Intra-laboratory quality control for chicken raw materials

. In recent times, food and water laboratories have aimed to improve their processes to meet market demands. In order to contribute to the efficacy and development of these processes, new methods must be sought, among which is intra-laboratory quality control. This research study portrays the use of this method in a food and water laboratory, in order to analyze the intra-laboratory statistical control of chicken raw materials. To that end, the methodology applied was: Exploratory data analysis; Cochran’s test, to assess the homogeneity of variances; Dixon’s Test, to verify whether the means are part of the same distribution; calculation of repeatability and reproducibility; calculation of the uncertainty of the mean; Snedecor’s F method, to compare the variances; and Dixon’s test, to compare laboratory variances. The analysis parameter analyzed in the studied raw material was the proficiency of technicians with regard to the accuracy of analyses of total coliform count in the raw materials prior to chilling. The final results showed low data variation around the mean, as well as significant reliability of the obtained results.


Introduction
Laboratory Best Practices are part of the integrated quality management system used to assess organizational procedures and the conditions under which laboratory research studies are planned, performed, tested, monitored, recorded, inspected and archived.The objective of intra-laboratory quality control is to meet the requirements of rule NBR ISO 17025:2005 and the demands of Brazil's Ministry of Agriculture, Livestock and Supply -MAPA.
According to studies proposed by Henry (1999), Quality Control for Food and Water Laboratories are systems devised to recognize and minimize lab analysis errors; they provide analysts with the criteria to identify and analyze laboratory performance, in order to obtain reliable and safe results.
The same author also reports that quality control programs in clinical laboratories were introduced in Brazil during the 1970s-1980s through the National Quality Program (PNQ) of the Brazilian Society of Clinical Analysis (SBAC).
Quality control may be defined as every systematic action to support and maintain laboratory services in order to meet client needs.It is regarded as a component of the Quality Management System (MOTTA et al., 2001).
Acta Scientiarum.Technology Maringá, v. 35, n. 3, p. 439-443, July-Sept., 2013 The view of statistical control of the process is associated with lower variability and provides improved quality levels in production results.This becomes important when dealing with best processes -meaning not only better quality, but lower costs as well.Costs decrease for two reasons, namely: sampling inspection and waste reduction (PALADINI, 2004).
Intra-laboratory trials are assays performed by several technicians on a standard sample, in order to evaluate proficiency and obtain the grouped repeatability variance value of the laboratory for each assay technique (LOPES, 2003).
Based on the above considerations, it can be observed that there currently is great concern by food and water laboratories regarding the accuracy of the results obtained in routine laboratory analysis procedures.In that regard, statistical techniques meet that need.Intra-laboratory quality control emerges as a prospect for better evaluating and demonstrating the reliability of the obtained results.
The objective of this work was to perform an intra-laboratory statistical control study on chicken raw materials at a food and water laboratory, in order to assess the efficiency and accuracy of the obtained results.

Material and methods
The methodology used in this study was quantitative and descriptive.The following statistical methods were used to analyze the research in question: 1. Exploratory data analysis (Cochran's test) (LOPES, 2003), to assess the homogeneity of variances; 2. Dixon's test (LOPES, 2003), to verify whether the means are part of the same distribution; 3. Calculation of repeatability and reproducibility (LOPES, 2003); 4. Calculation of the uncertainty of the mean, using the formula: In which: Ux = uncertainty of the mean; St = total standard deviation; n = total number of determinations; t (n-1; 0.025) = Student's t variable.5. Snedecor's F method to compare the variances; 6. Dixon's test to compare Laboratory variances.Using Cochran's test, the homoscedasticity of variances is compared; that is, it is possible to assess whether the variance of the obtained results by a group is excessive compared to the other groups; it is a one-sided test.The C test is performed according to the following formula: : is the maximum variance of the group.
2 i S  : is the sum of all sample variances in the group.
The decision rule is as follows: if the calculated value of C is lower than the obtained C value (LOPES, 2003), the variances are equal to 5% significance.
Dixon's test compares the individual values obtained by a single operator or the mean values obtained by several operators, or yet differences between two results obtained by several operators; it is a two-sided test (LOPES, 2003).
In which Z (H), Z (H-1) and Z (1) are minimum and maximum order statistics, respectively.
Repeatability refers to tests performed under conditions as constant as possible, known as repeatability conditions.The results of mutually independent tests are obtained using the same assay method, identical materials, by the laboratory, same operator, and using the same equipment over short intervals.The standard deviation of the test result obtained under repeatability conditions is known as repeatability standard deviation.It is a parameter of the dispersion of the distribution of test results.With the repeatability standard deviation, the socalled repeatability r value is calculated: the modulus of the difference between the two test results obtained under repeatability conditions is then calculated.The likelihood that this difference is lower than the repeatability r value is 95% (LOPES, 2003).
In which t is the value found in Student's t-table, with alpha= 0.05 (two-sided) and with the residual degree of freedom: Sr: is the repeatability standard deviation.
Reproducibility refers to tests performed under varying conditions, named reproducibility conditions.The results are obtained using the same assay method and identical materials, but in different laboratories, by different operators and using different equipment, with large time intervals between tests.The standard deviation of the test result obtained under reproducibility conditions is known as reproducibility standard deviation.It is a parameter of the dispersion of the distribution of test results.Com o reproducibility standard deviation, the so-called reproducibility R value is calculated: the modulus of the difference between the two test results obtained under reproducibility conditions is then calculated.The likelihood that this difference is lower than the reproducibility R value is 95% (LOPES, 2003).
In which t is the value found in Student's t-table, with alpha= 0.05 (two-sided) and with the degree of freedom SR: is the reproducibility standard deviation.
There are several tests involving the parameters of a multiple regression model or multiple correlation coefficient.The standard test is the analysis of variance, in which the explained variation is compared with the unexplained variation of the dependent variable.This relationship features Snedecor's F distribution with k and (n-k-1) degrees of freedom, with k equal to the number of regressors and n equal to sample size.The calculated statistical parameter F calculado is then compared to table parameter F (k,n-k-1) .If F calculado > F (k,n-k-1) , the null hypothesis of non-existence of a linear relationship is rejected, according to the indications of  % significance (1% or 5%) of evaluation norm (in order to reach Degree III); in other words, the regression equation is accepted.For the obtained model, that test is performed using the F calculado equation, calculated by the formula: In which: SQM: Sum of the Squares of Technicians; SQM: Sum of the Squares of Residuals.

Results and discussion
In order to apply the intra-laboratory techniques in the food and water laboratory used in the study, it was verified that the laboratory has adopted a complete management system for laboratory best practices in all procedures, from the moment samples are received to the process of issuing analysis reports.
Table 1 shows a summary of the results for mean and variance per technician.
In Table 1 are explained the results of the analyses obtained by each technician, the mean of results, standard deviation and variance.From those data, it can be observed in the obtained results that the variation was greater for technicians B and C, compared to the mean and variance of the other technicians.This indicates there was greater variability, which may be related to different factors inherent to the performed analysis.As such, certain aspects can be taken into account, such as: skill and experience of each technician, the manner in which the procedure was interpreted, and the condition of the sample at the time of analysis.Other factors that can favor the variability of a given process or procedure have to do with the environment, calibration and accuracy of equipment, among others.
Cochran's test was performed to assess whether the variances are equal (homoscedasticity) in each technician.The obtained result demonstrates that .
Thus, the variances have 5% de significance, as follows: Table 1 also includes the summary of means; it can be observed that the distribution of the means is in ascending order, considering that the highest mean is for technicians C and D.
Thus, the statistical calculation using Dixon's method to verify whether the means are part of the same distribution:  According to the calculations of Repeatability and Reproducibility and from the obtained results, the laboratory achieved repeatability variance of r = 2.09563E + 8 and reproducibility variance of R = 3.82917E + 7. The 173354 difference is acceptable and both analyses are valid, because that difference is lower than r and R.; thus, both values are acceptable.Werkema (1996) affirms that R and r are indexes linked to the accuracy of results, and other measurements are necessary for these values to be expressed correctly, without losing the physical significant that should be attributed to those numbers.That author also argues that certain factors inherent to measurement instruments may not show appropriate repeatability, such as lack of cleaning or maintenance of the instrument, environmental conditions, wear and tear of measurement components, among others.
When comparing the means among technicians, there are: When calculating Repeatability (r), Reproducibility and Uncertainty of the mean, according to the analyzed data, 16 degrees of freedom were accurately determined, obtained with four replications and four analysts; one of the means was within the interval of another mean, so they are compatible.The calculation of the uncertainty of the mean aims to inform the proximity of the values found, and show whether there is compatibility between them.
Therefore, it can be concluded that the means are compatible, because one of the means is within the interval of the other.
In the comparison between the variances of the technicians, the result of . Thus, the F-test shows that the variances are similar with regard to the accuracy of the analyses, because the calculated F was higher than table F.
Considering the context of the analyses, laboratory variance also showed similarities; that is, are similar, because they showed low data dispersion.For the analysis of the coefficient of variation (Standard CV < = 20%), the behavior of the data shows that, for carcass analysis prior to chiller, the coefficient of variation is significantly representative of the mean.This indicates low data dispersion (low variation); the analysis resulted in mean coefficient of variation = 12.71%.
The analysis of the obtained results evidenced the importance of the accuracy of laboratory instruments to issue analysis reports.Currently, customers require increasing levels of guarantee in rendered services, as an analysis report can be the basis for decisions taken by companies and the general population.
Mazzuchetti et al. ( 2010) comments that companies must be aware of the cost of devising their products seeking efficacy, so that they are compatible with their target, as some customers do not acquire what they desire for economic reasons.When the need arises, customers rely on means to collect precise information.
Likewise, Cima and Uribe-Opazo (2009) highlight that one of the fundamental aspects to be considered in managing companies is the manner in which work is organized, the understanding of the functions of each person, and clear assignments of responsibility.Those aspects bring necessary elements so that each person can perform their activities -highlighting the importance of the normality of a process, calibration and gauging of equipment.Quality management techniques are modern concepts that aim to contribute efficiently to the survival of companies in ever-changing markets (CIMA; URIBE-OPAZO, 2010).

Conclusion
The study performed herein demonstrated what technician proficiency with regard to the accuracy of analyses is within a statistically efficient normality.The analyses of variance were higher for technicians B and C, and the results of technicians C and D showed the greatest homogeneity -that is, lowest data dispersion -featuring a higher representativeness of the mean, as their coefficients of variation were near CV= 20%.In an overall evaluation, a small variation of the data was detected around the mean, as well as significant representativeness of the obtained results.
In that sense, the study also showed that the application of statistical techniques proved efficient in a context of continuous improvement, considering that the use of the scientific method is important in decision-making.Therefore, it was demonstrated that the use of information technology through statistical techniques is of fundamental importance for the process of integrated laboratory management.

Table 1 .
Summary of the results for mean and variance per technician.

Table 2 .
Summary of the results obtained from the statistics.