We searched the MR Base database (http://www.mrbase.org/) (14), which houses a large collection of summary statistic data from hundreds of genome‐wide association studies (GWASs). We used the publicly available summary statistics data sets of GWAS meta‐analyses for BMI in individuals of European descent (n = 77482; Genetic Investigation of Anthropometric Traits consortium (GIANT consortium); https://gwas.mrcieu.ac.uk/datasets/ebi-a-GCST90095041/) (15) as the exposure. The genetic association studies databases for HL can be accessed on the FinnGen website (https://www.finngen.fi/en) (16). The genetic data for non-Hodgkin’s lymphoma (NHL) encompassed various subtypes, such as diffuse large B-cell lymphoma (DLBCL), follicular lymphoma (FL), mature T/NK-cell lymphomas, and other unspecified NHL types. This information has been rephrased for conciseness. These GWAS datasets are openly accessible and downloadable from the OPEN GWAS website (https://gwas.mrcieu.ac.uk/) (17). Table 1 presented the relevant information for the outcomes.

In conducting MR studies, adherence to three fundamental assumptions is imperative: 1) a robust correlation between instrumental variants (IVs) and exposure variables exists; 2) the IVs are free from any influence of confounding factors about the exposure-outcome association; 3) genetic variants solely impact the outcome via the exposure route, excluding other pathways (18). The core assumptions depicted in Figure 1 were addressed by following a systematic approach. Initially, IVs with genome-wide significance (P <5e–8) were extracted from the BMI GWAS. Subsequently, linkage disequilibrium (LD) was considered with a threshold of r² = 0.001 and a clumping distance of 10,000. Proxy single nucleotide polymorphisms (SNPs) with LD greater than 0.8 were then identified, while palindromic SNPs were excluded while harmonizing the BMI and HL GWAS datasets. Finally, the MR Pleiotropy RESidual Sum was employed to detect potential outlier SNPs and correct for any horizontal pleiotropy (19). The remaining SNPs were utilized for MR analysis in the subsequent stage. The F-statistic was calculated for each of the remaining SNPs to determine the strength of the genetic instruments. An F-statistic exceeding 10 signifies a robust genetic instrument essential for trustworthy MR analysis.

We employed five methods to tackle the issue of horizontal pleiotropy. These methods consisted of the inverse-variance weighted (IVW) model, weighted median (WM), MR-Egger, simple mode, and weighted mode. The primary analyses in this study utilized the multiplicative random-effects IVW method, known for providing accurate estimates when all SNPs are considered valid instruments. The weighted median method can also generate consistent estimates if more than 50% of the weight comes from valid instrument variants (20). The MR-Egger regression model was represented by the following formula:

In this equation, the components represented the following meanings: β_IV : Effect size estimate for the instrumental variable (IV); α: Intercept term represented the average pleiotropic effect across all genetic variants; β: Slope term represented the causal effect estimate; SE: Standard error of the exposure; ε: Error term accounting for residual pleiotropy. MR-Egger regression offered estimates while accounting for horizontal pleiotropy, albeit with slightly reduced precision (21). MR-Egger was considered to be supportive when the effect estimate was consistent with MR-IVW. Through the utilization of these various techniques, we were successful in identifying and addressing the potential effects of horizontal pleiotropy in our study. The β Value acted as the effect size for determining the direction of causality, where β > 0 signifies the exposure as a risk factor for the outcome. Statistical significance was indicated by P< 0.05. The heterogeneity among genetic IVs was evaluated using Cochran’s Q test. Sensitivity analysis predominantly utilized the leave-one-out method. The causal link between BMI and lymphoma remained robust even when systematically removing one SNP at a time. The MR-Egger intercept test was employed to detect directional pleiotropy, with a P< 0.05 indicating its presence. Funnel plots were employed to assess directional pleiotropy, similar to their role in meta-analysis for detecting publication bias. The MR analysis was carried out using R software (version 4.3.2) with the package “TwoSampleMR” (version 0.5.10). The study’s flowchart is illustrated in Figure 2 .

A retrospective study was conducted at Fujian Medical University Union Hospital from 2020 to 2021, involving 124 newly diagnosed HL patients [Diagnostic criteria reference Hodgkin lymphoma: 2023 update on diagnosis, risk-stratification, and management (3)]. The validation study protocol was approved by the ethics committee of Fujian Medical University Union Hospital (Research project ethics approval number: 2021KJCX053). Our study excluded any personally identifiable information, and utilized BMI data gathered during routine clinical procedures. Minimal risks were associated with the study, safeguarding the subjects’ rights and interests and eliminating the necessity for individual informed consent. Control group data was sourced from the National Center for Health Statistics (NHANSE). NHANES, conducted by the National Center for Health Statistics of the Centers for Disease Control, is a comprehensive survey that gathers health and nutrition data from a nationally representative sample of individuals in the United States. The survey employs a rigorous stratified and multi-stage probability cluster design to ensure accurate and reliable results. All participants in the survey have given their informed consent to participate. In this study, we analyzed publicly available data from NHANES collected between 2009 and 2010. We randomly selected 125 people to be included in the external verification control group. Data analysis was performed using R software (version 4.3.2), and the correlation between BMI and HL was confirmed through a Logistic regression analysis. Statistical significance was established at P< 0.05. The study’s methodology is delineated in Figure 3 .

We have selected 9 independent SNPs (rs10876528, rs1106529, rs1294409, rs1482852, rs224333, rs459193, rs7251505, rs7766106, rs979012) from GWASs on BMI as IVs ( Table 2 ). For further details regarding these SNPs, researchers can refer to the National Center for Biotechnology Information (NCBI) Single Nucleotide Polymorphism Database (dbSNP: https://www.ncbi.nlm.nih.gov/snp/) (22).

The relationship between BMI and HL has been confirmed through collective causal assessments ( Table 3 ). Individuals with a higher BMI show a 91.65% reduced likelihood of developing HL (OR_IVW = 0.084, 95% CI 0.015 – 0.473, P = 0.005). Other Mendelian Randomization models also support this causal effect (OR_WM = 0.092, 95% CI 0.009 – 0.945, P = 0.045; OR_MR-Egger = 0.011, 95% CI 0.001 – 749.53, P = 0.455; OR_Simple-mode = 0.088, 95% CI 0.003 – 2.629, P = 0.198; OR_{Weighted-mode} = 0.094, 95% CI 0.002 – 3.656, P = 0.241). The consistency in the causal direction shown by these methods is remarkable, even though not all analysis methods yielded statistically significant results. However, with the P-values of IVW and WM both below 0.05, we maintain confidence in the MR analysis confirming the association between BMI and HL. Graphical representations like forest plots ( Figure 4 ) and scatter plots ( Figure 5 ) further illustrate the increased risk of HL in individuals with lower BMI. These findings were based on SNPs from the BMI GWAS and causal estimates from MR models.

The results of the statistical analysis showed no significant heterogeneity in the study on MR, with P-values of 0.564 for P_MR-Egger and 0.656 for P_IVW. The MR-Egger intercept test indicated a p-value above 0.05, suggesting the absence of gene-directional pleiotropy. The sensitivity analysis, conducted through the leave-one-out method illustrated in Figure 6 , consistently supported the conclusion on the causal relationship between BMI and HL, even when individual SNPs were excluded. Additionally, the symmetrical funnel plot in Figure 7 indicated the absence of directional pleiotropy.

Malignant lymphomas are a complex group of cancers that can be classified into Hodgkin’s lymphoma (HL) and non-Hodgkin’s lymphoma (NHL) based on histopathological characteristics. NHL in particular exhibits a wide range of pathological patterns, immunophenotypes, clinical presentations, and responses to treatment (23). The outcome GWASs were classified lymphomas into specific subtypes of HL and NHL. These subtypes included diffuse large B-cell lymphoma (DLBCL), follicular lymphoma (FL), mature T/NK-cell lymphoma, and other unspecified types of NHL.

Two-sample MR analysis results indicated that there is no causal relationship between BMI and NHL ( Table 4 ).

External validation has confirmed a strong link between BMI and the onset of HL. Studies have indicated that individuals with a lower BMI face an increased risk of HL (OR = 0.871, 95% CI 0.8260 – 0.9180, P< 0.001) ( Figure 8 ).