Causal relationship between inflammatory markers and left ventricle geometry and function: A 2-sample Mendelian randomization study

Abstract

Studies have shown that some inflammatory markers can predict the risk of cardiovascular disease (CVD) and affect the structure and function of the heart. However, a causal relationship between inflammatory markers and the cardiac structure and function has not yet been established. Thus, we conducted a 2-sample Mendelian randomization (MR) study to explore the potential causal relationship between inflammatory markers and prognostically-related left ventricular (LV) parameters. Instrumental variables (IVs) for C-reactive protein (CRP), interleukin-6 (IL-6), and myeloperoxidase (MPO) levels were selected from the databases of large genome-wide association studies (GWAS). Summary statistics for LV parameters, including LV mass, ejection fraction, end-diastolic and systolic volumes, and the ratio of LV mass to end-diastolic volume, were obtained from cardiovascular magnetic resonance studies of the UK Biobank (n = 16923). The inverse-variance weighted (IVW) method was the primary analytical method used, and was complemented with the MR-Egger, weighted median, simple mode, weighted mode, and MR pleiotropy residual sum and outlier (MR-PRESSO) methods. Sensitivity analysis was performed to evaluate the robustness of the results. CRP was significantly associated with the LV mass in the IVW method (β = −0.13 g [95% confidence interval [CI], 0.78 g–1.00 g], P = .046). A higher standard deviation of genetically-predicted CRP levels was associated with a 0.13 ± 0.06 g lower LV mass. No causal relationships of IL-6 and MPO with LV parameters were found. No evidence of heterogeneity and pleiotropy was detected. Sensitivity analyses confirmed the robustness of the results. Two-sample MR analysis revealed a causal association between increased CRP level and decreased LV mass, whereas IL-6 and MPO levels did not influence the LV parameters. However, further research is required to validate our findings.

1. Introduction

Cardiovascular disease (CVD) is the leading cause of morbidity and mortality globally.^[1] In particular, ischemic heart disease is one of the most prevalent conditions worldwide.^[2] The incidence of and its severe sequelae, such as heart failure, is increasing.^[3,4]

Cardiovascular magnetic resonance imaging is crucial for the diagnosis and treatment of CVD.^[5] Left ventricular (LV) parameters measured using cardiac imaging techniques, such as end-diastolic volume, end-systole volume, ejection fraction, and mass, have been reported to have prognostic value for subsequent major cardiovascular events and cardiovascular death.^[6,7] Previous observational and experimental studies have shown that persistent inflammation plays a crucial role in the pathogenesis of CVD.^[8,9] C-reactive protein (CRP), an inflammatory mediator, is widely recognized as a potent risk indicator that can independently predict future cardiovascular events. According to the available evidence, CRP is a proatherogenic factor in all stages of CVD, from the development of fatty streaks to clinical events.^[10,11] Interleukin-6 (IL-6), another inflammatory marker, is an immunomodulatory factor that is thought to influence the development of coronary heart disease.^[12–14] According to previous studies, myeloperoxidase (MPO), an abundant heme peroxidase found in azurophilic granules of neutrophils and monocytes, may also be involved in the development and prognosis of CVD.^[15,16] However, these studies may have had a high risk of confounding factors and reverse causality. To the best of our knowledge, no study has established the causative effect of these inflammatory markers on LV structure and function. Therefore, it is unclear whether a causal relationship exists between them.

The Mendelian randomization (MR) method, a natural genetic equivalent of the randomized controlled trial, is frequently used to explore the causal relationship between exposures and outcomes. Relying on the random distribution of genetic variants during meiosis, MR imitates the naturally occurring “randomized trials” in the population.^[17,18] This approach uses genetic variants with a specific effect on a trait (exposure) as the instrumental variables (IVs) to assess the causal effects of exposure factors on the outcomes.^[19] Because of the special advantage of using IVs, MR analysis is not affected by traditional confounding factors^[20] and provides a robust understanding of the causal relationship between exposures and outcomes. Therefore, in this study, we investigated the causal relationship of CRP, IL-6, and MPO levels with changes in the LV parameters using 2-sample MR analysis.

2. Materials and methods

2.1. Study design

This 2-sample MR study was designed to estimate the causal associations between inflammatory markers (CRP, IL-6, and MPO) and LV parameters, including LV ejection fraction, LV end-diastolic volume, LV end-systolic volume, LV mass, and LV mass-to-end-diastolic volume ratio, following the latest guidelines for MR analysis.^[21] This MR study was based on the following 3 assumptions: IVs are closely associated with serum CRP, IL-6, and MPO; IVs are not associated with any potential confounders; and IVs directly affect the outcomes (LV parameters) through exposure rather than through other pathways (Fig. 1).

Figure 1.

Study design flowchart of the Mendelian randomization study.

The data used in this study are publicly available and restricted to European populations, and informed consent and ethical approval were obtained during the original studies.

2.2. GWAS data for inflammation markers

Summary statistics for serum CRP were obtained from a meta-analysis of genome-wide association studies (GWAS) including individuals of European ancestry within the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Inflammation Working Group (CIWG) (GWAS ID: ieu-b-35). The GWAS adjusted for the body mass index and sex. Serum CRP levels (mg/L) were measured using standard laboratory techniques.^[22] In addition, the genetic variations in IL-6 and MPO were derived from a GWAS of 21,758 European individuals from the SCALLOP consortium (GWAS ID: ebi-a-GCST90012005, ebi-a-GCST90012031).^[23]

2.3. IV selection

All genetic variants significantly associated with the CRP were selected as IVs (P < 5 × 10⁻⁸). However, given the limited sample size and number of single nucleotide polymorphisms (SNPs), we relaxed the threshold to P < 5 × 10⁻⁶ when selecting SNPs associated with IL-6 and MPO. We then performed clumping using the Two-sample MR R package (version 0.5.6) to select genetic variants without linkage disequilibrium (LD) (R² < 0.001 across a 10,000 kb window).^[24] To exclude potential horizontal pleiotropy, SNPs associated with confounders or risk factors for the outcome (diabetes, lipids, body mass index, and CVD) were also excluded (threshold of P = 1E-5, R² = 0.8).^[25] In addition, we performed a strength assessment to determine the valid IVs. The overall F-statistic and R² (proportion of explained variance) were analyzed in this process.^[26,27] If the total F was > 10, the study was less likely to be affected by weak instrumental bias. A flowchart of the selection process is shown in Figure 2.

Figure 2.

The flowchart of instrumental variables selection. LVEF = left ventricle ejection fraction, LVEDV = left ventricular end-diastolic volume, LVESV = left ventricular end-systolic volume, LVM = left ventricular mass, LVMVR = left ventricular mass to end-diastolic volume ratio; CRP = C-reactive protein; IL-6 = Interleukin-6; MPO = myeloperoxidase.

2.4. GWAS data for LV parameters

The GWAS summary statistics for LV parameters were collected from cardiovascular magnetic resonance studies of the UK Biobank and comprised 16923 European UK Biobank participants (mean age 62.5 years; 45.8% men) without myocardial infarction or heart failure.^[28] Information on all the genetic datasets used in this study is shown in Table 1. The results were presented as β ± SE change in the LV parameters per one standard deviation increase in the genetically-predicted circulating CRP, IL-6, MPO levels.

Table 1.

Data sources and instrumental variables strength assessment.

Trait	Data sources	Sample size	Ancestry	R2 (total)	F-statistic (total)
Exposure
C-reactive protein Interleukin-6 Myeloperoxidase	CIWG SCALLOP SCALLOP	204,402 21,758 21,758	European European European	0.005 0.007 0.042 (0.039*)	35.98 25.07 56.47 (52.29*)
Outcome
LV parameters	UK Biobank	16,923	European

CIWG: the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Inflammation Working Group; SCALLOP: Systematic and Combined Analysis of Olink Proteins; F =  R²(N-K-1)/[K(1– R²)], R² = 2 × (1–MAF) × MAF × (β/SD)2, SD = SE × N1/2, where MAF is the effect allele frequency, β is the estimated effect on C-reactive protein, Interleukin-6 and Myeloperoxidase; N is the sample size of the GWAS and SE is the standard error of the estimated effect.

*The total R² and F-statistic for Myeloperoxidase and LVEDV.

2.5. Statistical analysis

We used inverse-variance weighted (IVW) for the major MR analysis, which can provide a robust causal estimate even in the presence of heterogeneity.^[29] In addition, the MR-Egger method,^[30] weighted median, simple mode, and weighted mode^[31] were used to complement the IVW. The odds ratios (OR) and 95% confidence intervals (CIs) for the 5 approaches were obtained. An MR-Egger regression test was conducted to detect horizontal pleiotropy based on the P value for its intercept and generate estimates after correcting for pleiotropy.^[30] The MR pleiotropy residual sum and outlier (MR-PRESSO) can detect and correct horizontal pleiotropic outliers using the IVW method and explore significant differences in causal assessments before and after excluding the outliers.^[32] In previous studies on MR analyses, IVW and MR-Egger regression were often used as the basic MR methods, while the other methods are innovative MR approaches that have been introduced in recent years.

2.6. Sensitivity analyses

In this study, we used several approaches for the sensitivity analysis. First, Cochran Q test was used to assess heterogeneity. If the P value was <.05 in the Cochran Q test, the results of MR were referred to as the IVW multiplicative random effects; otherwise, the IVW method with a fixed-effects model was applied.^[29,33] Secondly, we used the MR-Egger intercept to evaluate horizontal pleiotropy. If the P value of the MR-Egger intercept was <.05, the IVs failed to meet Assumptions 2 and 3, and were considered heavily influenced by horizontal pleiotropy, indicating the unreliability of the result.^[30] Meanwhile, horizontal pleiotropy was assessed by using funnel plots. Third, we used MR-PRESSO to detect and correct outliers and make the results of IVW reliable.^[32] In addition, the forest diagram of the IVs reflected the association between exposure and outcome for each SNP. All the MR analyses were performed using the Two-sample MR and MR-PRESSO packages in R (version 4.2.2).

3. Results

3.1. Causal association between CRP and LV parameters

A total of 54 SNPs related to CRP-related genetic variation were obtained. We excluded SNPs associated with LV parameters or its confounders (rs10512597, rs1051338, rs10832027, rs11108056, rs12202641, rs1260326, rs12960928, rs12995480, rs13233571, rs13409371, rs1490384, rs1558902, rs2293476, rs2794520, rs3134899, rs387976, rs4129267, rs4420638, rs4841132, rs644234, rs6485751, rs6601302, rs7310409, rs9271608, rs178810, rs2315008, and rs2352975), as well as palindromic SNPs (rs10240168 and rs10778215). The remaining 25 SNPs were considered as the final IVs with regard to CRP (Supplementary Tables 1 to 5, http://links.lww.com/MD/N152, http://links.lww.com/MD/N153, http://links.lww.com/MD/N154, http://links.lww.com/MD/N155, http://links.lww.com/MD/N156).

Figure 3 shows the association between genetically-determined serum CRP levels and LV parameters in the primary MR analyses. We found that there was a causal association between CRP and LV mass using the IVW approach (β = −0.13 g, OR = 0.88, 95% CI, 0.78 g, 0.99 g, P = .046), which means per standard deviation increase in genetically-predicted CRP levels was associated with 0.13 ± 0.06 g lower LV mass. However, IVW estimates showed that genetically-predicted serum CRP levels were not significantly associated with other LV parameters (LV ejection fraction: OR = 1.03, 95% CI, 0.89, 1.18, P = .733; LV end-diastolic volume: OR = 1.00, 95% CI, 0.89 mL, 1.14 mL, P = .944; LV end-systolic volume: OR = 0.98, 95% CI, 0.85 mL, 1.12 mL, P = .729; LV mass-to-end-diastolic volume ratio: OR = 0.89, 95% CI: 0.78–1.00, P = .054; Fig. 3).

Figure 3.

Causal effects are estimated using the inverse-variance weighted method; CRP, C-reactive protein; IL-6, Interleukin-6; MPO, Myeloperoxidase; LVEF, left ventricle ejection fraction; LVEDV, left ventricular end-diastolic volume; LVESV, left ventricular end-systolic volume; LVM, left ventricular mass; LVMVR, left ventricular mass to end-diastolic volume ratio; The evaluated causal effects of circulating levels of CRP, IL-6 and MPO (per 1-SD increase) on LV parameters. SE, standard error, OR, odds ratio and CI, confidence interval.

Supplementary MR analyses, including the weighted median, MR-Egger, Simple mode, and weighted mode also showed no association between genetically-predicted serum CRP levels and all LV parameters (Supplementary Table 16, http://links.lww.com/MD/N167). No outliers were identified in the MR-PRESSO analysis.

In our study, no evidence of heterogeneity was found in the MR analyses between CRP and LV parameters (Table 2). Therefore, the IVW method used in the primary MR analyses can provide robust causal estimates. The MR-Egger intercept p-values were larger than 0.05 for all LV parameters, indicating that there was no horizontal pleiotropy in the aforementioned results, which satisfies Assumptions 2 and 3. The leave-one-out method suggested that the association between serum CRP levels and LV parameters was not driven by a single SNP (Supplementary Fig. 2, http://links.lww.com/MD/N171). Scatter, forest and funnel plots are shown in Supplementary Fig. 1, http://links.lww.com/MD/N170, 3, http://links.lww.com/MD/N172 and 4, http://links.lww.com/MD/N173. In addition, the overall F-statistic for each parameter was >10, indicating the validity of the IVs (Table 1).

Table 2.

Pleiotropy and heterogeneity test of the CRP, IL-6, and MPO from LV parameters GWAS.

Exposure	Outcomes	Pleiotropy test			Heterogeneity test
		MR-Egger			MR-Egger			Inverse-variance weighted
		Intercept	SE	P	Q	Q_df	Q_pval	Q	Q_df	Q_pval
CRP	LVEF	−0.005	0.006	.35	30.33	23	0.14	31.50	24	0.14
	LVEDV	−0.005	0.006	.36	30.34	23	0.14	31.51	24	0.14
	LVESV	0.004	0.006	.47	30.74	23	0.13	31.47	24	0.14
	LVM	−0.002	0.005	.71	17.62	23	0.78	17.77	24	0.81
	LVMVR	−0.004	0.005	.40	18.01	23	0.76	18.74	24	0.77
IL-6	LVEF	−0.014	0.011	.24	3.41	5	0.64	5.21	6	0.52
	LVEDV	0.005	0.012	.70	6.66	5	0.25	6.88	6	0.33
	LVESV	0.009	0.013	.52	7.20	5	0.21	7.90	6	0.25
	LVM	0.008	0.011	.47	2.68	5	0.75	3.29	6	0.77
	LVMVR	−0.007	0.011	.53	4.02	5	0.55	4.45	6	0.61
MPO	LVEF	−0.003	0.007	.67	23.07	15	0.08	23.37	16	0.10
	LVEDV	−0.012	0.006	.06	12.67	14	0.55	16.83	15	0.33
	LVESV	−0.006	0.007	.41	22.02	15	0.17	21.06	16	0.18
	LVM	−0.007	0.007	.32	16.28	15	0.11	23.54	16	0.10
	LVMVR	0.007	0.006	.30	16.28	15	0.36	17.52	16	0.35

CRP = C-reactive protein, df = degree of freedom, IL-6 = Interleukin-6, LVEDV = left ventricular end-diastolic volume, LVEF = left ventricle ejection fraction, LVESV = left ventricular end-systolic volume, LVM = left ventricular mass, LVMVR = left ventricular mass to end-diastolic volume ratio, MPO = Myeloperoxidase, Q = heterogeneity statistic Q.

3.2. Causal association of IL-6 and MPO with LV parameters

For IL-6 and MPO, no evidence was found to support the hypothesis that they were causally associated with the LV parameters using the main IVW method after the exclusion of SNPs (Supplementary Table 6–15, http://links.lww.com/MD/N157, http://links.lww.com/MD/N158, http://links.lww.com/MD/N159, http://links.lww.com/MD/N160, http://links.lww.com/MD/N161, http://links.lww.com/MD/N162, http://links.lww.com/MD/N163, http://links.lww.com/MD/N164, http://links.lww.com/MD/N165, http://links.lww.com/MD/N166) associated with the confounders (rs1150754 and rs13107325 for MPO). In addition, the weighted median, simple mode, weighted mode, and MR-Egger methods used in the sensitivity analyses demonstrated the same result, as shown in Supplementary Tables 17, http://links.lww.com/MD/N168 and 18, http://links.lww.com/MD/N169. Forest, funnel and scatter plots also showed MR estimates of the association between each IL-6- and MPO-related SNP and changes in the LV parameters (Supplementary Fig. 5, http://links.lww.com/MD/N174, 7, http://links.lww.com/MD/N176, 8, http://links.lww.com/MD/N177, 9, http://links.lww.com/MD/N178, 11, http://links.lww.com/MD/N180, 12, http://links.lww.com/MD/N181). The MR-Egger intercept demonstrated that IL-6 and MPO were not directionally pleiotropic with respect to the LV parameters (Table 2). The Cochrane Q statistics for IL-6 and MPO revealed no heterogeneity (Table 2). The MR-PRESSO method identified one outlier (rs6034875) for the association between MPO and LV end-diastolic volume, and the non-causal association remained after excluding this outlier. Leave-one-out analysis showed that a single SNP did not influence the estimates of IL-6 and MPO (Supplementary Fig. 6, http://links.lww.com/MD/N175, 10, http://links.lww.com/MD/N179).

4. Discussion

Our MR study revealed an association between increased serum CRP levels and lower LV mass. However, IL-6 and MPO levels were not found to be significantly associated with the LV geometry and function. Importantly, these results remained reliable when validated by various sensitivity analyses. To the best of our knowledge, this study is the first to use 2-sample MR analysis to examine the association between inflammatory markers and prognostically-related LV parameters from a genetic perspective.

Recent studies demonstrated that chronic inflammation-induced coronary microvascular dysfunction plays an important role in the pathophysiology of CVD.^[34,35] It is known that coronary artery disease is characterized by a high inflammatory burden, and several studies demonstrated that these hematological parameters are associated with the severity and prognosis of coronary artery disease.^[36,37] Chronic inflammation not only contributes to the progression of CVD but also affects its prognosis. As the disease progresses, the risk of coronary artery damage and ischemia increases, promoting atherosclerosis.^[38] LV remodeling is a clinical characterization of the genesis and progression of morphological alterations in the LV that result in ventricular dysfunction.^[39] CRP is a sensitive marker of nonspecific systemic inflammation, and elevated serum CRP levels are associated with coronary heart disease.^[40] As a pivotal cytokine in innate immunity, IL-6 is also associated with LV remodeling after myocardial infarction.^[41] MPO is a heme enzyme produced by polymorphonuclear neutrophils, which has been identified as an important modulator of postischemic cardiac remodeling.^[42] The significance of LV mass as a CVD biomarker has been proven in previous studies.^[43,44] In our study, the 2-sample MR analysis demonstrated that serum CRP had a potential causal association with decreased LV mass, whereas IL-6 and MPO showed no association. CRP appears to be a causative factor in myocardial remodeling by affecting the LV mass. However, previous observational studies have yielded inconsistent results regarding the association between the CRP level and LV mass.

An observational study showed that the LV mass index increased in a stepwise manner with increasing CRP levels in both men and women.^[45] Similarly, a cohort study of 2633 adults revealed that LV hypertrophy was associated with an inflammatory state, as reflected by elevated CRP levels.^[46] However, a randomized controlled trial did not find a statistically significant association between high-sensitivity CRP levels and LV mass in patients with end-stage kidney disease.^[47] The inconsistent results of these observational studies may have been influenced by potential confounders and reverse causality. In addition, a large cohort study showed that the beta coefficient changed from positive to negative for the association between CRP and LV mass after further adjustment for weight,^[48] which was consistent with the trend shown in our study.

Several possible underlying mechanisms have been proposed for this association. First, the GWAS data used in our study were adjusted for the body mass index and weight.^[22,28] Obesity may serve as a mediator between the CRP level and LV mass. In other words, weight affects both inflammation and LV mass, and adjustment for weight significantly modifies the relationship between serum CRP levels and LV mass.^[48] Another possibility is that in clinically healthy individuals with normal LV mass and CRP levels, when LV hypertrophy has just started but remains in the preclinical stage, there is a compensatory drop in the CRP concentration. As an acute-phase protein, CRP has been studied extensively in the context of LV mass and geometry. Previous studies have suggested that the relationship between CRP levels and LV mass is most likely indirect,^[46,49] possibly reflecting the relationship between body size and LV mass. Therefore, further studies are warranted to determine the causal relationships among CRP levels, obesity, and LV mass.

However, because the majority of conventional studies to date have been observational or cohort studies, which are vulnerable to potential confounders and reverse causality, as well as biases caused by varying disease diagnostic criteria, small sample size, and different populations that restrict the interpretation of results, it is still unknown whether these 3 inflammatory markers have an impact on LV structure and function.

In this study, we used the MR method for the first time to evaluate the causative relationship between the inflammatory markers and LV structure and function. Because the MR analyses were restricted to individuals of European ancestry, our results were less likely to be biased by population stratification. Additionally, a 2-sample MR analysis was used to provide causal evidence that CRP is associated with changes in LV mass, which eliminated the limitations of previous studies.

However, this study has some limitations. First, the results of the GWAS pertaining to IL-6 and MPO may be biased because of the relatively small sample size. Therefore, caution should be exercised when interpreting the negative effects of IL-6 and MPO levels on LV changes. Second, as our dataset included only European populations, the conclusions cannot be applied to non-European populations. Further studies are required to confirm the applicability of these results to different populations and races.

In summary, our 2-sample MR analysis provides preliminary genetic evidence that CRP is associated with decreased LV mass; however, the causal relationship of IL-6 and MPO with LV structures and function could not be demonstrated. These findings provide new insights into the relationship between inflammatory markers and LV structure and function. Future studies should explore the mechanisms underlying the effects of these inflammatory markers.

Acknowledgments

We thank the UK Biobank, CIWG and SCALLOP consortium, We also thank Taylor & Francis (https://china.taylorandfrancis.com/) for English language editing.

Author contributions

Conceptualization: Bolin Lai, Li Li.

Data curation: Bolin Lai, Bin Huang.

Formal analysis: Bolin Lai, Bin Huang.

Funding acquisition: Li Li.

Investigation: Bolin Lai.

Software: Bolin Lai, Bin Huang.

Supervision: Li Li.

Writing – original draft: Bolin Lai.

Writing – review & editing: Li Li.

Supplementary Material