL and theability to distinguish between samples having similarconcentrations are becoming increasingly important forclinical and research purposes.

Based on proficiency testing results, laboratory per-formance has been steadily improving over the last 10years, such that experienced reference laboratories canachieve results for lyophilized proficiency samples thatare accurate to ±0.097 jmo]/L (±2 g/dL) (3). In gen-eral, however, consensus ranges for the CDC proficiencyparticipants are ± 0.193 1imoJfL (±4 tg/dL) for lead con-

1 Kennedy Krieger Institute, Trace Metals Laboratory, 3001#{189}E. Biddle St., Baltimore, MD 21213.

2J0j Hopkins University School of Hygiene and PublicHealth, Baltimore, MD.

Author for correspondence. Fax 410-550-8190; E-mailbannon@welchlink.welch.jhu.edu.

Nonstandard abbreviations: CDC, Centers for Disease Controland Prevention; GFAAS, graphite furnace atomic absorption spec-troscopy; NIST, National Institute of Standards and Technology;ASV, anodic stripping voltanimetry; and m0, characteristic mass.

Received April 4, 1994; accepted May 31, 1994.

centrations � 1.93 .&mol/L (40 p.g/dL) and ± 10% forthose >1.93 jmoWL (4).

Analysis of whole blood for lead is generally done witheither graphite furnace atomic absorption spectroscopy(GFAAS) or anodic stripping voltammetry (ASV). A re-cent CDC proficiency summary showed that 69% of par-ticipating laboratories used GFAAS and 23% used ASVfor the analysis of lead in whole-blood samples (4).Other methods (5-8) either have not been well estab-

lished or, like ASV, do not lend themselves easily toautomation.

Recent efforts by other researchers to increase theaccuracy of either ASV or GFAAS have involved in-creases in sample preparation time (9) or data-analysistime (10). Our current GFAAS technologr can use aque-ous standards for calibration (11), which is fast andadequate for screening, but we wanted a method withimproved accuracy and precision over that in previouslypublished work, particularly at lower concentrations.

The ASV method has been the method of analysis forblood specimens drawn in the weekly lead clinics at theKennedy Krieger Institute for nearly 20 years. Duringthis time, our laboratory has been participating success-fully in external proficiency programs and has served asa reference laboratory for several of these programs.

We have now developed a method for blood lead mea-surement with our graphite furnace instrument; ourapproach is based on currently used methods (12, 13),into which we have introduced a novel technique forcalibration. Here we report our comparison of thismethod with established methods, in terms of perfor-mance in an external proficiency testing program andaccording to previously published GFAAS studies. Wealso assessed method performance with human bloodsamples and compared the results with those by ASV forthe same set of samples.

MaterIals and MethodsApparatus

We used a Zeeman/5 100 PC atomic absorption spec-trophotometer with HGA-600 graphite furnace andAS-60 autosampler (all from Perkin-Elmer, Norwalk,CT). The instrument was interfaced to a DEC station316 SX computer with MS DOS 3.3, GEM Version 2.20(Rev A) and 5100 PC Version 6.2 (Rev E) software pack-ages. An Intensitron#{174}hollow-cathode lead lamp (Per-kin-Elmer) was the light source, and we used pyrolyti-cally coated graphite tubes with L’vov platforms(Perkin-Elmer) for sample atomization. Dilutions ofstandards and samples were carried out with a Digi-

flex” automatic pipette (ICN Biomedicals, Costa Mesa,

Stage

DryPre-ashAshCoolAtomizeClean

Temp, #{176}C

12026080020

16002700

Time, S

Ramp Hold5 151 5

5 271 40 51 2

Gas flow,LImln300300300300

0

300a The injection volume was 12 L; the flow gas was argon; and the injection

temperature was 70#{176}C.

CLINICALCHEMISTRY, Vol. 40, No. 9, 1994 1731

CA). For anodic stripping analysis (ASV) we used theESA 3010A Trace Metals Analyzer (Environmental Sci-ence Associates, Bedford, MA) with a mercury-coatedgraphite electrode, an Ag/AgC1 reference electrode, anda platinum counterelectrode.

ReagentsNitric acid was Baker Instra-Analyzed”; ammonium

dihydrogen phosphate was Ultrex”‘; Triton X-100 wasBaker Analyzed” (all from J. T. Baker Chemical Co.,Phillipsburg, NJ). “PE PURE” atomic absorption spec-troscopy lead standard was from Perkin-Elmer. We ob-tained Standard Reference Material 955a1 Lead inBlood, from the National Institute of Standards andTechnology (NIST; Gaithersburg, MD). Human bloodsamples measured for lead by isotope dilution massspectroscopy (by W. I. Manton, University of Texas)were used as primary reference standards for ASV. Sec-ondary standards for ASV were made from lead-supple-mented bovine blood. Monthly proficiency samples camefrom participation in the CDC/ Wisconsin State Labora-tory testing program. Human blood samples came fromthe routine lead clinics of Julian Chisolm at the Ken-nedy Krieger Institute.

For ASV analyses, we used a specially prepared lab-oratory reagent instead of the Metexchange” reagentrecommended by ESA. Our reagent contained 52.6 g ofKC1 and 16.7 mL of 6 molIL HC1 (GFS, Columbus, OH),76 mg of HgC12 (Spex Industries, Edison, NJ), 320 mgofNiCl2 (Johnson Matthey, Ward Hill, MA), 0.4 mL ofTriton X-100 (Kodak, Rochester, NY), and 0.2 mL of2-octanol (Sigma, St. Louis, MO) added to -500 mL ofdeionized water, mixed thoroughly, and diluted to atotal volume of 2 L with deionized water. The pH wasbetween 1.3 and 1.4 (adjusted by adding small quanti-ties of HC1). The KC1, in solution, was passed through acation-exchange column (Chelex 100; Bio-Rad, Rich-mond, CA) to remove any trace quantities of lead beforeuse.

Sample Collectionand StorageHuman blood samples were drawn from children who

had been referred to the lead clinic at Kennedy KriegerInstitute. The blood samples were taken as part of aroutine weekly lead clinic held at the Institute, andsampling adhered to the policy and protocols outlined byKennedy Krieger Institute.

Venous blood samples were collected into purple-cap(EDTA-containing) Vacutainer Tubes (Becton Dickin-son, Rutherford, NJ) after scrubbing of the arm withalcohol swabs followed by venipuncture. The collectedsamples were analyzed within the hour by ASV (whichwas available on-site) and later by GFAAS (in a differ-

ent location). Between these analyses, samples werestored at -20#{176}C.

ProceduresGlasswashing procedure. To ensure lead-free glass-

ware/plasticware, we washed our containers with Aca-tionox detergent, then soaked them in dilute (250 milL)

nitric acid overnight. After soaking, the glassware wasrinsed with deionized water (Hydro, Research TrianglePark, NC; prepared by reverse osmosis with two polish-ing columns) and air-dried until used.

Instrument settings. We optimized the furnace set-tings by using the procedures recommended by the man-ufacturer. Absorbance was measured at 283.3 nm. Theresulting time-temperature program (Table 1) lasted 71s, the injection volume was 12 L, and the cycle time foreach determination was 102 s.

Sample preparation and standardization. Blood wasdiluted 10-fold in a matrix-modifier solution ofNH4H2PO4 (2 g/L) + HNO3 (2 milL) + Triton X-100(0.5 milL) in deionized water and analyzed against ma-trix-matched standards. Instead of subtracting the ab-sorbance of the baseline blood by autozeroing, we in-cluded it as part of the calibration curve. We usedpyrolytically coated graphite tubes, L’vov platforms, in-tegrated absorbance, and Zeeman background correc-tion. A 12-L aliquot of the diluted blood sample wasdeposited on the platform, and each sample was ana-lyzed in duplicate.

On the Digiflex pipette the left syringe was set to 300hL and the right syringe to 100 L; these settings wereused to dilute (10-fold) the well-mixed low-lead bloodinto disposable culture tubes with caps, which were vor-tex-mixed for 10-20 s until the blood cells were com-pletely lysed. The diluted blood was then used in thepreparation of calibration standards each week as fol-lows. Stock solutions of lead-SO, 100, 300, 400, and 500gfL prepared in 2 milL nitric acid-were diluted 10-fold with the diluted, lysed blood in 2-mL autosamplercups and vortex-mixed. The resulting calibration stan-dards had concentrations of 5 + n, 10 + n, 30 + n, 40 +n, and 50 + n g/L, where n was the concentration of thediluted low-lead blood. The characteristic mass value(mo) for the instrument and the measured absorbance ofthe low-lead blood sample were used to calculate thevalue of the base blood, [ni. The base blood was fromblood samples delivered to the laboratory either frozenor fresh; samples were thawed if necessary and vortex-mixed thoroughly for 10 s before dilution with matrix-modifier solution. Before calibration, the instrumentwas autozeroed with the matrix-modifier solution as theblank. We assumed that m0 remained fairly constantfrom day to day. For instance, the lead m0 for this

Table 1. Graphite furnace settings for blood leaddetermlnatlon.a

70 y=0.984x+0.476R2=0.982

SO

150

j4066

*

10 #{163}

10 30 30 40 00 50 70 SO S

Ccn.snsu. V.111.. W9IdL Pb)

SO

y=0.977x+0.292

R2=0.996

#{149}0

&

150

I40

a

30

10

0 10 30 30 40 50 00 70 00 50

Con..nsu. Values(ag/dL Pb)

Fig. 1. Regressionanalysis of consensusvalues for blood leadconcentrationsvs (top) ASV measurementsand (bottom) GFAASmeasurements.

1732 CLINICALCHEMISTRY, Vol. 40, No. 9, 1994

instrument was defined by the manufacturer as themass of lead (12 pg) that gave an absorbance of 0.0044A’s (i.e., 1% absorbance). The mean m0 for our instru-ment over a 1-year period was 12.5 pg (SD O.7pg#{176},n =

18). For a characteristic mass of 12.5 pg and a sampleabsorbance value of 0.010 A’s, the base blood lead con-centration, [n], would be 0.114 .tmol/L (2.36 g/dL), ascalculated with the following equation:

s x m0 x dfBlood lead (pg/L) =

aXv

where s = blank-corrected absorbance reading for baseblood (A’s), a = 0.0044 As (1% absorbance), m0 = 12.5pg/O.OO44 A’s, df = factor for dilution of base blood withmatrix modifier (i.e., 10), and v = injection volume (formolar units, 10 pg/dL = 0.483 moI/L). For a base bloodlead concentration of 2.36 g/dL (see example above), acalibration curve was constructed with standards of7.27, 12.27, 32.27, 42.27, and 52.27 g/L (r � 0.999).

Proficiency samples and patients’ samples were pre-pared in duplicate by diluting 100 .&Lof sample with900 L of modifier solution and vortex-mixing for 10 s.For quality control we analyzed two reference samplesof high (1.478 moUL; 30.63 pg/dL) and low (0.242mol/L; 5.01 pg/dL) lead concentration. All sampleswere first analyzed by ASV, as is routine, and then byGFAAS with the new method.

Calibration of the ASV method used was based on thelead concentration of human blood-based primary stan-dards, analyzed by thermal ionization mass spectrome-try by William I. Manton at the University of Texas.These standards were used to set the values for a set ofsecondary calibrations made from lead added to bovineblood in concentrations ranging from 0.097 to 3.185moI/L (2 to 66 1g/dL). These secondary standards werethen used for routine calibration of the ASV duringanalysis of clinical and proficiency samples. The methodof analysis was based on the standard recommendedoperating procedures in the manufacturer’s manual(ESA; Dec. 1975).

Quality Control

After calibration we used NIST 955a samples, onehigh concentration and one low, for continuing verifica-tion of linearity. The low standard was 0.242 moI/L(5.01 g/dL) until our supply of this standard was ex-hausted, after which we used an equivolume dilution ofthe 0.653 .tmo]/L (13.53 g/dL) standard (i.e., a 0.327zmol/L low standard). Our high standard was 1.478moI/L (30.63 1.g/dL). Low and high standards wereanalyzed after calibration and after every four sampledeterminations. Our requirement for process controlwas a measured value within 0.097 mol/L (±2 g/dL)of the certified value for each high and low quality-control sample.

ResultsThe individual lead concentration values for the 52

proficiency blood samples analyzed by ASV and GFAAS

are plotted with their fitted least squares regressionlines in Fig. 1. Results of the GFAAS and ASV methods

correlated well with the consensus values (Pearson r =

0.996 and 0.982, respectively). The regression interceptfor each method was not significantly different from 0(ASV intercept = 0.48, SE = 0.60; GFAAS intercept =

0.29, SE = 0.29), indicating an absence of significantbias in the methods. For both ASV and GFAAS methodsthe hypothesis 13 = 1 was not rejected ((3= 0.984, SE =

0.02; (3 = 0.977, SE = 0.01, respectively), indicating thateach method is a good predictor of the consensus value.

The GFAAS method is more precise than the ASVmethod, according to an F-test for common variancebetween two samples (homogeneity of variance test).

Ref.

9

GFAAS

879298

100NA

ASV

52768894

100

NIST-c.rtlfledvalue

6.77j,

13.5330.63

6.9813.5929.59

101010

Within-run

CV,2.781.690.68

CLINICAL CHEMISTRY, Vol. 40, No. 9, 1994 1733

The Fratjo of 4.4 (P <0.001) indicates that the GFAASmethod is more precise (less variable) than the ASVmethod.

Moreover, 87% of measured values were within 0.048/.LmoI/L (± 1 g/dL) of the consensus value by the newGFAAS method vs 52% by the ASV method (see Table2). The relative standard deviations (CV) for replicateanalyses of three certified blood lead standards byGFAAS are listed in Table 3.

For 39 clinical whole-blood samples analyzed by bothmethods, regression of ASV values on GFA.AS valuesyielded the following estimates: slope = 1.01 (SE =

0.04), intercept = 0.20 (SE = 2.3).GFAAS measurements of a NIST 955a at two lead

concentrations, 0.242 jtmoIJL (5.01 p.gfdL) and 1.478hmoIJL (30.63 g/dL), were made on 20 different days

over a 1-year period. The mean measured value for thehigh standard was 1.466 jmol/L (30.38 p,g/dL, SD =

0.411 ,ag/dL; CV = 1.35%) and 0.240 pmo1/L (4.99 p.g/dL, SD = 0.245 g/dL; CV = 4.9%) for the low standard.

DiscussionThe new method performed well compared with pre-

viously published graphite furnace methods (9, 12, 13)(see Table 4). Although the methods listed in Table 4used a variety of replicates for determining precision,our method compares well with all of them both inaccuracy and precision. In particular, the within-run

precision and accuracy of our method shows that lowconcentrations of lead in blood can be measured withimproved accuracy and precision.

Although other methods of establishing the concen-tration of lead in the base blood could be used, e.g.,standard additions or thermal ionization mass spec-

Table 2. ASV and GFAAS results compared withconsensus values for 52 proficIency testing samples.

% of observations withindifference limitDifferences between observed

value and consensusvalue,pmol/L (ag/dL)

± 0.048 (± 1)± 0.096 (± 2)± 0.144 (± 3)

± 0.192 (± 4)± 0.384 (± 8)

NA, not applicable.

Table 3. Replicate measurements by GFAAS ofcertlf led blood lead standards.

Measuredmean

No. ofPb, ag/dL replicates

NIST Standard Reference Material 955a, Lead in Blood; replicates weremeasured In 1 day.

b Made by diluting 13.3 g/dL standard twofold.

Table 4. Comparison of selected results by variousGFAAS methods.

Certified Measuredvalue mean

Pb, og/dL

NANA

12 5.0130.6

13 5.530.42

This report 6.7730.63

This report 5.0130.63

a Within-run.Duplicates across 5 days.Across 20 analytical runs.

d Across 20 days.

replicates CV, %

5.17 ha 3.277.83 ha 1.449 .,0b 14

30.0 100 3.0NA 4#{216}C 7.0NA 40c 1.96.98 ba 2.8

29.59 ba 0.74.99 20d 49

30.38 1.4

trometry, these methods would be either time consum-ing or costly or both. We found that our daily measuredvalue of m0 did not vary enough to introduce significantimprecision into the method. For example, with ourmean m0 of 12.5 and a measured absorbance of 0.010 A’sthe estimated value of the base blood could have variedbetween 0.108 j.moI/L (2.23 g/dL) and 0.121 mol/L(2.50 g/dL), a difference that we did not consider sig-nificant because it was so close to the detection limit ofthe instrument. A furnace kept in good condition withregular cleaning and maintenance visits should havereasonably repeatable performance characteristics, in-cluding the characteristic mass.

Recent methods have also shown that aqueous stan-

dards can be used as calibrators for measuring lead inblood (11), given current innovations in furnace tech-nology. With older technology, however, our methodshould allow improved precision and accuracy and per-haps less dependence on very low concentrations ofblood lead for calibration, because the value of the low-lead blood used to make up the calibrators will be in-cluded in the calibration line.

Overall, the GFAAS method had better precision andtook a shorter analytical time than did ASV, althoughnewer-generation ASVs will probably show improvedperformance in time and precision. We found no evi-dence of systematic biases between the two methods ona set of 39 clinical blood samples ranging from 0.097 to2.220 /.LmoI/L (2-46 g/dL); thus, like ASV, the GFAASmethod could be used to provide a rapid analysis in aclinical setting.

In conclusion, we present a method that has betterperformance characteristics than those of an establishedASV method, has a matrix-based method of calibration,and has the ability to meet both clinical and researchneeds for improved blood lead measurement at low con-centrations. This is especially critical where researchinto the subclinical effects of low concentrations of blood

1734 CLINICAL CHEMISTRY, Vol. 40, No. 9, 1994

lead on children is being undertaken. We have used themethod for 9 months with a history of good performancein the CDC proficiency program, in which 24 of 27 mea-surements within the range 0.097-2.799 jimol/L (2-58g/dL) fell within 0.048 /LmolJL (1 g/dL) of the consen-sus values.

This work was funded by the Lead Poisoning Prevention Pro-gram at Kennedy Krieger Institute. We thank Charles Rohde forinvaluable help with statistics and Brian Rooney for data manage-ment, both of Johns Hopkins University. We also thank William I.Manton at the University of Texas for analytical assistance.

References1. Measuring lead exposure in infants, children, and other sensi-tive populations. Washington, DC: National Academic Press, 1993:16-39.2. Landrigan PJ. Toxicity of lead at low dose. Br J md Med1989;46:593-6.3. Parsons PJ. Monitoring human exposure to lead: an assessmentof current laboratory performance for the determination of bloodlead. Environ Res 1992;57:149-62.4. US Centers for Disease Control and Prevention (CDC). Bloodlead proficiency testing. Atlanta, GA: USDHHSIPHS, March 1993:1-2.5. Ost.apczuk P. Direct determination of cadmium and lead in

whole blood by potentiometric stripping analysis. Clin Cheni1992;38:1995-2001.6. Wang ST, Demshar HP. Determination of blood lead in driedblood-spot specimens by Zeeman-effect background-correctedatomic absorption spectrometry. Analyst 1992;117:959-61.7. Verebey K, Eng YM, Davidow B, Ramon A. Rapid, sensitivemicro blood lead analysis: a mass screening technique for leadpoisoning. J Anal Toxicol 1991;15:237-40.8. Omenetto N, Human HG, Cavalli P, Rossi G. Direct determi-nation of lead in blood by laser-excited flame atomic-fluorescencespectrometry. Analyst 1984;109:1067-70.9. Jacobson BE, Lockitch G, Quigley G. Improved sample prepara-tion for accurate determination of low concentrations of lead in wholeblood by graphite furnace analysis. Clin Chem 1991;37:515-9.10. Roda SM, Greenland RD, Bornschein RL, Hammond PB.Anodic stripping voltammetry procedure modified for improvedaccuracy of blood lead analysis. Cliii Chem 1988;34:563-7.11. Parsons PJ, Slavin W. A rapid Zeeman graphite furnaceatomic absorption spectrometric method for the determination oflead in blood. Spectrochim Acta 1993;48B:925-39.12. Shuttler IL, Delves HT. Determination of lead in blood byatomic absorption spectrometry with electrothermal atomization.Analyst 1986;111:651-6.13. Miller DT, Paschal DC, Gunter EW, Stroud PE, D’Angelo J.Determination of lead in blood using electrothermal atomizationatomic absorption spectrometry with a L’vov platform and matrix

modifier. Analyst 1987;112:1701-4.