Sage Journals: Discover world-class research

Abstract

Objectives

To explore the causal associations between oxidative stress markers and the occurrence of various types of lymphomas.

Methods

This two-sample Mendelian randomization (MR) study employed summary data from genome-wide association studies of oxidative stress markers and various lymphoma types. Analysis was conducted using the inverse variance weighted, with confirmation using the weighted median, weighted mode, and MR-Egger regression methods. Heterogeneity, horizontal pleiotropy, outliers, and robustness were tested using the Cochran Q, MR-Egger regression, MR pleiotropy residual sum and outlier (MR-PRESSO), and leave-one-out methods.

Results

Genetically predicted myeloperoxidase (MPO) was causally associated with follicular lymphoma (odds ratio (OR) = 1.33, P = 0.0173). Uric acid showed a causal link to diffuse large B-cell lymphoma (DLBCL) (OR = 1.003, P = 0.043). Glutathione peroxidase (GPX) was associated with follicular lymphoma (OR = 1.16, P = 0.033). Catalase (CAT) was inversely associated with non-Hodgkin lymphoma (NHL) (OR = 0.8921), non-follicular lymphoma (OR = 0.8755), and follicular lymphoma (OR = 0.8364) (all P < 0.05). Glutathione-S transferase (GST) was inversely associated with NHL (OR = 0.9248) and follicular lymphoma (OR = 0.8313) (all P < 0.05). Heterogeneity was noted for uric acid and DLBCL, but no pleiotropy was detected; after outlier removal, the association with uric acid was non-significant.

Conclusions

The findings suggest causal relationships between oxidative stress markers and lymphoma risk, notably MPO and GPX with follicular lymphoma, and inverse associations for GST and CAT with different lymphoma types. Limitations include heterogeneity for some markers, requiring further validation.

Keywords

blood biomarkers Hodgkin lymphoma non-Hodgkin lymphoma Mendelian randomization causality

Introduction

Lymphoma is a broad and heterogeneous family of cancers of the lymphatic system arising from lymphocytes, representing approximately 5% of all cancers.¹ The two main types of lymphoma are Hodgkin lymphoma (HL) and non-Hodgkin lymphoma (NHL). NHLs represent the 10th most common cause of cancer, with an estimated 553,010 new cases in 2022, and the 11th cause of cancer-related death, with 250,475 deaths in 2022.² HL is the 26th cause of cancer (82,409 new cases in 2022), and the 28th cause of cancer-related death (22,701 deaths in 2022).² The prognosis of newly diagnosed HL is generally excellent in patients who have received and responded to therapy.^3,4 NHLs include several histological types: FL, diffuse large B-cell lymphoma (DLBCL), chronic lymphocytic leukemia/small lymphocytic lymphoma, mantle cell lymphoma, marginal zone lymphoma, extranodal marginal zone lymphoma of mucosa-associated lymphoid tissue, mature T and natural killer (NK)-cell lymphoma, and adult T-cell leukemia-lymphoma, among others.^5,6 The prognosis of NHL is highly variable and depends upon the exact subtype, ranging from indolent (e.g. follicular lymphoma (FL)) to very aggressive diseases (e.g. DLBCL).^5,6

HL and NHL can be suspected in patients with painless persistent lymphadenopathy with or without B-symptoms (i.e., weight loss of >10%, night sweats, and/or fever). The diagnosis of HL and NHL is usually established based on blood test results, a biopsy of the lymph nodes and bone marrow, and immunophenotyping by flow cytometry and/or immunohistochemistry of the surgical specimen.^3–6 Several blood parameters are disrupted in patients with lymphoma, including complete blood count, erythrocyte sedimentation rate, lactate dehydrogenase levels, liver and kidney function markers, and β2-microglobulin.^7–9

Oxidative stress is involved in a wide variety of diseases, including cancer, cardiovascular diseases, diabetes, and neurological, kidney, respiratory, psychiatric, and gastrointestinal diseases.¹⁰ Oxidative stress is involved in lymphoma pathogenesis because it participates in DNA, protein, and lipid damage, and develops somatic oncogenic mutations.^11–13 Still, several biomarkers commonly evaluated in clinical practice, including markers of oxidative stress, could be used to screen for patients at risk of lymphoma, but their association with lymphoma is often biased by the confounding and risk of reverse causality often observed in epidemiological studies,¹⁴ and their exact causality in relation to lymphoma remain unknown and probably differ among subtypes of lymphomas.⁹ Determining biomarkers that could predict the occurrence of lymphoma could have public health importance for identifying and monitoring patients at risk. Still, lymphomas are uncommon, and a longitudinal study would require hundreds of thousands of patients followed over decades to reach an answer.

Genome-wide association studies (GWASs) revolutionized the field of genetic medicine by providing data on millions of genetic variations and their association with various phenotypes, especially for complex traits and diseases.^15,16 Mendelian randomization (MR) is a statistical method based on the random distribution of alleles during gamete generation and embryo conception. MR is based on the associations of common single-nucleotide polymorphisms (SNPs) as instrumental variables (IVs) with different exposures and the association of the identified IVs with the outcomes to explore possible causal associations between exposures and outcomes (or diseases).^17,18 Upon specific assumptions, MR studies can help determine causality without bias from confounders and reverse causality.¹⁴ Two-sample MR uses the associations between SNPs and exposures and between SNPs and outcomes from different GWASs to combine them into a causality estimation.

Therefore, this study aimed to explore the causal associations between oxidative stress markers and the occurrence of various types of lymphomas. The results could help provide additional evidence for the screening of lymphoma in the population.

Materials and methods

Study design

This study used publicly available data from GWASs to investigate the causal association between oxidative stress and lymphoma (Figure 1). Because this study used publicly available anonymized data, no additional ethical approval was necessary. This study assumed that (a) the SNPs used as IVs for oxidative stress are associated with oxidative stress (relevance assumption); (b) there are no common causes between the SNPs and lymphoma (independence assumption); and (c) there are no independent pathways between the SNPs and lymphoma other than through oxidative stress (no horizontal pleiotropy assumption).¹⁴

Figure 1.

Schematic representation of the Mendelian randomization study.

Data source

The outcome datasets were obtained from the R12 cohort of the FinnGen Project, which is a GWAS of about half a million Finnish residents. The outcomes included HLs, NHLs, DLBCLs, FLs, non-FLs, mature T or NK cell lymphomas, and other unspecified NHLs (Supplementary Table S1). The exposure datasets related to oxidative stress were from published papers^19–23 and the UK Biobank, a large GWAS of individuals of European ancestry living in the United Kingdom (Supplementary Table S1).

Instrumental variable selection

The IVs included in this study had to meet the following criteria. SNPs significantly associated with each exposure were initially selected at the conventional genome-wide significance level of P < 5 × 10⁻⁸, which is widely used in GWAS to control for multiple testing and to minimize false positive findings. However, for several exposures (cerebral dopamine neurotrophic factor, neurotrophin-3, neurotrophin-4, alpha-linolenic acid measurement, and beta-nerve growth factor levels), very few or no variants met this stringent threshold, which would have resulted in insufficient instrument strength and loss of statistical power in the MR analyses. To retain an adequate number of instruments while still focusing on reasonably strong and biologically plausible signals, a less stringent threshold of P < 5 × 10⁻⁶ was therefore applied, consistent with previous MR studies that have relaxed the genome-wide cut-off in settings with limited genome-wide significant variants.^24–27 The SNPs with a minimum minor allele frequency of >0.01 were screened out.²⁸ Linkage disequilibrium (LD) between SNPs was removed based on R² < 0.001 and a window size of 10,000 kb.²⁹ When a selected IV was not present in the summary data of the outcome, SNPs with high LD (R² > 0.8) were searched with the IV as proxy SNPs for replacement.³⁰ The F-value was calculated for each SNP in the IV to assess the strength of the IV. The calculation formula was F = R² × (N − 2)/(1 − R²), where R² is the proportion of exposure variance explained by the SNP in the IV. The requirement for the F-value to exclude potential weak instrumental bias between the IV and the exposure factor was >10.³¹

Mendelian randomization analysis

In the present study, the inverse variance weighted (IVW) method was used as the primary MR method to assess the causal association between exposure and the outcome risk by calculating the odds ratio (OR) and 95% confidence interval (CI).³² IVW is the main method for MR interpretation, and it calculates the inverse variance of each SNP as weights and computes the weighted average of the effect sizes.³² IVW is a random effect model that can accommodate a certain degree of heterogeneity.³² In addition, the MR-Egger,³³ weighted median,³⁴ and weighted mode³⁵ were used to examine the robustness of IVW results. The MR-Egger method examines the presence of an intercept and provides an accurate causal effect estimate when considering pleiotropic bias.³³ In the weighted median method, half of the IVs are assumed to be valid for analyzing the causal association between the exposure and the outcome.³⁴ All analyses were performed using the “TwoSampleMR” package in R.³⁶ Visualization was performed using scatter plots and sensitivity analysis plots.

Sensitivity analysis

Sensitivity analyses were performed to detect potential heterogeneity and pleiotropy. Cochran's Q-test was used to assess heterogeneity among IVs. A Q-test P-value of >0.05 indicated low heterogeneity and that the estimates among IVs varied randomly and had a minimal impact on the IVW results.³⁷ Because of the risk of an influence of genetic variation heterogeneity on the estimation of association effects and to meet the third MR assumption, the MR-Egger regression method was used to determine the presence of horizontal pleiotropy.³⁷ The absence of pleiotropy was considered when the intercept term of the MR-Egger regression tended toward zero or lacked statistical significance. In addition, an MR pleiotropy residual sum and outlier (MR-PRESSO) analysis was used to detect potential outliers (i.e. SNPs with P < 0.05).^37,38 The outlier SNPs were removed to recalibrate the causal associations and correct for horizontal pleiotropy. A leave-one-out analysis was used to assess the robustness and consistency of the results.³⁹

Results

Instrument variable selection

A total of 619 instrumental variables related to oxidative stress were extracted. The number of IVs ranged from 5 to 239, with all F statistics being >10, varying from 20.91 to 142,042.3 (Supplementary Table S2), indicating the reliable strength of the instrumental variables. The R² values for all exposures varied between 0.00329 and 0.3390 (with the lowest value for retinol and the highest for total bilirubin) (Supplementary Table S2), suggesting that the genetic explanatory power differs across various exposures. Supplementary Table S3 shows the results of SNPs after exposure and outcome matching, including details of the SNPs that were not matched in the outcomes and the proxy SNPs (48 SNPs were not matched in the summary data, of which 18 proxy SNPs were found).

Table 1.

Mendelian randomization analysis results (significant analyses).

Outcomes	Exposures	SNP (n)	Methods	OR (CI)	P	Q statistic (IVW)	P	MR-Egger intercept	P
Follicular lymphoma	MPO	10	Inverse variance weighted	1.3251 (1.0509–1.6709)	0.0173	4.624	0.866	0.006	0.8179
Follicular lymphoma	MPO	10	MR Egger	1.2792 (0.8822–1.8549)	0.2301
Follicular lymphoma	MPO	10	Weighted median	1.2161 (0.9109–1.6236)	0.1845
Follicular lymphoma	MPO	10	Weighted mode	1.2107 (0.8687–1.6875)	0.2882
Diffuse large B-cell lymphoma	Uric acid	217	Inverse variance weighted	1.0026 (1.0001–1.0052)	0.0429	255.841	0.0327	0.005	0.2329
Diffuse large B-cell lymphoma	Uric acid	217	MR Egger	1.0011 (0.9976–1.0047)	0.5333
Diffuse large B-cell lymphoma	Uric acid	217	Weighted median	1.0012 (0.9977–1.0049)	0.4975
Diffuse large B-cell lymphoma	Uric acid	217	Weighted mode	1.0019 (0.9989–1.0048)	0.218
Follicular lymphoma	GPX	14	Inverse variance weighted	1.1595 (1.0118–1.3286)	0.0332	14.628	0.331	−0.046	0.1329
Follicular lymphoma	GPX	14	MR Egger	1.3623 (1.0777–1.7222)	0.0239
Follicular lymphoma	GPX	14	Weighted median	1.2625 (1.066–1.4951)	0.0069
Follicular lymphoma	GPX	14	Weighted mode	1.2774 (1.0864–1.5019)	0.011
Non-Hodgkin lymphoma	CAT	14	Inverse variance weighted	0.8921 (0.823–0.9671)	0.0056	10.343	0.666	0	0.9951
Non-Hodgkin lymphoma	CAT	14	MR Egger	0.8928 (0.7001–1.1385)	0.3786
Non-Hodgkin lymphoma	CAT	14	Weighted median	0.9237 (0.8293–1.0287)	0.1486
Non-Hodgkin lymphoma	CAT	14	Weighted mode	0.9429 (0.8058–1.1033)	0.4764
Non-follicular lymphoma	CAT	14	Inverse variance weighted	0.8755 (0.7936–0.9658)	0.008	10.001	0.694	0.004	0.8679
Non-follicular lymphoma	CAT	14	MR Egger	0.8545 (0.6353–1.1493)	0.3189
Non-follicular lymphoma	CAT	14	Weighted median	0.9182 (0.8019–1.0514)	0.217
Non-follicular lymphoma	CAT	14	Weighted mode	0.9667 (0.7926–1.1792)	0.7439
Follicular lymphoma	CAT	14	Inverse variance weighted	0.8364 (0.7075–0.9889)	0.0365	15.725	0.264	−0.079	0.0738
Follicular lymphoma	CAT	14	MR Egger	1.2866 (0.8147–2.0319)	0.301
Follicular lymphoma	CAT	14	Weighted median	0.9177 (0.7431–1.1333)	0.425
Follicular lymphoma	CAT	14	Weighted mode	0.9239 (0.7023–1.2154)	0.5813
Non-Hodgkin lymphoma	GST	11	Inverse variance weighted	0.9248 (0.8689–0.9843)	0.014	8.058	0.623	−0.017	0.4173
Non-Hodgkin lymphoma	GST	11	MR Egger	0.97 (0.8548–1.1007)	0.6479
Non-Hodgkin lymphoma	GST	11	Weighted median	0.9333 (0.8597–1.0133)	0.0999
Non-Hodgkin lymphoma	GST	11	Weighted mode	0.9338 (0.8586–1.0157)	0.1415
Follicular lymphoma	GST	11	Inverse variance weighted	0.8313 (0.7316–0.9445)	0.0046	11.701	0.306	−0.034	0.4273
Follicular lymphoma	GST	11	MR Egger	0.9159 (0.7042–1.1913)	0.529
Follicular lymphoma	GST	11	Weighted median	0.8454 (0.7253–0.9855)	0.0318
Follicular lymphoma	GST	11	Weighted mode	0.8551 (0.7353–0.9945)	0.0696

CAT: catalase; CI: confidence interval; GPX: glutathione peroxidase; GST: glutathione-S transferase; IVW: inverse variance weighted; MPO: myeloperoxidase; MR: Mendelian randomization; OR: odds ratio; SNP: single-nucleotide polymorphisms.

Table 2.

MR-PRESSO (significant analyses).

Outcome	Exposure	Raw		Outlier corrected		Global P	Outliers	Distortion P
Outcome	Exposure	OR (CI)	P	OR (CI)	P
Diffuse large B-cell lymphoma	Uric acid	1.0025 (0.9999–1.005)	0.062	1.0021 (0.9997–1.0045)	0.0881	0.018	3	0.695
Follicular lymphoma	CAT	0.8364 (0.7075–0.9889)	0.0567	NA (NA-NA)	NA	0.338	NA	NA
Follicular lymphoma	GPX	1.1595 (1.0118–1.3286)	0.0529	NA (NA-NA)	NA	0.331	NA	NA
Follicular lymphoma	MPO	1.3251 (1.1222–1.5647)	0.0089	NA (NA-NA)	NA	0.857	NA	NA
Non-follicular lymphoma	CAT	0.8755 (0.8032–0.9542)	0.0097	NA (NA-NA)	NA	0.708	NA	NA
Non-Hodgkin lymphoma	CAT	0.8921 (0.8302–0.9587)	0.0083	NA (NA-NA)	NA	0.71	NA	NA
Non-Hodgkin lymphoma	GST	0.9248 (0.8745–0.978)	0.0209	NA (NA-NA)	NA	0.713	NA	NA
Follicular lymphoma	GST	0.8313 (0.7316–0.9445)	0.0177	NA (NA-NA)	NA	0.407	NA	NA

CAT: catalase; CI: confidence interval; GST: glutathione-S transferase; MPO: myeloperoxidase; NA: OR: odds ratio.

Mendelian randomization results

Genetically predicted myeloperoxidase (MPO) was causally associated with the risk of FL (OR = 1.3251, 95% CI: 1.0509–1.6709, P = 0.0173; other MR analyses P > 0.05) (Figures 2(a) and 3(a)). Uric acid was causally associated with DLBCL (OR = 1.0026, 95%CI: 1.0001–1.0052, P = 0.0429; other MR analyses P > 0.05) (Figures 2(b) and 3(b)). Glutathione peroxidase (GPX) was causally associated with FL (OR = 1.1595, 95% CI: 1.0118–1.3286, P = 0.0332; MR-Egger P = 0.0239; weighted median P = 0.0069; weighted mode P = 0.0110) (Figures 2(c) and 3(c)). CAT was causally associated with NHL (OR = 0.8921, 95% CI: 0.8230–0.9671, P = 0.0056; other MR analyses P > 0.05) (Figures 2(d) and 3(d)), non-FL (OR = 0.8755, 95% CI: 0.7936–0.9658, P = 0.0080; other MR analyses P > 0.05) (Figures 2(e) and 3(e)), and FL (OR = 0.8364, 95% CI: 0.7075–0.9889, P = 0.0365; other MR analyses P > 0.05) (Figures 2(f) and 3(f)). GST was causally associated with NHL (OR = 0.9248, 95% CI: 0.8689–0.9843, P = 0.014; other MR analyses P > 0.05) (Figures 2(g) and 3(g)) and FL (OR = 0.8313, 95% CI: 0.7316–0.9445, P = 0.0046; MR-Egger P = 0.5290; weighted median P = 0.0318; weighted mode P = 0.0696) (Figures 2(h) and 3(h)) (Table 1). Among the statistically significant IVW results, Cochran's Q-test revealed heterogeneity for only the analysis of uric acid and DLBCL (P = 0.033). No horizontal pleiotropy was observed using the MR-Egger regression analysis (Table 1 and Figure S1(a) to (g)). The leave-one-out analyses did not reveal SNPs driving the results (Figure S2(a) to (g)). The MR-PRESSO analysis revealed three outliers for the analysis of uric acid and DLBCL. After removing the three outliers, the causal association was no longer statistically significant (Table 2).

Figure 2.

Scatter plots for the Mendelian randomization analyses of (a) myeloperoxidase (MPO) and follicular lymphoma, (b) uric acid and diffuse large B-cell lymphoma (DLBCL), (c) glutathione peroxidase (GPX) and follicular lymphoma, (d) catalase (CAT) and non-Hodgkin lymphoma (NHL), (e) CAT and non-follicular lymphoma, (f) CAT and follicular lymphoma, (g) glutathione-S transferase (GST) and NHL, and (h) GST and follicular lymphoma.

Figure 3.

Forest plots for the Mendelian randomization analyses of (a) myeloperoxidase (MPO) and follicular lymphoma, (b) uric acid and diffuse large B-cell lymphoma (DLBCL), (c) glutathione peroxidase (GPX), and follicular lymphoma, (d) catalase (CAT) and non-Hodgkin lymphoma (NHL), (e) CAT and non-follicular lymphoma, (f) CAT and follicular lymphoma, (g) glutathione-S transferase (GST) and NHL, and (h) GST and follicular lymphoma.

All other analyses had negative IVW results (Supplementary Table S4). The Cochran Q test showed heterogeneity in the analysis of 18 negative IVW results. No horizontal pleiotropy was observed using the MR-Egger regression analysis (Supplementary Table S5). The MR-PRESSO analysis revealed that removing the outliers did not change the negative results (Supplementary Table S6).

Discussion

This two-sample MR study explored the causal associations between oxidative stress markers and the occurrence of various types of lymphomas. The results suggested direct causal associations between MPO and FL and GPX and FL, and inverse causal associations between GST and NHL and FL, and between CAT and NHL, non-FL, and FL. These results suggest that oxidative stress is involved in lymphoma pathogenesis. Monitoring the blood levels of those markers could eventually assist in screening patients and risk assessment, but additional studies are necessary before clinical application.

Oxidative stress is involved in cancer pathogenesis,¹⁰ including lymphoma, due to DNA, protein, and lipid damage and the occurrence of somatic oncogenic mutations.^11–13 Unfortunately, the available data are mainly from epidemiological studies, which are often biased by confounders and reverse causality.¹⁴ In the present study, the exposures MPO, GPX, GST, and CAT showed causal associations with various types of lymphoma. MPO is an enzyme that contributes to oxidative stress and inflammation. MPO is found in neutrophils and monocytes and is a key part of the innate immune system. MPO produces reactive oxygen species (ROS) and reactive nitrogen species, which can damage cells and tissues.⁴⁰ MPO plays a role in oxidative stress and cancer through DNA damage and promotion of tumor growth and progression.⁴¹ GPX is an antioxidant enzyme that protects cells from oxidative stress by reducing hydrogen peroxide and other hydroperoxides to water.⁴² Despite its antioxidant role, GPX appears to have dual roles in cancers and can act either as a suppressor or promoter, and is upregulated in lymphoma.⁴³ Those known properties of MPO and GPX support the direct causal associations of MPO and GPX with FL observed in the present study. On the other hand, GSTs are a group of enzymes that protect cells from oxidative stress by binding glutathione to toxic molecules. This process makes the molecules less toxic and easier for the cell to remove.^44–46 GST protects DNA from oxidative damage that can lead to DNA mutations and carcinogenesis.⁴⁵ On the other hand, once cancer has developed, GST appears to promote cancer cell survival, proliferation, and drug resistance.⁴⁷ CAT is an antioxidant enzyme that helps to neutralize oxidative stress by breaking down hydrogen peroxide into water and oxygen and is a key part of the body's defense against oxidative stress.^48,49 CAT also possesses a dual role in cancer, being protective before cancer development and being harmful once cancer develops.^50,51 Nevertheless, the present study examined the genetically predicted risk of lymphoma development, and both GST and CAT appeared protective against lymphoma development.

Still, lymphoma is a very heterogeneous group of diseases.¹ Although a wide variety of oxidative stress markers were tested in the present study, only a few showed associations with a small number of lymphoma types. Disease pathogenesis is often influenced by a combination of genetic and environmental factors.^52,53 Therefore, in the presence of negative results in MR studies, environmental factors should be investigated. Negative MR results do not mean that there is no link between exposure and outcome. The absence of association can be because the IVs are insufficient to model the effect of exposure on the outcome, which is a known limitation of MR.

In this study, many oxidative stress-related markers showed no evidence of a causal association with specific lymphoma subtypes. On the one hand, if the genetic instruments are sufficiently strong and the MR assumptions are reasonably satisfied, such results may indicate that lifelong differences in these biomarkers do not materially influence lymphoma risk, at least within the detectable effect size range of our design.^54,55 On the other hand, several methodological limitations of MR could also contribute to the absence of detectable effects, including limited power for modest or subtype-specific effects, weak or sparse instruments, measurement error in the underlying GWAS, and residual horizontal pleiotropy, all of which tend to bias estimates toward the null.^27,33,56 Moreover, oxidative stress biology is strongly shaped by environmental and lifestyle exposures (such as chemical solvents, air pollution, and other pro-oxidant agents), and accumulating evidence links these exposures—via oxidative damage and chronic inflammation—to lymphomagenesis.^57,58 Because conventional MR using germline variants cannot fully capture context-specific, time-varying, or exposure-intensity–dependent effects, the absence of an MR signal should be interpreted as “no association detected in this genetic framework” rather than definitive evidence of no association, and future work integrating detailed environmental exposure data with genetic instruments is needed to clarify these complex relationships.^54,59

In the analysis of genetically predicted uric acid and DLBCL, we observed substantial heterogeneity and detected several outlier variants using the MR-PRESSO framework. Biologically, this heterogeneity is plausible because uric acid metabolism is tightly linked to a broad spectrum of cardiometabolic disorders and chronic inflammatory pathways, reflecting a complex and pleiotropic genetic architecture.^60–62 At the same time, DLBCL is a highly heterogeneous lymphoma with distinct metabolic subtypes and diverse immune microenvironmental states, and multiple studies have highlighted extensive metabolic and immunologic reprogramming in DLBCL cells.^63–65 Consequently, some MR-PRESSO outlier SNPs may represent true horizontal pleiotropy; for example, variants that influence inflammatory, metabolic, or obesity-related traits in addition to uric acid, whereas others could reflect technical artifacts such as batch effects or subtle differences in genotyping, imputation, or sample structure between the exposure and outcome GWAS.^38,66 Notably, the association between uric acid and DLBCL was no longer statistically significant after removal of MR-PRESSO outliers, and the causal effect estimates were attenuated, indicating that the signal is either weak or sensitive to pleiotropic instruments and should therefore be interpreted with particular caution.

FL being the only subtype associated with MPO and GPX can be framed as biologically plausible because FL is a germinal center (GC)-derived lymphoma that depends heavily on redox-regulated survival and selection signals within a ROS-rich microenvironment, whereas other B-cell lymphoma subtypes are more strongly driven by alternative oncogenic pathways and may be less dependent on MPO/GPX-mediated oxidative balance.^11,67 FL arises from GC B cells (centrocytes/centroblasts), which undergo intense proliferation, somatic hypermutation, and class-switch recombination in a structurally and functionally specialized GC microenvironment. These processes generate high levels of ROS, necessitating tight control of oxidative stress and redox homeostasis to allow survival of GC B cells that carry extensive DNA damage yet avoid catastrophic apoptosis.^11,67,68 MPO, predominantly expressed in myeloid cells within the GC niche, contributes to local generation and modulation of ROS and hypochlorous species, shaping B-cell selection, antigen presentation, and inflammatory signaling that can influence GC B-cell survival and clonal evolution. GPX family members, particularly GPX1/GPX4, detoxify hydrogen peroxide and lipid peroxides, buffering oxidative damage and supporting anti-apoptotic and drug-resistance programs in B-cell malignancies; polymorphisms reducing GPX1 activity have been linked to increased mature B-cell lymphoma risk, with signals enriched in FL, and GPX overexpression has been implicated in therapy resistance and ferroptosis suppression in aggressive lymphomas.^11,68–71 Because FL cells remain GC-like and continue to rely on GC-style redox regulation and microenvironmental support, genetically determined variation in MPO/GPX activity may more directly perturb the homeostasis, selection thresholds, and survival of GC B cells that give rise to FL. In contrast, subtypes such as DLBCL, mantle cell lymphoma, and marginal zone lymphoma are frequently driven by alternative pathogenic programs (e.g. NF-κB activation, BCR signaling dependence, chronic antigenic stimulation, or immune evasion), and although oxidative stress pathways are still relevant for their biology and prognosis, their initiation and maintenance may be less specifically governed by MPO/GPX-mediated redox balance, attenuating subtype-specific genetic associations for these enzymes.^11,67–69

This study used a two-sample MR design and GWAS data from hundreds of thousands of patients. On the other hand, the study also had limitations. The present study was based on GWAS summary statistics derived from populations of European ancestry, which substantially constrains the generalizability of the findings to non-European groups. Because allele frequencies, linkage disequilibrium structure, environmental exposures, and gene-environment interactions can differ across ancestral backgrounds, the effect estimates and inferred genetic architecture observed here may not accurately reflect risk patterns in other populations, and any extrapolation should be made with caution. Future studies are needed to replicate and extend these results in large, ancestrally diverse cohorts or in cohorts of different ancestry (e.g. Asians) to validate the robustness of the conclusions and to enhance their relevance for global populations. Second, several SNPs could be selected as IVs, affecting the causal associations. Third, for some exposures, only a small number of genome-wide significant variants were available, so a more relaxed significance threshold had to be adopted to select IVs to retain sufficient statistical power. This strategy inevitably increases the risk of including weaker or potentially invalid instruments, which could bias effect estimates toward the null or introduce pleiotropy, although it was mitigated by LD clumping and applying robust sensitivity methods such as MR-Egger, weighted median, and related approaches. Alternative approaches specifically developed for settings with many weak instruments (e.g. MR-RAPS, debiased IVW or Egger-type estimators, and methods that explicitly model pleiotropy) exist but were not implemented here, and future work should evaluate whether these techniques materially change the observed associations.

In conclusion, this study suggests direct causal associations between MPO and FL, and GPX and FL, and inverse causal associations between GST and NHL and FL, and between CAT and NHL, non-FL, and FL. These results suggest that oxidative stress is involved in lymphoma pathogenesis. Monitoring such biomarkers could eventually help lymphoma risk stratification. Future studies could support the public health importance of such biomarkers.

Supplemental Material

sj-tif-1-jbm-10.1177_03936155261428137 - Supplemental material for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study

Supplemental material, sj-tif-1-jbm-10.1177_03936155261428137 for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study by Bin Zhu and Wenli Zhao in The International Journal of Biological Markers

Supplemental Material

sj-tif-2-jbm-10.1177_03936155261428137 - Supplemental material for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study

Supplemental material, sj-tif-2-jbm-10.1177_03936155261428137 for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study by Bin Zhu and Wenli Zhao in The International Journal of Biological Markers

Supplemental Material

sj-docx-3-jbm-10.1177_03936155261428137 - Supplemental material for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study

Supplemental material, sj-docx-3-jbm-10.1177_03936155261428137 for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study by Bin Zhu and Wenli Zhao in The International Journal of Biological Markers

Supplemental Material

sj-docx-4-jbm-10.1177_03936155261428137 - Supplemental material for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study

Supplemental material, sj-docx-4-jbm-10.1177_03936155261428137 for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study by Bin Zhu and Wenli Zhao in The International Journal of Biological Markers

Supplemental Material

sj-xlsx-5-jbm-10.1177_03936155261428137 - Supplemental material for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study

Supplemental material, sj-xlsx-5-jbm-10.1177_03936155261428137 for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study by Bin Zhu and Wenli Zhao in The International Journal of Biological Markers

Supplemental Material

sj-docx-6-jbm-10.1177_03936155261428137 - Supplemental material for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study

Supplemental material, sj-docx-6-jbm-10.1177_03936155261428137 for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study by Bin Zhu and Wenli Zhao in The International Journal of Biological Markers

Supplemental Material

sj-docx-7-jbm-10.1177_03936155261428137 - Supplemental material for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study

Supplemental material, sj-docx-7-jbm-10.1177_03936155261428137 for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study by Bin Zhu and Wenli Zhao in The International Journal of Biological Markers

Supplemental Material

sj-docx-8-jbm-10.1177_03936155261428137 - Supplemental material for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study

Supplemental material, sj-docx-8-jbm-10.1177_03936155261428137 for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study by Bin Zhu and Wenli Zhao in The International Journal of Biological Markers

Supplemental Material

sj-docx-9-jbm-10.1177_03936155261428137 - Supplemental material for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study

Supplemental material, sj-docx-9-jbm-10.1177_03936155261428137 for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study by Bin Zhu and Wenli Zhao in The International Journal of Biological Markers

Supplemental Material

sj-docx-10-jbm-10.1177_03936155261428137 - Supplemental material for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study

Supplemental material, sj-docx-10-jbm-10.1177_03936155261428137 for Causal association between oxidative stress and lymphomas: A two-sample Mendelian randomization study by Bin Zhu and Wenli Zhao in The International Journal of Biological Markers

Footnotes

Acknowledgments

Not applicable.

ORCID iD

Wenli Zhao

Ethical considerations

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Funding

The authors disclosed the receipt of the following financial support for the research, authorship, and/or publication of this article: The work was supported by Shanghai Fengxian District Second Batch of Clinical Treatment Center Construction Project under Grant [number fxlczlzx-a-202105].

Authors’ contributions

Bin Zhu carried out the studies, participated in collecting data, and drafted the manuscript. Bin Zhu and Wenli Zhao performed the statistical analysis and participated in its design. Bin Zhu participated in the acquisition, analysis, or interpretation of data and drafted the manuscript. All authors read and approved the final manuscript.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability

All data generated or analyzed during this study are included in this published article.

Supplemental material

Supplemental material for this article is available online.

References

Jamil

Mukkamalla

SKR

. Lymphoma. In: StatPearls. Treasure Island (FL): StatPearls Publishing, 2025. Ineligible companies. Disclosure: Shiva Kumar Mukkamalla declares no relevant financial relationships with ineligible companies.

Bray

Laversanne

Sung

, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024; 74: 229–263.

Connors

Cozen

Steidl

, et al. Hodgkin lymphoma. Nat Rev Dis Primers 2020; 6: 61.

Brice

de Kerviler

Friedberg

. Classical Hodgkin lymphoma. Lancet 2021; 398: 1518–1527.

Bowzyk Al-Naeeb

Ajithkumar

Behan

, et al. Non-Hodgkin lymphoma. Br Med J 2018; 362: k3204.

Armitage

Gascoyne

Lunning

, et al. Non-Hodgkin lymphoma. Lancet 2017; 390: 298–310.

Shen

Young

, et al. A nomogram-based prognostic model for lymphoma patients initially presenting with fever of unknown origin. J Inflamm Res 2024; 17: 8445–8469.

Yuan

, et al. High serum lactate dehydrogenase predicts an unfavorable outcome in Chinese elderly patients with multiple myeloma. Oncotarget 2017; 8: 48350–48361.

Paquin

Oyogoa

McMurry

, et al. The diagnosis and management of suspected lymphoma in general practice. Eur J Haematol 2023; 110: 3–13.

10.

Reddy

. Oxidative stress in health and disease. Biomedicines 2023; 11: 2925.

11.

Sousa-Pimenta

Estevinho

Sousa Dias

, et al. Oxidative stress and inflammation in B-cell lymphomas. Antioxidants (Basel) 2023; 12: 936.

12.

Zahra

Lefter

Ali

, et al. The involvement of the oxidative stress status in cancer pathology: a double view on the role of the antioxidants. Oxid Med Cell Longev 2021; 2021: 9965916.

13.

Reuter

Gupta

Chaturvedi

, et al.

Oxidative stress, inflammation, and cancer: how are they linked?

Free Radic Biol Med 2010; 49: 1603–1616.

14.

Haycock

Burgess

Wade

, et al. Best (but oft-forgotten) practices: the design, analysis, and interpretation of Mendelian randomization studies. Am J Clin Nutr 2016; 103: 965–978.

15.

Visscher

Wray

Zhang

, et al. 10 Years of GWAS discovery: biology, function, and translation. Am J Hum Genet 2017; 101: 5–22.

16.

Sud

Kinnersley

Houlston

. Genome-wide association studies of cancer: current insights and future perspectives. Nat Rev Cancer 2017; 17: 692–704.

17.

Davey Smith

Hemani

. Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum Mol Genet 2014; 23: R89–R98.

18.

Sekula

Del Greco

Pattaro

, et al. Mendelian randomization as an approach to assess causality using observational data. J Am Soc Nephrol 2016; 27: 3253–3265.

19.

Sun

Maranville

Peters

, et al. Genomic atlas of the human plasma proteome. Nature 2018; 558: 73–79.

20.

Folkersen

Gustafsson

Wang

, et al. Genomic and drug target evaluation of 90 cardiovascular proteins in 30,931 individuals. Nat Metab 2020; 2: 1135–1148.

21.

Gilly

Park

Png

, et al. Whole-genome sequencing analysis of the cardiometabolic proteome. Nat Commun 2020; 11: 6336.

22.

Evans

Zhu

, et al. Genome-wide association study identifies loci affecting blood copper, selenium and zinc. Hum Mol Genet 2013; 22: 3998–4006.

23.

Shin

Fauman

Petersen

, et al. An atlas of genetic influences on human blood metabolites. Nat Genet 2014; 46: 543–550.

24.

Fadista

Manning

Florez

, et al. The (in)famous GWAS P-value threshold revisited and updated for low-frequency variants. Eur J Hum Genet 2016; 24: 1202–1205.

25.

Chen

Boehnke

Wen

, et al. Revisiting the genome-wide significance threshold for common variant GWAS. G3 (Bethesda) 2021; 11. DOI: 10.1093/g3journal/jkaa056.

26.

Cai

Xiao

, et al. Benchmarking Mendelian randomization methods for causal inference using genome-wide association study summary statistics. Am J Hum Genet 2024; 111: 1717–1735.

27.

Sanderson

Glymour

Holmes

, et al. Mendelian randomization. Nat Rev Methods Primers 2022; 2.

28.

Charon

Allodji

Meyer

, et al. Impact of pre- and post-variant filtration strategies on imputation. Sci Rep 2021; 11: 6214.

29.

Raghavan

Vardarajan

Mayeux

. Genomic variation in educational attainment modifies Alzheimer disease risk. Neurol Genet 2019; 5: e310.

30.

Timmers

Tiys

Sakaue

, et al. Mendelian randomization of genetically independent aging phenotypes identifies LPA and VCAM1 as biological targets for human aging. Nat Aging 2022; 2: 19–30.

31.

Wang

Liu

, et al. Association between circulating unsaturated fatty acid and preeclampsia: a two-sample Mendelian randomization study. J Matern Fetal Neonatal Med 2024; 37: 2294691.

32.

Burgess

Butterworth

Thompson

. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol 2013; 37: 658–665.

33.

Bowden

Davey Smith

Burgess

. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol 2015; 44: 512–525.

34.

Bowden

Davey Smith

Haycock

, et al. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol 2016; 40: 304–314.

35.

Hartwig

Davey Smith

Bowden

. Robust inference in summary data Mendelian randomization via the zero modal pleiotropy assumption. Int J Epidemiol 2017; 46: 1985–1998.

36.

Rasooly

Patel

. Conducting a reproducible Mendelian randomization analysis using the r analytic statistical environment. Curr Protoc Hum Genet 2019; 101: e82.

37.

Bowden

Del Greco

Minelli

, et al. A framework for the investigation of pleiotropy in two-sample summary data Mendelian randomization. Stat Med 2017; 36: 1783–1802.

38.

Verbanck

Chen

Neale

, et al. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 2018; 50: 693–698.

39.

Burgess

Bowden

Fall

, et al. Sensitivity analyses for robust causal inference from Mendelian randomization analyses with multiple genetic variants. Epidemiology 2017; 28: 30–42.

40.

Chen

, et al. Targeting myeloperoxidase (MPO) mediated oxidative stress and inflammation for reducing brain ischemia injury: potential application of natural compounds. Front Physiol 2020; 11: 433.

41.

Valadez-Cosmes

Raftopoulou

Mihalic

, et al. Myeloperoxidase: growing importance in cancer pathogenesis and potential drug target. Pharmacol Ther 2022; 236: 108052.

42.

Pei

Pan

Wei

, et al. Research progress of glutathione peroxidase family (GPX) in redoxidation. Front Pharmacol 2023; 14: 1147414.

43.

Zhao

Wang

Zhou

, et al. Glutathione peroxidase GPX1 and its dichotomous roles in cancer. Cancers (Basel) 2022; 14: 2560.

44.

Veal

Toone

Jones

, et al. Distinct roles for glutathione S-transferases in the oxidative stress response in Schizosaccharomyces pombe. J Biol Chem 2002; 277: 35523–35531.

45.

Allocati

Masulli

Di Ilio

, et al. Glutathione transferases: substrates, inihibitors and pro-drugs in cancer and neurodegenerative diseases. Oncogenesis 2018; 7: 8.

46.

Mazari

AMA

Zhang

, et al. The multifaceted role of glutathione S-transferases in health and disease. Biomolecules 2023; 13: 688.

47.

Singh

Reindl

. Glutathione S-transferases in cancer. Antioxidants (Basel) 2021; 10: 701.

48.

Nandi

Yan

Jana

, et al. Role of catalase in oxidative stress- and age-associated degenerative diseases. Oxid Med Cell Longev 2019; 2019: 9613090.

49.

Anwar

Alrumaihi

Sarwar

, et al. Exploring therapeutic potential of catalase: strategies in disease prevention and management. Biomolecules 2024; 14: 697.

50.

Glorieux

Buc Calderon

. Targeting catalase in cancer. Redox Biol 2024; 77: 103404.

51.

Galasso

Gambino

Romanelli

, et al. Browsing the oldest antioxidant enzyme: catalase and its multiple regulation in cancer. Free Radic Biol Med 2021; 172: 264–272.

52.

Farhud

. Hypothetical strategies of gene and environmental influence on life expectancy: a brief review. Iran J Public Health 2022; 51: 2382–2387.

53.

Klimek

Aichberger

Thurner

. Disentangling genetic and environmental risk factors for individual diseases from multiplex comorbidity networks. Sci Rep 2016; 6: 39658.

54.

Davies

Holmes

Davey Smith

. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians. Br Med J 2018; 362: k601.

55.

Burgess

Thompson

. Interpreting findings from Mendelian randomization using the MR-Egger method. Eur J Epidemiol 2017; 32: 377–389.

56.

Pierce

VanderWeele

. The effect of non-differential measurement error on bias, precision and power in Mendelian randomization studies. Int J Epidemiol 2012; 41: 1383–1393.

57.

Pizzino

Irrera

Cucinotta

, et al. Oxidative stress: harms and benefits for human health. Oxid Med Cell Longev 2017; 2017: 8416763.

58.

Zheng

Gonçalves

Abiko

, et al. Redox toxicology of environmental chemicals causing oxidative stress. Redox Biol 2020; 34: 101475.

59.

Burgess

Butterworth

Thompson

. Beyond Mendelian randomization: how to interpret evidence of shared genetic predictors. J Clin Epidemiol 2016; 69: 208–216.

60.

Cho

Kim

, et al. Large-scale cross-ancestry genome-wide meta-analysis of serum urate. Nat Commun 2024; 15: 3441.

61.

Wang

Mesa-Eguiagaray

Campbell

, et al. A phenome-wide association and factorial Mendelian randomization study on the repurposing of uric acid-lowering drugs for cardiovascular outcomes. Eur J Epidemiol 2024; 39: 869–880.

62.

, et al. Serum uric acid and its metabolism-a vital factor in the inflammatory transformation of cancer. J Adv Res 2025; Aug 22: S2090-1232(25)00646-0. Online ahead of print. DOI: 10.2991/978-94-6463-932-2.

63.

Miao

Medeiros

, et al. Diffuse large B-cell lymphoma with molecular variations more than ABC and GCB classification. Precision Cancer Medicine 2018; 1: 4–4.

64.

Wang

Liu

Fei

, et al. Metabolic pathway-based subtyping reveals distinct microenvironmental states associated with diffuse large B-cell lymphoma outcomes. Hematol Oncol 2024; 42: e3279.

65.

Lordello

Nuan-Aliman

Kielbassa-Elkadi

, et al. Metabolic heterogeneity in diffuse large B-cell lymphoma cells reveals an innovative antimetabolic combination strategy. Cancers (Basel) 2025; 17: 394.

66.

Shen

Pan

, et al. Genetic interactions and pleiotropy in metabolic diseases: insights from a comprehensive GWAS analysis. J Cell Mol Med 2024; 28: e70045.

67.

Jaffe

. Follicular lymphomas: a tapestry of common and contrasting threads. Haematologica 2013; 98: 1163–1165.

68.

Domka

Goral

Firczuk

. cROSsing the line: between beneficial and harmful effects of reactive oxygen species in B-cell malignancies. Front Immunol 2020; 11: 1538.

69.

Peroja

Pasanen

Haapasaari

, et al. Oxidative stress and redox state-regulating enzymes have prognostic relevance in diffuse large B-cell lymphoma. Exp Hematol Oncol 2012; 1: 2.

70.

Andreadis

Gimotty

Wahl

, et al. Members of the glutathione and ABC-transporter families are associated with clinical outcome in patients with diffuse large B-cell lymphoma. Blood 2007; 109: 3409–3416.

71.

Schmitt

Grimm

Kreienkamp

, et al. BRD4 Inhibition sensitizes diffuse large B-cell lymphoma cells to ferroptosis. Blood 2023; 142: 1143–1155. doi:10.1182/blood.2022019274