Validation of cuvette tests for drinking water analysis

Transcription

1 PRACTICE REPORT LABORATORY ANALYSIS CUVETTE TESTS DRINKING WATER Validation of cuvette tests for drinking water analysis In 2007 the IWW (Rheinisch-Westfälische Institut für Wasser Beratungs- und Entwicklungsgesellschaft mbh) was commissioned to carry out a study to determine the formal permissibility of the use of cuvette tests for the analysis of drinking water in conformity with the 2001 Drinking Water Ordinance (Trinkwasserverordnung 2001). The study concluded that: the precision, trueness and limit of detection of the studied LANGE cuvette tests correspond to the requirements of the Drinking Water Ordinance (DWO) concerning analysis methods. They are therefore suitable in principle for official analyses of drinking water. Author: Ralf König - Diplom-Chemiker - Vertical Market Manager HACH LANGE

2 2 DRINKING WATER ANALYSIS_REQUIREMENTS Do your cuvette tests meet drinking water legislation? Accreditation In principle, non-standardised methods can be included in the accreditation, if they have been proven to be suitable for the user and have been validated accordingly. The characteristics of the analysis method must satisfy the requirements specified in Annex 5 of the Drinking Water Ordinance. As well as the basic proof of the suitability of the methods (e.g. through this study), a laboratory must also provide the simplified proof that it satisfies the requirements with regard to the characteristics. The test method must be approved by the accreditation body as part of the accreditation process (ISO 17025) and must be included in the certificate (or the annex to the certificate). In this context, proof of the fitness of the method for the intended purpose must be provided (in accordance with section 5.4 of ISO 17025). The necessary validation work must be carried out by the laboratory and if necessary, the manufacturer of the method. The laboratory carries out the stipulated internal (control cards) and external quality assurance (successful participation in round robin tests) for the method. Finally, compliance with the above mentioned requirements must be formally confirmed or recognised by the notification body in accordance with section 15 (5) of the Drinking Water Ordinance. If these criteria are satisfied, there is no legal impediment to the inclusion of cuvette tests in the laboratory s internal method portfolio. Background In Germany, official analyses of drinking water are regulated by law. They may only be carried out by specially authorised, notified laboratories, using methods that are specified in conformity with the Drinking Water Ordinance. As the German Drinking Water Ordinance of 2001 almost completely embodies the content of the European Drinking Water Directive, most of its conditions and requirements apply in similar form in the other Member States of the EU. In contrast to the situation with regard to wastewater, the drinking water legislation specifies no concrete methodological requirements. In principle, any method that yields sufficiently reliable results and whose performance characteristics conform to those defined in Annex 5, section 15 of the Drinking Water Ordinance, can be used. The performance characteristics (PC) are trueness, precision and limit of detection, which are expressed as a percentage (in this case 10 %) of the limit value. Another, no less important authorisation criterion is the requirement that the alternative method must conform to the generally acknowledged technical standards. Also, a laboratory that wants to use cuvette tests as alternatives to standardised methods must possess a valid accreditation in conformity with DIN ISO Requirements concerning accreditation If no method is specified by the client, an analysis laboratory is obliged to use the most appropriate analysis methods [3]. These appropriate methods must have been published either in international, regional or national standards or by reputable technical organisations, or in relevant scientific journals, or have been clearly described by the manufacturer of the methods (e.g. cuvette tests). Methods developed or adopted by the laboratory may also be used if they are suitable for the intended use and appropriately validated. The client must be informed about the method chosen. Basic validation of the cuvette tests A test method is validated by examining its: - measurement uncertainty - limit of detection and limit of quantification - selectivity - linearity - precision - trueness - ruggedness against external influences (cross-sensitivities, influences in the sample matrix) Furthermore, the alternative method should give results that are sufficiently comparable to those obtained using the corresponding standard method.

3 3 All measurements were carried out with a HACH LANGE DR 5000 UV-VIS spectrophotometer (see Fig. 1). The basic calibration was carried out using calibration solutions prepared by fortifying matrix-free solutions (ultrapure water) with the analytes of interest [1]. Certified reference materials were used to fortify the analytes. The limit of quantification was determined by measuring ten calibration solutions with equidistant concentration levels starting from the lowest concentration of the measuring range of the cuvette test over 10 days. The total measuring range was validated by measuring from two to five additional calibration solutions up to the highest concentration of the measuring range. For the validation of the cuvette tests, the basic calibration data were evaluated and used to calculate the following method specifications. Calibration function and performance characteristics, - Gradient (b, measure of sensitivity), - Intercept - Standard deviation of the residuals (s y, scatter of the measured values around the line of regression) - Method standard deviation (s x0, absolute measure of the precision of the calibration) - Method variation coefficient (V x0, relative method standard deviation, relative measure of the precision of the calibration) Verification of the linearity and Determination of the limits of detection and quantification in accordance with DIN In addition, the homogeneity of the variances over the measuring range was examined. The performance characteristics were calculated with Software for the statistical control of analytical data SQS (Version 1.51, Perkin Elmer, Rodgau-Jügesheim). Fig. 1: DR 5000 spectrophotometer with touchscreen, rotational barcode measurement and automatic error detection for reliable evaluation of all LANGE cuvette tests No. Parameter LANGE cuvette test Standard Measuring range of DWO limit value PC (%) PC (mg/l) cuvette test 1 Aluminium LCK301 DIN ISO mg/l 0.2 mg/l (Chromazurol S) (Pyrocatechol violet) 2 Ammonium LCK304 DIN mg/l 0.5 mg/l (Indophenol blue) (Indophenol blue) 3 Chlorine, free LCK310 DIN EN ISO mg/l min. 0.1 mg/l, - - (after disinfection) (DPD) (DPD) max. 0.3 mg/l 4 Chlorine, total LCK310 DIN EN ISO mg/l (DPD) (DPD) 5 Iron, trace LCK521 DIN E mg/l 0.2 mg/l (1,10-Phenanthroline) (1,10-Phenanthroline) 6 Manganese, trace LCW532 DIN E mg/l 0.05 mg/l (PAN indicator) (Formaldoxim) 7 Nitrate LCK339 DIN38405 D mg/l 50 mg/l 10 5 (2,6-Dimethylphenol) (2,6-Dimethylphenol) 8 Nitrite, trace LCK541 (Diazotisation) DIN EN (Diazotisation) mg/l 0.5 mg/l, or 0.1 mg/l at water treatment plant outlet Ortho-phosphate LCK349 (Molybdenum blue) DIN EN ISO 6878 (Molybdenum blue) mg/l 6.7 mg/l - - Table 1: Methodology, limit values and performance characteristics (PC) of the selected analysis parameters in conformity with the 2001 Drinking Water Ordinance (DWO)

4 4 DRINKING WATER ANALYSIS_VALIDATION Validation of cuvette tests: precision Fig. 2: 5 cm rectangular cuvette for the precise determination of trace amounts of iron, manganese and nitrite V x0 (%) 3 2,5 2 1,5 1 0,5 0 LCK301 Aluminium Method variation coefficients of cuvette tests Precision Evaluating the precision of the method using the method variation coefficient The method variation coefficient (relative method standard deviation, V x0 ) is used to evaluate the precision of a calibration. The method variation coefficient is a measure of the quality of a test method. Furthermore, it enables analysis methods with different measuring ranges to be compared. The method variation coefficients of the examined cuvette tests (Fig. 3) are all below 3 %. The methods therefore exhibit good to very good precision. Especially noteworthy are the cuvette tests for ammonium, iron, nitrate and phosphate, whose method variation coefficients are clearly below 1 %, indicating that they are very precise indeed. 2,2 2,21 2,13 1,84 LCK304 Ammonium 0,42 LCK521 Iron, trace 0,70 1,19 LCW532 Manganese, trace LCK339 Nitrate LCK541 Nitrite, trace LCK349 ortho-phosphate LCK310 Chlorine, free LCK310 Chlorine, total Fig. 3: Method variation coefficients (V x0 ) of the LANGE cuvette tests (basic calibration) 0,51 0,74 Analysis precision at the limit value ( 10 %) In analytical chemistry, the precision of a method depends on the concentration of the analytes. It decreases as the lower limit of the measuring range is approached. Under the drinking water legislation, the results of authorised measurement methods must not deviate by more than 10 % from the limit values specified in the legislation. The performance characteristics specified for precision in the Drinking Water Ordinance relate to the concentration of the parameter at the limit value. This is intended to ensure that the analysis method employed is at least suitable for detecting the concentration of the analyte reliably at the limit value. The analytical precision of the cuvette tests was calculated with Software for the statistical control of analytical data SQS (Version 1.51, Perkin Elmer, Rodgau-Jügesheim) and the data of the basic validation. All validated cuvette tests for which performance characteristics are specified in the Drinking Water Ordinance (aluminium, ammonium, iron, manganese, nitrite and nitrate) have a high level of analytical precision at the limit value. They all give results that do not exceed the maximum deviation of 10 %. In the case of ammonium, aluminium, iron and nitrate, the scatter of the results is less than 3 %. The Drinking Water Ordinance defines no performance characteristics for the parameters chlorine (free/total) and ortho-phosphate.

5 5 Precision under repeatability conditions Another quality objective of routine analysis is to ensure the constant precision and trueness of the analysis results over a long period of time. The precision of the method was examined and assessed over a period of 2 to 3 weeks by carrying out systematically repeated control analyses (ten in total, with double determination) in fortified real drinking water samples and on different measurement days,. For this purpose the within series standard deviation (double determination per measurement day) and the between series standard deviation were calculated, as was the total standard deviation in accordance with Funk et al. [5]. Under repeated conditions, all the examined cuvette tests exhibit good to very good precision. The analysis results were not found to vary as a function of time. Ruggedness: The influence of different reagent batches on precision Ruggedness is a measure of the relative insensitiveness of an analysis method to changes in the general analytical conditions. The ruggedness of the cuvette tests was examined using reagents from different production batches. To do this, triple determinations of a standard solution in a matrix-free solution (ultrapure water) were carried out under repeatability conditions with cuvette tests from two different production batches. The mean values of the two measurement series (in each case N=3) with different reagent batches differ only slightly. For all examined cuvette tests, the relative standard deviation over all measured values (N=6), as a measure of scatter, is good to very good. For all methods, the relative standard deviation is below 3 %. The reagent batch has no detectable influence on precision. Trueness Determination of the recovery rates Trueness is a measure of the deviation of the measured value (or the mean of several measured values) from the correct (true) value due to a systematic error. To test the trueness of the cuvette tests, the analytes were fortified with certified reference materials to create real drinking water matrices [1]. The fortified test solutions were analysed six times under repeatability conditions using the cuvette tests. The trueness was determined via the recovery rate after the measured values had been checked for outliers (Grubbs method, P=99 %). The recovery rate corresponds to the mean of the measured values (N=6) expressed as a percentage of the target concentration of the fortified analytes. Fig. 4: Cuvette test LCK310 Chlorine (free and total) Fig. 5: Cuvette test LCK349 ortho-phosphate

6 6 DRINKING WATER ANALYSIS_VALIDATION Validation of cuvette tests: trueness The recovery rates of the examined cuvette tests lie between 92.5 and 102 %. over the total range of matrices were achieved with the cuvette tests for iron and ortho-phosphate ( %). Fig. 6: Ideal for determining low concentrations of manganese LCW532 LCK/Parameter Recovery rate (%) LCK301 Aluminium 91.8 LCK304 Ammonium 98.2 LCK310 Chlorine, free 94.9 LCK310 Chlorine, total 91.5 LCK521 Iron, trace 99.2 LCW532 Manganese, trace 98.0 LCK339 Nitrate LCK541 Nitrite, trace 91.2 LCK349 ortho-phosphate 97.6 Table 2: Recovery rate of the LANGE cuvette tests in ultrapure water They therefore meet the requirements of the Drinking Water Ordinance with regard to the required trueness (10 % deviation relative to the limit value). Recovery rates in different matrices One of the key quality criteria of a test method is its suitability for use with real samples. Matrix effects and individual process steps can express themselves in increased imprecision and/or as constant or proportional systematic deviation of the analysis results from the true values. Calculation of the recovery function in a special drinking water matrix is a suitable way to determine a matrix effect. As a matrix, real drinking water was fortified with hypochlorite (0.8 mg/l free Cl 2 ), calcium (365 mg/l CaCO 3 = 20 dh), iron (2 mg/l Fe) and manganese (1 mg/l Mn) standard solutions. For each cuvette test, a recovery function was generated over the total measuring range using the four drinking water matrices. The recovery rates of all examined cuvette tests were between 90 and 100 % and meet the requirements of the Drinking Water Ordinance of 2001 with regard to the performance characteristic trueness. The aluminium cuvette test, which gave a recovery rate of only 85.1 % in the manganese matrix, formed an exception. The highest recovery rates The recovery rates for all methods are good to very good, depending on the matrix. Checking the limit of quantification (in accordance with DIN 32645) Annex 5 of the Drinking Water Ordinance defines performance characteristics for the limit of detection as a percentage of the limit value. The limit of detection is the lowest concentration of an analyte that can be determined qualitatively, while the limit of quantification is the lowest concentration of an analyte that can be determined quantitatively. The limit of quantification is the lowest concentration at which the measurement satisfies a specified precision requirement. As drinking water analyses involve determining the concentration of the analytes quantitatively, the limit of quantification rather than the limit of detection was used for the purposes of this validation, to assess the performance of the test method. The limit of quantification was determined in accordance with DIN (calibration line method) by measuring 10 standard calibration solutions with equidistant concentrations, starting at the lowest concentration of the measuring range of the HACH LANGE test method, over a period of ten days.

7 7 The cuvette tests for determining aluminium, ammonium, iron (trace), manganese (trace), nitrate and nitrite (trace) satisfy the performance characteristics specification for the limit of detection. For the parameters chlorine (free and total) and ortho-phosphate, no performance characteristics are specified in the Drinking Water Ordinance of The limits of quantification of these methods are in the range achieved using the corresponding standard photometric methods. Comparison of the cuvette tests with standard photometric methods This study is based on a comparison of the performance characteristics of both methods (cuvette tests v. standard photometric methods). The performance characteristics of the photometric test methods were determined separately within the framework of the study [1]. In addition, for the purpose of assessing the comparability of analysis results from real samples, 6-fold analyses of fortified drinking water matrix samples were carried out under repeatability conditions using the cuvette tests and the standardised photometric methods. All the examined standard photometric methods and all the examined cuvette tests gave recovery rates in the drinking water matrix, as a measure of trueness, between 91.0 and %. For all the examined methods, the within No. Examined cuvette test Measuring range of cuvette test in mg/l series relative standard deviation under repeatability conditions, as a measure of precision, was in the range between 0.29 and 2.6 %, and can therefore be described as very good to good The standard photometric test method for nitrate is an exception, as the method variation coefficient of 4.9 % lies clearly below the 3 % limit which all the other values comply. It is noticeable that the precision of the cuvette tests for manganese and nitrate is better than that of the standard photometric method. In general, the limits of quantification of the cuvette tests are of the same order of magnitude as those of the standardised photometric methods. Limit of quantification (mg/l) Limit of detection (%) 1 LCK301 Aluminium LCK304 Ammonium LCK310 Chlorine, free LCK310 Chlorine, total LCK521 Iron, trace LCW532 Manganese, trace LCK339 Nitrate LCK541 Nitrite, trace LCK349 ortho-phosphate PC of the DWO limit of detection (%) Table 3: Limits of detection of the examined LANGE cuvette tests in comparison with the specified PCs of the Drinking Water Ordinance Fig. 7: Cuvette test LCK521 Iron, trace

8 DRINKING WATER ANALYSIS_RESUMEE Final assessment of the validation In a validation study by the IWW, nine photometric LANGE cuvette tests were experimentally validated and their performance characteristics were examined and assessed against the requirements of the 2001 Drinking Water Ordinance. The cuvette tests for the determination of aluminium, ammonium, iron (trace), manganese (trace), nitrate and nitrite (trace) satisfy the requirements of the 2001 Drinking Water Ordinance with regard to the performance characteristics trueness, precision and limit of detection. The performance characteristics are all below 10 %. On the basis of the determined performance characteristics, these cuvette tests are basically suitable for use in the analysis of drinking and raw water. The 2001 Drinking Water Ordinance specifies no performance characteristics for three of the examined HACH LANGE cuvette tests: chlorine (free), chlorine (total) and ortho-phosphate. The performance characteristics for the ortho-phosphate cuvette test (trueness, precision and limit of quantification) are all below 10 %. This cuvette test would therefore also comply with the three performance characteristics and fulfil the criteria. In the case of the cuvette tests for chlorine (free) and chlorine (total), the LANGE cuvette test Measurement principle similar to that of standard Trueness deviation 10 % Precision deviation 10 % LCK301 Aluminium No Yes Yes Yes LCK304 Ammonium Yes Yes Yes Yes LCK521 Iron, trace Yes Yes Yes Yes LCW532 Manganese, trace No Yes Yes Yes LCK339 Nitrate Yes Yes Yes Yes LCK541 Nitrite, trace Yes Yes Yes Yes Table 4: Performance characteristics of the LANGE cuvette tests Fig. 8: Cuvette test LCK339 Nitrate precision and limit of quantification are above 10 %, and the trueness is below 10 %. This result shows that the performance characteristics are mostly of the order of magnitude of those for standard photometric methods. The reason for this is the limited stability of the standard control solutions, which have to be prepared freshly, and is therefore independent of the analysis method. Limit of quantification deviation 10 % Literature [1] Prüfung und Bewertung ausgewählter HACH LANGE Prüfverfahren für die Untersuchung von Trinkwasser gemäß Trinkwasserverordnung 2001, 2008, IWW Rheinisch-Westfälisches Institut für Wasser Beratungs- und Entwicklungsgesellschaft mbh, Moritzstraße 26, Mülheim an der Ruhr [2] Verordnung über die Qualität von Wasser für den menschlichen Gebrauch (Trinkwasserverordnung TrinkwV 2001) vom 21. Mai 2001; BGBl., Teil I, Nr. 24 vom , S [3] DIN EN ISO/IEC 17025: ; Allgemeine Anforderungen an die Kompetenz von Prüf- und Kalibrierlaboratorien (ISO/IEC 17025:2005); Deutsche und Englische Fassung EN ISO/IEC 17025:2005 [4] Liste der Aufbereitungsstoffe und Desinfektionsverfahren gemäß 11 Trinkwasserverordnung 2001, Umwelt Bundesamt 2004 [5] Qualitätssicherung in der Analytischen Chemie, Funk W., Dammann V., Donnevert G, 2005; Wiley-VCH, Weinheim DOC Apr09