BMI to BC: SNPs from publicly available BMI databases (p < 5×10− 8) were extracted. Additionally, SNPs (r^2 ≥ 0.001, clumping window ≤ 10,000 kb) were excluded, retaining only eligible SNPs. This method resulted in the extraction of 152 eligible SNPs. Upon merging with BC data, it was found that 149 of the same SNPs overlapped. Subsequently, after aligning allele genes and removing palindrome sequences, a total of 145 eligible SNPs meeting the criteria remained (rs11074452, rs1558902, rs809439, rs815610 palindrome sequences were removed). After removing 3 SNPs (rs13021737, rs6265, rs7629375) related to smoking behavior, 142 SNPs remained, ensuring that the F-statistic for each SNP was > 10 (with the minimum F-statistic exceeding 29), ensuring their strong relevance. There exists a positive causal relationship between BMI and BC, with IVW as the most important indicator (OR = 1.655, 95% CI = 1.122–2.441, p = 0.011). Additionally, the Weighted Median method (OR = 1.842, 95% CI = 1.022–3.319, p = 0.041) yielded the same results. There is no evidence of heterogeneity in the Heterogeneity statistics (IVW Q = 153.724, p = 0.219; MR Egger Q = 153.647, p = 0.203), and no horizontal pleiotropy was detected in the Horizontal pleiotropy test (Egger intercept = 0.003, p = 0.791,Fig. 3). Furthermore, we conducted Mendelian Randomization Pleiotropy RESidual Sum and Outlier (MR-PRESSO) test and contamination mixture detection. MR-PRESSO was used to identify outlier SNP that may affect causal estimates due to horizontal pleiotropy. The results of MR-PRESSO did not find any abnormal SNPs, and the Global TEST's P value was 0.222, indicating the absence of pleiotropy(Table 2). The contamination mixture represents the contribution of each genetic variant to being an effective instrument or an invalid instrument. MRConMix yielded a result of p = 0.022, further corroborating the previous conclusion.
BMI to CRP: Similarly, using the same method (p < 5×10− 8, r^2 ≥ 0.001, and clumping window ≤ 10,000 kb), we merged the 152 eligible SNPs extracted for BMI with CRP data, resulting in the same 146 SNPs. Subsequently, after aligning allele genes and removing palindrome sequences, 137 eligible SNPs meeting the criteria remained (removing rs10883521, rs11074452, rs1558902, rs17016673, rs4776970, rs6471932, rs809439, rs815610, rs9835772). Further removal of SNPs related to inflammation (rs13107325, rs7143963) resulted in a total of 135 SNPs, ensuring that the F-statistic for each SNP was > 10 (with the minimum F-statistic exceeding 29), confirming their strong relevance. This indicates a positive causal relationship between BMI and CRP, with IVW (OR = 1.237, 95% CI = 1.175–1.304, p = 9.417×10^-16). Additionally, the Weighted Median method (OR = 1.216, 95% CI = 1.128–1.312, p = 3.603×10^-7), and MR Egger method (OR = 1.381, 95% CI = 1.179–1.618, p = 1.055×10^-4) yielded consistent results. There is no evidence of heterogeneity in the Heterogeneity statistics (IVW Q = 150.050, p = 0.162; MR Egger Q = 147.75, p = 0.180), and no horizontal pleiotropy was detected in the Horizontal pleiotropy test (Egger intercept = -0.002, p = 0.153,Fig. 4). In MR-PRESSO, the Global TEST's P value was 0.166, indicating the absence of pleiotropy(Table 2).
CRP to BC: We used the same method (p < 5×10 − 8, r^2 ≥ 0.001, and clumping window ≤ 10,000 kb) to extract eligible CRP SNPs, resulting in 38 SNPs. These were then merged with BC SNPs, resulting in a total of 37 SNPs. Subsequently, after aligning allele genes and removing palindrome sequences, 30 eligible SNPs meeting the criteria remained (removing rs12675647, rs142393537, rs183157863, rs190946753, rs200241668, rs73074854, rs78934647). No SNPs related to smoking were present, and each SNP had an F-statistic > 10 (with the minimum F-statistic exceeding 29), confirming their strong relevance. The IVW method, as the primary analysis, indicates a positive causal relationship (OR = 1.401, 95% CI = 1.017–1.929, p = 0.038), and the OR values from MR Egger, Weighted Median, Simple Mode, and Weighted Mode are consistent in direction. There is no evidence of heterogeneity in the Heterogeneity statistics (IVW Q = 20.694, p = 0.870; MR Egger Q = 20.456, p = 0.847), and no horizontal pleiotropy was detected in the Horizontal Pleiotropy test (Egger intercept = 0.015, p = 0.629,Fig. 5). In MR-PRESSO, no outlier SNPs were found, and the Global TEST's P value was 0.898, further confirming the absence of pleiotropy(Table 2).
Table 2
The results of the heterogeneity and horizontal pleiotropy
Method | SNP | Heterogeneity statistics | Horizontal pleiotropy | MR-PRESSO(Global TEST) |
BMI-BC | 142 | 0.219 | 0.791 | 0.222 |
BMI-CRP | 134 | 0.162 | 0.153 | 0.166 |
CRP-BC | 30 | 0.840 | 0.629 | 0.898 |
In the previous step, the two-sample Mendelian randomization method could only suggest that CRP is a potential mediator. Finally, we used the MVMR method to demonstrate that CRP is the mediator through which BMI influences BC, and calculated its numerical value. When controlling for CRP, it was found that there is no causal relationship between BMI and BC (OR = 1.413, 95% CI = 0.959–2.081, p = 0.079). However, when controlling for BMI, it was still observed that there is a positive causal relationship between CRP and BC (OR = 1.434, 95% CI = 1.042–1.973, p = 0.026). Originally, the IVW value for BMI to BC was significant, but after including CRP through MVMR, the effect of BMI on BC decreased, providing stronger evidence that CRP is the mediator(Fig. 1,2). According to the calculation, the mediation effect is 15.9% (OR = 1.079, 95% CI = 1.009–1.161).