Of the 478 men with phthalate metabolites measured in urine, 422 had free T4, total T3, and TSH levels measured in serum. An additional 14 subjects taking hormone medications (e.g., propecia, finasteride, cabergoline, clomid, gonadotropin-releasing hormone, testosterone, prednisone taper) were excluded from the present analysis. In addition, none of the men reported taking medications that may alter the thyroid axis (i.e., amidoarone, carbamazepine, chlorpropamide, carbidopa/levodopa, heparin, interferon, lithium, phenytoin, phenobarbital, propylthiouracil, sulfasalazine, synthroid). Among the remaining 408 subjects (), most were white (85%) and had never smoked (72%). The mean (± SD) age and BMI were 36 ± 5.3 years and 28 ± 4.5, respectively. Distributions of SG-adjusted phthalate metabolite concentrations are presented in , and distributions of the thyroid hormones and TSH measured in serum are presented in . Among the 408 urine samples, MEP, a metabolite of diethyl phthalate, was detected in 100% of the samples, whereas MBP and MBzP were detected in > 97% and > 94% of the samples, respectively. Of the samples, 83% had detectable concentrations of MEHP. The sample size for MEOHP and MEHHP was 208 because analytical methods for the quantification of these analytes were only recently implemented in this study. More than 95% of these samples had detectable concentrations of MEHHP and MEOHP. Spearman correlations between MEHP and MEHHP or MEOHP were 0.74 and 0.71, respectively.
Subject demographics (n = 408).
Distribution of SG-adjusted phthalate metabolites in urine (ng/mL).
Distribution of thyroid hormones and TSH in serum (n = 408).
Age was inversely associated with both free T4
and total T3
(Spearman correlation coefficients were –0.2 and –0.1, respectively; p
< 0.05 for both), whereas there were suggestive positive weak associations of BMI with T3
, TSH, MBzP, and both oxidative DEHP metabolites (all Spearman correlation coefficients were 0.1; all p
-values were between 0.05 and 0.1). Current smokers had higher median T3
levels (1.04 ng/mL) and lower median TSH (1.1 μIU/mL) than never-smokers (0.96 ng/mL and 1.5 μIU/mL, respectively). Smoking status was not associated with T4
. Samples collected in winter had median T4
concentrations slightly lower than those collected in spring, summer, or fall (1.1 vs. 1.2 ng/dL), and median TSH was higher in samples collected in the morning compared with samples collected in the afternoon (1.5 vs. 1.4 μIU/mL). For the SG-adjusted phthalate metabolites, current smokers had higher median concentrations of MEP (215 ng/mL) but lower concentrations of MEHHP and MEOHP (21 ng/mL and 16 ng/mL, respectively) compared with never-smokers (140, 45, and 32 ng/mL respectively). As previously observed (Silva et al. 2004a
), median MEHP concentrations were also higher among men whose urine samples were collected in the afternoon (9.4 ng/mL) compared with men who provided urine samples in the morning (6.9 ng/mL).
Crude regression results were similar to the adjusted results, with the exception of MEP, where there was a suggestive inverse association with total T3 (p = 0.07) and a suggestive positive association with TSH (p = 0.1) that were no longer evident when covariates were included in the models (p = 0.2). Results from the multivariate regression analyses are presented in . All models were adjusted for age, BMI, smoking, and the time of day blood/urine samples were collected. We found an inverse association between SG-adjusted urinary MEHP concentration and serum total T3 levels, where an IQR increase in MEHP was associated with a 0.021-ng/mL decrease in T3 [95% confidence interval (CI), –0.042 to –0.001 ng/mL; p = 0.04]. For the median level of T3 (0.96 ng/mL), this represents a 2.2% decrease in T3 for an IQR increase in MEHP (3.16–21.3 ng/mL). In sensitivity analyses, effect estimates from the multivariate models were similar when men with SG outside the acceptable range were excluded (n = 339; results not shown).
Adjusteda regression coefficients (95% CI) for thyroid hormones associated with an interquartile range (IQR) increase in SG-adjusted urinary phthalate metabolite concentrations (n = 408).
When the analysis was limited to the subset of men with oxidative DEHP metabolite measures (n = 208), the inverse association between MEHP and T3 became weaker. An IQR increase in MEHP was associated with a 0.011-ng/mL decrease in T3 (95% CI, –0.021 to 0.009 ng/mL; p = 0.4). However, there was an inverse association between MEHP% and free T4 (). An IQR increase in MEHP% was associated with a 0.030-ng/dL decrease in free T4 (95% CI –0.055 to –0.005 ng/dL; p = 0.02). For the median T4 level (1.2 ng/dL), this represents a 2.5% decrease in T4 for an IQR increase in MEHP% (6% to 17% MEHP). When both MEHP% and MEHP were included in the models, the inverse association between MEHP% and T4 remained (). When MEHHP and MEHP were both included in the models, T4 was inversely associated with MEHP but positively associated with MEHHP. There was no evidence of collinearity between MEHP and MEHHP (i.e., the SEs and 95% CIs were not inflated). Results were identical when MEOHP was included in the models instead of MEHHP (data not shown), because MEHHP and MEOHP concentrations were highly correlated. Results for T3 followed the same pattern as free T4 when both MEHP and MEHHP were included in the model, although associations were weaker and not statistically significant (). We explored the interaction terms (MEHP × MEHHP or MEHP × MEHP%), but we found no evidence of interaction when these terms were added to the multivariate models.
Adjusteda regression coefficients for a change in thyroid hormones associated with an IQR increase in MEHP when also adjusted for MEHHP or MEHP% (n = 208).
To assess the robustness of the associations and potential nonlinear relationships, we categorized SG-adjusted phthalate concentrations into quintiles (n = 408). We found a suggestive inverse trend for MEHP quintiles and T3 (p = 0.07), whereas we unexpectedly found a significant inverse trend for MEHP quintiles and free T4 (p = 0.04). The regression coefficients for increasing quintiles of MEHP appeared to plateau at quintile 4 ( and ). Among the subset of men with MEHP oxidative metabolite measures (n = 208), metabolite concentrations were categorized by tertiles because of the smaller sample size. The tertile analysis resulted in a suggestive inverse trend between MEHP% and free T4 (p = 0.07; ). When tertiles of both MEHP and MEHHP were included, there was an inverse association between free T4 and medium and high tertiles of MEHP compared with the lowest MEHP tertile (). Similar to the quintile analysis among all the men, the relationships did not appear to be linear but were suggestive of having a plateau. Results were similar when MEHP and MEHP% were included in the same model (). In addition, the suggestive association between MEHP% and free T4 was no longer evident.
Regression coefficients (95% CIs) for a change in total T3 associated with increasing quintiles of SG-adjusted MEHP. Adjusted for age, BMI, smoking, and time of day (n = 408).
Regression coefficients (95% CIs) for a change in free T4 associated with increasing quintiles of SG-adjusted MEHP. Adjusted for age, BMI, smoking, and time of day (n = 408).
Regression coefficients (95% CIs) for a change in free T4 associated with increasing tertiles of MEHP%. Adjusted for age, BMI, smoking, and time of day (n = 208).
Regression coefficients (95% CIs) for a change in free T4 associated with increasing tertiles of MEHP and MEHHP. Adjusted for age, BMI, smoking, and time of day (n = 208).
Regression coefficients (95% CIs) for a change in free T4 associated with increasing tertiles of MEHP and MEHP%. Adjusted for age, BMI, smoking, and time of day (n = 208).
We compared concentrations of (unadjusted) urinary phthalate metabolites measured in the present study with those among U.S. males published in the Third National Report on Human Exposure to Environmental Chemicals
). Metabolite distributions were generally similar between the present study and the national data, although we found slightly lower concentrations of MEP, MBP, and MBzP. Conversely, urinary concentrations of MEHP, MEHHP, and MEOHP were somewhat higher in the present study than those from the Third National Report. For example, the median and 95th percentile values for MEHP (unadjusted for SG) in the present study were 6.3 and 112 ng/mL, respectively, compared with 4.3 and 37.9 ng/mL in the Third National Report (CDC 2005