Lean Six Sigma is a combination of certain tools and techniques that provides laboratory quality practitioners with a means to improve processes and reduce cycle times. This approach incorporates the use of Six Sigma methodology, which inherently focuses on gathering data, analysing the collected data, and thereafter improving the process yield, as well as the Lean methodology which identifies key areas of variation (Khalil, Khan & Mahmood 2006).

An approach to improve the Selenium Analysis Process at WCPVL led to the statement of the following research hypothesis, ‘H₀: variation in process, time and control procedures has a direct impact on the disparity in selenium testing results’. In this hypothesis, variation in process, time and control serve as the dependent variable, whilst disparity in selenium testing results serves as the independent variable.

Scott (2007) contends that a plethora of process improvement approaches exist and is of the opinion that ‘continuous process improvement methods include define, measure, analyse, improve and control, plan-do-study-act, Six Sigma and total quality management’. (Foster 2007), however, points out that there are seven basic tools, known also as the seven basic tools, typically used for process improvement. Whilst the basic tools are commonly used in a logical order, they can, however, be used in any order. These quality-tools are Ishikawa diagrams (Root Cause and Effect diagrams), Pareto charts, Process Maps (Flowcharts), Check sheets, Histograms, Scatter plots and Control charts. Berte’s (2007) opinion is that a means of improving laboratory test processes at bench level are accomplished by quality improvement programmes. Programmes include the use of statistical quality-tools, which provide a visual means to understand quality-control data, so that timely action can be taken when method problems are detected.

The merger between Six Sigma and Lean results in a powerful Lean Six Sigma tool, because the strengths from each individual method are secured (Table 1).

Wu, Pearn and Kotz (2009:339) consider process capability studies to be ‘an important and well-defined tool in applications of statistical process control (SPC) to continuous improvement of quality and productivity’. The authors state that the relationship between specification limits and tolerance limits of a process, and the actual process performance, may be quantified by using suitable process capability indices. Foster (2007) maintains that a process can be considered capable if individual results consistently meet specification; however, it is important to note that a process is considered stable only if special cause variation is not present in the process. As a result process capability studies are arguably seen to be the most fundamental toolset available for quality-improvement purposes on a process, as summarised by Marquis (2011), stating:

The clinical chemist must check whether uncertainty of all his analytical methods is compatible with the tolerance required by medical needs. It is the aim of the capability index. If the uncertainty interval of a method is greater than the tolerance interval, the analytical method must be discarded or improved. (n.p.)

For process capability studies to be conducted on an analytical chemistry process, it is important to determine and obtain the appropriate result measures (also known as R-criteria) from the process being studied. R-criteria represent the true process results, and as a result R-criteria are data, which are collected of overall process performance (Process Management International 2009). R-criteria closely track how well customer requirements are being met. This is executed by the determination of accuracy and precision of the analytical chemistry process. A controlled experiment is therefore designed and conducted in order to obtain quantitative result measures (R-criteria) or process results. A further measure known as process measures (P-criteria), represents process characteristics that are expected to affect the ‘process results’ (Dahlgaard, Kristensen & Gopal 2002). From a Lean Six Sigma perspective, R-criteria are considered to be quality-control points, whilst P-criteria are quality checkpoints. P-criteria are data measured or obtained from practical process steps. The state of the process may therefore be measured by a quality checkpoint or P-criteria.

Failure Modes and Effects Analysis (FMEA) is described as a reliability analysis tool, or a methodology to make process designs more reliable. Failure Modes and Effects Analysis is the systematic consideration of each component in a system, and involves the identification, analysis and documentation of the possible failure modes within a system, as well as the effects of each potential failure in the system (Foster 2007). Some of the benefits of FMEA are listed below:

• improvement of safety, quality and reliability
• improvement of a company’s image and its competitiveness
• increased satisfaction from a user standpoint
• reduction in development cost
• records of actions taken to reduce risk.

Van Leeuwen et al. (2009) argue that, as a quality-tool, FMEA has high improvement potential when applied to clinical laboratory processes. Failure Modes and Effects Analysis is a highly structured systematic assessment, which generates a comprehensive review to safeguard against system performance problems. Failure Modes and Effects Analysis is able to assure process and quality features, which are designed into the modified process (FMEA Information Centre 2011).

Blumberg, Cooper and Schindler (2011:13) are of the opinion that ‘good research follows the structure of the scientific method’, from which the analogy can be drawn that the structured Lean Six Sigma approach, which is inherently scientific and comprises distinct phases, provides a suitable platform to conduct research in a scientific laboratory environment. Allen (2006) regards Six Sigma as:

... an organised and systematic problem-solving method for strategic system improvement that relies on statistical methods and the scientific method to make dramatic reductions in customer defined defect rates and/or improvements in key outputs variables. (p. 8)

Following the Lean Six Sigma process map makes it possible for value to be identified within a process. Process efficiency of both current state and future state can be determined. This drives improvement actions and tasks to a future state process. The Lean Six Sigma process map also addresses process waste through Kaizen analysis, and identifies muda (non-value activities), muri (volatile work-in-progress) and mura (un-standardised working practice during sample processing) present in a process. Process standardisation is regarded as a form of process improvement (American Society for Quality 2011; Anvari, Ismail & Hojjati 2011; Byrne, Lubowe & Blitz 2008; Hubbard 2010). As a result, the analogy can be drawn that Lean Six Sigma provides a dominant and innovative tool, developed on the basis provided by the scientific method and capable of process improvement and concurrent reduction of process variation on an analytical chemistry process.

Many techniques have been developed for selenium analysis, including fluorimetry, Electrothermal Atomic Absorption (ET-AAS), HG-AAS, Neutron Activation Analysis (NAA), and different mass spectrophotometry methods. These methods include Gas Chromatography Spectrophotometry (GC-MS), thermal ionisation mass spectrophotometry and, more recently, Inductively Coupled Plasma Mass Spectrophometry (ICP-MS). All of these methods (except NAA and ET-AAS for some sample types) require a prior sample digestion to decompose organic matter before detection can occur (Ducros et al. 1994).

The technique of choice depends on the sample matrix, sample concentration and the type of information required (e.g. isotope or selenium species). The selenium analysis process thus predominantly involves two main sequential steps, namely, digestion of samples and thereafter, detection of the selenium in samples. Any changes to improve any one part of the process cannot be made without consideration for the other, because of the complex chemical nature of the element selenium (Tarin 2006).

As a result of the very narrow margin between deficient and toxic levels of selenium in healthy organisms, it is crucial that any laboratory rendering the service of selenium analysis takes steps to ensure the accuracy of their diagnostic service. Thus, during the development of an improved selenium analysis technique, it is important to take into consideration the specific process requirements, namely, total decomposition of organic material (chemically releasing selenium from the selenomethionine and selenocysteine form in biological samples), and the preservation of selenium yield (as selenium volatilises from samples at a temperature exceeding 70 °C).

Sample digestion in a closed system, such as a microwave digestion system, was believed to be capable of overcoming this challenge. For fluorometric detection of selenium, however, the oxidation state of selenium, as well as interference from nitrate ions and analyte pH which occurs as a result of microwave digestion, severely inhibits accurate selenium detection (Kurkova et al. 2008). Microwave digestion, as a means to overcome challenges associated with open heat-block digestion, is therefore only possible with an alternative detection method such as hydride generation or HG-AAS.

Laboratory trials conducted, returned that successful modification of the digestion method could not be accomplished without the additional modification of the detection technique. It was found that fluorometric detection of selenium was impossible in samples which had not undergone the required chemical redox reaction. The open-block digestion process enabled chemical redox reaction but microwave digestion did not return the same results; therefore, further processing of microwave digested samples was required. Furthermore, fluorometric selenium detection was also inhibited by a low sample pH and excess nitric acid, which remained in samples after microwave digestion. As a result, a Standard Operating Procedure (SOP), which employed the use of both microwave digestion and HG-AAS detection, was required.

• define(d) the problem or selenium analysis improvement opportunity phase
• measure(d) the process criteria phase
• analyse(d) the result criteria phase
• improve(d) and optimise(d) the selenium analysis process phase
• control(led) the selenium analysis process.

Systematic data analysis thus took place in the order which is specified by the Lean Six Sigma process map. During the Define phase of Lean Six Sigma, the purpose of the process was established. Furthermore, the extent to which the process met the requirements (customer) was defined during this phase, as well as how the process was monitored. The influence and extent of variation present in the process was also determined during this phase. Research design operational implications of the Define phase involved:

• the identification of process steps from a process map
• evaluation of a root cause analysis diagram
• assignable causes of process problems were listed
• Pareto analysis was conducted to highlight the most critical problem(s) in the process
• a detailed current state value map was drafted
• the improvement opportunity to be addressed was selected.

During the Measure phase both types of measures which were required, namely process measures (P-criteria) and result measures (R-criteria), were collected. Measures were analysed to provide an assessment of the current performance of the process, and to obtain information of changes to the process. Signals of potential problems in the process were also identified through the evaluation of P-criteria. A controlled experiment was conducted to obtain R-criteria. The controlled experiment involved the processing of 10 non-probability, purposively selected liver samples from an archived population of all the previously processed samples processed by the laboratory. In addition to these, an in-house control sample was added to the sample group. All samples in this sample group were processed with the current state selenium analysis method, and replicated 13 times. Furthermore, during the six trial runs, an international Certified Reference Material (CRM) sample was included to determine the deviation from trueness from the true accurate value (concentration). The confidence intervals of the CRM, and the quality of the standard curve fit, served as a guideline for the determination of analytical accuracy.

A blank sample, as well as a set of five standard calibration samples, was included with each process trial run. The experiment was designed so that measurement data obtained from the calibration standards served the purpose of assessing the analytical precision of the process.

During the Analyse phase of the Lean Six Sigma process, the primary tools used to analyse data are inherently Six Sigma tools, namely statistical tools such as SPC, including Analysis of Variation (ANOVA) and confidence intervals, as well as regression analysis. This phase is referred to as ’Listening to the voice of the process‘ (Process Management International 2009).

The analysis and interpretation of R-criteria was thus conducted in the following order:

• Total Regression uncertainty analysis. During the Total Regression uncertainty analysis, Quantitative method characteristics of the standard calibration curve of the current state fluorimetry selenium analysis process were identified through the analysis of raw data R-criteria.

• Certified Reference Material analysis. The accuracy and precision of CRM measurements were analysed also with respect to average linearity and characteristics established for total regression analysis of the average process standard calibration curve.

• Precision, Analytical Bias and HorRat (Horwitz ratio) analysis. The precision or closeness with which results of the replicated analyses of a sample agree, the bias or consistent deviation of analytical results from true value’, as well as the HorRat ratio or method’s performance with regard to precision, of the current state selenium analysis process were determined.

• Analysis of individual standard curves of process runs. Data collected from each individual process run, were analysed with statistical tools by plotting the data into scatter plots and histograms and with ANOVA hypothesis testing.

• Systematic error analysis. A Cochran variant of the t-hypothesis test was conducted as a means to evaluate systematic error, related to accuracy in the process. This determined whether systematic error present could be considered as statistically significant by assessing whether or not the average CRM measurement obtained from process runs was different from the value stated by the reputed CRM supplier.

• Process capability analysis. This involved statistical analysis that determined whether the analytical test process was capable of performing within the tolerance level as specified by the internationally certified CRM supplier. Divergent from process capability studies conducted by manufacturing quality practitioners, in the specialist area of clinical chemistry, an important factor known as total analytical uncertainty, has to be taken into account. As a result, based on literature review, process capability indices were adapted accordingly to make provision for the important laboratory consideration.

During the Improve phase, focus shifted to the evaluation of P-criteria once again. An identifiable characteristic of this phase involved improvement solutions being developed to address root causes of problems identified. As a result, alternative remedies were evaluated to secure a quality improvement on the process. A quality-tool known as Kaizen analysis, taken from the Lean bouquet of quality-tools, was the primary tool used in this phase. The Improve phase provided improvement direction, and led to the redesign of the selenium analysis process.

The Optimise process sub-phase as employed to secure the best possible process improvement solution and this sub-phase relied on R-criteria data, in addition to P-criteria. The P-criteria were used to develop improvement remedies as part of the research conducted to improve the selenium analysis process, and R-criteria obtained from the improved designs were used to evaluate the effectiveness of the solutions developed. The following R-criteria analysis and interpretation formed part of the Optimise sub-phase:

• Process variance analysis. Statistical analysis determined whether the variance observed in the two analytical processes differed with statistical significance from each other. An independent t-test was used to compare the variance around the means of the two laboratory methods.

• Non-parametric difference in mean analysis. This type of analysis was performed to determine the extent of the quality improvement, as a result of the Lean Six Sigma process modification, whereby the R-criteria process means of the two analytical processes were statistically compared. The variable of interest in this type of analysis was the difference between the values of replicate observations, rather than the values of the observations themselves.

The Control phase of Lean Six Sigma utilised a feedback loop, whereby the standard performance of the improvement process was compared to actual performance, and action was taken to maintain the gains of improvements secured. In this phase, the quality-tool, FMEA, was used as the final step in the approach to the improvement of the selenium analysis process of the Western Cape Provincial Veterinary Laboratory. The steps of this phase were, (1) design controls and document the improved process, (2) validate the measurement system, (3) determine the final process capability and (4) implement and monitor controls.

The DMAIC method emphasises data analysis and fact-based decision making. The essence of the DMAIC method is to reduce variation in a process in order to achieve high conformance to customer requirements. As a result, the Lean Six Sigma methodology provides an exceptional research structure and design, as used for the purpose of improving an analytical process.

A schematic of results obtained during the Define phase is seen in the form of an Ishikawa chart (Figure 1), a Pareto chart (Figure 2, Table 2 and 3) and a Current State Value Stream (Figure 3 and Table 4). The phase entailed a detailed evaluation of the current state selenium analysis process from data obtained from the process steps. The use of basic quality-tools enabled the analysis of the primary process data, or P-criteria.

FIGURE 1: Ishikawa root cause analysis diagram.

FIGURE 2: Pareto chart – Pareto analysis: Selenium analysis process.

FIGURE 3: Current state value stream map.

FIGURE 4: Future state value stream map.

• in the current state selenium analysis process, result quality was variable
• in the current state process, accuracy and process precision could be improved
• the overall current state process performance was poor
• systematic error was present in the current state selenium analysis process
• the current state selenium analysis process was experiencing a disparate loss in selenium yield from process run to process run.

The statistical analysis of research data centred on the determination of the following: The calculation of the t_stat, known as the calculated t statistic, also known as t_calc. In statistics, the t distribution is used to estimate the population mean when ơ is unknown. The t_crit, or critical value was obtained from the t-distribution table. The calculated f-statistic, or f_calc, also known as f_stat was calculated. In statistics, the f-distribution is used for testing the equality between the variances of populations. The f_crit, or critical value was obtained from the f-distribution table. Analysis of Variation, a statistical methodology, was used to compare the means of groups. The regression SS, also known as Regression Sum of Squares (SSR), which is equal to the sum of the squared differences between the predicted value of Y and the mean of Y was determined, whereas the residual SS, also known as the Error Sum of Squares (SSE), which is equal to the sum of the squared differences between the observed value of Y and predicted value of Y, was also determined. Significant linearity exists where SSR > SSE. Furthermore, the Regression MS was determined. The Regression MS is

also known as Regression Mean Square (MSR) = , where SSR is Regression Sum of Squares or explained variation and residual MS, which is

where k = number of independent or explanatory variables in the regression model. Furthermore, the S_y/x was calculated. The standard error of regression, also known as Standard Error of the Estimate, or Residual Standard Deviation or Random Calibration Uncertainty, is a measure of the amount of error accrued in predicting a y-value for each given x-value. This statistic measures the variability of the actual y-values from the predicted y-values. It is a measure of variation around the fitted line of regression and is measured in units of the dependent variable y, and was determined.

The S_b, also known as Standard Error of Slope or S_b1, is a statistic that determines the sensitivity of the calibration function or the rate at which the signal changes with concentration. The S_a, Standard Error of the Intercept or S_a1, is the error associated with the actual intercept point of the line. The intercept of the regression line has implications for the smallest detectable signal (measured response) and the corresponding lowest concentration. These were also calculated during statistical analysis of the research data. Confidence Levels (CL), is the probability that the interval estimate will include the population parameter. b, also known as the corresponding confidence level for b associated with S_b (interval of the slope), was calculated by using the t-statistic for (n-2) degrees of freedom. t_Sb, calculated as the product of t_crit and S_b, whilst a, or the corresponding confidence level for a, associated with S_a (y-intercept), was calculated in the same way as that of the slope. t_Sa, calculated by using the t_crit and S_a; the Limit Of Detection (LOD) were determined as X_lod or the concentration representing LOD, and Y_lod or the response representing the LOD. The Limit Of Quantitation (LOQ) was calculated where Y_loq, known as the signal representing the LOQ, was calculated.

The Improve phase involved the Kaizen analysis of P-criteria, in addition to further R-criteria analysis of data obtained from the improved analytical method process (Table 7).

• sample digestion was the root cause of problems associated with the process of selenium analysis
• a value stream of activities was identifiable in the current state process
• muda were identified as a form of process waste to be removed through modification
• muri was identified as a form of process waste to be removed through modification
• mura was identified as a form of process waste to be removed through modification
• heijunka (process improvement) was possible in the form of process modification
• the design of process control features assures the quality output of the modified process design.

Analysis that took place in the Control phase of Lean Six Sigma entailed FMEA, a quality-tool, whereby P-criteria, or information on potential failure points of the improved process, were analysed (Table 8).

The current state selenium analysis process was found to be reasonably accurate and reasonably precise, but a definite systematic error is present in the process. Systematic error is a known cause of poor accuracy and precision. Unstandardised process steps resulted in a systematic error in the process. This process performance is unacceptable and the process requires heavy control and inspection. With a process capability index of 1.02 the process is, however, potentially capable.

Process variation was out of statistical control, and data analysis found evidence of disparate loss in selenium yield, as process results varied quantitatively from weekly process runs. Process redesign offered a remedy in the form of a modified selenium process, assumed to produce superior results in terms of accuracy and precision. Modified process design enabled control features to be built into the design.

Data error was categorised as random error or systematic error. Random error is derived from unpredictable variation, and therefore cannot be controlled. Reducing uncertainty in the system, however, helps to correct systematic error as well as to minimise random error. Increasing the amount of observations of a measurement is considered to be one avenue of reducing uncertainty, although uncertainty may never be reduced to zero (Carpi & Egger 2003). As a consequence of error, scientific measurements were not reported as single values, but rather as ranges, or averages with estimates of the error surrounding the value, after repeated measurement of the value. All analytical data for the research conducted were reported with confidence levels to account for error, as well as the associated confidence limits, to reflect the precision of statistical estimates computed during research.

Furthermore, the inclusion of CRM analysis conducted as part of the research process also supports the validity of research data results.

A secondary objective ‘to determine if a modification of the existing hot plate digestion to microwave digestion is capable of reducing the risks associated with the digestion procedure’, was stated. The result of the process capability studies during Lean Six Sigma highlighted that the elimination of quality risks, such as process variation, results in an improved analytical process. The literature review (Desert Research Institute 2004; University of Alberta, Department of Environmental Health and Safety 2011) highlighted that the use of perchloric acid is hazardous to the process operator and to the environment. As a result, a modification of the selenium analysis method away from the use of perchloric acid may be regarded empirically as a form of process improvement. Furthermore, the use of the FMEA quality-tool culminated in further functional practical value to the Biochemistry operational system by ensuring that risks associated with the modified selenium analysis method, were minimal.

Key research findings which focussed on the secondary objective stated as ‘determining if a modification of the existing digestion method will result in the optimisation of selenium yield of the process’, was generated from result data, which served as evidence of the disparate loss of selenium yield during the current state selenium analysis process. This selenium loss was attributed to process variation, and a standardised modification of the selenium analysis process overcomes this quality problem. Foster (2007) contends that the Improve, or Kaizen, phase of Lean Six Sigma involves off-line experimentation, whereby the factors affecting process performance are analysed. Kurkova et al. (2008) stated that, at temperatures exceeding 70 °C, organic forms of selenium volatilise from samples. Optimisation of the process to microwave digestion and HG-AAS detection was therefore regarded as a quality improvement of the process.

The final research objective stated, namely ‘to determine whether a modification of existing digestion procedure will ultimately translate into an improvement in terms of reliability of results’, had clear tangent planes to the independent variable of the research hypothesis, that is, quality improvement and laboratory results. Key findings, based on the analysis of research data, were that both accuracy and precision of the current state selenium process would be improved by a modified selenium analysis process. Both process capability studies and process performance studies conducted supported this finding, which further emphasised the importance of this practical improvement to biochemistry laboratory processes. Reliability of an analytical method is the ability of the method to be both accurate and precise. Therefore, an improvement in both accuracy and precision with the modified selenium analysis process, would ultimately improve the reliability of the analytical process, and thereby the quality of the process.

• the current state selenium analysis process must be modified in order to assure the reliability of process results
• process modification is necessary to address process variability that resulted from process steps, time and control procedures in the process, which were found to be responsible for disparate results
• process modification from open heat-block to microwave digestion is recommended
• process modification from fluorometric selenium detection to hydride generation detection is recommended
• it is recommended that the modified process be monitored, and that continuous evaluation take place according to the Lean Six Sigma cycle to maintain a focus on continuous improvement of the selenium analysis process.

In addition, a modified process design would ultimately translate into an improvement in quality in terms of the reliability of process results. On the basis of conclusions drawn from key research findings, it can be stated with confidence that:

there is significant evidence to accept the stated null-hypothesis, and it was found that variability of the dependant variable of the research hypothesis, associated with the process, had a significant and detrimental impact on the independent variable of the research hypothesis, namely, the result quality of the selenium analysis process. (Cloete 2011, n.p.)

As a result, the research hypothesis was accepted as true.

American Association for Clinical Chemistry, 2011, How reliable is laboratory testing?, viewed 03 July 2011, from http://labtestsonline.org/understanding/features/reliability/?start=0

American Society for Quality, 2011, Lean Kaizen: A simplified approach to process improvement, viewed 07 October 2011, from http://asq.org/training/lean-kaizen-a-simplified-approach-to-process-improvement_KAIZEN.html

Anvari, A., Ismail, Y. & Hojjati, S., 2011, ‘A study on total quality management and Lean manufacturing: Through Lean thinking approach’, World Applied Sciences Journal 12(9), 1585–1596.

Berte, L., 2007, CD-ROM, ‘Laboratory quality management: A roadmap’, Clinical Chemistry and Laboratory Medicine 27, 771–790.

Blumberg, B., Cooper, D. & Schindler, P., 2011, Business research methods, third European edition, McGraw-Hill Education, Berkshire.

Byrne, G., Lubowe, D. & Blitz, A., 2008, Driving operational innovation using Lean Six Sigma, viewed 27 September 2011, from http://www-935.ibm.com/services/at/bcs/pdf/br-stragchan-driving-inno.pdf

Cawley, J., 2000, Process capability analysis in laboratory quality control, viewed 10 March 2011, from http://www.nwasoft.com/press/mag_sci0110

Carpi, A. & Egger, A., 2003, The scientific method, viewed 26 June 2011, from http://www.visionlearning.com/library/module_viewer.php?mid=45

Cloete, B., 2011, ‘An approach to the improvement of the selenium analysis process of the Western Cape Provincial Veterinary Laboratory’, M. Tech thesis, Dept. Industrial and Systems Engineering, Cape Peninsula University of Technology.

Dahlgaard, J., Kristensen, K. & Gopal, K., 2002, Fundamentals of total quality management: Process analysis and improvement, Taylor & Francis, London.

Desert Research Institute, 2004, Guidelines for using and storing perchloric acid, viewed 10 October 2011, from http://www.safety.dri.edu/.../PerchloricAcidGuidelines.pdf

Ducros, V., Ruffieux, D., Belin, N. & Favier, A., 1994, ‘Comparison of two digestion methods for the determination of selenium in biological samples‘, Analyst 119, 1715–1717. http://dx.doi.org/10.1039/an9941901715, PMid:7978324

Foster, S., 2007, Managing quality: Integrating the supply chain, Pearson Prentice Hall, New Jersey.

FMEA Information Centre, 2011, Failure modes effects analysis, viewed 05 October 2011, from http://fmeainfocentre.com/index.htm

Fraser, A., 2011, Email, 08 September, allan@allanfraserandassociates.co.za

Gryna, F., Chua, R. & DeFeo, J., 2007, Juran’s quality planning and analysis: For enterprise quality, McGraw-Hill, Singapore.

Hare, L., 2001, Chicken soup for processes: Understanding variation as a prerequisite for SPC, viewed 10 July 2011, from http://asq.org/pub/qualityprogress/past/0801/qp0801stats.pdf

Hubbard, B., 2010, Muda, Mura and Muri, viewed on 11 October 2011, from http://bobsleanlearning.wordpress.com/2010/01/14/muda

Jiang, W., Murphy, T. & Tsui, K., 2006, ‘Statistical methods for quality and productivity improvement’, in H. Pham (ed.), The Springer Handbook of Engineering Statistics, pp. 173–174, Springer, New Jersey.

Khalil, M., Khan, M. & Mahmood, T., 2006, ‘Lean Six Sigma – A tool to improve productivity, quality and efficiency in manufacturing and industrial sector’, paper presented at the Official Conference of the International Association for Management Technology, P. R. China, Beijing, 22–26 May.

Kurkova, T., Skrypnik, L. & Zalieckiene, E., 2008, ‘Features of plant material pre-treatment for selenium determination by atomic absorption and fluorimetric methods’, Chemija 19(1), 40–43.

Marquis, P., 2011, Email, 21 July, phil@multiqc.com

Moen, R. & Norman, C., 2011, Evolution of the PDCA cycle, viewed 02 July 2011, from pkpinc.com/files/NA01MoenNormanFullpaper.pdf

Nevalainen, D., 1999, CD-ROM, ‘The quality systems approach’, Archives of Pathology & Laboratory Medicine, 123.

Park, S. & Anthony, J., 2008, Robust design for quality engineering and Six Sigma, World Scientific Publishing Co. Pte. Ltd., Singapore. http://dx.doi.org/10.1142/6655

Process Management International Ltd., 2009, Green Belt Workshop: Part 1, Published in-house, Meriden.

SANS 17025, 2005, General requirements for competence of testing and calibration laboratories, Standards South Africa, Pretoria.

Scott, J., 2007, CD-ROM, ‘Process optimization for service organisations isn’t rocket science’, Quality Progress, 5.

Tarin, P., 2006, ‘Distribution, speciation and geochemistry of selenium contaminated marine sediments – Port Kembla Harbour, NSW, Australia’, PhD thesis, School of Earth and Environmental Sciences, University of Wollongong.

University of Alberta, Department of Environmental Health and Safety, 2011, CCIS 2 Fume Hood Standard Operating Procedure, Published in-house, Canada.

Van Leeuwen, J., Nauta, M., De Kaste, D., Odekerken-Rombouts, Y., Oldenhof, M., Vredenbregt, M. et al., 2009, ‘Risk analysis by FMEA as an element of analytical validation’, Journal of Pharmaceutical and Biomedical Analysis 50(5), 1085–1087. http://dx.doi.org/10.1016/j.jpba.2009.06.049, PMid:19640668

Wang, J., 2008, ‘Quality process improvement’, in S. Gad (ed.), Pharmaceutical Manufacturing Handbook, Regulations and Quality, n.p., John Wiley & Sons Inc., New Jersey.

Woodall, W., 2000, ‘Controversies and contradictions in statistical process control’, Journal of Quality Technology 32(4), 341–350.

Wu, C., Pearn, W. & Kotz, S., 2009, ‘An overview on theory and practice on process capability indices for quality assurance’, International journal of production economics 117(2009), 338–359.