Elsevier

Reproductive Toxicology

Volume 136, September 2025, 108959
Reproductive Toxicology

Analyzing the impact of occupational exposures on male fertility indicators: A machine learning approach

https://doi.org/10.1016/j.reprotox.2025.108959Get rights and content

Highlights

  • •
    Combines occupational epidemiology with cutting-edge machine learning for risk stratification.
  • •
    Provides actionable thresholds for workplace exposure limits (e.g., WBGT ≥ 28°C linked to decreased testosterone levels).
  • •
    Bridges methodological gaps in reproductive health research through model interpretability (SHAP, feature importance analysis).

Abstract

Occupational exposures are critical factors affecting workers' reproductive health. This study investigates the impact of magnetic fields, electric fields, whole-body vibration, noise levels, and heat stress on male reproductive indicators using advanced machine learning models. The aim is to identify key risk factors and provide predictive insights into workers' reproductive health over the next decade. Data were collected from 80 male workers in an automobile part manufacturing plant, capturing demographic characteristics, occupational exposures, biochemical markers, hormone levels, and sperm parameters. Five machine learning models logistic regression, bagging classifier, extreme gradient boosting, random forest, and support vector machine were trained and evaluated using 5-fold cross-validation to determine effective predictors of reproductive health outcomes. Exposure to whole-body vibration, magnetic fields, electric fields, and heat stress closely affected free testosterone levels, with SHAP importance indicating: Magnetic Field Exposure (0.339) and Wet Bulb Globe Temperature (0.138). Worker Age (0.244) was the most influential demographic factor negatively impacting Free Testosterone. The XGBoost and random forest achieved the highest AUC (0.99), outperforming other models in predictive accuracy. The Random Forest model Importance (% Increase in MSE) predicted that Electric Field Exposure (5 %) and Magnetic Field Exposure (4.7 %) would have the most substantial negative impact on Free Testosterone, followed by Worker Age (4.1 %). This study underscores the need for targeted interventions, such as improved workplace safety protocols and regular health monitoring, to protect workers’ reproductive health.

Introduction

The decline in reproductive health, especially among working populations, has become an important public health issue in recent decades [1]. Increasing evidence links occupational exposures to various physical and environmental hazards, along with demographic factors, to negative reproductive health outcomes [2]. Despite progress in occupational health and safety, reproductive health issues among workers continue to be a significant concern, particularly in industrial and high-risk environments [3]. Of particular worry is the decline in male reproductive health, as measured by semen quality and hormone levels, which has been observed worldwide [4]. Research has shown that sperm counts and motility have decreased by more than 50 % over the past 40 years, with occupational exposures and lifestyle factors playing a crucial role in this trend [5], [6]. This decline not only affects individual fertility but also has far-reaching consequences for population sustainability and workforce productivity [7].
The male reproductive system is particularly sensitive to environmental and occupational hazards. Sperm production requires optimal conditions, including a scrotal temperature approximately 2–3°C lower than the body's core temperature [8]. Even slight increases in scrotal temperature, common in occupations such as furnace operators, drivers, and laundromat workers, can lead to reduced sperm production, abnormal sperm morphology, and disorders like oligozoospermia, azoospermia, and teratozoospermia [9], [10].
Similarly, exposure to electromagnetic fields (EMFs), particularly in the extremely low-frequency (ELF) range (3–300 Hz), has been linked to disruptions in testosterone synthesis and secretion[11]. EMF exposure can affect the polarization of cell membranes, impair Leydig cell function, and reduce the responsiveness of these cells to luteinizing hormone (LH), ultimately leading to decreased spermatogenesis and infertility [12].
In addition to heat and electromagnetic fields, sound and vibration are prevalent occupational hazards that can adversely affect male reproductive health[13], [14]. Occupations such as construction, manufacturing, and transportation often expose workers to high levels of noise and vibration. Prolonged exposure to these factors has been associated with increased risks of reproductive disorders, including decreased sperm quality and hormonal imbalances[15], [16].
The prevalence of exposure to these hazards varies by industry and region. For instance, a study by the International Labour Organization reported Central Asia have experienced a rapid increase in excessive heat exposure, with a 17.3 % rise between 2000 and 2020, almost double the global average increase of 8.8 %. This region has seen a 16.4 % increase in the proportion of heat-related occupational injuries since 2000[17].Similarly, Research indicates that workers in the electrical and construction industries are at higher risk of EMF exposure, with average magnetic field exposure ranging from 0.4–0.6 microtesla (μT) for electricians and electrical engineers to approximately 1.0 μT for power line workers. In the electrical supply industry, workers may be exposed to magnetic fields which can exceed 2000 μT and electric fields up to 30 kilovolts per metre (kV/m)[18].
Addressing these occupational hazards is crucial for protecting male reproductive health. Implementing safety measures such as proper ventilation, protective equipment, and regular health monitoring can help mitigate the adverse effects of heat, electromagnetic fields, sound, and vibration in the workplace.
The hormonal regulatory network of the reproductive system is complex and interconnected, with key hormones such as testosterone, luteinizing hormone (LH), follicle-stimulating hormone (FSH), and estradiol playing critical roles in maintaining reproductive function [19]. Testosterone, synthesized and secreted by Leydig cells in response to LH, is essential for spermatogenesis and overall male fertility [20]. However, exposure to environmental stressors like EMF and mechanical vibrations can disrupt this delicate balance, leading to hormonal imbalances and impaired reproductive health [21], [22]. Despite the growing body of evidence linking occupational exposures to reproductive dysfunction, the combined effects of multiple stressors on hormonal levels and sperm quality remain poorly understood.
Previous studies have primarily focused on the individual effects of occupational exposures, such as heat stress or EMF, on reproductive health [23], [24]. However, workers in industrial settings are often exposed to multiple stressors simultaneously, including noise, vibrations, and electromagnetic fields, and heat stress which may have synergistic or additive effects on reproductive outcomes [25].
Traditional statistical methods, such as multiple linear regression (MLR), often fail to capture the non-linear relationships and intricate interactions between risk factors and reproductive outcomes [26]. MLR relies on strict assumptions, such as linearity and the absence of multicollinearity, which are frequently violated in real-world datasets [27], [28]. In contrast, machine learning (ML) algorithms, such as random forest (RF), gradient boosting machines (GBM), and support vector machines (SVM), offer a more flexible and robust approach to modeling complex data [29]. These algorithms can identify non-linear patterns, handle multicollinearity, and provide accurate predictions without relying on restrictive statistical assumptions[30]. Despite their potential, the application of ML in occupational reproductive health research remains limited, particularly in predicting the impact of occupational exposures and demographic factors on reproductive outcomes.
This study aims to bridge the gap in understanding the impact of occupational exposures and demographic factors on male reproductive health by leveraging advanced machine learning (ML) techniques. A comprehensive dataset was used, incorporating occupational exposure metrics, demographic variables, and reproductive health indicators. We constructed predictive models to identify key risk factors and their interactions with reproductive health outcomes, such as sperm count, motility, and hormonal balance.
Five machine learning algorithms—random forest, extreme gradient boosting (XGBoost), support vector regression (SVR), logistic regression, and bagging classifier—were applied to explore occupational exposures and reproductive health. Random forest effectively handles large datasets and nonlinear relationships, while XGBoost offers high accuracy and efficiency, managing complex interactions and missing data. SVR performs well with small, high-dimensional datasets. Logistic regression provided a baseline for comparison, and the bagging classifier enhanced stability and accuracy by aggregating multiple decision-tree models. These methods were chosen not only for their proven effectiveness in predicting complex outcomes, but also for their ability to provide interpretable results, which is crucial for understanding the underlying factors affecting reproductive health. Each model brings unique strengths in handling different aspects of the data and offering insights into the most significant risk factors.[31].
The primary contributions of this study are as follows:
I. We identified the underlying structure of the relationships between occupational exposures (magnetic field, electric field, whole-body vibration, and equivalent noise level), demographic factors (age, work experience, BMI, WHR, education, smoking status, and regular exercise), and reproductive health indicators (free testosterone, FSH, LH, sperm count, normal morphology, sluggish motility, low motility, immotile sperm, malondialdehyde, superoxide dismutase, and total antioxidant capacity) using advanced machine learning techniques. We identified the underlying structure of the relationships between occupational exposures (magnetic field, electric field, whole-body vibration, and equivalent noise level), demographic factors (age, work experience, BMI, WHR, education, smoking status, and regular exercise), and reproductive health indicators (free testosterone, FSH, LH, sperm count, normal morphology, sluggish motility, low motility, immotile sperm, malondialdehyde, superoxide dismutase, and total antioxidant capacity) using advanced machine learning techniques.
II. We evaluated multiple ML classifiers using performance metrics to select the most accurate algorithm for predicting reproductive health outcomes in occupational settings.
III. We applied explainable artificial intelligence (XAI) techniques like SHAP to interpret the results and identify the major risk factors influencing reproductive health among workers.
By applying these advanced analytical techniques, this study not only enhances our understanding of the factors influencing reproductive health but also provides actionable insights for policymakers and healthcare providers [32]. The findings can inform the development of targeted interventions to mitigate occupational risks and improve reproductive outcomes among workers [33].

Section snippets

Study population

This study was conducted on workers employed at an automobile parts manufacturing facility, where foundry operations exposed them to various environmental hazards, including magnetic fields, electric fields, whole-body vibration, and high noise levels. Participants were selected based on predefined inclusion criteria and their voluntary agreement to take part in the research. After screening for eligibility, 80 workers were included in the study. The participants were between 20 and 50 years

Results

This study used a two-step analytical approach to explore the connections between demographic factors, occupational exposures, and reproductive health indicators, with the broader aim of forecasting workers' health trends over the next decade. First, univariate analyzes were performed to examine the linear relationships between biological responses (such as sperm parameters, hormone levels, and oxidative stress markers) and key predictors, including demographic characteristics and workplace

Discussion

The main goal of this study was to examine how occupational exposures and demographic factors influence male reproductive health, using machine learning models to predict reproductive outcomes. We aimed to pinpoint key risk factors and investigate the interactions between various workplace hazards—such as noise, vibration, and electromagnetic fields—and reproductive health indicators like sperm count, motility, and hormone levels.

Conclusion

This study concludes that machine learning, particularly tree-based models like Random Forest and XGBoost, can effectively identify key occupational and demographic factors influencing male reproductive health. Electric and magnetic field exposures, age, work experience, and oxidative stress biomarkers emerged as the most critical predictors. Explainable AI methods revealed complex interactions among these factors. The 10-year forecast highlighted electric field exposure as the most significant

CRediT authorship contribution statement

Somayeh Farhang Dehghan: Writing – review & editing, Writing – original draft, Validation, Supervision, Project administration, Methodology, Conceptualization. Farideh Golbabaei: Writing – review & editing, Project administration, Investigation, Conceptualization. Shayan Khoddam: Writing – review & editing, Writing – original draft, Methodology, Investigation, Formal analysis. Hamzeh Mohammadi: Writing – review & editing, Writing – original draft, Methodology, Investigation, Formal analysis,

Ethics approval and consent to participate

Ethical approval for this study was obtained from the School of Public Health & Allied Medical Sciences- Tehran University of Medical Science (IR.TUMS.SPH.REC.1398.297) and School of Public Health & Neuroscience Research Center, Shahid Beheshti University of Medical Sciences (IR.SBMU.PHNS.REC.1399.157). All participants filled out consent form and participated voluntarily in the study. All experiments were performed in accordance with the Declaration of Helsinki and national guidelines.

Consent for publication

Not applicable.

Declaration of Generative AI and AI-assisted technologies in the writing process

This article transparently discloses the selective use of artificial intelligence technologies to enhance research efficiency and communication quality. The applications were strictly limited to non-substantive tasks, with human authors maintaining full scholarly responsibility.
  • 1.
    DeepSeek (AI-Powered Research Assistant)
    • •
      Supported conceptual framing and initial structuring of the article
    • •
      Assisted in refining academic language and simplifying complex technical concepts
    • •
      Proposed organizational

Funding

This study was part of the research projects supported by the Tehran University of Medical Sciences (Grant no. 98–3–99–45128) and Shahid Beheshti University of Medical Sciences (Grant no. 25821).

Declaration of Competing Interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Somayeh Farhang Dehghan reports financial support was provided by Tehran University of Medical Sciences and Shahid Beheshti University of Medical Sciences.

Acknowledgements

The authors would like to thank the HSE office of the automobile parts manufacturing industry, especially Maryam Afzali Rad who helped us conducting the biological and environmental assessment.

Cited by (0)

1
ORCID ID: 0000–0002-6607–6396
View full text
© 2025 Elsevier Inc. All rights are reserved, including those for text and data mining, AI training, and similar technologies.