The importance of selecting an accredited laboratory to undertake testing is impressed upon in almost every aspect of site investigation, but what does it mean to be an accredited laboratory and more importantly what information is available to defend analytical data and explain the uncertainty associated with test results?
First it is important to consider what accreditation is and how it is granted and monitored.
The management system
Laboratories may be accredited to the internationally recognised standard ISO/IEC 17025:2005. Countries may have one or more accreditation bodies capable of issuing accreditation and within the UK this accreditation standard is granted by the United Kingdom Accreditation Service (UKAS). Under the International Laboratory Accreditation Cooperation Mutual Recognition Agreement (ILAC MRA) all signatories agree to maintain assessment of accreditation to the same high technical standard, and in turn deliver confidence in the results generated by accredited laboratories. Assessment bodies signed up to the ILAC MRA share a scheme of mutual recognition in that laboratories accredited to ISO/IEC 17025:2005, regardless of their geographical location, in that an accredited facility in one part of the world shall be carrying out testing to the same standard as any other accredited facility.
Within the UK, the Environment Agency’s MCERTS performance standard builds on ISO/IEC 17025:2005 impressing additional technical requirements on laboratories. This standard goes so far as to stipulate the manner of reporting test results, the maximum uncertainty budgets for specific determinands tested and in some cases the methodologies to be used.
In the UK, UKAS performs annual assessments of laboratories, during which conformance of the laboratory’s systems to ISO/IEC 17025:2005 is reviewed and accredited test methods are technically assessed. Any issues identified by the accreditation body are flagged and the appropriate action is agreed to remedy the issues, which must be completed within an agreed time frame for the laboratory to maintain their accreditation. During the interim period it is the responsibility of the laboratory to ensure the management system is maintained and that test results issued are of the highest quality. This done by a process of auditing, addressing any non-conforming work in an effective manner, and by a programme of continuous improvement.
So if the management system is all in place, how does a laboratory gain accreditation for a specific test? And how is the quality and uncertainty of the data assured?
For a laboratory to accredit a test method they must define the target analyte(s) in a matrix; and then fully document an analytical method which can be followed by a competent analyst to produce consistent high quality data. The method should be shown to take into account any interferences which may be present (and avoid them as far as is practicable); and to be robust and fit for purpose.
Defining the method
As part of the initial method development process the range of application of the method is defined. This should be fit for purpose with respect to the technique and analyses, and the levels of interest in the samples to be tested. For example, a method used for determining benzene in soil which has a linear range up to 200 µg/kg may be not be fit for purpose when samples contaminated with fuels are submitted for analysis that may contain levels up to 100 times greater than the upper range of the method. Equally if the requirement is for a highly sensitive method capable of determining this analyte at the 1 µg/kg, this needs to be considered when setting up the method. The linear range and sensitivity of the method is defined by the calibration range, but this is ultimately confirmed by the validation process.
For each calibration a “curve” (although this is generally a straight line relationship) is plotted with a line of best fit set. There is usually a minimum correlation criteria set that must be met in order for the calibration to be valid to use. (R2 or the Pearson Coefficient should typically be > 0.99 for most analytical methods). The correlation equation is then used to determine the concentration of an unknown sample.
The backbone to developing a method is the validation stage, which must be undertaken prior to seeking accreditation and defines the performance of the method. As a minimum, the validation involves taking a blank sample and spiking it with low and high concentrations of the parameter(s) of interest. This is undertaken in duplicate, and repeated a number of times. The recovery (bias) and precision of the method can then be estimated, and the significance of any difference between the required performance and achieved performance assessed. The validation should be undertaken in each matrix accreditation is sought for and should take into account all steps in the analytical method, thus also validating the sample preparation process.
Certified reference materials (CRMs) are available for a wide range of determinands in environmental matrices. These materials are well characterised homogenised samples and have been tested a number of times and show to give repeatable results for a parameter of interest. Suppliers should be accredited to ISO Guide 34, which is a standard for producing high quality traceable reference materials. CRMs are the preferred validation materials and are useful for a laboratory to assess their repeatability by repeated analysis of the same certified material; however, they are more useful for assessing external bias and only through analysing an external reference sample can a true bias check be performed.
In some cases (e.g. MCERTS), the maximum precision and bias targets for analyses are assigned, but it is quite common for the laboratory to define their own target quality criteria. Where the laboratory has defined its own criteria it is reviewed by the accreditation body to ensure they are meaningful and representative of the target analytes and test methods used. Generally, a laboratory will define its own criteria where necessary, but largely base the data on recognised performance criteria.
An important aspect of the validation is to ensure the measurement traceability chain meets all necessary requirements to produce high quality analytical data. If poor quality materials are used for calibration or recovery standards the method will appear to perform poorly. The quality of materials used throughout the method and particularly with respect to validation is heavily controlled within the Quality Management System and associated procedures.
So once the method has been defined, calibration ranges set and a validation is carried out in each appropriate matrix, a further two areas of quality control need to be evaluated:
• Internal Quality Control – or Analytical Quality Control (AQC)
This is the analysis of a sample of known concentration with every batch of samples (these may be, but are not limited to, a combination of CRMs, spiked materials akin to validation samples, spiked samples, re-run calibration standards and blanks).
• External Quality Control – Proficiency Testing Schemes
These are materials supplied to the laboratory for analysis. Their concentration is known, but not by the laboratory. The laboratory tests the samples blind and is assessed by an external body based on the results produced.
Internal quality control
The most commonplace form of internal quality control is the use of an AQC sample. Typically these are a matrix matched sample type, spiked with the analyte(s) of interest at a level equivalent to somewhere mid-range in the calibration curve. This concentration is known and should be consistent. Ideally the sample is analysed once per batch of samples, and is taken through the entire extraction process and analysis in the same way as a sample. The results are then calculated and plotted. Using statistically derived limits the performance of the result is assessed.
Ideally the result should be within ± 2 standard deviations of the mean/target value. Values outside of this constitute warnings, and results outside of ± 3 standard deviations are failures and require investigation. Given a pool of data with a normal distribution, it would be expected that a good size data set would have around 90-95 % of data passing between ± 2 standard deviations, with the remaining data points being classed as warnings or failures. The data points are recorded on a chart – commonly known as an AQC or Shewhart Chart – and monitored for trends.
Generally each laboratory will define their own criteria for investigating a trend and when failed data may be reported, but some standards give clarity to this – such as MCERTS for soils. Charts must be submitted as part of a request for accreditation so that the assessment body can review the performance of the method. They are also pivotal in monitoring ongoing performance of a method within the laboratory environment.
External quality control
In order for a test method to be granted accreditation some form of external quality control should be submitted – usually this would be proficiency testing data. Proficiency testing data is considered to be the gold standard for external quality control. A wide range of schemes exists, a useful library of suppliers and schemes can be found on the EPTIS website. A laboratory should participate in schemes for all accredited analytes and their matrices, and it is good practice to participate in schemes for routine determinands which are not accredited.
The simplest way to explain the scheme and use of data is to consider an example. LGC group provide a number of schemes – one of which is contest, which is specifically designed for laboratories undertaking contaminated land analysis (soil samples). Within the scheme, separate soil samples are supplied for analysis for metals, inorganics and organic components. These samples are homogenised soils that have been well characterised by the scheme provider, and then sub samples for each laboratory participating in the scheme. The laboratory tests each sample for a range of chemicals and then, along with all other participating laboratories, submits the data to the scheme provider for assessment. The provider then circulates a report to each participating laboratory with details of how they fared in the analysis. They also provide a summary report showing results from all labs under an anonymised lab code.
This data is perhaps the most useful in terms of assessing a laboratory’s performance as the samples tested are of an unknown concentration just like real samples and represent a complex matrix.
Below is an example of results achieved for a CONTEST scheme sample 3c (soil), tested for PAHs.
A further addition to the quality requirements for accrediting a test and maintaining accreditation is staff training and competency. Generally an analyst will be signed off as competent if they can follow a method and provide good reliable data. This can be checked by reviewing AQC performance, proficiency testing results for an analyst and even through original or continuing validation. Each laboratory defines their own criteria for assessing competence; however, these criteria are reviewed by the accreditation body to ensure they are fit for purpose and provide objective evidence to ensure an analyst is competent to carry out a particular method or technique.
Once a laboratory has gained accreditation to a standard for a particular analyte or method, the accreditation has to be maintained. This means that as well as the initial method definition and validation, ongoing internal and external quality control must be monitored by the laboratory. This data, as well as evidence of ongoing competence, is reviewed by the accreditation body to ensure that the quality of results reported by the laboratory is of a sufficient standard; and at least once in every accreditation cycle the test method itself is witnessed by a technical assessor. With each assessment the laboratory will usually receive some minor improvement actions that must be completed in order for accreditation to be maintained.
Quality data for site investigations
It is often the case that analytical data for a particular project is challenged or requires some additional supporting evidence, perhaps in relation to the uncertainty of a result. Within various sampling guidelines (e.g. BS 10175:2011 “Investigation of Potentially Contaminated Sites – Code of Practice”) there are specifications for duplicate samples that do give an idea of the sampling uncertainty; and the laboratory can also duplicate samples either prior to preparation or post preparation to give further information on how representative a sample is for contaminants of interest. All of this information can give confidence in analytical data; however, what may be more of use is to assess the performance of a laboratory test by reviewing AQC data and proficiency testing data.
Shewhart charts are straight forward to report and give an idea of trending within a method; however reviewing the plots may not easily give an indication of the precision or bias of a method. Generally a laboratory will be able to provide the precision and bias (and therefore overall measurement uncertainty) budgets in terms of a percent.
Proficiency testing data may also be useful to include in a report on the quality of laboratory data. Proficiency testing samples are usually run with other commercial samples and therefore represent an external quality check on the data associated with that method at the time. A summary is typically the most useful way to include this in a report.
Ultimately the laboratory has lots of supporting data behind each analytical result produced, but often including each and every one of these just leads to a mass of data which is little understood. An audit of the laboratory with a walk through of each of the steps for gaining and maintaining accreditation is an excellent way to grasp a sense of data quality, but often this is not practical; and therefore a quality report including Shewhart charts and perhaps proficiency testing summaries is fit for purpose, giving an overall idea as to the quality of the data produced by the laboratory for a project.
A word of caution
These points have heavily laboured the laboratory quality traceability chain and international and national quality standards for environmental laboratory testing, but little detail has been given here to the uncertainty surrounding sampling. This is undoubtedly an area that must be considered when assessing the overall uncertainty of data produced. In fact, uncertainty associated with analytical data is perhaps the most widely understood and characterised, and if the sample tested is not representative of the conditions on site then even the most advanced techniques cannot produce meaningful data.