Trueness

The accuracy of an analytical procedure expresses the closeness of agreement between the value, which is accepted either as a conventional true value or an accepted reference value and the value found. This Is sometimes termed trueness . 

The methods It is advised that, prior to the certification measurements, the participants discuss their methods so that all participants have confidence in each others methods and there is a good level of agreement between laboratories. As it is preferred to certify on the basis of the agreement between different methods applied in different laboratories, a proposal should indude, where relevant and possible, a group of laboratories offering a range of widely different measurement methods. Each laboratory should use well established method(s), with which it can demonstrate adequate performance in terms of trueness and in terms of reproducibility. 

CsuRos M (1997) Environmental Sampling and Analysis. Laboratory Manual. CRC Press. Doerfeel K (1994) Assuring trueness of analytical results. Fresenius J Anal Chem 348 183-184... 

There are various approaches to determine the trueness of methods. The most common is the performance of recovery experiments. According to the guidance document SANCO/825/00, the mean recovery should be in the range of 70-110%. In justified cases, recoveries outside this range can be acceptable. 

The assessment of validation data of CEN methods does not differ significantly from other validation schemes. The most important quantitative performance characteristics are trueness and precision. Additionally, some information about sensitivity... 

Komit6 for Levnedsmidler (NMKL)]. The standard presents a universal validation approach for chemical analytical methods in the food sector. This includes methods for the main constituents and also for trace components. Therefore, the NMKL procedure focuses on primary validation parameters, such as specificity, calibration, trueness, precision, LOD or LOQ and does not refer to special requirements of pesticide residue analysis. 

Even if most examples and procedures presented apply to in-house validation, the procedure does not distinguish between validations conducted in a single laboratory and those carried out within inter-laboratory method performance studies. A preference for inter-laboratory studies can be concluded from the statement that laboratories should always give priority to methods which have been tested in method performance studies. Within the procedure a profound overview of different categories of analytical methods according to the available documentation and previous external validation is given. For example, if a method is externally validated in a method performance study, it should be tested for trueness and precision only. On the other hand, a full validation is recommended for those methods which are published in the scientific literature without complete presentation of essential performance characteristics (Table 9). 

Verification implies that the laboratory investigates trueness and precision in particular. Elements which should be included in a full validation of an analytical method are specificity, calibration curve, precision between laboratories and/or precision within laboratories, trueness, measuring range, LOD, LOQ, robustness and sensitivity. The numbers of analyses required by the NMKL standard and the criteria for the adoption of quantitative methods are summarized in Table 10. 

Verification of trueness and precision, possibly also detection limit... 

No fixed criteria, because requirements on trueness depend on concentration a mean recovery of 80-110% generally seems sufficient... 

Trueness At LCL and two higher levels >1 Rephcate in >3 laboratories at each level >4 commodities (typical matrices) Mean recovery 70-110% calculated for each analyte-commodity-concentration combination Outhers should be removed (the number of deleted suspicious recovery data must not exceed 20%)... 

In summary, official German analytical methods for pesticide residues are always validated in several laboratories. These inter-laboratory studies avoid the acceptance of methods which cannot readily be reproduced in further laboratories and they do improve the ruggedness of analytical procedures applied. The recently introduced calibration with standards in matrix improves the trueness of the reported recovery data. Other aspects of validation (sample processing, analyte stability, extraction efficiency) are not considered. 

If analytical methods are validated in inter-laboratory validation studies, documentation should follow the requirements of the harmonized protocol of lUPAC. " However, multi-matrix/multi-residue methods are applicable to hundreds of pesticides in dozens of commodities and have to be validated at several concentration levels. Any complete documentation of validation results is impossible in that case. Some performance characteristics, e.g., the specificity of analyte detection, an appropriate calibration range and sufficient detection sensitivity, are prerequisites for the determination of acceptable trueness and precision and their publication is less important. The LOD and LOQ depend on special instmmentation, analysts involved, time, batches of chemicals, etc., and cannot easily be reproduced. Therefore, these characteristics are less important. A practical, frequently applied alternative is the publication only of trueness (most often in terms of recovery) and precision for each analyte at each level. No consensus seems to exist as to whether these analyte-parameter sets should be documented, e.g., separately for each commodity or accumulated for all experiments done with the same analyte. In the latter case, the applicability of methods with regard to commodities can be documented in separate tables without performance characteristics. 

When specifying atomic coordinates, interatomic distances etc., the corresponding standard deviations should also be given, which serve to express the precision of their experimental determination. The commonly used notation, such as d = 235.1(4) pm states a standard deviation of 4 units for the last digit, i.e. the standard deviation in this case amounts to 0.4 pm. Standard deviation is a term in statistics. When a standard deviation a is linked to some value, the probability of the true value being within the limits 0 of the stated value is 68.3 %. The probability of being within 2cj is 95.4 %, and within 3ct is 99.7 %. The standard deviation gives no reliable information about the trueness of a value, because it only takes into account statistical errors, and not systematic errors. 

The reliability of multispecies analysis has to be validated according to the usual criteria selectivity, accuracy (trueness) and precision, confidence and prediction intervals and, calculated from these, multivariate critical values and limits of detection. In multivariate calibration collinearities of variables caused by correlated concentrations in calibration samples should be avoided. Therefore, the composition of the calibration mixtures should not be varied randomly but by principles of experimental design (Deming and Morgan [1993] Morgan [1991]). 

Trueness. Absence of systematic errors can be tested traditionally by means of recovery functions see Sect. 6.1.2, Fig. 6.3C Burns et al. [2002]. For this reason the concentration estimated by the model, x, is compared with the true concentration value, xtrue, by a regression model... 

For the characterization of the reliability of analytical measurements the terms precision, accuracy, and trueness have a definite meaning. 

The ISO recommendation [1993] should be followed and accuracy used only as a qualitative term. In case of quantitative characterization (by means of the bias), a problem may appear which is similar to that of precision, namely that a quality criterion is quantified by a measure that has a reverse attribute regarding the property which have to be characterized. If the basic idea of measures can be accepted, which is that a high quality becomes a high value and vice versa, bias is an unsuited measure of accuracy (and trueness). In this sense, accuracy could be defined by means of a measure proposed in the next paragraph. 

Precision, accuracy and trueness are important performance characteristics in analytical chemistry. Each of them is well-defined in a positive sense ( closeness of agreement... ). However, their quantifying is done by means of unfavourable measures, namely by error quantities like, e.g., standard deviation and bias, respectively, which indeed do quantify imprecision and... 

Accuracy and trueness have been defined above and it was mentioned that these terms base on qualitative concepts (ISO 3534-1 [1993]). If it is necessary to have quantitative information, the bias, which is a measure of inaccuracy, should not be used to quantify accuracy and trueness, respectively. Instead of this, the following measures might be applied... 

Analyte Reference values/ umol/L Analytical results x Ax/(umol/L) Recovery Precision VrecOc) Trueness accCx) Comment The result is... 

The example given in Table 7.2 is taken from a study to verify the trueness of clinical analyses (Streck [2004]). Recovery rates have been used as the criterion to accept a good agreement between the measured results and the reference values as it is frequently done by analysts. 

It should be noted that the term sensitivity sometimes may alternatively be used, namely in analytical chemistry and other disciplines. Frequently the term sensitivity is associated with detection limit or detection capability. This and other misuses are not recommended by IUPAC (Orange Book [1997, 2000]). In clinical chemistry and medicine another matter is denoted by sensitivity , namely the ability of a method to detect truly positive samples as positive (O Rangers and Condon [2000], cited according to Trullols et al. [2004]). However, this seems to be more a problem of trueness than of sensitivity. 

The trueness of the analytical results or their inaccuracy, respectively, characterized by the bias 6 = x — xtrue see Sect. 7.1.3. 

Classical information theory according to Shannon [1948] and Bril-louin [1963] consider only items (1) and (2) the trueness of information has not been taken into account. 

The situation becomes more complex when aspects of the trueness of analytical results are included in the assessment. Trueness of information cannot be considered neither by the classical Shannon model nor by Kullback s divergence measure if information. Instead, a model that takes account of three distributions, viz the uniform expectation range, po(x), the distribution of the measured values, p(x), and that of the true value, r(x), as shown in Fig. 9.5, must be applied. 

The consideration of both precision and trueness by means of the Kerridge-Bongard model can be generalized as follows ... 

It goes without saying that you should make all measurements to the best of your ability. However, a value to the highest level of precision and trueness is not always required. The aim is that the result produced should be accurate enough to be of use to the customer, for the intended purpose (see Chapter 4). Customers may want the technical details of the method used but more often this will not be... 

Measurements are subject to systematic errors as well as the random errors covered in Section 4.3.2. Bias is the difference between the mean value of a large number of test results and an accepted reference value for the test material. The bias is a measure of trueness of the method. It can be expressed in a number of ways, i.e. simply as a difference or as a ratio of the observed value to the accepted value. This latter representation, when expressed as a percentage, is often termed recovery. This represents how much of the analyte of interest has been extracted from the matrix and measured. This is dealt with in Section 4.6.3. 

---