Biological Trace Element Research - American Chemical Society

---

Chapter 10

Accuracy and Precision of Trace Metal Determinations in Biological Fluids Interlaboratory Comparison Program

Downloaded by UNIV MASSACHUSETTS AMHERST on August 2, 2012 | http://pubs.acs.org Publication Date: December 26, 1991 | doi: 10.1021/bk-1991-0445.ch010

Jean-Phillipe Weber Centre de Toxicologie du Quebec, Le Centre Hospitalier de l'Université Laval, 2705 Boulevard Laurier, Quebec G1V 4G2, Canada

Since 1979, we have been conducting an interlaboratory comparison program for several toxic elements in blood and urine with a view to validate the accuracy and precision of toxicological trace element analyses. Presently, over 120 North American and European laboratories participate. Samples are prepared by pooling the specimens obtained from exposed workers and patients and are sent bimonthly to participants. The target values determined from the results of reference laboratories are used to evaluate proficiency of the participating laboratories. The participants are ranked according to their accuracy and reproducibility. Analytical performance has improved over time. Comparison of methods has enabled us to identify problems in the determination of several analytes, especially serum aluminum and urine mercury. The accurate determination of trace metals in human biological fluids is not a trivial task. Many problems confront the analyst among which are contamination during the sampling and analytical process, and the complexity of the biological matrix. The absence of standardized methodology adds to the burden of the analyst who must validate his chosen method and ensure its reliability over time. Within the laboratory, it is feasible to verify the reproducibility of results generated by a given analytical technique, simply by repetitive analysis of the same sample (e.g., a patient sample) over the course of time. The results, obtained within a preset window of tolerance will indicate whether the desired level of precision has been achieved. This however does not address the question of the accuracy of the measurement, for which a representative reference standard of known concentration is needed. Aqueous standard solutions of metals, used to prepare calibration curves,are not suitable since they do not take into account matrix effects. Available reference materials, such as NIST (US National Institute of Standards and Technology) bovine liver are better but still have drawbacks, e.g., the concentration of the analyte is not necessarily in the desired range and the matrix is not identical, necessitating a different analytical procedure. One possible solution is an interlaboratory comparison program in which participants analyze aliquots of the same representative sample. The results are used to estimate the true concentration or target value. By comparing their own results with the target value, participants can then assess their accuracy and precision. We have operated such a program for several toxic metals in human blood and urine over the course of the past eleven years. The data generated by participants have enabled us to identify problems in the determination of serum aluminum and urine mercury and thus to suggest alternative techniques. 0097-6156/91Λ)445-0120$06.00Α) © 1991 American Chemical Society In Biological Trace Element Research; Subramanian, K., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1991.

10. WEBER

Accuracy and Precision of Trace Metal Déterminations

121

P R O G R A M DESCRIPTION This program has been described elsewhere (1). It is based on the analysis of control samples by participants. A s shown in Fig. 1, the number of participants has increased steadily over the course of time to the present level of more than 120 laboratories. The international composition of the program is illustrated in Fig. 2.

Downloaded by UNIV MASSACHUSETTS AMHERST on August 2, 2012 | http://pubs.acs.org Publication Date: December 26, 1991 | doi: 10.1021/bk-1991-0445.ch010

Available analytes and matrices include lead and cadmium in blood; aluminum, copper, zinc and selenium in serum; and arsenic, cadmium, mercury, chromium and fluoride in urine. Control samples are prepared either by pooling material obtained from exposed persons, or if temporarily unavailable, by adding a known quantity of the trace metal to normal human blood or urine. These samples are thus very similar to the real samples. Blood and serum specimens testing positive for H I V and Australia antigen are rejected prior to pooling. Each sample is then divided into aliquots (usually about 5 m l for blood and 20 m l for urine). The aliquots are sent to participants who are thus allowed five weeks to analyze the sample and report the results to us. Laboratories are identified only by a code number in order to ensure confidentiality. Each of these six annual runs includes three samples per analyte/matrix pair. Several samples are sent in duplicate over the course of a year in order to evaluate long-term reproducibility. W e compile results, and perform statistical calculations including determination of mean, standard deviation and median for all results and according to the analytical method used.

140 η 120-

Ώ ο Φ

10080-

Ώ

60 H

Ε C

40-j «

.ι

.Ι

Ι

200

JJ—U—U—U—U—U—U—U—U—L— 80

81

82

83

84 85 year

86

87

88

89

Figure 1. Participation in the program as a function of time (for all analytes and matrices).

In Biological Trace Element Research; Subramanian, K., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1991.

Downloaded by UNIV MASSACHUSETTS AMHERST on August 2, 2012 | http://pubs.acs.org Publication Date: December 26, 1991 | doi: 10.1021/bk-1991-0445.ch010

122

BIOLOGICAL TRACE E L E M E N T RESEARCH

Figure 2. Geographical location of participants (for all analytes and matrices)

In Biological Trace Element Research; Subramanian, K., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1991.

10· WEBER

Accuracy and Precision of Trace Metal Determinations

123

Downloaded by UNIV MASSACHUSETTS AMHERST on August 2, 2012 | http://pubs.acs.org Publication Date: December 26, 1991 | doi: 10.1021/bk-1991-0445.ch010

A target value is set, based on the results obtained by a subgroup of "reference" laboratories identified by their past reliable performance. Limits of acceptability are chosen at a high and low level for each analyte using criteria adapted from those developed by Yeoman [2] and Taylor and Briggs [3]. "Outer" limits are based on what is necessary for clinical purposes, and what can be achieved using currently available techniques (e.g., for blood lead ± 0.15 μ π ι ο ΐ / ΐ at 0.5 μ π ι ο ΐ / ΐ and ± 0.3 μ π ι ο ΐ / ΐ at 3.0 μ ι η ο ΐ / ΐ ) . Limits at intermediate concentrations are obtained by linear interpolation. M o r e stringent "inner" limits allow discrimination between highly and moderately proficient laboratories. These criteria are shown in Table I. A report is sent back to the participants within two weeks. The performance of each participant is evaluated annually using a scoring system which takes into account both accuracy and reproducibility. For accuracy, each result within the outer limits scores 1 point. A n additional point is awarded if the result is also within the inner limits. Samples sent in duplicate are used to evaluate reproducibility. For each pair, the difference between the two results is used to compute a "reproducibility score", using the same criteria. A n individual performance summary is prepared for each participant. The data gathered during the course of the program are also used to evaluate the various analytical methods.

W e have previously reported our findings for blood lead and cadmium, and urine arsenic and cadmium [ l ^ ] . Performance trends for aluminum in serum and mercury in urine are discussed below. ~* A L U M I N U M IN S E R U M The performance for aluminum in serum was examined over a three-year period (1985-1987). A s shown in Table 1, only a few participants were able to attain a reasonable proficiency level (e.g. for 1987, only 26 % of laboratories scored over 75 %). In an attempt to shed some light on the causes of this situation, we sent a questionnaire to all laboratories participating for serum aluminum asking for detailed information on analytical methodology and laboratory practices, with the following topics being addressed : sample preparation procedures (sample volume used, use of digestion, complexation, deproteinization, dilution, etc.); labware used (glass or plastic, acid wash of material prior to use, etc.); instrumental techniques (type, instrument model, specific operation conditions); calibration (type of calibration such as aqueous standards, matrix-matched standards, standard additions and the use of standard reference materials; experience in using the method (number of years, samples/year). The data obtained were compared with the observed performance to identify parameters linked to analytical proficiency. Performance for the year 1987 was quantified using the overall score obtained by each laboratory. Participants were then assigned to one of three groups as shown below: Performance Category good average poor

Overall 1987 Score > 75 % 50 - 75 % < 50 %

There were many common features in the analytical approach used by participants: all used graphite furnace atomic absorption spectrophotometry with 75 % preferring PerkinElmer instrumentation; a majority of participants used background correction (89 % , of which 56 % preferred the Zeeman system and 33 % chose the deuterium system), and pyrocoated graphite tubes (85 % , of which 60 % also used the pyrocoated platform); sample preparation was kept to a minimum, consisting mainly of dilution with water.

In Biological Trace Element Research; Subramanian, K., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1991.

124

BIOLOGICAL TRACE ELEMENT RESEARCH

Downloaded by UNIV MASSACHUSETTS AMHERST on August 2, 2012 | http://pubs.acs.org Publication Date: December 26, 1991 | doi: 10.1021/bk-1991-0445.ch010

Table I. Acceptability Criteria

Substance

Target Value

Inner limit

Outer limit

Blood Lead (μιηοΐ/ΐ)

0.5 3.0

± 0.05 ± 0.15

± 0.15 ± 0.30

Urine Arsenic (μπιοΐ/ΐ)

0.7 3.3

± 0.1 ± 0.3

± 0.3 ± 0.5

Urine Mercury (nmol/1)

50 1250

± 7 ± 125

± 15 ± 250

Urine Fluoride (μπιοΐ/ΐ)

50 525

± 5 ± 25

± 10 ± 50

Blood Cadmium (nmol/1)

20 220

± 4.5 ± 9

± 9 ± 18

Urine cadmium (nmol/1)

20 220

± 4.5 ± 9

± 9 ± 18

Serum Aluminum (μιηοΐ/ΐ)

1.0 4.0

± 0.2 ± 0.4

± 0.4 ± 0.8

Urine Chromium (nmol/1)

40 400

± 8 ± 18

± 15 ± 36

Serum Selenium (μπιοΐ/ΐ)

0.75 2.00

± 0.06 ± 0.10

± 0.12 ± 0.20

Serum Copper (μιηοΐ/ΐ)

4.0 20.0

± 0.5 ± 0.8

± 1.0 ± 1.5

Serum Z i n c (μιηοΐ/ΐ)

4.0 20.0

± 0.5 ± 0.8

± 1.0 ± 1.5

Notes :

1) 2)

Limits at other concentrations are obtained by linear interpolation. Results within the outer limits are considered acceptable.

In Biological Trace Element Research; Subramanian, K., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1991.

10. WEBER

Accuracy and Precision of Trace Metal Determinations

125

However, differences in methodology existed in the following areas : acid rinse of laboratory material (wash all containers and pipettes, wash only micropipette tips, wash only containers, wash nothing); the dilution factor varied between 2 and 10-fold; the calibration methods varied among aqueous standards (30 % of participants), matrix-matched standards (26 %) and the method of standard additions (44 % ) ; use of reference material (60 % of participants); variations in furnace temperature program (times and temperatures) were numerous; participants' experience in using the technique varied between 6 months and 5 years.

Downloaded by UNIV MASSACHUSETTS AMHERST on August 2, 2012 | http://pubs.acs.org Publication Date: December 26, 1991 | doi: 10.1021/bk-1991-0445.ch010

Major differences in analytical technique were found between "good" laboratories and others in the following areas: P R E W A S H I N G P R O C E D U R E : "Good" performers tended to either acid-wash both pipette tips and containers (50%), or do no prewashing (25%). In contrast "poor" performers always carried out at least a partial wash of labware, especially containers. This would seem to indicate that prewashing with a dilute acid solution can actually do more harm than good. Whether this is due to contamination of the acid used for cleaning or simply a consequence of additional handling is unclear. However, even somewhat contaminated acid would not contaminate containers as long as the subsequent water rinse was done with pure water. It would probably be a better idea to determine the necessity of cleaning the labware by running blanks on a sample from each new lot. C A L I B R A T I O N P R O C E D U R E : Three different calibration methods were used: aqueous standards, matrix-matched standards and standard additions. A s seen in F i g . 3, the "good" performers used all three methods in roughly equal proportion (38, 38, 24% respectively). Surprisingly, all "poor" performers used the method of standard additions. This would suggest that more complicated techniques do not guarantee better results; in fact the opposite seems true. E X P E R I E N C E : The picture for this parameter is more murky. However all "good" performers had been using their analytical method for at least one year. N o pattern could be distinguished for "poor" performers. A certain amount of experience thus seems necessary to achieve a satisfactory degree of proficiency. Experience is however no guarantee of proficiency. For other parameters, no relationship was observed. Methodological differences did not appear to influence performance. These included the type of graphite tube used (pyrocoated or not, with or without platform) and the background correction scheme (deuterium or Zeeman correctors). In this latter case, "good" performers were more frequent users of deuterium (50%) than Zeeman (40%). This probably reflects their greater experience and consequently older instruments which are more likely to be fitted with deuterium correctors. Our general conclusion is that simpler methods tend to yield more accurate and reproducible results. Complexity increases the likelihood of contamination in the case of trace metal determination. It would be a worthwhile exercise for each participant to examine his or her analytical method critically to assess the necessity of each step. O f course such a survey can only take into account objective parameters, and may not explain why two laboratories using apparently identical methods obtain divergent results. Conclusions must therefore be made in the light of this reality. M E R C U R Y IN U R I N E Traditionally, the determination of mercury in urine has been performed by cold vapor atomic absorption spectrophotometry ( C V A A S ) . The sample is usually pretreated by acid digestion, in order to generate H g (II) which is then reduced to its elemental form using an appropriate reductant (stannous chloride or sodium borohydride). However, it had been reported as early as 1977 (5) that the digestion step was unnecessary. During the past few years, participants in the program have increasingly switched to a non-digestion cold vapor technique as shown in Fig. 4.

In Biological Trace Element Research; Subramanian, K., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1991.

126

BIOLOGICAL TRACE ELEMENT RESEARCH

100"

60 H

2

60-

Downloaded by UNIV MASSACHUSETTS AMHERST on August 2, 2012 | http://pubs.acs.org Publication Date: December 26, 1991 | doi: 10.1021/bk-1991-0445.ch010

Ο Ο

40H

20 λ

aqueous

matrixmatched

standard addition

Figure 3. Serum Aluminum : Calibration methods used by participants, according to their performance category ( • : overall, ο : "good" performers, • : "poor" performers) 40

30

D

α 'ϋ

Έ 20

φ Δ Ε

10

87

88 year

89

Figure 4. Urine Mercury : Analytical methods used by participants as a function of time. ( • : cold vapor with digestion, Θ: cold vapor without digestion, • : gold film analyzer)

In Biological Trace Element Research; Subramanian, K., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1991.

10. WEBER

Accuracy and Precision of Trace Metal Determinations

Recently, a new instrumental technique, the gold film analyzer (Jerome Instruments, Jerome, A Z ) has emerged on the marketplace. A s in the C V A A S technique, elemental mercury vapor is generated. The mercury is then adsorbed onto a gold film. The electrical conductivity of the resulting amalgam varies with the amount of mercury present.

Downloaded by UNIV MASSACHUSETTS AMHERST on August 2, 2012 | http://pubs.acs.org Publication Date: December 26, 1991 | doi: 10.1021/bk-1991-0445.ch010

We have examined the performance of participants for the three-year period, 19871989, in using C V A A S methods with or without sample digestion. Due to the small number of laboratories using the gold film method, no comparison was made for this technique. A s seen in Fig. 5, participants using the digestion procedure initially scored much higher. The performance of laboratories which did not use a digestion procedure improved gradually. B y the end of 1989, 70% of these laboratories were able to generate acceptable results. The reasons for this improvement remain to be explained. However the necessity of digesting urine samples prior to the reduction of mercury can be questioned. Nevertheless, laboratories using a digestion procedure still performed somewhat better with 80% obtaining acceptable results.

1001

30 87

88 sample chronology

89

Figure 5. Urine Mercury : Observed performance over time according to the analytical method used (— O — ' cold vapor with digestion, — O — ' cold vapor without digestion) CONCLUSION Participation in this interlaboratory comparison program has provided feedback to participants, enabling them, when necessary, to revise their analytical methods and operating procedures. Data obtained over the duration of the program have allowed us to evaluate the different methods used by participants and to suggest modifications in specific instances. For both aluminum in serum and mercury in urine it seems acceptable results can be obtained with only limited sample pretreatment. Additional steps in sample preparation do not necessarily improve performance.

In Biological Trace Element Research; Subramanian, K., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1991.

127

128

BIOLOGICAL TRACE ELEMENT RESEARCH

ACKNOWLEDGMENTS The financial support of the Institut de Recherche en Santé et Sécurité du Travail du Q u é b e c is gratefully acknowledged. I thank A l a i n Beaudet, Suzanne M o r i n , Sergine Lapointe and Jacinthe Larochelle for their able technical assistance and Claudette Biais for her secretarial and administrative work.

Literature Cited

Downloaded by UNIV MASSACHUSETTS AMHERST on August 2, 2012 | http://pubs.acs.org Publication Date: December 26, 1991 | doi: 10.1021/bk-1991-0445.ch010

1. Weber, J.P. Sci. Total Environ. 1988, 71, 111-23. 2. Yeoman, W.B., In Analytical Techniques for Heavy Metals in Biological Fluids: Occupational and Environmental Commission of the European Communities Joint Research Centre, Ed., ISPRA, Italy. 1980. 3. Taylor, Α.; Briggs, R.J. J. Anal. At. Spectromet. 1986, 1, 391-94. 4. Savoie, J.Y.; Weber, J.P. In Chemical Toxicology and Clinical Chemistry ofMetals;Brown, S.S.; and Savory, J. Eds.; Academic Press : New York, 1983, pp 77-80. 5. Ebbestadt, V.; Gunderson, G.; Torgrimsen, T.A. At. Absorp. Newslett. 1977, 14, 142-43. RECEIVED August 6, 1990

In Biological Trace Element Research; Subramanian, K., et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 1991.