International Journal of Cosmetic Science, 2006, 28, 2133

A method for the determination of N-nitrosodiethanolamine in personal care products collaboratively evaluated by the CTPA Nitrosamines Working Group
C. Flower*, S. Carter, A. Earls, R. Fowler, S. Hewlins, S. Lalljie, M. Lefebvre**, J. Mavro**, D. Small and N. Volpe
*Cosmetic Toiletry and Perfumery Association, London, U.K., Consumer Safety Department, LGC Limited, Teddington, Middlesex, U.K., Quality Department, Boots Manufacturing, Nottingham, U.K., Procter and Gamble Technical Centre, U.K. Limited, Egham, U.K.,Safety and Environmental Assurance Centre, Unilever Colworth, Sharnbrook, Bedford, U.K., **Advanced Research Analytical Chemistry Physical & Chemical Sciences, LOreal Recherche, Aulnay sous Bois, France, STIEFEL Laboratories, High Wycombe, Buckinghamshire, U.K. and Christian Dior, Technical R & D, Laboratoire Chimie Fine, St Jean de Braye Cedex, France

Received 3 June 2005, Accepted 23 September 2005

Keywords: cosmetics, Griess reaction, N-nitrosodiethanolamine, post-column photolysis, reverse-phase liquid chromatography, solid-phase extraction

Synopsis A procedure for the determination of N-nitrosodiethanolamine (NDELA) in personal care products was evaluated in collaborative studies by member organizations of the United Kingdoms Cosmetic Toiletry and Perfumery Association (CTPA) and LGC Limited, formerly known as the Laboratory of the Government Chemist (LGC). Samples were prepared depending on the matrix of the cosmetic product: aqueous samples were prepared by diluting in water followed by solid-phase extraction; emulsions, oils and solid materials were dissolved in dichloromethane and extracted with water. NDELA was separated from the sample matrix using reverse-phase liquid chromatography. The N-nitroso bond was cleaved by photolysis to give nitrite, which was colorimetrically quantied. The nitrite functional group reacted with sulphanilamide in an acid medium to form a diazonium ion
Correspondence: Chris Flower, Cosmetic Toiletry and Perfumery Association, Josaron House, 5-7 John Princes Street, London W1G 0JN, U.K. Tel.: +44 20 7491 8891; fax: +44 20 7493 8061; e-mail: info@ctpa.org.uk

which was then coupled with N-(1-naphthyl)ethylenediamine dihydrochloride according to the Griess reaction to give a purple-coloured azo dye that absorbed at 540 nm. Compared with other published methods for NDELA, the method described here is quick and easy to use. It has the required sensitivity and specicity, and can accurately and reliably quantify NDELA in a wide range of personal care product matrices. sume Re thode de dosage de la N-nitrosodiethanolUne me tiques a amine (NDELA) dans les produits cosme ` es en collabor te e value e, au cours de tudes mene e sentants du CTPA (United ation avec les repre Kingdoms Cosmetic Toiletry and Perfumery Association, et du LGC (Laboratory of the Govern nomme LGC Limment Chemist, maintenant de chantillons sont pre pare s en fonction de ited). Les e ` analyser. Les tique a la matrice du produit cosme pare s par produits dispersables dans leau sont pre dilution dans leau, suivie dune extraction sur phase solide. Les produits non dispersables dans

te Franc tologie 2006 Society of Cosmetic Scientists and the Socie aise de Cosme

21

Determination of NDELA in personal care products

C. Flower et al.

mulsions, huiles ou e chantillons solides sont leau ; e thane, la NDELA est endissous dans le dichlorome pare e des suite extraite par leau. La NDELA est se autres constituants de la matrice par chromatographie liquide en phase inverse. La liaison N-nitroso en e sortie de colonne chromatographique est coupe ` 254 nm pour donner du nitrite. Le par photolyse a colorime triquement. Le nitrite est ensuite quantie re agit avec la sulfanilamide en milieu nitrite forme acide pour donner un ion diazonium. Celui-ci est avec le N-(1-naphthyl)ethyleneensuite couple action de Griess diamine dichlorhydrate selon la re tecte pour donner un colorant azo que pourpre de ` 540 nm. Compare aux autres me thodes de dosa es pour la NDELA, la me thode de crite age publie ` mettre en oeuvre. Cette me thest rapide et facile a cique, elle est capable de ode est sensible et spe quantier la NDELA de fac on reproductible et able dans une large gamme de produits cosme tiques. Introduction N-nitrosamines such as N-nitrosodiethanolamine (NDELA) are potent animal carcinogens that induce cancer by a genotoxic mechanism. Indeed, it has been demonstrated that nitrosamines are carcinogenic in more animal species than any other category of chemical carcinogen [1]. Although there are no adequate data from epidemiology studies to conrm such effects in humans, it is prudent to regard nitrosamines as potential human carcinogens that act by a genotoxic mechanism. When assessing the risks from genotoxic carcinogens it is the widely adopted policy of regulatory authorities to make the assumption that such compounds have the potential to damage DNA at any level of exposure, and that it is impossible to identify a safe level. However, it is recognized that exposure to low levels of nitrosamines is unavoidable. Cosmetics represent a relatively minor route of exposure, the main routes being food and the environment. Nitrosamine contamination of cosmetics may result through the use of nitrosamine-contaminated cosmetic ingredients or through the nitrosation of precursors (principally secondary amines) present in nished cosmetic products by nitrosating agents such as nitrite or oxides of nitrogen. Under certain conditions, nitrosation of diethanolamine yields the nitrosamine, NDELA, a polar, non-volatile, N-nitrosamine compound.

Within the European Community, the Fifteenth Commission Directive relating to cosmetic products (92/86/EEC) [2] formally prohibits the marketing of cosmetic products that contain nitrosamines. An allowance is made for trace levels if they are technically unavoidable, so long as the product cannot cause damage to human health when applied under normal or reasonably foreseeable conditions of use. Essentially, this requires nitrosamine levels to be kept as low as reasonably practicable, although no specic level has been set for nished cosmetic products. However, this Directive also set a limit of 50 lg kg)1 (ppb) for the N-nitrosodialkanolamine content of fatty-acid dialkanolamides, monoalkanolamines and trialkanolamines used as raw materials in the manufacture of cosmetics. A similar limit (50 lg kg)1) has been set for the N-nitrosodialkylamine content of fatty-acid dialkylamides, monoalkylamines and trialkalyamines and their salts because the properties of these compounds are similar to their respective alkanolamine analogues with respect to their potential as precursors of nitrosamine formation (2003/83/EC) [3]. Cosmetics manufacturers are also required not to use the above-mentioned raw materials with nitrosating systems. Essentially, this means that ingredients that contain or release nitrite ions, should not be used in conjunction with these raw materials, unless effective inhibition systems are employed. In addition, the use of secondary dialkanolamines, e.g. diethanolamine, that have been shown to be a major source of nitrosamine contamination in cosmetics, has also been banned. Because of their perceived carcinogenic potential in several animal species, minimization of exposure to N-nitrosamines from cosmetic products is important for the protection of public health. It is therefore essential to have methods of analysis that can determine N-nitrosamines with the required sensitivity and specicity in cosmetic matrices. Screening methods for apparent total nitrosamines have been published [46] but they suffer from false-positive results. Specic analytical methods for each possible N-nitrosamines should not be so prone to false-positive results but performing a number of analyses on each sample would not be practicable, even if such methods had been developed. In practice, the most likely source of N-nitrosamines and their precursors for cosmetic products are the dialkanolamines used in the production of dialkanolamides, of which diethanolamide is the most widely used in cosmetic products. The corres-

22

2006 International Journal of Cosmetic Science, 28, 2133

Determination of NDELA in personal care products

C. Flower et al.

ponding N-nitrosamine, NDELA, is also generally regarded as one of the more potent carcinogens amongst the N-nitrosamine family of chemicals. For that reason, there is a requirement for analytical procedures that are capable of reliably detecting and quantifying NDELA in cosmetic products and their associated raw material. Analytical method Existing methodologies for the determination of NDELA To date, two generic approaches have been used for the determination of NDELA. The rst involves extraction of NDELA from the sample matrix, followed by derivatization. The extracted NDELA is derivatizated, the separation of the analyte from matrix components is achieved by gas chromatography then the derivatizated NDELA is detected by using a range of detectors [electron capture, electrolytic conductivity, mass spectrometry or thermal energy analyser (TEA)] [79]. The second approach involves the extraction of NDELA from the sample matrix, followed by separation by liquid chromatography (LC) and detection using a TEA [10]. Both of these approaches involve lengthy sample preparation steps and have potential for false positives, including in-situ formation of NDELA. A commonly used procedure for the determination of N-nitrosodialkanolamines in cosmetics is based on the method described by Sommer and Eisenbrand [7]. The analysis involves the extraction of NDELA and other N-nitrosodialkanolamines from the sample matrix by solid-phase extraction on kieselguhr and further clean-up on a silica gel column, followed by a derivatization of the N-nitrosodialkanolamine (silylation of the dried eluate from the column). After, the trimethylsilyl derivatives of N-nitrosodialkanolamines are determined by gas chromatography TEA. Like the other methods, a major drawback with this procedure is its complexity and the potential it presents for in-situ formation of N-nitrosodialkanolamine. Analysis of nitrosamines in cosmetic products containing the corresponding precursors involves the risk of artefactual nitrosamine formation during the clean-up procedure. Particularly, the solid-phase extraction is a critical step because traces of NOx adsorbed on the kieselguhr column might cause

substantial nitrosamine formation. Losses during clean-up are possible and recoveries should be checked by the addition of an internal standard with properties similar to NDELA. Suitable compounds are N-nitrosodiisopropanolamine and N-nitroso(2-hydroxyethyl)(2-hydroxypropyl)amine. Screening for N-nitroso compounds in cosmetic products Several procedures for the determination of total N-nitroso compounds have been described [4, 5]. Most are based on the denitrosation of N-nitroso compounds and subsequent measurement of the liberated NO using a chemiluminescence detector. The analytical procedure currently employed was published in the International Journal of Cosmetic Science in 1995 [6]. This procedure was chosen because it is highly sensitive and responds to a wide range of N-nitroso compounds. Furthermore, it is the method recommended by the United Kingdoms Cosmetic Toiletry and Perfumery Association for the determination of total nitrosamines in personal care products. In this method, all N-nitrosamines, irrespective of their volatility, are determined as a group to yield a total nitrosamine content calculated with respect to an authentic nitrosamine calibration standard. The procedure is a validated screening technique that provides an estimate of the total concentration of all N-nitroso compounds present in a cosmetic sample. Nitrosamines are determined quantitatively, and the method is not prone to false-negative results. However, there is some potential for interference from non-N-nitroso compounds including, for example, nitrite salts and esters, nitrate salts, thionitrites, thionitrates, nitrolic acids, peroxides and iron oxides, as well as nitro-substituted hair dyes. Because the absence of potential interferences cannot easily be shown, the result obtained in the determination of total N-nitroso compounds is commonly referred to as the apparent total N-nitroso compounds content. Apart from N-nitrosamines, N-nitrosamides will also be determined, but most other nitroso and nitro species should give only a weak response. These include C-nitroso and C-nitro compounds, N-nitro compounds (nitramines) and nitrites, of which the latter are decomposed prior to introduction into the denitrosation agent. It is strongly recommended that conrmatory analysis of positive results is undertaken.

2006 International Journal of Cosmetic Science, 28, 2133

23

Determination of NDELA in personal care products

C. Flower et al.

Proposed method for the determination of N-nitrosodiethanolamine Shuker and Tannenbaun described a method that utilizes liquid chromatography with colorimetric detection of the nitrite generated by cleavage of the nitroso compounds through post-column photolysis detection for nonvolatile N-nitroso compounds in biological uids [11]. Light from a high intensity discharge lamp photolyses N-nitroso compounds in aqueous solution to give nitrite ion, which is determined colorimetrically with Griess reagent in a post-column reactor. Methods by Pignatelli et al. [12] and Bellec et al. [13] also rely on LC separation and subsequent photohydrolysis of NDELA to form nitrite. The liberated nitrite can then be reacted with sulphanilamide to give a diazo intermediate which can be coupled with N-(1-naphthyl)ethylenediamine dihydrochloride (NED) to form a purple azo dye that can be determined spectrophotometrically. As N-nitrosamines can easily undergo photolytically induced de-nitrosation, this three-step chemical reaction process of photolysis, diazotization and the coupling reaction (Fig. 1) has the potential to form the basis of an appropriate method for the analysis and detection of N-nitroso compounds such as NDELA in cosmetic products and raw materials. An alternative procedure based on the Shuker and Tannenbaun method and utilizing the three steps described above was developed by a member organization of the United Kingdoms Cosmetic Toiletry and Perfumery Association (CTPA) Nitrosamines Working Group for the quantitative determination of NDELA in cosmetic matrices and cosmetic raw materials. In this procedure, NDELA and other N-nitrosodialkanolamines are separated
Photolysis
C2H4OH O N N C2H4OH hv

from the cosmetic matrix by reverse-phase liquid chromatography. The N-nitroso bond is cleaved by UV photolysis with the formation of nitrite ion. According to the Griess reaction, the nitrite is diazotized with sulfanilamide in an acid medium and then coupled with NED to form a purple azo dye that is quantitatively determined spectrophotometrically at 540 nm. The presence of NDELA can be conrmed by repeating the analysis without photolysis (no nitrite ion is then produced because the N-nitroso bond is not cleaved). The absence of a chromatographic peak at the retention time of NDELA in the chromatogram conrms that the peak observed in the rst analysis is NDELA (Fig. 2a and b). In order to avoid any possible doubt in the identication of the peak at the retention time of NDELA, an irradiation at 254 nm followed by one at 366 nm gives a very specic determination for N-nitrosamines (constant ratio). This paper presents the ndings of two studies organized by the United Kingdoms CTPA Nitrosamines Working Group to evaluate collaboratively this alternative method for the determination of NDELA in personal care products. The method was assessed for accuracy (recovery of NDELA from spiked samples at different concentration levels) and precision (the inter-laboratory variability). Experimental Test samples In the rst collaborative trial, the test samples consisted of aliquots of a shampoo spiked at four different levels (0, 20, 50, 100 ng mL)1 [ppb]) and a sample of cocodiethanolamide (unspiked).

C 2H 4OH

H2O

254nm

HN C 2H 4OH

HNO 2

Diazotization Reaction
H+

H2NO2S

NH 2

2H+

NO 2-

H2NO2S

2H2 O

Coupling Reaction

H 2NO2S

H2NO2S

NHC 2H 4NH2

NHC2C4NH 2

Azo dye

Figure 1 Reaction of nitrite with Griess Reagent to form an azo dye.

24

2006 International Journal of Cosmetic Science, 28, 2133

Determination of NDELA in personal care products

C. Flower et al.

Response
103.0

Photolysis at 254 nm (a)

102.5

NDELA
101.5 101.0 100.5 100.0 0 103.0 102.5 102.0 101.5 101.0 100.5 100.0 0 5 10 15 20 Retention time 5 10 15 20 Retention time

(b)

Without photolysis

Figure 2 (a) Chromatogram of a nished product spiked with 60 ppb of NDELA in water and with photolysis. (b) Chromatogram of a nished product spiked with 60 ppb of NDELA in water and without photolysis.

Additionally, a solution containing only NDELA was prepared at three concentration levels [20, 50, 100 lg mL)1 (ppm)] and sent blindly to participants. In the second collaborative trial, six different types of personal care products were prepared and fortied with NDELA at a range of levels against which the method could be evaluated, viz. a shampoo (0, 20, 50, 100 lg kg)1), a make-up remover milk (0, 20, 50, 100 lg kg)1), a moisturizing lotion (0, 50 lg kg)1), a cucumber cleansing lotion (0, 50 lg kg)1), a wheat germ hair conditioner (0, 20, 50, 100 lg kg)1) and a hair dye (0, 50 lg kg)1). In addition, test samples of a commonly used cosmetic ingredient, cocodiethanolamide (unspiked), and standard NDELA solutions (20, 50, 100 lg mL)1) were prepared and tested. Each sample was well mixed to ensure homogeneity, divided into portions, placed in amber-coloured (to protect against light) sample bottles and add as in the rst collaborative trial, was sent blindly to participating laboratories for assay. After distribution, the samples were refrigerated by the participants prior to NDELA analysis.

Reagents and reactants High-performance liquid chromatography (HPLC)grade dichloromethane, methanol, orthophosphoric acid 85%, ammonium acetate, potassium chloride, sulfanilamide and N-(1-naphthyl)ethylenediamine dihydrochloride (components of the Griess reagent) were obtained from Fisher Scientic (Loughborough, U.K.) and Sigma-Aldrich (Gillingham, U.K.). N-nitrosodiethanolamine was supplied by SigmaAldrich. Sep-Pak 386 C18 cartridges were supplied by Waters Ltd (Elstree, U.K.). The C18 Spherisorb ODS II column was supplied by Waters Ltd and the C18 Aqua phase column was supplied by Phenomenex (Maccleseld, U.K.). Griess reagent N-(1-naphthyl) ethylenediamine dihydrochloride (0.25 g) was dissolved in deionized water and made up to 250 mL in a volumetric ask; 4.0 g sulfanilamide was dissolved in 250 mL of a 5% aqueous solution of 85% orthophosphoric acid. The reagents were mixed together in an amber

2006 International Journal of Cosmetic Science, 28, 2133

25

Determination of NDELA in personal care products

C. Flower et al.

glass bottle, stoppered and refrigerated when not in use. Maximum shelf life was 5 days. Preparation of N-nitrosodiethanolamine standard solutions A stock standard solution containing 1 mg mL)1 of NDELA was prepared. This solution was diluted serially with deionized water to prepare working standard solutions covering the concentration range of 020 ng mL)1 (ppb). Working standard solutions were freshly prepared. Sample preparation The clean-up of cosmetic products for the NDELA analysis depends on the dispersibility of the samples in water. If the cosmetic product is dispersible in water, solid-phase extraction must be applied to the sample. If the cosmetic product is not dispersible in water but in dichloromethane, then the dichloromethane clean-up must be applied. Therefore, the dispersibility of the cosmetic matrix in water or dichloromethane must be established before a clean-up procedure is chosen. Solid-phase extraction clean-up for samples that are dispersible in water (e.g. lotions, bath foams, shampoos) Exactly 2.0 g of sample was weighed and dispersed in 20 mL water and centrifuged for 10 min at

3600 g. The Sep-Pak 386 cartridge was conditioned with 2 mL methanol followed by 2 mL of water. The cartridge was not allowed to dry out; 5 mL of the sample solution was loaded on to the Sep-Pak 386 cartridge and the rst 2 mL of eluant was discarded. The following 3 mL was collected for chromatographic analysis. When necessary, the collected solution was ltered through an appropriate lter. Dichloromethane clean-up for samples that are not dispersible in water (e.g. oils, lipsticks) Exactly 0.4 g sample was weighed and dispersed in 4.0 mL dichloromethane; 4 mL water was added, shaken and then centrifuged at 19 400 g for 10 min. The upper aqueous layer was retained for chromatographic analysis. With both type of sample matrices, the nal extract was analysed for NDELA by HPLC with UV photolysis and Griess reaction under the conditions specied in Table I. Apparatus The instrumental apparatus for the analysis comprises of ve principal components: Chromatographic system Post-column photolysis reactor Chemical reactor Detector UVVis Data processing station.

LC pump (1) Column type Pre-column type Injection volume Mobile phase Mobile phase ow rate Griess Reagent ow rate Temperature of post-column reactor Detection wavelength Photolysis unit (2)

Pumped equipped with a pulse damper 150 cm 4.6 mm, packed with 5 lm Spherisorb ODS II 1 cm 4.6 mm packed with 5 lm Spherisorb ODS II 50 or 100 lL sample loop 0.02 m aqueous ammonium acetate (pH 6.8) 0.5 mL min)1 0.5 mL min)1 50C 540 nm Photochemical reactor comprising a high-pressure knitted Teon reactor coil, 5 or 6 m 0.3 mm i.d. placed around a low-pressure UV lamp emitting at 254 nm (Beam Boost and Knauer Photoreactors) Low-pressure pump without pulsation, a second LC pump with the same characteristics as the rst may be used Column oven (or water bath) containing a low dead-volume mixing Tee connected to a 1 mL knitted reactor coil, 5 m 0.3 mm i.d. UV/visible detector equipped with a tungsten lamp, sensitivity 0.001 or 0.0005 AUFS (Shimadzu SPD-10AV).

Table I Chromatographic conditions and equipment for the determination of NDELA in personal care products

Griess Reagent delivery (3) Post-column reactor (4)

Detector (5)

26

2006 International Journal of Cosmetic Science, 28, 2133

Determination of NDELA in personal care products

C. Flower et al.

Pump
(Griess Reagent)

Pump (HPLC)

Photolysis unit 254 nm Low volume mixing T

Heater 50 C Detector 540 nm

Figure 3 Schematic illustration of the instrumental conguration for the analysis of NDELA.

A schematic illustration of the instrumental conguration for the analysis of NDELA is shown in Fig. 3. The chromatographic conditions and equipment for the determination of NDELA in personal care products are summarized in Table I. Calibration and calculation The external standard calibration method was used to quantify the concentration of NDELA present in a test sample. A least-squares linear regression of peak areas versus concentration of the NDELA reference standard (express in ppb) was constructed. The linear relationship of the increasing concentrations of the NDELA reference standard (expressed in ppb) versus their corresponding peak areas is expressed as y ax + b, where the slope of the regression line is represented by a and b is the intercept. From the chromatogram obtained with the formulation under investigation SF is the peak area of NDELA. The concentration, C (lg kg)1) (ppb) of NDELA in the formulation under investigation is calculated from the following equation: Clg kg1 SF b V a m

ility across the laboratories. The mean, the standard deviation and the coefcient of variation (CV) were calculated. The CV is a statistical measure of the deviation of the values from the mean. The smaller the CV value, the less is variation from the mean. Grubbs test was used to identify whether there were any outliers across the laboratories for each sample and spike combination. Grubbs test identies observations that may be considered outliers because they are signicantly higher than the others in the data set. Where outliers were identied, these results were excluded from the statistical analysis. In the cases where the level of NDELA found was quantied as less than a threshold value, i.e. limit of quantication, nd is recorded in the data tables. These results were excluded from the statistical analysis. Remarks concerning the method Before the collaborative trial, some laboratories investigated various method parameters. The following are some of the key elements of the method that were assessed for effect on performance. In the ring-trialled method we used a 0.02 m aqueous acetate buffer at pH 6.8 as mobile phase, a Spherisorb C18 LC column and an NED concentration of 1 g L)1. Detection of the azo complex is made at 540 nm. The concentration of NED was not found to be critical. The use of sulfamic acid (nitrite scavenger) was investigated in samples which produced nitritegenerating species. After treatment with acid, the pHs of the solutions were re-adjusted to that of the mobile phase with an ammonia solution before HPLC analysis.

where SF is the peak area of NDELA in the formulation under investigation, b the y-intercept of the calibration curve, a the slope of the calibration curve, m the weight of the formulation under investigation (g), V is the volume of the tested solution (mL). Statistical treatment of data Summary statistics were calculated for each sample and spike combination in order to assess the variab-

2006 International Journal of Cosmetic Science, 28, 2133

27

Determination of NDELA in personal care products

C. Flower et al.

Response

100.15 100.10 100.05 100.00

0 2 4 6 8 10 12 Retention time

(a)

NDELA

Response

(b)
100.15 100.10 100.05 100.00
0 2 4 6

NDELA

10

12 Retention time

Figure 4 (a) Chromatogram of a 2.34 ppb NDELA standard in water. (b) Chromatogram of 11.71 ppb NDELA standard in water.

A limit of detection of 10 ng mL)1 (10 ppb) was achieved initially by optimizing parameters such as ow rate and composition of mobile phase, residence time of NDELA in photolysis unit and postcolumn reactor, and ow rate of NED reagent. It became obvious that the tungsten lamp source in the UVVis detector gave better performance, stability and sensitivity at 540 nm than a deuterium lamp. A limit of detection of 1 ng mL)1 (1 ppb) for a standard NDELA solution was achieved with the tungsten lamp. Figure 4(a) and (b) shows chromatograms of 2.34 and 11.71 ppb of NDELA standard in water. An improvement in the separation of NDELA from nitrite was obtained using either a longer Spherisorb column or a column with an Aqua

C18 stationary phase. Compared with the conditions already mentioned above, both of these columns increased the retention and separation of NDELA from nitrite (Fig. 5ac). Under aqueous conditions, the octadecyl sidechains of conventional C18 columns can collapse. This can result in loss of retention and poor column performance. The C18 Aqua phase, however, is end capped with a hydrophilic (polar) reagent, which allows water in the eluent to wet the silica and prevents the C18 side-chains from collapsing. The Aqua packed column was also found to show greater stability over a longer time period than a conventional C18 column. Prior to analysis, the LC column must be allowed to equilibrate.

Figure 5 (a) Spherisorb 5 lm ODS2 (150 4.60) mm with 5 lm ODS2 (10 4.60) mm guard column, (b) aqua 5 lmC18 (150 4.60) mm with Aqua 5 lmC18 (30 4.60) mm guard column, (c) Spherisorb 5 lmODS2 (250 4.60) mm.

28

2006 International Journal of Cosmetic Science, 28, 2133

Determination of NDELA in personal care products

C. Flower et al.

Table II Repeatability of the chromatographic system, extraction methods and method recovery

Repeatability of the chromatographic system (standard solutions)* NDELA 20 ng mL)1 8 ng mL)1 4 ng mL)1 2 ng mL)1

Mean (ng mL)1) CV (%)

20.3 7.89 3.85 1.91

3.7 3.7 5.8 10.7

Repeatability of extraction methods shampoo Sep-Pak cartridges, NDELA (100 lg kg)1) Dichloromethane extraction, NDELA (20 lg kg)1)

Mean (lg kg)1) CV (%) 91.1 9.2 13.9 15.1

Method recovery Recovery (%) 82.4 Sep-Pak cartridges, NDELA recovery (10 ng mL)1) Dichloromethane extraction, NDELA recovery (10 ng mL)1) 96.1

CV (%) 11.9 1.9

*Repeatability was determined at four different concentrations by analysing six replicate injections at each concentration. The repeatability (n 6) of the extraction method was also determined with shampoos of known concentration of NDELA (100 and 20 lg kg)1). The recovery of both extraction methods was determined at 10 ng mL)1.

It should be noted that through out the analytical procedures, all solutions must be protected from light. Only amber-coloured vials or aluminium foil-coated volumetric asks should be used. In the discussion that follows, the two collaborative studies are referred to as Study 1 and Study 2. Results and discussion Validation of analytical method As part of their in-house validation, one of the participating laboratories investigated the repeatability and recovery of the method. The repeatability for this method describes the precision expected from a set of replicate measurements made by a single laboratory with the same analyst and the same equipment. The standard deviation is used to calculate a repeatability value r. Repeatability Repeatability (Table II) was determined for both the Sep-Pak and dichloromethane extraction procedures by analysing shampoo samples containing known amounts of NDELA. A CV of 9.2% was obtained using Sep-Pak cartridges and 15.1% after dichloromethane extraction. Repeatability of the chromatographic system and post-column: photolysis and Griess reaction technique was calculated with measurements at 2 ng mL)1 (CV

10.7%), 4 ng mL)1 (CV 5.8%), 8 ng mL)1 (CV 3.7%) and 20 ng mL)1 (CV 3.7%). Recovery Recoveries at 10 ng mL)1 were also determined for both the Sep-Pak and dichloromethane extraction procedures (Table II). Good agreement was obtained for both procedures. The small differences between the recoveries, 8092% for Sep-Pak Sep-cartridges and 9398% for dichloromethane extractions, can be attributed to variations in the fractions collected for analysis from the C18 Sep-Pak cartridge. Linearity In order to quantify the concentrations present, a least-squares linear regression of peak area against
1200
Peak area (mV)

1000 800 600 400 200 0 0 5 10 15 20 25

NDELA concentration (ppb)

Figure 6 Calibration graph showing good linearity of response of NDELA over the 120 ng mL)1 (ppb) range.

2006 International Journal of Cosmetic Science, 28, 2133

29

Determination of NDELA in personal care products

C. Flower et al.

concentration was constructed using the external standard calibration method. Good linearity of response was achieved over the 120 ng mL)1 (ppb) range with coefcient of correlation (r2) >0.999 (Fig. 6). Limit of detection and limit of quantication The limit of detection for a standard solution of NDELA using the method is 1 ng mL)1 (1 ppb). Results of a validation exercise indicates a spread in the limit of quantication in a range of products, including lipstick, sunscreens and hair shampoos, between 10 and 40 lg kg)1 (1040 ppb). Evaluation of personal care product matrices Additional investigations were undertaken on different sample matrices including sunscreens, shampoos, lipsticks and hair conditioners; some of them were more difcult to analyse than others. Recoveries from these products were determined after spiking samples with NDELA with the following results: sunscreens (73%), hair shampoos (83%) and lipsticks (80%). Emulsion formation required centrifugation of the sample solution before analysis by liquid chromatography. Sulphamic acid was added to remove nitrite. The sample preparation of hand and face lotions and bubble baths was performed by aqueous dissolution. Lipsticks and lip-gloss products were extracted with dichloromethane by water. An additional clean-up procedure was applied to sunscreen products where persistent emulsions formed. Addition of methanol and potassium chloride to the water/ dichloromethane solution followed by vigorous agitation for 30 s removed the emulsion. For per-

Table IV Inter-laboratory variability (precision) for the samples tested in the rst collaborative trial
NDELA spike level, lg kg)1 (ppb) Mean

Sample

SD

CV %

Shampoo

20 50 100 Cocodiethanolamide 0 NDELA solution 20 000 50 000 100 000

27.2 14.5 49.5 11.7 105.2 18.6 21.0 6.1 19 300 1800 42 600 4700 90 100 8400

53 24 18 29 9 11 9

sistent emulsions, solutions were centrifuged at 19 400 g for 10 min. Less interference was encountered with oil-based samples, presumably because much of the organic content is removed with dichloromethane. One sample, a shampoo, was found to contain a measurable amount of NDELA (>1000 lg kg)1). This result was conrmed by analysing without photohydrolysis (cleavage of the N-nitroso bond to give nitrite). No response was observed at the retention time of NDELA in the chromatogram. Collaborative trials In Study 1, to explore the potential of the method, six participating laboratories analysed a shampoo and the same shampoo spiked at three levels [20, 50, 100 lg kg)1 (ppb)]. Also analysed were a cosmetic raw material, cocodiethanolamide and standard solutions of NDELA at three different concentration levels [20, 50, 100 lg mL)1 (ppm)]. Table III shows the results of Study 1. Considering the low level of measurement, acceptable

Sample description

NDELA spike level, lg kg)1 (ppb) Lab 1

NDELA measured, lg kg)1 (ppb)

Lab 2

Lab 3

Lab 4

Lab 5

Lab 6

Shampoo

0 20 50 100 Cocodiethanolamide 0 NDELA solution 20 000 50 000 100 000

<25 <10 49 20 47 54 141 103 <20 17 16 900 18 900 18 39 300 44 300 50 82 500 94 800 92

38 <10 nd nd 51 67 94 102 <40 nd 700 21 600 20 300 39 500 39 700 80 500 100

nd <10 20 20 31 47 88 103 28 18 600 18 400 600 43 600 400 94 800

Table III Results of the rst collaborative study (Study 1) for the determination of NDELA in shampoos spiked at four levels, a cosmetic raw material and standard solutions of NDELA at three different concentration levels

30

2006 International Journal of Cosmetic Science, 28, 2133

Determination of NDELA in personal care products

C. Flower et al.

Table V Results of the second collaborative study (Study 2) for the determination of NDELA in shampoos, make-up remover, cosmetic raw material, moisture lotion, cucumber cleansing lotion, wheatgerm hair conditioner, hair dye, and standard solutions of NDELA at three different concentration levels

Sample description

NDELA measured, lg kg)1 (ppb) NDELA spike level, lg kg)1 (ppb) Lab 1 Lab 2 Lab 3 Lab 4 Lab 5

Shampoo

Make-up remover milk Cocodiethanolamide Moisture lotion

0 20 50 100 0 50

0 50 Cucumber cleansing lotion 0 50 Wheatgerm hair conditioner 0 20 50 100 Hair dye 0 50 NDELA solution 20 150 50 370 100 750

<10 26 51 93 <10 44 66 <10 73 <10 72 <10 30 63 114 130 163 16 300 20 45 000 50 68 500 100

nd 20 50 101 nd 52 52 nd 55 nd 56 nd 27 61 135 nd 90 100 18 600 62 600 132

38 <10 100 15 298 58 85 104 nd <20 123 <20 nd <50 0 <10 99 70 0 <10 109 61 0 <10 83 21 71 47 189 104 nd nd nd nd 000 18 800 20 000 47 700 53 000 93 400 110

nd 23 56 115 nd 46 50 nd 66 nd 74 nd 34 79 143 000 000 000

agreement was obtained between the laboratories for both the spiked shampoo and the standard solutions. CV at all spike levels is shown in Table IV. The highest CV (53%) was observed with the shampoo spiked at the 20 ppb level. However, this 20 ppb spike was below the limit of quantication for two of the six laboratories. The CVs for the NDELA solution were about 10% for all three spiked levels, in line with the reproducibility standard deviation predicted by the Horwitz [14] function at these levels. In general, these results indicated that the method has the potential to recover NDELA at very low concentrations from cosmetic matrices and the CV also indicates the robustness of the method. The raw material proved more difcult. However, it should be noted that fatty-acid dialkanolamide as a sample matrix is difcult to analyse by the other approaches described earlier. As a result of these promising initial ndings, it was decided to carry out a second collaborative study with a wider range of matrices. In the second study, ve laboratories analysed a total of six personal care products (shampoo, make-up remover milk, moisture lotion, cucumber cleansing lotion, wheatgerm hair conditioner, hair dye) that were spiked with NDELA at 20, 50 and 100 lg kg)1 (ppb). A raw material (cocodiethanol-

amide) was also analysed in duplicate for NDELA. In addition, three separate standard solutions containing 20 000, 50 000 and 100 000 lg kg)1 (ppb) of NDELA were also analysed. Each sample was analysed in duplicate. Grubbs test was used to identify whether there were any outliers across the laboratories for each sample and spiked combination. In view of the fact that the data set was very small and the outliers identied by the Grubbs test were all from the same laboratory, the ranking test was used to conrm the exclusion of all results from this laboratory. Results of Study 2 are shown in Table V. The data indicate that the method satisfactorily determined NDELA from ve of the six spiked formulations. Excellent results were obtained by four of the ve laboratories1 for the ve spiked formulations. Acceptable results were obtained for the three standard solutions (at relatively much higher concentrations). Recoveries of NDELA obtained by the ve laboratories at the different spike levels are shown in Table VI and the CV for all matrices at the different spike levels and the standard solutions are shown in Table VII. The results of laboratory 3 were excluded from the statistical analysis on the basis of the ranking test.
1

2006 International Journal of Cosmetic Science, 28, 2133

31

Determination of NDELA in personal care products

C. Flower et al.

Sample description

NDELA spike level, lg kg)1 (ppb)

% NDELA recovered

Lab 1

Lab 2

Lab 3

Lab 4

Lab 5

Table VI Results of the recoveries of NDELA obtained by the ve laboratories during the second collaborative study (Study 2)

Shampoo

Make-up remover milk Moisture lotion Cucumber cleansing lotion Wheatgerm hair conditioner

NDELA solution

20 50 100 50 50 50 20 50 100 20 150 50 440 100 700

130 102 93 88 146 144 150 126 114 81 89 68

100 100 101 104 110 112 135 122 135 100 100 105

(500) (596) 85 (246) 198 218 (415) 142 189 89 123 131

75 116 104 nd 140 122 105 94 104 93 95 93

115 112 115 92 132 148 170 158 143 99 105 99

Results in parenthesis are highly suspect.

Table VII Inter-laboratory variability (precision) for the samples tested in the second collaborative trial (Study 2)
NDELA spike Mean, level, lg kg)1 lg kg)1 (ppb) (ppb)

Sample

SD

CV (%)

Shampoo

Make-up remover milk Cocodiethanolamide Moisture lotion Cucumber cleansing lotion Wheatgerm hair conditioner NDELA solution

20 50 100 50 ) 50 50 20 50 100 20 150 50 440 100 700

21.0 53.8 99.6 47.3 56.0 66.0 59.7

4.7 3.9 9.1 4.2 8.7 7.9 9.3

22 7 9 9 16 12 16 20 14 14 10 16 18

24.5 4.8 55.2 7.9 113.2 15.6 18 800 1390 49 100 2800 93 100 1680

The hair dye sample was excluded because of the difculty the laboratories had in analysing this matrix.

Again, the data conrms the good accuracy, precision (agreement between laboratories) and robustness of the method observed in Study 1. However, the hair dye sample proved to be a difcult matrix to analyse because of matrix interference. Each of the laboratories had difculty with this matrix. The results from the collaborative trial are not in agreement with the expected amount of 50 lg kg)1. Sample clean-up is the most critical step in the analytical procedure for NDELA. This study shows

that not all types of personal care products can be analysed using the same sample preparation, chiey because of varying matrices: for example, lotions, shampoo, conditioner and hair dye are water-soluble and were therefore mixed with water before extracting through a Sep-Pak 386 cartridge. The presence of NDELA in cocodiethanolamide was determined by dissolving the material in dichloromethane followed by extraction into water and centrifuging at 19 400 g. The results of these collaborative trials indicate that NDELA was satisfactorily recovered, identied and quantied by all but one of the participating laboratories. It should be noted that prior to the rst ring trial most of the participating laboratories were not familiar with the method and had to purchase the instrumentation and set up the method in the laboratory in order to participate in the collaborative trial. Yet, in general, there was broad agreement in the results obtained by the laboratories. This clearly demonstrated the relative simplicity and potential of the method. The different analytical equipment used by the participating laboratories, lack of experience with the method and detector sensitivity can explain the differences in the results obtained. Conclusions The good inter-laboratory agreement between participating laboratories conrmed that the method is sensitive, robust and gives good accuracy and

32

2006 International Journal of Cosmetic Science, 28, 2133

Determination of NDELA in personal care products

C. Flower et al.

precision. It has been shown to be applicable to a range of different cosmetic formulation matrices and has the ability to conrm the presence or absence of NDELA without the sophistication of mass spectrometry, or the technical expertise required for mass spectrometric detection. It is therefore acceptable for determining NDELA in a broad range of personal care products. The method described employs a relatively rapid sample preparation providing faster analysis than is possible with existing methodologies, and can be adapted for routine analysis of large numbers of samples. Further work is required to extend the method to complex formulations such as hair dyes and lipstick. Acknowledgements The authors wish to express their sincere appreciation to the analysts from the participating Laboratories: Unilever (U.K.), Boots Contract al (France), Christian Manufacturing (U.K.), LOre Dior (France), and LGC Limited (U.K). The authors also wish to express their appreciation to LGC Limited for providing the data on their in-house validation of the method. References
1. International Agency for Research on Cancer. Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Human Beings, Vol. 17. IARC, Lyon (1978). 2. European Commission Directive (92/86/EEC). Ofcial J. Eur. Communities, L325. 1822 (1992). 3. European Commission Directive (2003/83/EC). Ofcial J. Eur. Communities, L 238. 2327 (2003). 4. Waters, C.L., Downes, M.J., Edwards, M.W. and Smith, P.L.R. Determination of non-volatile nitrosamines on a food matrix. Analyst 103, 11271133 (1978). 5. Chou, H.J., Yates, R.L. and Wenninger, J.A. Screening cosmetics products for N-nitroso compounds by

6.

7.

8.

9.

10. 11.

12.

13.

14.

chemiluminescence determination of nitric oxide. JOAC 70, 960963 (1987). Challis, B.C., Colling, J., Cromie, D.D.O. et al. A screening procedure for total N-nitroso contaminants in personal care products: results of collaborative studies undertaken by a CTPA Working Group. Int. J. Cosmet. Sci. 17, 219231 (1995). Sommer, H. and Eisenbrand, G. A method for the determination of N-nitrosoalkanolamines in cosmetics. Z. Lebensm. Unters. Forsch. 186, 235238 (1988). Rollman, B., Lombart, P. and Mercier, J. Determination of N-nitrosodiethanolamine in cosmetics by gas chromatography with electron capture detection. J. Chromatogr. 206, 158163 (1981). Collier, S.W., Milstein, S.R. and Orth, D. Quantitative assay of volatile and non-volatile N-nitrosamines by gas chromatography with an electrolytic conductivity detector. 1. Method development and assay of N-nitrodiethanolamine (NDELA) in creams and lotions. J. Soc. Cosmet. Chem. 39, 329346 (1988). Ikeda, K. and Migiorese, K. Analysis of nitrosamines in cosmetics. J. Soc. Cosmet. Chem. 41, 308 (1990). Shuker, D.E.G. and Tannenbaum, S.R. Determination of nonvolatile N-nitroso compounds in biological uids by liquid chromatography with postcolumn photohydrolysis detection. Anal. Chem. 55, 21522155 (1983). Pignatelli, B., Malaveille, C., Thuillier, P., Hautefeuille, A. and Bartsch, H. Improved methods for analysis of N-nitroso compounds and application in human biomonitoring in Nitrosamine and related Compounds, Chemistry and Biochemistry, eds R.N. Loeppley & C.J. Micheda. ACS Symp. Ser. 553, ACS, Washington DC, 102118 (1992). Bellec, G., Cauvin, J.M., Salaun, M.C. et al. Analysis of N-nitrosamines by high-performance liquid chromatography with post-column photohydrolysis and colorimetric detection. J. Chromatogr. A 727, 8392 (1996). Horwitz W., Kamps, L.R. and Boyer K.W. Quality assurance in the analysis of foods for trace constituents. JAOAC 63, 13441453 (1980).

2006 International Journal of Cosmetic Science, 28, 2133

33