Genetic Study Finds Limited Evidence That Inflammatory Cytokines Drive Blood Cancers

A large genetic analysis has found that circulating inflammatory cytokines show little evidence of a causal role in the risk of several blood cancers. In a two-sample Mendelian randomization study published in Blood Research, researchers examined 41 inflammatory cytokines and C-reactive protein against selected hematologic malignancies, including myeloproliferative neoplasms, Hodgkin lymphoma, follicular lymphoma, diffuse large B-cell lymphoma, mature NK/T-cell lymphoma, chronic lymphocytic leukemia, and acute lymphoblastic leukemia. Although several nominally significant associations appeared in the primary analysis, none survived correction for multiple testing, leaving the authors to conclude that genetically predicted cytokine levels did not have a significant effect on the risk of the malignancies studied.

The question matters because observational studies have repeatedly linked inflammation to blood cancers. Cytokine profiling work has reported that rising levels of interleukin-8, interleukin-2 receptor, interleukin-12, interleukin-15, and interferon gamma-induced protein 10 were associated with inferior survival in patients with primary myelofibrosis. Elevated tumor necrosis factor-alpha has been shown to facilitate clonal expansion and contribute to myeloproliferative neoplasm progression, and increased levels of interleukin-6, interleukin-8, interleukin-10, and tumor necrosis factor-alpha have been found in non-Hodgkin lymphoma, where they may serve as predictive markers for risk stratification. C-reactive protein, a classical acute-phase reactant of the pentraxin family, has been associated in observational research with increased risk of extranodal NK/T-cell lymphoma, myelodysplastic syndromes, follicular lymphoma, and diffuse large B-cell lymphoma.

Yet associations like these cannot settle whether inflammation causes cancer or whether emerging malignancies drive the inflammation. Tumor burden, immune activation secondary to the malignancy, infection, treatment, comorbidities, and other confounding factors can all shape circulating cytokine levels, and reverse causality remains a persistent threat in observational data. Mendelian randomization offers a way around this problem by using genetic variants as proxies for the exposure of interest. Because alleles are randomly assigned during gamete formation, the approach mitigates the effects of confounders and reverse causation bias, allowing researchers to ask whether lifelong genetically predicted differences in cytokine levels influence cancer risk.

In the new study, the team drew summary statistics from genome-wide association studies of the 41 cytokines and C-reactive protein, obtained from the YFS and FINRISK cohorts and their meta-analyses, comprising 8,293 and 204,402 participants of European ancestry respectively. Outcome data came from the UK Biobank, which supplied figures for myeloproliferative neoplasms covering 1,086 cases and 407,155 controls, and from the FinnGen consortium, which provided data for Hodgkin lymphoma with 780 cases and 299,952 controls, follicular lymphoma with 1,081 cases and 299,952 controls, diffuse large B-cell lymphoma with 1,010 cases and 287,137 controls, mature NK/T-cell lymphoma with 335 cases and 299,952 controls, chronic lymphocytic leukemia with 624 cases and 287,133 controls, and acute lymphoblastic leukemia with 184 cases and 287,136 controls.

Instrument selection followed the standard assumptions of Mendelian randomization: genetic variants strongly associated with the exposure, unrelated to confounders, and affecting the outcome only through the exposure. The researchers extracted single nucleotide polymorphisms associated with the cytokines at genome-wide significance, using a relaxed cutoff of P less than 5 × 10−6 for a few cytokines with limited numbers of variants. To remove linkage disequilibrium, they retained SNPs with an R2 below 0.001 within a 10,000 kilobase window, and all included SNPs showed no association with the outcomes at P greater than 1 × 10−5. In total, 414 exposure-associated SNPs entered the analysis, and F-statistics above 10 indicated that weak instrument bias was not a concern. Statistical power values ranged from 1 to 100 percent across the tested pairs.

The inverse-variance weighted method served as the primary analysis, with false discovery rate adjustment applied across the many cytokine-outcome tests. Supplementary methods included weighted median estimation, MR-Egger regression, the MR-Robust Adjusted Profile Score, and MR-Pleiotropy Residual Sum and Outlier analysis. The MR-Egger intercept and Cochran’s Q test assessed pleiotropy and heterogeneity, while leave-one-out analysis and the MR Steiger directionality test checked the reliability and causal direction of the results. All computations used the TwoSampleMR and MR-PRESSO packages in R, with an FDR below 0.05 considered statistically significant and a P-value below 0.05 treated as merely suggestive.

The primary analysis produced a scatter of suggestive signals. Inverse-variance weighted analysis revealed a negative causal relationship between macrophage migration inhibitory factor and myeloproliferative neoplasms, with an odds ratio of 0.58 and a 95 percent confidence interval of 0.41 to 0.81 at P equal to 0.001, but this disappeared after adjustment, with an FDR of 0.062. A positive effect of RANTES on Hodgkin lymphoma, with an odds ratio of 1.32 and a confidence interval of 1.03 to 1.70, faded from an unadjusted P of 0.029 to an FDR of 0.766. Beta-nerve growth factor and interleukin-1 receptor antagonist showed positive effects on follicular lymphoma that vanished after adjustment, and monokine induced by gamma interferon, macrophage inflammatory protein-1 alpha, and stem cell factor lost their links to diffuse large B-cell lymphoma once FDR correction was applied.

The same pattern held for the remaining malignancies. Interferon gamma-induced protein 10, monocyte chemoattractant protein 3, basic fibroblast growth factor, interleukin-16, interleukin-1 receptor antagonist, and platelet-derived growth factor BB showed effects on mature NK/T-cell lymphoma that were absent after adjustment, while RANTES and interleukin-8 associations with chronic lymphocytic leukemia and negative associations of interferon gamma-induced protein 10 and stem cell factor with acute lymphoblastic leukemia likewise disappeared. No causal effects were observed for other cytokines or for C-reactive protein in any of the malignancies at FDR above 0.05. Sensitivity analyses reinforced the picture: after FDR adjustment, none of the 41 cytokines or C-reactive protein showed causal effects by MR-Egger regression, weighted median, MR-RAPS, or MR-PRESSO. Leave-one-out analysis showed no result was driven by a single variant, the MR Steiger test confirmed the hypothesized causal direction for all analyses, and pleiotropy appeared only for a few pairs, including macrophage inflammatory protein-1 beta, RANTES, and granulocyte colony stimulating factor with follicular lymphoma, and C-reactive protein with diffuse large B-cell lymphoma.

The authors interpret the findings as suggestive evidence rather than definitive proof of causal effects, and they caution against reading the result as evidence that inflammatory cytokines are irrelevant to hematologic malignancies. If causal effects exist, they are likely to be modest, subtype-specific, or difficult to detect with currently available genetic sample sizes, particularly since several outcome datasets contained relatively few cases. The study also carries structural limitations: the summary statistics came mainly from individuals of European ancestry, limiting generalizability; the number of genetic instruments varied across cytokines, reducing power for some; genetically predicted circulating levels cannot capture local cytokine activity in the bone marrow, lymph nodes, or tumor tissue; and Mendelian randomization mainly estimates linear lifelong effects, potentially missing nonlinear or threshold relationships such as U-shaped associations. Detailed pathological subtypes could not be analyzed, and important entities including acute myeloid leukemia, multiple myeloma, and myelodysplastic syndromes were excluded because suitable genome-wide association data were unavailable.

What the study does establish is a careful negative result with a clear roadmap for what comes next. Chemokines such as interferon gamma-induced protein 10, monokine induced by gamma interferon, macrophage inflammatory protein-1 alpha, monocyte chemoattractant protein 3, RANTES, and interleukin-8 remain biologically plausible players because they govern leukocyte recruitment, immune-cell trafficking, and tumor microenvironment regulation, while stem cell factor drives stem and progenitor cell survival through the c-KIT pathway. Because the nominally significant associations were not robust after correction, the authors frame them as hypothesis generating, requiring validation in larger genetic studies and functional experiments. Future work with larger genome-wide association datasets, more detailed disease subtypes, multi-ancestry populations, and experimental validation will be needed to clarify whether inflammatory cytokines play any causal role in the development of blood cancers.

Subject of Research: Mendelian randomization analysis of causal effects of circulating inflammatory cytokines on hematologic malignancy risk

Article Title: Limited evidence for causal effects of circulating inflammatory cytokines on the risk of selected hematologic malignancies: a two-sample Mendelian randomization study

Article References: Zhang, H., Zhang, L., Zhao, D., Ma, B., Wang, J., Liu, X., Yin, W., Shen, Y., Yang, L., Dong, W., & Wang, C. (2026). Limited evidence for causal effects of circulating inflammatory cytokines on the risk of selected hematologic malignancies: a two-sample Mendelian randomization study. Blood Research, 61(1), Article 53. https://doi.org/10.1007/s44313-026-00164-z

Image Credits: AI Generated

DOI: 10.1007/s44313-026-00164-z

Keywords: Mendelian randomization, inflammatory cytokines, hematologic malignancies, C-reactive protein, lymphoma, leukemia, myeloproliferative neoplasms, genetics, GWAS, UK Biobank, FinnGen, causal inference