Urinary phthalate concentrations and mortality risk: A population-based study

Phthalates are widely used as plasticizers. Laboratory-based mechanistic and epidemiological studies suggest that phthalates are detrimental to human health. Here, we present prospective analyses on phthalate exposure and all-cause, as well as cause-specific, mortality from the National Health and Nutrition Examination Survey (NHANES), a population-based cohort. Between 1999 and 2018, urinary concentrations of 12 phthalate metabolites were measured by high-performance liquid chromatography-electrospray ionization tandem mass spectrometry in spot urine samples of 10,881 adults aged 40 – 85 years, of which 2382 died over a median duration of 8.9 years after sample provision. Multivariable Cox regression analyses adjusted for a wide range of lifestyle factors and comorbidities showed that higher concentrations of mono-benzyl phthalate (MBzP) and Mono-n-butyl phthalate (MnBP) were associated with increased mortality. The hazard ratios for participants in the highest quartiles of MBzP and MnBP concentrations were at 1.27 [95% confidence interval: 1.08, 1.49; p linear trend = 0.002] and 1.35 [1.13, 1.62; p linear trend = 0.005). These findings reinforce the need for monitoring of phthalate exposure in relation to health outcomes.


Introduction
Phthalates are synthetic products from the esterification of phthalic acids with various alcohols. Depending on the particular alcohol used, different chemical properties can be derived. Phthalates are used as plasticizers to improve a material's softness and flexibility and to increase the durability and elongation of rigid polymers. They are widely applied in a variety of products such as flooring, cable sheathing, textured wallpaper and children's toys as well as in medical devices such as intravenous tubing and blood bags. Phthalates are further used by the cosmetics industry as ingredients in many personal care products, perfumes, creams, and makeup (also see Table 1) (Katsikantami et al., 2016). Through direct skin contact to cosmetics, but also via packaging materials and sealants, they can be directly absorbed. Moreover, due to their fat-soluble characteristics, phthalates can enter the body via packaged foods, particularly if these contain lipids (Dong et al., 2017). However, it is not only through consuming industrially manufactured products that people come into contact with phthalates, as these are now also detected throughout the environment, for example in air, water, and soil (Wormuth et al., 2006). Thus, it is almost impossible for consumers to completely avoid phthalates, as they are ubiquitous and present in goods that are an integral part of everyday life.   (Katsikantami et al., 2016(Katsikantami et al., , 2020Trasande et al., 2013;Kasper-Sonnenberg et al., 2012;Koch et al., 2013;Cheon, 2020). b According to the U.S. Consumer Product Safety Improvement Act (CPSIA) the listed phthalates must not exceed a proportion of 0.1%, April 25, 2018. c Maximum allowable dose levels according to Californiás proposition 65 Save Harbor levels list, inclusion of phthalates since 1988.
Phthalates have long been suspected to act as endocrine and metabolic disruptors (Eales et al., 2021) with hormone-like properties and harmful effects on reproductive ability (Panagiotou et al., 2021). For example, experimental evidence suggests that neonatal exposure to Di-2-ethylhexylphthalate (DEHP) leads to dysfunction of the male reproductive system (Lee et al., 2021). DEHP metabolites have also been associated with an increased odds for breast cancer in case-control studies , although a prospective study did not show associations between phthalate and breast cancer risk (Reeves et al., 2019). Cross-sectional epidemiological studies further indicate associations between higher urinary phthalates levels and obesity (Song et al., 2014), insulin resistance (Stahlhut et al., 2007), diabetes (Nam et al., 2020), and cardiovascular diseases (CVD) (Zhu et al., 2021).
While the harmful effects of phthalates have been described in many experimental studies, prospective epidemiological studies on phthalate exposure and risks of chronic diseases and mortality are lacking. Therefore, we examined the associations between 12 urinary phthalate metabolites, which reflect the exposure to common phthalates (Table 1), and all-cause as well as cancer-and CVD-specific mortality in the general American adult population, using data from the National Health and Nutrition Examination Survey (NHANES) 1999-2018.

Study population
NHANES is funded by the U.S. Centers for Disease Control and Prevention's (CDC) National Center for Health Statistics (NCHS). The study program examines and represents the health and nutritional status of the civilian noninstitutionalized U.S. population. The periodic routine collection of data by NHANES allows current public health issues to be addressed and monitored.
The survey is carried out in 2-year survey cycles since 1999. Representative sampling is achieved by a complex multistep process (Centers for Disease Control and Prevention (CDC), 1999). In each cycle, around 10,000 participants from 30 selected counties across the U.S. take part in the data collection, which includes interviews and physical examinations. Interviews are designed to obtain detailed information on lifestyle factors, socioeconomic and demographic characteristics, and dietary habits (Ahluwalia et al., 2016). Medical examinations are performed in mobile examination centers. The analyses presented here are based on an analytical sample of 10,881 participants aged ≥40 years, who participated in the ten NHANES cycles between 1999 and 2018, with data on urinary phthalates and mortality (see next paragraph). Data on phthalates, covariates and mortality were downloaded from the NHANES website (Centers for Disease Control and Prevention (CDC), a) and merged into one dataset. The NHANES survey protocols were approved by the NCHS Research Ethics Review Board, and all participants provided written informed consent (Centers for Disease Control and Prevention (CDC), b).
Out of the 101,316 participants of the NHANES cycles 1999-2018, 55,081 were adults older than 19 years, of which 54,945 with mortality data. Data on phthalate concentrations were available for 16,733 of these participants. We excluded participants younger than 40 years at urine sampling, given their low risk for mortality due to chronic diseases. Thus, the analytical sample for the present study consisted of 10,811 NHANES participants ( Supplementary Fig. 1).

Measurement of urinary phthalate metabolites
Commonly used phthalic acid diesters are converted to monoester metabolites and oxidative products (Frederiksen et al., 2007), which can be quantitatively detected in spot urine samples. A High-Performance Liquid Chromatography-Electrospray Ionization tandem mass spectrometry (HPLC-ESI-MS/MS) was used for analyzing urine samples in NHANES for the cycles 2003-2018 (Silva et al., 2007), while phthalate metabolites from the 1999-2000 and 2001-2002 survey cycles were analyzed using high-performance liquid chromatography-atmospheric pressure chemical ionization tandem mass spectrometry (HPLC-AP-CI-MS/MS) (Silva et al., 2005). Further information on the detailed test principles is provided in the Laboratory Procedures Manual provided by the NHANES (Centers for Disease Control and Prevention (CDC), c).
Urine samples were screened for fifteen phthalate monoester metabolites, which reflect the exposure to the phthalates that are most common in consumer products (Table 1)

Mortality data
Information on vital status, time between NHANES participation and death or censoring date, and cause of death was extracted from the 2019 Linked Mortality Files provided by the National Center for Health and merged with NHANES data using probabilistic linkage (National Center for Health Statistics, 2021). This linkage procedure is based on a combination of matching criteria (Social Security number; first name/initial, last name/initial, and maiden name; year, month, and day of birth; sex; race; state of birth) and has been shown to provide accurate results, with a sensitivity of around 97 to 98% (National Center for Health Statistics, 2021; Calle and Terrell, 1993;Rich-Edwards et al., 1994). NHANES participants are matched to deceased persons, from whose death certificates the date and cause of death according to the International Classification of Diseases (ICD-10) are retrieved using the National Death Index (NDI). Time between participation in NHANES and death or the time between participation and December 31, 2019, for participants who were not matched to a death certificate and were therefore d According to REACH, the European Regulation on Registration, Evaluation, Authorization and Restriction of Chemicals, phthalate levels are limited to a maximum level of 0.1 percent by weight of the plasticized material in a product (individually or in combination), entered into force on July 7, 2020. considered alive.

Covariates
After a comprehensive literature search, we included the following socio-demographic factors as covariates in our analyses (also see chapter 2.5): Age (years), sex, race/ethnicity (non-Hispanic white, non-Hispanic black, other), education level (college education or higher/high school degree or lower), and marital status (married or living with partner, not married or living with partner). We further included information on chronic diseases at the time of NHANES participation and sample provision, i.e. self-reported history of CVD (heart failure, coronary artery disease, angina pectoris, myocardial infarction, stroke), prevalent diabetes (yes, borderline, no), history of cancer, history of liver disease, and history of kidney disease. As health-related lifestyle factors we included data on self-reported physical activity, alcohol consumption, and smoking, as well as measured body mass index (BMI). Smokers were classified as "non-smokers" (smoked < 100 cigarettes in a lifetime), "current smokers" (currently smoked daily or on some days) and "former smokers" (smoked > 100 cigarettes in lifetime but do not currently smoke) based on self-reported smoking habits. Regarding their alcohol consumption participants were classified as "non-drinkers," "moderate drinkers," "binge drinkers," and "heavy drinkers" based on sex-specific criteria, as suggested by Agrawal et al. (2018). Physical activity was classified based on categories (moderate/vigorous vs. none) proposed by Vasquez et al. (Vásquez et al., 2014). Participants were classified according to their BMIs as "underweight" (<18.5 kg/m 2 ), "normal weight" (18.5-24.9 kg/m 2 ), "overweight" (25-29.9 kg/m 2 ), or "obese" (>30 kg/m 2 ).

Statistical analyses
To describe characteristics of the study participants, surveyweighted frequencies (categorical variables) and mean values as well as standard errors (SE, continuous variables) were used. For individual phthalate monoesters, we used survey-weighted medians and percentiles, given that proportions of measured values below the LOD were high for some metabolites.
Survey-weighted Cox proportional hazards regression analyses were performed to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for mortality risk across quartiles of phthalate levels. Before creating quartiles, phthalate levels were divided by urinary creatinine levels to adjust for urine dilution (Barr et al., 2005). Cox regression model 1 was adjusted for age, sex, and survey cycle. Model 2 was additionally adjusted for smoking status, physical activity, BMI, education level, alcohol consumption, marital status, race/ethnicity as well as history of cardiovascular diseases, diabetes, cancer, liver diseases, or kidney diseases at the time of urine sample provision. Linear trends were assessed via modelling phthalate/creatinine ratios on the log2 scale as continuous exposures. In addition to our main Cox regression analyses on single phthalate metabolites and mortality, analyses on total LMW, HMW, and DEHP metabolite concentrations were carried out. For these analyses, only data from the NHANES cycles 2003-2018 were used, as several metabolites necessary to create LMW, HMW, and DEHP exposure variables were not measured between 1999 and 2002 (MECCP) or 1999 and 2000 (MiBP, MCPP, MEHHP, and MEOHP).
Cause-specific hazards from a competing risk model were estimated in ancillary analyses on mortality due to cardiovascular diseases and cancer. In sensitivity analyses, we excluded participants with a history of cancer and CVD from our main Cox regression models. We also re-ran our main analyses adjusting for urinary creatinine rather than standardizing phthalate levels by creatinine. In further sensitivity analyses, we used cycle-specific LODs rather than applying the highest cyclespecific LOD to all cycles. Inspection of Schoenfeld residuals from Cox regression analyses did not point to deviations from the proportionality assumption (Hess, 1995). Associations were considered statistically significant at p-values <0.05. SAS 9.4 (Cary, NC, USA) was used for all statistical analyses.

Characteristics of the study population
Survey-weighted baseline characteristics of the study population are shown in Table 2. Of the 10,881 participants who took part in the NHANES cycles 1999-2018, 8499 were still alive on December 31, 2019, while 2382 had died. The median age of the study participants was 54.8 years (52.8 years among those, who survived, and 70.7 years among those who died between their the NHANES examinations and December 2019). Study participants who died after their participation in one of the NHANES cycles between 1999 and 2018 were more likely to be men (51.2% vs. 47.4% among those who were alive in 2019) and non- Unweighted counts; survey-weighted frequencies, medians and percentiles (p25 and p75) *Information on prevalent diabetes was assessed during interviews, with the option for interviewers to assign a 'borderline' status based on the participants' information; Hispanic white (79.9% vs. 71.6%). Among the deceased participants a somewhat less favorable pattern of socioeconomic, lifestyle, and health indicators was evident. For example, these participants were less likely to report never smoking (39.0% vs. 53.5%), less likely to live in a relationship (55.1% vs. 70.8%), and less likely to be physically active (39.4% vs. 56.1%), but had a higher prevalence of diabetes (21.3% vs. 11.4%) and other chronic diseases at the time of their participation in NHANES.
With regard to raw values of the phthalate monoesters (Supplementary Table 1), median MnBP, MEP, MBzP, MEHHP, MEOHP, and MECPP concentrations among participants, who died between NHANES examinations and closure date, were slightly higher than among the participants, who were still alive in 2019. By contrast, levels of MiBP and MCOP were slightly lower among the study participants who died.
Creatinine-standardized levels of phthalate metabolites showed moderate to strong positive inter-correlations, with the strongest among MEHHP, MECPP, and MEOHP ( Supplementary Fig. 2). The majority of MCHP (97.9%), MiNP (84.4%), MOP (96.8%), and MMP (50.8%) levels were below the limit of detection (Supplementary Table 2). Given the high proportions of values below the LOD, statistical analyses on MCHP, MiNP as well as MOP and mortality risk were not carried out.

Phthalate concentrations and mortality risk
Median time from NHANES assessment to death or study closure was 8.9 years. Age-and sex-adjusted Cox regression analyses (Model 1) showed a statistically significant positive association between higher MBzP levels and all-cause mortality risk, with an HR of 1.48 [95% CI: 1.26, 1.72] between extreme quartiles and a p-value for linear trend (p linear trend ) of <0.0001 (Table 3). This association was attenuated following multivariable adjustment (HR Q4vs.Q1 : 1.27 [1.08, 1.49], p linear trend = 0.002), but remained statistically significant, also when correcting the p-value for multiple testing (corrected p linear trend = 0.024). Higher levels of MnBP (HR Q4vs.Q1 : 1.52 [1.27, 1.82], p linear trend < 0.0001) were associated with increased all-cause mortality in the ageand sex-adjusted Cox regression model as well. This association was also attenuated after multivariable adjustment in Model 2 (HR Q4vs.Q1 : 1.35 [1.13, 1.62], p linear trend = 0.005), with only a borderline significant trend for a linear association upon correction for multiple testing (corrected p linear trend = 0.06).
In ancillary analyses restricted to the NHANES cycles 2003-2018, the sum of LMW metabolites was not significantly associated with mortality (HR Q4vs  Table 3).

Sensitivity analyses
Sensitivity analyses excluding individuals with cancer or CVD at baseline showed similar results compared to our main analyses (Supplementary Table 6), although risk associations became slightly stronger for MBzP and MnBP levels. The use of quartiles of raw phthalate levels with adjustment for urinary creatinine in Cox regression models instead of prior standardization for creatinine only led to marginal changes of our main results (Supplementary Table 7). Our results also remained similar when using cycle-specific LOD values instead of uniform LOD values across all cycles (Supplementary Table 8). We did not adjust for intermediate risk factors such as blood lipids or blood pressure, as including these as covariates in our multivariable Cox regression models only very marginally changed the associations between phthalate levels and mortality. For example, when adjusting for total cholesterol, triglycerides and systolic blood pressure, the association between MBzP levels and all-cause mortality changed from an HR Q4vs.Q1 of 1.27 [95% CI: 1.08, 1.49] to an HR Q4vs.Q1 of 1.26 [95% CI: 1.07, 1.48].

Discussion
In the present study, we observed positive associations between urinary concentrations of MBzP as well MnBP and all-cause mortality risk in the population-based NHANES study after comprehensive multivariable adjustment. These associations were slightly stronger regarding cancer mortality compared with cardiovascular mortality.
The parent phthalate of MBzP, Benzyl butyl phthalate (BBzP), is used as a plasticizer in a wide range of products, such as sealants, food packaging, furniture, care products, or artificial leather, of which many contain polyvinyl chloride (PVC) (Trasande et al., 2013;Begum et al., 2020). MnBP is a metabolite of Di-n-butyl (DnBP), a phthalate that is used in cosmetic products and cellulose plastics (Trasande et al., 2013;Begum et al., 2020). Several mechanisms may explain the increased mortality risk, particularly due to cancer, among individuals with higher MBzP and MnBP levels in this study. For example, BBzP may alter DNA methylation in the estrogen receptor promoter regions (Kang and Lee, 2005). In addition, a previous analysis of NHANES data showed that MBzP was associated with longer leukocyte telomere length (Scinicariello et al., 2016), and both longer and shorter telomere length are associated with an increased cancer risk (Qu et al., 2013;McNally et al., 2019;Protsenko et al., 2020). MBzP, MnBP, and their parent phthalates may affect the expression of tumor suppressor genes and protooncogenes in leucocytes (Sicińska et al., 2022), and may also have mutagenic properties (Wang et al., 2012). However, while in vitro and animal studies suggest that different phthalates cause DNA damage (Caldwell, 2012), impair DNA methylation (Singh and Li, 2012), modify the expression of particular genes (Caldwell, 2012;Roy et al., 2015), and increase cell proliferation (Caldwell, 2012), epidemiological evidence on a link between phthalates and cancer risk is generally lacking (Hlisníková et al., 2020) and a few recent prospective studies have shown mixed results. A prospective study from the US on urinary phthalate metabolites and breast cancer risk did not show associations (Reeves et al., 2019). In another US cohort, urinary DEHP levels were positively associated with breast cancer risk, while an inverse association was observed for MBzP levels. By contrast, DnBP exposure via prescription drugs was associated with increased breast cancer risk in a nationwide cohort from Denmark (Ahern et al., 2019), and there was a trend for an association between DnBP metabolites and increased endometrial cancer risk in a smaller prospective study from the US (Sarink et al., 2021). The evidence on phthalates and cancer outlined above may explain the findings of the present study to some degree, although the associations with MBzP and MnBP levels were similar for cancer and cardiovascular mortality. This may in part be due to additional cardiovascular effects of phthalates, such as increased intima media thickness and plaque formation (Wiberg et al., 2014), and MBzP levels were associated with prevalent cardiovascular diseases in a previous NHANES analyses (Zhu et al., 2021). Alternatively, BBzP and DnBP exposure may be related to a greater risk of cardiovascular diseases and cancer via effects on body weight (Song et al., 2014), insulin resistance (Stahlhut et al., 2007;Nam et al., 2020), and inflammation (Liu et al., 2022). However, more prospective epidemiological studies on phthalates and risks of major chronic diseases are needed to better understand the present associations.
NHANES is an open resource that is available to researchers worldwide, and three previous analyses on phthalates and mortality risks have been carried out based on data from the study (Sturgeon et al., 2016;Trasande et al., 2021;Zeng et al., 2022). An earlier analysis by Sturgeon et al. from 2016 did not provide evidence for associations between urinary phthalate levels and cardiovascular mortality risk (Sturgeon et al., 2016), while all-cause and cancer mortality were not investigated. In a recent study by Trasande et al. from 2022, a positive association between MEP and cardiovascular mortality was observed, but no further significant associations between individual phthalate metabolites and mortality risks (Trasande et al., 2021). In a similar study by Zeng et al. from 2022(Zeng et al., 2022, the strongest association was between MBzP and all-cause mortality risk, which is in line with our study. Zeng et al. further reported a significant positive association between MECPP levels and CVD mortality, while cancer mortality was not investigated. Our study was different to the mentioned previous ones in that we used a more recent update of the mortality data provided by NHANES, with a closure date in 2019 vs. 2015. We also adjusted for history of CVD and cancer at baseline, and used exposure data from more NHANES cycles (1999-2018 compared to 2001-2010 in the study by Trasande et al. and2003-2014 in the study by Zeng et al.). Notably, despite differences with respect to the included NHANES cycles and case numbers, findings on phthalate exposure, especially composite LMW and DEHP variables in relation to all-cause mortality risk, were similar across the studies.
Phthalate concentrations analyzed in this study were from urine samples collected in the USA between 1999 and 2018. During this time, legal regulations for phthalate use in many countries worldwide have changed. In the USA, threshold levels for the most widely used phthalates including DEHP, DnBP, and BBzP were defined in the Consumer Product Safety Improvement Act of 2008 (CPSIA , Table 1). However, threshold levels for phthalates were only established in the CPSIA for children's toys and childcare products. Further nation-wide restrictions are in place for the use of phthalates in food packages, but not in other products. In the European Union, threshold levels for common phthalates have been introduced for a wider range of products in 2019 (Table 1). It remains to be seen whether these regulations are sufficient to protect consumers against adverse health effects, and continuous monitoring of phthalate exposure is needed to assess whether the present restrictions are sufficient to reduce phthalate exposure on the population level. In this context it should acknowledged that exposures to different phthalates and related health effects may greatly vary across different countries and populations, which is why more international studies are needed (Benjamin et al., 2017). Whether 'non-phthalate' plasticizers, derivatives of organic acids that are increasingly used instead of phthalates, may be phthalate alternatives that are less harmful to human health also needs to be assessed and monitored (Schwedler et al., 2020).
A major strength of the present analyses is the sample size of the NHANES study. Notwithstanding the large case number for analyses on the primary endpoint, all-cause mortality, case numbers were somewhat smaller for the analyses on cancer and cardiovascular mortality, which may have precluded from detecting moderate risk associations for these endpoints. Analyses on individual causes of death were not possible due to limited data availability and it should also be noted that the mortality data were derived from probabilistic linkage. Another strength of NHANES is the measurement of objective phthalate biomarkers in urine. However, intra-individual urine biomarker concentrations can fluctuate and only a single spontaneous sample was taken from each study participant in NHANES. Ideally, average phthalate values from repeatedly collected urine samples should be used for a more precise exposure assessment, as in a mother-child cohort from France (Guilbert et al., 2021). Nevertheless, considering intra-individual fluctuations, it is likely that the true effect of phthalate exposure on mortality is stronger, given regression dilution indicated by moderate correlations of phthalate concentrations over time from previous studies (Hoppin et al., 2002;Townsend et al., 2013). At the same time, we acknowledge that the associations between phthalate concentrations and mortality risks observed in our study were moderate in magnitude and that data on potential confounders was only available from one time point. Thus, and given the observational design of the study, residual confounding cannot be ruled out. Finally, we carried out a high number of tests, and chance findings cannot be ruled out despite the correction for multiple testing we applied in the analyses on total mortality. Given that longer-term randomized controlled trials on phthalates are not feasible due to ethical and practical reasons, a combination of laboratory-based experimental and prospective cohort studies is needed to better investigate the effects of phthalates on human health in the future.
In summary, we observed that higher urinary concentrations of the phthalate metabolites MBzP and MnBP were associated with increased risks of all-cause mortality in a large population-based study. These findings underline the need for monitoring phthalate exposure in relation to health outcomes in epidemiological studies.