A causal relationship between bone mineral density and breast cancer risk: a mendelian randomization study based on east Asian population

Abstract

Background

Breast cancer (BC) poses significant burdens on women globally. While past research suggests a potential link between bone mineral density (BMD) and BC risk, findings remain inconsistent. Our study aims to elucidate the causal relationship between BMD and BC in East Asians using bidirectional Mendelian randomization (MR).

Methods

Genetic association data for bone mineral density T-scores (BMD-T) and Z-scores (BMD-Z) (Sample size = 92,615) and BC from two different sources (Sample size1 = 98,283; Sample size2 = 79,550) were collected from publicly available genome-wide association studies (GWAS). Single-nucleotide polymorphisms (SNPs) associated with BMD-T and BMD-Z as phenotype-related instrumental variables (IVs) were used, with BC as the outcome. As the primary means of causal inference, the inverse variance weighted (IVW) approach was employed. Heterogeneity analysis was conducted using Cochran’s Q test, while MR-Egger regression analysis was implemented to assess the pleiotropic effects of the IVs. Sensitivity analyses were performed using methods such as MR-Egger, weighted median, and weighted mode to analyze the robustness and reliability of the results. The MR-PRESSO method and the RadialMR were used to detect and remove outliers. The PhenoScanner V2 website was utilized to exclude confounding factors shared between BMD and BC. Besides, the Bonferroni correction was also used to adjust the significance threshold. Then, the meta-analysis method was applied to combine the MR analysis results from the two BC sources. Finally, a reverse MR analysis was conducted.

Results

The results of the IVW method were consolidated through meta-analysis, revealing a positive correlation between genetically predicted BMD-T ($\:OR\:=\:1.22$, $\:95\%\:CI:\:1.13-1.33$, $\:P<0.001$) and BMD-Z ($\:OR\:=\:1.17$,$\:\:95\%\:CI:\:1.09-1.26$, $\:P<0.001$) with increased BC risk. The Cochran’s $\:Q$ test and MR-Egger regression suggested that neither of these causal relationships was affected by heterogeneity or horizontal pleiotropy. The sensitivity analyses supported the IVW results, indicating the robustness of the findings. Reverse MR analysis showed no causal relationship between BC and BMD.

Conclusion

Our MR study results provide evidence for the causal relationship between BMD and BC risk in East Asian populations, suggesting that BMD screening is of great significance in detecting and preventing BC.

Background

Low bone mineral density (BMD) is a key feature of osteoporosis, which subsequently increases the risk of fracture [1]. BMD shows a concomitant decline with age. About 50% of women suffer from fragility fracture after the age of 50 [2]. Postmenopausal women often experience osteoporosis, and they experience a decline in estrogen levels, leading to a weakening of bone synthesis and metabolism [3]. Low estrogen levels also affect the proliferation and differentiation of immune cells, thereby influencing the function of bone cells [4]. Breast cancer (BC) represents a substantial threat to global public health, ranking as a primary cause of mortality among female populations. In the past few decades, the incidence of BC in Asian countries has been gradually increasing and may ultimately surpass that in Western countries [5]. In East Asian countries and regions such as China, Japan, and Singapore, BC is more prevalent among cancers and has become the second leading cause of cancer-related mortality among women [6]. However, age is also a significant risk factor for BC, with the incidence highly correlated with age, and the risk of developing BC increases as people get older [7]. Early diagnosis or prevention of BC can help reduce mortality rates, thus highlighting the crucial importance of BC screening [8].

Currently, it appears that both osteoporosis, characterized by low BMD, and the risk of developing BC increase with age. However, previous studies have yielded inconsistent results regarding whether there is an association between BMD and BC risk. A study showed that women diagnosed with BC clinically had higher bone mineral density compared to women without BC who had similar characteristics, suggesting that BMD may be a biomarker for BC risk [9]. A previous prospective study based on elderly French women found that older women with high BMD had a higher risk of developing BC compared with those with low BMD [10]. Similar results were also observed in a prospective study among women over 65 years old [11]. However, a cross-sectional study of women over 28 years old found no association between hip or spinal BMD and the BC risk [12]. A long-term follow-up study conducted in Austria showed that BMD could not predict the BC risk among young postmenopausal women [13]. Additionally, a prospective study based on early postmenopausal women in France found there is no significant association between BMD measured in the post-menopausal years and BC risk [14]. The current evidence is inconclusive regarding the potential link between BMD and the risk of developing BC. Furthermore, these observational studies cannot explain the causal relationship between the two. A Mendelian randomization (MR) study by Wu et al. [15], based on individuals of European descent, suggested that BMD is not the risk factor for BC, rather, BC can lead to decreased bone density. Most of the observational studies collected on BMD and the risk of BC are based on European or Caucasian populations, while the incidence and mortality rates of BC vary significantly in different regions of the world [16]. It remains unknown whether similar results can be obtained from samples of different ethnic groups. Therefore, it remains unclear whether there is a causal relationship between BMD and BC in East Asian populations.

The research methodology of MR serves to elucidate the causal link between exposure and outcome. In MR studies, single nucleotide polymorphisms (SNPs), recognized as genetic variations, are employed as instrumental variables (IVs) to investigate the potential causal link [17]. Due to the random allocation of genes at conception and their independence from external environmental and social factors, MR circumvents confounding factors and reversed causality encountered in observational studies, thereby offering a relatively precise epidemiological approach [18]. The Genome-wide Association Study (GWAS) serves as a method to identify gene variations linked to intricate human diseases or traits. Its ultimate goal is to expose the influence of genetic variations on the predisposition to complex diseases, thus offering fresh insights and approaches for preventative measures, diagnosis, and treatment [19]. Within this investigation, we explored the relationships between BMD and BC using a two-sample MR study through GWAS summary data.

Materials and methods

Accessibility of data

The entire data utilized in this research is publicly accessible and available on designated websites. The exposure data came from a GWAS published in 2023 [20], a prospective study that incorporates multi-omics genomic data along with longitudinal phenotypic and environmental measurements. This study involved 102,900 individuals from the population of Taiwan, discovering 36 quantitative traits and identifying 968 novel loci. Among these, we selected Bone mineral density T-score (BMD-T) and Bone mineral density Z-score (BMD-Z) as our exposures, with a sample size of 92,615. All the data applied in this study can be accessed at the web address provided in Table 1.

Table 1 Sample sources in MR studies

We utilized two GWAS specific to East Asian populations as sources for our outcome data. One GWAS was conducted in 2020 among 212,453 Japanese participants, covering 42 diseases [21]. The participants were recruited from 66 affiliated hospitals of 12 medical institutions between 2003 and 2007. Among these, we selected BC as outcome, which included 5,552 cases and 89,731 controls, totaling a sample size of 95,283. The second dataset for BC was derived from an expanded GWAS conducted in 2021 [22], which examined 220 traits and included 108 phenotypes that had never been GWAS-assessed in East Asian populations. Within this dataset, there were 6,325 cases of BC and 73,225 controls (Table 1). The identification of BC cases adhered to the coding standards of the International Classification of Diseases (ICD-10). Notably, the datasets pertaining to BMD-T, BMD-Z and BC did not exhibit any overlap in samples.

Study design

We followed the STROBE-MR guidelines (STROBE-MR: Supplement) [23]. Using BMD-T and BMD-Z as exposure factors, BC as the outcome, and genetic variations as IVs, we performed two-sample MR analyses successively to assess the association between genetically predicted exposures and genetic susceptibility to the outcome. In a reverse MR analysis, the GWAS for BC was used as the exposure to explore its relationship with BMD.

To properly interpret the MR analysis, three core assumptions need to be satisfied [24]. Assumption 1: There must be a strong correlation between the genetic variants and the exposure. Assumption 2: The genetic variants should be unaffected by any confounding factors that could potentially skew the association between the exposure and the outcome. Assumption 3: The influence of the genetic variants on the outcome should be mediated solely through the exposure. These assumptions are crucial for the validity of the MR analysis, ensuring that the observed associations are due to the genetic effects on the exposure, rather than being confounded by other factors or having direct effects on the outcome. (Fig. 1)

Fig. 1

Overview of the overall MR design. Assumption 1, instrument variables are robustly related to exposure; Assumption 2, instrument variables are not related to confounders; Assumption 3, instrument variables are related to outcome only through exposure. BMD-T, Bone mineral density T-score; BMD-Z, Bone mineral density Z-score; BMI, body mass index; WHR, waist-hip ratio; HDL-C, high-density lipoprotein cholesterol; IVW, inverse variance weighted; MR, Mendelian randomization

Selecting methods for instrumental variables

Genetic variants, known as SNPs, are extracted as IVs from relevant GWAS studies (Table 1). The selection of these IVs must satisfy the following criteria:

(1) SNPs must have a strong association with the exposure at the genome-wide significance level, with a set threshold of $\:P\:<\:5\times\:{10}^{-8}$.

(2) SNPs in linkage disequilibrium are removed, using a threshold of$\:\:{R}^{2}\:=\:0.001$ and a genetic distance of 10,000 kb. This means SNPs with an $\:{R}^{2}$ value greater than 0.001 within a 10,000 kb range of the most significant SNPs are excluded.

(3) SNPs significantly associated with the outcome are filtered out, using a threshold of $\:P\:=\:5\times\:{10}^{-5}$.

(4) SNPs with palindromic structures are excluded when integrating exposure and outcome data.

(5) SNPs with an F-statistic less than 10 are removed to exclude weak IVs. Based on traditional experience, when $\:F\:>\:10$, the bias caused by weak IVs is relatively small [25]. Weak instruments, while associated with the exposure, provide low statistical power to test hypotheses due to their low ability to explain the exposure, potentially leading to imprecise estimates of causal effects and increasing the probability of type I errors [26]. The formula for calculating the F-statistic for a single SNP is: $\:F\:=\:{\beta\:}^{2}/{se}^{2}$ [27], where$\:\:\beta\:$ denotes the magnitude of the SNP’s impact on the exposure, and $\:se$ signifies the standard error associated with this effect size.

Statistical analysis

The primary methodology chosen for causal inference is inverse-variance weighting (IVW). This method assumes that all SNPs are effective and do not exhibit horizontal pleiotropy, leading to high statistical power. However, in the presence of horizontal pleiotropy among SNPs, significant bias may occur in the results [28]. The Cochran $\:Q$ test is used to assess the heterogeneity across the chosen SNPs. This test calculates the $\:Q$ statistic by summing the squared differences from the mean, each weighted by its standard deviation. A significant $\:P$-value of less than $\:0.05$ indicates the presence of heterogeneity among the SNPs, leading to the use of a random-effects model to determine the causal link. In contrast, a fixed-effects model is used [29]. Genetic pleiotropy can influence the estimation of association effects. To detect the existence of horizontal pleiotropy, MR-Egger regression is employed as a diagnostic tool. If the intercept term in the MR-Egger regression model is zero ($\:P\:>\:0.05$), it indicates that horizontal pleiotropy is absent [30]. Additionally, the weighted median method, the MR-Egger method and the weighted mode are employed to assess the robustness of the analytical results. The MR-Egger method is predominantly employed in MR causal inference when there is potential horizontal pleiotropy among genetic variants [30]. The weighted median approach mandates that no less than 50% of the allocated weight must be derived from IVs. When there exists heterogeneity but no pleiotropy among genetic variants, it’s the best choice [31]. The weighted mode identifies multiple variables as effective IVs to detect similar causal relationships [32]. When IVW shows significant results ($\:P\:<\:0.05$), and the other three methods align with the direction of IVW, it can be considered that there is a causal relationship.

The MR pleiotropy residual sum and outlier (MR-PRESSO) method [33]and the RadialMR [34] approach are employed to detect outliers. If outliers are identified, they are excluded, and the causal association will be re-estimated. Furthermore, to minimize the interference of horizontal pleiotropy on the results, we manually search for each SNP in the human genotype-phenotype database, PhenoScanner V2 [35], to identify and exclude confounding factors that affect both BMD and BC, such as body mass index (BMI) [36, 37], high-density lipoprotein cholesterol (HDL-C) [38, 39]. SNPs exhibiting genome-wide significant associations ($\:P\:<\:5\times\:{10}^{-8}$) pertaining to these confounders are eliminated from consideration, and causal inference is re-conducted. The meta-analysis approach is utilized to combine the MR analysis results from diverse data sources. The I² statistic is calculated to assess the heterogeneity among the results from different data sources. An I² value greater than 50% indicates heterogeneity, then a random-effects model is adopted. Conversely, an I² value less than 50% suggests homogeneity, then a fixed-effects model is employed [40].

In this study, MR analysis was conducted successively on the genetic predictions of BMD-T and BMD-Z in relation to BC. To reduce the probability of false positive results arising from multiple hypothesis testing, the Bonferroni method was used to adjust the significance threshold in the results obtained from the primary IVW approach. A significant causal relationship was indicated when $\:P\:<\:0.025$ ($\:0.05/2$). We used Manhattan plot to display the association results of each SNP or gene locus. The above methods were performed using R 4.3.2. The statistical analysis was conducted using the Mendelian Randomization R Package (version 0.9.0), Two Sample MR (version 0.5.8) and meta (version 6.5.0). We utilized the R packages MRPRESSO (version 1.0) and RadialMR (version 1.1) to remove outliers. Data visualization was performed using the R Package forestploter (version 1.1.1) and CMplot (version 4.5.1).

Results

Characteristics of instrumental variables

After initial screening, 99 to 104 SNPs were included in each analysis (Supplementary Tables 1–4). After removing outliers and excluding confounding SNPs using the PhenoScanner V2 website, the final numbers of SNPs included in the MR analyses for BMD-T with two BC outcomes were 90 and 89, respectively, while the number of SNPs included in the MR analyses for BMD-Z with the same two BC outcomes was 92 for both. The screening process is detailed in Fig. 2. The association of each SNP or gene locus were showed in Fig. 3.

Fig. 2

Steps for selecting instrumental variables. BMD-T, Bone mineral density T-score; BMD-Z, Bone mineral density Z-score; SNP, single nucleotide polymorphisms

Fig. 3

Manhattan plot. A: Manhattan plot of$\:\:-log10$ values using GWAS summary statistics of BMD-T. B: Manhattan plot of$\:\:-log10$ values using GWAS summary statistics of BMD-Z. The horizontal axis represents the chromosome number. The dashed line indicates the $\:P<5\times\:{10}^{-8}$ threshold. GWAS, genome-wide association study; BMD-T, Bone mineral density T-score; BMD-Z, Bone mineral density Z-score

Causal association with BMD and BC

The primary analytical method, IVW, revealed a positive association between genetically predicted BMD-T ($\:OR\:=\:1.23$, $\:95\%\:CI:\:1.10-1.39$, $\:P\:<\:0.001$) and BMD-Z ($\:OR\:=\:1.17$, $\:95\%\:CI:\:1.05-1.30$, $\:P\:=\:0.004$) with the risk of BC (bbj-a-160). Similarly, BMD-T ($\:OR\:=\:1.22$, $\:95\%\:CI:\:1.09-1.36$, $\:P\:<\:0.001$) and BMD-Z ($\:OR\:=\:1.17$, $\:95\%\:CI:\:1.06-1.30$, $\:P\:=\:0.002$) were also positively associated with the risk of BC (ebi-a-GCST90018579). It is worth mentioning that both the MR-Egger method ($\:OR\:=\:1.41$, $\:95\%\:CI:\:1.02-1.95$, $\:P\:<\:0.05$) and the weighted median method ($\:OR\:=\:1.25$, $\:95\%\:CI:\:1.03-1.52$, $\:P\:<\:0.05$) showed significant associations between BMD-T and BC(bbj-a-160) risk. Both the MR-Egger ($\:OR\:=\:1.37$, $\:95\%\:CI:\:1.03-1.81$, $\:P\:<\:0.05$) and weighted median ($\:OR\:=\:1.22$, $\:95\%\:CI:\:1.02-1.45$, $\:P\:<\:0.05$) also demonstrated a correlation between BMD-Z and BC (ebi-a-GCST90018579). The findings derived from the IVW approach were corroborated by the weighted median, MR-Egger, and weighted mode (Fig. 4). Scatter plots for each MR analysis are presented in Fig. 5.

Fig. 4

Utilizing diverse methodologies for MR analysis to investigate the relationship between BMD and BC. BMD-T, Bone mineral density T-score; BMD-Z, Bone mineral density Z-score; SNPs, single nucleotide polymorphisms; IVW, inverse-variance weighting; OR, odds ratio; CI, confidence interval

Fig. 5

Scatter plots of the MR analysis between BMD and breast cancer. A: Scatter plot of the MR analysis between exposure to BMD-T and outcome of breast cancer (bbj-a-160); B: Scatter plot of the MR analysis between exposure to BMD-T and outcome of breast cancer (ebi-a-GCST90018579); C: Scatter plot of the MR analysis between exposure to BMD-Z and outcome of breast cancer (bbj-a-160); D: Scatter plot of the MR analysis between exposure to BMD-Z and outcome of breast cancer (ebi-a-GCST90018579)

It is evident that the P-values of the IVW for each MR analysis are all less than the threshold corrected by the Bonferroni method ($\:P=0.025$), thus the causal relationship between BMD and BC remains statistically significant.

The results of the MR analysis between BMD-T and BC from two data sources were combined using the meta-analysis method. The heterogeneity test $\:{I}^{2}$ was less than 50%, so a fixed-effects model was adopted. The results still showed a positive causal relationship between BMD-T and BC risk ($\:OR=1.22$, $\:95\%\:CI:\:1.13-1.33$, $\:P<0.001$) (Fig. 6A). Similarly, the results of the MR analysis between BMD-Z and BC from the same two data sources were combined. The heterogeneity test $\:{I}^{2}$ was also less than 50%, thus a fixed-effects model was used. The results again demonstrated a positive causal relationship between BMD-Z and BC ($\:OR=1.17$, $\:95\%\:CI:\:1.09-1.26$,$\:\:P<0.001$) (Fig. 6B).

Fig. 6

A: Meta-analysis results of BMD-T and breast cancer from two data sources after MR analysis. B: Meta-analysis results of BMD-Z and breast cancer from two data sources after MR analysis

Sensitivity analysis

The Cochran $\:Q$ test did not detect any heterogeneity, therefore, a fixed-effects model was employed for the IVW analysis. The P-values of the MR-Egger intercept test for each analysis did not provide evidence of horizontal pleiotropy (Table 2).

Table 2 Results of heterogeneity and pleiotropy tests for instrumental variables

Reverse mendelian randomization analysis

When performing MR analysis with two GWAS-derived BC phenotypes as exposures and BMD-T and BMD-Z as outcomes, no evidence of causality was found between them (Fig. 7). The $\:P$-values from both the Cochran $\:Q\:$test and the MR-Egger intercept test exceeded the threshold of 0.05, suggesting the absence of heterogeneity or horizontal pleiotropy.

Fig. 7

Reverse Mendelian randomization analysis. SNPs, single nucleotide polymorphisms; BMD-T, Bone mineral density T-score; BMD-Z, Bone mineral density Z-score; IVW, inverse-variance weighting; OR, odds ratio; CI, confidence interval

Discussion

In this study, we utilized the GWAS summary statistics of BMD and BC in East Asian populations for analysis. Through various MR methods (IVW, MR Egger, Weighted median, Weighted mode), we analyzed and investigated the relationship between BMD and the risk of BC in East Asian populations. It has been ultimately shown that genetically predicted BMD-T and BMD-Z exhibit a significant positive causal relationship with BC risk. Additionally, a bidirectional MR analysis was conducted, and the resulting evidence indicated that BC had no causal impact on BMD. Among these methods, the IVW approach is the primary one in MR studies, and its results indicate a significant positive correlation between genetically predicted BMD-T and BMD-Z with the risk of BC. The results of this study differ from those of previous research examining the causal relationship between the two. Furthermore, a sensitivity analysis was conducted, and no heterogeneity or pleiotropy was detected, demonstrating the robustness of the results. Additionally, reverse MR analysis revealed that BC did not have a significant impact on BMD. This further enriches and complements the findings of previous studies. As our MR analysis selected two GWAS datasets for BC as outcomes, we combined the MR analysis results of BMD T-scores and Z-scores for these two sources through meta-analysis. The results still indicate a causal relationship between BMD and BC risk. MR, distinct from observational studies, utilizes genetic variants that satisfy the IV assumptions to investigate causal relationships in epidemiological research, thereby eliminating the influence of confounding factors as well as biases arising from reverse causality and other factors [17]. Furthermore, compared to randomized controlled trials, MR greatly reduces research costs and is more feasible, holding significant implications for disease prevention and treatment strategies.

Prior observational investigations point to a potential link between BMD and BC, yet controversies have persisted among these studies. A prospective cohort study conducted in North America involving 8,203 postmenopausal women found that women exhibiting higher BMD seemed to be at a higher risk for developing BC [41]. Grenier et al. [42] observed a link between elevated BMD in the vertebra lumbalis and collum femoris and a heightened risk of BC among women aged 50 and above. A retrospective study conducted in southern Israel observed that women diagnosed with BC had higher mean BMD Z-scores across all skeletal sites compared to women without BC [43]. A case-control study conducted by Kim et al. on Korean women found that postmenopausal women with BC had significantly higher BMD in the femoral neck and lumbar vertebra compared to the control group, indicating a relationship between higher BMD in these regions and a higher risk of BC. This finding also provides us with inspiration to consider BMD as a potential predictive factor for BC [44]. Although these observational studies cannot establish causality, they provide evidence for a relationship between BMD and BC. However, there remains considerable variability among studies. A long-term cohort study conducted in Austria with a follow-up of 20.7 years did not find a positive correlation between BMD and BC risk [13], The systematic review conducted by Chen et al. noted that no association was found between BMD and BC risk, suggesting a lack of definitive evidence for BMD as an indicator of BC risk [45]. Prior MR analysis in European populations revealed no genetic evidence supporting a causal link between BMD and BC risk. The observed association between the two was explained by the pleiotropic effects of genetic variants [15, 46]. This is inconsistent with the results of our MR analysis. We have analyzed the GWAS data of East Asian population and concluded that there is a positive causal relationship between BMD and the risk of BC, which is a brand-new discovery and has enlightening significance for the screening and prevention of BMD and breast cancer in East Asian population. This difference may be caused by a variety of factors. First, genetic factors play an important role in the susceptibility to diseases among different ethnic groups, thus the ethnic differences may be the primary reason for the discrepancy between this study and the MR analysis on European populations. Second, environmental factors, including dietary habits, lifestyle, and climate, can also affect the incidence of diseases in populations from different regions. In addition, socio-economic status is also an important factor affecting the incidence of diseases. Therefore, more research and conclusive evidence are necessary to establish the causal relationship between BMD and BC risk, which holds significance for early screening and prevention of the disease.

Several factors can explain the relationship between BMD and BC. BMD serves as a pivotal metric in evaluating bone quality, assessing the likelihood of osteoporotic fractures, and tracking the inherent evolution and sustenance of bone mass. Currently, BMD is considered a marker of lifelong estrogen exposure [47]. The maintenance of BMD is closely associated with the antagonistic roles played by estrogen receptors ERα and ERβ in the body [48], and Estrogen plays a significant role in maintaining bone integrity, and its deficiency can expedite bone loss, reduce BMD, and even lead to osteoporotic fractures [49]. Estrogen serves as a crucial regulatory factor in maintaining breast development and function, and it is also a key player in the occurrence and progression of BC. Excessively high levels of estrogen can increase the risk of developing BC [50]. Furthermore, studies have demonstrated that estrogen can upregulate the anti-apoptotic proto-oncogene Bcl-2 to inhibit the apoptosis of cancerous cells in mammary gland, which plays a crucial role in the progression and proliferation of BC [51]. A prior investigation revealed that women exhibiting lower bone turnover rates, quantified by CTX-1 and osteocalcin, are at a higher risk of developing cancer [52]. Notably, the bone turnover markers CTX-1 and osteocalcin are both negatively correlated with BMD, and they can reflect the level of BMD [53]. However, the underlying mechanisms require further exploration. Additionally, among BC patients, lower BMD increases the risk of bone metastasis of BC [54]. During BC treatment, adjuvant chemotherapy increases the risk of osteoporosis in postmenopausal women, which is accompanied by a decrease in estrogen levels [55]. Research has found that chemotherapy is associated with a high risk of low BMD among postmenopausal BC survivors [56], suggesting that it is meaningful to further monitor BMD and maintain normal bone mass during the prognosis process of BC patients.

The strength of this study lies in exploring the causal relationship between BMD and BC risk through MR analysis. The findings of this study represent a novel discovery, providing robust evidence for the association between BMD and BC and offering new insights for the prevention and early screening of BC. Compared to previous observational studies on BMD and BC, MR studies are less susceptible to confounding factors and reverse causality. To enhance the credibility of the research findings, various MR methods, meta-analyses and sensitivity analyses were employed. The findings of this study suggest that doctors should be vigilant in screening and preventing BC for patients with high T-scores or Z-scores of BMD. For patients who have already been diagnosed with BC, monitoring BMD during treatment to maintain bone health and prevent osteoporosis is also essential.

Nonetheless, this study is subject to certain limitations that merit consideration. Firstly, despite various sensitivity analyses have been conducted, the presence of horizontal pleiotropy still cannot be entirely ruled out. Secondly, the BMD GWAS data used in this study in East Asian populations is relatively limited, with a smaller sample size and number of cases compared to European populations. Although we have selected and analyzed all appropriate GWAS summary data of BC in East Asian populations, we still hope that in the future, GWAS summary data with larger sample sizes, more cases, and more diverse ethnic groups will be established to help us further explore the deeper connections between diseases. Thirdly, due to the lack of specific GWAS data on age and BC subtypes, we were unable to perform stratified analyses for age and different BC subtypes.

Future research could further expand the GWAS databases of Asian and other populations to further explore the relationship between BMD and BC, and other diseases in different populations. Additionally, while MR studies can provide robust causal evidence, their results still need to be validated in larger-scale clinical trials. Future research should further explore the underlying biological mechanisms of the association between BMD and BC, and strive to explain the reasons for the controversies observed in different ethnic groups, which will provide references for BC prevention and treatment worldwide.

Conclusion

In conclusion, our study provides support for the relationship between BMD and BC. Our evidence suggests a positive causal relationship between genetically predicted BMD-T and BMD-Z and BC risk in East Asian populations. Therefore, patients with high or above-normal BMD should be attentive to BC screening and prevention. Additional research is crucial to gain a comprehensive understanding of the underlying mechanisms and offer valuable insights for BC prevention across diverse ethnic populations.