Causal associations between diet and thromboembolism: A 2-sample Mendelian randomization study

Abstract

Studies have revealed links between diet and cardiovascular conditions such as venous thromboembolism (VTE). The potential causal relationship between dietary habits and VTE remains ambiguous. This study aimed to assess the causal association of diets with VTE using a 2-sample Mendelian randomization (MR) method. The UK biobank and FinnGen biobanks were used for data acquisition. genome-wide association studies presented at UK biobank provide a comprehensive dataset of more than 139 dietary intakes. The FinnGen biobank was used to acquire genome-wide association studies data on VTE, including both pulmonary embolism and deep-vein thrombosis. Statistical approaches included MR. The primary analysis in the MR study employed the Inverse-Variance Weighted method. Cochran Q test, MR-Egger intercept test, Mendelian Randomized Polymorphism Residual and Outlier test, radial-MR test, and a leave-one-out analysis were used to assess heterogeneity and horizontal pleiotropy. 8 diets associated with reduced risk of VTE, including black pepper, chili pepper, curry, globe artichoke, herring, mackerel, sardines and alcohol consumption. Conversely, ham intake significantly increased risk of VTE. Food preferences for cheesecake and coffee with sugar were associated with higher pulmonary embolism risk, meanwhile gherkins preference was correlated with higher deep-vein thrombosis risk. No causal relationships were found between VTE and other dietary factors. There was no evidence of heterogeneity or pleiotropy in the sensitivity analysis. The MR analysis shows a genetic causal connection between dietary habits and the onset of VTE, highlighting the importance of dietary adjustments in preventing and managing VTE.

1. Introduction

Venous thromboembolism (VTE), including pulmonary embolism (PE) and deep-vein thrombosis (DVT), is a prevalent vascular condition globally. It is a leading cause of illness and death related to cardiovascular disease (CVD). Annually, around 10 million individuals are affected by VTE.^[1] The prevalence of VTE has been gradually rising during the last decades, posing significantly to the global economic burden,^[2] which has raised public concern on its related risk factors. The formation of VTE results from numerous factors sharing risk factors related to CVD. These factors include lack of physical exercise, trauma, surgery, cancer, infection, hormonal therapy and obesity.^[3]

The consumption of different foods can promote or inhibit blood clotting, indicating a potential link to venous thrombosis through certain nutrients. Over the past decade, several studies have confirmed that dietary factors relate to VTE.^[4–9] For instance, consuming more “prudent” diet, rich in fruits, vegetables, yogurt, poultry, and seafood, was linked to a lower VTE risk by prospective cohort study. Conversely, a higher consumption of the “Western” diet, known for its high levels of red and processed meat, and trans fatty acids, was associated with an elevated VTE risk.^[4] A moderate consumption of fatty fish was connected to a decreased risk of idiopathic VTE when compared to a low consumption of fatty fish for both genders.^[10] In a Danish cohort the study found a correlation between alcohol consumption and VTE: drinking an average intake of 4 to 13 standard drinks weekly (1–2 drinks per day) was linked to a reduced risk of VTE.^[11] Despite substantial evidence supporting a connection between dietary habits and VTE, inconsistencies in findings arise from constraints like limited sample sizes and the influence of confounding factors. A definitive causal link between specific dietary components and VTE remains inconclusive.

Utilizing genetic variants as instruments, Mendelian randomization (MR) estimates potential causal associations between exposures and outcomes.^[12] This approach offers a robust framework for evaluating causal links between genetic predispositions and phenotypic results from a genetic standpoint effectively mitigating the impact of confounding variables and the phenomenon of reverse causation often encountered in observational research. Leveraging this method genome-wide association studies (GWAS) with extensive sample sizes have uncovered numerous genetic variations associated with food preference traits, thereby supplying a rich pool of genetic instrumental variables (IVs) for causal inference.^[13]

This research involved 2-sample MR analyses to explore the causal relationships between food preferences and VTE, DVT, and PE. The food preferences traits were conducted on 161,625 participants including over 139 specific foods.^[14] We aimed to analyze the link between dietary elements and VTE from a genetic perspective, highlighting the importance of diet as a changeable risk factor in preventing and managing this condition.

2. Materials and methods

2.1. Study design

A 2-sample MR study was conducted to investigate the potential causal impacts of 139 distinct food phenotypes on VTE, DVT, and PE, respectively. The genetic proxies for food exposures were determined utilizing information from the UK Biobank (https://www.ukbiobank.ac.uk/), a comprehensive biomedical database and research platform. This resource encompasses extensive genetic, health, and biological sample data collected from about 500,000 participants across the United Kingdom. The data source on VTE, including PE and DVT, was the FinnGen study (https://www.finngen.fi/en) - a substantial collaboration between the public and private sectors that has gathered and scrutinized genetic and health information from 500,000 contributors to the Finnish biobank to elucidate the genetic underpinnings of illnesses. Single nucleotide polymorphisms (SNPs) were used as IVs.

Following the identification of suitable SNPs, multiple analytical approaches were applied, encompassing inverse-variance weighted (IVW), MR-Egger regression, weighted median, weighted mode, and MR-Robust Adjusted Profile Score (MR-RAPS). Sensitivity analyses, including the MR-Egger intercept test, Cochran Q test, and leave-one-out analysis, were performed to assess the robustness of the results. The genetic variations used in this analysis must comply with the 3 MR assumptions. First, the IVs linked to dietary choices show notable correlations. Second, the association between these genetic IVs and VTE is not influenced by confounding variables. Third, the genetic IVs exclusively impact the VTE risk through the exposure pathway. MR assumption and the study design depicted in Figure 1. Importantly, this study followed the reporting standards detailed in Strengthening observational studies using MR (STROBE-MR).^[16] The STROBE-MR checklist is available in the Supplementary Materials, Supplemental Digital Content, https://links.lww.com/MD/Q814.

Figure 1.

Research hypothesis and design flowchart. (A) Three assumptions of MR study. Assumption1: IVs must be strongly correlated with exposure; Assumption2: IVs cannot be correlated with confounding factors; Assumption3: IVs can only affect outcomes through exposure factors; (B) The design flow chart of this study. DVT = deep-vein thrombosis, GWAS = genome-wide association study, IVs = instrumental variables, IVW = inverse-variance weighted, MR = Mendelian randomization, MR-RAPS = Mendelian randomization-robust adjusted profile score, MR-PRESSO = MR-pleiotropy residual sum and outlier, PE = pulmonary embolism, VTE = venous thromboembolism. Adapted from Zhou et al.^[15]

2.2. Data sources

The GWAS data of 139 specific foods phenotypes were obtained from UK biobank (https://www.ukbiobank.ac.uk/) involving 161,625 participants. These participants, aged 37 to 73 years, were enrolled across the United Kingdom from 2006 to 2010.^[14] The study used an online questionnaire comprising 152 items to obtain information on participants about food-liking. Participants rated their food-liking on a 9-point scale ranging from 1 for “Extremely dislike” to 9 for “Extremely like,” Additional choices comprised “Have never tried it” and “Prefer not to answer.” The questionnaire specifics are available at https://biobank.ndph.ox.ac.uk/showcase/showcase/docs/foodpref.pdf. Finally, 1401 significant food-liking genetic associations were identified.

The GWAS summary data for VTE, DVT and PE that met the diagnostic criteria based on the International Classification of the diseases codes, are available from the latest FinnGen study,^[17] including 500,348 individuals (26,333 cases and 474,0015 controls), 440,357 individuals (8134 cases and 432,223 controls) and 499,081 individuals (12,762 cases and 486,319 controls), respectively. The GWAS summary data are available online (https://www.finngen.fi/en/access_results). The study involved participants of European descent. Table S1, Supplemental Digital Content, https://links.lww.com/MD/Q815 provides comprehensive information on the data sources and diagnostic codes utilized. All the documents for this project were sourced from the project website and were publicly available. During the original research, all the participants gave informed consent. Thus, no further ethical approval or additional informed consent is necessary.

2.3. Instrumental variables selection

Eligible SNPs were chosen as IVs according to the following criteria: SNPs had a strong association with exposures, achieving genome-wide significance at P < 5 × 10⁻⁸. SNPs were independently associated with exposures, with a threshold of linkage disequilibrium at R² <0.001 and clumping window size of 10,000 Kb. SNPs with F-statistics exceeding 10 were selected to guarantee a robust correlation between IVs and exposures. To meet the second assumption, we utilized the GWAS Catalog tools (https://www.ebi.ac.uk/gwas/) were used to identify SNPs linked to potential confounders, including smoking, cancer, obesity, alcohol, T2DM (type 2 diabetes),^[1–3] body height, hip circumference, waist-hip ratio, waist circumference,^[18,19] education,^[20] cardiovascular disease,^[21] which were validated, then excluding IVs linked to these confounders. The Radial-MR test^[22] and MR-pleiotropy residual sum and outlier (MR-PRESSO) methods^[23] were used to remove outlier SNP, then the outliers were removed from further MR analysis.

2.4. Statistical analysis

The primary method employed in this study to assess the causal association between food preference and VTE was the IVW method.^[24] This approach provides a reliable estimation of causality when all SNPs are valid and without directed pleiotropy. In addition, we conducted various analyses to validate the robustness of our findings, including the Weighted median, Weighted mode, MR-Egger and MR-Robust Adjusted Profile Score (MR-RAPS) methods. Consistent estimates were obtained by the weighted median model when more than half of the weight comes from valid IVs.^[25] The true causal effect is identified by the weighted mode when there is no larger set of invalid instruments with a comparable ratio estimation in cases where less than half of the variables are valid instruments.^[26] The MR-Egger technology allows for assessing pleiotropic effects within a causal framework by testing the null hypothesis of causality under the assumption of instrument strength independent of direct effect (InSIDE), which posits that instrument strength is independent of direct effects. A P-value for the MR-Egger intercept below .05 indicated that the influence of SNPs associated with exposure factors on outcomes was considered unreliable.^[27] By employing MR-RAPS, it is feasible to include multiple weak instruments below the standard GWAS threshold, improving the reliability of Cochran Q-statistic for identifying heterogeneity due to pleiotropy, thus reducing the rate of false positives (or type I errors).^[28] To mitigate false positives in multiple hypothesis testing, false discovery rate correction was used to determine causal links according to adjusted P-values below .05.^[29]

2.5. Sensitivity analysis

The sensitivity analysis utilized Cochran Q test to evaluate heterogeneity across IVs. This method computes the weighted sum of squared deviations for both individual estimates and the pooled estimate, a P-value >.05 signifies the absence of heterogeneity.^[23] Horizontal pleiotropy was assessed through MR-Egger regression. A P-value for the MR-Egger intercept below .05 suggested that the effects of SNPs linked to exposure variables on the results might lack reliability.^[27] Additionally, the outliers were identified through the MR-PRESSO method, enabling a comprehensive assessment of pleiotropic effects and the identification of anomalous SNPs. This method also provided causal estimates by excluding relevant outliers.^[30] Furthermore, Leave-one-out was performed to assess the impact of individual SNP on the association results by systematically excluding each SNP. The Steiger test was also utilized to determine the causal relationship, with P-value threshold of <.05 indicating the correct causal direction in the analysis.^[22] Finally, we generated scatter plots, forest plots, and funnel plots to provide further clarification. All analyses in this study were performed using the TwoSampleMR (version 0.6.0), MendelianRandomization (version 0.8.0), and MR-PRESSO package (1.0) in R Software 4.3.3 (https://www.R-project.org).

3. Results

3.1. Instrumental variables selection

Following the exclusion of SNPs with linkage disequilibrium effects, potential confounding associations, palindromic sequences, or outlier characteristics, the analysis proceeded. Among these, 176 SNPs from the GWAS Catalog were excluded due to associations with confounding factors such as smoking (9 SNPs), cancer (3 SNPs), obesity (2 SNPs), BMI (43 SNPs), alcohol (41 SNPs), hip circumference (8 SNPs), waist circumference (10 SNPs), waist-hip ratio (9 SNPs), body height (20 SNPs), CVD (5 SNPs), educational (14 SNPs) and T2DM (12 SNPs). The MR-PRESSO analysis detected 65 SNPs displaying horizontal pleiotropy. After the Radial-MR test, 46 outlier SNPs were removed. At the end, a total of 3468 (P < 5 × 10^-8) SNPs were chosen as IVs. In this study, the F-statistics >10 showed no evidence of weak instrumental bias. Detailed information of IVs is available in Tables S2 and 3, Supplemental Digital Content, https://links.lww.com/MD/Q815.

3.2. The Mendelian randomization estimates of food-related phenotypes on venous thromboembolism

Using the IVW method, we identified 9 food preferences causally associated with VTE, as shown in Figure 2. Among them, black pepper (OR, 0.85; 95% CI, 0.74–0.99; P = .034), chili pepper (OR, 0.88; 95% CI, 0.80–0.97; P = .008), curry (OR, 0.75; 95% CI, 0.60–0.95; P = .016), globe artichoke (OR, 0.90, 95% CI, 0.82–0.99; P = .020), herring (OR, 0.90, 95% CI, 0.82–0.98; P = .015), mackerel (OR, 0.85, 95% CI, 0.74–0.97; P = .018), sardines (OR, 0.87; 95% CI, 0.78–0.96; P = .006) and alcohol consumption (OR, 0.81; 95% CI, 0.72–0.91; P = .001) had protective effects on VTE. On the contrast, although there was no evidence of an overall causal relationship between meat liking and the risk of VTE, ham intake (OR, 1.27; 95% CI, 1.08–1.49; P = .003) seemed to be promote effect for VTE. The results of MR-RAPS method for these traits showed a suggestive causality to DVT (P <.05) as well. Besides, we also found that spicy food (OR 0.86; 95% CI, 0.75–0.98; P = .026) showed protective effects against VTE whereas beer consumption (OR, 1.14; 95% CI, 1.02–1.27; P = .017) and liver intake (OR, 1.23; 95% CI, 1.06–1.42; P = .007) had a positive effect using the WM method, although the IVW method showed nonsignificant effects (spicy food: OR, 0.91; 95% CI, 0.83–1.01; P = .071, beer: OR, 1.08; 95% CI, 0.92–1.28; P = .353, liver: OR,1.10; 95% CI, 0.96–1.26; P = .172) (Table S4, Supplemental Digital Content, https://links.lww.com/MD/Q815).

Figure 2.

Associations of genetically food preferences with VTE, PE and DVT risk using IVW methods. DVT = deep-vein thrombosis, IVW = inverse-variance weighted, PE, pulmonary embolism, VTE = venous thromboembolism.

3.3. The Mendelian randomization estimates of food-related phenotypes on pulmonary embolism

Genetic causal relationships with PE were observed for 5 food preference traits based on IVW results, as shown in Figure 2. Preferences for alcohol consumption (OR, 0.79; 95% CI, 0.66–0.94; P = .007), chili pepper (OR, 0.83; 95% CI, 0.83–0.94; P = .005) and globe artichoke (OR, 0.86; 95% CI, 0.75–0.98; P = .021) were link to reduced PE risk, whereas cheesecake (OR, 1.40; 95% CI, 1.09–1.79; P = .008) and coffee with sugar (OR, 1.39, 95% CI,1.07–1.80; P = .012) were suggestive to be causally correlated with a higher PE risk. These findings were confirmed by MR-RAPS (5/5), MR-Egger method (1/5) and weighted median (1/5) methods (Table S4, Supplemental Digital Content, https://links.lww.com/MD/Q815).

3.4. The Mendelian randomization estimates of food-related phenotypes on deep-vein thrombosis

In Figure 2, for DVT, we found that preferences for gherkins (OR, 1.18; 95% CI, 1.04–1.34; P = .012) was positively correlated with DVT using IVW, while alcohol consumption (OR, 0.80; 95% CI, 0.65–0.99; P = .042) suggested marginal protective effects against DVT. We found no causal associations of the remaining food with VTE in all MR methods (Table S4, Supplemental Digital Content, https://links.lww.com/MD/Q815).

3.5. Sensitivity analysis

There was no evidence of pleiotropy and heterogeneity in our findings (all P-values >.05) according to MR-Egger intercept and Cochran Q statistics (Table 1). The Steiger test confirmed all the causal directions of food preferences (Table S5, Supplemental Digital Content, https://links.lww.com/MD/Q815). Based on leave-one-out analysis, we did not find any SNP with a dominant effect on the overall estimates in all marginally or statistically significant causal relationships. Figures 3–6 and Figure S1–8, Supplemental Digital Content, https://links.lww.com/MD/Q814 display the scatter plots, forest plots, funnel plots, and leave-one-out plots.

Table 1.

MR analysis of food preferences on VTE, PE, and DVT.

Exposure	Outcome	Method	SNPs	P	OR (95% CI)	P (Cochran’ Q heterogeneity)	P (MR-egger intercept Pleiotropy)	P (MR-PRESSO globle test)
Black pepper	VTE	IVW	7	.03	0.85 (0.74–0.99)	.611	.274	.599
		MR-Egger	–	.18	0.48 (0.19–1.21)	–	–	–
		Weighted median	–	.03	0.81 (0.67–0.98)	–	–	–
		Weighted mode	–	.10	0.78 (0.60–1.01)	–	–	–
		MR-RAPS	–	.04	0.85 (0.73–1.00)	–	–	–
Chilli pepper	VTE	IVW	10	.01	0.88 (0.80–0.97)	.935	.276	.929
		MR-Egger	–	.16	0.60 (0.31–1.15)	–	–	–
		Weighted median	–	.02	0.87 (0.77–0.98)	–	–	–
		Weighted mode	–	.09	0.83 (0.69–1.01)	–	–	–
		MR-RAPS	–	.01	0.88 (0.80–0.97)	–	–	–
Curry	VTE	IVW	4	.02	0.75 (0.60–0.95)	.314	.236	.395
		MR-Egger	–	.58	1.22 (0.67–2.22)	–	–	–
		Weighted median	–	.06	0.78 (0.60–1.01)	–	–	–
		Weighted mode	–	.64	0.89 (0.59–1.36)	–	–	–
		MR-RAPS	–	.02	0.75 (0.60–0.95)	–	–	–
Globe artichoke	VTE	IVW	9	.02	0.90 (0.82–0.98)	.459	.520	.488
		MR-Egger	–	.26	0.79 (0.54–1.16)	–	–	–
		Weighted median	–	.02	0.86 (0.76–0.98)	–	–	–
		Weighted mode	–	.08	0.85 (0.72–0.99)	–	–	–
		MR-RAPS	–	.01	0.89 (0.80–0.98)	–	–	–
Ham	VTE	IVW	6	.00	1.27 (1.08–1.49)	.424	.296	.469
		MR-Egger	–	.70	0.88 (0.47–1.64)	–	–	–
		Weighted median	–	.02	1.29 (1.04–1.60)	–	–	–
		Weighted mode	–	.10	1.43 (1.01–2.04)	–	–	–
		MR-RAPS	–	.01	1.27 (1.07–1.52)	–	–	–
Herring	VTE	IVW	12	.01	0.90 (0.82–0.98)	.678	.289	.703
		MR-Egger	–	.14	0.71 (0.47–1.08)	–	–	–
		Weighted median	–	.02	0.87 (0.78–0.98)	–	–	–
		Weighted mode	–	.16	0.87 (0.72–1.04)	–	–	–
		MR-RAPS	–	.02	0.89 (0.81–0.98)	–	–	–
Mackerel	VTE	IVW	6	.02	0.85 (0.74–0.97)	.756	.902	.767
		MR-Egger	–	.64	0.80 (0.34–1.89)	–	–	–
		Weighted median	–	.11	0.88 (0.75–1.03)	–	–	–
		Weighted mode	–	.39	0.89 (0.70–1.13)	–	–	–
		MR-RAPS	–	.03	0.85 (0.74–0.98)	–	–	–
Sardines	VTE	IVW	10	.01	0.87 (0.78–0.96)	.646	.541	.720
		MR-Egger	–	.93	1.02 (0.60–1.74)	–	–	–
		Weighted median	–	.02	0.86 (0.75–0.98)	–	–	–
		Weighted mode	–	.11	0.85 (0.72–1.01)	–	–	–
		MR-RAPS	–	.01	0.86 (0.77–0.96)	–	–	–
alcohol	VTE	IVW	3	.00	0.81 (0.72–0.91)	.668	.949	NA
		MR-Egger	–	.75	0.77 (0.23–2.62)	–	–	–
		Weighted median	–	.00	0.80 (0.69–0.92)	–	–	–
		Weighted mode	–	.11	0.77 (0.65–0.93)	–	–	–
		MR-RAPS	–	.00	0.81 (0.70–0.93)	–	–	–
Cheesecake	PE	IVW	6	.01	1.40 (1.09–1.79)	.186	.065	.198
		MR-Egger	–	.03	4.34 (1.76–10.72)	–	–	–
		Weighted median	–	.14	1.25 (0.93–1.66)	–	–	–
		Weighted mode	–	.66	1.13 (0.68–1.87)	–	–	–
		MR-RAPS	–	.01	1.36 (1.06–1.75)	–	–	–
Chilli pepper	PE	IVW	10	.00	0.83 (0.72–0.94)	.970	.861	.977
		MR-Egger	–	.57	0.76 (0.31–1.89)	–	–	–
		Weighted median	–	.02	0.81 (0.69–0.96)	–	–	–
		Weighted mode	–	.09	0.80 (0.64–1.01)	–	–	–
		MR-RAPS	–	.01	0.83 (0.72–0.95)	–	–	–
Coffee with sugar	PE	IVW	3	.01	1.39 (1.07–1.80)	.476	.878	NA
		MR-Egger	–	.81	2.48 (0.01–858.70)	–	–	–
		Weighted median	–	.05	1.40 (0.99–1.98)	–	–	–
		Weighted mode	–	.21	1.46 (0.98–2.19)	–	–	–
		MR-RAPS	–	.02	1.40 (1.05–1.86)	–	–	–
Globe artichoke	PE	IVW	9	.02	0.86 (0.75–0.98)	.950	.268	.937
		MR-Egger	–	.13	0.63 (0.37–1.06)	–	–	–
		Weighted median	–	.06	0.85 (0.72–1.01)	–	–	–
		Weighted mode	–	.13	0.82 (0.64–1.04)	–	–	–
		MR-RAPS	–	.03	0.86 (0.75–0.98)	–	–	–
Alcohol	PE	IVW	3	.01	0.79 (0.66–0.94)	.404	.458	NA
		MR-Egger	–	.39	0.29 (0.05–1.63)	–	–	–
		Weighted median	–	.12	0.84 (0.68–1.05)	–	–	–
		Weighted mode	–	.33	0.85 (0.67–1.09)	–	–	–
		MR-RAPS	–	.01	0.78 (0.65–0.95)	–	–	–
Gherkins	DVT	IVW	15	.01	1.18 (1.04–1.34)	.835	.928	.856
		MR-Egger	–	.89	0.96 (0.57–1.63)	–	–	–
		Weighted median	–	.09	1.16 (0.98–1.38)	–	–	–
		Weighted mode	–	.36	1.12 (0.89–1.41)	–	–	–
		MR-RAPS	–	.01	1.19 (1.04–1.36)	–	–	–
Oily fish	DVT	IVW	9	.04	0.87 (0.76–1.00)	.610	.521	.626
		MR-Egger	–	.70	1.24 (0.44–3.53)	–	–	c
		Weighted median	–	.06	0.84 (0.71–1.01)	–	–	–
		Weighted mode	–	.33	0.88 (0.69–1.12)	–	–	–
		MR-RAPS	–	.09	0.88 (0.76–1.02)	–	–	–
Alcohol	DVT	IVW	3	.04	0.80 (0.65–0.99)	.574	.532	NA
		MR-Egger	–	.61	2.15 (0.25–18.51)	–	–	–
		Weighted median	–	.24	0.85 (0.65–1.12)	–	–	–
		Weighted mode	–	.44	0.86 (0.63–1.18)	–	–	–
		MR-RAPS	–	.06	0.80 (0.63–1.01)	–	–	–

CI = confidence interval, DVT = deep-vein thrombosis, IVW = inverse-variance weighted, MR = Mendelian randomization, MR-RAPS = Mendelian randomization-Robust adjusted profile score, NSNPs = number of single nucleotide polymorphism, OR = odds ratio, PE = pulmonary embolism, Q_P value = results of Q test, VTE = venous thromboembolism, WM = weighted median.

Figure 3.

The forest plots of the causal effect of food preferences on VTE. (A) Black pepper liking; (B) Chilli pepper liking; (C) Curry liking; (D) Globe artichoke liking; (E) Ham liking; (F) Herring liking; (G) Mackerel liking; (H) Sardines liking; (I) F-strong alcohol liking. VTE = venous thromboembolism.

Figure 6.

The leave-one-out plots of the causal effect of food preferences on VTE. (A) Black pepper liking; (B) Chilli pepper liking; (C) Curry liking; (D) Globe artichoke liking; (E) Ham liking; (F) Herring liking; (G) Mackerel liking; (H) Sardines liking; (I) F-strong alcohol liking. VTE = venous thromboembolism.

Figure 4.

The scatter plots of the causal effect of food preferences on VTE. (A) Black pepper liking; (B) Chilli pepper liking; (C) Curry liking; (D) Globe artichoke liking; (E) Ham liking; (F) Herring liking; (G) Mackerel liking; (H) Sardines liking; (I) F-strong alcohol liking. VTE = venous thromboembolism.

Figure 5.

The funnel plots of the causal effect of food preferences on VTE: (A) Black pepper liking; (B) Chilli pepper liking; (C) Curry liking; (D) Globe artichoke liking; (E) Ham liking; (F) Herring liking; (G) Mackerel liking; (H) Sardines liking; (I) F-strong alcohol liking. VTE = venous thromboembolism.

4. Discussions

A 2-sample MR method was used to explore potential causal links between food preferences on VTE, including DVT and PE. Our findings unveiled causal connections between various dietary habits and different types of VTE. These findings suggested that black pepper, chili pepper, curry, globe artichoke, herring, mackerel, sardines and alcohol consumption were associated with lower VTE. Conversely, ham intake was associated with higher VTE. Additionally, we have shown several possibly risky food preferences for PE and DVT, such as cheesecake, coffee with sugar and gherkins. This research provides a basis for investigating the genetic underpinnings of dietary risk factors associated with VTE, DVT and PE, offering way for future dietary recommendations tailored to patients.

In the present study, we found that spicy foods (such as black pepper, chili pepper and curry) might be a protective factor in VTE, which have not been specifically reported previously. Although all these foods have the function of preventing VTE, the mechanism of their action varies. Black pepper and its main active ingredient piperine have been acknowledged to have beneficial effects on human health.^[31] Obesity and hyperlipidemia, as confirmed by previous MR analysis,^[32] are risk factors for VTE. Evidence has shown that taking capsules containing the active ingredients of pepper (including piperine) can effectively alleviate obesity-related complications and suppress inflammatory responses for overweight individuals.^[33] Unlike pepper, curry is rich in curcumin, which has been confirmed by modern researches to play an antithrombotic role by inhibiting platelet aggregation and prolonging the clotting time, together with anti-inflammatory effects.^[34,35] However, studies on the role of those spicy foods in VTE were limited. Current work was the first to reveal that preference for spicy foods could be a protective factor for VTE. We suggest continuing studies to fully evaluate the untapped potential of those foods in VTE and to better treat CVD patients.

Alcohol intake was identified as a slightly protective factor for VTE in this MR investigation, consistent with most of the prior studies. A recent meta-analysis of fourteen cohorts and 4 case-control studies indicated that modest alcohol consumption is linked to reduced VTE risk.^[36] Comparable findings were reported in a separate meta-analysis.^[37] Published studies showed that alcohol intake was linked to reduced DVT occurrence and decreased mortality risk at 90 days and 1-year post-surgery.^[38,39] The mechanisms of the protective effects of alcohol on VTE are not fully understood and may be related to the following factors. Firstly, alcohol can directly disrupt the synthesis of coagulants and anticoagulants by altering liver function.^[40] The studies have shown that low to moderate alcohol intake is linked to decreased levels of fibrinogen, plasma viscosity, von Willebrand factor, and factor VII.^[41] Besides, higher alcohol levels progressively impair blood clotting, significantly prolonging blood clot formation time.^[42] Secondly, alcohol impacts platelet number and function by inducing apoptosis and inducing Ca2 + mobilization.^[43,44] In addition, evidence suggests that alcohol effects on VTE formation involve not only the coagulation system but also the immune and inflammatory responses. Some alcohol compounds, like polyphenols, possess anti-inflammatory and antioxidant properties that may protect against VTE.^[45] Further studies on molecular mechanisms are required to elucidate the relationship between alcohol and VTE.

Data on the associations of coffee consumption with VTE in general are scarce. We observed no genetical association between coffee consumption and VTE, which was in line with the previous studies.^[5,8,46] However, an unexpected finding in our study was that preference for sugar-sweetened coffee and cheesecake, which are high in sugar, might increase the risk of PE. VTE has been strongly linked to inflammation.^[47] Prior research has indicated that elevated blood sugar levels may play a role in the onset of VTE by triggering inflammation.^[48] Mraovic et al provided evidence supporting this association, revealing that a significant link between high blood glucose levels (>200 mg/dL) and the occurrence of PE following orthopedic procedures.^[49] Another retrospective study using an USA cohort found that positive association between elevated admission BG and increased mortality with PE (OR,1.60; 95% CI, 1.26–2.03).^[50] Similar, a cohort study from Israel population obtained the same results (HR, 2.3; 95% CI, 1.2–4.5).^[51] This reminds that VTE can be prevented by reducing sugar intake or treating diabetes.

Seafood, with its high-quality proteins, unsaturated fatty acids, minerals, and essential vitamins, is crucial for human health.^[52] The studies have shown that regular intake of seafood is linked to a lower risk of cardiovascular disease-related mortality.^[53] An inverse association of fish consumption, especially marine fish (herring, mackerel and sardines) with VTE was detected in our MR analysis. Several published studies have examined the link between fish consumption and VTE, yielding conflicting results.^{[5–7,10,54]} Recent meta-analysis of 8 prospective cohort studies conducted by Zhang et al^[10] found no significant association between fish intake and VTE risk (RR, 1.02; 95% CI, 0.93–1.11; P = .709). However, the consumption of omega-3 fatty acids was linked to a reduced risk of VTE (RR, 0.89; 95% CI, 0.80–0.98; P = .024) and recurrent VTE (RR, 0.45; 95% CI, 0.25–0.81; P = .008).^[10] Thus, the mechanisms of the protective effects of fish consumption on VTE may be related to omega-3 fatty acids contained in fish. A randomized control trial has demonstrated that omega-3 fatty acids supplementation can reduce the incidence of VTE after orthopedic surgery.^[55] The impact of omega-3 fatty acids on VTE likely involves their ability to modulate inflammation,^[56] platelet activity,^[57] interactions between platelets and endothelium^[58] and expression of tissue factor^[59,60] – critical pathways in VTE development. It necessitates additional investigation of the underlying molecular mechanisms.

The outcomes derived from observational studies of meat consumption and the incidence of VTE were still controversial to date. Our findings that red meat is not related to risk of VTE are consistent with the most observational studies and meta-analysis.^[9] However, we found that ham intake, which a type of processed meat, was associated with an increased risk of VTE (OR, 1.27; 95% CI,1.08–1.49; P = .003). Red or processed meat is rich in saturated fatty acids and other unhealthy substances, and excessive intakes of saturated fats can contribute to chronic inflammation in the body, which is associated with thrombosis profile.^[47] Although meat intake was not found to be associated with VTE in our study, based on earlier research by Steffen et al in 2007, there has been broad consensus that decreasing red or processed meat intake could lower the likelihood of VTE.^[6] Indeed, further studies are needed to evaluate the potential VTE risk associated with various types of meat intake in the future.

Interestingly, our findings revealed a previously unrecognized causal link between globe artichoke intake and VTE, demonstrating that its increased consumption correlates with a reduced risk. The precise biological pathways driving this relationship are not yet fully understood, though it may be attributed to the elevated levels of polyphenols and terpenes present in globe artichoke, which could contribute to lowering lipid levels and mitigating VTE risk.^[61,62]

Our study has several advantages. It is the inaugural MR investigation examining the causal impact of diets on various VTE subcategories, effectively mitigating confounding variables and reverse causation through MR analysis. Following adjustment for established risk factors like BMI, we observed persistent associations between specific dietary patterns and VTE subtypes, suggesting additional potential mechanisms warranting exploration in future research. Furthermore, our study leveraged MR to encompass 139 food preferences and 3 VTE phenotypes utilizing data from the most extensive dataset available and employing various sensitivity analyses to increase the robustness and validity of our findings.

However, this study also has certain limitations. Firstly, while the GWAS in our study featured the most extensive sample size for VTE, it was limited by exclusive focus on the European population, potentially impacting the broader applicability of the findings across all demographic groups. Secondly, VTE is the result of multiple factors, we could not rule out the influence of factors other than diet, such as physical activity, which might have an impact on our results. Thirdly, after false discovery rate adjustment, no significant causal link was observed between food preference and VTE, indicating the necessity for further research to validate their association. Finally, although we employed multiple sensitivity analyses such as MR-Egger and MR-PRESSO and detected no significant horizontal pleiotropy, residual pleiotropy remains an inherent limitation that cannot be entirely ruled out. It is worth noting that for the primary findings of this study, the results from the IVW, weighted median, and MR-PRESSO methods demonstrated a high degree of consistency. This stable estimation across methods with different robustness assumptions for pleiotropy strengthens the reliability of our conclusions to some extent, suggesting that the main causal estimates are unlikely to be entirely driven by directional pleiotropy. Nonetheless, these associations should still be interpreted with caution and await further validation through more detailed biological mechanism exploration in future research.

5. Conclusion

Our study provides evidence for a causal relationship between dietary factors and VTE. Most importantly, we identified that black pepper, chili pepper, curry, globe artichoke, herring, mackerel, sardines and alcohol consumption were negatively associated with VTE risk while ham intake was associated with higher VTE. The findings hold significant relevance for preventing, managing, and addressing VTE in prevalent locations by implementing appropriate dietary adjustments.

Acknowledgments

We thank all the consortium studies for making the summary association statistics data publicly available.

Author contributions

Conceptualization: Shuanghui Hu, Guoliang Huang.

Data curation: Shuanghui Hu, Hanlin Li, Lujie Wu.

Formal analysis: Shuanghui Hu.

Methodology: Jinshi Lin, Jiangping Li, Zuxiong Su, Bingwen Li.

Project administration: Hanlin Li, Lujie Wu, Limei Ye.

Supervision: Limei Ye, Jinshi Lin, Po Mao, Jiangping Li.

Writing – original draft: Shuanghui Hu.

Writing – review & editing: Shuanghui Hu, Guoliang Huang.

Supplementary Material

medi-104-e46296-s001.pdf ^{(1.3MB, pdf)}

medi-104-e46296-s002.xlsx ^{(1.1MB, xlsx)}