VAM BULLETIN

An LGC publication in support of the National Measurement System Issue Nº 19 Autumn 1998

Are you prepared for the change of the century? VAM and the dipstick – consumer analytical products Industrial analytical measurement – the reality DNA chips – a taste of the future VAM in the brewing industry EDITOR’S NOTE AND CONTENTS

Alison Gillespie Editor Editor’s note 0181-943 7563

Welcome to VAM Bulletin 19. technology is the area of DNA microarrays which General enquiries about VAM to: VAM Helpdesk The VAM programme comprises a large is described in one of our contributed articles. 0181-943 7393 amount of work related to biologically-based As we approach the millennium, there is analytical measurement. In this issue, there are timely reminder in our editorial on how the LGC’s address: two articles describing new developments in the Century Date Change problem may affect your LGC rapidly expanding area of DNA technology and laboratory and what you should do if you have Queens Road their use in bioanalysis. Dr Valerie Owen has not already made plans. Teddington written the guest column for us on how the VAM I would like to thank all the contributors for Middlesex TW11 0LY principles may be applied to the use of the the time and effort they have spent preparing the increasing number of consumer analytical articles for this edition which I hope you will enjoy The DTI products that are available over the counter. One reading. As always contributions from any of our of the most exciting developments in DNA readers are welcome. VAM programme: The DTI’s programme on Valid Analytical Measurement (VAM) is an integral part of the UK National Measurement System. The VAM Contents programme aims to help analytical Editorial Statistics in context laboratories demonstrate the validity of their data and to facilitate mutual Are you prepared for the change Missing values, outliers, robust recognition of the results of analytical of the century?...... 3 statistics and non-parametric measurements. methods ...... 22 The VAM programme sets out the following six principles of good analytical Guest column practice, backed up by technical support VAM in education VAM and the dipstick – consumer and management guidance, to enable laboratories to deliver reliable results analytical products...... 4 PT scheme for schools results...... 28 consistently and thereby improve performance. Focus on principles VAM news

1. Analytical measurements should be VAM principle 6 – The Consultative Committee made to satisfy an agreed requirement. Well defined QC and QA...... 6 on Amount of Substance...... 29 2. Analytical measurements should be made using methods and equipment Contributed articles Reference materials update ...... 29 which have been tested to ensure they are fit for their purpose. Customers are more concerned VAM products and services...... 30 3. Staff making analytical measurements about price of analysis than quality… should be both qualified and competent aren’t they?...... 7 Chemical nomenclature to undertake the task. Industrial analytical measurement – How to get it wrong! ...... 33 4. There should be a regular independent the reality ...... 12 assessment of the technical performance of a laboratory. DNA chips – a taste of the future ...... 14 Forthcoming events...... 34 5. Analytical measurements made in one location should be consistent with those Case study Contact points...... 36 elsewhere. VAM in the brewing industry ...... 18 6. Organisations making analytical measurements should have well defined quality control and quality assurance procedures.

Cover photograph by Andrew Brookes 2 VAM BULLETIN EDITORIAL Are you prepared for the change of the century?

implications for the laboratory it is helpful to larger instruments but even on some simple Mike Sargent, break the problem into three main areas: appliances such as balances, uses date and/or • central systems time information. It seems likely that the LGC • scientific systems: instruments, only viable solution for some of these workstations and applications systems will be direct replacement or s we approach the year 2000, the • embedded systems rationalisation of requirements to fewer ACentury Date Change problem (or Central systems may include legacy systems. The cost and disruption to work is ‘Millennium Bug’) could bring chaos to your mainframe or central mini-computer potentially very high. laboratory. By now there has been so much applications (such as a laboratory information publicity regarding the problem that few can management system (LIMS), a finance Scientific systems are likely doubt its existence or serious nature. The system, and a personnel database) and a to be the major problem key question for managers, however, is how local area network (LAN) infrastructure with for most laboratories vulnerable is their own business? If you have workstations and office applications not yet made an assessment of the potential software. The main problem is likely to arise Embedded systems could, potentially, pose risks to your own operations it is vital that with the legacy systems and applications. the biggest problem for managing you do so as soon as possible – and take the Unless these are compliant already (possibly millennium compliance. Some idea of the appropriate corrective measures. with recent upgrades or minor software possible scope can be gained by noting that All businesses should be aware that the ‘patches’ from the supplier if they are still such systems are used in automated building problem extends much more widely than supporting the system) the time and management systems, office equipment such mainframe, networked or personal knowledge needed to provide bespoke fixes as photocopiers, building systems such as computers. It embraces not only ‘real’ may be prohibitive. In this case the only lifts or switchboards, and most of the computers but also the microchips which are viable option is a new system – which may in ‘minor’ laboratory equipment. Unless they embedded in many products or facilities, any case offer other benefits. Most of the have already started work on a year 2000 including most modern analytical widely used LAN systems are relatively project, few laboratories have any knowledge instruments. Many of these microchips modern with support for millennium of the extent or size of their potential incorporate a date-logic function. As a compliance available from suppliers or third problem. Without this knowledge, it is result, the equipment may cause data to be parties. The main problem is likely to be old impossible to determine the likely corrupted, may malfunction, or may fail MS-DOS applications still in use on some consequences or the cost of the fixes that are completely when the Century Date Change office workstations. Nevertheless, a required, if indeed any are required. occurs. On the other hand, there may be no significant effort may still be needed to There are also other Century Date problem at all – it depends on how the chip implement any upgrades or replacements Change issues common to all businesses. In is used, on how it was programmed, or even and to confirm the claims made by suppliers. particular, a company may be at risk because on the production version available at the Scientific systems are likely to be the of problems experienced by its suppliers, for time of manufacture. major problem for most laboratories. example, urgently needed reagents or services the problem extends Frequently there is a large number of these may be unavailable for several weeks after 1 systems, their age, type and origin vary January 2000. There may even be a problem much more widely than enormously, there is often no detailed with customers, for example, their inability mainframe, networked inventory of the equipment (irrespective of to settle invoices could cause cash flow or personal computers any knowledge of potential year 2000 problems. In addition, the crisis may be problems), and a large part of the company’s much closer than expected, for example, This situation poses a major headache business will depend on them. Many of the some financial systems are already refusing for analytical laboratories, most of which instruments use bespoke software, some to accept next-century dates which are have a large collection of instrumentation produced by defunct suppliers, some needed to input tax, pension, or contract acquired over many years. One such item produced in-house and often inadequately data. It is also believed that many systems may be ‘year 2000 proof’ whilst another, documented. As well as future operations with a problem may generate erroneous almost identical, piece of equipment from managers also need to consider forensic or dates from 1 January 1999 rather than 2000. the same manufacturer could have a date- accreditation requirements for integrity of So what can be done in the very limited related problem even though manufactured and access to archived electronic data. time available? The best advice to management more recently. In reviewing the year 2000 Much of the software, particularly on is: DON’T PANIC! There is much to be

3 VAM BULLETIN EDITORIAL gained by preparing an inventory of all • do establish priorities for action based independent review of satisfactory equipment, services, and business partners on an assessment of commercial and conformity on completion which may be affected. Many items can then technical risks • do obtain independent advice on legal be checked off as either obviously safe or not • do ensure unambiguous management aspects, accreditation requirements, posing any risk to the business. The next responsibility for the project, use of insurance issues, etc. stage is to review those that remain and set resources, and rapid authorisation of • do pilot the compliance process for one priorities: it is essential to identify what is expenditure or two systems of each type and also one critical to the business rather than • do identify specific individuals to take or two suppliers attempting to fix all possible problems. It is responsibility for conformity of each • do turn to suppliers for assistance, but… important to recognise that some equipment critical system • don’t assume that suppliers will offer will inevitably need to be replaced and to • do keep records to show accountants, financial or other warranties allow adequate budget provision and time auditors, clients or other interested • do make contingency plans where for purchase and testing. The latter is parties that the task has been properly compliance of critical equipment or particularly important for the analytical undertaken suppliers cannot be proven laboratory – a change of instrumentation is • do remember that staff will need time to • do set up a procedure to ensure all likely to require expensive and time undertake the extra work and possibly new purchases (especially those intended consuming revalidation of the methods expert advice or assistance to overcome a year 2000 problem!) which use it. • do use expert advice and assistance in are compliant Finally, some do’s and don’ts: identifying problems, providing specific • DON’T LEAVE EVERYTHING TO • do plan and budget a year 2000 project solutions where necessary, and for an THE VERY LAST MINUTE.

GUEST COLUMN VAM and the dipstick – consumer analytical products

Most OTC analytical kits are single-shot manufacture (principle 4 – independent Valerie Owen, dipsticks which are either assessed visually or assessment of a laboratory, principle 5 – inserted into some sort of reader/detection measurements made in one location should be SciTec device. The construction and use of consistent with those elsewhere and principle 6 these kits is different from laboratory – having well defined quality control and quality Management instrumentation and thus may pose different assurance procedures.) One principle presents validation requirements. Is there any way of a greater challenge (principle 3 – analytical Consultants ensuring that the user of a single-shot device staff should be both qualified and competent to is getting a reliable, useful analytical result? undertake the task.) he VAM principles for achieving Is it possible to achieve the aims of the six The choice of the right measurements, Treliable analytical measurements were VAM principles when the general public the right tools and the right procedures are originally designed for analyses carried out in uses single-shot devices for analysis? in the hands of the designers and a laboratory by trained staff. However, there manufacturers of the device. Hence the user is an increasing trend for analysis to be The trend to depends upon the manufacturer to ensure performed out-of-the-laboratory by people ‘consumer analytical that their products are fit for purpose. The with no specialist analytical training. Over- products’ looks set criteria applied to a laboratory-performed the-counter (OTC) devices are available for to continue assay should be applied to single-shot pregnancy testing, contraceptive planning, devices for use by the non-specialist person. cholesterol measurement, water testing and Of the six principles it should be possible The analyte must be a proven, acceptable a range of other analyses. The trend to to satisfy two using the same criteria as are indicator of the condition for which the ‘consumer analytical products’ looks set to used in laboratory analysis (principle 1 – device is to be used. The analytical system continue as the public demand increases to satisfying agreed requirements and principle 2 – must be thoroughly checked against the know more about their state of health and using tested methods and equipment.) Three usual criteria used in a laboratory assay, e.g. the quality of their environment. principles may be satisfied at the point of accuracy, precision and freedom from

4 VAM BULLETIN GUEST COLUMN interference. The equipment and analytical dipstick or individual analysis. The quality analytical result is of modest socioeconomic procedure will probably differ from that used control must be incorporated into the device importance, such as checking the pH of the in the laboratory as these must be itself. For example, the device might have garden soil, then the implications of an appropriate for use by the non-analyst and three channels, one of which is a low or incorrect analysis are relatively trivial. must be sufficiently robust to operate ‘in the negative control, one is a high or positive However, as out-of-the-laboratory/OTC field’. Nevertheless, the developer/ control, while the third channel is for the dipstick technology moves into areas of very manufacturer must thoroughly verify their analysis of the sample. These channels significant socioeconomic importance, such suitability and reliability. Also the onus is on should become active by the addition of the as monitoring drinking water quality or as the developer/manufacturer to check the sample or by the analytical process with little part of contraception, the implications of an device against accepted international or no extra handling steps by the user. inaccurate analysis could be tragic. It is at standards and against reference methods and this level we need to ensure that a valid other established analytical procedures. This Can the dipstick give measurement is made by good design, checking should extend to the use of the an analytical result which extensive testing, and carefully controlled device outside the development laboratory, is fit for purpose? manufacture. into the environment in which the device will be expected to function. Principle 3, that measurement staff Demonstrable consistency of devices can should be competent, properly qualified and The opinions expressed in this article be achieved by thorough quality control well informed, cannot be achieved if the are those of the author and not necessarily during manufacture so that every item device is to be used by the general public. the opinions of other organisations with which leaving the manufacturer is within defined Competent staff should in theory ensure that she is associated. limits. This level of quality control and the correct sample is used in the analysis, the reliability of supply is being achieved as the sample is handled correctly, an ‘out of line’ result of heavy investment (research, result is spotted and checked, and the result development and financial) by the of the analysis is reported correctly. There manufacturers of the world’s leading should also be some capacity to interpret the devices. It has involved development and result and spot an analytical failure/error. selection of extremely well defined, Some advice to address these aspects can be consistent raw materials, extensive batch given to the user even when the devices are testing and so on. In areas such as clinical purchased over the counter. For instance, analysis, manufacture is also subject to the user could be advised to read the regulatory control. instructions. These instructions should Principles 4 and 6 call for both internal provide clear guidelines as to how to use the and independent assessment of the technical kit and explain problems or common errors performance of measurement facilities and which may arise. However, even if the procedures. Internal quality control used in guidelines are well written and clear, will the routine laboratories is usually statistical user interpret them in the way that they are quality control (SQC) designed to check the meant? If instructions include ‘looking at the precision and accuracy of an analytical window’, does the user understand that the process which is carried out frequently, window is a small panel on the device and repeatedly and often using mechanical/ not the aperture in the room where the automated instrumentation. This form of analysis is taking place! QC is designed to look for systematic errors. Is it possible for the manufacturers of In the case of a single-shot device, each such devices to engineer out the human analysis is separate and once a QC sample is propensity to err? The designers should run the device is discarded and thus is not attempt to minimise the opportunities for available for the subsequent sample analysis. misuse or misinterpretation. Validation Also the manufacturer should have had facilities could be built into devices to ensure adequate QC in place to check each batch of a dipstick from an appropriate batch is being devices to ensure total consistency. inserted correctly. Some new state-of-the-art However, can we ensure that the device has devices have gone a long way to achieving been handled appropriately after release by this goal with internal data checks to detect the manufacturer? (Devices containing for analytical anomalies, plus physical design biological reagents of modest stability may to reduce the chance of device mishandling. need to be refrigerated.) Can we detect the The ultimate question is whether the one-off error in handling? Any QC, at the dipstick can give an analytical result which is time of analysis, must be on the individual fit for purpose. When the use of the

5 VAM BULLETIN FOCUS ON PRINCIPLES VAM principle 6 – Well defined QC and QA

and the defined quality control procedures. how well the organisation performs there will Elizabeth Over a period of time a laboratory will always be some cost but some of this should have developed a range of procedures to be recouped in the efficiency savings Prichard, LGC ensure that it meets its clients’ needs. produced. There will also be running costs, Frequently these procedures are assumed to one new cost will be for internal audits that n previous issues of the VAM Bulletin1 - 5 be common knowledge. However, rarely are make sure the written procedures are being Ifive of the six VAM principles have been all these procedures written down and hardly followed. If the system is externally assessed highlighted and an explanation given of how ever will they all be known by any one there will be further costs to cover the and why they are part of good laboratory person. All this is fine until there is a dispute external audits. It is difficult to quantify all practice. The goal of adopting the VAM or a key person is absent or leaves suddenly the costs but the operating costs are principles is to demonstrate that all the and a problem cannot be resolved. This is a considered to be in the region of 5 - 10% of analytical measurements made by a good reason for having a quality system. The total costs. It is possible that this will be set laboratory are accurate, reliable and fit for top level document of a quality system is the off against business that would not have their intended purpose. Applying VAM Quality Manual, in which essential policy been obtained without a formal quality principles 1 to 5 is in itself not sufficient to matters are written down. This is supported system. Measurements will be repeated only achieve this goal. The missing component is by other documentation containing the when there is a good and clear reason for VAM principle 6, “Organisations making procedures to be implemented. This ensures doing so. In addition, instrumentation and analytical measurements should have well that everyone in the laboratory knows what other equipment may last longer because it defined quality control and quality assurance is expected of them. The process of writing is properly maintained. procedures”. These components you will a Quality Manual often shows up Increasingly clients are asking for recognise as the elements of formal quality inconsistencies in how things are done and evidence that a laboratory’s results are management systems. It is recommended occasionally that some things are not done at reliable. An effective way of satisfying this that a laboratory should have some form of all! The size of this document will depend on requirement is to be compliant with a quality system through which its declared the size of the organisation and on the range National or International Quality Standard, quality policy can be implemented regardless of their activities. Writing the manual will e.g. ISO 90006, EN 450017 (NAMAS M10 of whether the laboratory is formally not be a painless activity but the resulting in the UK)8 or GLP9. Such laboratories will recognised as compliant with a National or documentation should go a long way towards have been assessed by an independent body International Quality Standard. the smooth running of an organisation. It can and shown to be meeting the quality Quality Control (QC) is the set of be compared with running a comb through a standard. It has to be borne in mind that in procedures undertaken by laboratory staff tangled mass of hair; a process that may this context ‘quality’ means ‘fitness for for the continuous monitoring of operations produce a few tears but the result is easier to purpose’. Quality is not necessarily synonymous and results of measurements in order to manage and very pleasing to the beholder. with excellence and for the analyst it is often decide whether results are sufficiently a balance between excellence, time and cost. reliable. Quality Control activities ensure reliable measurements A result with an uncertainty of 1% that has that the procedures used to produce the require the implementation taken a month to achieve is of little value to analytical results are within acceptable of a system for the clinician treating a very sick patient, or a boundaries. For example, check samples of manager requiring to know if a batch of raw known concentration are taken through the assuring quality in material is acceptable. The requirements of whole analytical method along with the the laboratory the client define the fitness for purpose. In samples being tested to check that the same way the quality system has to be fit acceptable results are being achieved. There is often the belief that introducing for purpose. This is why each laboratory has Quality Assurance (QA) is the essential a quality system does not add value but to establish which procedures and standards organisational infrastructure that underpins merely introduces another cost. It is true most closely matches its needs and those of all reliable results. It includes such things as: that setting up a quality system can be very its clients. use of validated methods; calibration of expensive. It is not uncommon to encounter No sector should be excluded from the instruments; use of staff shown to be trained a figure of 20% of turnover where it has need for a formal or informal quality system. and competent to carry out specific tasks; an been necessary to introduce new QC and Reliable data are achievable through a environment and equipment suitable for QA procedures as well as documenting combination of sound analytical science and making the measurements; record keeping existing procedures. However, no matter the application of quality assurance

6 VAM BULLETIN FOCUS ON PRINCIPLES principles. It cannot be emphasised too REFERENCES systems – model for quality assurance strongly that, no matter how good the in production and installation; 1. ‘VAM principle 3 – Qualified and science, reliable measurements require the ISO 9003:1994, Quality systems – competent’, VAM Bulletin, 14, 5–7, implementation of a system for assuring model for quality assurance in Spring 1996. quality in the laboratory. It is equally production and installation). applicable to an industrial research and 2. ‘VAM principle 1 – An agreed 7. EN 45001, General criteria for development laboratory, or a university requirement’, VAM Bulletin, 15, 8–10, the operation of testing laboratories, laboratory as it is to a testing laboratory. The Autumn 1996. ISBN 0 580 17939 7. type and extent of the documentation may be different but the essential elements of 3. ‘VAM principle 2 – Fitness for purpose’, 8. M10, NAMAS Accreditation standard – quality control and quality assurance should VAM Bulletin, 16, 9–10, Spring 1997. general criteria for competence for be covered. It will be most effective when a calibration and testing laboratories. quality culture exists in the organisation. 4. ‘VAM principle 4 – Independent The processes become second nature, assessment of laboratory performance’, 9. The OECD Principles of Good striving for improvement is the norm and each VAM Bulletin, 17, 8, Autumn 1997. Laboratory Practice, Environmental person takes responsibility for their results. Monograph No. 45, OCDE/GD(92)32, 5. ‘VAM principle 5 – Comparable Standards are not just measures of efficiency, Organisation for Economic Co- analytical measurements’, VAM Bulletin, productivity and quality. They are also about operation and Development, 1992. 18, 5–6, Spring 1998. the point and purpose of what we do. The six VAM principles are a package 6. ISO 9000 series of standards with each one making its own contribution. (ISO 9001:1994, Quality systems – To achieve measurements fit for purpose we model for quality assurance in design/ need to apply them all, along with a good development, production installation helping of common sense. and servicing; ISO 9002:1994, Quality

CONTRIBUTED ARTICLES Customers are more concerned about price of analysis than quality…aren’t they?

Summary is it to the end users of the data? More to the Paul point, are end users prepared to pay for the This article establishes the concept of high cost of analytical accuracy? Berryman, ‘critical success factors’ for laboratories. It explores whether laboratory heads and their are end users prepared customers attach the same degree of to pay for the high cost of Hampshire importance to valid analytical measurement. Research in the Public Analyst sector is analytical accuracy? County discussed, in particular looking at the quality Modern laboratories spend huge sums of – delivery – price dilemma. The LABKEY money each year on quality control systems Council model is proposed as a tool for solving to ensure that analytical results are of the this predicament. highest accuracy and precision. Individual Scientific tests are replicated, compared to results Introduction found on certified reference materials, Service The reliability of analytical results is checked and double checked. Laboratories always of utmost concern to the producers of seek UKAS accreditation to show to their analytical measurements but how important customers that they are regularly inspected

7 VAM BULLETIN CONTRIBUTED ARTICLES by third parties. Many laboratories now use It particularly examines the quality controlled, whereas the latter are business proficiency testing (PT) schemes as a versus price dilemma and suggests a novel competencies that can be built up in fundamental part of their quality systems. approach to ensuring that laboratories response to environmental changes. The list goes on. exceed client expectations on both fronts. The study Critical Success Factors Although a literature survey showed that Critical Success Factors (CSFs) are the CSFs were commonly used in many limited number of areas in which satisfactory industries, including banking, manu- results will ensure successful competitive facturing, engineering and computing, there performance for the individual, department or was a distinct lack of any reference relating organisation. CSFs are the few key areas where to the contract scientific service industry. ‘things must go right’ for the business to flourish The main aim of the study was to and for the manager’s goals to be attained. establish the CSFs for scientific laboratories The CSF concept has been used in a involved in the analysis of food. It was multitude of ways but basically falls into two then intended to use this to develop a Is reliability of measurement a critical success main categories: novel strategic planning framework for use factor? • as a tool for identifying Management by this sector. However, does the customer give a hoot Information Systems Literature research and interviews with about all this? Is reliability the most critical • as an integral part of the strategic laboratory heads and client groups revealed factor in ensuring the success of the planning process. a bewildering array of CSFs. These were laboratory or is the client more interested in This study concentrates on the sorted into groups based on laboratory when the report arrives, how much the work latter category. functions and are presented in Figure 1. cost and what colour paper it is printed on? Research has identified over 50 critical Research identified four main types of These were the sorts of questions I asked success factors for various industries and laboratory involved in food analysis: when formulating a research proposal to examples include: sales volume, profit, • local government laboratories discover what factors were critical to success image, quality, political stability, economic • central government laboratories in the contract scientific service industry. factors (e.g. interest rates, inflation), • quangos and research associations This article introduces the concept of legislative changes and technological change. • private sector laboratories ‘Critical Success Factors’ (CSFs) and Note that there is an inherent difference In the first instance, it was decided to summarises the main findings of my PhD between environmental CSFs (e.g. interest carry out a pilot study using the Public research thesis which investigated the rates, legislation changes), and internal Analyst part of the local government sector. importance of CSFs in strategic planning CSFs (e.g. quality, delivery, price). The This was chosen because there within Public Analyst laboratories. former can be monitored, but not is a clearly defined number of Public

Laboratory Functions SCIENCE & SERVICE LABORATORY CUSTOMER CUSTOMER TECHNOLOGY DEVELOPMENT OPERATIONS COMMUNICATIONS SUPPORT Modern Quality systems Capacity Prices Expert witness instrumentation capability New methodology Service design Vertical integration Terms Interpretation (e.g. sampling, of results prosecuting) R&D resources Accreditation Analytical portfolio Marketing & Complaints handling sales expertise procedures Modern IT & LIMS Reliability of results Skilled scientists Channels Courier system Critical Success Innovation Size & critical mass Flexibility Promotions Financial systems Factors & campaigns Capacity Geographical Image & reputation Reporting systems location Specialisms Staff qualifications Access to technical Emergency handling expertise procedures Strategic vision Productivity Personal Partnerships & relationships networks with other labs. Staff training & Project Independence development co-ordination Turnaround times Regular review meetings Figure 1: Laboratory success factors

8 VAM BULLETIN CONTRIBUTED ARTICLES

Analyst laboratories (33 in all), and money for a service, you don’t expect to get • To laboratory heads: “What are the three most ‘Critical Success a clearly defined local government the wrong answers! Factors’ for your type of business? customer (the Trading Standards Nevertheless, this was deemed critical Please rank your final selection 1, 2 department of each authority). because results needed to stand up to cross and 3 (where 1 = most important).” Two laboratory heads and ten senior examination in a court of law. Contrast this Trading Standards Officers were with results required for internal quality • To Chief Trading Standards Officers: “If you were selecting a interviewed. The information obtained control in a food factory, where timeliness Public Analyst Laboratory, what enabled the formulation of separate but and trend spotting are more crucial than would be the three most important linked questionnaires for each. The Public absolute accuracy. criteria for selection? Please Analyst questionnaire was sent to 26 However, second most critical factor indicate in the following table laboratories. The Trading Standards Officer from a customer viewpoint was ‘expert whether each factor is critical or questionnaire was sent to 130 Trading witness capability’. This supports the not critical. Then rank your top Standards Officers. viewpoint that results must be reliable three 1, 2 and 3. (where 1 = Figure 2 summarises the key question in enough to stand up in court but must also be most important).” both questionnaires that asked decision presented in a convincing way. In contrast • The list of CSFs makers in both laboratories and Trading laboratories put this aspect almost last on • Price Standards departments to prioritise a list their list! (11th). • Interpretation of results of factors that could be deemed critical Customers deemed turnaround time and • Accreditation to success. price next most important whereas • Personal relationship with client laboratories felt accreditation was crucial. • Geographical location Results Perhaps not surprisingly, clients listed • Reliability of results The response rate was high with 11 accreditation as 7th. Was this because, • Turnaround time (42%) of the laboratories replying and unlike EN ISO 9001 registration (which puts • Qualifications of staff 40 (31%) of the Trading Standards the needs of the client and regular contract • Technical innovation departments replying. Figure 3 summarises review firmly at the centre of the quality • Marketing and sales expertise the main findings. system), UKAS accreditation almost ignores • Independence First the good news! Both laboratories the customer in favour of technical detail in • Expert witness capability and Trading Standards departments listed methods, calibration systems and training • Productivity ‘reliability of results’ as their number one procedures? (Inputs versus outputs?) • Strategic vision priority. An interesting rider to this was that A particularly worrying result was that • Other (please specify) many of the Trading Standards respondents laboratories ranked ‘technical innovation’ pointed out that reliability of results was last on their list! Perhaps all extra capacity is Figure 2: The questionnaire taken for granted. After all, if you pay good being used on accrediting current methods

Public Trading % of total Analyst Standards % of total score Score ranking Critical success factor ranking Score score 8.8 12 4 Price 4 57 10.8 9.6 13 3 Interpretation of results 6 42 7.9 8.8 12 4 Accreditation 7 24 4.5 7.4 10 6 Personal relationship with client 10 15 2.8 5.1 7 8 Geographical location 12 11 2.1 19.1 26 1 Reliability of results 1 124 23.4 13.2 18 2 Turnaround time 3 72 13.6 2.9 4 12 Qualifications of staff 8 23 4.3 0.7 1 15 Technical innovation 13 6 1.1 2.2 3 14 Marketing & sales expertise 15 0 0.0 6.6 9 7 Independence 5 51 9.6 3.7 5 11 Expert witness capability 2 73 13.8 4.4 6 9 Productivity 11 13 2.5 4.4 6 9 Strategic vision 14 2 0.4 2.9 4 12 Other 9 16 3.0

100.0 136 529 100.0 Figure 3: Main findings of study

9 VAM BULLETIN CONTRIBUTED ARTICLES

data. However, subsequent studies within As a result of my study, I have developed private sector laboratories and end-users the LABKEY (Laboratory Users Key indicate that price and delivery play a much Requirements) model to customise bigger part in the equation, with reliability of laboratory services based on customer results taken for granted. There is also much perceptions of CSFs. This model has more emphasis placed on fitness for purpose resulted in a large increase in market share where screening and semi-quantitative for Hampshire Scientific Service, increasing testing is adequate for routine quality control its number of Public Analyst appointments purposes. This approach would not be from three in 1994 to fifteen and rising in acceptable for enforcement purposes. 1998. Figure 4 represents the LABKEY

Are staff skills critical to success? In the water industry, throughput of strategic planning model showing the links samples is the critical success factor. One between laboratory management activities, with little time left for developing new ones? major water laboratory did not rank laboratory operations, the customer and the Customers also ranked this very low, but I reliability of results as critical at all! Due to external environment. would suggest that as accreditation becomes the relative simplicity of the water sample In summary, the LABKEY model the norm (offering no competitive matrix when compared to say food, there is contains six crucial activities: advantage) laboratories offering new services more importance in having dedicated 1. POLICY – setting a vision, mission, and products would steal a march on the automated equipment for each test values and key corporate objectives laid competition. parameter and very high sample throughput. down in a Business Plan. Last on the list for customers were For foods, smaller batches of samples dictate 2. COMMUNICATIONS – with ‘strategic vision’ with two points and flexible use of expensive equipment such as customers (to establish their CSFs) ‘marketing and sales expertise’, scoring no HPLC, GC, GC-MS and ICP. This is why and External Environment (to points at all! This set me thinking about the smaller contract laboratories find it virtually take account of scientific, legal, advisability of carrying out a PhD in impossible to compete with the large water political, technological, social and strategic planning for laboratories! authority laboratories when it comes to economic factors). ‘LABKEY’ – A model for water analysis. 3. CONTROL - the CSFs are used to build solving the Quality-Delivery- However with the advent of ‘Next Steps up the LABKEY Agreement which Price dilemma Agencies’ and privatisation of central dictates performance criteria for Laboratory government laboratories, compulsory Operations. Other control mechanisms This study showed that in the Public competitive tendering and ‘best value’ such as terms and conditions of Analyst sector, reliability of results comes regimes in local government sectors, price, is employment, health & safety procedures, first from the perspective of both the becoming more and more the Critical quality systems and statements of policy producer and the end user of the analytical Success Factor. are grouped together as ‘rules’.

Laboratory Management

POLICY (1)

Liaison and Marketing COMMUNICATIONS (2) Research and Networking

CONTROL (3)

Customer External Environment LABKEY RULES

AUDIT (5) (6) COORDINATION

LABORATORY OPERATIONS Day to day liaison (4)

Emergency Services Research and Networking

Figure 4: The LABKEY Model

10 VAM BULLETIN CONTRIBUTED ARTICLES

4. LABORATORY OPERATIONS – For example, in response to the above …and finally The staff, equipment, operating systems survey all scientists and technicians at and services of the laboratory. Covers all Hampshire Scientific Service are trained in So coming back to my original analysis, consultancy and fee earning courtroom technique. They are shown how question… do customers care more about operations. Translates customer to construct statements of witness and how price than quality? requirements into outputs in line with to explain ‘reliability of measurement’ to the Well, no. Not in the case of Trading the LABKEY Agreement. layperson. Furthermore, court appearance Standards departments anyway. But with 5. AUDIT – sporadic checks by manage- and follow up statements can be offered free increasing pressure on Local Authority ment to ensure that Laboratory of charge as part of a LABKEY Agreement. budgets, price is becoming more important. Operations are being carried out in This focuses on one of the customer’s key Laboratories need to find ways of ensuring accordance with (4), e.g. financial, requirements and removes the worry of extra quality and driving down costs. Alternatively quality and delivery audits. expenditure from enforcement authorities laboratories need to satisfy end users that it 6. CO-ORDINATION – mechanisms when considering prosecutions. is worth paying more for high quality results. to ensure that work for different clients Some clients require very fast This can only be done through a dialogue is timetabled correctly and that turnaround. For example, Port Health between the laboratory community and end resources are available to satisfy the Authorities may detain consignments of users of analytical data to establish CSFs terms of the LABKEY Agreement. foods from third world countries pending within each market sector. Includes sampling plans, reporting analytical results. These results must be The next step is to convince systems, courier system and resource turned around within 6 days. In some allocation plans. cases screening tests can be used. In other end users that valid The LABKEY model uses a service level cases higher charges are made to enable analytical measurement agreement approach which itemises key queue jumping. performance indicators in terms of quality, Each LABKEY Agreement is can save money delivery and price. It lays down expectations customised to ensure the optimum balance Until recently, the VAM programme has of the laboratory and customer in a written between quality, delivery, price or some concentrated on educating producers of agreement after a series of conversations to other key customer requirement. analytical data about the importance of establish critical success factors from the All this sounds pretty obvious…but it’s reliable analytical measurement. The next customer’s viewpoint. This document is the surprising how many laboratories… step is to convince end users that valid LABKEY Agreement. Figure 5 shows the • try to gain accreditation for analytical measurement can save money and main components of the LABKEY an unrealistically wide portfolio of prevent business mistakes. This is a much Agreement. tests rather than striking up partnerships harder task than convincing a group of like with other laboratories having minded scientists that VAM makes sense! • Overall contract value complementary portfolios The 1997–2000 VAM programme has • Itemised work schedule • take part in proficiency schemes for tests end user education as one of its key • Guaranteed turnaround times that are never carried out for clients! objectives. Perhaps the LABKEY model • Consultancy and advice • take over 2 months to turn around could prove useful? • Payment terms samples and then wonder why • Quality criteria clients grumble • Review mechanisms • produce excellent, highly accurate • Performance indicators results but cannot explain their results in court Figure 5: • offer tests at such a high price so that no The LABKEY Agreement one can afford to use the facility.

11 VAM BULLETIN CONTRIBUTED ARTICLES Industrial analytical measurement – the reality

Albright & Wilson, Courtaulds, Croda Type of quantitative analysis Steve Groome, Chemicals, Hickson & Welch Ltd, LGC, Assays 36% Thomas Swan and Vickers Laboratories. Impurities 23% Albright The group was set up to represent SMEs, Other chemical analyses 18% although some of the larger companies test Physical measurements 18% the definition of ‘medium-sized’. This group & Wilson Other 5% decided it would carry out some Average total number of benchmarking on analytical error and method quantitative analyses per month: 8000 Introduction validation to see how the VAM principles “ nalytical data is the language of the were being practically applied in SMEs. AChemical Industry”. When I was told this Seven of the companies participated in Sample types by the Analytical Manager of one of Britain’s the study; it was agreed beforehand that Inorganic chemicals 20% largest chemical companies I had my doubts. because LGC is a contract analysis Organic chemicals 58% I cast my mind back to a myriad of company laboratory rather than a chemical company, Polymers 12% reports, each referring to profit and loss it would not take part. Waste 2% accounts, sales volumes, contribution and Benchmarking – the process of Environmental samples 7% cash flow – not an analytical measurement identifying and learning from best practices Biological samples 1% amongst the lot of them. anywhere in the world – is a powerful tool in the quest for continuous improvement. The Solids 46% Benchmarking…is a SOCSA benchmarking was carried out by Liquids 44% powerful tool in the quest for sending a questionnaire to the seven Gases 10% continuous improvement participating companies. The questionnaire was split into three sections. The first section Analytical techniques (time basis) Then I thought about it in more practical was used to profile the analysis performed in terms. We sell chemicals. No-one buys a each of the Group members’ departments, the Chromatography – GC 22% chemical unless they know what they are second section covered method validation and Chromatography – LC 17% buying. No-one makes or sells a chemical the third covered analytical error. As agreed at Chromatography – other 6% unless they know it’s the right stuff, of the the start of the exercise, confidentiality has Molecular spectroscopy – UV/Vis 5% right quality and their process is safe and been maintained and none of the companies Molecular spectroscopy – other 2% under control. We couldn’t do any of these are individually identified in this article. Atomic spectroscopy – AA/ICP 7% things without specifications, certificates of Atomic spectroscopy – other 0% analysis or in-process measurements. We Section 1: Titrimetry 18% back up advertising claims with analysis, we Department profiles Colorimetry 9% define our patents using analysis and we are The company representatives were Gravimetry 2% regulated by others using analysis. involved in a variety of analytical Microscopy 1% So if analysis is so important, it must be departments with a variety of functions, Physical methods 8% equally important that we get it right. In ranging from quality control to R&D. The Others 3% particular we need to ensure that our average of all the profiles was as follows: analytical measurements are fit for the It may be concluded that a typical purpose that they are intended. The DTI’s analysis is a QC assay determination on an Sample sources VAM programme is aimed at helping us to organic liquid or solid using GC or HPLC. get our analyses right. Production (QC) 58% The Specialised Organic Chemicals Sector Production (Problems) 9% Section 2: Method validation Association (SOCSA) set up an analytical R&D 19% There are several general points that networking group following initiatives from Marketing/Technical service 5% may be made. LGC under the DTI’s Analytical Innovation Environmental/IPC 6% • All participants agreed that a suitable programme. This Group is known as SANG Other regulatory work 3% definition of validation is ‘The process by (SOCSA Analytical Networking Group) and it Average number of which it is established, by laboratory studies, samples per month: 2300 includes representatives from A H Marks, that the performance characteristics of the

12 VAM BULLETIN CONTRIBUTED ARTICLES

method meet the requirements for the and others doing 5–10 different • There was general agreement in the level intended applications’. concentrations. of analytical error deemed to be • Nearly all of the seven companies said • There was general agreement that 5 acceptable for different methods. The they worked to ISO9000, but few concentration levels should be used to average figures were: worked to NAMAS, GLP or GMP. determine linearity. Assay method Only one company specifically used • Approaches to robustness were very variable. (nominal 95.0 %w/w) 1.5% or ± 0.75% VAM (see later discussion). Generally it seemed to depend how critical • Nearly all had procedures for method the method was, but some companies Trace impurity method validation but few had procedures for said they never determined robustness. (nominal 100 ppm) 13% or ± 6.5% method review or revalidation. • Most used a single internal or external • There were a variety of ways to decide • Validation of analytical methods is control standard in methods where a whether a result was ‘out of spec’ and a variable. When asked to give an indication standard was appropriate. variety of actions were taken as a result. of the percentage of their methods subject • Most companies did not use the results The general consensus was to reanalyse to each aspect of validation the results in of analytical validation in the setting of and then to resample and reanalyse if the Figure 1 were obtained. product specifications. result was still out of spec. This would • Only some companies used system Analytical error generally be done in duplicate. The suitability checks. Most did not. accepted result would be an average • Most used 10 repeats when performing a Several general points may be made. which might or might not include the precision experiment, but the number of • Each person was asked to define total original result. repeat experiments varied from 3 to 15. analytical error. A variety of responses • Most companies did not use statistical • When determining the recovery of a were forthcoming but they all conveyed methods for the detection of ‘outliers’. method, a variety of different concen- a similar meaning. Most people regarded Where they did, the Dixon’s Q test trations were employed with some doing analytical error and measurement was used. the experiment at only one concentration uncertainty as two different things. Discussion Company Benchmarking exercises of this nature Test A B C D E F G inevitably give variable results because the Precision/Repeatibility 100% 10% 100% 50% 30% 20% 25% situation within one company is always Accuracy/Recovery 10% 10% 80% 30% 20% 40% different to that in another. In this case there Linearity/Range 100% 10% 100% 50% 30% 10% 20% was some similarity in that all companies LOQ (Limit of Quantitation) 1% 30% 30% 40% were members of SOCSA and consequently LOD (Limit of Detection) 1% 50% 30% 20% 40% most were involved in organic analysis. Most Specificity/Selectivity 90% 100% 30% 5% 60% laboratories were small or medium sized. Ruggedness 100% 10% 30% From post benchmarking discussions it Robustness 100% 30% is obvious that there is a good knowledge of practical analytical chemistry in all the % Use of Validation in Analytical Methods companies. Although most said they do not 120 A specifically use VAM, they all apply the 110 VAM principles to a greater or lesser extent. 100 B Most had a knowledge of method 90 validation and why it was important, but the 80 C benchmarking data suggests that many analytical methods are not validated 70 sufficiently for users to be fully confident 60 D that they are fit for purpose. 50 All analytical departments in SMEs are E 40 under great time and resource pressure and 30 analysts are likely to do the amount of 20 F validation their customers are prepared to 10 pay for. This may not be consistent with 0 G their professional judgement. Perhaps VAM is directed too much at the analyst and not LOD LOQ

Range enough at his customers? Linearty/ Recovery Precision/ Accuracy/ Selectivity

Specificity/ Most companies did not use the results Robustness Ruggedness Repeatability of analytical validation in the setting of Figure 1: Use of validation in analytical methods specifications. This may be a problem with

13 VAM BULLETIN CONTRIBUTED ARTICLES the way chemical companies are structured. product to be within specification. If it to a poorly validated method will they attach Very often a specification is agreed between proves not to be, it would seem reasonable a realistic value to the validation process. the customer and a salesperson before the to check any result. The way in which this is analyst becomes involved. There is a done could easily be the subject for another Steve Groome is a Project Manager in growing trend for sales people to be trained article on its own. Albright & Wilson’s International Technical in chemistry but it does not necessarily Conclusions Centre in the West Midlands. He is responsible follow that they are trained in analytical for analytical science, a project team working on method validation. Sometimes the analyst is This was the first time the SOCSA chemicals for water treatment applications, and presented with the customer’s draft method Analytical Networking Group had also for regulatory affairs. The Analytical Group and sometimes he has to invent one for participated in a benchmarking exercise and is 15 strong and provides a service to the himself. It is not unknown for the error of it was well worthwhile. We learned that we company world-wide, but primarily to the R&D the method to be wider than the have some good analysts who often work and pilot plant activities on site. specification for that measurement. under difficult resource and time pressures. As well as SANG he is also a member of The question of out of specification We are conscious of good practice in method LABS-NW, a similar Group which covers larger analyses has always been a difficult one validation but often find it difficult to apply. chemical companies in the North West of England. which has become a major issue for the There is a need for our customers, He has represented LABS-NW in the Analytical pharmaceutical industry. In the chemical internal and external, to be educated in the Science Forum, a group working under the industry, if a process is in control it is value of the information they receive and act Foresight banner, which brings together all the normally quite reasonable to expect the upon. Only when they learn to link a problem disparate analytical interests in the UK.

DNA chips – a taste of the future

analysis, most significant of which have been by multinational (pharmaceutical) sponsors, Helen Parkes, the advances made in miniaturisation and and are currently primarily targeted at automation of the diagnostic processes and research applications. LGC the development of microarrays commonly referred to as ‘DNA chips’1. DNA chip What is a DNA chip? he VAM DNA technology project technology, is the bio-molecular equivalent An array is an orderly arrangement of Tundertaken at LGC has a provision to of a computer microprocessor. Hailed as DNA target sequences which can be identify, research and investigate technical ‘Star-Trek’ technology, microchips are used interrogated with a DNA probe. In general, developments which are set to impact upon on to which 10,000–100,000 DNA fragments arrays are described as macroarrays or the rapidly evolving area of bioanalysis in the are immobilised containing different microarrays, the difference primarily residing not-too-distant future. combinations and permutations of sequence in the sample size applied to the array. One such exciting area is the potential variation. The DNA sequence, length and Microarrays usually consist of thousands of application of DNA microarrays, or DNA location of each oligonucleotide within the DNA ‘spots’, typically less than 200µm in chip technology, which have been developed array is known. Sample DNA, for example from diameter, arrayed on a microchip, which to accommodate the technological challenge an individual, can then be ‘matched’ against require specialised robotics and imaging of high throughput genetic analysis. Existing these variants by hybridisation analysis. equipment for hybridisation analysis. These DNA technologies were not specifically These DNA microarrays permit the are generally referred to as ‘DNA chips’. designed to cope with the analytical rapid interrogation of information expressed demands of detecting and identifying the in different cell types or cell states, enabling Current applications complex interactions between multiple hybridisation based expression monitoring, Gene expression analysis has, to date, genetic loci and numerous sites of DNA polymorphism detection and genotyping on been the main application of microarray mutation which characterise common health a genomic scale. DNA chips will therefore analysis, with the major driver being disorders such as cardiovascular disease provide unprecedented access to key areas of ‘functional genomics’. Extensive research is and cancer. human health, including disease prognosis being carried out into the function of the The International Human Genome and diagnosis, toxicology and drug discovery. genes identified as a result of the Human Mapping Project has fuelled the recent To date these high cost, high technology Genome Mapping Project. DNA and oligo notable advances in high throughput solutions have been generally heavily backed chips have been used to measure expression

14 VAM BULLETIN CONTRIBUTED ARTICLES levels of genes in plant2, bacterial3, yeast4–6, murine7 and human samples8,9. Identification of DNA polymorphisms and mutations is a major potential application for DNA microarrays in genetic analysis. By hybridising a labelled target DNA to a microarray of oligonucleotides containing all possible sequence deviations, such analyses can be performed extremely rapidly. The use of DNA chip technology for mutation detection has been reported in a number of studies10 including HIV, cystic fibrosis, exon 11 of the BRCA1 breast and ovarian cancer gene, genes encoding the cytochrome p450 enzymes and ß-thallassemia. DNA microarray production The challenge in production of a DNA microarray is in laying down DNA quickly, with sufficient fidelity and reproducibility11. Three main categories of DNA microarray Figure 1: Microarray deposition technologies13 production are emerging employing the deposition technologies illustrated in Figure 1. Different approaches to microarray production – • In-situ (on chip synthesis) of oligonu- 1) photolithography 2) mechanical microspotting and ink jetting cleotides or peptide nucleic acid (PNA) – two approaches have been adopted. • Spotting of DNA fragments – cDNAs are synthesised, stored and separated before i) Photolithography, or light directed (complementary DNAs derived from being individually spotted onto arrays are synthesis of oligonucleotides, in which messenger RNA) are spotted in an array ponderous and necessitate the use of successive bases are photoactivated in pattern directly onto the chip surface. complex high throughput robotics. accordance with an order determined by This is cheap and offers flexible assay DNA microarray design a photolithographic mask. Successive design. Typically glass slides are overlaid rounds of deprotection and chemistry with a positively charged coating e.g. The design of the microarray assay yield synthesised oligos of up to 30 bases poly-L-lysine or amino silane, and cDNA depends on the application. Gene expression in length. Commercially synthesised ‘off fragments, contained in a denaturing monitoring experiments can employ microarrays the shelf’ chips may be available through solution, are printed directly onto the in which sets of oligonucleotides are this route. surface using mechanical microspotting. constructed that identify unique motifs in ii) Controlled oligonucleotide synthesis, genes. Alternatively, spotted cDNA microarrays in which piezoelectric ‘ink jet’ printers Advantages of the in-situ synthetic are used which can determine a time course sequentially deposit successive reagents methods include the direct manufacture of of expression for thousands of genes or for oligonucleotide synthesis, deprotection chips from sequence databases, removing differential gene expression by the competitive and washing. Successive cycles of the need for sample handling and tracking, hybridisation of two labelled samples. synthesis yields oligonucleotides or PNA and the use of synthetic reagents minimises Typically, parallel analysis of gene expression14 of up to 50 bases in length. chip-to-chip variation by ensuring a high is carried out by isolating total mRNA from • Arraying of prefabricated oligonucleotides degree of precision in each coupling cycle. cells and hybridising it to an array of or PNAs – typically, oligonucleotides or The complexity and expense of photo- molecules that represents all the genes of an PNA are synthesised by traditional lithography technology means that chip organism. Through analysis of the binding means and printed onto the array using manufacture must be carried out ‘in-house’ positions and signal intensities, the relative ink jetting or mechanical microspotting. by companies such as Affymetrix, although levels of gene expression under different An alternative approach of ‘electronic this does result in commercial availability of conditions/tissue states can be determined. addressing’ has been developed by ‘off the shelf’ chips. The key advantages of Genetic mutation and/or DNA Nanogen12 in which biotinylated oligo- deposition of presynthesised DNA polymorphism analyses require exact nucleotides are directed to individual oligonucleotides, PNA or cDNA include: sequence information. Genotyping or spots by polarisation of the charge at greater flexibility in experimental design, polymorphism screening chips are essentially that spot and then anchored in place via a ease of prototyping, low cost and longer very large arrays of allele specific streptavidin-containing agarose permeation probe length possibilities. However, these oligonucleotides. Every nucleotide position layer that covers the chip surface. approaches in which the cDNAs or oligos in the gene exon or mutational ‘hot-spot’

15 VAM BULLETIN CONTRIBUTED ARTICLES under study has to be interrogated. Typically Target sequence: 5’…GGGCTTCATACAACAGTGCCAAA…3’ a probe ‘tiling’ strategy is adopted in the “A”probe: 3’- AAGTATAT TGTCA -5’ microarray15 (see Figure 2). A ‘frame’ is “C”probe: 3’- AAGTATCT TGTCA -5’ selected around a section of the target “G”probe: 3’- AAGTATGT TGTCA -5’ (correct match) sequence, representing the perfect match “T”probe: 3’- AAGTATT T TGTCA -5’ allele complementary to the wild type probe.

For each interrogation site a set of four Target sequence: 5’…GGGCTTCATACAACAGTGCCAAA…3’ probe features are generated that are position: n AAGTATGTTGTCA complementary to the frame but with a n+1 AAGTATGTTGTCAC central nucleotide substitution. A second set n+2 AAGTATGTTGTCACG of four probe features are generated by translating the frame by one base and Figure 2: Probe tiling strategy for sequencing arrays carrying out the substitutions to query the next interrogation site. The relative hybridisation intensities of each series of Fluorescence labelled probes at a particular location identifies the nucleic acids correct base. Figure 3 illustrates the general principal behind both oligonucleotide and DNA microarray hybridisation analysis. DNA microarray assays A microarray assay could typically consist of the following experimental steps. 1. Sample preparation which includes hybridisation isolation and, typically, fluorescent labelling of the probe target DNA. 2. Hybridisation, in which matching of the Specific hybridisation to an Data analysis by labelled target probe DNA to the chip- TACGGTCTACAGT TACGGTCTACAGC ATG CCAGATGTC C TACGGTCTACAGG TACGGTCTACAGA TACGGTCTACA TT ordered matrix of sequence fluorescent pattern representatives recognition based DNA sequences, with the formation of a duplex, takes place. Figure 3: General principle of oligonucleotide-chip technology11 3. Detection of target probe hybridised to Sample nucleic acids are exposed to an ordered grid of array-bound oligomer molecules. At chip-based sequences. Currently, samples each position, a distinct sequence is attached to the support medium, enabling detection of for hybridisation are predominantly single base differences. Subsequent to specific target hybridisation with fluorescently labelled fluorescently labelled requiring optical probes, analysis of the pattern fluorescent signal colour and intensity reveals mismatch detection. Duplex formation is analysed hybridisation indicative of for example, polymorphisms. by measurement of relative fluorescence intensity. Fluorescent scanners are commercially available, and are being GeneChip CYP450 Hybridized Probe Cell Probe Array developed by equipment manufacturers Single stranded, fluorescently along with data analysis software labeled DNA target specifically for DNA chip applications. Oligonucleotide probe 50µm For example, Molecular Dynamics and Hewlett Packard offer scanning confocal Each probe cell contains microscopy instruments that can detect millions of copies of a specific thousands of probes hybridised to the oligonucleotide probe chip surface at micrometre resolution. 5.25 mm Current status Analysis block

The first generation of chips now Over 8,000 different probes complementary to sense and reaching the market are still primarily for anti-sense p450 genomic DNA Probes range in length from research applications. Affymetrix, a leading 18mer to 22mer US company in DNA chip technology, Images of Hybridized Probe Array has developed proprietary GeneChip® ® oligonucleotide probe arrays that are Figure 4: A schematic of the Affymetrix GeneChip assay for manufactured using a high resolution cyp450 mutations (image courtesy of Affymetrix, Inc. (Santa Clara))

16 VAM BULLETIN CONTRIBUTED ARTICLES photolithographic fabrication process applications for a number of years. intrastrand duplex formation, or the adapted from the semiconductor industry. Hybridisation detection is by dual label formation of 3° structures in DNA in the Affymetrix is developing GeneChip® arrays fluorescence analysis. The site outlines parts, sample. Hoheisel claims “this is a big for HIV mutations, P53 cancer gene costs and construction details for the arrayer issue but no-one is acknowledging it”. mutations, and P450 gene mutations (Figure and scanner, in addition to protocols for Different sequence oligonucleotides can 4) in drug metabolising enzymes. probe preparation and hybridisation analysis. have large differences in hybridisation Affymetrix intends to set the industry This ‘DIY’ approach has the advantages of: efficiency and stability, due to different standard for commercial microarray • relative low cost base sequence composition (G-C content) technology. The company together with • great flexibility in application leading to different melting temperatures Molecular Dynamics, is a founder member • broad availability without patent problems for hybridisation. Therefore special of the Genetic Analysis Technology The most significant drawback could software is necessary to reliably interpret Consortium (GATC), created to ‘provide a potentially lie in the extreme variability of all the data obtained16. PNA may be of unified technology platform to design, process, parameters, leading to decreased validity. greater value in microarrays for read, and analyse DNA arrays’. Validation of promoting greater duplex stability, in An alternative commercial technology is microarray technology addition to allowing hybridisation to the Nanogen bioelectronic chip. Nanogen, a occur in low ionic strength conditions US company, uses controlled electric fields In reviewing the current status of DNA which disfavour the formation of 2° and to immobilise prefabricated oligonucleotides chip technology it is encouraging that 3° structures. to spots (microelectrodes) on the validation issues have already been raised. • Detection issues – Fluorescence measure- microarray. The Nanogen technology allows Several concerns have been identified. ments of hybridisation patterns on the small sequences of DNA capture probes to • Standardisation issues – A need to array do not give any indication of the be electronically ‘addressed’ to specific sites determine suitable reference standards specificity of the binding event. Ideally on the chip. DNA has a strong negative with which to compare results was internal quality standards should be charge, and can be electronically moved to highlighted in a recent article1. In the designed to confirm that fluorescence an area of positive charge. Microarrays can US, the National Cancer Institute from a particular location is due to an be produced through the sequential (NCI), NHGRI and National Institute of exact binding match, not to a mismatch, positioning of DNA probes through Health (NIH) have stated a commitment a combination of the two or even non- electronic activation of sites, row by row. to the verification of a validated set of specific binding due to a faulty protocol Nanogen also exploits electronically sequenced human cDNA clones to be or operator. mediated active hybridisation. Rapid used as scientific community standards. • Quantitation issues – Quantitation of hybridisation is promoted through the • Microarray production issues – Sharing gene expression/signal as monitored by electronic concentration of sample DNA at of the technical information for cDNA/polylysine chips is affected by a each test site. Stringency control is also microarray and scanner production, and non-linear relationship between signal achieved electronically through the reversal the protocols for sample preparation and strength and the amount of target probe of the site polarity forcing unbound or non- hybridisation amongst scientists present, resulting in an underestimate of specifically bound sample DNA back into encourages uptake, but controls and the extent of large changes in expression, solution away from the capture probes. The standards should also ideally be made especially between abundant species of Nanogen ‘bioelectronic’ chip has several available to allow the interlaboratory mRNA. The lower end of the expression potential advantages over ‘passive’ chips. comparison of the performance and spectrum is also subject to distortion, Greater precision, control and accuracy of efficiency of the equipment and assay. which is influenced by the type of probe array and hybridisation may be achieved. There has also been a question over the used. This is reported as being The WWW site ‘The Microarray shelf-life and stability of the arrayed of particular concern1 as subtle changes Homepage’ (http://cmgm.stanford.edu/ DNA. Stability trials may be required. in genes with low level expression pbrown/array.html) is “designed to provide • Hybridisation issues – One worker in the are as likely to be important in information regarding the production of field, Jorg Hoheisel, has reported a phenotype determination as more microarrays to the scientific community”. This number of concerns with respect to dramatic differences. site is sponsored by the US National Human target hybridisation. After production of • Data interpretation issues – Participants Genome Research Institute (NHGRI), and the DNA chip, this “is the most important at the 2nd Workshop on Methods and encourages fellow research scientists to issue of chip based analyses”14. The issue Applications of DNA Microarray follow the construction of a cDNA of how much of a sample is readily Technology1 perceived the greatest microarray analysis system. This system accessible to hybridisation is not yet challenge facing the field as being “the employs microarrays made by ‘spotting’ resolved, problems such as intrastrand development of efficient methods and cDNAs directly onto the ‘chip’ surface – a 2° and 3° structure can prevent or limit standards for organising, distributing glass slide. This DNA deposition method is binding to the array. The conditions that and interpreting the large volumes simple and has been widely employed in favour good duplex formation on the of data generated by microarray molecular biology laboratories for different chip are also likely to promote experiments”.

17 VAM BULLETIN CONTRIBUTED ARTICLES

Conclusion 3. De Saizieu A, Certa U, Warrington J, 11. Marshall A and Hodgson J, Nature DNA arrays, and microarrays in Gray C, Keck W and Mous J, Nature Biotechnology, 16, 27-31, 1998. particular, have tremendous potential in Biotechnology, 16, 45-48, 1998. 12. Sosnowski R G, Tu G, Butler, W F, genomics and genetic diagnostic applications. 4. Shalon D, Smith J S and Brown P O, O’Connell J P and Heller M J, Proc. As reviewed above, there are many different Genome Res., 6, 639-645, 1996. Natl. Acad. Sci., 94, 1119-1123, 1997. approaches to array production and application. However, the validation issues 5. DeRisi J L, Lyer V R. and Brown P O, 13. Schena N, Heller R A, Theriault T P, identified are, to a certain degree, shared Science, 270, 680-686, 1997. Lachenmeier E and Davis R W, across the different microarray technology TIBTECH, 16, 301-306, 1998. platforms. These issues will need to be 6. Wodicka L, Dong H, Mittman M, Ho M-H addressed in order to ensure the valid and Lockhart D J, Nature Biotechnology, 14. Hoheisel J D, TIBTECH, 15, 465-469, application of this powerful technology. 15, 1-15, 1997. 1997.

7. Lockhart D J, Dong H, Byrne M C, Follettie 15. McGall, G.H. The fabrication of high For further information on the VAM M T, Gallo M V, Chee M S et al, Nature density oligonucleotide arrays for DNA technology project, or to request a Biotechnology, 14, 1675-1680, 1996. hybridisation-based sequence analysis copy of the more detailed report on DNA Ch.2 in Biochip arrays and integrated microarrays, please contact Ginny Saunders 8. Schena M, Shalon D, Heller R, Chai A, devices for clinical diagnostics IBC Tel: 0181-943 7346; Email: [email protected] Brown P O and Davis R W, Proc. Natl. Library Series (1997). Acad. Sci USA, 93, 10614-10619, 1996. REFERENCES 16. Blanchard A P and Hood L, Nature 9. DeRisi J L, Penland L, Brown P O, Biotechnology, 14, 1649, 1996. 1. Editorial ‘Getting Hip to the Chip’, Bittner M L, Meltzer P S, Ray M, et al, Nature Genetics, 18, 195-197, 1998. Nat. Genet., 14, 457-460, 1996.

2. Schena M, Shalon D, Davis R W and 10. Ramsay G, Nature Biotechnology, 16, Brown P O, Science, 270, 467-470, 1995. 40-44, 1998.

CASE STUDY VAM in the brewing industry

situation, or, as in the following, they may is a sophisticated analytical quality system. Ann Mundy, explain how VAM links with the core business. The input parameters for each product – It is hoped that these articles will provide materials, process conditions and analytical Bass Brewers enough detail to be of value beyond procedures – are fully specified because a the industrial sector directly involved, primary aspect of consumer satisfaction is Ltd to a wide range of businesses engaged in known to be quality consistency. Consumers analytical science. want a uniform product, without variation in Summary taste from day to day, between retail outlets Sinéad Doran and between packaging types. Means of n this case study, Bass Brewers achieving this goal include analytical method and John ILtd demonstrates the application of validation, training to ensure competence, the VAM principles to the high volume participation in proficiency testing, and a Francis, LGC brewing industry. formal quality management system. Bass Brewers is an increasingly profitable Benefits of applying the VAM principles The VAM Bulletin now presents the first of a business producing five of the leading UK at Bass Brewers have included the relocation of projected series of six case studies of the beer brands, and including Britain’s best- testing activities to production teams without business benefits of VAM. Companies often selling beer, Carling. The company aims to loss of analytical quality, and an improvement perceive the merit in sharing their own stories produce beers to delight the consumer, and in testing precision leading to improved of the tangible financial and technical benefits this requirement gives a sense of direction in quality and consistency of the final product. of the VAM approach. The resulting studies the choice and application of analytical Perhaps most importantly, quality may be examples of the resilience that VAM methods to the production process. management of analytical data has greatly confers in dealing with a specific analytical Underlying the success of the company reduced the risks and associated costs

18 VAM BULLETIN CASE STUDY of dealing with defective batches of the ‘One Brewery Concept’. The One Table 1. Analytical methods, too, are input product, production delays and unnecessary Brewery Concept aims for the production of parameters for the achievement of an repeat testing. any beer at any Bass brewery without acceptable product. variation in quality or organoleptic (sensory) Individual production sites have the Introduction properties. Currently, keg beers are brewed flexibility to vary in process specifications Bass Brewers Ltd is a part of Bass PLC, at eight different sites across the UK, provided that the final package specification which is active in the provision of a wide range whereas bottled and canned beers are is met. of leisure facilities including hotels, pubs, produced at fewer sites. Consequently, Analytical methods bottles and cans are sold alongside keg leisure venues, beer and soft drinks. The and quality control famous names in the Bass Brewers portfolio beers of the same brand but from a are Carling, Britain’s best-selling beer; different brewery. Under the One The analyses performed to ensure that Caffrey’s Irish Ale; Worthington Draught Brewery Concept, these products must be the product meets its specifications fall into Bitter; Tennent’s Lager, Scotland’s top beer organoleptically identical. three main groups: brand; and Draught Bass, a cask-conditioned The specification system • chemical ale. Bass Brewers’ 1997 increased operating • microbiological profit of £168m was supported by increased As a result of a quality management • organoleptic (flavour and appearance) beer volume and market share combined with approach, the input parameters of the Increasing emphasis is being placed on margin improvements. brewing process are tightly specified – this ensuring that specifications are met Bass Brewers currently employs over encompasses materials, production and throughout the production process 6,000, at eight UK sites. Burton-on-Trent packaging. Suppliers of all ingredients and rather than only at the point of packaging. has been the historic home of Bass packaging materials are required to carry In-line equipment plays an important Brewers since 1777. The Cape Hill out prescribed tests on the materials role in process monitoring, and this (Birmingham) site is highly flexible supplied at the final process point, and undergoes the same calibration and and produces all export products. are audited by Bass annually or at a verification procedures. All these sites produce keg beer; three frequency dependent on the criticality Bass Brewers has a documented and sites also produce small pack (bottles and of the particular material. Among the systematic approach to quality control (QC) cans), and three sites also produce cask ale. parameters of production are the timing, for chemical analysis, having evolved a Cask ale differs from the keg product in temperatures, pressures and fluid handling simple and effective terminology appropriate being unfiltered and unpasteurised, retaining steps for each stage. to the brewing process (Table 2). The the brewing yeasts in final package, but The analytical methods required to criteria of acceptability of the test results are resulting in a relatively short shelf life (20 assure product quality are specified as tightly strongly focused on consistency – that is to days). The short shelf life and risk of as the input material and process say, the results must not merely meet a spoilage has resulted in a move by Bass and parameters. An outline of the analytical predetermined target level, but must fall most other brewers away from cask ale scheme for Carling lager at one particular within an agreed tolerance of the target level. products to brewery conditioned, chilled and stage of production – in package – is given in Control charts are extensively used, and the pasteurised keg ales.

The analytical requirement Original gravity Total carbohydrate Present gravity Calorific value Bass Brewers has a strong perception Final gravity (attenuation limit) Pasteurisation that the consumer prefers a consistent Fermentable residue Nitrogen product, without variation in taste from % Alcohol by volume n-Propanol day to day, between retail outlets or between packaging types. This perception Colour Isobutanol sets up a strong component of the Haze 2-Methylbutanol α requirement for analytical science. Each Bitterness: iso- -acid 3-Methylbutanol product is defined by its own unique Vicinal diketones Total isoamyl alcohol series of analytical, flavour and Carbon dioxide Acetaldehyde microbiological specifications at all stages Dissolved oxygen Ethyl acetate in its production and processing. Total in pack oxygen Isoamyl acetate Consistency also implies, of course, that Foam stability Ethyl hexanoate the product must be stable in storage. Chloride ion Isobutyl acetate There is a structured and documented Sulphate ion Ethyl n-butyrate approach to research leading to the pH Dimethyl sulphide introduction of new product lines. A key driver in the development of Table 1: Analytes specified for Carling Black Label quality assurance at Bass Brewers has been in package (keg, can and bottle)

19 VAM BULLETIN CASE STUDY control chart approach can be extended to Internal audits are used to check that interchanged at random between QA and monitor between-analyst and between- the prescribed methods are understood and production teams. instrument variations. are followed in practice. Audited areas BAPS has proved to be flexible in that it Choice of analytical methods then respond by producing a list of actions reports results for Bass Brewers as a group, for improvement. and provides a similar report for Bass and their validation The requirement for microbiological contract brewers. The concept of a correct Bass Brewers uses analytical methods methods in the brewing industry is result assigned by a reference laboratory, or published by the Institute of Brewing (IOB). principally to identify the correct brewing even by consensus, is less useful to Bass International methodological harmonisation yeast and confirm the absence of undesirable Brewers than the knowledge that their own is a goal, in particular through working with organisms. Recently, adenosine triphosphate product is consistent. Furthermore, the European Brewing Convention. The (ATP) assay by a bioluminescence method differences in instrumentation for IOB methods are documented in a Bass has supplemented conventional culture parameters such as haze result in large Methods Manual in a fuller form for internal methods to give a real time hygiene indicator differences in the reported results between use; for instance, a particular instrument for plant such as fermentation vessels. The companies. However, results generated may be specified for the analysis. case for such rapid methods is not within the company, with known diminished by a sophisticated quality instruments and standard methods, can be the most important assurance system, but is in fact strengthened meaningfully compared. aspect of method validation because the place of the rapid methods is in the brewing industry clearly defined in the overall microbiological testing programme. is precision testing Proficiency testing It is vital that analytical methods are shown to be sufficiently precise to indicate Bass Brewers has operated a monthly efficiently whether a parameter conforms to internal PT scheme for many years. its set quality limits (QLs). Consequently, the Company and contract breweries were tested most important aspect of method validation with typical beer samples on a monthly in the brewing industry is precision testing. basis. Bass Brewers strongly supported the All new analytical methods are precision introduction of the independent PT scheme, tested by Bass Brewers to determine Brewing Analytes Proficiency Testing repeatability and reproducibility. Precision Scheme (BAPS), which is co-ordinated from testing within the company is required for LGC, because this relieved the costs arising method validation even in the case of a widely from participation of the contract brewers accepted and validated method; one reason in several different corporate schemes, The BAPS results are received for this is that tighter QLs – better product and its own costs in managing such a in Burton-on-Trent and summarised to consistency – will be achievable if scheme. Internal PT now operates for the regional sites. Results are handled reproducibility within the company is better specific purposes only, such as the rapid positively as a competitive activity between than that recorded for the industry as a whole. comparison of analytical results on samples the sites, and an annual prize for excellence is awarded. Term Definition Turnaround time is seen as a priority Specification The quality target laid down by Bass Brewers for for a large PT scheme such as BAPS, process control and product analysis parameters. since early notification of any unsatisfactory Accept Quality Limit (AQL) The limits of variation from specification which should result can avoid the costs associated with be achievable within a well controlled process. retesting of product. Reject Quality Limit (RQL) The maximum variation from specification which would PT has assisted in demonstrating the not be detected by a discerning consumer (at any stage within the product shelf life), and which is therefore importance of instrument specification in permitted by Bass Brewers. analysis. In the early days of BAPS, it was Specification Range The range about the specification defined either by noted that the colour of beers produced at AQLs or RQLs. Bass differed from that of others in the Analytical Parameters Quality parameters measured by in-line equipment or as industry. It was found that the interference documented in Bass Brewers Analytical Methods. filters in the colorimeters used had Microbiological Parameters Quality parameters measured as documented in deteriorated, and the discrepancy was Bass Brewers Handbook of Practical Microbiology. rectified when the filters were replaced. Process Parameters Parameters controlled by process measurements such as time, temperature and pressure. Bass Brewers are aware of the ability of a PT scheme to pick up methodological Table 2: QC terminology appropriate to a large variations that may not yet be recognised as scale brewing process significant. An example was cited in which

20 VAM BULLETIN CASE STUDY

BAPS participants reported alcohol and original gravity results from a highly defined catalytic combustion method relying on a specified commercial analyser, the SCABA. Results for a strong beer were distributed over two peaks, i.e. bimodally. Upon investigation, it became apparent that the method was non-linear with respect to sample dilution. Those participants who had diluted the sample twofold before analysis, and had corrected back afterwards by the same factor of two, produced different results from those who had used the undiluted sample or had made allowance for the method non-linearity. Formal quality systems There is a strong central lead and monitoring of quality assurance at Bass Brewers. A direct alarm system operates in the Burton headquarters when a predetermined limit for complaints on a given product line is exceeded locally. In addition to immediate action with the locality concerned, a weekly incident report is generated at Burton. The business benefits • Participation in industry-wide Bass Brewers Quality Management of VAM to Bass Brewers proficiency testing enhances awareness System is accredited to ISO 9002. ISO 5725 of the range of analytical methods for method validation by precision testing is • VAM is not prescriptive, but enables available, encouraging the uptake of the also followed. an industry-specific approach to most efficient methods. British Standard 6001 has been analytical quality. • Decentralisation of testing facilities introduced by Bass Brewers to determine the • The analytical activities carried out are to the production teams has been lot sizes for validation of quality control directed toward a common analytical achieved through planned sampling and testing. Decentralisation of these tests from requirement, which is to produce beers retesting activities, including random the core laboratory to the production area to delight the consumer. Within this checks by interchange of samples expedites the release of urgently required requirement, a major component of between the production team and the products from the factory gates. But there is consumer satisfaction is product QA core laboratories. Decentralisation also strong customer-driven pressure to consistency, and recognition of this fact has has led to opportunities to satisfy release the product at this point, entailing shaped a cost-effective analytical quality customer demand flexibly by speeding the risk of an oversight in quality control. control and quality assurance system. up product release and reducing Consequently, this policy of decentralisation • A rigorous approach to the specification testing costs. has been supplemented by a system of of production parameters and their • Total quality management, including random checks of the production testing by tolerances has provided high product fully documented quality control the laboratory team. If the rejection rate rises consistency. Alternative approaches to procedures, provides clarity in too high, the quality control function reverts product consistency, such as the production management, helps to to the core laboratory facility. relocation of all brewing activities to a reduce unanticipated variations in The annual Bass Brewers Quality single site, would be costly and process and product, and minimises the Awards involves the assessment of each of impossible to achieve. costs of retesting, waste and reworking the eight breweries in the company on a • Method validation, with an emphasis of product. number of points, the most important of on precision testing, leads to confidence • Performance against quality standards which is Right First Time. Right First Time in the level of product uniformity can be handled positively. Sound aims to prevent repeat processing of batches, achievable. technical training builds confident with its associated costs and possible risks to • Transfer to the UK-wide proficiency people as well as competent analysts, product quality. There is an award for best testing scheme, BAPS, eliminated the while a well-run awards system can performance, and a second award for the costs associated with management of the inspire staff to a proactive role in quality greatest improvement over the year. internal Bass Brewers scheme. product delivery.

21 VAM BULLETIN STATISTICS IN CONTEXT Missing values, outliers, robust statistics and non-parametric methods

is not feasible, particularly where readings situations where parameters (correlation Shaun Burke, are taken at set times or the cost of re-testing coefficients, for example) are calculated is prohibitive, so alternative ways of on successive pairs of variables (e.g. in a LGC addressing this problem are needed. recovery experiment we may be Current statistical software packages interested in the correlations between This is the fourth article in a series of short typically deal with missing data by one of material recovered and extraction time, papers introducing basic statistical methods of use three methods: temperature, particle size, solvent in analytical science. 1. Casewise deletion excludes all examples polarity, etc. With pairwise deletion if Introduction (cases) that have missing data in at least one solvent polarity measurement was one of the selected variables. For missing (see Table 2) only the one n the three previous papers1–3 we have example, in ICP-AAS (inductively solvent polarity/recovery pair would be Iassumed the data have been ‘tidy’ i.e. coupled plasma – atomic absorption deleted from the correlation (the normally distributed with no anomalous spectroscopy) calibrated with a number correlations for recovery versus and/or missing results. In the real world, of standard solutions each containing extraction time and particle size would however, we often need to deal with ‘messy several metal ions at different be unaffected. data’, for example data that contain concentrations, if, as shown in Table 1, Pairwise deletion can, however, lead to transcription errors, unexpected extreme some of the aluminium values are serious problems. For example, if there results or is skewed. How we deal with data missing for particular test portions, all is a ‘hidden’ systematic distribution of of this type is the subject of this article. the results for those test portions would missing points then a bias may result be disregarded. when calculating a correlation matrix Transcription errors (i.e. different correlation coefficients in Transcription errors can normally be Al B Fe Ni the matrix can be based on different corrected by implementing good quality Solution 1 94.5 578 23.1 subsets of cases). control procedures before statistical analysis Solution 2 567 72.1 673 7.6 3. Mean substitution replaces all missing is carried out. For example, the data can be Solution 3 34.0 674 44.7 data in a variable by the mean value for independently checked or more rarely the Solution 4 234 97.4 429 82.9 that variable (see Table 3). This data set now ‘appears’ to be complete, but mean data can be entered again, independently, Case wise deletion. Statistical analysis substitution has its own disadvantages. into two separate files and the files compared only carried out on electronically to highlight any discrepancies. the reduced data set. The variability in the data set is There are also a number of outlier tests (see Al B Fe Ni artificially decreased in direct proportion below) which can be used to highlight Solution 2 567 72.1 673 7.6 to the number of missing data points, anomalous values before other statistics are Solution 4 234 97.4 429 82.9 leading to under-estimates of dispersion calculated. These tests do not remove the (the spread of the data). Mean need for good quality assurance; rather they Table 1: Missing data handled substitution may also considerably should be seen as an additional quality check. by casewise deletion change the values of some other statistics, such as linear regression Missing data This is the usual way of dealing with statistics3, particularly where correlations missing data, but it does not guarantee are strong. No matter how well our experiments are correct answers. This is particularly so, Examples of these three approaches are planned there will always be times when in complex (multivariate) data sets illustrated further, for the calculation of something goes wrong, resulting in gaps in where it is possible to end up deleting a correlation matrix, where the the data. Some statistical procedures will not the majority of your data if the missing correlation coefficient3 (r) is determined work as well, or at all, when some data are data are randomly distributed across for each paired combination of the missing. The best recourse is always to cases and variables. 5 variables, A to E (see Figure 1). repeat the experiment to generate the 2. Pairwise deletion can be used as an Note, how the r value can increase, complete dataset. Sometimes, however, this alternative to casewise deletion in diminish or even reverse sign depending

22 VAM BULLETIN STATISTICS IN CONTEXT

Recovery Extraction Particle size Solvent polarity 4,5 (%) time (min) (µm) index Imputation Sample 1 93 20 90 This is a method which is Sample 2 105 120 150 1.8 increasingly being used to handle missing Sample 3 99 180 50 1.0 data, but, is not yet widely available in Sample 4 73 10 500 1.5 statistical software packages. In its simplest ad hoc form an imputed value is Pair wise deletion. Statistical analysis unaffected except for when one of a pair of data points is missing. substituted for the missing value (e.g. mean substitution, already discussed, is a Recovery vs Recovery vs Recovery vs Extraction time Particle size Solvent polarity form of imputation). In its more r 0.728886 -0.87495 0.033942 general/systematic form, however, the (number of data points (4) (4) (3) imputed values are predicted from in the correlation) patterns in the real (non-missing) data. A Table 2: Missing data handled by pairwise deletion total of m possible imputed values are calculated for each missing value and on which method is chosen to handle the Al B Fe Ni then m possible complete data sets are missing data (see the A, B correlation Solution 1 94.5 578 23.1 analysed in turn by the selected statistical coefficients). Solution 2 567 72.1 673 7.6 method. The m intermediate results are Solution 3 34.0 674 44.7 Extreme values, then pooled to yield the final result and Solution 4 234 97.4 429 82.9 stragglers and outliers an estimate of the statistics uncertainty. This method works well providing that Mean substitution. Extreme values are defined as observations Statistical analysis the missing data are randomly distributed carried out on pseudo in a sample, so far separated in value from the completed data with no allowance made for errors remainder as to suggest that they may be from and the model used to predict the in estimated values. a different population, or the result of an error imputed values is sensible. Al B Fe Ni in measurement6. Extreme values can also be Solution 1 400.5 94.5 578 23.1 subdivided in to stragglers, extreme values e.g. increase the estimate of variance (a Solution 2 567 72.1 673 7.6 detected between the 95% and 99% measure of spread), or possibly introduce a Solution 3 400.5 34.0 674 44.7 confidence level, and outliers, extreme values at bias in the calculated mean. There is Solution 4 234 97.4 429 82.9 > 99% confidence level. one golden rule, however, no value should It is tempting to remove extreme values be removed from a data set on statistical Table 3: Missing data from a data set, because it is believed they grounds alone. ‘Statistical grounds’ include handled by mean substitution will incorrectly alter the calculated statistics, outlier testing.

Figure 1: Effect of missing data on a correlation matrix

23 VAM BULLETIN STATISTICS IN CONTEXT

Outlier tests tell you, on the basis of some simple assumptions, where you are   |x- - x| x - x (n - 3) x s2 most likely to have a technical error; they do G = _____i G = _____n 1 G = 1 -______n - 2  1 ss2 3 2 not tell you that the point is ‘wrong’. No  (n - 1) x s  matter how extreme a value is in a set of where, s is the standard deviation for the whole data set, x is the suspected single outlier data, the suspect value could nonetheless be i i.e. the value furthest away from the mean, || is the modulus – the value of a calculation a correct piece of information1. Only with ignoring the sign of the result, x- is the mean, n is the number of data points, x and x are experience or the identification of a n 1 the most extreme values, s is the standard deviation for the data set excluding the particular cause can data be declared n-2 suspected pair of outlier values i.e. the pair of values furthest away from the mean. ‘wrong’ and removed. So, given that we understand that the Figure 2: Grubbs’s tests tests only tell us where to look, how do we test for outliers? Outlier tests The appropriate outlier tests for the If we have good grounds for believing three situations illustrated in Figure 3 are: our data are normally distributed then a In analytical chemistry it is rare that we (3a) Grubbs 1 and Dixon, (3b) Grubbs 2, number of ‘outlier tests’ (sometimes called have large numbers of replicate data, and (3c) Grubbs 37. Q-tests) are available which identify extreme small data sets often show fortuitous We will concentrate on the three values in an objective way7. Good grounds grouping and consequent apparent outliers. Grubbs’s tests. The test values are calculated for believing the data are normal are: Outlier tests should, therefore, be used with using the formulae in Figure 2, after • past experience of similar data care and, of course, identified data points arranging the data in ascending order. • passing normality tests e.g. Kolmogrov- should only be removed if a technical reason If the test values (G1, G2, G3) are greater Smirnov-Lillefors test, Shapiro-Wilk’s can be found for their aberrant behaviour. than the critical value obtained from tables test, skewness test, kurtosis test7,8 etc. Most outlier tests look at some measure (see Table 4) then the extreme value(s) are • plots of the data, e.g. frequency of the relative distance of a suspect point unlikely to have occurred by chance at the histogram normal probability plots1,7 etc. from the mean value. This measure is then stated confidence level. Note that the tests used to check assessed to see if the extreme value could normality usually require a significant reasonably be expected to have arisen by Pitfalls of outlier tests amount of data (a minimum of 10–15 results chance. Most of the tests look for single are recommended depending on the extreme values (Figure 3a), but sometimes it Figure 4 shows three situations where normality test applied). For this reason there is possible for several ‘outliers’ to be present outlier tests can misleadingly identify an will be many examples in analytical science in the same dataset. These can be identified extreme value. where either it will be impractical to carry in one of two ways. Figure 4a shows a situation common in out such tests, or the tests will not tell us 1. By iteratively applying the outlier test. chemical analysis. Because of limited measure- anything meaningful. 2. By using tests which look for pairs of ment precision (rounding errors) it is possible If we are not sure the data set is normally extreme values, i.e. outliers that are to end up comparing a result which, no matter distributed then robust statistics and/or non- masking each other (see Figure 3b and 3c). how close it is to the others’ values, is an infinite parametric (distribution independent) tests Note, as a rule of thumb if more than number of standard deviations away from the can be applied to the data. These three 20% of the data are identified as outlying mean of the remaining results. This value will approaches (outlier tests, robust estimates you should start to question your therefore always be flagged as an outlier. and non-parametric methods) are examined assumption about the data distribution In Figure 4b there is a genuine long tail in more detail below. and/or the quality of the data collected. on the distribution which may cause

Figure 3: Outliers and masking

24 VAM BULLETIN STATISTICS IN CONTEXT

13 replicates are ordered in ascending order

x 1 x n 47.876 47.997 48.065 48.118 48.151 48.211 48.251 48.559 48.634 48.711 49.005 49.166 49.484

2 n = 13, mean = 48.479, s = 0.498, sn-2 = 0.123   49.484 - 48.479 49.484 - 47.876 10 x 0.123 G = ______= 2.02 G = ______= 3.23 G = 1 -______= 0.587 1 0.4982 0.498 3  12 x 0.4982

Grubbs’s critical values for 13 values are G1 = 2.331 and 2.607, G2 = 4.00 and 4.24, G3 = 0.6705 and 0.7667 for the 95% and 99% confidence levels respectively. Since, in all cases, the test values are less than their respective critical value, it can be concluded there are no outlying values.

Worked example 1: Grubbs’s tests

Figure 4: Pitfalls of outlier testing successive outlying points to be identified. 95% confidence level 99% confidence level This type of distribution is surprisingly nGG G G G G common in some types of chemical analysis, 1 2 3 1 2 3 3 1.153 2.00 — 1.155 2.00 — e.g. pesticide residues. 4 1.463 2.43 0.9992 1.492 2.44 1.0000 If there are few measurement results 5 1.672 2.75 0.9817 1.749 2.80 0.9965 (Figure 4c) an outlier can be identified by 6 1.822 3.01 0.9436 1.944 3.10 0.9814 chance. In this situation it is possible that the identified point is closer to the ‘true 7 1.938 3.22 0.8980 2.097 3.34 0.9560 value’ and it is the other values that are the 8 2.032 3.40 0.8522 2.221 3.54 0.9250 outliers. This occurs more often than we 9 2.110 3.55 0.8091 2.323 3.72 0.8918 would like to admit; how many times do 10 2.176 3.68 0.7695 2.410 3.88 0.8586 your method procedures state “average the 12 2.285 3.91 0.7004 2.550 4.13 0.7957 best two out of three determinations”? 13 2.331 4.00 0.6705 2.607 4.24 0.7667 15 2.409 4.17 0.6182 2.705 4.43 0.7141 Outliers by variance 20 2.557 4.49 0.5196 2.884 4.79 0.6091 When the data are from different 25 2.663 4.73 0.4505 3.009 5.03 0.5320 groups (for example when comparing test 30 2.745 4.89 0.3992 3.103 5.19 0.4732 methods via interlaboratory comparison) 35 2.811 5.026 0.3595 3.178 5.326 0.4270 it is not only possible for individual points 40 2.866 5.150 0.3276 3.240 5.450 0.3896 within a group to be outlying but also for 50 2.956 5.350 0.2797 3.336 5.650 0.3328 the group means to have outliers with 60 3.025 5.500 0.2450 3.411 5.800 0.2914 respect to each other. Another type of 70 3.082 5.638 0.2187 3.471 5.938 0.2599 ‘outlier’ which can occur is when the spread 80 3.130 5.730 0.1979 3.521 6.030 0.2350 of data within one particular group is 90 3.171 5.820 0.1810 3.563 6.120 0.2147 unusually small or large when compared 100 3.207 5.900 0.1671 3.600 6.200 0.1980 with the spread of the other groups (see 110 3.239 5.968 0.1553 3.632 6.268 0.1838 Figure 5). 120 3.267 6.030 0.1452 3.662 6.330 0.1716 130 3.294 6.086 0.1364 3.688 6.386 0.1611 140 3.318 6.137 0.1288 3.712 6.437 0.1519 Table 4: Grubbs’s critical value table 7

25 VAM BULLETIN STATISTICS IN CONTEXT

In such cases, therefore, two types of outlier test are required. An inter laboratory study was carried out by 13 laboratories to determine the amount of 1. The Grubbs’s tests which are used to cotton in a cotton/polyester fabric, 85 determinations were carried out in total. determine the presence of within group outlying replicates and also to test for The standard deviations of the data obtained by each of the 13 laboratories were as follows: suspected outlying means. 2. The Cochran test used to test for Std. Dev. 0.202 0.402 0.332 0.236 0.318 0.452 0.210 0.074 0.525 0.067 0.609 0.246 a suspected outlying variance. [The 0.198 variance is a measure of spread and is simply the square of the 85 0.6092 0.371 1 - __ ≈ ______standard deviation .] n = = 6.54 7C- = = = 0.252 13n 0.2022 + 0.4022 +…+ 0.2462 + 0.1982 1.474 To carry out the Cochran test, the suspect variance is ratioed with the sum of all group variances. Cochran’s critical value for n-- = 7 and g = 13 is 0.23 at the 95% confidence level7.

g Since the test value is greater than the critical value it can be concluded that the laboratory ∑ with the highest standard deviation (0.609) has an outlying spread of replicates and the suspected (s2) ni C = ______and n-- = _____i = 1 results from this laboratory therefore, need to be investigated further. It is normal practice n- g g ∑ 2 in inter-laboratory comparisons not to test for low variance outliers, i.e. laboratories si i = 1 reporting unusually precise results.

where g is the number of groups and ni is the number of replicates in the ith Worked example 2: Cochran test

group. If the calculated ratio, Cn-- exceeds the critical value obtained from the statistical table7 then the suspect group spread is extreme.

The Cochran test assumes the number of replicates within the groups are the same or at least similar (± 1). It also assumes that none of the data has been rounded and that there are sufficient replicates to get a reasonable estimate of the variance. The Cochran test should not be used iteratively as this could lead to a large percentage of data being removed. Robust statistics Robust statistics include methods which are largely unaffected by the presence of extreme values. The most commonly used of these statistics are the median and median absolute deviation (MAD). Figure 5: Different types of outlier in grouped data

Median Median Absolute Deviation (MAD) The median is a measure of central The MAD value is an estimate of the 1 x when n is odd 1, 3, 5,… tendency and can be used instead of the mean.  m spread of the data like the standard deviation. ◆◆ x + x To calculate the median (x) the data are x = _____m m+1 For n values arranged in order of magnitude and the median  2 when n is even 2, 4, 6,… ◆◆ is then the central member of the series (or the MAD = median [x - x ] i i = 1, 2, …, n mean of the two central members when there are If the MAD value is scaled by a factor of an even number of data, i.e. there are equal n where m = round 1.483 it becomes comparable with a numbers of observations smaller and greater 2 standard deviation, this is the MAD value. than the median). For a symmetrical distribution E the mean and median have the same value. MADE = 1.483 x MAD

26 VAM BULLETIN STATISTICS IN CONTEXT

Other robust statistical estimates include Types of comparison Parametric methods Non-parametric methods11,12 trimmed mean and deviations, Winsorized Differences between t-test for independent groups 2 Wald-Wolfowitz runs test mean and deviation, least median of squares independent groups of data Mann-Whitney U test (robust regression), Levene’s test Kolmogorov-Smirnov two-sample test (heterogeneity in ANOVA) etc. A discussion 13 of robust statistics in analytical chemistry can (ANOVA/MANOVA) Kruskal-Wallis analysis of ranks be found elsewhere 9,10. Median test Differences between t-test for dependent groups2 Sign test Non-parametric tests dependent groups of data Wilcoxon’s matched pairs test McNemar’s test Typical statistical tests incorporate χ2 (Chi-square) test assumptions about the underlying ANOVA with replication13 Friedman’s two-way ANOVA distribution of data (such as normality), and Cochran Q test hence rely on distribution parameters. ‘Non- Relationships between Linear regression3 Spearman R continuous variables Correlation coefficient3 Kendall parametric’ tests are so called because they Tau make few or no assumptions about the Homogeneity of variance Bartlett’s test7 Levene’s test, distributions, and do not rely on distribution Brown & Forsythe parameters. Their chief advantage is Relationships between Coefficient Gamma improved reliability when the distribution is counted variables χ2 (Chi-square) test Phi coefficient unknown. There is at least one non- Fisher exact test parametric equivalent for each parametric Kendall coefficient type of test (see Table 5). In a short article, of concordance such as this, it is impossible to describe the Table 5: Non-parametric alternatives to parametric statistical tests methodology for all these tests but more information can be found in other REFERENCES 7. Practical statistics for the analytical 11,12 publications . scientist: A bench guide, T J Farrant, 1. ‘Statistics in context: Exploring and The Royal Society of Chemistry, Conclusions summarising the results of measure- Cambridge ISBN 0 85404 442 6 (1997). ments’, VAM Bulletin, 16, 20 – 22, • Always check your data for transcription Spring 1997. 8. International Encyclopaedia of errors. Outlier tests can help to identify Statistics, William H Kruskal, Judith M them as part of a quality control check. 2. ‘Statistics in context: Significance Tanur, Collier Macmillian Publishers, • Delete extreme values only when a testing’, VAM Bulletin, 17, 18 – 21, New York ISBN 0 02 917960 2 (1978). technical reason for their aberrant Autumn 1997. behaviour can be found. 3. ‘Statistics in context: Regression & 9. Analytical Methods Committee, Analyst, • Missing data can result in misinter- calibration’, VAM Bulletin, 18, 18 – 21, 114, 1693-7, 1989. pretation of the resulting statistics so Spring 1998. 10. Understanding Robust and Exploratory care should be taken on the method 4. Monographs on Statistics and Applied data analysis, D C Hoaglin, F Mosteller, chosen to handle the gaps. If at all Probability 72 – Analysis of Incomplete J W Tukey, John Wiley & Sons, New possible, further experiments should be Multivariate Data, J L Schafer, Chapman York ISBN 0-471-09777-2 (1983). carried out to fill in the missing points. & Hall, London ISBN 0 412 04061 1 • Outlier tests assume the data 11. Non-parametric statistical methods, (1997). distribution is known. This assumption M Hollander, D A Wolf, Wiley & Sons, should be checked for validity before 5. Statistical analysis with missing data, New York ISBN 0 47140 635 X (1973). these tests are applied. R J A Little, D B Rubin, John Wiley & 12. Applied non-parametric statistics, • Robust statistics avoid the need to use Sons, New York ISBN 0 471 80254 9 W W Daniel, Houghton Mifflin, Boston outlier tests by down-weighting the (1987). ISBN 0 53491 976 6 (1978). effect of extreme values. 6. ISO 3534. Statistics – Vocabulary and • When knowledge about the underlying Symbols. Part 1: Probability and general 13. S Burke, ‘Statistics Refresher 2 Analysis data distribution is limited, non- statistical terms, section 2.64. Geneva of Variance’, Scientific Data Management, parametric methods should be used. (1993). 2 (1), 36–41, 1998.

27 VAM BULLETIN VAM IN EDUCATION PT schemes for schools results

roficiency Testing (PT) schemes Pprovide analytical laboratories with an independent check on their capability to perform certain analyses and provide a benchmark for that laboratory. Many laboratories use these schemes as part of their Quality Control procedures. As reported previously Nuffield Curriculum projects and LGC, as part of the VAM programme, have organised a PT scheme for schools. The scheme is run as a competition and, as organised, it is a measure of two things. It measures the teachers’ ability to teach good practice as well as the ability of the students to learn and use the techniques required. It also provides an opportunity to show how chemical measurements fit into the National Measurement System. were caused by calculation or transcription The exercise fits in well with the 16+ Between January and April of this year errors. However, overall the reports syllabuses and provides the teachers with an 26 centres took part in the PT scheme, submitted were of a higher standard than external assessment. The length of time involving 346 students in the competition. those submitted in the last competition required for the task fits in with normal Students were required to determine the (1996) and choosing a winner was more lesson times so is not too disruptive. concentration of the ethanoic acid supplied difficult. The winning centre was Halesowen The teachers said that the information and, using a primary standard, determine College (West Midlands) with Alcester provided was “sufficient to make the students the concentration of the sodium hydroxide Grammar School (Warwickshire) and Ridge think and easy to follow”. Although the used. From the teachers’ reports the Danyers College (Stockport) runners up. It number of centres involved is only a small students enjoyed the exercise and its should be noted that Ridge Danyers were percentage of the whole 16+ population competitive nature. The students thought it the winners for the 1996 competition and it studying chemistry it is encouraging to see important to take care with their is good to see that they maintained their high such good performances. It is to be hoped measurements and reports, because they standard with a new group of students. that these students will go on into disciplines were to be externally assessed. The rules At two events in July the prizes were which involve chemical measurements. were slightly tougher this time and more awarded to the winning centres and the Another competition is planned for January comparable with professional schemes. The students received their certificates. Mr to March 1999 when hopefully more centres Z – scores, Graham Hall of Halesowen College and Mr will be involved. Jeremy Slater of Alcester Grammar school The success of the competition involves reported result – assigned value received the prizes on behalf of their a number of people and organisations. We Z =,______ target standard deviation  departments at the Albright & Wilson, would like to thank the following Oldbury Works, after they and their students organisations for their contributions: were based on the reported results and not had been given a very informative tour of the Nuffield Curriculum Projects (organisation just the titration values. A result gaining a Z – site. The prizes and certificates were presented and half the prize money); Kodak Research score in the range +2 to - 2 is deemed by Dr Bob Crowte of Albright & Wilson, and Development, Harrow, (production and satisfactory, a result in the range |2 to 3| representing the local section of The Royal donation of acid samples); Albright & indicates more care is required and a score in Society of Chemistry. Mr Alan Todd received Wilson (site tour, staff and venue for excess of |3| is a poor result, something has the award on behalf of Ridge Danyers College, presentation); The Royal Society of gone wrong with either the estimation or the at the college, following a lunch provided by Chemistry local section committees – calculation. During this round, 79% of the the Manchester section of The Royal Society Birmingham & West Midlands and participating groups achieved Z – scores in the of Chemistry. The chairman of the section, Dr Manchester (personnel and refreshments for range -2 to +2 and were awarded a distinction, Gordon Thomson of BASF, presented the the prize giving); the laboratory staff at LGC 2% between |2 and 3| gained merit and the certificates to the students. The students at who carried out homogeneity testing studies remaining 19% with scores in excess of |3| both venues spoke enthusiastically about the and the statistical evaluation and the DTI received a certificate indicating they had competition and were keen to know more for funding the VAM Education and participated. Most of the high Z – scores about analytical science. Training programme.

28 VAM BULLETIN VAM NEWS The Consultative Committee on Amount of Substance

he fourth annual meeting of the Comitè Working Group Chairman Organisation Consultatif pour la Quantitè de Matière T Gas analysis Dr Anton Alink NMI (NL) (CCQM) took place on 19 and 20 February Organic analysis Dr Willie May NIST(USA) 1998 at the Bureau International des Poids et Inorganic analysis Dr Mike Sargent LGC(UK) Mesures (BIPM) which is located at Sèvres pH measurement Dr Wolfgang Richter PTB(D) in the suburbs of Paris. The joint UK delegates to the meeting were Mike Sargent Table 1: CCQM Working Groups of LGC and Martin Milton of NPL. The CCQM was formed primarily to Reference Topic Organising Body organise international comparisons of Study I Metals in water NIST chemical analysis methodology and Study II Gas analysis NMI standards, with great emphasis on attaining CCQM-1 Pb in water by IDMS NIST the highest possible levels of accuracy. To CCQM-2 Gas analysis NMI date there have been two preliminary studies CCQM-3 pp’ DDE in solvent LGC and three comparisons but interest in the CCQM-4 NMR analysis of mixtures BAM work of the Committee has grown rapidly CCQM-5 pp’DDE in corn oil LGC and this year it was attended by 40 members CCQM-6 Characterisation of NIST or invited observers – a record number for a pure substances Consultative Committee. As a result it was CCQM-7 Clinical analysis NIST possible to rationalise the work of the CCQM-8 Purity determination NIST CCQM with the confirmation of four formal CCQM-9 Cd and Pb in natural water IRMM working groups which will take activities by IDMS forward between the Paris meetings. The CCQM-10 Automobile emission gases NMI in nitrogen current arrangements for these working CCQM-11 Ethanol in air or nitrogen NPL groups are summarised in Table 1. The delegates expressed a wide variety of views on Table 2: CCQM studies and comparisons priorities for future work but good progress was made towards a long term strategy and If you would like further information analysis) or Martin Milton, Tel: 0181-943 6826; eight new comparisons were agreed. These about the CCQM activities please contact Email: [email protected] (gas analysis). are summarised in Table 2, together with the either Mike Sargent, Tel: 0181-943 7360; The next meeting of the CCQM will take existing studies or comparisons. Email: [email protected] (organic and inorganic place on 10–12 February 1999.

REFERENCE MATERIALS UPDATE Reference materials update

GC’s Office of Reference Materials source for CRMs and contains details of the LGC’s range L(ORM) continues to supply an materials available from the world’s leading continues to expand expanding array of certified reference CRM producers. LGC’s Reference materials (CRMs) in order to enhance the Materials Advisory Service (REMAS) will VAM support has enabled LGC to analytical capability of laboratories be pleased to offer free advice if you launch a number of new CRMs, which are worldwide. The fourth edition of the have problems locating the desired now available from the ORM. These are catalogue is still available and has proved material. Contact the ORM for a free copy listed in Table 1. Table 2 shows CRMs extremely popular. This provides a single of the catalogue. currently under production at LGC and

29 VAM BULLETIN REFERENCE MATERIALS UPDATE these will be available soon. All LGC’s being available separately or as a set. CRMs currently being produced by LGC CRMs are available either directly from the Contact the ORM for further information. Meat and fish based foods (proximates ORM or through LGC’s growing network of and trace metals) The Office of Reference Materials is here international distributors. (Details are Distilled spirit (congeners) to help you purchase the materials you available on request). Frying oil (BHT and BHA) require and has made this easier by accepting Animal feed (proximates and credit and debit cards. Free advice is available What do you need? trace metals) to assist you in finding the right materials for It is important that CRMs are produced Canned pet food (proximates and whatever techniques you are using. trace metals) which will be of use to the analytical Estuarine water (major and A comprehensive community. We are always pleased to hear trace elements) your needs. Is there a CRM you want which & personal service Estuarine sediment (heavy metals) is unavailable? Please contact us and we will Soils (anions, PAHs, PCBs) The ORM aims to provide a global focal endeavour to produce the materials which Sewage sludges (total & point from where you can purchase the are of the greatest importance to the leachable metals) reference materials you require and obtain analytical community. Waste Oil (PCBs) free advice to assist you in finding the right Pulverised fuel ash (heavy metals) material to use. We offer you a New CRMs produced by LGC Freeze-dried urine containing drugs comprehensive personal service aimed at Micro-organisms in freeze-dried sugar of abuse (LSD/ecstasy) meeting all your CRM needs. solutions UV/VIS absorbance standard for HPLC Biscuit based food product (proximates detectors and trace metals) UV/VIS stray light standard LGCs Office of Reference Materials Soft drinks (sweetners – acesulphame-k, Paint, textile, plastic and paper Brian Brookman Marketing & Sales aspartame and saccharin) (toxic elements) Manager α Refined olive oil ( -tocopherol) Bio-compatible materials Alison Jones Marketing & Sales Potato powder (sulphur dioxide) Executive Biscuit pet food (proximates Table 2 Kevin Thurlow Marketing & Sales and chloride) Executive Hard and soft drinking water certified New soil CRMs Allen Brown Marketing & Sales for trace elements and nutrients Administrator Soils and sediments (total & Soil analysis is becoming increasingly Jane Hayman Customer Support Manager leachable metals) important owing to environmental concerns. Debra Roker Customer Support Landfill leachate (metals and anions) To ensure that the analytical community can Mary Bourne CRM Despatch Non-equilibrium flash point standards easily obtain the most appropriate CRMs, Low melting point standard LGC continues to expand its ‘one stop- LGC’s Office of Reference Materials Pesticide standards shop’ portfolio. The ORM is pleased to Queens Road, Teddington announce the availability of a new set of Middlesex TW11 0LY Table 1 CRMs, certified by the Czech Metrological Tel: 44 (0) 181 943 7565 Fax: 44 (0) 181 943 7554 Institute. There are four different soils, with Email: [email protected] varying elemental composition, the CRMs

VAM PRODUCTS AND SERVICES VAM advisory services

GC operates a number of advisory Analytical Quality Assurance organisations that support them Lservices within the VAM programme. In Advisory Service (AQAAS) • helping laboratories to comply with future, all calls will now come in through a quality standards and assess their readiness common contact point – our aim is to match Getting to the right source of for assessment against those standards your enquiry to the right source of advice information on analytical quality can be an • providing independent quality auditing and for it to be answered as quickly and awkward business. Assistance provided facilities painlessly as possible. through the AQAAS covers a wide variety of • providing advice on specific technical issues, for example: aspects of QA such as measurement • directing users to the quality system that is uncertainty, method validation, calibration most appropriate for their needs and the and quality control

30 VAM BULLETIN VAM PRODUCTS AND SERVICES

• advising on developments arising Proficiency testing materials. Information includes the from the ever-changing international advisory service (PTAS) name and address of the producer, the quality scene. form of the material, its certified values, The Proficiency Testing Advisory Service and their uncertainties, date of certification Chemical nomenclature (PTAS) provides advice and information on and its field of application. LGC represents advisory service (CNAS) a wide range of proficiency testing issues. UK on the COMAR Council which Callers seek information on the availability of The proliferation of new chemicals has includes representatives from all over PT schemes in their particular field of given rise to a number of systems for naming the world. interest. They may also need to know how to them. The lack of a unified system of use the schemes, how to interpret data, how Statistics advisory service nomenclature has caused confusion in many to gain accreditation etc. areas, for example in the purchase of This service provides advice on statistics Callers may require help and advice to set chemicals, or in interpreting legislation. to analysts. It is aimed at answering questions up a new PT scheme. CNAS specialises in the use of a logical and such as “which method(s) should I use to unambiguous form of nomenclature, based Reference materials analyse my data?”, “where can I find out on the recommendations of the International advisory service (REMAS) more about method X?”, “do you know of a Union of Pure and Applied Chemistry software package that will do Y?”, “who (IUPAC). CNAS can provide (inter alia) This service provides the analytical should I contact for specialist advice/ names for notification schemes, check community with a means for obtaining training?”. If you have a statistics/chemometrics chemical nomenclature in documents and access to current information on related question, contact this service. also offer advice on the use of different Certified Reference Materials (CRMs), systems of nomenclature. This may include their certified values and their availability We offer 30 minutes free advice for use of Chemical Abstracts Service (CAS) from producers round the world. all services. Longer enquiries can be numbers and International Nonpropietary REMAS makes full use of the information handled, but may attract charges. Names (INN) and International available on the international CRM Whatever your question, please contact the Organisation for Standardisation (ISO) database, COMAR. The COMAR database VAM Helpdesk on 0181-943 7393 or names for pesticides. contains data on over 10000 reference via Email: [email protected]. Assuring quality in qualitative analysis

he work performed by analytical As a part of the VAM programme, LGC After analysing the importance of Tchemists traditionally falls into two has produced guidance to assist laboratories establishing the customer requirements, the broad categories. Analysts may be asked to in achieving an appropriate standard of Guide then looks at issues surrounding the identify an unknown material (qualitative quality in the execution of qualitative analysis. receipt and processing of samples, the analysis) or they may be required to This guide has recently been published by analytical process itself and aspects of quality determine the amount present of a stated The RSC under the title ‘Qualitative Analysis: assurance. It will be of interest to substance (quantitative analysis) or indeed A Guide to Best Practice’. The Guide was laboratories wishing to reinforce their quality to do both. There are anecdotal accounts developed with the assistance of senior programmes and should find use as a (laboratories seldom want to admit it chemists from industry, UKAS and the training aid. publicly) of analysts quantifying one Forensic Science Service. It sets out a number Further information on the Guide or compound in the mistaken belief that it was of principles of best practice and amplifies other aspects of qualitative analysis is something else. While there are many aids to these with short paragraphs of explanatory available from the VAM Helpdesk at LGC. achieving quality in quantitative analysis, the text. Each of the principles advocated is practice of qualitative analysis does not related to one or more of the six VAM generally have such a high profile. principles which underpin all of analysis.

31 VAM BULLETIN VAM PRODUCTS AND SERVICES Quality Standards

K analytical laboratories seeking particular sectors. Beyond this basic choice system options and information on the roles Uexternal assessment of their quality of systems laboratories also find themselves and aims, membership and contact details systems face a number of choices. NAMAS faced with a bewildering assortment of of more than 50 organisations, involved in, accreditation provides a high level of organisations, related either closely or or with interests in, various aspects of technical assurance but for a limited scope of distantly to quality, and often only identified quality and conformity assessment, in the activities. Assurance that is more widely by acronyms. UK, Europe and internationally. The applicable but less technically valid is To clarify this situation LGC, under the profiles of the quality standards include provided by certification to ISO 9000. VAM programme, has just published a information on the initiatives in hand to Regulatory requirements may demand quality standards report. The report update ISO Guide 25 and the ISO 9000 compliance with Good Laboratory Practice provides an objective snapshot of the family of standards. whilst work that is clinically based may be options for quality systems currently This report is available from LGC’s most suitably covered by Clinical Pathology available for analytical laboratories in the Office of Reference Materials, price £30. Accreditation. Other options are possible in UK. It lists profiles of the main quality (ISBN 0-948926-13-9). Analytical Molecular Biology: Quality and validation

NA technology is revolutionising the to priorities lying with innovation, rather For further details and to order a copy, Dway in which healthcare, than consideration being given to the more please contact: pharmaceutical production, diagnostics, routine applicability, reliability and Sales and Promotion Department agriculture, animal husbandry, food, reproducibility of the methods employed. The Royal Society of Chemistry pollution abatement, archaeology, forensic There is a need to evaluate these factors. Thomas Graham House analysis, and many other industrial and Further validation of the methods is Science Park regulatory sectors operate. Analytical necessary before the successful transition Milton Road molecular biology has been typically from research laboratory to analytical Cambridge CB4 4WF developed in the academic and medical laboratory can be made. Analytical Molecular Tel: 01223 420066 research environments. Here, the general Biology: Quality and Validation addresses Fax: 01223 423429 excitement of at last being able to measure these issues, providing guidelines to ease Email: [email protected] what has never been possible before, has led the transition.

32 VAM BULLETIN CHEMICAL NOMENCLATURE How to get it wrong!

document contains a bizarre reference to, Kevin Thurlow, ‘carbon bisulphide, or a compound of carbon bisulphide, or a substance containing carbon LGC bisulphide’. The first and third of those are understandable, but what is ‘a compound of hemical nomenclature can be difficult. carbon bisulphide’? ‘Carbon bisulphide’ is a You are presented with a complicated compound. It could mean a derivative of CS , C 2 diagram depicting a structure with rings of perhaps, but where would you draw the line? atoms, charged atoms and large numbers of Heptanedithioic acid (Figure 2) could be substituents. Maybe it is a polymer with regarded as a derivative of CS . Note that 2 Figure 1 unknown molecular weight. Naming such ‘heptanebis(dithioic acid)’ would cover the structures can be a nightmare. There is only compound with -CS2H at each end. The one thing worse than a complicated department concerned admitted some years structure and that is no structure at all. It is ago that they had made a mistake, but have interesting trying to create structures from yet to correct it. The same organisation the names provided in books and papers, described pyridine (Figure 3) as a when you have no idea of whether the ‘chlorinated nitro-paraffin’. Unfortunately, author has used the terms correctly. pyridine has double bonds, so it is scarcely a It is always unfortunate if published paraffin, and there is a complete absence of names are ambiguous or incorrect, but it is ‘nitro’ and ‘chloro’ groups. Apart from that, even more alarming when names in official the description is perfect. Figure 2 documents are wrong. This is perhaps not There are many common errors in names. surprising as many names are used by non- How often do we see ‘isopropanol’ and specialists, who have only a hazy idea of the ‘propionitrile’? ‘Isopropane’ does not exist, so subject. An European Communities legal ‘isopropanol’ cannot be justified. document referred to ‘methylene’ added to ‘Propionitrile’ looks all right, but nitrile names ethanol. Methylene should be ‘-CH2-’, so it of this sort are formed by removing the ‘-ic’ was not clear how it could be added to from trivial acid names, and adding ‘-onitrile’. anything. Did they mean ‘methylene blue’, Hence, ‘acetic acid’ goes to ‘acetonitrile’ and which is used as a dye? The authors had done ‘propionic acid’ goes to ‘propiononitrile’. something properly – they defined the terms Scientists are frequently confused by used in the document. This is a very chemical names so it is perhaps not surprising Figure 3 important principle which too many people that non-scientists are as well. Somebody overlook. Not everybody knows what you are complained that he had been poisoned by LGC colleague was puzzled to receive a talking about, so it is essential to clarify any ‘methyl chloride’ (CH3Cl), but after investigation telephone enquiry about what sounded like potentially confusing terminology. Thus we it was found the claimant had been exposed to ‘oksilene’ content of a sample. After a while it had the definition of ‘methylene’ as ‘actual ‘methylene chloride’ (CH2Cl2). This was pointed became clear that it was ‘o-xylene’. Telephone methylene, that is to say raw methyl alcohol…’ out, but the claimant was not fazed. He said that enquiries are fraught with difficulties especially At least this explained matters, but the ‘methyl’ appeared in ‘methylene’, so obviously if non-specialists are talking. Ask a non- dangers of using a proper chemical term in an that was covered as well. Another example chemist to say ‘ethylenediaminetetraacetic incorrect way should be obvious. It reminds occurred when someone claimed that ‘platinum acid’! Another enquiry was about ‘pentan- one of Humpty Dumpty who declared that bleach’ as used in a hairdressing salon was 2,1’. The initial reaction is that something is words meant whatever he said they did. included in ‘platinum salts’. These beliefs seem missing, but the written confirmation showed National legal documents have a few ridiculous, but are at least understandable. It ‘pentan-2-one’. ‘Propanediol’ becomes problems as well. One referred to is up to scientists to educate the general ‘proppa-needy-ol’, instead of ‘propane-di-ol’. ‘azobenzene’ (Figure 1), but it became clear public. Many years ago, some dock workers in There are many other examples which that ‘azobenzene’ was only a fragment of the London refused to handle cylinders of demonstrate equally well the need for clarity. structure. The legal experts were somewhat hydrogen chloride, because hydrogen was If you are transmitting information about a unwilling to change it, because ‘azobenzene’ inflammable. They were not even placated chemical, give as much detail as you can. had been used in another document. Another when it was explained that sodium was If you are having problems with document referred to ‘carbon bisulphide’, dangerous, but sodium chloride was edible. chemical nomenclature, we are able to offer which should be carbon disulfide, CS2. The Spoken names cause problems too. An advice (see back cover for details.)

3333 VAM VAM BULLETINBULLETIN FORTHCOMING EVENTS Forthcoming events

Measurement Uncertainty ChemAnalysis handle complaints, finance, and the training course perceptions of customers and upper 19–20 November 1998 management. Particular attention will be 20–21 October 1998 Brussels, Belgium given to people issues and future trends in Scarman House, University ChemAnalysis 1998 represents a forum technology, workforce and documentation. of Warwick for analytical chemists and associated New managers, as well as experienced LGC and UKAS join forces to provide instrument manufacturers to discuss recent managers with new responsibility for the training in measurement uncertainty. developments within the separation science chemical analysis function, will benefit from Lectures and workshops will demonstrate market of the pharmaceutical, food and the workshop. how to calculate and apply measurement drink, environmental and biotechnology For registration please contact: uncertainty by using step-by-step instructions industries. Incorporating the Fifth Byrg Godtfredsen and clear worked examples. The course will PharmAnalysis Europe Conference, Analytical Sciences R&D explain the concepts of measurement ChemAnalysis will feature a unique Nycomed Imaging AS uncertainty as stated in the ISO ‘Guide to the conference programme and comprehensive PO Box 4220, Torshov expression of uncertainty in measurement’. technical sessions, with key presentations on N-0401 Oslo, Norway This course is ideally suited to analytical new developments in HPLC, GC, CE/CEC Tel: 47 2318 5506 chemists who are involved in method and hyphenated techniques. Fax: 47 2318 6014 development and method validation. The forum is dedicated to the latest technologies, products and applications in the Quality forum For further details or reservations please separation science market and it provides an contact: excellent networking environment for everyone 10 December 1998 Elaine Jaundrell connected with the industry. The conference is 12 Errol St, London EC1 LGC preceded by Workshops on 16-18 November. The Quality Improvement Committee of Queens Road, Teddington the Royal Statistical Society is holding a Middlesex TW11 OLY For further information and registration Quality Forum on the subject of measurement Tel: 0181-943 7341 please contact: uncertainty and how its importance is often Fax: 0181-943 2767 Margaret Kermode ignored. Everyone is welcome to attend and Email: [email protected] Support Services Manager Advanstar Communications contribute their views and experience. Automation in the Advanstar House, Park West For further information, please contact: pharmaceutical industry – Sealand Road, Chester CH1 4RN Shirley Coleman Current update Tel: 01244 378888 Royal Statistical Society Fax: 01244 370011 12 Errol Street 21 October 1998 Email: [email protected] London EC1Y 8LX Scientific Societies Lecture Theatre, Email: [email protected] London European Analytical The Royal Society of Chemistry’s Laboratory Manager’s Continued on back page Analytical Division Automatic Methods Association (ALMA) Group is organising this update meeting conference which will cover a number of topics related to the automation of analytical measurements. 20–22 November 1998 The topics include drug safety, speciation and Oslo, Norway managing an analytical resource. In this workshop, the analytical process, For further information, please contact: measurement systems and the activities of Mr D G Porter practising analytical professionals will be Honorary Secretary of the presented along with ways to improve the Automatic Methods Group quality, performance and evaluation of the Willowford, Fir Tree Lane laboratory. Included will be discussions of West Chiltington the lab mission and functions, ways to West Sussex RH20 2RA organise/restructure, lab design, selecting Tel: 01798 812383 and servicing costly instruments, good Email: [email protected] laboratory practices, ways to avoid crises and

34 VAM BULLETIN 35 VAM BULLETIN FORTHCOMING EVENTS

Thermal desorption Laboratory software SAC ‘99 into the millennium and data validation 25-30 July 1999 14–15 December 1998 25 February 1999 Dublin City University, Dublin Scientific Societies Lecture Theatre, Scientific Societies Lecture Theatre, The main themes of the Analytical London London Division of The Royal Society of Chemistry’s This meeting will review new The Royal Society of Chemistry’s 1999 conference will be analysis for the public developments in automated instrumentation Analytical Division Automatic Methods good; process analysis and control; molecular as well as applications in a range of areas. Group are organising a seminar on software recognition; separation science; new There will be a number of brief papers by used in the laboratory and data validation. developments in analytical instrumentation; users on interesting applications and novel mass spectrometry; surface analysis; analytical For further information, please contact: techniques as well as a poster session. biotechnology; atomic spectroscopy; Mr D G Porter chemometrics; environmental analysis; For further information, please contact: Honorary Secretary of the molecular spectroscopy; electroanalysis; Mr D G Porter Automatic Methods Group archaeometry; other methods. Honorary Secretary of the Willowford, Fir Tree Lane Automatic Methods Group West Chiltington For further information please contact: Willowford, Fir Tree Lane West Sussex RH20 2RA Diana Hort West Chiltington Tel: 01798 812383 Conference Organiser West Sussex RH20 2RA Email: [email protected] The Royal Society of Chemistry Tel: 01798 812383 Burlington House Email: [email protected] London W1V 0BN Tel: 0171-437 8656 Fax: 0171-734 1227 Email: [email protected]

CONTACTS Contact points

General VAM contact points Proficiency testing advisory service LGC Nick Boley, LGC Queens Road, Teddington VAM Helpdesk 0181-943 7311 Middlesex TW11 0LY 0181-943 7393 Tel: 0181-943 7000 (switchboard) Gas analysis advisory service Fax: 0181-943 2767 Advisory services Dr Paul Quincey, NPL Email: [email protected] 0181-943 6788 Reference materials advisory service Web: http://www.lgc.co.uk (REMAS) Chemical nomenclature advisory service Alison Jones, LGC (CNAS) National Physical Laboratory (NPL) 0181-943 7621 Kevin Thurlow, LGC Queens Road, Teddington Analytical QA advisory service 0181-943 7424 Middlesex TW11 0LN David Holcombe, LGC Tel: 0181-977 3222 (switchboard) 0181-943 7613 Fax: 0181-943 2155 Web: http://www.npl.co.uk Statistics advisory service Trevor Farrant, LGC Aerosol Science Centre 0181-943 7374 AEA Technology plc E6 Culham, Abingdon Oxfordshire OX14 3DB Tel: 01235 463677 Fax: 01235 463205 Email: [email protected]

Produced by Horrex Davis Design Associates 9/98

36 VAM BULLETIN