Associations between phthalate metabolites and two novel systemic inflammatory indexes: a cross-sectional analysis of NHANES data

Abstract

Background

The potentially risky effects of metabolites of phthalates (mPAEs) on inflammation and immune function have attracted much attention in recent years. However, direct studies on the relationship between these metabolites and the systemic immune inflammatory index (SII) and systemic inflammatory response index (SIRI) are limited.

Methods

This cross-sectional study used generalized linear regression models (GLM), restricted cubic splines (RCS), weighted quantile sum (WQS), and Bayesian kernel-machine regression (BKMR) to analyze data from 2,763 U.S. adults aged between 20 and 80 years, obtained from the U.S. National Health and Nutrition Examination Survey (NHANES) conducted between 2013 and 2018. The study aimed to investigate the relationship between urine samples of nine mPAEs and levels of SII/SIRI in a single, nonlinear, and mixed relationship and explored the robustness of the findings under single and mixed effects using two sensitivity analyses for completeness. In addition, the effects of six variables (age, sex, BMI, the percentage of total daily energy intake from ultra-processed foods (UPFs), total vegetable intake, and dietary supplements) on the association results were explored through subgroup analyses to identify potentially important confounders.

Results

In single exposure analyses, mono-n-butyl phthalate (MnBP), mono-ethyl phthalate (MEP), and monobenzyl phthalate (MBzP) were positively associated with SII/SIRI. The findings from the two mixed exposure models demonstrated a positive association between the collective concentrations of mPAEs and levels of SII/SIRI, with MBzP being identified as a significant contributor to the urinary levels of mPAEs. The subgroup analysis results of the effects of single and mixed exposures show that the association between mPAEs and SII/SIRI is more significant in females, overweight/obese populations, young/middle-aged populations, and populations with high levels of intake of UPFs.

Conclusion

Positive associations were identified between mPAEs and SII/SIRI. MBzP was determined to have the most significant impact. The association between mPAEs and SII/SIRI is significantly influenced by female groups, young and middle-aged populations, overweight and obese individuals, as well as those with a higher intake of UPFs.

HIGHLIGHTS

1. Based on routine blood test data in clinical settings, SII and SIRI can quickly assess the systemic inflammatory response.

2. Multiple models (GLM, WQS, BKMR) show a significant association between mPAEs and SII/SIRI in both single and mixed exposures.

3. The association between mPAE and SII/SIRI is more pronounced in women, young and middle-aged individuals, overweight and obese individuals, and those with high UPF intake.

1. Introduction

Phthalates (PAEs) represent a category of environmental chemicals extensively utilized in various industries for their effectiveness as plasticizers. They are commonly integrated into a wide range of products, including furniture materials, cosmetics, plastic packaging, as well as personal care items like shampoos, lotions, and cosmetics [1,2]. The primary role of PAEs is to improve the pliability, sealing, and protective properties of materials, especially polyvinyl chloride (PVC) and other polymers [3]. Nevertheless, a notable concern associated with PAEs is their lack of covalent bonding to these materials, which allows for easy migration, leaching, and volatilization from products into the environment [4,5]. This non-covalent bonding enables the dispersion of these compounds into the air, water bodies, and soil, leading to extensive human exposure via ingestion, inhalation, and skin contact [6].

Once PAEs are introduced into the human body, they undergo rapid metabolism to form monoesters and various secondary oxidative byproducts, which are commonly referred to as metabolites of PAEs (mPAEs). These metabolites are commonly detected in biological fluids such as urine and blood, making them reliable biomarkers for evaluating human exposure to PAEs [7]. The omnipresence of PAEs and their metabolites in the environment and biological systems has raised significant public health concerns. Numerous epidemiological and experimental studies have linked these compounds to adverse health outcomes, including endocrine disruption, reproductive toxicity, and metabolic disorders [8,9]. Research indicates that PEMs possess anti-androgenic properties through interference with androgen signaling transduction pathways, potentially disrupting male reproductive system ontogenesis [8,9]. In females in the United States, urinary monobenzyl phthalate (MBzP) and mono-ethyl phthalate (MEP) concentrations exhibit a positive association with endometriosis risk elevation [10]. Adult epidemiological investigations further identify dose-dependent relationships between elevated mPEMs concentrations and adiposity prevalence [11]. Adolescent cohorts demonstrate significant di(2-ethylhexyl) phthalate (DEHP) exposure associations with impaired insulin sensitivity [12]. Significantly, recent research suggests that mPAEs may play a role in promoting systemic inflammation, a crucial pathway in the pathogenesis of numerous chronic illnesses [13].

Inflammation represents a multifaceted biological response to detrimental stimuli, such as pathogens, impaired cells, or irritants. The primary objective of inflammation is to eradicate the initial cause of cellular damage, eliminate necrotic cells, and commence tissue restoration [14]. The process commences with the identification of harmful agents by pattern recognition receptors (PRRs) on immune cells, triggering the activation of signaling pathways like Nuclear Factor-kappa B (NF-κB) and the secretion of pro-inflammatory cytokines such as Interleukin 6 (IL-6) and Tumor Necrosis Factor alpha (TNF-α). These agents coordinate the recruitment of leukocytes to the injury site, heightened vascular permeability, and the stimulation of endothelial cells and platelets [15]. While acute inflammation is pivotal for recovery, persistent inflammation can result in prolonged tissue harm and is associated with conditions like cardiovascular disease, diabetes, cancer [16], asthma [17], non-alcoholic fatty liver disease [18], depression [19], and arthritis [20].

Inflammatory markers are indispensable tools for the diagnosis and monitoring of inflammation-related conditions. Commonly utilized markers include C-Reactive Protein (CRP), the Neutrophil-lymphocyte ratio (NLR), the Platelet-lymphocyte ratio (PLR), and various cytokines such as IL-6 and TNF-α [21,22]. However, these conventional markers often assess only a single aspect of inflammation and may lack specificity or sensitivity in diagnosing specific conditions [23]. The novel inflammation indices, particularly the systemic immune inflammatory index (SII) and the systemic inflammatory response index (SIRI), serve as significant indicators of the equilibrium between inflammation and immune response. These indices are derived from a combination of lymphocyte, monocyte, neutrophil, and platelet counts [20]. SII was originally suggested as a stand-alone prognostic factor for postoperative recurrence and overall survival in patients with hepatocellular carcinoma (HCC). It was later broadened to serve as a marker of inflammatory-immune equilibrium [24]. SIRI, which was initially unveiled in 2016, functions as a prognostic tool for predicting survival rates, monitoring immune responses, and assessing systemic inflammation in individuals with pancreatic cancer post-chemotherapy treatment [25]. SII/SIRI integrate multidimensional information from neutrophils (pro-inflammatory), lymphocytes (immune regulation), platelets (coagulation/inflammation), and monocytes (chronic inflammation) [26]. Compared to other inflammation markers that target specific inflammatory pathways or molecules and reflect local inflammatory responses, SII/SIRI can more systematically and comprehensively reflect the overall immune-inflammatory balance state of the body [23,27]. In addition, SII/SIRI are based on routine blood test data and do not require additional testing, making them suitable for rapid clinical assessment [28,29]. They show higher sensitivity and specificity in tumor prognosis and infectious disease monitoring [30,31].

Current evidence indicates that mPAEs induce inflammation primarily by stimulating an inflammatory response through the activation of NF-κB in B cells. This activation leads to the upregulation of genes encoding pro-inflammatory cytokines and chemokines, resulting in the production of inflammatory markers. However, research on the association between mPAEs and blood cell-based inflammatory markers in the context of mixed exposure is still insufficient, with a greater focus on exploring the mediating effects of inflammation. The findings of Xu et al. indicate that the systemic inflammation represented by SII mediates the singular and mixed effects of mPAEs on biological aging [32]. Niu et al. also found that the association between mPAEs and hyperuricemia may be related to the mediating role of SII [33]. Additionally, a study based on community data from Wuhan, China, revealed that during seasonal changes, the levels of inflammatory markers such as leukocytes and lymphocytes increased more significantly after winter exposure to PAEs [34]. Zhao et al.’s research focused on the association between mPAEs and oxidative stress and inflammatory markers in children, such as SII, PLR, and others [13]. These studies preliminarily reveal the potential association between mPAEs and inflammatory markers, but there is little research that delves into the specific associations between mPAEs, SII, and SIRI in adult populations under mixed exposure conditions. Furthermore, various confounding factors, such as age, sex, and BMI, may significantly influence these associations. Among these, dietary intake, as a major source of PAEs and an important trigger for inflammation, may also impact this association. Colacino et al. analyzed the relationship between various dietary intake levels and human PAEs levels, finding an association between monoethyl phthalate, a metabolite of diethyl phthalate (DEP), and vegetable intake [35]. Romano et al.’s research found a significant association between the intake of dietary supplements and MEP concentration [36]. Ultra-processed foods (UPFs), due to their plastic packaging and the use of food additives, contribute significantly to mPAEs exposure [37–39]. In addition, the impact of dietary intake on inflammation has been confirmed in multiple studies [40–42]. Therefore, based on the exploration of the association between mPAEs and SII/SIRI, a thorough analysis of the effects of various potential confounding factors will be a necessary and valuable research direction.

This study utilized four different models: generalized Linear Regression Models (GLM), weighted Quantile Sum (WQS), Bayesian Kernel-Machine Regression (BKMR), and restricted Cubic Splines (RCS) to analyze the effects of single and mixed exposures to nine mPAEs on SII/SIRI. Based on this, subgroup analyses were conducted on six confounding factors, including age, sex, BMI, and three dietary intake variables: the percentage of total energy intake from UPFs, total vegetable intake, and use of dietary supplements, to identify potential influencing variables. Additionally, sensitivity tests were performed to assess the robustness of the results. The findings of this study may provide new epidemiological insights into the association between exposure to mPAEs and susceptibility to immune-mediated inflammation.

2. Method

2.1. Study population

The National Health and Nutrition Examination Survey (NHANES) is a major survey administered by the National Center for Health Statistics (NCHS) to elucidate the health and nutritional status of the U.S. population. NHANES information is obtained primarily through interviews, physical assessments, and the collection of biological samples. Since its inception in the 1960s, the survey has played a critical role in shaping public health policy and evaluating health interventions in the U.S. The NHANES survey was approved by the NCHS Ethics Review Board.

The research incorporated information from a sample of 17,057 individuals aged 20 years and above, spanning the years 2013 to 2018. A total of 5,204 participants had both phthalate and complete blood count data after excluding those with missing mPAEs (11,853) and inflammatory indexes (184). Additionally, participants with missing covariates (2,257) were further excluded. Ultimately, 2,763 participants with complete data were recruited. The detailed procedure is illustrated in Figure 1.

Figure 1.

Flowchart of participants included in our final analysis (N = 2,763), NHANES, US, 2013–2018.

2.2. Measurement of mPAEs

Urine samples collected by the NHANES team were preserved by cryopreservation. Subsequently, mPAEs were extracted and concentrated from mixed organic solvents using solid-phase extraction. Finally, these metabolites were quantified using high-performance liquid chromatography-electrospray tandem mass spectrometry (HPLC-ESI-MS/MS). Detailed information on laboratory procedures has been provided in a related study [43]. To ensure the quality and improve the accuracy of the study, nine mPAEs were incorporated into the study, each exhibiting a detection rate exceeding 90%. These metabolites include monocarboxyisooctyl phthalate (MCOP), monocarboxy-isononyl phthalate (MCNP), MEP, MnBP, mono(2-ethyl-5-carboxypentyl) phthalate (MECPP), mono(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP), mono(2-ethyl-5-oxohexyl) phthalate (MEOHP), mono-isobutyl phthalate (MiBP), and MBzP. The information on the precursors and distribution of these nine mPAEs is shown in Table S1.

2.3. Measurement and calculation of systemic indicators of inflammation

The NHANES Mobile Examination Center (MEC) uses the Beckman Coulter DxH 800 instrument for comprehensive blood cell counting of blood samples. This instrument integrates an automatic dilution and mixing device for sample processing, as well as a single-beam photometer for hemoglobin analysis to determine the blood cell distribution of the subjects. In this survey, we extracted four variables from the CBC module, all measured in units of 1,000 cells/µL: lymphocyte count, monocyte count, neutrophil count, and platelet count. These were used to calculate two new systemic inflammation indices: SII = platelet count × neutrophil count/lymphocyte count [24], SIRI = neutrophil count × monocyte count/lymphocyte count [25].

2.4. Selection of covariates

To control for potential confounding effects, this study constructed a Directed Acyclic Graph (DAG) based on prior knowledge, domain expertise, and existing research evidence. According to the variable influence structure revealed by the DAG, potential confounding factors that may obscure the association between mPAEs and SII/SIRI were identified. These include age, sex, BMI, physical activity, race, drinking status, hypertension, diabetes, intake of UPFs, total vegetable intake, dietary supplements, and serum cotinine.

The types of UPFs were derived from the food information provided by participants in the NHANES dietary interview records and the NOVA classification system. Based on this, the percentage of total energy intake from UPFs was calculated using food energy codes provided by What We Eat in America (WWEIA), NHANES, and the Food and Nutrition Database (FNDDS). Total vegetable intake was calculated based on food types and consumption frequency information obtained from two 24-hour dietary recall questionnaires and one food frequency questionnaire, measured in commonly used cups in the United States. In addition, we also converted the non-normally distributed intake of UPFs and total vegetable intake into binary classifications using the weighted median based on the complex sampling design of NHANES. The data on dietary supplements come from the ‘Dietary Supplement Use 30 Days – Total Dietary Supplements’ dietary data from three rounds of NHANES conducted from 2013 to 2018. Whether dietary supplements were used is determined based on the results of the question coded as DSD010, which asks: ‘In the past month, have you used or taken any vitamins, minerals, or other dietary supplements?’

Physical activity was classified into three levels based on the MET range determined by the International Physical Activity Questionnaire (IPAQ): low intensity (≤600 MET minutes per week), moderate intensity (601–3000 MET minutes per week), and high intensity (>3000 MET minutes per week) [44]. Due to differences in the questions asked in the Alcohol Questionnaire (ALQ) in the NHANES surveys conducted in 2013–2016 and 2017–2018, alcohol consumption was divided into four groups based on the responses to the questions ALQ101 (frequency of drinking at least 12 times a year), ALQ120U (frequency of drinking per week, per month, or per year), and ALQ121 (frequency of drinking in the past 12 months) [45].

2.5. Statistical analysis

Due to the complex sampling design of NHANES, it is necessary to carefully consider weights and stratification variables in the statistical analysis process of this study. Specifically, the weight variable included in this study is the environmental subsample weight WTSB2YR. To assess the normality of continuous variables, we used the Kolmogorov-Smirnov test (KS). As nine mPAEs and two systemic indexes exhibited right-skewed distributions, an In-transformation was applied to achieve normality. Continuous variables were presented as means and standard deviations, and comparisons between groups were performed using the Wilcoxon rank-sum test. Categorical variables were represented as frequencies in percentage form and subjected to analysis utilizing the chi-square (χ²) test. Additionally, Spearman correlation coefficients were calculated for the nine mPAEs after In-transformation. All data analysis and visualization in this study were performed using R software (version 4.3.1), with statistical significance set at a p-value below 0.05.

2.6. mPAEs single exposure analysis

In this study, we employed generalized linear models to evaluate the impact of single mPAEs on SII/SIRI. The metabolite levels were categorized into quartiles, with the upper three quartiles compared to the lowest quartile (Q1). Two models were constructed for this study: Model 1 without any covariate adjustments and Model 2 with adjustments for age, sex, race, BMI, physical activity, serum cotinine, drinking status, hypertension, the intake of UPFs, total vegetable intake, dietary supplements, and diabetes. In addition, subgroup analyses explored the effects of sex, age, BMI, total vegetable intake, the intake of UPFs, and dietary supplements on the single and mixed associations between mPAEs and SII/SIRI.

Second, to verify the stability of the results, we performed sensitivity analyses to interpolate missing covariates to explore the robustness of the associations in the full dataset. We generated five interpolated datasets after several iterations using the Jomo package in R and interpolated unknown data hierarchically based on known covariate data with sampling units to ensure independence across datasets [46]. As both systemic inflammation indexes are associated with cancer prediction, we also verified the stability of the results by excluding the population of cancer patients (n = 271).

Finally, we utilized RCS to investigate the nonlinear dose-response association between mPAEs and SII/SIRI within a comprehensive model. The RCS model is a nonparametric regression technique that estimates the continuous functional relationship using a segmented cubic polynomial. It offers high flexibility and smoothness, making it a common choice for modeling nonlinear relationships between predictor and outcome variables [47]. In this research, the median value of nine mPAEs served as the reference values, and four nodes were chosen to construct the curves, specifically at the 5th, 35th, 65th, and 95th percentiles.

2.7. mPAEs mixed exposure analysis

In the present investigation, we used two complementary statistical models of mixture exposure: the WQS and the BKMR model. These models were utilized to examine the mixed impact of mPAEs on SII/SIRI.

The WQS model evaluates the general pattern of environmental exposure and the significance of each constituent in the mixture towards the overall pattern through the creation of a WQS index [48]. The expression of the WQS model is as follows:

$$g\left(\mu \right)={\beta }_0+{\beta }_1\left(\sum _{i=1}^c{\omega }_i{q}_i\right)+z\phi$$

$$g\left(\mu \right)={\beta }_0+{\beta }_1\left(\mathit{\text{WQS}}\right)+z\phi$$

where $g\left(\mu \right)$ is the generalized linear model link function; ${\beta }_0$ is the intercept of the model; ${\beta }_1$ represents the coefficients of the WQS index, which responds to the total exposure effect; in the WQS index $\mathit{c}$ represents the quantity of exposure variables; ${\omega }_i$ represents the exposure weights, which are constrained to a range of [0,1]. These weights correspond to the relative importance of the effect on the total effect; $q_i$ represents the exposure factor after the variable is quartiled; and $z$ and $\phi$ represent the covariate matrices with coefficients, respectively [49]. In the course of this study, the data will be partitioned into a 40% training dataset and a 60% validation dataset. The values of each mPAE can be derived through bootstrapping the training set, and the impact of each exposure will be constrained to a singular direction (either positive or negative). To improve the stability of the model, the WQS index will be averaged across multiple bootstrap samples. In this study, the bootstrap sample was set to 1000, and the number of seeds was 100. The sensitivity analysis of the WQS model utilized the same source population as the single exposure model. This decision was made to investigate the inherent association between the sensitivity analysis outcomes of the single exposure model and the mixed exposure model, and to evaluate their alignment with the results of mixed exposure throughout the entire population.

In addition, we used the BKMR model to evaluate the combined effects of assessing mPAEs on inflammation indexes. Unlike the WQS model, the BKMR not only has the capability to estimate the combined effects of mixed exposures on outcomes but also to determine the interactive effects among mixed exposures. This ability effectively tackles issues related to nonlinearity, nonadditivity, and intricate interactions among mixed exposures [50]. Given the significant association among mPAEs, the research employed a hierarchical variable selection approach to develop the model. This method constructs the model by dividing the highly correlated variables into groups. Its specific formula is shown below:

$$g\left(\mu \right)=h\left[\mathit{\text{Group}}1\left(\mathit{\text{MECPP}},\mathit{\text{MEHHP}},\mathit{\text{ME}}\mathrm{О}\mathit{\text{HP}}\right),\mathit{\text{Group}}2\left(\mathit{\text{MC}}\mathrm{О}P,\mathit{\text{MCNP}},\mathit{\text{MEP}},\mathit{\text{MnBP}},\mathit{\text{MiBP}},\mathit{\text{MBzP}}\right)\right]+x_i\beta +{ϵ}_i$$

where $g\left(\mu \right)$ is the generalized linear model link function; h[] represents the exposure response function framework; $x_i$ and $\beta$ denote covariates and effect coefficients, respectively; and ${ϵ}_i$ denotes the residuals [51]. In this study, the Markov Chain Monte Carlo (MCMC) sampler in the BKMR model was run through 10,000 iterations to provide reliability of the results. Similar to the sensitivity analyses performed on the WQS model, the BKMR model integrates both methods of sensitivity analysis to explore the variability of results across populations.

3. Results

3.1. Basic information about the population

Table 1 displays the fundamental demographic characteristics of adults aged between 20 and 80 years in the NHANES dataset from 2013 to 2018. Among them, the female population numbered 1,365 (49.40%). As indicated in Table 1, significant sex-based variations were observed in BMI, physical activity, dietary supplements, smoking habits, serum cotinine, and total vegetable intake (all p < 0.05), while no statistically significant differences were found in terms of age, race, the intake of UPFs, diabetes, and hypertension (all p > 0.05). The distribution of mPAEs in the study cohort from 2013 to 2018 is detailed in Table S1, with all nine mPAEs exhibiting detection rates exceeding 90%. Notably, MECPP had the highest detection rate (99.68%), whereas MBzP had the lowest (97.26%). Pearson’s correlation coefficients for these nine mPAEs revealed strong correlations between MEHHP, MEOHP, and MECHP (r > 0.9), while the correlations between the remaining metabolites were all below 0.3 (Figure 2).

Table 1.

Variables of study subjects in NHANES, 2013–2018.

Characteristic	Overall	Female	Male	p-Value
	N = 2,763	N = 1,365	N = 1,398
Age, n (%)				0.711
20–69 Years	2,390 (86.50%)	1,204 (88.21%)	1,186 (84.84%)
70–80 Years	373 (13.50%)	161 (11.79%)	212 (15.16%)
BMI, n (%)				<0.001
<25 kg/m²	755 (27.33%)	403 (29.52%)	352 (25.18%)
≥25 kg/m²	2,008 (72.67%)	962 (70.48%)	1,046 (74.82%)
Physical activity, n (%)				<0.001
low exercise intensity	495 (17.92%)	290 (21.25%)	205 (14.66%)
Moderate exercise intensity	1,086 (39.30%)	593 (43.44%)	493 (35.26%)
High exercise intensity	1,182 (42.78%)	482 (35.31%)	700 (50.08%)
Race, n (%)				0.718
Hispanic	636 (23.02%)	329 (24.10%)	307 (21.96%)
Non-Hispanic White	1,114 (40.32%)	536 (39.27%)	578 (41.34%)
Non-Hispanic Black	629 (22.78%)	311 (22.78%)	318 (22.75%)
Non-Hispanic Asian	262 (9.48%)	130 (9.52%)	132 (9.44%)
Other	122 (4.42%)	59 (4.33%)	63 (4.51%)
Drinking status, n (%)				<0.001
Non-drinker	883 (31.96%)	550 (40.29%)	333 (23.81%)
1–5 Drinks/month	1,223 (44.26%)	584 (42.78%)	639 (45.71%)
5–10 Drinks/month	246 (8.90%)	105 (7.69%)	141 (10.09%)
10+ Drinks/month	411 (14.88%)	126 (9.24%)	285 (20.39%)
Hypertension, n (%)				0.443
Yes	977 (35.36%)	486 (35.60%)	491 (35.12%)
No	1,786 (64.64%)	879 (64.40%)	907 (64.88%)
Diabetes, n (%)				0.436
Yes	416 (15.06%)	199 (14.58%)	217 (15.52%)
No	2,347 (84.94%)	1,166 (85.42%)	1,181 (84.48%)
Dietary supplements, n (%)				<0.001
Yes	1,549 (56.06%)	868 (63.59%)	681 (48.71%)
No	1,214 (43.94%)	497 (36.41%)	717 (51.29%)
Ultra-processed foods (%)				0.312
<54.82%	1473 (53.31%)	760 (55.68%)	713 (51.00%)
≥54.82%	1290 (46.69%)	605 (44.32%)	685 (49.00%)
Total vegetable intake (cup)				0.008
<1.39 Cups	1410 (51.03%)	743 (54.43%)	667 (47.71%)
≥1.39 Cups	1353 (48.97%)	622 (45.57%)	731 (52.29%)
Serum cotinine (ng/mL)	49.03 ± 115.12	35.24 ± 92.48	62.38 ± 132.09	<0.001

BMI: Body mass index; SII: Systemic immune inflammation index; SIRI: Systemic inflammatory response Index. Cup: US customary cup. Categorical variables are represented by n (%), continuous variables are represented by Mean ± SD.

Figure 2.

Correlation coefficients of spearman for concentrations of nine urinary mPAEs.

3.2. Single effect analysis of mPAEs on SII/SIRI

In the context of GLM models incorporating unadjusted and fully adjusted covariates, MnBP and MBzP exhibited positive associations with SII/SIRI (Figures 3(A,B) and 4(A,B)). Upon comparing the highest and lowest quartiles of the nine mPAEs within the GLM fully adjusted for covariates, it was observed that MnBP, MEP, and MBzP were positively linked to SII/SIRI (Figures 3(B) and 4(B)). Notably, in comparison to the first quartile (Q1), the second quartile (Q2) of MEOHP displayed a positive association with SIRI (Figure 4(B)).

Figure 3.

Associations of single mPAEs with SII in the NHANES 2013–2018 cycles. (A) Model 1 was not corrected for any covariates. (B) Model 2 adjusted for sex, age, race, BMI, physical activity, serum cotinine, drinking status, hypertension, diabetes, dietary supplements, ultra-processed foods, total vegetable intake. • p < 0.05; **p < 0.01; ***p < 0.001.

Figure 4.

Associations of single mPAEs with SIRI in the NHANES 2013–2018 cycles. (A) Model 1 was not corrected for any covariates. (B) Model 2 adjusted for sex, age, race, BMI, physical activity, serum cotinine, drinking status, hypertension, diabetes, dietary supplements, the percentage of total daily energy intake from UPFs and total vegetable intake. • p < 0.05; **p < 0.01; ***p < 0.001.

The subgroup analysis results show that the strength of the association between mPAEs and SII/SIRI varies across different population subgroups. Specifically, this association is more pronounced in the youth and middle-aged population aged 20–69, females, and individuals who are overweight and obese (Table S2). The dietary intake subgroup analysis results in Table S3 indicate that in populations with a high level of intake of UPFs and those not taking dietary supplements, the association between high levels of mPAEs (Q4) and SIRI is particularly prominent. Additionally, in populations with low total vegetable intake, the association between mPAEs and SII/SIRI is also significantly enhanced.

In both sensitivity analyses involving multiple interpolated covariates and the exclusion of cancer patients, the findings remained consistent with those obtained through linear regression analysis. MnBP, MBzP, and MEP were all found to elevate the levels of SII/SIRI (Figures S2(A), S2(B), S3(A), and S3(B)). The RCS models fully adjusted for covariates showed dose-response relationships between mPAEs and SII/SIRI. Statistically significant nonlinear relationships were observed between MBzP, MnBP, and SII/SIRI (p < 0.05) (Figures S4(D), S5(D), S4(H), and S5(H)), and between MiBP and SIRI (p < 0.05) (Figure S5(I)). There is a linear relationship between the remaining metabolites and SII/SIRI (p > 0.05) (Figures S4(A–C), S4(A–C), S4(E–G), S4(I), S5(A–C), and S5(E–G)).

3.3. Mixed effect analysis of mPAEs on SII/SIRI

The WQS indices of the nine mPAEs were positively correlated (p < 0.05) with SII/SIRI in the three WQS regression models. In the fully adjusted covariate model, for each quartile increase in the WQS index, the levels of SII/SIRI increased by 0.073 (95% CI: 0.017, 0.129) and 0.106 (95% CI: 0.050, 0.161), respectively (Figure 5). The weight results of the WQS model show that the mPAEs, MBzP and MnBP, have a positive impact on SII/SIRI, contributing the most to this effect, followed by the three phthalates MEP, MCOP, and MEOHP (Figure 6). In the two sensitivity analyses, the WQS index was positively correlated with SII/SIRI, consistent with the results for the overall population (Table S4).

Figure 5.

Association between WQS index and SII/SIRI, NHANES, United States, 2013–2018. Crude model was not corrected for any covariates. Model 1: adjusted for sex, age, race. Model 2: further adjusted for BMI, physical activity, serum cotinine, drinking status, hypertension, diabetes, dietary supplements, the percentage of total daily energy intake from UPFs and total vegetable intake. All mPAEs and two systemic inflammatory indexes were ln-transformed before analysis.

Figure 6.

WQS model regression index weights for the SII/SIRI. Estimated weights of urinary mPAEs for SII/SIRI by WQS models adjusted for sex, age, race, BMI, physical activity, serum cotinine, drinking status, hypertension, diabetes, dietary supplements, the percentage of total daily energy intake from UPFs and total vegetable intake.

The subgroup analysis results of the WQS model indicate that there are significant population differences in the association strength between the WQS index and SII/SIRI. In the population of young and middle-aged individuals aged 20 to 60, there is a significant positive correlation between the WQS index and SII/SIRI (p < 0.05) (Figure S6). In the female subgroup and among those who are overweight and obese, the association between the WQS index and SIRI is statistically significant. Furthermore, the dietary intake subgroup analysis in Figure S7 indicates that in the population with a higher intake of UPFs, the WQS index is positively correlated with both SII/SIRI (p < 0.05).

The results of the fully adjusted BKMR model showed no statistically significant differences in the levels of SII compared to the 50th percentile concentration at the high concentration of the mPAEs (Figure 7(A)). In contrast, SIRI levels were statistically different at all concentrations compared to the 50th percentile concentration (Figure 7(B)). From the overall trend, the levels of SII/SIRI showed an increasing trend when the concentration of the mPAEs increased (Figure 7). Table S5 displays group PIP versus cond PIP for each mPAE to assess the relative significance of mPAEs for SII versus SIRI. The groups with the highest PIP were all in Group I (group PIP = 0.944/0.965), with MBzP being the most influential in Group I (cond PIP = 0.9574/0.9401).

Figure 7.

The combined effect of SII (a), SIRI (B) and mixtures. The Y-axis represents the estimated difference when all mPAEs are fixed at a specific percentile (range from 0.25 to 0.75) compared to when mPAEs are at the 50th percentile. The dots represent the estimated values, and the black vertical line. The model was adjusted according to sex, age, race, BMI, physical activity, serum cotinine, drinking status, hypertension, diabetes, dietary supplements, the percentage of total daily energy intake from UPFs and total vegetable intake.

Figure S8 summarizes the univariate exposure-response functions for the nine mPAEs. Except for MECPP, the exposure-response trends for the remaining eight mPAEs and SII with SIRI were basically the same, with a positive dose-response relationship between MEOHP, MCOP, MnBP, MEP with MBzP, and SII with SIRI, and a negative dose-response relationship between MEHHP and MiBP. MECPP showed a positive relationship with SII (Figure S8(A)) and a negative relationship with SIRI (Figure S8(B)). When the other mPAEs were fixed at the 25th, 50th, and 75th percentiles, assessment of the effects of individual metabolites on SII/SIRI revealed that as MBzP increased from the 25th to the 75th percentile, the level of SIRI increased while the level of SII declined (Figure S9(A)). Similarly, as MEOHP increased from the 25th percentile to the 75th percentile, the levels of SII/SIRI continued to rise (Figure S9(B)). When MEHHP increased from the 25th to the 75th percentile, both SII/SIRI levels decreased (Figure S9). In a further interaction study, the binary exposure dose-response curves showed similar slopes of the exposure-response curves for one metabolite at different quartiles (10th, 50th, and 90th percentiles) of the other metabolite when the other mPAEs were fixed at the 50th percentile. There are no overlapping exposure-response curves between the various mPAEs (Figures S10(A) and S10(B)).

Similarly, sensitivity analyses were performed, excluding cancer patients and after multiple interpolations. The results remained consistent between the two sensitivity analyses and with the total population. The table of posterior inclusion probabilities (PIPs) indicated that MBzP continued to be the metabolite with the highest contribution in the model. Furthermore, the nine mPAEs in the overall effects of the mixtures also showed positive associations with SII/SIRI (Figures S11(A–F) and S12(A–F)).

4. Discussion

The current study comprehensively analyzed the effects of single and mixed exposures to nine mPAEs on two systemic inflammation indexes, SII/SIRI, using four distinct models. Results from the single-effect model showed that MBzP, MnBP, and MEP all positively correlated with SII/SIRI. In the WQS and BKMR models, overall mPAEs demonstrated positive associations with both indexes, with MBzP showing the strongest association. This study also analyzed the differences in the single and combined effects of mPAEs on SII/SIRI across different subgroups. The results showed that in the young and middle-aged population (ages 20–69), among females, overweight and obese individuals, and those with higher intake of UPFs, the association between mPAEs and SII/SIRI was significantly stronger.

MBzP is a metabolite of benzylbutyl phthalate (BzBP), a widely used plasticizer in the manufacturing of PVC, vinyl oils, synthetic leathers, and adhesives. BzBP has the potential to be released into the environment, accumulating on crops designated for human consumption or being absorbed into the human daily intake [52]. Several studies have been conducted to show a plausible link between BzBP and its metabolite MBzP and the inflammatory response. For instance, occupational exposure to BBP-containing mPAEs has been linked to increased rates of respiratory issues, neurological disorders, and cancer [53]. Furthermore, MBzP has been positively associated with elevated levels of the inflammatory marker C-reactive protein (CRP) [54]. This association is believed to be closely tied to the modulation of inflammatory matrix signaling pathways such as PI3K/AKT, Nrf2, NF-κB, and NLRP3, with NF-κB shown to stimulate CRP production [55,56]. Additionally, MBzP has been found to exhibit a negative association with serum bilirubin concentration, which possesses antioxidant properties capable of scavenging free radicals in the body. Free radicals are highly reactive molecules that induce cellular and tissue damage, leading to inflammation and oxidative stress [57]. Experimental evidence indicates that BzBP inhibits CpG-induced IFN-α and IFN-β expression in plasmacytoid dendritic cells (PDCs), potentially disrupting the immune system and exacerbating allergic inflammation [58]. Animal studies have demonstrated the toxic and inflammatory effects of BzBP on reproductive system development in mice [59]. Apart from MBzP, metabolites such as MnBP and MEP have been individually linked to SII/SIRI. Positive associations have been observed between MEHHP, MEOHP, and MnBP with potential inflammatory markers like alkaline phosphatase (ALP) and absolute neutrophil count (ANC), as well as significant positive associations between MnBP and MBzP with CRP [57]. In addition, exposure to MnBP may significantly increase the risk of inflammation in mice with neutrophilic asthma (NA) [60]. The risk of asthma in the adolescent population is also positively associated with the levels of urinary MnBP [42]. A study on a population of children in Brazil indicated that MEP, MEOHP, and MEHHP showed a significant positive association with the oxidative stress marker 8-hydroxy-2′-deoxyguanosine (8OHDG) [61].

The association between mPAEs and these two indicators varies across different subgroups. Recent studies suggest that the significant association of mPAEs with inflammation may be related to female-specific physiological processes, such as childbirth, pregnancy, the use of beauty products, and preterm birth. A study conducted on females in Puerto Rico during early pregnancy revealed that heightened levels of the metabolites of DEHP, MEHHP, MEOHP, and MECPP were connected to a slight elevation in IL-6 cytokine levels. IL-6 is a crucial mediator in the inflammatory cascade, facilitating the initiation and perpetuation of the inflammatory response [62]. Similarly, in a study involving a cohort of pregnant females in Puerto Rico, certain high molecular weight mPAEs exhibited positive associations with a majority of eicosanoids in both single- and multi-contaminant analyses. Eicosanoid metabolites like prostaglandin E2 (PGE2) are potent inflammatory mediators that could contribute to the progression of an inflammatory response [63]. Furthermore, a study in China demonstrated positive associations between various mPAEs and inflammatory markers such as leukocytes, neutrophils, lymphocytes, and monocytes in a pregnant population [64]. A comprehensive investigation involving females attempting to conceive in a representative U.S. study indicated that mPAEs were not only linked to markers of inflammation and oxidative stress but also suggested that preconception exposure could have adverse effects on females’ reproductive health [65]. Research evidence indicates that in overweight and obese populations, the positive associations between MEP and CRP, as well as between MEHHP and IL-6, show a significantly enhanced trend [66]. Notably, the association between di(2-ethylhexyl) phthalate (∑DEHP) and its metabolites with TNF-α is more pronounced in individuals with a higher BMI [67]. A study from Beijing further reveals that young and middle-aged individuals have a higher total concentration of mPAEs in their bodies compared to the elderly, due to factors such as occupational characteristics and lifestyle habits [68]. The association between MCNP and hsCRP in young and middle-aged populations is more significant compared to other age groups [69].

UPFs often use plastic packaging materials to extend shelf life and meet packaging demands, which allows PAEs to migrate from packaging materials and food processing equipment (such as plastic pipes and conveyor belts) into food matrices, leading to potentially higher PAE levels in UPFs compared to fresh or minimally processed foods [38]. Epidemiological studies also indicate that high levels of ultra-processed food intake are significantly associated with elevated serum inflammatory markers (such as TNF-α and IL-6) [70]. UPFs are rich in pro-inflammatory components (such as saturated fatty acids, trans fatty acids, refined carbohydrates, and food additives) and typically lack anti-inflammatory components (such as dietary fiber and polyphenols), which may induce inflammatory responses through various pathways [71–74]. Although there is currently no definitive research exploring the impact of ultra-processed food intake on the association between mPAEs and inflammation, given that UPFs are a major route of PAE exposure and may also directly induce inflammatory responses, ultra-processed food intake is likely to act as an important confounding factor in studies examining the inflammatory effects of mPAEs, necessitating consideration and control. The observed differences in the associations between mPAE and SII/SIRI due to varying intakes of UPFs in this study provide new clues and empirical evidence for understanding the potential role of UPFs in the complex relationship between mPAEs and human inflammatory responses.

mPAEs are continuously present in the external environment, sharing similar chemical properties and often having unclear origins and complex interactions. The potential for additive, antagonistic, synergistic, and complementary interactions among metabolites can result in correlative effects that differ from those of a single exposure.

For example, some congeners (such as DEP/DMP) can produce competitive inhibition through shared cytochrome P450 metabolic enzymes, leading to a decrease in toxicity thresholds or apparent inhibitory effects [3]. Additionally, due to their similar hydrophobicity and bioaccumulation characteristics, MnBP and MiBP, which have phthalate monoester homologous structures, can interfere with the functions of biomolecules through non-covalent binding [75]. Therefore, investigating mixed metabolite exposure analysis is crucial in research. In this study, the outcomes of two mixed exposure models, the WQS model and the BKMR model, indicated an overall positive association between mPAEs and SII/SIRI, demonstrating a positive trend in the collective impact of metabolites on inflammation. Seonyoung Park et al. concluded that the relationships between exposure to an individual phthalate and immune inflammation in pregnant females during pregnancy may not necessarily align with those of a combination of phthalates [63]. Furthermore, findings from a study in Guangzhou, China, suggested that single or mixed exposure to PAE was positively correlated with the oxidative stress biomarker 8-OHdG, but not significantly associated with various inflammatory markers [22]. This outcome contradicts the results of the current study, possibly due to variations in mPAEs exposure levels and regional disparities among the study populations. Although the above studies have demonstrated an association between mixed exposure to mPAEs and inflammation, studies delving into the specific mechanisms of action between mixtures and indicators of inflammation are still limited, and the potential associative effects are not yet rigorously explained and still need to be explored through a large number of scientific studies.

The current investigation delved into single versus mixed exposure analyses of mPAEs and two systemic inflammation indexes through four complementary models, which produced mostly consistent results. Nevertheless, the study has certain limitations that need to be acknowledged. Firstly, it relied on a cross-sectional survey design, which hindered the exploration of causal relationships between mPAEs and the two systemic indexes and posed challenges in controlling for bias. Secondly, since the NHANES data used in this study represent the American population, it may not fully reflect the exposure patterns and health characteristics of other regions or global populations. Due to differences in dietary habits, product usage, industrial practices, and environmental regulations among different countries and regions, the exposure levels and patterns of PAEs may vary significantly. Additionally, although this study has made efforts to control for various known confounding factors to assess the association between mPAE and SII/SIRI, there remain potential unmeasured confounders due to limitations in data availability and measurement methods, such as genetic factors and occupational exposure. Lastly, in the validation of the WQS model against the BKMR model, biases were observed in results such as the relative importance of all phthalate mixtures, preventing direct correspondence between them. It is imperative to employ complementary statistical methods to better assess the significance of each mPAE within the mixture of exposures.

5. Conclusion

In this study, the investigation focused on examining the impact of nine different mPAEs on the SII/SIRI. The results revealed that MnBP, MEP, and MBzP exhibited positive associations with both SII and SIRI. Subsequent subgroup analyses showed that the association between mPAEs and these two indicators was more significant in females, overweight and obese individuals, young and middle-aged populations, and those with high levels of ultra-processed food intake. The research results of the RCS model indicate that there is a nonlinear relationship between MnBP, MEP, and MBzP with SII/SIRI, as well as between MiBP and SIRI. Moreover, when considering the combined effects of exposure, mPAE mixtures generally showed positive associations with SII/SIRI. Notably, MBzP emerged as the primary positive influencing factor in this context. These research outcomes offer valuable insights into the connection between mPAEs and two systemic inflammation indexes, contributing to a more comprehensive understanding of the risks associated with phthalate exposure and their potential impact on inflammation. Therefore, it is recommended that relevant authorities develop and implement evidence-based environmental public health guidelines and feasible and effective intervention strategies to reduce human exposure to phthalate-generating products.

Funding Statement

This study was supported by the China National Center for Food Safety Risk Assessment research project (LH2022GG03).

Ethical approval

This study received an ethical exemption from the Ethics Committee of Shandong Second Medical University. We confirm that this study was conducted in accordance with the 1964 Declaration of Helsinki and its subsequent amendments. This study’s informed consent has been exempted by the Institutional Review Board of Shandong Second Medical University.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

All data are open access and available for download at url: https://www.cdc.gov/nchs/nhanes/?CDC_AAref_Val=https://www.cdc.gov/nchs/nhanes/index.htm. If reasonable requests are made, further data can be obtained from the corresponding author.