Causal associations between circulation β-carotene and cardiovascular disease: A Mendelian randomization study

Abstract

The causal association between circulating β-carotene concentrations and cardiovascular disease (CVD) remains controversial. We conducted a Mendelian randomization study to explore the effects of β-carotene on various cardiovascular diseases, including myocardial infarction, atrial fibrillation, heart failure, and stroke. Three single nucleotide polymorphisms (SNPs) associated with the β-carotene levels were obtained by searching published data and used as instrumental variables. Genetic association estimates for 4 CVDs (including myocardial infarction, atrial fibrillation, heart failure, and stroke) in the primary analysis, blood pressure and serum lipids (high-density lipoprotein [HDL] cholesterol, LDL cholesterol, and triglycerides) in the secondary analysis were obtained from large-scale genome-wide association studies (GWASs). We applied inverse variance-weighted as the primary analysis method, and 3 others were used to verify as sensitivity analysis. Genetically predicted circulating β-carotene levels (natural log-transformed, µg/L) were positively associated with myocardial infarction (odds ratio [OR] 1.10, 95% confidence interval [CI] 1.02–1.18, P = .011) after Bonferroni correction. No evidence supported the causal effect of β-carotene on atrial fibrillation (OR 1.02, 95% CI 0.96–1.09, P = .464), heart failure (OR 1.07, 95% CI 0.97–1.19, P = .187), stroke (OR 1.03, 95% CI 0.93–1.15, P = .540), blood pressure (P > .372) and serum lipids (P > .239). Sensitivity analysis produced consistent results. This study provides evidence for a causal relationship between circulating β-carotene and myocardial infarction. These findings have important implications for understanding the role of β-carotene in CVD and may inform dietary recommendations and intervention strategies for preventing myocardial infarction.

1. Introduction

Cardiovascular disease (CVD) is a significant global health issue affecting millions worldwide, causing significant morbidity, mortality, and healthcare costs.^[1] Oxidative stress is a known risk factor for CVD and promotes endothelial damage by excessive reactive oxygen species production, alongside common factors such as smoking, obesity, and metabolic abnormalities.^[2–5] This damage can cause significant changes in the vascular system, including increased vasodilation, platelet aggregation, increased endothelial permeability, and the formation of local lesions that increase the risk of CVD.^[6]

Antioxidants can reduce oxidative stress-induced damage by scavenging oxygen free radicals, which may help prevent or mitigate the development of CVD.^[7] However, the relationship between specific antioxidants and CVD is complex and not yet fully understood. β-carotene is a common dietary source of antioxidants that has been extensively investigated in relation to CVD. Some observational studies suggest that β-carotene dietary supplements/blood concentrations reduce the risk of CVD.^[8–10] However, a meta-analysis reported that β-carotene supplementation increases the risk of cardiovascular incidence and mortality.^[11] The inconsistency in these findings could be due to the different biological effects of exposure throughout the lifespan compared to short-term exposure, given the chronic course of CVD.

Understanding the relationship between carotenoids and CVD is important to develop effective preventive measures and treatments. However, the presence of unmeasured confounding factors and reverse causality limits the inference of causality. Therefore, the relationship between carotenoids and CVD remains unclear. This ambiguity underscores the need for robust study designs that can provide reliable estimates of causality. Mendelian randomization (MR) is a study design that utilizes genetic variants associated with exposure to estimate causality between exposure and outcome while avoiding confounding and reverse causality.^[12]

To investigate the causal relationship between circulating β-carotene and CVD, we used a MR study design that utilized single nucleotide polymorphisms (SNPs) as unconfounded proxies for exposure. We examined the effects of β-carotene on 4 CVDs (myocardial infarction [MI], atrial fibrillation, heart failure, and stroke) in our primary analysis and explored its impact on blood pressure and blood lipids in our secondary analyses.

2. Methods and materials

2.1. Study design

To determine the causal relationship between genetically predicted β-carotene concentrations and 4 common CVDs, we utilized a 2-sample Mendelian randomization (MR) study with single nucleotide polymorphisms (SNPs) as instrumental variables. We obtained genetic associations from published large-scale genome-wide association studies (GWASs) and illustrated the MR design principles and assumptions we followed in Figure 1. The study design ensured that the incorporated studies were approved by the respective institutional review boards and ethics committees.

Figure 1.

Schematic overview of the study design and 3 assumptions for the Mendelian randomization. SNP = single nucleotide polymorphism.

2.2. Selection of instrumental variables

We searched the NHGRI-EBI GWAS catalog and PubMed for published GWAS, and gene-carotene associations were obtained from a GWAS that included 2344 participants of European ancestry.^[13] The SNPs included in the instrumental variables need to meet the following requirements: associated with circulating β-carotene levels at a genome-wide significance level (P < 1 × 10⁻⁵); 2) associated with β-carotene independently (linkage disequilibrium, r² < 0.01); 3) not palindromic or with ambiguous information (minor allele frequency, MAF > 0.45). Three SNPs (rs6564851, rs7501331, and rs12934922) met the criteria and were included as instrument variables in the study (Detail information in Table S1, http://links.lww.com/MD/K931). If instrument variables were not available in the outcome dataset, a proxy SNP (r² > 0.8) was used through the LDlink tool based on the European 1000 Genomes data.^[14]

2.3. Data sources for instrument-outcome associations

Detailed information on outcome datasets was showed in Table 1. Gene-outcome associations were extracted from large-scale summary GWAS data of MI,^[15] atrial fibrillation,^[16] heart failure,^[16] stroke,^[17] blood pressure,^[18] high-density lipoprotein (HDL) cholesterol, LDL cholesterol, and triglycerides.^[19] Cohorts or studies included in used GWAS were shown in Table 1. MI, atrial fibrillation, heart failure, and stroke were defined according to the International Classification of Diseases (ICD) code (MI: ICD-10 code I21, I22, I23, I25.2; atrial fibrillation: ICD-9 code 427.3 and ICD-10 code I48; Heart failure: ICD-10 code I42.0; and stroke ICD-9 codes 430 to 436 and ICD-10 codes I60-I64, G45).

Table 1.

Details of outcomes datasets.

Outcomes	Unit	Cohort/studies	Case number	Sample size	Population	Reference (PMID)
Primary
Myocardial infarction	One-unit in logtransformed odds ratio	UK Biobank and Coronary ARtery DIsease Genome-wide Replication and Meta-analysis plus The Coronary Artery Disease (CARDIoGRAMplusC4D)	14,825	395,795	European	33532862
Atrial fibrillation		The Nord-Trøndelag Health Study (HUNT), deCODE, the Michigan Genomics Initiative (MGI), DiscovEHR, UK Biobank, and the AFGen Consortium	60,620	1030,836	European	30061737
Heart failure		26 studies from the Heart Failure Molecular Epidemiology for Therapeutic Targets (HERMES) Consortium	47,309	977,323	European	31919418
Stroke		MEGASTROKE consortium	40,585	446,696	European	29531354
Secondary
Systolic blood pressure	mm Hg	UK Biobank and the International Consortium of Blood Pressure-Genome Wide Association Studies (ICBP)	-	757,601	European	30224653
Diastolic blood pressure	mm Hg		-	757,601	European	30224653
HDL cholesterol	SD (15.5 mg/dL)	45 studies (37 and 8 studies consisted primarily of individuals of European ancestry and non-European ancestry, respectively)	-	187,167	Mixed	24097068
LDL cholesterol	SD (38.7 mg/dL)		-	173,082	Mixed	24097068
Triglycerides	SD (90.7 mg/dL)		-	177,861	Mixed	24097068

SD = standard deviation.

2.4. Statistical analysis

Inverse variance-weighted (IVW) was used as main analysis method, and other 7 MR methods were used for sensitivity analysis, including MR-Egger, maximum likelihood, weighted median, robust adjusted profile score, simple mode, weighted mode, and penalized weighted median.^[20] In order to confirm whether the genetically predicted β-carotene concentration is associated with CVD-related confounders, we also estimated the relationship between β-carotene and blood pressure and blood lipids, and the analysis method was the same as the main analysis. To further confirm the robustness of the results, we used a stricter genome-wide significance threshold (P < 5 × 10⁻⁸) to select IV, and 2 SNPs (rs6564851 and rs12934922) were included. Due to the limited number of SNPs, we only performed IVW analysis in the fixed-effect model. We also calculated the effect of a single SNP and performed the leave-one-out analysis. All tests were 2-sided and the P value <.05/4 (Bonferroni correction) was deemed significant. All analyses were performed using the TwoSampleMR package in R software (Version 4.2.1).^[21]

2.5. Instrument strength and power calculation

The variance explained by instruments was 9.0% based on the formula (${\mathrm{R}}^2=(\beta \times \sqrt{2\times \mathrm{M}\mathrm{A}\mathrm{F}(1\text{-}\mathrm{M}\mathrm{A}\mathrm{F})}{)}^2$).^[22] The F-statistic was 21.12, and an F-statistic > 10 suggests sufficiently strong instruments to mitigate weak instrument bias. We estimated statistical power by a web application (https://cnsgenomics.shinyapps.io/mRnd/),^[23] and we had 80% power (α = 0.05) to detect an OR of 1.08 for MI, 1.04 for atrial fibrillation,1.04 for heart failure, and 1.05 for stroke.

3. Results

After Bonferroni correction, the OR per 1 unit increase of natural log-transformed β-carotene level for MI was 1.10 (95% CI 1.02–1.18, P = .011, Fig. 2). Maximum likelihood (OR 1.10, 95% CI 1.02–1.18), weighted median (OR 1.10, 95% CI 1.01–1.19), robust adjusted profile score (OR 1.10, 95% CI 1.02–1.19), and penalized weighted median (OR 1.10, 95% CI 1.02–1.18) provided similar estimates (Table S2, http://links.lww.com/MD/K932). The results of MR-Egger, simple mode, and weighted mode were consistent with the directionality of the main analysis results, but not statistically significant, which may be due to its low precision. The results of the 4 methods are plotted on a scatter plot (Figure S1, http://links.lww.com/MD/K930). No evidence supported the causal effect of β-carotene on atrial fibrillation (OR 1.02, 95% CI 0.96–1.09, P = .464), heart failure (OR 1.07, 95% CI 0.97–1.19, P = .187), stroke (OR 1.03, 95% CI 0.93–1.15, P = .540) (Fig. 2). When a stricter genome-wide significance threshold was used, the association between β-carotene and MI remained stable (OR 1.10, 95% CI 1.02–1.18, P = .017), and the estimates for other CVDs were also similar to primary results. We did not detect statistically significant heterogeneity (P value for Cochrane Q test = 0.127) and pleiotropy (intercept P value = .308) (Table S2, http://links.lww.com/MD/K932). The single effect of each SNP and leave-one-out analysis corroborates that there is no apparent heterogeneity in the estimates among the 3 SNPs (Table S3, http://links.lww.com/MD/K933).

Figure 2.

Associations of genetically predicted β-Carotene with CVDs. The results are estimated by the random-effects inverse variance–weighted method. P < .05/4 after Bonferroni correction is considered significant. CI = confidence interval, OR = odds ratio.

In the secondary analyses, we observed no causal relationship between β-carotene and CVD-related confounders (P > .239). The OR per 1 unit increase of natural log-transformed β-carotene level was 1.02 (95% CI 0.77–1.36) for systolic blood pressure, 0.93 (95% CI 0.79–1.10) for diastolic blood pressure, 1.02 (95% CI 0.94–1.09) for HDL cholesterol, 0.97 (95% CI 0.92–1.03) for LDL cholesterol, 0.98 (95% CI 0.92–1.01) for triglycerides (Fig. 3).

Figure 3.

Associations of genetically predicted β-Carotene with blood pressure and blood lipids. Results are estimated by the random-effects inverse variance–weighted method. CI = confidence interval, OR = odds ratio.

4. Discussion

We investigated the causal relationship between carotenoids and CVD using MR. The results of this study suggest that higher levels of circulating carotenoids may increase the risk of myocardial infarction. We did not observe evidence of a causal association of β-carotene with atrial fibrillation, heart failure, and stroke. There is no exact causal relationship between β-carotene and blood pressure and blood lipids.

A meta-analysis that included 69 prospective studies noted the protective effect of β-carotene on CHD and stroke. The pooled risk ratio (RR) of CHD per 25 µg/dL of blood β-carotene was 0.76, 95% CI 0.62–0.93, I² = 22%) and the RR of stroke was 0.85 (95% CI 0.74–0.97, I² = 0%).^[8] A meta- analysis that included 16 randomized controlled trials (RCTs) with 182,788 participants instead suggested a positive association between β-carotene supplementation and CVD risk (RR 1.04, 95% CI 1.00–1.08; P = .05; I² = 0%).^[11] We need to be cautious about the differences in results between observational and intervention studies. Observational studies are susceptible to confounding factors and it is difficult to focus purely on the effect of a single carotenoid concentration. In contrast, the intervention length, dose, and duration can affect the effects obtained in interventional studies. CVD is a disease with a long-term course rather than an acute onset, and the ongoing effects of exposure over the life course may have different biological effects than short-term exposure. A feature of MR studies is that they can respond to ongoing effects that reflect lifetime exposure. The results of our study suggest that exposure to higher levels of β-carotene does not reduce the risk of CVD but increases the risk of developing MI instead. The associations of antioxidants with ischemic stroke and atrial fibrillation have been separately studied in 2 previous MR studies. The results suggested no causal relationship between total carotene and ischemic stroke (OR 0.89, 95% CI 0.73–1.09, P = .266), which was similar to our results but did not distinguish between the careful classification in carotene.^[24] Results for atrial fibrillation were consistent with us (OR 1.01, 95% CI 0.97–1.07, P = .560) in another Mendelian randomization study.^[25]

Oxidative stress is associated with many chronic diseases such as cardiovascular disease, diabetes and cancer, and antioxidants are thought to reduce the damage caused by oxidative stress.^[26] However, it has gradually been suggested that excessive use of antioxidants may be harmful.^[27] Certain basal levels of reactive oxygen species are essential for certain functions,^[28] such as apoptosis and phagocytosis, and an excessive reduction in the number of free radicals in the body may affect the performance of these functions, with consequent adverse effects on the organism.^[29,30] This is one possible explanation for the harmful results of short-term β-carotene supplementation in RCTs. Furthermore, it has been shown that carotenoids exhibit pro-oxidant effects under certain conditions and vitamin C can resist that effect.^[31,32] This suggests that interactions between different antioxidants may have an impact on the actual effects of antioxidants.^[32–34] Antioxidants may act in the body in a stepwise fashion or networks in which multiple antioxidants are required.^[35] This may partly explain the lack of beneficial effects of individual β-carotene.

The strengths of this study are the use of data from large studies of circulating β-carotene and CVD, and the MR methodology that reduces bias from reverse causation and measured or unmeasured confounding factors. The data used have a large sample size and enough cases number and the consistency of results obtained from multiple methods improves the precision and robustness of our results. We also explored the relationship between β-carotene and CVD-related confounding or mediating factors. One limitation of this study is that it was restricted to individuals of European ancestry in analysis for MI, atrial fibrillation, heart failure, stroke and blood pressure in order to avoid the effects of population stratification, which makes it possible that our results cannot be extrapolated to other populations. Secondly, as we used published summary-level data, we could not test the possibility of a non-linear effect of circulating β-carotene concentrations on CVD, such as a U- or J-type association. Third, we were unable to explore the association between β-carotene and CVD in individuals with different susceptibilities to CVD. Furthermore, our study only discussed the effects of a single β-carotene, and we could not discuss the differential effects of multiple antioxidants when combined.

5. Conclusion

The findings of this study do not support the notion that elevated circulating carotenoid concentrations provide a protective effect against CVD. Specifically, our results showed that increased levels of circulating β-carotene did not significantly impact the risk of atrial fibrillation, heart failure, or stroke, and may even increase the risk of myocardial infarction (MI). As a result, we do not recommend β-carotene supplementation alone as a preventative measure against CVD. However, additional studies are necessary to examine the effects of combining multiple antioxidants and their impact on various populations.

Author contributions

Conceptualization: Shuangyan Liu.

Formal analysis: Shuangyan Liu, Qiaoyu Wu, Shangshang Wang, Ying He.

Methodology: Shuangyan Liu, Qiaoyu Wu, Shangshang Wang, Ying He.

Writing – original draft: Shuangyan Liu, Qiaoyu Wu, Shangshang Wang, Ying He.

Writing – review & editing: Shuangyan Liu, Shangshang Wang, Ying He.