Accuracy and Traceability in Speciation Analysis

PHILIPPE QUEVAUVILLER

1 Introduction

Traceability issues are of increasing concern in all fields where chemical measurements form the basis for decisions. The concepts of accuracy and traceability as applied to environmental analysis are, however, still prone to misunderstanding. Some years ago, Horwitz stated that 'considerable evidence exists in the literature that few analytical chemists pay attention to the question of the reliability of the analytical results they produce. These chemists believe that a natural law exists in measurement science, that if the directions for conducting a measurement are followed, the true value necessarily results'. This corresponds to the long-term debate about precision or reproducibility over accuracy, which is now relayed by ongoing discussions on accuracy and traceability: while accuracy refers to the closeness of analytical values to 'true values' (trueness) and among various repetitions (precision), the term traceability implies a link between the data obtained and established references, through an unbroken chain of comparisons, all with stated uncertainties. Recent controversial discussions have illustrated the misunderstanding, which may occur among the analytical community with respect to accuracy and traceability issues in the area of speciation analysis, with possible consequences on environmental data interpretation.

Speciation analysis is no longer a new feature. IUPAC defines this term as 'the analytical activity of identifying and measuring the quantity of one or more individual chemical species in a sample'. The speciation of an element is defined as 'the distribution of defined chemical species of an element in a system'. Chemical species of some elements (e.g. organotins, organomercury compounds) are now included in the list of substances to be determined in the frame of international environmental programmes, requiring an increasing knowledge and care with respect to the quality control (including traceability issues) of all monitoring steps (from sampling to reporting data).

Analytical techniques used for the determination of chemical species are generally based on a succession of steps (e.g. extraction, derivatisation, separation, detection) all of which are prone to various sources of systematic errors. Within the last decade, international collaborative efforts (through inter-laboratory studies and certification of reference materials) have enabled the systematic study of hyphenated techniques used for the determination of chemical species of, e.g., arsenic, chromium, mercury, lead, tin and selenium in environmental matrices (water, fish or mussel tissues, sediments). The determination of operationally-defined element fractions (extractable forms of trace elements) in sediment and soil matrices has also been collaboratively studied, mainly to harmonise and standardise extraction schemes, in order to improve the comparability of data (stressing that, strictly speaking, this type of determination should not be covered by the term 'speciation'). In this context, all of these collaborative efforts have been understood as being directed towards the drive for accuracy (trueness and precision as defined below). It has been recognised recently that these achievements have actually led primarily to the establishment of reference points (e.g. certified values in reference materials), which do not necessarily correspond to 'true values', but offer a mean with which laboratories may compare their data internationally and, hence, achieve traceability. This ambiguity still generates confusion and misunderstanding among the scientific community. This chapter discusses this issue, focusing on analytical measurements only. Extending discussions on general traceability issues would imply an examination of steps prior to laboratory work (sampling, storage, etc.) which is beyond the scope of this contribution.

2 Accuracy

The accuracy concept covers the terms trueness and precision. Trueness is defined as 'the closeness of agreement between the 'true value' and the measured value', whereas precision is 'the closeness of agreement between the results obtained by applying the same experimental procedure several times under prescribed conditions'. Trueness relies on the true value of the substance to be measured, which is defined as 'a value, which would be obtained by measurement, if the quantity could be completely defined and if all measurement imperfections could be eliminated'. We will discuss later in this chapter what the practical implications for speciation analysis are. There are three recognised ways to evaluate accuracy:

1 comparison with an 'independent' method (i.e. with different measurement principles and different sources of errors),

2 comparison with other laboratories, and

3 use of Certified Reference Materials.

The principle of establishing certified values of reference materials on the basis of comparisons of independent methods used by different (independent) laboratories has been followed by the European Commission (BCR and its successors). The certified values were long considered to reflect the best estimates of the true values of the certified substances. As discussed below, some may argue that certified values actually represent reference points to achieve traceability, but that they are not necessarily to be considered as true values for the verification of accuracy. Strictly speaking, measurement results are accurate when both the result and its uncertainty are described in units from the 'Systeme International' (SI units), i.e. the kg or the mole for analytical measurements. Previous discussions have underlined that the use of SI units in chemical measurements is quite unrealistic and is only applicable if primary methods ('method having the highest metrological quality, for which a complete uncertainty statement can be written down in terms of SI units', e.g. gravimetry, titrimetry, coulometry, IDMS) are involved.' This will be discussed later in this chapter.

3 The Traceability Concept as Applied to Speciation

Traceability is defined as 'the property of the result of a measurement or the value of a standard whereby it can be related to stated references, usually national or international standards, through an unbroken chain of comparisons all having stated uncertainties'. The application of this concept to chemical measurements has been extensively discussed over the past ten years. The practice often differs from theory: indeed, analytical chemists usually describe their results (amounts of chemical compounds) in terms of weight or mass, whereas metrologists underline that weighing ignores the chemical nature of the measurand and the fact that particles interact with each other in chemical reactions and not masses of matter. It is recognised that the actual measurement of the 'amount of substances' correspond to approximations, consisting of measuring ratios of 'weight' and converting them into ratios of amount by means of 'atomic weights' or 'molecular weights', which is considered accurate enough for most chemical purposes. In the strict metrological sense, the 'approximations' do not allow one to demonstrate that the measurements are traceable to the relevant SI unit, i.e. the mole. The arguments developed by metrologists are that, in any chemical reaction, the masses of reacting compounds change (even if the effect is extremely small) since there is energy uptake or production. Furthermore, mass is a property of matter with is basically inert, which is not the case for (amount of) particles, which explains why the SI system distinguishes 'mass' from 'amount of substance'. As stressed above, the practice is far removed from theory, and a recent discussion of the key elements of traceability was conducted between analytical chemists and metrologists, to understand the concept and make it applicable to routine chemical measurements. The three key elements concern (1) the link to stated references, (2) the unbroken chain of comparison and (3) the stated uncertainties. The following paragraphs examine how these concepts apply to speciation analysis.

3.1 Stated References for Speciation Analysis

3.1.1 Generalities

The 'stated references' may be reference methods, reference materials or, as already said above, the units of the Systeme International (SI). The mole and the kg are the SI units which underpin chemical measurements; the mole relates atomic/molecular entities to a macro-scale via classical chemical reactions and provides the basis for analytical techniques such as, for example, gravimetry, titrimetry and electrochemical measurements (keeping in mind that the kg is necessary to define the mole). In theory, perfect traceability could be established if each atom/molecule of a certain substance could be counted one by one on a microscopic scale. In practice, the measurements correspond to approximations, e.g. via comparisons of amounts, of instrumental response generated by a number of particles, etc. Basically, establishing SI traceability nowadays implies the ability to demonstrate to what extent the approximations made are clearly related to the stated references. Typically, many chemical measurements are actually traceable to either a reference material or to a (reference) method. In the field of speciation analysis, the 'stated references' can hardly be established to the mole given the present state of knowledge. Indeed, as discussed below, the techniques used involve a series of analytical steps, which multiplies approximations that are still not under control. The references, in this case, are either pure substances or Certified Reference Materials (when they exist). At present, there are no real 'reference methods' in speciation to which results can be traceable, with the exception of operationally-defined parameters (i.e. 'forms' of elements defined according to an extraction protocol, e.g. a single or a sequential extraction protocol, which represents in this case the reference. As stressed above, however, this type of measurement is not really considered as being covered by the term 'speciation').

3.1.2 CRMs as Stated References

It has been emphasised that the 'reference' represented by a CRM is not always reliable since, in many cases, the RM does not have the 'same' matrix as the unknown sample. In most cases, CRMs represent a compromise with respect to the matrix of the unknown sample, which will be useful as a quality control tool. Analytical chemists should not expect more from CRMs than they can offer, i.e. they are to be considered as useful tools for validation and not calibration tools for applying 'correction factors' to measurement results. In other words, if an error is detected in a method when analysing a CRM, this error has to be removed before the analysis of the unknown sample, and not corrected for on the basis of the deviation observed for the CRM results. This may sound trivial, but the validation process of a method, involving the use of CRM(s), is required prior to measuring unknown samples! Furthermore, it should be stressed that a correct result obtained with a CRM does not give a full assurance that 'correct results' will be achieved when analysing unknown samples, due to differences in matrix composition.

In addition, the question of traceability of CRMs, representing complex chemical systems, to SI units is still an open debate. The new ISO definition of a CRM implies 'traceability to an accurate realisation of the unit in which the property values are expressed'. As with the traceability to SI units, this 'accurate realisation' is often difficult to demonstrate in the field of speciation analysis. In practice, the approximations made at different analytical steps do not give proof that a 100% recovery has been obtained (e.g. extraction recovery, derivatisation yield). The approximations are actually more valid when the certified values have been obtained in the frame of inter-laboratory studies involving several (independent) laboratories using a variety of different techniques. Even then, in the absence of a 'definitive method', the collaboratively obtained value is considered to reflect the 'state-of-the-art' of a given method (hence a good reference point), but not necessarily the 'true value' of the measured chemical compound.

As stressed before, traceability of chemical measurements to the SI unit (i.e. to the mole) is achievable for relatively 'simple' determinations such as trace elements in seawater. This has been made possible through measurement rounds between a few metrological laboratories using a so-called 'primary reference method' (in this case isotope dilution mass spectrometry). The traceability of certified values of a RM to the mole is, therefore, theoretically achievable. What is demonstrated for water analysis, however, is far from being achievable for complex matrices requiring chemical pre-treatment, extraction, clean-up, separations, etc. In this case, the 'chain' will be broken at several stages and the traceability will rely on approximations, i.e. recovery estimates. The better these estimates, the closer is the achievement of traceability to the relevant SI unit, the mole. This goes hand-in-hand with the possible achievement of accuracy, i.e. the closeness to the 'true value', which is intimately linked with the possibility of achieving, and demonstrating, a 100% recovery of the measurand at each analytical step, where loss or contamination might occur.

Examples of CRMs developed over the past few years to serve as 'references' for speciation analysis are given in Table I.

3.1.3 Reference to Well-defined Species

As a general comment, one should consider that a given compound may have different species and/or different ways of being reported. An example is tributyltin, (C4H9)3Sn, which may be reported as the cation TBT+, or with its respective anion (e.g. chloride, acetate, oxide), or even as Sn, depending on the analytical techniques used. Therefore, besides the definition of 'stated references' as physical entities (e.g. pure substances or CRMs), the results of speciation analysis require to be traceable to well-defined units and chemical forms. Potentially serious errors are made because results, submitted according to one chemical form, are compared with results reported in a different unit (e.g. comparison of TBT results reported as TBTCl with results reported as Sn). Strictly speaking the results of chemical determinations should be reported as 'amount (of substance) measurements' to comply with the requirements of the SI system.