Front Nutr. 2025 ;12 1611848
Yanyang Jin,
Dong-Shan Lv,
Li-Po Zhou,
Li Xiao,
Guang-Quan Tong,
Abdallah Karim,
Shuai Xue,
Hongyang Tian,
Cheng-Cai Wang,
Kun Feng,
Ding-Ming Song,
You-Liang Guan.
Background: Oxidative stress and dietary micronutrient imbalances have been implicated in prostate cancer (PCa) development and progression. Although flavonoids and antioxidants show promise in experimental models, evidence from population-based studies remains limited.
Objectives: This research aimed to investigate the relationship between the consumption of antioxidants and flavonoids in the diet and the risk and survival of PCa, as well as to assess the potential of machine learning models in identifying significant dietary factors.
Methods: Data from 2,629 male participants aged ≥40 years from National Health and Nutrition Examination Survey (NHANES) 2007-2010 were analyzed. Dietary intake was estimated using two 24-h recalls linked to the USDA Flavonoid Database. PCa status was self-reported. Survey-weighted logistic regression and Cox models evaluated associations with PCa prevalence and all-cause mortality, adjusting for demographic, lifestyle, and clinical covariates. Nine supervised machine learning models, including random forest (RF), were developed and validated. Shapley Additive Explanations (SHAP) values identified key predictors and visualized their effects.
Results: Among 2,629 U.S. male participants from NHANES 2007-2010, 144 reported a history of PCa. Compared with non-cancer individuals, cases had lower intake of selenium, magnesium, quercetin, kaempferol, epicatechin, epigallocatechin, total flavones, and total flavonoids (all P < 0.05). Higher intake of selenium, magnesium, catechin, and myricetin was associated with reduced PCa risk in weighted regression models, with selenium remaining significant after multivariable adjustment [odds ratio (OR) = 0.50, 95% confidence interval (CI): 0.33-0.76]. Lower intake of selenium, magnesium, luteolin, quercetin, kaempferol, and total flavones was linked to increased mortality risk, and selenium independently predicted improved survival [hazard ratio (HR) = 0.69, 95% CI: 0.54-0.88]. The RF model showed superior predictive performance [area under the curve (AUC) = 0.740], identifying selenium, luteolin, total flavones, myricetin, catechin, and magnesium as key features. SHAP analysis revealed U-shaped associations for selenium, catechin, and myricetin, and dose-dependent protective effects for luteolin and magnesium.
Conclusion: Our results highlight selenium, magnesium, and select flavonoids as promising dietary factors in reducing PCa risk and improving prognosis. These insights support the development of evidence-based, individualized nutritional strategies and call for further mechanistic and clinical investigations.
Keywords: NHANES; antioxidants; flavonoids; machine learning; magnesium; prostate cancer; selenium