Our main findings were associations of MEP and low MWP with BMI and waist circumference among overweight children. For example, there was a 2 unit increase in BMI and a 5cm increase in waist circumference for both MEP and low MWP going from the from 1st (<1μm) to 3rd (>3μm) micromolar biomarker quantiles. This corresponds approximately to a tenfold increase in concentrations (median MEP 100μg/gC in <1μm quantile and 1211μg/gC in >3μm quantile). Of note, MEP was the highest concentration individual metabolite in our study. We did not observe associations between the other phthalates and body size among normal weight girls and boys. To our knowledge, this is the first longitudinal study to examine the association between phthalate exposure and anthropometric measures of children.
Our results offer interesting comparisons with those from two cross-sectional NHANES studies that examined the urinary concentrations of several phthalates and body size (Stahlhut et al. 2007
;Hatch et al. 2008
). Consistent with findings in adolescents and older women (Hatch et al. 2008
), we found a positive association between MEP concentrations with both BMI and WC. Although our findings were not significant, the magnitude of associations were comparable to Hatch, who reported a difference of about 2 units in BMI for adolescent girls: 22.9 to 24.7 kg/m2
with a 2-3 μM change in MEP (quartile 1-4; 110-694 μg/gC). We observed a similar increase in our high BMI girls (21 to 23 kg/m2
) for a 1-3 μM change. For waist circumference, they found an increase of about 4 cm (77.4 to 81.6 cm) with a 1-3 μM change in MEP (quartile 1-4); we also report about a 5 cm difference in our high BMI girls (see and ). However, among the most comparable NHANES female age group (children ages 6-11 years), no phthalate metabolite-body size characteristic associations were observed in boys (N=329) or girls (N=327) (Hatch et al. 2008
). We saw no significant associations in boys, whereas among older NHANES males, positive associations between the concentrations of some phthalate metabolites and body size measures were observed, both in adolescents and adults (Stahlhut et al. 2007
;Hatch et al. 2008
Our failure to see other similar relationships with those found in NHANES may be due to differences in the distribution of the exposure, assessed from the magnitude and range of the urinary concentrations, and body size measures resulting from the lack of similarity between the populations studied. Our sample had a narrow range of exposures (ex. MEP range in quartiles: 67-948 μg/gC) and outcome (ex. our percentage of girls in the overweight and obese categories compared to the general population 40.2% vs. 32.6%). Furthermore, associations seen in Stahlhut and Hatch were not monotonic increases. Reports of nonmonotonic (e.g., U-shaped) dose-response relationships, ultra-low dose effects and nonthreshold effects for endocrine disrupting chemicals continue to challenge some of the basic assumptions of toxicology and risk assessment (Hotchkiss et al. 2008
). It is possible that these low level exposures could be spurious associations due to residual confounding (see below discussion).
With biomarkers of phthalates having limited range of exposure, population-based quartiles may not create a sufficient gradient as biomarker concentrations are low (95%<1 μM) (, Supplemental Table 1
). The exposure range exhibited in our children may have been too low to elicit biologic effects. Population quartiles are assumed to represent increasing dose, when it is possible that there is no real biologic difference in concentration across the low quartiles. We created biologically relevant exposure categories based on molar concentrations to provide a possible parallel with biological doses in experimental studies. In our study and NHANES, few individuals have biomarker levels above 1 μM.
Among Danish children aged 4 to 9 years, the only associations observed were inverse phthalate-height among boys and girls regardless of the metabolite examined (Boas et al. 2010
); urinary phthalate metabolite concentrations among the Danish were fairly similar to those reported for the NHANES children except that MEP was lower, (median MEP for girls was 36 μg/gC and boys 31 μg/gC whereas NHANES median MEP for girls was 137.7 μg/gC and 93.9 μg/gC for boys). We observed significant inverse height associations in girls for MBP and MCPP when using the log-linear model. The Danish study did not examine MCPP, a non specific metabolite of high molecular-weight phthalates and a minor metabolite of di-n-butyl phthalate. In our girls when we created cutpoints based on μM, there were no individuals with medium and high concentrations of MCPP or MBP. Inverse associations across a low concentration range may reflect increased random variation (noise) at the lower end of the detection range and may not be biologically meaningful. Another possibility is residual confounding by creatinine may underlie these inverse associations between body size measures; creatinine correction on adjustment of the urinary concentrations of the phthalate biomarkers may not be adequate or may introduce error. As a possible improved approach, we used the residual method to adjust for creatinine (Willett 1998
). Results were similar to the main models implying that creatinine itself does not appear to account for these inverse associations (not shown). In girls, the inverse height relationships were observed primarily in normal-weight but not in overweight girls, possibly due to a general decline in muscularity and glomerular filtration rate (Engel et al. 2007
). Excretion of creatinine is related to muscle mass, and thus in children strongly correlated with age and anthropometric measures such as height and weight (Skinner et al. 1996
). Therefore, creatinine-corrected phthalate metabolite urinary concentrations would tend to decrease as age and body size increase, which could create a negative association between phthalates and anthropometric measures (Boas et al. 2010
). Other investigators have utilized specific gravity to normalize urinary dilution; it was not available to us. However, specific gravity and creatinine are highly correlated, and thus similar relationships may hold (Carrieri et al. 2001
The biological basis of increased body size with phthalate exposure rests most strongly on their anti-androgenic effects, which have been observed in both epidemiologic and animal research. The best evidence, on testicular and gonadal toxicity, suggests reduced testosterone levels (Parks et al. 2000
;Pan et al. 2006
). A possible mechanism supported by clinical and epidemiologic research may be the role of androgens in regulation of adipose tissue distribution (Mayes and Watson 2004
). The observed association between MEP and increased body size among the overweight and obese girls is consistent with the idea that obesity is associated with alterations in androgen secretion, metabolism, and action (Pasquali 2006
). Differences in hormonal makeup between the sexes (Wajchenberg 2000
) support the findings that some of the effects of phthalates on body size differ among girls and boys. We could not examine boys in great detail as >50% of our boys were overweight/obese and our sample size was small. Other emerging mechanistic pathways that implicate phthalates and obesity include thyroid hormone (Meeker et al. 2007
) and peroxisome proliferator-activated receptors (PPARs) (Bility et al. 2004
;Feige et al. 2007
). These hormonal pathways may affect adipocyte formation and/or function (Feige et al. 2007
;Casals-Casas et al. 2008
;Boberg et al. 2008
The overweight/obese stratum-specific associations could be due to the presence of higher phthalate metabolite concentrations in larger children, if phthalate metabolism differed by body size or if greater exposure to phthalates occurred through increased dietary consumption. Mean MEP and ΣDEHP urinary concentrations did not differ by body size in our sample, although it is possible that phthalate metabolites in obese children lengthen exposure time. The absence of statistically significant associations for the majority of the phthalate metabolites may be due to a lack of power given the size of our study population; however, the small magnitude of the non-significant regression coefficients may also indicate that there is no association between body size and exposure to these particular analytes.
A strength of this study is the availability of detailed anthropometric measures specific to adiposity. BMI and WC have been studied much more extensively than any of the other body size measures and have been shown to be risk factors for cardiovascular diseases in adults and a means of identifying children susceptible to developing cardiovascular disease in adulthood (Freedman et al. 2007
). However, body fat distribution, particularly an increase in abdominal adipose tissue, is associated more strongly with cardiovascular disease risk than is BMI (Goran and Gower 1999
;Maffeis et al. 2003
). Several studies in children have shown WC, a measure of visceral fat deposits, to be associated with cardiovascular disease risk independent of BMI (US Department of Health and Human Service 2000
;Janssen et al. 2002
). We found BMI, WC and percent body fat all to be highly correlated with each other (radj
>.90). An additional strength is that we have longitudinal analyses as body size measures were obtained one year after phthalate exposure was assessed. The availability of a single spot urine could be considered a limitation because urinary phthalate metabolite concentrations are short-term biomarkers. However, sources of phthalate exposures are likely to be reasonably steady because patterns of use for personal care products, diet and other daily activities are relatively constant. In fact, we undertook this study only after showing that the temporal stability of urinary concentrations of phthalate metabolites was sufficient to categorize exposure level based on a single urine sample over at least a 6 month period (Teitelbaum et al. 2008
). However, it may not represent exposure earlier in life during more susceptible periods.
We and others find associations of MEP and low MWP but not other metabolites with obesity, which raises several questions. MEP is not active in classic androgen insufficiency bioassays, while other phthalates including the MBP precursor DBP exhibit a range of activity (e.g., five-fold for testicular testosterone depletion). As noted in the NAS report (National Academy of Sciences 2008
), MEP in epidemiologic research may be a surrogate for other coexisting phthalates; this is consistent in our data with the correlation of MEP and MBP (rS
0.23 for girls, 0.32 for boys). In our children, MEP and MBP represent 68% and 22% of low MWP among girls, but the range is wide (IQR 52-84% and 9.5-51%, respectively). Thus MBP, if more potent, might influence the association for both MEP and low MWP with obesity. This further raises the neglected area of mixed, multiple exposures, a problem that low MWP and high MWP molar sums do not solve. Other statistical methods, including factor analysis and principal components analysis, could be unduly influenced by MEP. Thus a more useful approach incorporating differential potency may be needed for phthalate exposures. This method would be similar to the toxic equivalent factors used for dioxin-like exposures or the functional toxicity suggested by McLachlan specifically for hormonally active agents (McLachlan 1993
). Such factors do not yet exist for phthalates, and they might include more than one (National Academy of Sciences 2008
). The approach has been examined in detail for androgen insufficiency in experimental data (National Academy of Sciences 2008
). Finally, the associations of low MWP with obesity in our and other studies may also be due to differences in pharmacokinetics, distribution, and excretion of different phthalate monoesters. For example, MEP may represent more recent and MBP less recent exposure, and either metabolite or their sums may misclassify exposure. As discussed recently using DEHP metabolites, metabolites with longer or shorter elimination half-lives may underestimate or overestimate exposure over an interval (“near” vs. “distant”) (Lorber et al. 2007
). Taking these factors into account, in addition to biologic potency and dilution correction (see (McLachlan 1993
)), could improve risk estimates using exposure biomarkers.
Phthalates are increasingly being removed from products used by children (US Consumer Product Safety Commission 2010
;The European Parliament and the Council of the European Union 2010
) mainly DEHP in the U.S. and dibutyl phthalate in the European Union. Phthalates are found in common consumer products such as perfumes, medications, deodorants, nail polish, shampoos, hair sprays, and cosmetics as well as vinyl household products such as floor covering and shower curtains. A major question among health and environmental scientists is whether phthalates are safe enough for use in common products that are not tested for safety (Koo et al. 2002
). In 2008, the FDA determined that there was insufficient evidence upon which to take regulatory action (US Food and Drug Administration 2010
). The current evidence is not sufficient for attributing a causal effect for phthalates on increased body size, but the urinary concentrations of some phthalate metabolites among children and the epidemic of obesity may merit voluntary reductions in exposure where possible.