Environ. Sci. Technol. 1994, 28,338-345
Quality Control Level: An Alternative to Detection Levels David E. Klmbrough’ and Janice Wakakuwa
Department of Toxic Substances Control, California Environmental Protection Agency, Hazardous Material Laboratory-Southern California, 1449 West Temple Street, Los Angeles, California 90026-5698 Existing methods for the determination of reporting limits (e.g., MDL, LOD, etc) are based on a binary procedure that determines that lowest concentration of an analyte that is either detected or not detected within specified confidence limits. There is no assessment of the accuracy or precision of the results detected. An alternative procedure is presented, the Quality Control Level which determines the lowest concentration that meets the data quality objects of the data user in term of the minimum acceptable precision and accuracy. To examine this hypothesis, a series of 15 soils and aqueous liquids were prepared with successively smaller concentrations of 16 toxic regulated elements. The range of the concentrations were over four orders of magnitude for both. Each of these liquids and soils were analyzed eight times and the accuracy and precision of each analyte was measured against concentration. This paper will show that it is possible to use a quality control approach to reporting levels.
Introduction Environmental laboratories are faced with the problem of determining the smallest concentration of an analyte that can be reported in a given sample. A number of statisticians and chemists over the last 25 years have developed a number of methods to determine this concentration. The most common approach is to determine the minimum concentration that can be reliably detected. This has been posedas the question: ”What is the smallest number that is not zero within a certain confidence interval?”. In the environmental community, the most widely accepted procedure for answering this question is the Method Detection Limit (MDL) procedure developed by Glazer et al. (I)and adopted by the U.S. EPA (2,3).The usefulness of this procedure has been questioned on both theoretical and practical grounds (4-8). In response, the U S . EPA, ASTM, ACS, AOAC, and others have advanced a new procedure to supplement the MDL procedure, the Reliable Detection Level (RDL) procedure, which is partially based on Glazer et al. and on Clayton et al. (5). What is still not addressed is the quality of the corresponding result. Analytical environmental chemists ought to be concerned with the accuracy and precision of a result. In this paper, we will examine the question of “less than” value determination from the empirical and quality control approach. Our approach will be to empirically determine the relationship between concentration and the quality control parameters of accuracy (bias) and precision. Our hypothesis is that there is a direct correlation between concentration, accuracy, and precision. This being the case, the lowest reportable concentration is that which has acceptable precision and accuracy.
Definitions Percent bias (% bias) is the absolute difference between the mean measured concentration of the replicate deter338
Environ. Sci. Technol., Vol. 26, No. 2, 1994
Table 1. Components of Artificial Soil for Reporting Limit Samples
element
spike material
aluminum iron calcium sodium potassium manganese magnesium silicon
A1203 Fen03 CaO NaKC4H40~4H20 NaKC4H40~4H20 MnSOgH20 MgC1.6H20 Ottawa sand (80mesh) cellulose acetate dextrin see above see above total
carbon oxygen hydrogen
concn in soil mass of mass (g) element (g) (mg/kg) 282 14.3 14.0 7.19 7.19
3.08 8.33 200 239 232
1000
100 10 10
0.58 1.0 1.0 1.0 93 100 100
100 000 10 000 10 OOO
580 lo00 1 000 1 000 93 OOO 200 000
550 000 33 000 1 000 000
minations of a sample and the true concentration of the sample divided by the true concentration and multiplied by 100. This is a measure of accuracy. Percent bias does not distinguish positive from negative bias. Percent Relative Standard Deviation (% RSD) is the measured standard deviation of replicate determinations of the soil divided by the mean measured concentration and multiplied by 100. Instrument Detection Level (IDL) is the smallest concentration of an analytical standard that is interference free and not zero with 99% confidence. This would be indentical to the Glazer et a1.lU.S. EPA MDL procedure using an interference-free analytical standard as the sample matrix.
Experimental Section
Study Design. The most obvious approach is to analyze a single matrix type using a series of successively lower analyte concentrations. Previous work shows that change in precision and accuracy versus concentration can be very abrupt (8). Therefore, any series of materials used for this type of study would need a large number of sample concentrations that varied over a fairly narrow range. No such series of naturally occurring materials can easily be found. To get such a series, it is necessary to spike a suitable base material. Soil is a matrix commonly analyzed for “metals”. (Many of these analytes are often referred to as “metals” or “heavy metals” although they are not technically metals, e.g., arsenic and selenium. In the rest of this paper, these analytes will be referred to as elements.) Preparing a series of spiked soils a t very low concentrations would be difficult, as most elemental analytes of environmental interest are naturally occurring in soils. To get around this, an artificial soil was prepared (see Materials). This artificial soil was then separated into 15 aliquots spiked at 14 different concentrations with one unspiked aliquot. Each aliquot was then digested and 0013-936X/94/0928-0336$04.50/0
0 1994 American Chemical Society
Table 2. Mean Measured Results of Target Elements in Liquid Standards without Baseline Corrections Cd co Ba Be As concn of target elements standard ID Ag
s1
S 13 S 14 S 15
2.00 1.50 1.00 0.50 0.20 0.15 0.10 0.05 0.02 0.015 0.010 0.005 0.002 0.001