A fTlM United States tfVPKA Envlroninerta! Protection !¦£ ,e*A§siicy Procedure for Safe Drinking Water Act Program Detection Limits for Radionuclides Office of Water (MS-140) EPA 815-B-17-003 April 2017 ------- Questions concerning this document should be addressed to: Glynda A. Smith, Ph.D. U.S. EPA, Office of Ground Water and Drinking Water, Standards and Risk Management Division, Technical Support Center, 26 W. Martin Luther King Dr., Cincinnati, OH 45268 Phone:(513)569-7652 smith.glvnda(a)epa.gov Authors Katie Adams, Inorganic Chemistry Technical Lead, USEPA Region 10 Laboratory, 7411 Beach Drive East, Port Orchard, WA 98366 Gerald Dodo, Supervisory Chemist, USEPA Region 10 Laboratory, 7411 Beach Drive East, Port Orchard, WA 98366 Barry V. Pepich, Ph.D., Director, USEPA Region 10 Laboratory, 7411 Beach Drive East, Port Orchard, WA 98366 Glynda A. Smith, Ph.D., U.S. EPA (Cincinnati, OH) Acknowledgements The following served as peer reviewers for the document: John G. Griggs, Ph.D., National Analytical Radiation Environmental Laboratory (NAREL), Office of Radiation and Indoor Air (ORIA), USEPA Keith McCroan, Ph.D., National Analytical Radiation Environmental Laboratory (NAREL), Office of Radiation and Indoor Air (ORIA), USEPA Bahman Parsa, Ph.D., Environmental and Chemical Laboratory Services, PHEL/PHILEP, New Jersey Department of Health Bob Read, Ph.D., Environmental Chemistry Laboratory, Tennessee Department of Health, Division of Laboratory Services Andy Eaton, Ph.D., Eurofins Eaton Analytical, Inc., Monrovia, California ------- TABLE OF CONTENTS Contents 1.0 Introduction 1 1.1 Background and Objectives 1 1.2 Scope and Application 1 2.0 Overview 1 3.0 Calculating Detection Limits for Radiochemical Measurements 2 3.1 Definition of the Detection Limit for SDWA Radiochemical Measurements 2 3.2 Derivation of the SDWA Detection Limit Calculation 2 4.0 Performing Experimental Confirmation of SDWA Detection Limits for Radiochemical Measurements 4 4.1 Experimental SDWA Detection Limit Studies 4 4.2 Statistical Evaluation of Detection Limit Studies 5 5.0 References 5 Appendix A: Example Calculations 6 Table 1. Experimental Values for Seven Spiked Replicates 7 Appendix B: Chi-Square Values at the 99th Percentile 9 Table 2. Chi-Square Values (99th Percentile) 9 Appendix C: Abbreviations and Acronyms 10 ------- 1.0 I xluction 1.1 Background and Objectives When analyzing radionuclides for the drinking water program, it is important to carefully evaluate method performance at the lowest concentrations attainable for the method. Critical water testing and treatment requirements impacting public health are made based on results that are often near the limits of method detection capability. The Code of Federal Regulations (CFR) specifies Required Detection Limits (RDLs) for radionuclides. Laboratories must demonstrate their performance at those levels. Many radiochemistry laboratories are accustomed to using a Minimum Detectable Activity (MDA) to achieve this requirement. The MDA is a calculation that is based on counting precision that is scaled by multipliers to account for such factors as sample volumes, chemical yields, and counting times, which may vary. It is therefore a useful, sample- specific tool. However, MDA equations vary and may or may not account for the variability of the whole system (including, for example, the sample separation steps, which often precede instrument counting). Consequently, the Office of Ground Water and Drinking Water (OGWDW), in administering the National Drinking Water Program, emphasizes the need for laboratories to capably and reproducibly demonstrate system performance through detection limit studies. These experimental studies seek to confirm that the system does, in fact, meet the method performance that can be derived mathematically. Because most radiochemistry methods are based on Poisson distributions rather than Gaussian distributions (as in other chemistry fields), the mechanism of calculating the detection limit for radionuclides differs from that described in 40 CFR 136 Appendix B, which is applied for inorganic and organic analytes. This document provides the derivation of the Safe Drinking Water Act (SDWA) program's radionuclide Detection Limit (herein after referred to as the "SDWA DL"), as well as practical steps for executing the experimental DL study. 1.2 Scope and Application The procedure provided in this document describes the basis for the SDWA DL for radionuclides and provides an example calculation (see Appendix A) intended to assist laboratories conducting the DL determination for the first time. This procedure describes in detail the calculations associated with the radionuclide detection limit that is defined in 40 CFR 141.25(c). The DL procedure is one part of demonstrating method capability. The evaluation and monitoring of laboratory reagent blanks (LRBs) are also required to verify low system background, and method accuracy and precision are demonstrated through the evaluation of laboratory fortified blanks (LFBs). 2.0 Overview As an initial estimate, laboratories should calculate their theoretical ability to meet the DL requirement of 40 CFR 141.25(c). Subsequently, they experimentally verify that their analytical system does actually perform consistently with what has been demonstrated in theory. The experimental verification consists of the analysis of at least seven standards spiked at or near the concentration of the RDL. These standards are taken through the entire analytical process, and the results are evaluated against a Chi- l ------- square ( x2) distribution to determine if the experimental results compare favorably with the expected values. 3.0 1 Ululating Detection Lin-ii t »r Radiochemical Measurements 3.1 Definition of the Detection Limit for 5DWA Radiochemical Measurements The detection capability of radiochemical measurements used for SDWA drinking water compliance monitoring is defined at 40 CFR part 141.25(c) as a detection limit with the following conditions: "The detection limit shall be that concentration which can be counted with a precision of plus or minus 100 percent at the 95 percent confidence level (1.96a, where a is the standard deviation of the net counting rate of the sample)." The SDWA Detection Limit according to this definition differs from other "detection limits/' such as the method detection limit or MDL (defined in 40 CFR part 136, Appendix B), and the Minimum Detectable Activity (MDA), which is commonly used by radiochemists. The RDLs for SDWA drinking water compliance monitoring of radionuclides are expressed in terms of the definition given in 40 CFR 141.25(c). For measurements involving simple nuclear counting with Poisson counting statistics, the procedure given in Section 3.2 below is used to obtain a preliminary estimate of the SDWA DL. 3.2 Derivation of the SDWA Detection Limit Calculation The definition of the SDWA DL may be expressed mathematically as follows: RDl = 1-96 x aDL (1) Where: Rdl is the mean net count rate for a sample with concentration at the detection limit oDL is the standard deviation of the net count rate The relationship for the standard deviation of a radiochemical measurement is centered around the fact the gross rate has a background rate subtracted from it to derive a net count rate: Rdl = Rg ~ Rb (2) Where: Rq is the mean gross count rate for a sample (with concentration at the DL) Rb is the mean background count rate for a sample measurement However, each count rate is a calculated quantity as specified below: RG = Cf and RB = C-f- (3) Where: Rq is the mean gross count rate for a sample (with concentration at the detection limit) Rb is the mean background count rate for a sample measurement CG is the mean total (gross) sample count CB is the mean total background count 2 ------- tG is the time of the measurement used to accumulate the sample count tB is the time of the measurement used to accumulate the background count The standard deviation of a count rate is proportional to the square root of the mean of a measurement. Assuming Poisson counting statistics, the standard deviations of the measured values of RG and RB are given by: oG=1—=— and oB=*—= — (4) rG y rG y Where: oG is the standard deviation of the measured gross count rate oB is the standard deviation of the measured background count rate Since the net count rate, RDL, is the difference between Rc and RB, its standard deviation is given by: °dl= JOg+(Tb) (5) Where: oDL is the standard deviation of the net count rate Combining equations (4) and (5), one arrives at: + (6) Substituting equation (6) into equation (1), one arrives at: R»' = 1-96xJ(t+S) (7) Equation (2) may now be used to eliminate the variable, RG, from the equation. Since RG = RDL + RB, equation (7) may be rewritten as: Rn, = 1.96 X "L " + (Rdl+Rb i rb "i) <8> ,e ¦ Equation (8) is then solved algebraically for the value of RDL. First, rewrite the radicand: (v+R»x£+i)) (9) RDl — 1-96 x M Squaring each side of the equation, one arrives at: S1 tG " \tG tBj Collecting all items on the left-hand side to put the equation in standard quadratic form, one arrives at: Rdl =^xRDL + 1.962RBx{^ + f) (10) p2 1^96^ KDL xRDL- 1.962RBx(j- + f) = 0 (11) 2 tG D \tG tB The quadratic formula gives two solutions to equation (11), one of which is positive and one of which is negative. The positive solution is required and it is given by the following equation: 3 ------- x 1+J1+SxR»x(t + i) (12) J Equation (12) provides a reasonable estimate of the count rate at the DL for the net activity that is based on counting statistics alone. This count rate is then divided by the product of the experimental factors, H, which can include the following items: the method of detection's counting efficiency, the sample volume, chemical recoveries (measured by gravimetric or tracer techniques), conversion factors to picocuries, etc. The result is used to derive a specific DL of the radioanalyte of interest for a radiochemical method of analysis that is used for SDWA compliance monitoring: Where: H is the product of the experimental factors (see example calculations in Appendix A) SDWA DL is the SDWA Detection Limit This SDWA DL is mathematically equivalent to the detection limit specified in 40 CFR part 141.25(c). It is expected that the experimental factors will vary with specific method and sample conditions. If an estimate of the SDWA DL described in equation (13) does not exceed the required DL, a DL study is performed as described below to verify that laboratory performance in practice can be demonstrated prior to analyzing drinking water samples for compliance. However, if the estimate of the DL exceeds the required DL, the performance will be considered inadequate and there will be little value in completing the experimental DL study. Conditions would need to be adjusted to meet the required DL before proceeding to confirm the DL experimentally. This may entail using a larger sample volume or longer sample counting time. NOTE: Typical drinking water compliance samples will have very low activity levels and compliance samples should be run under the same conditions as those used to confirm the DL. 4.0 Performing Exp* i if i a i ii. -1» «. >i,nrmation of - II11 II >i > ctlion Limits for Radiochemical Measurements 4.1 Experimental SDWA Detection Limit Studies The experimental SDWA DL study will verify that the method is capable of routinely achieving the required detection capability. The experimental SDWA DL study consists of seven replicate samples. Each sample is prepared with ASTM II grade reagent water, or other blank matrix as appropriate for the method, and using the sample volume described in the method. For example, gross alpha analyses are highly dependent on the total dissolved solids content in the sample matrix. Reagent water can yield artificially low DLs due to higher detector efficiencies. Thus, more realistic gross alpha DLs will be obtained using either laboratory tap water or a synthetic water solids matrix to prepare the DL study samples. Each DL study sample is spiked with NIST traceable source(s) of the method target radionuclide(s) to an activity concentration at or near their RDL. The sample is mixed and then processed through sample preparation, processing and analysis per the test method. The measurements of the DL study samples are then assessed by calculating a precision statistic. SDWA DL = — H (13) 4 ------- 4.2 Statistical Evaluation of Detection Limit Studies The assessment of the replicate results for each radionuclide uses a chi-square statistic to test whether the relative standard deviation of the results exceeds the maximum value allowed at the RDL. Where: n is the number of replicate measurements (> 7) ju is the spike concentration (at or near the RDL) Xj is the result of the fh replicate measurement (j = 1,2,..., n) To be deemed acceptable, the value of x2 must be less than or equal to the 99th percentile of the x2 distribution with (n-1) degrees of freedom. When n = 7, the value of this percentile is 16.812. NOTE: Refer to Appendix A - Example Calculations. Refer to Appendix B for a table of Chi-square values. 5.0 References 1. 40 CFR 141: National Primary Drinking Water Regulations 2. ASTM D1193-99E01: Standard Specifications for Reagent Water. American Society for Testing and Materials. March 1999, with editorial change made in October 2001. 3. MARLAP 2004. Multi-Agency Radiological Laboratory Analytical Protocols Manual. NUREG-1576, EPA 402-B-04-001C. 4. Chapter VI, Critical Elements for Radiochemistry. The Manual for the Certification of Laboratories Analyzing Drinking Water. (EPA/815-R-05-004). Calculate the mean, X, and a chi-square statistic, x2, as follows: 71 y=i 5 ------- Appendix A: Examp culations The following section provides example calculations for the estimation and experimental confirmation of the SDWA Detection Limit for radionuclide activity. The example uses gross alpha results obtained using EPA Method 900.0. The data was generated by the New Jersey Department of Health (NJDOH) Radioanalytical Services Laboratory, and is used with their permission1. 1.0 Example Detection Limit Calculation Equations (12) and (13) in Section 3.2 state: 1.962 RDL = —o X °L 2 tr. 1 + 4 tl /II 1 + , rZl X RB X b — 1.962 B \tG tB And SDWA DL = R DL H Combining these equations and considering the experimental factors relevant for gross alpha determination, the following equation is obtained: DL c!"Ci/L) = X Ru X (Efficiency) (Volume) (Chemical Recovery) (2.22) Where: Rb is the mean background count rate for a sample measurement tG is the time of the measurement used to accumulate the sample count tB is the time of the measurement used to accumulate the background count 2.22 is the conversion factor from dpm to pCi For this DL study, gross alpha recovery is assumed to be 100%. RB = 0.03 cpm, Volume = 1.0 L, and tG =tB = 200 minutes. The detection efficiency was 0.177 cpm/dpm. Substituting these values into the equation produces the following: DL (PCi/L) = 1.96 (2 X 200) x 1 + . 1 + 4(200)' 1.962 x 0.03 x (200 + 20o) (0.177)(1)(1)(2.22) r\ r -1 r\— 3 l+Vl + 12.5 = 9.6x10 d x 0.393 = 2.44xl0-2 X 4.7 = 0.11 pCi/L The Required Detection Limit (RDL) for gross alpha is 3 pCi/L. Because 0.11 pCi/L is a smaller quantity than 3 pCi/L, it is theoretically true that the counting times, volumes, and efficiencies assumed for this example would lead to acceptable precision at the RDL concentration. 2.0 Example Experimental SDWA Detection Limit Study 6 ------- The instructions for performing an experimental SDWA DL study are given in Sections 4.1 and 4.2. The following example illustrates how the evaluation criteria are applied. 1, Experimental Values for Seven Spiked Replicates Replicates Measured Gross Alpha (Th-230) Activity (pCi/L) Spike Amount (pCi/L) BS 1 2.89 + 0.30 3.0 BS 2 5.51 + 0.45 3.2 BS 3 2.88 + 0.31 3.3 BS 4 3.72 + 0.36 3.2 BS 5 3.42 + 0.34 3.0 BS 6 3.11 + 0.32 3.1 BS 7 3.17 + 0.32 3.1 Average: 3.53 3.13 The mean gross alpha activity is calculated using the equation: n y=i Substituting the data, this produces: X = ^(2.89 + 5.51 + 2.88 + 3.72 + 3.42 + 3.11 + 3.17) = 3.53 pCi/L The Chi-square statistic is calculated using the equation: „ n 1.962\~> 2 x2=—2Sx<~x) 7 = 1 Where: n is the number of replicate measurements (7) ju is the spike concentration (at or near the RDL; in this case 3.13 pCi/L) Xj is the result of the jth replicate measurement Substituting the data, this produces: 1 962 X2 = ——7 X [(2.89 - 3.53)2 + (5.51 - 3.53)2 + (2.88 - 3.53)2 + (3.72 - 3.53)2 3.13/ + (3.42 - 3.53)2 + (3.11 - 3.53)2 + (3.17 - 3.53)2] 3.84 ^ -v = — x (5.1) 9.8 v J = 2.0 This data set has seven replicates and thus, six degrees of freedom. So, the critical value for the statistic is the 99th percentile of the x2 distribution with six degrees of freedom, which equals 7 ------- 16.812. (see Chi-Square Table provided in Appendix B). Since the calculated x2 value of 2.0 does not exceed 16.812, the method passes the experimental DL study. 1. Detection Limit Study, Gross Alpha, Evaporation, EPA Method 900.0. Dr. Bahman Parsa, NJDOH Laboratory, 3 Schwarzkopf Drive, West Trenton, NJ 08628. June 14, 2011. 8 ------- Appeimdi > V ». Li-Sqi>»[ * * ''lne* jf 111. 33th Percentile Table 2. Chi-Square Values (99th Percentile) Degrees of Freedom *2 1 6.635 2 9.210 3 11.345 4 13.277 5 15.086 6 16.812 7 18.475 8 20.090 9 21.666 10 23.209 11 24.725 12 26.217 13 27.688 14 29.141 15 30.578 16 32.000 17 33.409 18 34.805 19 36.191 20 37.566 9 ------- Appenfii i -hbreviation Acronyms ASTM ASTM International CFR Code of Federal Regulations DL Detection Limit EPA U.S. Environmental Protection Agency MARLAP Multi-Agency Radiological Laboratory Analytical Protocols Manual MDA Minimum Detectable Activity NJDOH New Jersey Department of Health NIST National Institute of Standards and Technology OGWDW Office of Groundwater and Drinking Water RB Reagent Blank RDL Required Detection Limit SDWA Safe Drinking Water Act 10 ------- |