Application of Handheld X-Ray Fluorescence Analyzers to Identify Mercury in Skin-Whitening Creams in Cambodia

Subscription required to view. DOI:10.1053/ ajkd.2003.50017 21. Glenn EN. Yearning for lightness: transnational circuits in the marketing and consumption of skin lighteners. Gender Soc [Internet]. 2008 Jun [cited 2012 Jun13]; 22(3):281-302. Available from: http://gas. sagepub.com/content/22/3/281.short. Subscription required to view. 22. Olumide YM, Akinkugbe AO, Altraide D, Mohammed T, Ahamefule N, Ayanlowo S, Onyekonwu C, Essen N. Complications of chronic use of skin lightening cosmetics. Int J Dermatol [Internet]. 2008 Mar 28 [cited 2012 Jun 13]; 47(4):344-53. Available from: http://onlinelibrary. wiley.com/doi/10.1111/j.1365-4632.2008.02719.x/full Subscription required to view. DOI: 10.1111/j.13654632.2008.02719. 23. Murphy T, Slotton D, Irvine K, Sukontason K, Goldman C. Mercury contamination of skin whiteners in Cambodia. Hum Ecol Risk Assess [Internet]. 2009 Nov 6 {cited 2012 Jun 13];15(6):1286303. Available from: http://www.tandfonline.com/ doi/abs/10.1080/10807030903306877 Subscription required to view. DOI: 10.1080/10807030903306877 24. Circular control groups, community groups, the media, compilers. Association of Southeast Asian Nations Cosmetic Documents [Internet]. Mueang Nonthaburi District, Thailand. Food and Drug Administration, Ministry of Public Health; 2008 Jan 1. Appendix VI: ASEAN guidelines for cosmetic good manufacturing practice; [cited 2012 Jun 13]; 24-41. Available from: http://elib.fda.moph.go.th/fulltext2/ word/14697/1.pdf 25. Ingredients prohibited & restricted by FDA regulations [Internet]. Silver Spring, Maryland: U.S. Food and Drug Administration; 1996 Jun 22 [updated 2000 May 30; cited 2012 Jun 13]. Available from: http://www.fda. gov/Cosmetics/ProductandIngredientSafety/ SelectedCosmeticIngredients/ucm127406.htm 26. The European Parliament, The Council of the European Union. Regulation (EC) no 1102/2008 of the European Parliament and of the council of 22 october 2008 on the banning of exports of metallic mercury and certain mercury compounds and mixtures and the safe storage of metallic mercury. Off J Eur Union [Internet]. 2008 Nov 14 [cited 2012 Jun 14]; 304(75). Available from: http://eur-lex.europa.eu/ LexUriServ/LexUriServ.do?uri=OJ:L:2008:304:0075: 0079:EN:PDF 27. Global mercury assessment: [Internet]. Geneva, Swizerland: United Environment Programme Chemicals; 2002 Dec. Chapter 9: 764: cosmetics; [cited 2012 Jun 15]; 179. Available from: http://www. chem.unep.ch/mercury/report/Final%20report/final-


Introduction
Mercury is a toxic metal that, even in low concentrations, can impair fertility, suppress the immune system, damage kidneys, cause nerve damage, and impair learning. 1,2Frequently, the potential for health impacts due to mercury contamination is assessed using human hair or blood. Yokoo et al. reported significant negative impact on neuropsychological tests from adults in fishing villages in Brazil when mercury levels in hair exceeded 6 μg/g. 5Agusa et al. found 10 out of 60 hair samples from Phnom Penh had more than 10 μg/g of mercury and three of these samples had extreme levels of mercury (190, 70, and 60 μg/g). 6Furthermore, Agusa et al. reported that some female subjects in Phnom Penh had mercury levels in their hair 7 that exceeded World Health Organization (WHO) threshold levels of 50 μg/g for neurotoxic effects. 8The same authors found serum estrone and estradiol levels were positively correlated with blood mercury level for both males and females, indicating possible induction of female hormones by mercury exposure in Cambodians.These observations are consistent with studies indicating mercury is an endocrine disruptor in mercury exposure experiments with birds, 9 fish, 10 and human cell cultures. 11Agusa et al. ruled out the Sihanoukville mercury dumpsite storage as the source of mercury for the hair and blood samples collected in Phnom Penh. 12Although it is a documented and large environmental hazard, Sihanoukville is far enough from Phnom Penh that it was not considered the source of the city's mercury problem.Levels of mercury in fish in Phnom Penh markets were similar to those found elsewhere, so fish were also ruled out as the source of mercury.Unfortunately, the source of the mercury contamination has not been definitively resolved.
One well-documented source of mercury contamination is skinwhitening creams, into which the heavy metal is sometimes added to block production of melanin and thus produce whiter skin. 13While these mercury-containing skin creams are a health concern internationally, 14-21 Murphy et   The primary objective of this study was to evaluate the utility of a handheld XRF analyzer for detection of mercury in skin-whitening creams in developing countries with few resources for stringent monitoring and testing.

Sampling
Sampling protocols were approved by the Ethical Review Committee of the Cambodian Ministry of Health.
In accordance with the committee's recommendations, the researchers recruited volunteers from a variety of socio-economic backgrounds to provide samples of skin-whitening creams.The researchers also purchased, in early 2011, 66 skinwhitening creams from the Central, Laak3, Olympic, and Orussey markets in Phnom Penh and a local, peri-urban market to test product variability.
For the purposes of this study, 610 skin-whitening creams were donated by: i) 279 female and 4 male students The study took place in UHS, both Don Bosco Institutes, two coffee shops near the factories, a bar and a massage parlor.Each participant completed, on-site, a questionnaire that provided socio-economic and skin-cream usage information (most of these data points are not assessed in this paper).In each case, the participants were able to witness their sample being processed, an important step for several reasons.First, it provided a level of transparency, the lack of which is a problem that continues to face the country as it struggles to recover from the tragic Khmer Rouge period.Second, the demonstration facilitated the educational portion of the study.The study was preceded by a seminar during which the provider of each sample was questioned on her awareness of the dangers of mercury toxicity, provided information on the subject, allowed to ask questions and return later for more discussion.
The origins of some of the samples provided by participants were not as obvious as of those purchased directly from markets, as study participants seldom brought in original packaging.However, participants were able to provide insight into their products.For example, students from UHS had several samples purchased while traveling abroad, particularly from Thailand.

Analysis
The Niton XL3t handheld XRF unit (Thermo Fisher Scientific, Waltham, MA) used in this study allowed us to take readings of skin creams in 30 seconds with no extraction and minimal sample preparation.For skin creams, the XRF analyzer was used in plastics mode with the algorithm and calibration done at the Thermo Fisher Scientific office in Billerica, Massachusetts, USA.The XRF analyzer had a built-in algorithm to calculate the standard deviation of individual analysis.Other statistical calculations were done with Microsoft Excel software.Because of the presence of a variety of volatiles other than water, the research team followed the same protocols as Murphy et al. 23 and processed wet creams.Samples 3 to 5 ml in volume were taken by spatula from the top 5mm layer of the cream and placed on a piece of Mylar film (6 µm in thickness) that was then folded.Next, the detector was placed directly against the film.If the cream jar was full, the Mylar film was placed over the mouth of the container, directly on the cream, and the detector placed against the film.It was crucial to minimize space between the sample and the detector to avoid reduction of signal.The general directions provided by the XRF manufacturer stated that the instrument detection limit was 20 μg/g. 34,35This is in agreement with our sample processing in that we could not detect mercury amounts below 20 μg/g in 579 skin cream samples.Increasing the analysis time would reduce the detection limits by the square root of the increased time. 34,35However, increasing the analysis time to approach guideline values of mercury was not feasible for these educational clinics as researchers chose to process samples in front of participants, requiring the analysis to be completed within a tolerable time frame.Also, the skin cream products were returned afterwards to study participants, especially if they were mercury free.
Samples were not stirred because the researchers wanted to use a process that could be effective in the field when used by a regulatory agent.The need for processing speed was also a factor in the decision.After verification (Table 1), the researchers accepted that the products were relatively homogeneous, requiring no mixing before analysis.Since the relative standard deviation of triplicate analysis of the same samples averaged less than 5%, the sampling error was modest (Table 1).To stir the creams well required placing the samples into a container about one-third full so that a vortex mixer can mix the cream, leading to concerns over possible contamination and slow-down of results.

Quality Assurance/Quality Control
Prior to the study, Essco Safety Check of Seattle, WA, used a handheld XRF to process six samples collected from Cambodia. 36The same creams were also processed by Thermo Scientific in Billerica, MA.Another set of 12 skin creams purchased in Phnom Penh markets was processed at the National University of Singapore, using inductively coupled plasma optical emission spectrometry (ICP-OES).
The results of the Singapore analysis was then compared to the XRF study done by Thermo Scientific.
Eight samples containing less than 1000 μg/g of mercury in duplicate were spiked with additional mercury to a

Research
final load of 2000 μg/g of mercury.
The mercury content of four samples in duplicate that originally contained more than 1000 μg/g of mercury was raised to 4000 μg/g.Before ICP-OES analysis, 1 mg/L of trace metals grade acid (Fluka, Sigma Aldrich, Singapore) was added to all samples, calibration standards, and blanks for amalgamation of mercury. 37For ICP-OES analysis, samples of 50 mg were extracted using the protocols outlined in Murphy et al. 23 Cold vapour atomic absorption spectrometry (CVAAS) analysis was done in 2008 at the University of California, Davis to provide certified analyses as a framework for using kits to measure mercury in skin-whitening creams.

Research
for better replication of sampling.In part because the reagent costs were minimal and analysis time short, with XRF analysis we were able to conduct a variety of preliminary experiments to optimize analytical protocols with the DMSO spikes.Certified reference materials supplied by Thermo Fisher Scientific were used for daily pre-and post-analysis to confirm that the analyzer was within 5% of manufacturer's specifications.

Analysis Verification
The replication between laboratories with XRF analysis was good (Table 2).Samples with high mercury concentrations caused serious equipment contamination and closure of the California laboratory for more than a month.This result is unusual but can happen when the client and chemist are unaware of an extreme concentration.It is important that there is consistency with respect to laboratories agreeing that a cream has insignificant or high concentrations of mercury.For the analysis of skinwhitening creams, the XRF sensor is more than a qualitative screening tool and has the utility of more traditional laboratory equipment.For concentrations of mercury <4500 μg/g, both XRF and ICP-OES processes produced similar results, with an r value of 0.97 and slope of 0.954 in a linear regression analysis (Figure 1).About 80% of the skin-whitening creams in this study had <4500 μg/g mercury.The quality control of ICP-OES showed good recovery in spiking trials.For samples spiked at 2000 μg/g, recoveries were 86±29% (n=8), whereas for samples spiked at 4000 μg/g, recoveries were 114±18% (n=4).
There was too little data and too much variation to justify plotting the ICP-OES analysis as biphasic but the data could plot into two regions.
The spiking experiments with DMSO produced a relationship with two regions that are more distinct than in the ICP-OES analysis.Using DMSO, we found that there was up to a 22% suppression of the signal in XRF response with the highest levels of mercury.It is possible to put a line through all data and get a strong r value of 0.999 but recovery of mercury is less in the higher spikes with a lower slope (0.785) for the full data set (Figure 2a) than with the slope of

Research
Mercury in Skin-Whitening Creams samples containing less than 2000 μg/g (0.903, Figure 2b).The mean recovery of the lowest five mercury spikes was 97.1% with a standard deviation of 7.3% (n=10).The mean recovery of the five highest spikes was 82.2% with a standard deviation of 4.2% (n=10).
The difference in recovery between the lowest five and highest five set of spikes was significant with 99% confidence limits.The replication was such that in figures 2a and 2b, the duplicates often show as one sample.
The replication in the samples spiked with DMSO was by far the best in this study, with an average relative standard deviation of 2.5%.This is about half the relative standard deviation of all other analyses in this project.The accuracy and precision of the XRF analysis with mercury concentrations below 2000 μg/g were good whereas above 2000 μg/g of mercury, the results were reading about 18% too low.
In general, the replication of samples with mercury was good.Table 3 shows the variability in mercury levels in various products collected at different markets that were analyzed using the Thermo XRF.Sample replication of some commercial products with high mercury content, such as the ones apparently from the U.S. and one from China (China4), was good but sample replication of four others apparently from China and the sample apparently from Taiwan was poor.Throughout this paper researchers use the term "apparently" because in this region bootlegging of all commercial products is common and the real source of creams is often uncertain.The variability in the mercury content of some products may reflect different batches possibly containing other whitening agents such as retinoic acid or hydroquinone.The ten-fold variation in mercury content of five different products from China (Table 3) likely reflects inconsistent chemistry or varying formulations.Table 1 illustrates that reproducibility of sampling from the same cream was good.The relative standard deviation of triplicate analysis of the same samples with the Thermo XRF was less than 11% and averaged less than 5%.

Sample Analysis in Education Clinics
Six hundred and ten creams brought to us by 279 young women and 4 men were analyzed and 15.9% contained more than 20 μg/g of mercury, which is approximately the detection level of mercury with the XRF.

Research
To impute the concentration of mercury in samples with levels below the detection limit for XRF analysis, we used the mean of 11 skin-whitening creams collected in Phnom Penh in 2008, processed by CVAAS and containing less than 20 μg/g of mercury.The low rate of matching between the sampling of the market and the educational clinics and the continued collection of new products in the most recent samplings indicates that there may be more than 400 different skin cream products in Cambodia.

Discussion
Sampling Sampling at educational institutes was more efficient, safer, and more conducive to follow-up sampling or consultation than sampling with factory or hospitality workers.Factory workers were only available before or after shifts and hospitality workers were only available when their establishments were not busy.One hospitality manager instructed his staff not to provide researchers with whitening creams containing mercury that were sold on-site from a producer in Phnom Penh.In this instance, the study participants waited for an opportunity to explain and to give the creams to the researchers privately.There also was some concern about potential theft of the XRF unit, which could be useful for mineral exploration.The sampling of workers in the hospitality sector proved difficult and they may have been underrepresented in this study.
The goal of sampling creams provided by participants from a range of socioeconomic backgrounds was partly satisfied by the selection of different categories of educational institutions.In all cases, lectures were given to introduce the risks of mercury and each participant was able to directly observe the analysis of their cream.This led to great enthusiasm and usually initiated questions and requests for more analysis.Toxic creams were set out in an exhibit and usually when a participant realized that their cream contained high levels of mercury and had not yet been identified, the item was donated to the exhibit.As a teaching tool, the educational clinics seemed to be effective.Interviews indicated that before the clinics, only 9.6% of the participating students knew that skin-whitening creams could contain mercury.
The sampling in 2011 confirmed an earlier study 24 showing mercury contamination in skin-whitening creams is common in Phnom Penh and a handheld XRF analyzer can readily detect mercury.The researchers processed all skin creams, including ones that were not whitening, that were brought to them and therefore they observed less mercury than in the 2008 sampling of skin-whitening creams selected from markets.Studies from around the world have linked mercury-loaded skin-bleaching creams with negative health impacts.In Hong Kong, patients in the Prince of Wales and Jinling hospitals who had used

Research
Mercury in Skin-Whitening Creams skin-whitening creams containing up to 30,000 μg/g mercury were shown to be suffering from kidney damage. 39-41A U.S. study reported on a Mexican skin-whitener with 60,000 to 100,000 μg/g mercury, resulting in high urinary mercury concentrations and a prevalance of mercury-poisoning symptoms among users. 15e reported exposure period of patients hurt by mercury in skinwhitening creams varies considerably: from 2-60 months, 41 and 1-60 months 14 (with a mean of 13 months).The frequency of use of these creams by patients is rarely reported.Users of the toxic Mexican skin-whitening cream reported the following frequency of use: less than once per day, 18%; once per day, 30%; 2-3 times per day, 52%. 15 2007 Nielson survey of the general public in Asia reported that weekly or daily use of skin-whitening creams ranged between 8% and 30%.21 Handheld XRF analyzers lack the sensitivity of some laboratory analyzers but are able to measure mercury levels in skin whitening creams close to regulatory standards.The comparison of XRF and ICP-OES illustrates the utility of XRF analysis.The XRF we used was about 100 times more sensitive than a mercury kit developed by the Thai government's Department of Medical Science and marketed by JSP Pharmaceutical Manufactory (Thailand) Co. Ltd.The ASEAN cosmetics committee suggested in January 2008 the use of screening kits that would detect 1500 μg/g of mercury, or about 38% (37 out of 98) of the contaminated samples the researchers detected by XRF in this study. 42e fact that the ASEAN committee proposed such a kit at all suggests that regional governments understand that local laboratories lack the analytical capacity to process samples and that a screening tool is needed to focus on the most contaminated creams.
The 18-22% suppression of signal observed in spiking of creams with mercury/DSMO deviates slightly from a study by Shackley 43 who found that XRF was linear for 17 international standards containing strontium with no suppression of signal over three orders of magnitude.Possibly our sample matrix was unusual being a cream but the validation with the reference materials supplied by Thermo Fisher Scientific was within manufacturer's specifications.
With respect to detection of toxic creams, the 18-22% suppression of response in the samples with more than 2000 μg/g of mercury has no applied significance.ICP-MS, AFShydride, CVAAS etc. are designed for much lower concentrations of mercury and without awareness and sample dilution they would be badly contaminated by the worst skinwhitening creams.Chemists that are aware of the high concentration of mercury in skin creams would dilute but they sometimes require a series of dilutions and must take care with the worst of their contaminated glassware.With some AFS-hydride systems, the detection range without dilution is only one order of magnitude, whereas the XRF was useful over four orders of magnitude without dilution.
The comparison of skin-whitening creams purchased from retail centers and beauty shops was complicated by the lack of standard product labeling and the perception of what actually constituted a skin-whitening cream.
Beauty shop creams had minimal to no labeling, and the percentage of samples having > 20 μg/g was higher (34.4%) than in the 424 commercial products tested (7.8%).The products containing mercury work by suppressing melanin synthesis, while weaker products contain non-toxic coloring agents like zinc oxide, barium oxide or titanium oxide.These do not block melanin production but reflect UV light, are white in color and suppress tanning.However, three sun-block products were found with detectable mercury levels.Although these products were mainly advertised as UV-blockers, in small print they did purport to be whitening creams.

Treatment and Control
Presently, the only treatment possible in Cambodia for mercury poisoning from skin-whitening cream is simply to stop using it.There currently are no resources in Cambodia for chelation therapy to remove mercury from people.With some exception, there is good recovery of most subjects who stop using these toxic creams.Prevention and education is the best available strategy for this public health problem.Before the educational clinics that ran parallel with this study, more than 90% of participants did not know that skin-whitening creams could contain mercury.In all cases where mercury was detected, users were cautioned to discontinue using their creams.
It seemed that though most participants took the warnings of toxic contamination seriously, the majority seemed determined to keep using skin-whitening creams, albeit switching products.A few individuals planned to continue using toxic products, despite seeming to understand the risks.

Conclusions
The handheld Thermo Fisher Scientific Niton XL3t analyzer is appropriate for screening for mercury in skin creams, and is useful in Cambodia where levels higher than the instrument detection limit of 20 μg/g are often found.The XRF requires no sample preparation or extraction, can be used by staff with little technical background (e.g.customs officials), and, as in the case of this study, facilitates on-site education on the subject.It also has a much lower detection limit than available kits from Thailand (1,500 μg/g).The XRF analyzer is portable and samples do not have to touch the analyzer, therefore rendering contamination unlikely.As opposed to other methods of measuring metals, expensive reagents are not required for XRF analysis.Like ICP-MS, but unlike most CVAAS systems, XRF can measure many types of metals.The capital cost of a handheld XRF is one-half to a fifth of other methods of measuring metals.Furthermore, the handheld units have their own power supply; this is significant because power supplies in Cambodia are often unreliable and backup power supplies for laboratory units (ICP-MS etc.) are essential.
Handheld XRF analyzers lack the sensitivity of some laboratory analyzers but are able to measure mercury levels in skin-whitening creams close to regulatory standards.It is possible to use a laboratory-based XRF unit which Variability in Mercury (μg/g) Levels in Market Skin-Whitening Creams Measured by XRFIn this experiment each replicate came from a different market and none were provided by students.1 Hg = mercury; 2 SD = standard deviation; 3 RSD%= relative standard deviation in percentage; 4 <LOD = less than detection limit of 20 μg/g al. found in 2007 and

Table 2 -Comparison of XRF Results to ICP-OES and CVAAS
4 Thermo = Thermo Fisher Scientific; 2 2σ = standard deviation; 3 Essco = Essco Safety Check;4ICP-OES = National University of Singapore; 5The ICP_OES result is a mean of 2 samples whereas the XRF results are instrument-generated means; 6 CVAAS = University of California; 7 ND = no detection