On the whole, urine phthalate metabolite concentrations in this study were comparable to those from previous studies conducted in other countries (Table S8 in Supporting Information File S1), although the profiles of urinary phthalate metabolites demonstrated some differences in terms of quantity and species of metabolites. For example, MBzP was detected only in 38.6% of subjects in this study, but detected in almost 100% in the studies of U.S., German, and Spanish children 
. These differences may primarily reflect regional differences in use of phthalates. Additionally, different race and urine type (first morning urine in this study) may contribute to these differences as well.
As a result of an extensive use of phthalate-related products, human can be exposed to phthalates through almost all possible pathways, such as ingestion, inhalation, dermal contact, and placenta transmission 
. Generally speaking, ingestion of food is the major exposure route of phthalates, in particular for high molecular weight phthalates, such as DEHP and BBzP 
. For low molecular weight phthalates, such as DBP, DiBP, and DMP, inhalation of indoor air or dermal contact may be relevant exposure pathways next to food ingestion 
. For DEP, one of low molecular weight phthalates, the major exposure route may largely be through dermal contact (especially by use of personal-care products) 
. It was expected that exposure pathway-related variables, such as caloric intake and food composition related to food ingestion, and daily indoor time related to inhalation of indoor air, were positively associated with corresponding phthalate exposures, and then urine phthalate metabolites. Sum of DEHP and MCHP may be proxies for food ingestion-related variables, such as caloric intake and food composition, MMP may be for daily indoor time, and MEP may be for use of personal-care products, and sum of DBP or MiBP may be related to both food ingestion and daily indoor time.
In the present study we found nine urine phthalate metabolites and five sums were positively associated with BMI or WC in Chinese school children when adjusted for individual SG, age, and sex (in adjusted model 1), but most of these significant associations disappeared and some even reversed when further adjusted for other urine phthalate metabolites (in adjusted model 2). Only MEHP and MEP remained significant. Close correlations among phthalate metabolite measurements may be one reason for these association changes (Table S4 in Supporting Information File S1), and these changes may mean other possibilities. One possibility was that some phthalate metabolites might act as confounders. When these metabolites were included as covariables, they corrected the associations of metabolites of interest with BMI or WC. Another possibility was that some phthalates might have modifying effects on BMI or WC 
. A simple adjustment might diminish true associations of metabolites of interest with BMI or WC. So it was expected that a still significant association might be more reliable after adjustment for other urine phthalate metabolites.
Given the cross-sectional nature of this study, the possibility of reverse causation cannot be ruled out. It is possible that increased overall caloric intake, different diet composition, and longer daily indoor time for obese individuals might increase the risk of phthalate exposures, compared to normal weight ones 
. These variables related to obesity were also confounders for the relationships between phthalate exposures and obesity. Lack of information on these obesity-related variables was an important limitation for this study. Fortunately, since urine phthalate metabolites could be proxies for these confounders as discussed above, the inclusion of urine phthalate metabolites as covariables in multiple linear regression model (adjusted model 2) corrected potential effects of these confounders unavailable in this study to some extent. Therefore, it should be more valid for the presence of significant associations of MEHP and MEP with BMI or WC after adjustment for age, sex as well as other urine phthalate metabolites.
An underlying assumption in this study and other similar studies is that one single measurement of spot urine could represent a long-term exposure of corresponding participant to no-persistent environmental chemicals 
. Given short biologic half-lives of phthalate metabolites (12 hours or less), multiple sources and exposure routes of phthalates, varying diet composition and behavior of participants, urine collection time in a day and urine dilution, one single measurement of spot urine is not likely to perfectly represent a long-term exposure 
. Some researches assessed the prediction of single spot urine for a long-term exposure and reported to have a moderate predictabilityapp:ds:power for long-term exposure and this ability differed by phthalate metabolites 
. For example, Hauser et al. reported that single spot urine was moderately predictive of each subject’s exposure tertile categorization over three months with the sensitivities ranging from 0.56 to 0.74, and most predictive for MEP and least predictive for MEHP among five metabolites in 11 adult men 
. Teitelbaum et al. assessed the variability of eight creatinine-corrected phthalate metabolites over six months and found moderate reproducibilities with intraclass correlation coefficients (ICCs) ranging from 0.21 (mono(3-carboxypropyl) phthalate, MCPP) to 0.62 (MBzP) in 35 children 6–10 years of age 
. Additionally, misclassification due to one measurement of spot urine might result in an attenuated estimate of association of phthalate metabolite exposures with BMI or WC, which meant that significant associations based on spot urine were conservative 
. In this study, we used first morning urine instead of spot one. Because the urine collection time is similar in a day for first morning urine, it decreases the variations of urine phthalate metabolite concentration. This strengthened the predictive ability of one measurement for long-term exposure than pure spot urine 
In our studies, we observed some age- or sex-related differences in the associations of phthalate metabolites with BMI or WC. MEHHP and MEOHP were found to be significantly associated with BMI only in the 8–11 year age group, not in all the subjects. These age- or sex-related differences were also reported in some previous studies 
. These differences might be related to different status of sex hormone endocrine of participants between sex or age subgroups (8–11 or 12–15 year age group). Puberty usually begins between 10 and 13 years of age, and serum levels of endogenous sex hormone in puberty significantly rise, especially for estradiol in girls and testosterone in boys. The differences in sex hormone function may lead to different susceptibility of participants to sex hormone perturbation caused by phthalate exposure 
. More accurate measures of sex hormone levels and puberty status are needed to further explore the biologic mechanisms for the interrelationship among sex hormone, phthalate exposures and BMI.
There were some epidemiologic studies supporting the associations of postnatal phthalate exposures and body size measures in children. Using data from the National Health and Nutrition Examination Survey (NHANES) 1999–2002, Hatch et al. found that BMI increased with urinary MEP quartiles and inversely related to MEHP quartiles in girls 12–19 years of age, but no metabolite quartiles were significantly associated with BMI in the 6–11 year age group 
. Boas et al. observed that some metabolites, such as MEP, MEOHP, MECPP, and sum of DEHP metabolites, negatively correlated with BMI in 845 Danish children 4–9 years of age although this study was primarily designed to investigate the association of urine phthalate metabolites and thyroid function, insulin-like factor 1, and growth 
. Teitelbaum et al. reported no significant associations of urinary nine phthalate metabolites at baseline with BMI or WC measured one year later in 387 Hispanic and Black children 6–8 years of age in New York City; however, after divided into normal weight and overweight groups, the data showed significant associations of MEP and sum of low molecular-weight (MEP, MBP, and MiBP) with BMI and WC in overweight children 
. Our study results were comparable to those from Teitelbaum et al. and Hatch et al., except for an inverse association of MEHP with BMI reported by Hatch et al., but distinctly different from those by Boas et al.
There are some plausible explanations for these discrepancies. Firstly, creatinine correction for urine phthalate metabolite concentrations can weaken or even reverse the association due to the positive associations of urine phthalate metabolite concentrations and BMI with urinary creatinine 
. In this circumference, SG correction may be a better choice 
. The studies conducted by Teitelbaum et al. and Hatch et al. used creatinine to correct the urine dilution. Secondly, some studies including the study by Boas et al. reported that phthalates, such as DEHP and DiNP, might suppress thyroid hormone levels and insulin-like growth factor I (IGF-I) in serum 
. This suppression would decrease the growth of children, and therefore BMI values. When this trend exceeded the adipogenesis effect caused by phthalates, phthalate exposures would be negatively associated with BMI. Because younger children experience more rapid growth, their growth is more likely to be affected by external disturbance. This might explain the negative associations of phthalate exposures with BMI in the study of younger study population by Boas et al. 
. There are some mice studies that have shown negative associations between prenatal phthalate exposures, such as DBP, BBP, and DEHP, and offspring weight 
. If these explanations are reasonable, DEP, DEHP, DBP, and DiBP are main phthalates associated with BMI or WC.
In this study we found nine urinary phthalate metabolites and three sums were positively associated with BMI or WC in Chinese school children, and it was more evident for MEHP and MEP. Some associations showed sex and/or age related difference. Given the cross-sectional nature of this study and lack of some important obesity-related variables, the associations discovered here need to be further investigated.