Phthalates and Sex Steroid Hormones Among Men From NHANES, 2013–2016

Abstract

Context

Phthalates are commonly found in commercial packaging, solvents, vinyl, and personal care products, and there is concern for potential endocrine-disrupting effects in males. The commonly used di-2-ethylhexyl phthalate (DEHP) has progressively been replaced by seldom studied compounds, such as bis-2-ethylhexyl terephthalate and 1,2-cyclohexane dicarboxylic acid di-isononyl ester (DINCH).

Objective

To investigate the associations between the urinary phthalate metabolites and serum sex steroid hormone concentrations in a nationally representative sample of adult males.

Design, Setting, Participants, and Intervention

This was a cross-sectional analysis of data from the 2013–2016 National Health and Nutrition Examination Survey among 1420 male participants aged ≥20 years.

Main Outcome Measures

Serum levels of total testosterone, estradiol, SHBG, and derived sex hormone measurements of free testosterone, bioavailable testosterone, and free androgen index were examined as log-transformed continuous variables.

Results

Phthalate metabolites were not statistically significantly associated with sex hormone concentrations among all men. However, associations varied by age. High molecular weight phthalates were associated with lower total, free, and bioavailable testosterone among men age ≥60. Specifically, each doubling of ΣDEHP was associated with 7.72% lower total testosterone among older men (95% confidence interval, -12.76% to -2.39%). Low molecular phthalates were associated with lower total, free, and bioavailable testosterone among men age 20 to 39 and ∑DINCH was associated with lower total testosterone among men age ≥40.

Conclusions

Our results indicate that males may be vulnerable to different phthalate metabolites in age-specific ways. These results support further investigation into the endocrine-disrupting effects of phthalates.

Phthalates are a diverse class of synthetic chemicals present in cosmetics, personal care products, and food packaging (1). They are used as softeners in polyvinyl chloride (PVC) and other plastics, and as lubricants and fragrances (2). Widespread use of phthalates in this diverse array of contexts results in pervasive human exposure (2, 3).

Evidence from both experimental and observational studies supports a connection between exposure to phthalate esters and endocrine function, particularly steroid hormone dysregulation (4). Prior evidence suggests that phthalates may have antiandrogenic effects, such as reduced testosterone production and bioavailability, decreased sperm production and motility, shortened anogenital distance, and increased odds of genital anomalies (3–9). In animal studies, phthalates have also been shown to alter the expression of genes encoding enzymes responsible for testosterone biosynthesis through PPAR-α activation (10). These PPARs are also known to downregulate nuclear receptors involved in testis development, such as estrogen receptors, providing possible insight into the mechanism of phthalate-induced endocrine disruption.

Decrements in testosterone function have broad consequences across the life course. In newborns and children, lower testosterone levels have been associated with a range of genital anomalies, such as hypospadias, hydrocele, and decreased anogenital distance in infant boys (6–9, 11). In adult men, low testosterone has been documented to be a risk factor for several chronic health conditions: metabolic syndrome (12), diabetes (13), neurological and cognitive functioning (14), bone loss (15), cardiovascular disease (16, 17), and premature mortality (17–19). One study estimated that as many as 10,700 deaths annually among US men may be attributable to phthalate-induced decreases in testosterone (20). These broad health sequelae of decreased testosterone reinforce the need to identify preventable causes of male hypogonadism, including environmental exposures to possible endocrine disrupting chemicals.

Research showing the endocrine-disrupting effects of common phthalates has led to increased regulation of some phthalates and consequent efforts to establish replacement chemicals (21). Specifically, in response to concerns about di-2-ethylhexyl phthalate (DEHP), manufacturers have introduced replacement chemicals such as diisononyl phthalate (DINP), di-isobutyl phthalate, diethyl phthalate, and polyethylene terephthalates (PET), including bis-2-ethylhexyl terephthalate (DEHTP), and 1, 2-cyclohexane dicarboxylic acid di-isononyl ester (DINCH) (22). PETs are commonly used in food packaging, such as water bottles, polycarbonate containers, and ready-to-eat meals (23–25). However, the health effects of these substitutes have not been comprehensively studied, and in fact, DEHTP and DINCH metabolites were measured for the first time in the National Health and Nutrition Examination Survey (NHANES) in 2015–2016 (26). As industries shift which chemicals are used, population-level exposure patterns change, and more people are likely to be exposed to phthalates about which there is little known (27).

Using data from the 2011–2012 NHANES, Meeker and Ferguson identified decreases in serum testosterone in relation to increases in urinary phthalate metabolite levels, specifically those of DEHP, among older men (28). We sought to update and extend their analysis to examine possible associations between phthalates, including these newer PET metabolites, and sex steroid hormone levels using more recent NHANES data from 2013–2016 (29). Specifically, we evaluated cross-sectional associations between urinary phthalate metabolite groupings (i.e., metabolites of DEHP, DINP, DINCH, DEHTP, low molecular weight [LMW] phthalates, and high molecular weight [HMW] phthalates) and serum sex steroid hormone concentrations (ie, total testosterone, SHBG, and estradiol) as well as derived endocrine outcome measures (ie, free testosterone [FT], bioavailable testosterone [BAT], testosterone/estradiol [T/E2] ratio, and free androgen index [FAI]) among adult men. Furthermore, we examined whether these associations varied by age.

Methods

Study sample

NHANES is a nationally representative survey of the noninstitutionalized US population, released in 2-year cycles. This analysis used the 2013–2014 and 2015–2016 releases. The study population was restricted to male respondents age ≥20 years who were not taking sex hormone medication including testosterone, progesterone, estrogen, or “other sex hormones” noted in the NHANES questionnaire, and had measured values for urinary phthalate metabolite concentrations (ie, NHANES environmental subsample B), as well as serum sex hormone measures.

Exposures

NHANES laboratories used spot urine samples to assess phthalate metabolite levels. Laboratory methods for this assessment have been described in detail elsewhere (30). According to NHANES guidelines, metabolites were grouped by parent phthalate molecule and by the weight of the parent molecule. Micromolar sums of individual metabolites were calculated for the following parent phthalates: DEHP, DINP, DINCH, and DEHTP; and were grouped by weight into LMW phthalates and HMW phthalates. Table 1 lists the individual metabolites assessed and how they were grouped for our analysis. Metabolites that were only available for the 2015–2016 sample and those that were below the lower limit of detection (LLOD) in >50% of the sample are noted. Measures below LLOD were imputed by $LLOD/\surd 2$ and included in the molar sums.

Table 1.

Phthalate metabolites measured in NHANES, 2013–2016

Phthalate Metabolite	Grouping	LLOD (ng/mL)
Mono-ethyl phthalate (MEP)	LMW	1.2
Mono-isobutyl phthalate (MIBP)	LMW	0.8
Mono-2-methyl-2-hydroxypropyl phthalate (MHIBP)a	LMW	0.4
Mono-n-butyl phthalate (MBP)	LMW	0.4
Mono-3-hydroxybutyl phthalate (MHBP)a	LMW	0.4
Monobenzyl phthalate (MBZP)	HMW	0.3
Mono(2-ethylhexyl) phthalate (MEHP)	DEHP/HMW	0.8
Mono(2-ethyl-5-carboxypentyl) phthalate (MECPP)	DEHP/HMW	0.4
Mono(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP)	DEHP/HMW	0.4
Mono(2-ethyl-5-oxohexyl) phthalate (MEOHP)	DEHP/HMW	0.2
Mono-isononyl phthalate (MNP)b	DINP/HMW	0.9
Monocarboxyoctyl phthalate (MCOP)	DINP/HMW	0.3
Mono-oxo-isononyl phthalate (MONP)a	DINP/HMW	0.4
Monocarboxy-isononly phthalate (MCNP)	HMW	0.2
Mono (3-carboxypropyl) phthalate (MCPP)	HMW	0.4
Cyclohexane-1,2-dicarboxylic acid mono(hydroxy-isononyl) ester (MHINCH)b	DINCH/HMW	0.4
Cyclohexane-1,2-dicarboxylic acid mono(carboxyoctyl) ester (MCOCH)a,b	DINCH/HMW	0.5
Mono(2-ethyl-5-hydroxyhexyl) terephthalate (MEHHTP)a	DEHTP/HMW	0.4
Mono(2-ethyl-5-carboxypentyl) terephthalate (MECPTP)a	DEHTP/HMW	0.2

Abbreviations: DEHP: di-2-ethylhexyl phthalate; DEHTP: dioctyl terephthalate; DINCH: cyclohexane-1,2-dicarboxylic acid diisononyl ester; DINP: diisononyl phthalate; HMW, high molecular weight; LLOD, lower limit of detection; LMW, low molecular weight.

^aOnly available for 2015–2016 data.

^bDetected in <50%.

Outcomes

Total testosterone, estradiol, and SHBG were directly measured in serum samples according to laboratory methods following the National Institute of Standards and Technology’s guidelines (31). FT, BAT, and FAI were calculated based on hormone concentrations and serum albumin levels. Albumin concentrations were assessed following a bichromatic digital endpoint method in examined NHANES participants. FT, BAT, and FAI were calculated according to the Vermeulen et al. formula as described by Ho et al. and shown here (32, 33). Resulting FT and BAT estimates were converted from nmol/L to ng/dL for consistency with the total testosterone measure provided by NHANES. T/E2 ratio was calculated after converting testosterone and estradiol measures to the same units.

$${\displaystyle \begin{array}{l}{\displaystyle T_{free}=\frac{T_{total}-N-S+\sqrt{{(N+S-T_{total})}^2+4N{T}_{total}}}{2N}}\end{array}}$$

$${\displaystyle \begin{array}{l}{\displaystyle T_{bioavailable}=N\ast T_{free}}\end{array}}$$

$${\displaystyle \begin{array}{l}{\displaystyle FAI=T_{total}\ast \frac{100}{S}}\end{array}}$$

Where $N=0.5217\ast Albumin+1$ and $S=SHBG$

Total testosterone is a measure of all the testosterone in the blood at a given time, whereas FT is testosterone not currently bound to albumin or SHBG. BAT represents the readily available testosterone in the body at a given time point, theoretically combining the amount of free and albumin-bound testosterone and providing deeper insight into the sex hormone profile (34).

Statistical analysis

Data from the 2 cycles (2013–2014 and 2015–2016) were combined using the appropriate weighting methods (35). Following NHANES analytic guidelines, environmental subsample B sample weights were used in all analyses (36). The proper protocol for variance estimation was followed as per the guidelines for NHANES’s complex survey design, using the provided primary sampling unit and strata information (35). We conducted weighted linear regression to estimate the relation between grouped urinary phthalate concentrations and individual serum sex hormone levels controlling for various covariates: age, creatinine, poverty-to-income ratio (PIR), education, race/ethnicity, obesity, and time of day serum was sampled. Urinary creatinine was included as a covariate to account for differences in urinary dilution. PIR was divided into quartiles based on the study sample (first quartile: <1.15, second: ≥1.15 to <2.18, third: ≥2.18 to <4.16, and fourth: ≥4.16). Obesity was based on body mass index categories created from National Institutes of Health published standards (37, 38). Finally, because of the known diurnal rhythms of serum total testosterone, FT, and BAT, we controlled for the time of day each participant’s serum was sampled (39).

All hormone measures were natural log transformed to approximate a normal distribution. In addition, molar sums of phthalate metabolites were natural log transformed to reduce the influence of extreme outlier exposure values. To aid in interpretation, beta-coefficients were transformed to represent the percent change in a given hormonal outcome in relation to a doubling (i.e., 100% increase) in phthalate concentrations. Linear regression models were fit with individual phthalate groupings and each sex steroid hormone measure. Finally, models were fit again stratified by age groups 20 to 39, 40 to 59, and ≥60 years, controlling for the same covariates as the overall model. Because of the multiple comparisons, we adjusted the P value based on a modified Bonferroni approach (40–42). In short, despite 7 hormonal outcomes (3 measured and 4 calculated), the 4 that were calculated (i.e., FT, BAT, T/E2, and FAI) were derived from the measured hormones (i.e., total testosterone, estradiol, and SHBG); and of 6 exposures, 4 (i.e., DEHP, DINP, DINCH, and DEHTP) were subsets of HMW. Thus, we divided the conventional α = 0.05 by these 6 effective comparisons (3 outcomes × 2 exposures) = 0.0083. Last, because of the variance of hormonal measures by time of day, in addition to controlling for time of blood collection in our primary analyses, as a sensitivity analysis, we also fit all models among morning samples only. All analyses were performed in Stata 15 (College Station, TX).

Results

Table 2 shows sample characteristics for the NHANES participants included in this analysis. The total sample provided by NHANES from 2013 to 2016 was 20,146. After excluding those without urinary phthalate levels (n = 14,164), those without sex steroid hormone data (n = 2,171), females (n = 2,005), those younger than 20 (n = 374), and those taking sex hormone medication (n = 12), the resulting sample yielded 1,420 adult men who were included in the analytic sample. Of these, 488 were age 20 to 39, 459 were age 40 to 59, and 482 were age 60 or older. Most respondents in this sample were non-Hispanic white (65.29%). The median PIR was 3.0 (interquartile range, 1.6, 5.0). The median body mass index was 28.4 kg/m², and 38.10% of the sample was categorized as obese. Most of the sample (76.13%) had completed at least a high school or equivalent education.

Table 2.

Characteristics of study population, NHANES 2013–2016

Characteristic	Na	%b	Medianb	(Interquartile range)b
Total	1,420	100
Age categories (years)			47	(33, 60)
20–39	488	37.53
40–59	457	36.49
≥60	475	25.97
Race/ethnicity
Mexican American	219	9.35
Other Hispanic	154	6.29
Non-Hispanic white	534	65.29
Non-Hispanic black	289	9.74
Other/multi	224	9.33
Poverty: income (continuous)			3.0	(1.6, 5.0)
Pov:in quartiles
First	321	16.41
Second	318	20.40
Third	324	26.87
Fourth	322	36.32
Missing	135
Body mass index (kg/m2 continuous)			28.4	(24.9, 32.4)
Body mass index categories
Normal (<25 kg/m2)	382	25.86
Overweight (≥25-<30 kg/m2)	526	36.04
Obese (≥30 kg/m2)	493	38.10
Missing	19
Education
Less than high school	338	16.18
High school or greater	1,081	83.82
Missing	1
Time of day of blood draw
Morning	704	50.91
Afternoon	505	33.02
Evening	211	16.06

^aUnweighted.

^bWeighted with respect to environmental subsample B. Some percentages may not sum to 100% because of weighting and missing responses.

Median exposure levels and corresponding interquartile ranges in the total sample and stratified by age group are shown in Table 3. Exposures did not vary greatly by age across most phthalate groups except for HMW phthalates. Men age 20 to 39 had greater HMW concentrations compared with older men (median = 0.38 μmol/L vs. 0.29 μmol/L among those age 40 to 59 and 0.20 μmol/L among those age ≥60).

Table 3.

Phthalate metabolite group medians and interquartile ranges (IQR) stratified by age, NHANES 2013–2016

	All Years		20–39 Years		40–59 Years		≥60 Years
Phthalate metabolite groups (μmol/L)	Median	(IQR)	Median	(IQR)	Median	(IQR)	Median	(IQR)
Σ LMW	0.28	(0.14, 0.59)	0.32	(0.13, 0.57)	0.25	(0.13, 0.50)	0.30	(0.15, 0.76)
Σ HMW	0.29	(0.14, 0.59)	0.38	(0.18, 0.70)	0.29	(0.13, 0.61)	0.20	(0.13, 0.42)
Σ DEHP metabolites	0.07	(0.04, 0.13)	0.07	(0.04, 0.12)	0.08	(0.04, 0.15)	0.07	(0.04, 0.13)
Σ DINP metabolites	0.03	(0.02, 0.07)	0.03	(0.02, 0.07)	0.04	(0.02, 0.08)	0.03	(0.02, 0.05)
Σ DINCH metabolites	0.002	(0.003, 0.007)	0.004	(0.002, 0.008)	0.003	(0.002, 0.007)	0.003	(0.002, 0.005)
Σ DEHTP metabolites	0.06	(0.02, 0.23)	0.09	(0.03, 0.31)	0.06	(0.02, 0.20)	0.04	(0.02, 0.14)

Abbreviations: DEHP: di-2-ethylhexyl phthalate; DEHTP, dioctyl terephthalate; DINCH, cyclohexane-1,2-dicarboxylic acid diisononyl ester; DINP, diisononyl phthalate; HMW, high molecular weight; IQR, interquartile range; LMW, low molecular weight;

The results from models for associations between urinary phthalate groupings and serum hormone concentrations among the total sample showed small, nonstatistically significant decrements in testosterone (total, free, and bioavailable) and estradiol in relation to several exposure groups (Table 4). For example, a doubling of ∑LMW phthalates was associated with 2.41% lower FT (95% confidence interval [CI], -5.10% to 0.35%) and 2.91% lower FAI (95% CI, -5.45% to -0.30%), although no association with total testosterone. In contrast, DEHP was associated with 2.24% lower total testosterone (95% CI, -4.46% to 0.03), but not FT or BAT. Point estimates for all metabolite groupings and estradiol were consistently negative, as well, though not statistically significant.

Table 4.

Percent Change (95% CI) in Serum Hormone Concentrations Associated With a Doubling In Urinary Phthalate Group Concentrations, NHANES 2013–2016

	Measured Hormonal Outcomes										Calculated Hormonal Outcomes
	Total testosterone			Estradiol				SHBG			Free testosteronea			Bioavailable testosteronea			Testosterone/estradiol ratio		Free Androgen Indexa
	%	95% CI		%	95% CI		%	95% CI		%	95% CI		%	95% CI		%	95% CI		%	95% CI
Σ LMW	-0.47	-4.35	3.57	-1.91	-4.71	0.97	2.52	-1.24	6.42	-2.41	-5.10	0.35	-2.52	-5.26	0.30	1.47	-1.93	4.99	-2.91	-5.45	-0.30
Σ HMW	-1.23	-4.04	1.66	-2.00	-4.45	0.51	-1.02	-4.61	2.70	-0.96	-3.11	1.25	-1.16	-3.44	1.17	0.78	-0.15	3.13	-0.21	-2.71	2.36
Σ DEHP	-2.24	-4.46	0.03	-2.25	-4.76	0.33	-2.72	-5.77	0.43	-0.61	-2.79	1.63	-1.25	-3.44	0.99	0.01	-2.46	2.54	0.49	-2.21	3.27
Σ DINP	-1.03	-3.45	1.45	-1.12	-3.19	0.99	-1.29	-4.82	2.38	-0.48	-2.13	1.20	-0.68	-2.34	1.00	0.09	-2.25	2.50	0.26	-1.91	2.48
Σ DINCH	-1.31	-4.26	1.73	-1.19	-3.81	1.50	-3.33	-7.99	1.57	0.78	-2.51	4.19	0.51	-2.79	3.93	-0.13	-4.14	4.06	2.09	-2.30	6.67
Σ DEHTP	-0.28	-1.94	1.40	-1.01	-2.54	0.54	-0.55	-2.71	1.67	0.00	-1.41	1.44	-0.06	-1.50	1.42	0.73	-1.00	2.50	0.26	-1.46	2.02

Abbreviations: CI, confidence interval; DEHTP, dioctyl terephthalate; DEHP, di-2-ethylhexyl phthalate; DINCH, cyclohexane-1,2-dicarboxylic acid diisononyl ester; DINP, diisononyl phthalate; HMW, high molecular weight; LMW, low molecular weight.

* P ≤ 0.05.

# P ≤ 0.0083.

^aCalculated based on Vermeulen et al. formula from measured SHBG, testosterone, and albumin.

When the analyses were stratified by age, associations differed (Table 5). Among those age 20 to 39, LMW phthalates were significantly associated with lower FT (-4.76%; 95% CI, -8.62% to -0.75%) and BAT (-4.60%; 95% CI, -8.55% to -0.47%). However, among those age 40 to 59, LMW phthalates were associated with higher FT and BAT, and among those age ≥60, LMW phthalates were not associated with total testosterone, FT, or BAT. In contrast, among men age ≥60, HMW phthalates were associated with lower testosterone, FT, BAT, estradiol, and FAI. For example, for a doubling in HMW, we estimated 4.92% lower testosterone (95% CI, -9.79% to 0.21%), 4.68% lower FT (95% CI, -8.40% to -0.81%), 5.02% lower BAT (95% CI, -8.57% to -1.32%), 3.74% lower estradiol (95% CI, -7.65% to 0.34%), and 4.08% lower FAI (95% CI, -7.68% to -0.33%). Similar negative associations were noted in relation to DEHP in this age group, with the addition of lower SHBG (-4.35%; 95% CI, -8.61% to 0.11%). There were no significant associations with HMW or DEHP in other age groups.

Table 5.

Percent Change (95% CI) in Serum Hormone Concentrations Associated With a Doubling in Urinary Phthalate Group Concentrations Stratified by Age, NHANES 2013–2016

	Measured Hormonal Outcomes												Calculated Hormonal Outcomes
	Total Testosterone				Estradiol				SHBG				Free Testosteronea				Bioavailable Testosteronea				Testosterone/estradiol ratio				Free androgen indexa
	%	95% CI			%	95% CI			%	95% CI			%	95% CI			%	95% CI			%	95% CI			%	95% CI
20-39 years
Σ LMW	-3.64	-7.21	0.07		-3.04	-6.66	0.72		0.93	-2.92	4.94		-4.76	-8.62	-0.75	*	-4.60	-8.55	-0.47	*	-0.61	-4.40	3.33		-4.53	-8.88	0.02
Σ HMW	0.85	-2.31	4.11		-1.71	-5.05	1.74		-0.09	-5.17	5.26		0.55	-2.56	3.76		0.38	-2.83	3.69		2.60	-1.05	6.40		0.94	-3.35	5.41
Σ DEHP	-0.73	-3.94	2.59		0.39	-3.39	4.33		-1.90	-5.69	2.03		0.35	-2.88	3.68		-0.50	-3.72	2.83		-1.12	-4.65	2.54		1.20	-2.30	4.82
Σ DINP	-1.56	-4.63	1.62		-0.47	-3.80	2.99		-3.15	-7.58	1.50		-0.21	-2.81	2.46		-0.83	-3.38	1.80		-1.10	-3.62	1.49		1.65	-1.65	5.06
Σ DINCH	3.73	-0.23	7.85		-2.08	-6.67	2.74		2.13	-4.48	9.19		2.28	-1.70	6.43		2.17	-1.93	6.45		5.93	-0.01	12.23	*	1.57	-3.61	7.02
Σ DEHTP	1.78	0.19	3.39	*	-1.25	-3.46	1.00		1.27	-2.20	4.86		1.07	-0.62	2.79		1.19	-0.56	2.97		3.07	0.70	5.50	*	0.51	-2.39	3.49
40-59 years
Σ LMW	3.91	-0.62	8.64		-0.08	-4.22	4.24		3.52	-3.18	10.67		2.19	0.36	4.06	*	2.13	0.27	4.02	*	3.99	0.53	7.56	*	0.38	-2.27	3.10
Σ HMW	0.12	-6.11	6.77		-0.56	-3.44	3.45		0.41	-8.23	9.87		-0.21	-3.86	3.58		-0.46	-4.50	3.76		0.18	-3.87	4.40		-0.29	-5.37	5.07
Σ DEHP	0.89	-3.13	5.07		-1.20	-5.27	3.04		-0.50	-5.36	4.60		1.35	-2.51	5.37		0.98	-2.93	5.06		2.11	-1.29	5.64		1.40	-3.03	6.03
Σ DINP	1.85	-4.17	8.24		-0.56	-3.48	0.24		4.35	-3.72	13.10		-0.85	-3.57	1.94		-0.77	-3.64	2.18		2.42	-1.78	6.80		-2.40	-5.69	1.00
Σ DINCH	-3.45	-5.70	-1.14	#	0.05	-2.91	2.86		-5.76	-11.69	0.57		-0.42	-3.50	2.77		-0.38	-3.48	2.82		-3.50	-6.94	0.07		2.45	-3.35	8.60
Σ DEHTP	-1.51	-3.72	0.74		0.60	-1.58	2.82		-3.26	-6.85	0.48		0.48	-1.20	2.20		0.27	-1.53	2.12		-2.10	-3.75	-0.42	*	1.80	-0.78	4.45
≥60 years
Σ LMW	-1.08	-9.00	7.53		-2.93	-7.57	1.94		3.05	-3.96	10.58		-3.68	-9.18	2.16		-3.97	-9.87	2.32		1.91	-3.86	8.01		-4.01	-8.93	1.17
Σ HMW	-4.92	-9.79	0.21		-3.74	-7.65	0.34		-0.88	-6.28	4.83		-4.68	-8.40	-0.81	*	-5.02	-8.57	-1.32	*	-1.23	-7.27	5.21		-4.08	-7.68	-0.33
Σ DEHP	-7.72	-12.76	-2.39	#	-5.55	-9.81	-1.09	*	-4.35	-8.61	0.11		-5.17	-9.81	-0.30	*	-5.73	-10.53	-0.68	*	-2.30	-6.19	1.76		-3.53	-8.10	1.28
Σ DINP	-3.61	-8.91	2.00		-3.84	-7.46	-0.07	*	-4.49	-11.29	2.82		-0.25	-4.92	4.65		-0.50	-5.16	4.38		0.23	-5.33	6.13		0.93	-4.48	6.64
Σ DINCH	-7.87	-12.43	-3.06	#	-1.76	-5.29	1.89		-8.92	-14.38	-3.11	#	-1.52	-7.81	5.21		-2.01	-7.91	4.27		-6.21	-10.38	-1.86	*	1.15	-5.33	8.09
Σ DEHTP	-1.97	-5.81	2.04		-2.61	-5.49	0.37		0.41	-3.05	4.00		-2.42	-5.08	0.32		-2.54	-5.12	0.10		0.66	-3.48	4.97		-2.37	-4.48	-0.21

Abbreviations: CI, confidence interval; DEHP, di-2-ethylhexyl phthalate; DEHTP, dioctyl terephthalate; DINCH, cyclohexane-1,2-dicarboxylic acid diisononyl ester; DINP, diisononyl phthalate; HMW, high molecular weight; LMW, low molecular weight.

* P ≤ 0.05.

# P ≤ 0.0083.

^aCalculated based on Vermeulen et al. formula from measured sex hormone binding globulin, testosterone, and albumin.

Finally, DEHP substitute phthalates showed varying associations across age groups. Among men age 20 to 39, DINCH and DEHTP metabolites showed positive associations with total testosterone. In addition, point estimates for estradiol were negative, which translated into significantly greater testosterone/estradiol ratios in relation to DINCH and DEHTP. However, DINCH was associated with lower total testosterone and SHBG among older men (i.e., among both men age 40 to 59 and ≥60), although no associations with FT or BAT. Last, DEHTP was not associated with total testosterone among older men, but was associated with lower FT (-2.42; 95% CI, -5.08 to 0.32), BAT (-2.54%; 95% CI, -5.12% to 0.10%), estradiol (-2.61%; 95% CI, -5.49% to 0.37%), and FAI (-2.37%; 95% CI, -4.48% to -0.21%) in men ≥60.

Results did not change when the sample was restricted to those with morning blood collections only (data not shown). In addition, there was little variation in blood draw timing across strata of age or race/ethnicity, thus yielding a sample that was similarly distributed with respect to demographics and covariates compared with the total original sample (data not shown).

Discussion

In this study among a representative sample of US men, we identified age-specific associations of phthalates with lower testosterone. LMW phthalates, commonly found in cosmetics and lotions, were associated with lower testosterone in younger men; and HMW phthalates, commonly found in food packaging and PVC products, were associated with lower testosterone in older men. In contrast, among men age 40 to 59, LMW phthalates were associated with increased free and bioavailable testosterone. A secondary and important finding was lower testosterone among older men in relation to terephthalate metabolites (i.e., DEHTP) of polyethylene plastics and DINCH, which were previously thought not to have the same adverse effects as orthophthalates used to soften PVC plastics.

Previous work has indicated that a number of phthalates, such as DEHP and di-n-butyl phthalate (and its primary metabolite mono-n-butyl phthalate), are associated with lower testosterone levels (43). Results from this analysis were consistent with previous work on the subject and also found an association between DEHP exposure and testosterone levels, particularly among older men. This was previously reported by Meeker and Ferguson, specifically among men age 40 to 60 (28). In our study, we reported an association between DEHP and lower testosterone in men age ≥60, and additionally found that DEHP was associated with lower free and bioavailable testosterone in this age group.

Although we detected lower testosterone levels in relation to several phthalates, estimates were small and the clinical significance of these findings is unclear. Still, there were a number of estimates that were ≥5%, which is notable. For example, DEHP was associated with almost 8% lower total testosterone, 6% lower estradiol, and 5% to 6% lower FT and BAT among men aged ≥60. This older age group already has lower testosterone, suggesting that environmental chemical exposure might hasten naturally occurring testosterone decline and thereby accelerate any adverse physiological consequences (e.g., loss of bone and muscle mass).

Other studies have recommended against the use of PVCs that contain phthalates such as DEHP because of their broad systemic effects (43), yet our results suggest that their alternatives, such as PETs and DINCH, may not be free of health impacts. We found that both DEHTP and DINCH were associated with higher testosterone among younger men age 20 to 39, but that DINCH was associated with lower total testosterone among those age 40 to 59 and ≥60, and DEHTP was associated with lower FT, BAT, and FAI among those age ≥60. These results suggest that these newer derivatives merit further study, given their potential for similar endocrine-disrupting effects to their PVC predecessors. If replicated, this may warrant the recommendation to avoid all plastics, and not only those with recycling numbers 3, 6, and 7, which are indicative of PVC and polystyrene plastics (44).

The heterogeneity of associations across age groups that we observed may have several explanations, including an age-period-cohort effect. For example, different phthalates were predominantly used at different times over the latter half of the 20th century. Therefore, men of different ages may have been exposed to different combinations at different times in their biological development (45). In particular, exposure to phthalates during the prenatal, early childhood, and pubertal periods, when the testosterone-producing Leydig cells in the testes develop (46), may dysregulate testosterone production throughout the life course, with consequences for reproductive function (47). For example, a study of Swedish military recruits (48) whose mothers’ prenatal sera samples were analyzed for phthalates reported that prenatal DEHP and DINP exposure were associated with lower semen volume in their offspring. DINP was also associated with lower testicular volume and DINP metabolites were associated with higher FSH and LH, even after models were adjusted for current phthalate levels. Although phthalate concentrations in our study reflected recent exposure owing to rapid metabolism, experiencing different chemical exposures during critical stages of gonadal development may affect lifelong testosterone production and perhaps vulnerability to chemical exposures, which may be another potential reason for our observed differences in associations by age group. This is supported by findings showing a relationship between indicators of Leydig cell functioning and phthalate exposure levels among young and adult men. Research has pointed to some specific possible mechanisms for this, including decreased functioning of the endoplasmic reticulum of Leydig cells, and a decrease in the number of Leydig cells that successfully differentiate to reach maturity, thus producing steroid hormones (49–52).

Another potential reason for the different associations by age are differences in the distribution of phthalate metabolite concentrations. In particular, we observed differences in HMW concentrations by age, such that younger men were more highly exposed than older men. It is possible that phthalates may exert nonmonotonic affects hormone concentrations as has been noted with other endocrine-disrupting chemicals such as bisphenol A (53, 54). In other words, those with high exposure levels may not show the strongest associations or even associations in the same direction as those with lower concentrations of exposure. In addition, if variation in exposure differs by age group, that affects the ability to contrast different exposure levels within age strata. Finally, hormone function and hormone levels vary widely across the lifespan, with total testosterone and BAT decreasing and SHBG increasing with age (55, 56). It is possible that exogenous exposures, such as phthalates, may interact with hormones differently at different concentrations.

Strengths of this analysis include a large, nationally representative sample of men with individually quantified urinary phthalate metabolites and serum sex steroid hormones. In addition, our age-stratified approach, reflective of different testosterone levels and reproductive life stages, allowed us to observe differential associations. In addition, our study was unique in deriving additional markers of sex steroid hormone measures, such as FT, BAT, and FAI, which offer a more comprehensive picture of the endocrine profile (55). Specifically, studies have argued that BAT provides a more accurate reflection of androgen status than total testosterone (55, 57). Finally, the inclusion of SHBG in our study helped to elucidate potential mechanisms for the observed changes in total testosterone, FT, and BAT. For example, DINCH was associated with lower total testosterone among men age 40 to 59 and ≥60, and this may be explained by the significant associations with lower SHBG, given that no associations were observed with FT or BAT. In contrast, HMW was associated with decreased total testosterone, FT, and BAT among men age ≥60, without accompanying decreases in SHBG, suggesting that this may occur through other mechanisms.

However, our study did have some limitations. A particular weaknesses of this study is its cross-sectional nature, which limits our ability to make inference about the temporality, directionality, and causality of these relationships. In addition, although we were able to control for several important covariates in the analysis, there may have been other factors that may have been important to adjust for, but were not available. For example, certain medical conditions may both affect testosterone levels and require phthalate-containing medications or medical products, such as gel capsules and PVC tubing, respectively (58), thus presenting a potential confounding pathway. Finally, we conducted multiple statistical tests and it is possible that some of the statistically significant findings were due to chance. In an attempt to address this issue, we applied a modified Bonferroni “correction” to the P value.

It is important to bring greater scrutiny to any structural analogues that may be used to replace DEHP. Studies of association are by their nature limited when it comes to establishing causal relationships or mechanisms of action, but growing evidence from both toxicologic and epidemiologic studies creates a compelling case for endocrine disruption, targeting sex steroid hormones.

Glossary

Abbreviations

BAT

bioavailable testosterone

CI

confidence interval

DEHTP

bis-2-ethylhexyl terephthalate

DEHP

di-2-ethylhexyl phthalate

DINCH

1, 2-cyclohexane dicarboxylic acid di-isononyl ester

DINP

diisononyl phthalate

FAI

free androgen index

FT

free testosterone

HMW

high molecular weight

LLOD

lower limit of detection

LMW

low molecular weight

NHANES

National Health and Nutrition Examination Survey

PET

polyethylene terephthalate

PIR

poverty-to-income ratio

PVC

polyvinyl chloride

T/E2

testosterone/estradiol

Additional Information

Disclosure Summary: The authors have nothing to disclose.