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

Plain-language summary

This study asked which workplace exposures matter most for men's reproductive health. It took 80 men working in an automobile parts plant, measured their exposure to magnetic fields, electric fields, whole-body vibration, noise and heat stress, and matched that against their hormone levels and sperm measures. Rather than testing one exposure at a time, it trained five machine-learning models to work out which exposures carried the most information about the outcomes, and ranked them by SHAP importance, a score for how much each input moves the model's prediction. Vibration, magnetic fields, electric fields and heat stress all came out as closely tied to free testosterone, the fraction not bound to protein and therefore usable by the body. Magnetic field exposure ranked highest at 0.339 and heat stress at 0.138, with worker age at 0.244 the strongest of the personal factors and the one the paper explicitly describes as pushing free testosterone down. Projecting forward, the model put electric fields and magnetic fields as the largest future influences. With 80 men at a single plant this is a small, single-site study, and a model ranking is not the same as a measured effect size.

Key findings

• Whole-body vibration, magnetic fields, electric fields and heat stress were all closely tied to free testosterone levels.
• Magnetic field exposure ranked highest among the workplace exposures, at a SHAP importance of 0.339, with heat stress at 0.138.
• Worker age, at 0.244, was the most influential personal factor and is described as negatively affecting free testosterone.
• Projected forward, electric fields and magnetic fields were the largest predicted influences on free testosterone.