Phthalate metabolites and sex steroid hormones in relation to obesity in US adults: NHANES 2013-2016

Jiechang Zhang, Wen Gu, Shilei Zhai, Yumeng Liu, Chengcheng Yang, Lishun Xiao, Ding ChenFrontiers in Endocrinology 2024 · DOI

TL;DR

This study investigates the complex "toxic triangle" between environmental phthalates, sex steroid hormones, and obesity using data from over 7,700 US adults. By combining traditional epidemiology with machine learning, the researchers demonstrate that mixed exposure to phthalates disrupts hormonal balance—specifically lowering testosterone and raising estradiol—which in turn serves as a primary driver for obesity risk.

What Problem Does This Paper Address?

Obesity is a multifaceted public health crisis, but we are increasingly learning that it isn't just about "calories in vs. calories out." Endocrine-disrupting chemicals (EDCs), particularly phthalates (PAEs) found in plastics, solvents, and personal care products, are suspected "obesogens."

While previous research has linked phthalates to weight gain, the biological mechanism remains a bit of a "black box." This paper seeks to bridge that gap by examining whether phthalates drive obesity specifically by messing with our sex steroid hormones (testosterone, estradiol, and SHBG). The challenge lies in the "cocktail effect"—humans aren't exposed to one chemical at a time, but a complex mixture—and traditional statistical methods often struggle to untangle these overlapping influences.

Methods

The authors employed a robust, multi-stage analytical pipeline:

  • Principal Component Analysis (PCA): Used to group 11 different phthalate metabolites into "components" to handle multicollinearity (when variables are too closely related to study individually).
  • Weighted Quantile Sum (WQS) Regression: A specialized technique designed to evaluate the "mixture effect" of multiple chemicals, identifying which specific phthalates contribute most to hormonal changes.
  • Machine Learning (ML): The team tested several models, including Random Forest (RF) and eXtreme Gradient Boosting (XGBoost), to classify obesity status.
  • SHAP (SHapley Additive exPlanations): They used this framework to "peek inside" the black box of the XGBoost model, quantifying exactly how much each variable (hormones vs. chemicals) contributed to the model's prediction of obesity.

Datasets

The study utilized data from the National Health and Nutrition Examination Survey (NHANES) 2013-2016. This is a high-quality, nationally representative dataset from the CDC. The final analysis included 7,780 adult participants, providing significant statistical power.

Key Insights and Ideas

  • The "Good" and the "Bad" Components: PCA revealed that not all phthalates act the same. "PC2" (led by MCNP, MCOP, and MCPP) was a clear risk factor for obesity. Interestingly, "PC1" (led by MECPP, MEHHP, and MEOHP) showed a negative association, highlighting the complexity of how different chemical structures interact with human metabolism.
  • Hormonal Disruption is the Bridge: The WQS results showed that mixed phthalate exposure was significantly associated with decreased total testosterone (TT) and increased estradiol and SHBG. This suggests that phthalates act as "upstream" disruptors that trigger a hormonal cascade leading to weight gain.
  • Hormones Outrank Chemicals in Prediction: When the machine learning models were asked to predict who was obese, the sex hormones (especially SHBG and Testosterone) were the most important features. While phthalates were significant contributors, their impact was largely filtered through their effect on the endocrine system.

Key Results

  • Model Performance: The XGBoost model achieved the highest classification accuracy at 89.1% (F1-score: 0.879), followed closely by Random Forest at 88.2%.
  • The Testosterone Drop: Mixed phthalate exposure was strongly correlated with a decrease in total testosterone ($\beta$ = -20.85).
  • The Obesity Link: PC2 phthalates increased the odds of obesity (OR = 1.082), while hypertension and diabetes remained the strongest comorbid risk factors (ORs of 3.31 and 3.54, respectively).
  • Feature Importance: SHAP values confirmed that SHBG, Total Testosterone, and Estradiol were the top three most influential variables in predicting obesity status, outperforming age and even specific phthalate levels.

Limitations and Future Work

The authors acknowledge several key hurdles:

  1. Cross-Sectional Design: Because the data is a "snapshot" in time, they cannot definitively prove that phthalates cause obesity—only that they are strongly related.
  2. Single-Point Sampling: Phthalates pass through the body quickly. A single urine sample might not capture a person's long-term exposure levels.
  3. The "Male-Mid" Mystery: Subgroup analysis showed that the correlation between phthalates and estrogen was negative in middle-aged men but positive in other groups, a nuance that requires further biological investigation.

Future Work: The researchers call for longitudinal studies (following people over years) and in vivo (animal/cell) experiments to confirm the exact molecular pathways by which these plasticizers flip the "obesity switch" in the human endocrine system.

Summary updated

Abstract

Individuals with obesity exhibit lower levels of TT and SHBG, accompanied by elevated estradiol levels. Exposure to PAEs disrupts sex hormone levels, contributing to an increased risk of obesity in US adults. In the exploration of the interrelationships among these three factors, the RF and XGBoost algorithm models demonstrated superior performance, with sex hormones displaying higher variable importance.
BibTeX
@inproceedings{zhang2024phthalate,
  author = {Jiechang Zhang and Wen Gu and Shilei Zhai and Yumeng Liu and Chengcheng Yang and Lishun Xiao and Ding Chen},
  title = {{Phthalate metabolites and sex steroid hormones in relation to obesity in US adults: NHANES 2013-2016}},
  booktitle = {Frontiers in Endocrinology},
  year = {2024},
  doi = {10.3389/fendo.2024.1340664},
}
Figure