Classification of Iranian bottled waters as indicated by manufacturer’s labellings

Food Chemistry 120 (2010) 1218–1223

Contents lists available at ScienceDirect

Food Chemistry journal homepage: www.elsevier.com/locate/foodchem

Analytical Methods

Classification of Iranian bottled waters as indicated by manufacturer’s labellings K. Yekdeli Kermanshahi a, R. Tabaraki a,*, H. Karimi b, M. Nikorazm c, S. Abbasi c a

Chemometrics Lab, Department of Chemistry, Faculty of Science, Ilam University, Ilam, Iran Faculty of Agriculture, Ilam University, Ilam, Iran c Department of Chemistry, Faculty of Science, Ilam University, Ilam, Iran b

a r t i c l e

i n f o

Article history: Received 31 December 2008 Received in revised form 24 June 2009 Accepted 29 November 2009

Keywords: Bottled water Cluster analysis Principal component analysis Chemometrics

a b s t r a c t Water is the most important substance in our daily lives and contains minerals which play an important role in our nutrition. In this study, the chemical composition of Iranian bottled water brands were investigated by correlation analysis, principal component analysis and hierarchical cluster analysis. For this purpose, the chemical composition reported on the label of 73 Iranian bottled waters was used as data set. It was found out that only 26 brands had eight important parameters such as calcium, magnesium, potassium, sodium, chloride, sulphate, bicarbonate and fluoride and 20 brands had acceptable charge balance error. Results showed that Iranian bottled waters can be divided into 11 classes. Most of them were Ca–Mg–HCO3 type. The relationships among selected variables were also examined by Piper diagram. The best brands were introduced for common customers and kidney stone patients. The chemical content of Iranian bottled water brands was also compared with some world standards. It was observed that only one of the brands had fluoride in excess compare to that of standard values. Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction The world market of bottled water has grown quickly and is considered as a global billion dollar business (Güler, 2007a, 2007b; Ikem, Odueyungbo, Egiebor, & Nyavor, 2002; Versari, Parpinello, & Galassi, 2002). In the countries such as the United Arab Emirates, nearly 90% of the population drinks bottled mineral water (Nsanze, Babarinde, & Al Kohaly, 1999). The dramatic increase in the consumption of bottled water worldwide has been attributed to the consumers’ concern over increasing water pollution and their objection to offensive tastes and odours such as chlorine from municipal water supplies and bacterial contamination (Saleh, Ewane, Jones, & Wilson, 2001). Another reason is a common belief that mineral waters have beneficial medicinal and therapeutic effects (Warburton, Dodds, Burke, Johnston, & Laffey, 1992). Apart from the use of bottled water as drinking water, it has found wide usage in infant formula preparation, reconstituting other foods, also for cleaning contact lenses, skin care and filling humidifiers. The ideal bottled water should be rich in magnesium and calcium and have low sodium content (Garzon & Eisenberg, 1998). Epidemiologic and clinical studies suggest that magnesium may reduce the frequency of sudden death, sodium contributes to the occurrence of hypertension, and calcium may help prevent osteoporosis (Garzon & Eisenberg, 1998). This will help individuals to * Corresponding author. Tel./fax: +98 841 2227022. E-mail address: [email protected] (R. Tabaraki). 0308-8146/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.foodchem.2009.11.067

achieve the recommended daily allowances of these minerals. Because wide variations exist in the mineral contents of commercially available bottled waters, further information on the characterisation of Iranian bottled water is needed to protect the consumers and to draw guidelines for quality control and the regulation of this industry. Multivariate analysis is widely used for food quality evaluation and differentiation or classification of food samples. Among the different multivariate techniques, cluster analysis (CA) and principal component analysis (PCA) are great potential for classification of problems. The purpose of this paper is: (1) to investigate the chemical characteristic of domestic brands of bottled water sold in Iran (2) to classify them by utilising parameters reported on their government issued production licenses (3) to compare the chemical composition of water samples with world standards. 2. Materials and methods 2.1. Bottled water database The chemical compositions reported on the label of 73 bottled waters were used as data set for this study. This data was obtained generally by purchasing the bottled waters from different supermarkets in the Iranian cities and by telephone or electronic mail to manufacturers. The chemical parameter determinations were carried out and certified by official laboratories of analysis and their accuracy and precision were not questioned in this study.

K. Yekdeli Kermanshahi et al. / Food Chemistry 120 (2010) 1218–1223

Distribution of companies is shown in Fig. 1. The majority of water companies are found in western and north parts of Iran. Bottled water dataset is presented as Supplementary material. Out 73 brands, only 26 brands had eight important parameters  (Ca2+, Na+, Mg2+, K+, Cl, F, SO 4 and HCO3 ). Routine analysis of mineral water is carried out by each company on a daily basis, whereas a complete analytical control is scheduled, at least, with annual frequency. However, as an independent check on the quality of the chemical analysis in the data set they were tested for charge balance error (CBE) (Freeze & Cherry, 1979):

P P z  mc  z  ma P % CBE ¼ P  100 z  mc þ z  ma

ð1Þ

In Eq. (1), z is the absolute value of the ionic valence, mc the molality of cationic species and ma the molality of the anionic species. Calculated charge balance errors are less than 10% for 20 samples in the dataset, which is an acceptable error for the purpose of this study. The mean, standard deviation, minimum and maximum of charge balance error were 4.18, 2.14, 0.16 and 8.06, respectively. 2.2. Multivariate analysis 2.2.1. Correlation analysis Correlation analysis was applied to describe the degree of relation between two water chemistry parameters. In statistics, correlation coefficient indicates the strength and direction of a linear relationship between two variables. The correlation is 1 in the case of an increasing linear relationship, 1 in the case of a decreasing linear relationship, and some value in between in all other cases, indicating the degree of linear dependence between the variables.

1219

A correlation coefficient of zero suggests that the two variables are independent of each other. When the coefficient is closer to either 1 or 1, strong correlation occurs between the variables. A number of different coefficients are used for different situations. The best known is the Pearson r correlation coefficient which is obtained by dividing the covariance of the two variables by the product of their standard deviations. 2.2.2. Principal component analysis PCA is a well-known statistical method for reducing the dimensionality of data sets (Brereton, 2003). PCA is the simplest of the true eigenvector-based multivariate analyses. Often, its operation can be thought of as revealing the internal structure of the data in a way which best explains the variance in the data. If a multivariate data set is visualised as a set of coordinates in a high-dimensional data space (1 axis per variable), PCA supplies the user with a lower-dimensional picture, a ‘‘shadow” of this object when viewed from it is (in some sense) most informative viewpoint. PCA involves the calculation of the eigenvalue decomposition of a data covariance matrix or singular value decomposition of a data matrix, usually after mean centering the data for each attribute. The results of a PCA are usually discussed in terms of component scores and loadings. This approach has been used to extract related variables and infer the processes that control water chemistry (Güler, 2007b; Helena et al., 2000; Versari et al., 2002). A Varimax rotation is carried out in order to ensure that the resulting factors are uncorrelated and to facilitate the interpretation of the results. The number of PCs extracted is chosen by using Kaiser’s criterion where only the PCs with eigenvalues greater than unity are retained (Kaiser, 1960).

Fig. 1. Distribution of Iranian bottled water companies.

1220

K. Yekdeli Kermanshahi et al. / Food Chemistry 120 (2010) 1218–1223

Table 1 Chemical composition of major constituents reported on the label of 73 Iranian bottled waters. Parameter

Mineral water (53brands)

1

Calcium (mg L ) Magnesium (mg L1) Sodium (mg L1) Potassium (mg L1) Chloride (mg L1) Sulphate (mg L1) Bicarbonate (mg L1) Fluoride (mg L1) Iron (mg L1) Nitrate (mg L1) Phosphate (mg L1) pH a

Drinking water (20 brands)

a

n

Mean ± SD

Minimum–maximum

na

Mean ± SD

Minimum–maximum

52 52 47 48 45 45 31 38 8 28 6 39

50 ± 33 12 ± 7 8±1 1±2 12 ± 13 27 ± 40 179 ± 99 0.4 ± 0.5 0.04 ± 0.07 5±4 0.1 ± 0.1 7.4 ± 0.2

8–248 1.40–26.70 0.1–48 0.1–7 0.0001–60 0.33–195 0.44–585 0.04–2.9 0.01–0.18 0.1–19 0.01–0.31 7–7.96

20 20 16 16 19 19 8 9 – 10 – 14

33 ± 8 12 ± 8 17 ± 16 1 ± 0.8 36 ± 43 37 ± 30 106 ± 56 0.4 ± 0.3 – 8±5 – 7.4 ± 0.3

3.2–85.4 5–35 2–48 0.4–3 7–170 3.16–100 22.5–197.27 0.1–0.85 – 2–17 – 7–8.03

Number of sample.

2.2.3. Hierarchical cluster analysis Clustering is known as the classification of objects into different groups, or more precisely, the partitioning of a data set into subsets (clusters), so that the data in each subset (ideally) share some common trait – often proximity according to some defined distance measure. Data clustering is a common technique for statistical data analysis, which is used in many fields, including machine learning, data mining, pattern recognition, image analysis and bioinformatics (Beebe, Pell, & Seasholtz, 1998; Brereton, 2003). Hierarchical cluster analysis (HCA) is an unsupervised technique that examines the interpoint distances between all of the samples and represents that information in the form of a two dimensional plot called a dendrogram (Brereton, 2003). These dendrograms present the data from high dimensional row spaces in a form that facilitates the use of human pattern recognition abilities (Brereton, 2003). In order to generate the dendrogram, HCA methods form clusters of samples based on their nearness in row space. Initially every sample is considered as a cluster and join closest clusters together. This process is repeated until only one cluster remains. Variation of HCA use different approaches to measure distances between clusters (e.g., single, Euclidean and Mahalanobis distance). The linkage rules iteratively link nearby points (similar samples) by using the similarity (distance) matrix (Beebe et al., 1998; Brereton, 2003). The initial cluster is formed by linkage of the two samples with the greatest similarity. Ward’s method is distinct from all other methods because it uses an analysis of variance (ANOVA) approach to evaluate the distances between clusters and forms smaller distinct clusters than those formed by other methods (Ward, 1963). The Euclidian distance was selected as the similarity measurement, which is straight line distance between two points in c-dimensional space defined by c number of variables. 2.3. Software Statistics and chemometrics data analysis was performed by means of the MATLAB 7.1. MathWorks Inc. (2005) and MINITAB 14. Minitab Inc. (2003) softwares.

to the different geological origins of the waters. The chemical composition of natural waters is controlled by many factors such as chemistry of atmospheric precipitation, mineralogy of the rocks encountered along the underground flow path, residence time of the groundwater in the aquifer, climate and topography of the area (Güler, Thyne, McCray, & Turner, 2002). To get an insight into natural diversity in the composition of mineral waters, the type of water is defined by all ionic constituents that contribute at least 20% meq to the total anionic or cationic composition of water, where total equivalents of cations and anions were recorded as 100% (Table 2). 11 different water types were identified. The most frequently observed type is Ca– Mg–HCO3. However, there are no two drinking water brands that were classified in the same water type. The type of water does not give detailed information about the composition of a water sample. Different graphical and statistical techniques have been developed to describe the concentration or relative abundance of major and minor constituents and the pattern of variability in different water samples. These techniques are Collins bar diagram, pie diagram, Stiff pattern diagram, Schoeller plot, Piper diagram, hierarchical cluster analysis, principal component analysis (Güler et al., 2002). One of the most common graphical approaches to describe the abundance or relative abundance of ions in individual water samples is Piper diagram (Piper, 1944). To construct the Piper diagram, the relative abundance of cations with the % meq L1 of Na + K, Ca and Mg is first plotted on the cation triangle. The relative abundance of Cl, SO4, and HCO3 + CO3 is then plotted on the anion triangle. The two data point on the cation and anion triangles are then combined into the quadrilateral field that shows the overall chemical property of the water sample. Piper diagram of the 20 Iranian bottled waters was shown in Fig. 2. Most of the brands are calcium, magnesium and bicarbonate type waters. As shown in Table 2, all of the water classes had calcium and eight classes had bicarbonate. Multivariate pattern recognition methods were utilised to discriminate between separate groups (types or clusters) of water samples.

3. Results and discussion

3.2. Multivariate analysis

3.1. Chemical characteristics of bottled waters

Multivariate pattern recognition methods were utilised to sort bottled water brands into groups or clusters. Associations among variables can be demonstrated statistically by correlation analysis. In correlation analysis, Pearson r correlation coefficients are calculated for all possible pairs of variables. The results were shown that high significant correlation coefficient exist between Ca and HCO3 (r = 0.907) and Na and SO4 (r = 0.730).

The mean, standard deviation, minimum and maximum values of the 12 chemical parameters of 73 bottled water brands were determined and presented in Table 1. For most elements the difference between the lowest and the highest concentration was one to three orders of magnitude. These large variations can be attributed

1221

K. Yekdeli Kermanshahi et al. / Food Chemistry 120 (2010) 1218–1223 Table 2 Classification of the 20 bottled water brands based on total hardness.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Brand

Water type

Total hardness

Water quality based on total hardness

Aqua prime Bisheh Polur Dasani Damavand Surperize Sepidan Siva sar Saman Sheshpeer Kandowan Kooh rang Mah rooz Margoon Mabda Vata Pure life Persab Dimeh Gahar

Ca–HCO3 Ca–Cl–HCO3 Ca–Mg–HCO3 Ca–Mg–Na–Cl Ca–Mg–Cl–SO4 Ca–Na–Cl–SO4 Ca–HCO3 Ca–Mg–HCO3 Ca–Mg–HCO3 Ca–K–HCO3 Ca–Mg–Na–HCO3 Ca–Mg–HCO3 Ca–Mg–HCO3 Ca–Mg–HCO3 Ca–Na–SO4–HCO3 Ca–Mg–Na–SO4–HCO3 Ca–Mg–HCO3 Ca–SO4–HCO3 Ca–HCO3 Ca–Mg–HCO3

115 232 111 203 204 68 154 115 142 146 114 240 189 170 68 34 106 238 162 141

Hard Very hard Hard Very hard Very hard Moderately hard Very hard Hard Hard Hard Hard Very hard Very hard Very hard Moderately hard Soft Hard Very hard Very hard Hard

Bold names are drinking water brands.

Fig. 2. Piper diagram of the 20 Iranian bottled water.

PCA technique was used to reduce the number of dimensions present in the data matrix (reducing eight variables to three PCs in this study). Also this technique was used to select the most discriminating parameters, and to investigate the overall variation of data. Rotation of principal components was carried out using the Varimax normalised method and only factors with eigenvalues

greater than one were taken into consideration (Kaiser criterion). Varimax normalised procedure for eigenvector rotation resulted in three principal components (PC1, PC2 and PC3), which explained 81.5% of the total variance. The result of the PCA analysis of bottled water brands was shown in Fig. 3. The PC1, PC2 and PC3 contain 45.3%, 23.3% and 12.8% of total variance, respectively. As Fig. 3

1222

K. Yekdeli Kermanshahi et al. / Food Chemistry 120 (2010) 1218–1223

Ward’s method was used to obtain hierarchical associations. The result of the HCA is presented as a dendrogram (Fig. 4). The resulting dendrogram had four major groups based on a similarity of eight parameters. The first group is composed of brand 6 (mineral) and brand 4 (drinking). The second group is comprised three mineral waters and one drinking water. The third and forth groups are composed of the remaining brands. 3.3. Health implications of bottled water consumption

Fig. 3. Score plot: (a) PC1 vs. PC2 and (b) PC1 vs. PC3 of Iranian bottled waters.

Fig. 4. Hierarchical dendrogram from the HCA for the 20 bottled water brands.

shows, one of the mineral water (brand 6) and one drinking water (brand 4) are clearly different from the others. Hierarchical cluster analysis (HCA) was used for searching the natural grouping among bottled waters from different sources. The bottled water brands were classified according to their major ion composition. The data were standardised (z-scores) and the Euclidean distance was used as similarity measurement. The

Recommended dietary allowance has proposed a minimum daily required intake for magnesium (350 mg) and calcium (800 mg). For sodium, a maximum intake of 2400 mg has been recommended (Heany, Gallagher, & Johnston, 1982; Whitney, Corinne, & Sharon, 1991). Because the estimated minimum daily requirement of sodium for an adult (500 mg per day) is easily achieved in most diets, no minimum recommended intake of sodium has been set (Whitney et al., 1991). A highly significant relationship has been observed between average blood pressure and sodium intake in many populations around the world (McCarron, Holly, & Morris, 1982). Due to differences in absorbance, magnesium bioavailability may be greater from water than from food sources (Eisenberg, 1992; Jones, Manalo, & Flink, 1967). Cardiovascular disease rates may be inversely related to water hardness. Rates of cardiovascular mortality and sudden death are 10–30% greater in soft water areas (low in magnesium or calcium) than in hard areas (high in magnesium or calcium) (Anderson & LeRiche, 1971; Anderson, LeRiche, & MacKay, 1969). Naturally bioavailability calcium is found almost exclusively in milk, milk products, water and few vegetables such as parsley, broccoli and kale (Whitney et al., 1991). Nutritional surveys indicate that the calcium intake in many populations is below the daily recommendation (Whitney et al., 1991). The bioavailability of calcium in water is believed to be at least as high as that of milk and milk products (Couzy, Kastenmayer, & Vigo, 1995; Heany & Dowell, 1994). Therefore, selecting bottled water with high calcium content may help to achieve the daily recommended intake. On the other hand, ingesting too much calcium may lead to the formation of kidney stones (Coe, Parks, & Asplin, 1992). As low calcium diet (400–600 mg daily) has been suggested as a preventive treatment for patients with a history of kidney stone formation might benefit from avoiding bottled waters with high calcium content. Drinking water that is high in magnesium and calcium and low in sodium will help individuals achieve the recommended daily allowances of these minerals (Garzon & Eisenberg, 1998). As shown in Table 2, the water classes that have Ca and Mg and have not Na are the best. Therefore, Mahrooz, Sivasar, Polur, Koohrang, Saman, Margoon, Gahar and Damavand (mineral water brands) and Purelife (drinking water brand) are the best bottled waters. Since calcium carbonate is one of the more common causes of hardness, total hardness is usually reported in terms of calcium carbonate concentration (mg L1 as CaCO3). Total hardness of water samples was calculated by Eq. (2) (Crittenden, Rhodes Trussell, Hand, Howe, & Tchobanoglous, 2005) and listed in Table 2. Waters are typically classified as soft (0–50 mg L1 as CaCO3), moderately hard (50–100 mg L1 as CaCO3), hard (100–150 mg L1 as CaCO3) and very hard (above 150 mg L1 as CaCO3) (Crittenden et al., 2005).

Total hardness ¼ 2:5½Ca þ 4:1½Mg

ð2Þ

where [Ca] is calcium concentration (mg L1). The soft water is the best for kidney patients, since selected brand is Vata (mineral water). A review of the current regulations was made and compared to several standards around the world including European Economic

K. Yekdeli Kermanshahi et al. / Food Chemistry 120 (2010) 1218–1223

Community (EEC), World Health Organisation (WHO), US Environmental Protection Agency (EPA), International Bottled Water Association (IBWA) and US Food and Drug Administration (FDA) (Güler, 2007a). Results showed that all Iranian brands contain calcium, magnesium, potassium, sodium, chloride, sulphate and bicarbonate below the maximum allowable level. Only one of the brands (Surprize with 2.9 mg L1 fluoride) was exceeded from the standards. The maximum concentration allowed for fluoride are EEC (1.5 mg L1), WHO (1.5 mg L1), EPA (2 mg L1), IBWA (1.7 mg L1) and FDA (2.4 mg L1). 4. Conclusion The quality of water for human consumption has always been and still is one of the most serious challenges. In this study, 73 Iranian bottled water brands were characterised by descriptive statistical measures (mean, minimum and maximum). There is a large variation in the water composition of different brands. Twentysix brands of Iranian bottled waters had eight major ions data; therefore it is important to establish norms to regulate the quality and labelling of bottled waters. Twenty out of 26 brands had acceptable charge balance error, thus data of these 20 brands were subjected to different pattern recognition methods such as correlation analysis, PCA and HCA. Eleven different water types were identified in which the most frequently observed type were Ca– Mg–HCO3. Based on PCA results, one mineral water (brand 6) and one drinking water (brand 4), are completely separated from the others. On the basis of connecting distances between parameters, four clusters were distinguished. HCA confirmed the results of PCA. The best brands were introduced for common customers and kidney stone patients. The chemical content of Iranian bottled water brands was also compared with some world standards. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.foodchem.2009.11.067. References Anderson, T. W., & LeRiche, W. H. (1971). Sudden death and ischemic heart disease in Ontario and its correlation with water hardness and other factors. Canadian Medical Association Journal, 105, 155–160. Anderson, T. W., LeRiche, W. H., & MacKay, J. S. (1969). Sudden death and ischemic heart disease: Correlation with hardness of local water supply. New England Journal of Medicine, 280, 805–807. Beebe, K. R., Pell, R. J., & Seasholtz, M. B. (1998). Chemometrics: A practical guide. New York: John Wiley. Brereton, R. G. (2003). Chemometrics: Data analysis for the laboratory and chemical plant. Chichester: Springer.

1223

Coe, F. L., Parks, J. H., & Asplin, J. R. (1992). The pathogenesis and treatment of kidney stones. New England Journal of Medicine, 327, 1141–1152. Couzy, F., Kastenmayer, P., & Vigo, M. (1995). Calcium bioavailability from a calcium and sulfate rich mineral water, compared with milk, in young adult women. American Journal of Clinical Nutrition, 62, 1239–1244. Crittenden, J. C., Rhodes Trussell, R., Hand, D. W., Howe, K. J., & Tchobanoglous, G. (2005). Water treatment: Principles and design (2nd ed.). John Wiley. Eisenberg, M. J. (1992). Magnesium deficiency and sudden death. American Heart Journal, 124, 544–549. Freeze, R. A., & Cherry, J. A. (1979). Groundwater. Englewood Cliffs, NJ: Printice-Hall. Garzon, P., & Eisenberg, M. J. (1998). Variation in the mineral content of commercially available bottled waters: Implications for health and disease. The American Journal of Medicine, 105, 125–130. Güler, C. (2007a). Evaluation of maximum contaminant levels in Turkish bottled drinking waters utilizing parameters reported on manufacturer’s labeling and government-issued production licences. Journal of Food Composition and Analysis, 20(3–4), 262–272. Güler, C. (2007b). Characterization of Turkish bottled waters using pattern recognition methods. Chemometrics and Intelligent Laboratory Systems, 86, 86–94. Güler, C., Thyne, G. D., McCray, J. E., & Turner, A. K. (2002). Evaluation of graphical and multivariate statistical methods for classification of water chemistry data. Hydrogeology Journal, 10(4), 455–474. Heany, R. P., & Dowell, M. S. (1994). Absorbability of the calcium in a high-calcium mineral water. Osteoporos International, 4, 323–324. Heany, R. P., Gallagher, J. C., & Johnston, C. C. (1982). Calcium nutrition and bone health in the elderly. American Journal of Clinical Nutrition, 36, 986–1013. Helena, B., Pardo, R., Vega, M., Barrado, E., Fernandez, J. M., & Fernandez, L. (2000). Temporal evolution of ground water composition in an alluvial aquifer (Pisuerga River, Spain) by principal component analysis. Water Research, 34, 807–816. Ikem, A., Odueyungbo, S., Egiebor, N. O., & Nyavor, K. (2002). Chemical quality of bottled waters from three cities in eastern Alabama. The Science of the Total Environment, 285, 165–175. Jones, J. E., Manalo, R., & Flink, E. B. (1967). Magnesium requirements in adults. American Journal of Clinical Nutrition, 20, 632–635. Kaiser, H. F. (1960). The application of electronic computers to factor analysis. Educational and Psychological Measurement, 20, 141–151. MATLAB 7.1. MathWorks Inc. (2005). McCarron, D. A., Holly, H. J., & Morris, C. D. (1982). Human nutrition and blood pressure regulation: An integrated approach. Hypertention, 4, 2–13. MINITAB 14. Minitab Inc. (2003). Nsanze, H., Babarinde, Z., & Al Kohaly, H. (1999). Microbiological quality of bottled drinking water in the UAE and the effect of storage at different temperatures. Environment International, 25, 53–57. Piper, A. M. (1944). A graphic procedure in the geochemical interpretation of water analyses. Transactions of the American Geophysical Union, 25, 914–923. Saleh, M. A., Ewane, E., Jones, J., & Wilson, B. L. (2001). Chemical evaluation of commercial bottled drinking water from Egypt. Journal of Food Composition and Analysis, 14, 127–152. Versari, A., Parpinello, J. P., & Galassi, S. (2002). Chemometrics survey of Italian bottled mineral waters by means of their labeled physico-chemical and chemical composition. Journal of Food Composition and Analysis, 15, 251–264. Warburton, D. W., Dodds, K. L., Burke, R., Johnston, M. A., & Laffey, P. J. (1992). A review of the microbiological quality of bottled water sold in Canada between 1981 and 1989. Canadian Journal of Microbiology, 38, 12–19. Ward, J. H. (1963). Hierarchical grouping to optimize an objective function. Journal of American Statistics Association, 69, 236–244. Whitney, E. N., Corinne, B. C., & Sharon, R. R. (1991). Understanding normal and clinical nutrition (3rd ed.). St. Paul: West Publishing.