The current study confirms the initial hypothesis that mixed mineral intake is associated with a lower risk of hHcy. First, results from traditional multiple linear regression showing that eight minerals were associated with lower blood Hcy concentrations, as well as results from traditional multivariate logistic regression showing that eight minerals were associated with lower hHcy risk. Then, using three innovative machine learning methods, a mixed mineral intake was found to be associated with lower blood hcy concentrations and lower hHcy risk. Anyway, a growing body of literature focuses on the relationship between nutrients and health indicators, and the present study emphasizes the relationship between minerals in nutrients and hHcy risk.

In the traditional multifactorial model, the eight minerals associated with lower blood Hcy concentrations were calcium, phosphorus, potassium, magnesium, iron, zinc, copper, and manganese, whereas the eight minerals associated with lower hHcy risk were calcium, phosphorus, potassium, magnesium, iron, zinc, selenium, and copper. Traditional multifactorial models have been widely used in previous studies, with ease of data processing and ease of interpretability of results being the primary reasons. However, it is not appropriate to include each mineral simultaneously in one model because of the high correlation among these minerals. Therefore, the inclusion of only a single mineral in a model without considering the effects of the other minerals may cause the appearance of false positives or false negatives, with less confidence in the results. Furthermore, nonlinearities and interactions among minerals may exist, and the joint effect of mixed minerals on hHcy risk is not available in traditional models, with innovative methodologies needing to be used in this field as soon as possible.

WQS and Qg-comp are novel approaches that have been used to answer questions related to mixed exposure, and they allow for reporting the joint effect as well as the weighting of each exposure in the joint effect [32]. Both WQS and Qg-comp in the current study reported a negative association between higher mixed minerals’ intake and lower hHcy risk. The minerals with the highest negative weights in the WQS model are calcium and copper, and it is manganese in the Qg-comp model. It can be found that the two novel methods show a higher sensitivity to the results relative to the traditional model, after considering the influence of other minerals.

BKMR is another kind of machine learning method to analyze the relationship between mixed exposures and health. It additionally outputs univariate effects and interactions in parallel to reporting weights for joint effects and single exposures. The results of the BKMR model in the current study suggested that higher intakes of the nine minerals, except for sodium, were associated with a lower risk of hHcy and had a high weighting in the joint effect. Univariate effects findings also indicated that higher intakes of phosphorus, zinc, copper, and magnesium were associated with lower hHcy risk when levels of each of the remaining minerals were controlled. Also, there were some interactions between the ten minerals. Thus, the novel methods do more analyzing and interpreting on the data relative to the traditional models, elucidating the complex relationship between the mixed minerals and hHcy.

Previous studies have also used advanced statistical methods in exploring the relationship between mixed nutrients and health, despite the limited number of studies [33,34,35,36,37,38,39,40]. Li, RQ et al. [34] investigated the relationship between nutrient intake, inflammatory potential, and depressive symptoms in the elderly by using BKMR, identifying significant associations and revealing a collective impact of multiple nutrients on depression risk, with an offsetting effect between pro-inflammatory and anti-inflammatory diets. A Korean study evaluated the relationship between nutrients’ intake and MetS and its components in adults aged 19–80 years (n = 16807) using WQS regression, Qg-comp, and BKMR regression, and found that mixed nutrients’ intake was associated with a lower risk of MetS [33]. The results remained statistically significant after adjusting for basic social characteristics, smoking, alcohol consumption, and family history of disease. Minerals are an important group of nutrients, however, their joint role in the impact on health has not been sufficiently discussed nowadays. To the best of our knowledge, there is only one literature assessing the relationship between mixed minerals and health with advanced statistical methods [36]. It evaluated the relationship between six minerals (iodine, selenium, zinc, calcium, magnesium and iron) and maternal thyroid function in 489 pregnant women in Hangzhou, China. The results showed that mixed mineral concentrations were negatively correlated with TSH and positively correlated with FT3 and FT4, with iodine contributing the most. For the first time, the current study used three advanced statistical methods to find a negative joint association between mixed minerals and hHcy risk and to elucidate the existence of interactions among mixed minerals.

The normal metabolism of Hcy in the body typically maintains low levels. However, when Hcy metabolism is impaired, the concentration of Hcy in the blood increases, reaching a condition known as hHcy [41]. Disrupted Hcy metabolism can lead to a redox imbalance, heightened oxidative stress, endoplasmic reticulum stress, and altered DNA methylation, ultimately influencing the expression of genes associated with various diseases [42]. In the human body, Hcy is primarily converted through two pathways: the remethylation pathway and the trans-sulfuration pathway. The remethylation process involves the catalytic action of methionine synthase with its coenzyme vitamin B12, as well as betaine-Hcy methyltransferase. The trans-sulfurization pathway relies on cystathionine β-synthase and its coenzyme vitamin B6 for the production of cysteine. Factors affecting Hcy metabolism can indirectly contribute to its accumulation in the blood, resulting in hHcy.

Fortunately, some other literature has assessed the relationship between selected single minerals and hHcy and has elaborated on the possible mechanisms whereby minerals influence Hcy metabolism [14, 15, 43,44,45]. Elevating calcium intake was found to be associated with low hHcy risk in the multivariate regression, WQS regression, and BKMR in the current study. Paraoxonase 1 was involved in the metabolism of Hcy, which is a calcium ion-dependent enzyme [14]. Increasing calcium intake may promote Hcy metabolism through this pathway. It has also been found that blood calcium levels are positively correlated with Hcy [46], which is contrary to our conclusions and may be due to the fact that the correlation between blood calcium levels and dietary calcium intake may not be direct. The results of an 8-week zinc intervention trial in postmenopausal women implied that zinc supplementation may improve blood folate levels and reduce blood Hcy levels after zinc supplementation, and correlation analyses found a negative correlation between folate levels and blood Hcy levels after zinc supplementation [43]. The improvement of blood folate levels by a zinc intervention was also found in a study of an older Australian population (65–85 years) [47]. Betaine-Hcy methyltransferase, a zinc-containing catalase, plays an important role in the Hcy-to-methionine pathway [15], which may be explained as mechanisms by which zinc promotes Hcy metabolism. A study of an elderly Spanish population suggests that selenium in the blood may play a greater role in regulating Hcy levels [44], and similar connections were found in a British National Diet and Nutrition Survey [45]. This may be because the metabolic activity of methionine synthase is significantly reduced at low selenium levels, resulting in less Hcy methylation to methionine [44].

The current study has some distinct advantages. It is the first study to explore the relationship between the mixed minerals’ intake and blood Hcy levels, reporting the single and overall effects, and the contribution of each mineral to the joint effect. Moreover, three of the advanced machine learning methods were used, which provide a higher control over the complex interactions among minerals than traditional models, and the results are more interpretable. Meanwhile, we considered numerous covariates to adjust for in the model, and the results have a certain degree of confidence. However, there are limitations to the current study. First, this was a cross-sectional study and was unable to demonstrate the causal relationship between mixed minerals’ intake and blood Hcy levels or hHcy prevalence. Although it is widely recognized that the FFQ can reflect the dietary intake of participants in the long term, intake is similar to external exposure in risk assessment, which may not be as accurate as internal exposure, that is, blood mineral levels are unknown to us in this study. Therefore, perhaps in future studies, it may be possible to test the levels of various minerals in blood samples, first to see how the concentration of minerals in the blood relates to intake, and then to assess their relationship to blood Hcy, using a combined internal and external approach to give a more comprehensive explanation of the protective effect of minerals on blood Hcy. When considering the FFQ, it is important to acknowledge that the reliability and validity coefficients reported for the FFQ in our study were lower than desired. This might be due to the extended questionnaire entries and the lengthy recall time frames, which had made it challenging for participants to complete. Nevertheless, for a thorough evaluation of the population’s long-term dietary status, detailed food items and sufficiently lengthy recall time frames are indispensable. Additionally, sex imbalance in the current study should be mentioned. The proportion of females in the participants was higher than that of males, and the prevalence of hHcy was higher among males than females; sex bias was adjusted for as a covariate in models, but the effect on the results may have been substantial. Finally, the CFCT provides a relatively small variety of minerals, and although we have considered as many common minerals as possible, other minerals that we have not yet included may have some effect on outcomes. Overall, high intake of the ten minerals was associated with lower blood Hcy levels and lower hHcy risk, and different minerals had different weights in the joint effect.