Multiple Sclerosis and Hematologic Malignancies: A Bidirectional Mendelian Randomization Study

Summary

This bidirectional two-sample Mendelian randomization study used European GWAS to assess causal links between multiple sclerosis (MS) and hematologic malignancies (HM). Genetic liability to MS was associated with higher odds of Hodgkin lymphoma and unspecified leukemia; no associations for other subtypes or in reverse analyses.

Abstract

Observational studies have reported associations between multiple sclerosis (MS) and hematologic malignancies (HM), but findings remain inconsistent. We conducted a bidirectional two-sample Mendelian randomization (MR) analysis using publicly available genome-wide association summary statistics for MS (n = 115,803) and HM (n = 218,792). Primary causal estimates were obtained with the inverse-variance-weighted (IVW) estimator, complemented by prespecified sensitivity analyses for heterogeneity and pleiotropy. Genetically proxied liability to MS was associated with higher odds of leukemia (unspecified subtype) (odds ratio [OR] 1.311, 95% confidence interval [CI] 1.002-1.716, P = 0.048) and Hodgkin lymphoma (HL) (OR 1.224, 95% CI 1.052-1.425, P = 0.009), with no evidence for other leukemia, lymphoma, or plasma-cell neoplasm subtypes (all P > 0.05). Tests indicated no substantial heterogeneity or directional pleiotropy for the leukemia (unspecified subtype) or HL analyses. These results provide genetic evidence consistent with an elevated risk of select HM subtypes among individuals with higher genetic liability to MS; however, the signals should be interpreted cautiously and validated in larger, multi-ancestry datasets and with additional causal frameworks.

Introduction

Multiple sclerosis (MS) is a chronic autoimmune disorder of the central nervous system characterized by inflammatory demyelination and a relapsing-remitting course that can lead to disability and substantial societal and familial burden¹. Hematologic malignancies (HM) comprise heterogeneous neoplasms of hematopoietic or lymphoid origin and account for approximately 6.5% of all tumors worldwide². Although targeted and immunotherapies have improved outcomes for many patients, prognosis remains poor for a subset³. The relationship between MS and HM has been investigated for decades with divergent res....

Protocol

This study analyzed de-identified, summary-level genome-wide association study (GWAS) statistics that are publicly available. In accordance with repository policies and the approvals obtained by the original investigators, no new institutional review board approval or additional individual informed consent was required for this secondary analysis. All contributing GWAS reported ethics approval and consent procedures in their source publications. All analyses were conducted in compliance with institutional guidelines and the Declaration of Helsinki.

Overview and rationale

The study implemented a b....

Results

Selection of instrumental variants

Using a genome-wide significance threshold of P < 5×10⁻⁸, 22,466 SNPs were initially retrieved for MS. After LD clumping in PLINK (r² < 0.001, window 10,000 kb), 72 SNPs remained as instruments, and each had F > 10 (range 30-1,044). To minimize pleiotropy through known risk pathways, 11 SNPs associated with potential HM confounders (e.g., body mass index, waist circumference, weight, hip circumference, fat percentage, fat .......

Discussion

This two-sample Mendelian randomization (MR) study in a European-ancestry population provides genetic evidence consistent with a higher risk of Hodgkin lymphoma (HL) and leukemia of unspecified cell type among individuals with greater genetic liability to multiple sclerosis (MS), while showing no convincing association with other leukemia subtypes, non-Hodgkin lymphoma (NHL) subtypes, or multiple myeloma/plasma-cell neoplasms²⁷. Signals for HL and unspecified leukemia are suggestive rather than de.......

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Bonferroni Correction (α = 0.0056)	Statistical method reference	BC-0056	Multiple testing correction method to control the false positive rate
Cochran's Q Statistic	Cochran, 1954	Q-001 (https://www.jstor.org/stable/2334368)	Used to assess heterogeneity between MR instruments
F-statistic	TwoSampleMR documentation	F-049 (https://cran.r-project.org/web/packages/TwoSampleMR/vignettes/perform_mr.html)	Metric for assessing instrument strength in MR
HM Summary Statistics (GWAS)	FinnGen	https://www.finngen.fi/en	Genomic data for hematologic malignancies (n = 218,792), including Hodgkin lymphoma (HL), non-Hodgkin lymphoma (NHL), leukemia, etc.
I2GX statistic	Bowden et al., 2016	I2GX-074 (https://academic.oup.com/ije/article/44/2/512/753845)	Statistic to evaluate compliance with the "no measurement error" assumption
LD Clumping Parameters	PLINK	https://www.cog-genomics.org/plink/1.9	Parameters for LD clumping to ensure independence of genetic instruments
MR-PRESSO v1.0	MR-PRESSO	1445 (https://github.com/rondolab/MR-PRESSO)	Software for heterogeneity and outlier detection and correction
MS Summary Statistics (GWAS)	International MS Genetics Consortium	https://imsgc.net	Genomic data for multiple sclerosis (MS) (n = 115,803), from European cohort
PhenoScanner V2	PhenoScanner	1550 (http://www.phenoscanner.medschl.cam.ac.uk)	Database to screen genetic instruments for associations with known confounders (e.g., smoking, BMI)
PLINK v1.9	PLINK	SCR_001757 (https://www.cog-genomics.org/plink/1.9)	Software for genome-wide association analysis and linkage disequilibrium (LD) clumping
R version 4.3.1	R Foundation	SCR_001905 ( https://www.r-project.org)	Open-source statistical computing software
Steiger Test	TwoSampleMR methodology	STG-001 (https://cran.r-project.org/web/packages/TwoSampleMR)	Statistical test for directionality in Mendelian randomization
TwoSampleMR v0.5.7	RStudio / CRAN	1134 (https://cran.r-project.org/package=TwoSampleMR)	R package for two-sample Mendelian randomization (MR) analysis
Weighted Median and MR-Egger	MendelianRandomization methods	WM-026 ( https://cran.r-project.org/package=MendelianRandomization)	Alternative estimators for sensitivity analysis in MR