Temporal Variability of Urinary Di(2-Ethylhexyl) Phthalate Metabolites During a Dietary Intervention Study

ORIGINAL ARTICLE Temporal variability of urinary di(2-ethylhexyl) phthalate metabolites during a dietary intervention study

Janet M. Ackerman1, Robin E. Dodson1, Connie L. Engel2, Janet M. Gray3 and Ruthann A. Rudel1

Exposure to di(2-ethylhexyl) phthalate (DEHP) may be related to adverse health effects including developmental and reproductive disorders, prompting interest in strategies for reducing human exposure. We previously reported a reduction of DEHP metabolite levels in composite urine samples by more than 50% (geometric means) during a 3-day dietary intervention avoiding plastics in food packaging, preparation, and storage. In the present study, we analyzed individual spot urine samples before compositing in order to evaluate temporal variability. There were no meaningful changes in any of the previous findings when using individual rather than composited samples. Individual urine samples, like the composites, showed significant decreases of Z50% in all three measured DEHP metabolites during the intervention. Compositing urine samples provided sufficient information to observe the effect of the intervention, whereas reducing analytical expenses compared with analyzing multiple samples individually. Low intraclass correlations (ICCs) for samples collected from the same person before the intervention indicate the importance of collecting multiple samples per exposure condition. Substantially larger ICCs during and after the intervention suggest that much of the variability observed in DEHP metabolite levels originates from dietary exposure.

Journal of Exposure Science and Environmental Epidemiology (2014) 24, 595–601; doi:10.1038/jes.2013.93; published online 22 January 2014 Keywords: dietary exposure; biomonitoring; endocrine disruptors; phthalates; analytical methods

INTRODUCTION was a major contributor to one participant’s high DEHP exposure. 17 Di(2-ethylhexyl) phthalate (DEHP), a common additive to polyvinyl We previously reported that switching to a ‘‘fresh foods’’ diet, chloride (PVC), is present in a variety of consumer goods, including with minimal exposure to plastic or epoxy-based food packaging, food packaging,1,2 leading to widespread exposure in the United for 3 days reduced geometric mean concentrations of bisphenol A States and elsewhere. DEHP metabolites have been detected in (BPA) and DEHP metabolites by over half in the urine of five the urine of virtually 100% of a representative sample of the US California families. A larger-scale randomized trial of a similar population.3,4 Evidence in animal studies indicates that DEHP is an intervention18 unexpectedly found increased DEHP metabolite anti-androgenic endocrine disruptor, with some evidence for levels in subjects following a ‘‘no packaged food’’ diet because of effects on other pathways.5–8 Consistent with the animal evidence, unanticipated DEHP contamination in coriander and milk, cross-sectional studies in humans have linked DEHP exposure with demonstrating that food contamination with DEHP is another effects on male reproductive development,9 as well as effects on potentially important contributor to dietary exposure. adult male hormone levels and semen quality,10–12 although the Although using biomarkers such as urinary concentrations to associations described in these studies may be confounded by co- estimate exposure has many advantages, temporal variability of occurrences of exposures to other environmental chemicals or to biomarker concentrations can lead to exposure misclassification. dietary and lifestyle factors. Temporal variation is an important consideration for DEHP, as Exposure estimates based on concentrations in food, air, dust, DEHP metabolites have short half-lives14,15,19,20 and exposure and consumer products indicate that diet is likely to be a major patterns appear sporadic.16 Temporal variability in phthalate source of non-occupational exposure to DEHP.1,13 Recent studies metabolite concentrations from multiple urine samples from the measuring phthalate metabolite excretion after fasting or changes same individual has been well documented, with intraclass in diet have added to the evidence that non-occupational correlations (ICCs) for DEHP metabolite from multiple samples exposures to DEHP and di-isononyl phthalate are largely driven from the same individual ranging from 0.08 to 0.67,21,22 although by diet, whereas exposures to other phthalates are more most of the literature reports ICCs between 0.2 and 0.4.23–28 To influenced by other sources.14,15 minimize misclassification of typical phthalate exposure because Exposure studies have had limited ability to identify what foods of temporal variability, investigators have suggested including were responsible for participants’ phthalate exposure, with a few more participants in exposure studies23 or collecting repeated notable exceptions. Lorber and Calafat,16 using detailed diaries spot urine samples.24,29 Existing studies of temporal variability of and repeated urine measurements to model intake, identified phthalate metabolite levels have mostly focused on using urinary increases in phthalate metabolite concentrations after individual measurements to classify long-term phthalate exposure, for meals, including a single meal of prepared food (reported as example, for the purpose of investigating potential associations ‘‘bagel w/egg, sausage, cheese, coffee bought at gas station’’) that between phthalate exposures and health outcomes. The few

1Silent Spring Institute, Newton, Massachusetts, USA; 2Breast Cancer Fund, San Francisco, California, USA and 3Department of Psychology, Vassar College, Poughkeepsie, New York, USA. Correspondence to: Janet M. Ackerman, Silent Spring Institute, 29 Crafts Street, Newton, MA 02458, USA. Tel.: +617 332 4288x 231. Fax: +617 332 4288x 231. E-mail: [email protected] Received 8 July 2013; accepted 15 November 2013; published online 22 January 2014 Phthalate metabolites and diet Ackerman et al 596 studies that have measured phthalate metabolites during a urine with phthalate metabolites and their isotopically labeled surrogates period of fasting14,15 provide valuable information about pharm- added at known concentrations), one of which was analyzed in duplicate. acokinetics and exposure, but to our knowledge no published Data and analyses are presented only for metabolites of DEHP, with the work has directly addressed strategies for accurately esti- exception of summary data in Supplementary Table 1. mating exposure to phthalates during periods of altered expo- In addition to a limit of detection (LOD) for each analyte in each batch based on the lowest level calibrated, the laboratory calculated an LOD for sure, as in an intervention designed to reduce potential sources of each analyte in each sample, based on recovery of isotopically labeled exposure. 17 surrogates added to each sample. The laboratory reported the higher of Rudel et al. reported concentrations in composites of two these two limits for each sample.30 LODs for DEHP metabolites were urine samples collected after dinner on sequential days from mostly around 1 ng/ml (median LODs 1 ng/ml for all three metabolites), each participant before, during, and after the dietary intervention. with a few elevated values for each compound (MEHHP: 90th percentile BPA and DEHP metabolite levels dropped substantially and LOD 3.5 ng/ml, maximum LOD 14.2 ng/ml; MEOHP: 90th percentile significantly during the intervention. BPA levels rose significantly 3.0 ng/ml, maximum 10.2 ng/ml; MEHP: 90th percentile 1.98 ng/ml, after the end of the fresh foods intervention, whereas DEHP maximum 3.4 ng/ml). Elevated LODs were reported only in samples in metabolite levels did not. To provide a more detailed picture of which the analyte was quantified above the LOD, with the exception of three samples in which MEHP was not detected above an elevated exposure and variability over the period of the study, the present detection limit of 2.97 ng/ml. work measured DEHP metabolite levels in each individual sample. Eighteen MEHP measurements reported as non-detects were assigned This reanalysis allowed us to compare the relative benefits of the sample-specific LOD.30 This approach biases results up, which compositing samples as opposed to testing samples individually. constitutes a bias toward the null given that non-detects were most common during the intervention, when we observed a decrease in concentrations (Supplementary Table 2). METHODS We compared the concentrations of the DEHP metabolites measured in Intervention Study Design composite samples in 2010 and the concentrations measured in individual samples in 2011 using Spearman correlation tests. The design of the dietary intervention and urine sample collection has been described previously.17 Briefly, in January 2010, five families in the San Francisco Bay Area, each comprising two adults aged 36–46 years and two children aged 3–11 years, participated in the three phases of the Data Analysis and Statistics study: pre-intervention, when participants ate their typical diet; interven- We report phthalate metabolite concentrations in urine as ng/ml, tion, when they ate only foods provided by a caterer, prepared from fresh unadjusted for creatinine. As dietary protein levels influence urinary 31,32 17 ingredients (no canned or frozen foods) and packaged almost exclusively creatinine concentrations, changes in creatinine concentration without contact with plastic; and post-intervention, when the participants during the intervention were likely unrelated to changes in urine returned to preparing their own food and were instructed to return to their dilution. However, to address the influence of urine dilution, we included normal diet. With a few exceptions,17 all participants ate the fresh food creatinine as a variable in the mixed-effects models described below, as 33 (intervention) diet for all meals and snacks on days 3–5 of the 8-day study recommended by Barr et al. We used unadjusted concentrations for and collected a urine sample in the evenings of days 1 and 2 analyses comparing concentrations between phases, and used creatinine- (preintervention), 4 and 5 (intervention), and 7 and 8 (post-intervention; corrected concentrations for ICC calculations within phases, because we Figure 1). Each of the 20 participants contributed six samples over the expected that variation in creatinine concentration within a study phase course of the study. Informed consent was obtained from all adult study primarily reflect changes in urine dilution. Creatinine-adjusted and participants on behalf of themselves and their children. All minor unadjusted concentrations are summarized in Supplementary Table 2. participants, with the exception of one 3-year-old, provided verbal assent. We used the sums of the geometric means of the concentrations of each DEHP metabolite on each day as a simple means of comparison between days. To evaluate the consistency of results, we estimated correlations Chemical Analysis and Quality Assurance between concentrations measured in composite samples (analyzed in Urine specimen jars were stored, double bagged, in participants’ home 2010) and simulated composites (averages of concentrations in pairs of freezers until pickup within a week of the study’s conclusion. After pickup, individual samples analyzed in 2011). urine samples were stored in a freezer overnight and shipped overnight on We used log-transformed data for mixed-effects models and analysis of blue ice to the laboratory, where they were stored frozen at À 20 1C. variance models that require the assumption of normality, because the In February and March 2010 (o2 and o8 weeks after collection, distributions of analyte concentrations appear approximately log-normal. respectively), laboratory technicians thawed the samples and removed We used non-parametric statistics based on ranks of untransformed data 40 ml aliquots for quantification of BPA (February) and phthalate to test for differences (paired Wilcoxon signed-rank) and estimate metabolite (March) concentrations. These analyses used 1 ml subsamples correlation (Spearman). drawn from composites of equal volumes from the two samples from each We used a linear spline mixed-effects model to evaluate changes in participant during each phase of the study, except for three samples that DEHP metabolite concentrations in individual samples between the study were excluded because of ambiguous labeling (the other sample from phases. The model had a single knot placed at the intervention and the same person during the same study phase was used instead of included family and participant as multilevel random effects and creatinine a composite in these cases). After the removal of aliquots for testing, as a fixed effect. This model treated the two samples collected for each samples were refrozen at À 20 1C. In September–November 2011, participant during each phase of the study as independent. Although laboratory technicians thawed the samples and removed 1 ml aliquots the assumption of independence is violated in this case, making this for reanalysis as individual (non-composited) samples, with the exception assumption was preferable to overfitting the small data set by adding of the three ambiguously labeled samples, the three samples that had additional terms to the model. We compared the results of this model with been individually analyzed in previous analyses, and two samples of which a similar model using the concentrations detected in composite samples there was insufficient volume. (this differs slightly from the previously reported model17 in that samples The individual samples were analyzed for the DEHP metabolites mono- were excluded if we did not have both composite and individual data), and (2-ethyl-5-hydroxyhexyl) phthalate (MEHHP), mono-(2-ethyl-5-oxohexyl) with a model using simulated composites (averages of concentrations phthalate (MEOHP), and mono-2-ethylhexyl phthalate (MEHP). They were from paired individual samples). For purposes of comparison, we excluded also analyzed for seven other phthalate metabolites (Supplementary the three samples that were never composited and two with insufficient Table 1). As previously reported, composited samples were analyzed for volume. Similar models were used to evaluate changes between pairs of MEHHP, MEOHP, MEHP, and BPA, as well as four other phthalate consecutive phases considering only one day of each phase, for a total of metabolites. Composited and individual samples were also analyzed for eight comparisons (day 1 vs day 4, day 4 vs day 7, etc; four comparisons creatinine. between preintervention and during-intervention, four between during- Analysis was conducted at AXYS Laboratory, Sidney, British Columbia, by intervention and post-intervention). We also used pairwise Wilcoxon HPLC-Tandem Mass Spectroscopy using isotope dilution quantification as signed rank tests to compare concentrations between days and between previously described.17 Samples were analyzed in batches, each including phases, using creatinine-adjusted concentrations. We used the Holm34 a procedural blank (water) and two spiked reference samples (synthetic method to correct P-values to account for the multiple comparisons.

Journal of Exposure Science and Environmental Epidemiology (2014), 595 – 601 & 2014 Nature America, Inc. Phthalate metabolites and diet Ackerman et al 597 To evaluate the relative importance of temporal (within person, within o5x LOD, vs 8% MEHHP, 20% MEOHP), and most likely less phase) variability and interpersonal (between-person, within phase) precise because of their closeness to the LOD. variability, we calculated ICCs for the concentrations of each metabolite The sum of the geometric means of the concentrations of the during each phase, treating participant as a random effects predictor of log 35 three DEHP metabolites ranged from 29.1 ng/ml (49 mg/g creati- (creatinine-adjusted concentration). nine) on day 5 to 112 ng/ml (128 mg/g) on day 1. Geometric means All data analysis was performed in R 3.0.0.36 of DEHP metabolites calculated from individual sample data were slightly lower on each day than those calculated from composite or simulated composite (average of concentrations in paired RESULTS individual samples) data, consistent with the skewness of the data Method Performance (Table 2). Analytical methods performed well and results for individual samples were consistent with the previous study measuring phthalate metabolites in composited urine samples.17 None of the Mixed-Effects Models reported phthalate metabolites were detected in any of the Mixed-effects models using individual, composited, and simulated laboratory blanks. Recovery in spiked samples was within 10% of composite data all showed significant decreases (slopeo0, spiked concentration. Samples analyzed in duplicate agreed to Po0.05) in all three DEHP metabolite levels between the within 15% of the mean, with the exception of one pair in which preintervention and intervention study phases, although all slope [MEHHP] differed by 22% of the mean, and one pair of duplicates estimates were positive (Table 2). No model showed significant in which [MEOHP] differed by 16% of the mean. change in any phthalate metabolite concentration between the As would be expected, DEHP metabolite levels were correlated. intervention and post-intervention study phases. Slope estimates MEOHP and MEHHP concentrations were strongly correlated with for change between phases were similar whether using individual, each other (Spearman r ¼ 0.94, Po0.001), and well correlated composited, or simulated composited data, although 95% with [MEHP] (Spearman r ¼ 0.71–0.73, Po0.001). Creatinine and confidence intervals for slopes were smaller in the models using DEHP metabolite concentrations in composited samples (mea- individual samples because of the increased power associated sured 2010) and simulated composites (average of paired with the additional measurements being treated as independent. individual samples analyzed 2011) were highly correlated (Spear- As previously reported, age and sex did not strongly influence man r ¼ 40.97, Po0.001; Supplementary Figures 1 and 2). phthalate metabolite concentrations (data not shown).

Overall Results Day by Day Comparisons We detected MEHHP and MEOHP in 100% of individual samples and MEHP in 84% (Table 1). To evaluate the value of collecting multiple urine samples for each DEHP metabolite concentrations were generally lower during exposure condition, we modeled changes in DEHP metabolite the intervention than before the intervention, with intermediate concentrations between each preintervention day and each values after the intervention (Figure 2, Table 2; creatinine adjusted during-intervention day, and between each during-intervention values in Supplementary Figure 3). Medians were lower on both day and each post-intervention day, for a total of eight days during the dietary intervention than on either preinterven- comparisons for each of the three metabolites. These mixed- tion day. Median, mean, geometric mean, third quartile, and effects models, like the models described above, include maximum concentrations were highest on day 1 (preintervention) creatinine as a fixed effect and participant and family as nested and lowest on day 5 (during intervention). Creatinine-adjusted random effects. The concentrations of all DEHP metabolites were concentrations followed a similar pattern (Supplementary significantly lower (upper bound of 95% confidence interval for Figure 3). MEHHP and MEOHP concentrations were three to five slopeo0) on day 5 than on either day 1 or day 2. MEHHP was times higher than corresponding concentrations of MEHP. MEHP significantly lower on day 4 compared with day 1, but not levels were much closer to the LOD (65% of reported MEHP values compared with day 2, and day 4 concentrations of MEOHP and MEHP were not significantly different than concentrations on either preintervention day. Concentrations of all three metabolites were significantly higher on day 7 than on day 5, but no other combination of a during-intervention day and a post-intervention day showed a significant change. Similar models using composited and simulated composited data also showed significant reductions in all three DEHP metabolites between the preintervention and during-intervention study phases. We also compared daily levels using paired Wilcoxon tests corrected for multiple comparisons,34 with similar results: There were significant (Po0.005) differences in creatinine-uncorrected concentrations of all three DEHP metabolites between day 1 and Figure 1. Study design. Modified from Rudel et al.,17 with permission day 5. Between day 2 and day 5, there were significant differences of Environmental Health Perspectives. in MEHHP (Po0.01) and MEOHP (Po0.03) concentrations, but not

Table 1. Concentrations of DEHP metabolites detected in this study and in NHANES 2009–2010.

Metabolite Abbreviation NHANES 2009–2010 Median pre-intervention Percent detected median (mg/l) this study this study (n)

Mono-(2-ethyl-5-hydroxyhexyl) phthalate MEHHP 12.9 45.7 100% (112) Mono-(2-ethyl-5-oxohexyl) phthalate MEOHP 8.02 18.5 100% (112) Mono-2-ethylhexyl phthalate MEHP 1.51 5.1 84% (94) Abbreviations: DEHP, di(2-ethylhexyl) phthalate; NHANES, National Health and Nutrition Examination Survey.

& 2014 Nature America, Inc. Journal of Exposure Science and Environmental Epidemiology (2014), 595 – 601 Phthalate metabolites and diet Ackerman et al 598

Figure 2. DEHP metabolites in urine samples collected before, during, and after a dietary intervention. Concentrations of all metabolites were lower during the intervention than before, with intermediate levels after the intervention. Thick lines are medians. Boxes extend from 25th percentile to 75th percentile, with whiskers extending to the most extreme data point within 1.5* (interquartile range) of the 25th or 75th percentile. Blue line indicates geometric mean of concentrations within a day. n ¼ 18 on days 1, 2, 7, and 8; n ¼ 19 on days 4 and 5.

Table 2. Geometric means and results of mixed-effects models for DEHP metabolites.

Analyte Variable % Change in GM (GMs; ng/ml) 95% CI for slope estimate Analyzed as

MEHHP Pre- to during intervention change À 60 (59 to 24) ( À 1.2, À 0.53)a Individual À 61 (67 to 26) ( À 1.4, À 0.4)a Simulated composite À 60 (65 to 26) ( À 1.4, À 0.38)a Composite During to post-intervention change 15 (24 to 27) ( À 0.15, 0.54) Individual 14 (26 to 30) ( À 0.33, 0.68) Simulated composite 9 (26 to 28) ( À 0.37, 0.64) Composite MEOHP Pre- to during intervention change À 58 (25 to 11) ( À 1.2, À 0.46)a Individual À 58 (29 to 12) ( À 1.4, À 0.3)a Simulated composite À 59 (30 to 12) ( À 1.4, À 0.32)a Composite During to post-intervention change 6 (11 to 11) ( À 0.23, 0.48) Individual 4 (12 to 12) ( À 0.45, 0.62) Simulated composite 3 (12 to 13) ( À 0.44, 0.61) Composite MEHP Pre- to during intervention change À 50 (5.4 to 2.7) ( À 1.1, À 0.28)a Individual À 56 (6.6 to 2.9) ( À 1.4, À 0.17)b Simulated composite À 56 (7.7 to 3.4) ( À 1.3, À 0.2)b Composite During to post-intervention change 15 (2.7 to 3.1) ( À 0.17, 0.61) Individual 13 (2.9 to 3.3) ( À 0.39, 0.79) Simulated composite 8 (3.4 to 3.7) ( À 0.41, 0.71) Composite Abbreviations: CI, conﬁdence interval; DEHP, di(2-ethylhexyl) phthalate; GM, geometric mean; MEHP, mono-2-ethylhexyl phthalate; MEHHP, mono-(2-ethyl-5- hydroxyhexyl) phthalate; MEOHP, mono-(2-ethyl-5-oxohexyl) phthalate. aPo0.005. bPo0.05.

MEHP concentrations (P40.4). No other day-wise comparisons The two measurements from each participant in each phase showed significant differences (P40.1). were significantly correlated (Spearman correlation coefficient ¼ Z0.58, Po0.05), with highest coefficients after the intervention Correlations Between Concentrations in Samples from One and lowest before the intervention. The one exception was MEHP Participant during the intervention phase (Spearman correlation coefficient We used simple random effects linear models and Spearman’s ¼ 0.31, P40.19), when there was a particularly high proportion of tests to evaluate the correlation of creatinine-corrected concen- samples in which MEHP was not quantifiable. trations of each DEHP metabolite in the two samples from the same person during each phase of the study (Figure 3, Supplementary Table 3). DISCUSSION For each metabolite, ICCs based on a linear model of log Effects of the Dietary Intervention (creatinine-adjusted concentrations) predicted by participant were Mixed-effects models using concentrations of DEHP metabolites in highest after (0.70–0.87) and during (0.38–0.70) the intervention individual urine samples, in samples composited by participant and lowest during the preintervention phase (0.30–0.50). In each and study phase, and in simulated composite samples (averaged phase, MEHHP and MEOHP ICCs were similar, and MEHP ICCs were by participant and study phase) showed substantial and lower. Lower ICCs during preintervention compared with the other significant decreases in DEHP metabolite concentrations because two phases of this study primarily reflect decreased within person of the dietary intervention. This analysis strengthens the evidence variability associated with the intervention (Supplementary from our previous study of composited samples from the same Table 3). ICCs for using family as the grouping variable were intervention17 that food packaging is a major source of DEHP generally similar to those using participant, except in the cases of exposure in the United States. We observed this change using MEOHP and MEHHP during and after the intervention, in which multiple statistical approaches; however, owing to limited sample family ICCs were substantially lower (Supplementary Table 3). size, we cannot rule out the possible role of chance in our findings.

Journal of Exposure Science and Environmental Epidemiology (2014), 595 – 601 & 2014 Nature America, Inc. Phthalate metabolites and diet Ackerman et al 599

Figure 3. Temporal variability in DEHP metabolites concentrations in samples from the same person and from the same family before, during, and after the dietary intervention.

However, the longitudinal sampling in this intervention study adopted subsequent to the intervention, such as using glass allowed individuals to serve as their own controls, avoiding many storage containers or eating out less often, might have affected sources of confounding that can limit cross-sectional studies. sources of DEHP more than sources of BPA. However, As previously discussed,17 levels of DEHP and most other Sathyanarayana et al.18 did not see a decrease in urinary DEHP phthalate metabolites in this study were within the range of those metabolite or BPA levels in participants given advice on avoiding reported in recent National Health and Nutrition Examination dietary sources of plasticizers. Alternatively, the discrepancy Survey reports,3 although this population had somewhat higher between the increase in BPA and the lack of an increase in median levels of most metabolites excluding monoethyl DEHP metabolites could simply reflect a greater proportion of BPA phthalate, which was substantially lower in this group. This exposure than DEHP exposure coming from food packaging similarity suggests that our findings are likely to be broadly sources. This is consistent with the larger decrease in BPA than in relevant to American diets, despite the small number and DEHP metabolite levels because of the intervention.17 geographic homogeneity of our study participants. Notably, geometric means of urinary DEHP metabolite concen- Although concentrations of DEHP metabolites increased after trations observed during the intervention, at least 48 h after the the intervention, none of the increases were statistically sig- last exposure to packaged food, were still many times higher than nificant, excluding a few comparisons of individual days. In the concentrations reported by Koch et al.14 in urine from contrast, BPA levels in the same group nonetheless increased volunteers fasting for as few as 12–24 h. This suggests that significantly after the intervention, with a 200% increase in substantial dietary sources of DEHP remained even after eliminat- geometric mean.17 The half-life of the DEHP metabolites (5–10 h) ing exposure to food packaging, consistent with detections of is not long enough to explain the lack of a significant increase phthalates in processed and unprocessed food products.37 Some 30 þ hours after participants began eating their own food; in fact, of the difference could also reflect higher exposure to DEHP from we measured higher DEHP metabolite concentrations on the first non-dietary sources in our study population of Californian families post-intervention day than on the second. Although chance may than in the German adults studied by Koch et al. play a role, the findings also suggest that study participants may The higher concentrations observed on the first day of each have altered their dietary habits because of their participation in study phase most likely reflect random variability in exposure rather the study and that DEHP and BPA have different dietary sources. than any meaningful difference because of biological residence That is, the behavioral changes that the families might have time between the 2 days within any phase. Studies measuring

& 2014 Nature America, Inc. Journal of Exposure Science and Environmental Epidemiology (2014), 595 – 601 Phthalate metabolites and diet Ackerman et al 600 phthalate metabolites in urine from fasting subjects or from ACKNOWLEDGEMENTS volunteers exposed to D4-DEHP have established that MEHHP, We thank the Vassar College Institutional Review Board for assistance in study MEOHP, and MEHP have urinary half-lives of excretion between 4 planning and human subjects research oversight. This work was supported by and 10 h, with urinary concentrations of these metabolites leveling funding from Passport Foundation (Charlotte, North Carolina), the Susan S Bailis off between 12 and 24 h after dietary exposure.14,15,19,20 For Breast Cancer Research Fund at Silent Spring Institute, and the American Chemistry example, as participants in our study had gone at least 48 h Council Long-Range Research Initiative. without eating packaged food when they collected day 4 urine samples, it is unlikely that the higher DEHP metabolite con- DISCLAIMER centrations on day 4 compared with day 5 represent excretion of The authors were free to design, conduct, interpret, and publish research without DEHP from preintervention dietary exposures. interference or input from any funding organization. This publication has not been reviewed by the American Chemistry Council and views expressed are solely the authors’. The authors declare that they have no conflicts of interest related to this work. Variability and Implications for Sampling Strategies By analyzing urine samples individually rather than as composites, we were able to analyze temporal variability in phthalate REFERENCES metabolite levels. Analysis of multiple urine samples per phase 1 NTP-CERHR. NTP-CERHR Monograph on the Potential Human Reproductive and also yielded somewhat tighter confidence intervals for slope Developmental Effects of Di(2-Ethylhexyl) Phthalate. National Toxicology Program estimates in our mixed-effects models, but it did not change our Center for the Evaluation of Risks to Human Reproduction: Research Triangle Park, overall conclusions regarding the effect of the dietary intervention NC, 2006. Report No.: NIH 06-4476. on DEHP exposure. 2 EFSA. Opinion of the scientific panel on food additives, flavorings, processing aids Correlations (evaluated by ICC or Spearman’s correlation and materials in contact with food (AFC) on a request from the commission coefficient) of DEHP metabolite concentrations in multiple spot related to bis (2-ethylhexyl) phthalate (DEHP) for use in food contact materials. EFSA J 2005; 243: 1–20. urine samples from the same person before the dietary 3 Fourth National Report on Human Exposure to Environmental Chemicals, Updated intervention (Figure 3, Supplementary Table 3a) were squarely Tables, September 2013. Centers for Disease Control and Prevention National Center 23–28,38 within the range reported by the many investigators who for Health Statistics: Atlanta, GA, 2013. have previously measured DEHP metabolites in repeated urine 4 Meeker JD, Ferguson KK. Relationship between urinary phthalate and bisphenol A samples from the same people. However, correlations were much concentrations and serum thyroid measures in U.S. adults and adolescents from higher for samples collected from the same person during the the National Health and Nutrition Examination Survey (NHANES) 2007-2008. intervention (MEHHP and MEOHP) and after intervention (all Environ Health Perspect 2011; 119: 1396–1402. metabolites) than in past reports. 5 Gray Jr. LE, Ostby J, Furr J, Price M, Veeramachaneni DN, Parks L. Perinatal The higher ICCs and lower within-participant variance during exposure to the phthalates DEHP, BBP, and DINP, but not DEP, DMP, or DOTP, alters sexual differentiation of the male rat. Toxicol Sci 2000; 58: 350–365. the intervention, although all participants were eating essentially 6 Martinez-Arguelles DB, Campioli E, Culty M, Zirkin BR, Papadopoulos V. identical diets, suggest that residual exposures from non-food Fetal origin of endocrine dysfunction in the adult: the phthalate model. J Steroid sources vary less from day-to-day than do typical food-related Biochem Mol Biol 2013; 137: 5–17. exposures. Many non-dietary sources of exposure to DEHP (e.g., 7 Hannas BR, Lambright CS, Furr J, Howdeshell KL, Wilson VS, Gray Jr LE. shower curtain, flooring, car interior) would be shared between Dose-response assessment of fetal testosterone production and gene expression members of the same household, and the slightly increased family levels in rat testes following in utero exposure to diethylhexyl phthalate, ICCs for MEHHP and MEOHP during intervention (Figure 3, diisobutyl phthalate, diisoheptyl phthalate, and diisononyl phthalate. Toxicol Sci Supplementary Table 3b) support this interpretation. 2011; 123: 206–216. 8 National Academies. Phthalates and Cumulative Risk Assessment: the Tasks Ahead. National Academies Press: Washington, DC, 2008. 9 Swan SH. Environmental phthalate exposure in relation to reproductive outcomes CONCLUSION and other health endpoints in humans. Environ Res 2008; 108: 177–184. We have demonstrated that urinary metabolites of DEHP 10 Hauser R. Urinary phthalate metabolites and semen quality: a review of a decreased by more than 50% after 2–3 days of a dietary potential biomarker of susceptibility. Int J Androl 2008; 31: 112–117. intervention avoiding exposure to food packaging. This supports 11 Meeker JD, Calafat AM, Hauser R. Di(2-ethylhexyl) phthalate metabolites may alter earlier reports that DEHP exposures are dominated by DEHP in thyroid hormone levels in men. Environ Health Perspect 2007; 115: 1029–1034. food packaging. The conclusions are similar whether using 12 Meeker JD, Calafat AM, Hauser R. Urinary metabolites of di(2-ethylhexyl) phthalate are associated with decreased steroid hormone levels in adult men. J Androl 2009; composited or individual urine samples from 2 days in each 30: 287–297. exposure condition. The considerable expense of analyzing each 13 Fromme H, Gruber L, Schlummer M, Wolz G, Bohmer S, Angerer J et al. Intake of sample individually and relatively modest increase in statistical phthalates and di(2-ethylhexyl)adipate: Results of the Integrated Exposure power from individual rather than composite measurements, even Assessment Survey based on duplicate diet samples and biomonitoring data. when treating samples as independent within a study phase, Environ Int 2007; 33: 1012–1020. suggest that compositing repeat spot urine samples is an efficient 14 Koch HM, Lorber M, Christensen KL, Palmke C, Koslitz S, Bruning T. Identifying strategy for biomonitoring to assess non-occupational DEHP sources of phthalate exposure with human biomonitoring: Results of a 48h fasting exposure, especially in research settings with limited resources. study with urine collection and personal activity patterns. Int J Hyg Environ Health Based on the relatively low ICCs observed before the intervention 2013; 216: 672–681. 15 Aylward LL, Lorber M, Hays SM. Urinary DEHP metabolites and fasting time in in this study and by other researchers, compositing multiple spot NHANES. J Expo Sci Environ Epidemiol 2011; 21: 615–624. samples is also likely to give a more accurate estimate of exposure 16 Lorber M, Calafat AM. Dose reconstruction of di(2-ethylhexyl) phthalate using a in the general population than analysis of a single spot sample. simple pharmacokinetic model. Environ Health Perspect 2012; 120: 1705–1710. Further investigations of biomarker variability under different 17 Rudel RA, Gray JM, Engel CL, Rawsthorne TW, Dodson RE, Ackerman JM et al. Food exposure conditions could help to clarify when multiple samples packaging and bisphenol A and bis(2-ethylhexyl) phthalate exposure: findings are necessary for accurate exposure assessment, and provide from a dietary intervention. Environ Health Perspect 2011; 119: 914–920. additional information about the prevalence and variability of 18 Sathyanarayana S, Alcedo G, Saelens BE, Zhou C, Dills RL, Yu J et al. Unexpected different sources of phthalates and other chemicals of concern. results in a randomized dietary trial to reduce phthalate and bisphenol A exposures. J Expo Sci Environ Epidemiol 2013; 23: 378–384. 19 Anderson WA, Castle L, Hird S, Jeffery J, Scotter MJ. A twenty-volunteer study using deuterium labelling to determine the kinetics and fractional excretion of CONFLICT OF INTEREST primary and secondary urinary metabolites of di-2-ethylhexylphthalate and di-iso- The authors declare no conflict of interest. nonylphthalate. Food Chem Toxicol 2011; 49: 2022–2029.

Journal of Exposure Science and Environmental Epidemiology (2014), 595 – 601 & 2014 Nature America, Inc. Phthalate metabolites and diet Ackerman et al 601 20 Koch HM, Bolt HM, Preuss R, Angerer J. New metabolites of di(2-ethylhex- phenols among minority children in the United States. Environ Res 2008; 106: yl)phthalate (DEHP) in human urine and serum after single oral doses of deu- 257–269. terium-labelled DEHP. Arch Toxicol 2005; 79: 367–376. 29 Preau Jr JL, Wong LY, Silva MJ, Needham LL, Calafat AM. Variability over 1 week in 21 Hoppin JA, Brock JW, Davis BJ, Baird DD. Reproducibility of urinary phthalate meta- the urinary concentrations of metabolites of diethyl phthalate and di(2-ethyl- bolites in ﬁrst morning urine samples. Environ Health Perspect 2002; 110: 515–518. hexyl) phthalate among eight adults: an observational study. Environ Health 22 Braun JM, Smith KW, Williams PL, Calafat AM, Berry K, Ehrlich S et al. Variability of Perspect 2010; 118: 1748–1754. urinary phthalate metabolite and bisphenol A concentrations before and during 30 US EPA. Data Quality Assessment: Statistical Methods for Practitioners. US EPA: pregnancy. Environ Health Perspect 2012; 120: 739–745. Washington DC, 2006. Report No.: US EPA QA/G-9S. 23 Frederiksen H, Kranich SK, Jorgensen N, Taboureau O, Petersen JH, Andersson AM. 31 Kesteloot HE, Joossens JV. Relationship between dietary protein intake and Temporal variability in urinary phthalate metabolite excretion based on spot, serum urea, uric acid and creatinine, and 24-hour urinary creatinine excretion: the morning, and 24-h urine samples: considerations for epidemiological studies. BIRNH Study. J Am Coll Nutr 1993; 12: 42–46. Environ Sci Technol 2012; 47: 958–967. 32 Neubert A, Remer T. The impact of dietary protein intake on urinary creatinine 24 Hauser R, Meeker JD, Park S, Silva MJ, Calafat AM. Temporal variability of urinary excretion in a healthy pediatric population. J Pediatrics 1998; 133: 655–659. phthalate metabolite levels in men of reproductive age. Environ Health Perspect 33 Barr DB, Wilder LC, Caudill SP, Gonzalez AJ, Needham LL, Pirkle JL. Urinary crea- 2004; 112: 1734–1740. tinine concentrations in the U.S. population: implications for urinary biologic 25 Baird DD, Saldana TM, Nepomnaschy PA, Hoppin JA, Longnecker MP, Weinberg monitoring measurements. Environ Health Perspect 2005; 113: 192–200. CR et al. Within-person variability in urinary phthalate metabolite concentrations: 34 Holm S. A simple sequentially rejective multiple test procedure. Scand J Stat 1979; measurements from specimens after long-term frozen storage. J Expo Sci Environ 6: 65–70. Epidemiol 2010; 20: 169–175. 35 Rosner B. Fundamentals of Biostatistics, 5th edn. Duxbury Press, 2000. 26 Fromme H, Bolte G, Koch HM, Angerer J, Boehmer S, Drexler H et al. Occurrence 36 R Development Core Team. R: A Language and Environment for Statistical and daily variation of phthalate metabolites in the urine of an adult population. Computing, 3.0.0 ed. R Foundation for Statistical Computing: Vienna, Austria, Int J Hyg Environ Health 2007; 210: 21–33. 2013. 27 Adibi JJ, Whyatt RM, Williams PL, Calafat AM, Camann D, Herrick R et al. Char- 37 Cao XL. Phthalate esters in foods: sources, occurrence, and analytical methods. acterization of phthalate exposure among pregnant women assessed by repeat Compr Rev Food Sci Food Saf 2010; 9: 21–43. air and urine samples. Environ Health Perspect 2008; 116: 467–473. 38 Suzuki Y, Niwa M, Yoshinaga J, Watanabe C, Mizumoto Y, Serizawa S et al. 28 Teitelbaum SL, Britton JA, Calafat AM, Ye X, Silva MJ, Reidy JA et al. Temporal Exposure assessment of phthalate esters in Japanese pregnant women by using variability in urinary concentrations of phthalate metabolites, phytoestrogens and urinary metabolite analysis. Environ Health Prev Med 2009; 14: 180–187.

Supplementary Information accompanies the paper on the Journal of Exposure Science and Environmental Epidemiology website (http:// www.nature.com/jes)

& 2014 Nature America, Inc. Journal of Exposure Science and Environmental Epidemiology (2014), 595 – 601