Thromboembolic events and cardiovascular mortality in inflammatory bowel diseases: A meta-analysis of observational studies ☆

Objective: Patients with inflammatory bowel disease (IBD) are at increased risk of having venous thromboembolism. The magnitude of this risk has yet to be determined. The question of whether IBD patients have an increased risk of arterial thromboembolism and cardiovascular (CV) mortality remains controversial. Design: We searched MEDLINE, Cochrane Library, EMBASE and international conference abstracts and included all controlled observational studies that evaluated the incidence of venous and/or arterial thromboembolic events (TE) and CV mortality in adult IBD. Results: 33 studies enrolling 207,814 IBD patients and 5,774,898 controls and capturing 3,253,639 hospitalizations of IBD patients and 936,411,223 hospitalizations of controls reported Abbreviations: CD, CV, cardiovascular; DVT, deep venous thrombosis; PE, pulmonary embolism; IHD, ischemic heart disease; RR, relative risk; OR, odds ratio.


Introduction
Venous thromboembolism (TE) including deep venous thrombosis (DVT) and pulmonary embolism (PE) is a well-known and feared complication of inflammatory bowel diseases (IBDs). 1,2 The incidence of venous TE is estimated to be 0.26% per year in both Crohn's disease (CD) and ulcerative colitis (UC). 3 However, despite several large population-based studies, 4 the true magnitude of the risk remains unclear as a result of methodological differences and heterogeneity across studies. Unlike venous TE, the risk of arterial TE and cardiovascular events in IBD is not well understood. Inflammation is involved throughout all stages of atherosclerosis pathogenesis, from plaque initiation to rupture and subsequent thrombosis. 5 C-reactive protein, often elevated during IBD flares, has also been associated with an increased risk of coronary artery disease independent of traditional cardiovascular risk factors. 6 Of note, other chronic inflammatory diseases such as rheumatoid arthritis are also associated with an increased risk of arterial TE and cardiovascular mortality. 7,8 The aim of this meta-analysis was to determine the risk of venous and arterial TE, as well as the risk of cardio-vascular events and mortality in patients with IBD compared to the general population in referral center and population-based cohorts.

Materials and methods
We conducted a systematic review of the literature following the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines.

Literature search and selection criteria
We conducted a computerized search of English and non-English language publications listed in the electronic databases of PUBMED (1966 to September 2012), the Cochrane Library (to June 2012), and EMBASE (1980 to June 2012), by two independent researchers (MF, CX). We searched for the following terms: "Inflammatory bowel disease", "Crohn's disease", "ulcerative colitis", "thrombosis", "arterial thrombosis", "heart disease", "vascular disease", "atherosclerosis", "coronary artery disease", "myocardial infarction", "cerebrovascular disorders", "stroke", "mesenteric ischemia", "peripheral artery disease", "deep venous thrombosis", "pulmonary embolism", "venous thromboembolism", "mortality cause specific" and "mortality". We also hand-searched abstracts from the annual meetings of Digestive Disease Week (2009 to 2012), the United European Gastroenterology Week (2008 to 2011), and the European Crohn and Colitis Organization congress (2009 to 2012) over the past three years, as well as references from review articles, meta-analyses, and published observational studies in order to identify additional articles. We did not employ any search software. Data abstraction was carried out independently by two investigators (MF, CX) using standardized data collection form. Discrepancies in data interpretation were resolved by a discussion and re-review of the studies and by consultation with the other clinical authors (LPB, LD). We selected peer-reviewed observational controlled data (case-control and cohort studies) originating from referral center, hospital and population based-studies. If data from a single study was reported in more than one article, only the results from the most recent study were included in the meta-analysis. If a population contributed to more than one publication, each publication could be included only if different study periods were involved.

Selection criteria
Inclusion and exclusion criteria were defined before commencement of the literature search. Selected peer-reviewed studies (case control and cohort studies) were included for analyses if all participants in the study were adult patients with IBD and if they reported either (a) a risk of thrombotic events in IBD (UC or CD) patients and controls expressed as odds ratios (ORs) or relative risks (RRs) with associated 95% confidence intervals or data for calculating them or (b) cardiovascular-disease-specific Standardized Mortality Ratio (SMR) with 95% confidence interval for IBD (UC or CD).

Outcome measures
The outcome measures were defined a priori. The meta-analysis evaluated different outcome variables including venous and arterial thromboembolic events and cardiovascular mortality. We evaluated ORs or RRs of thrombosis among IBD patients versus controls, RRs or ORs of arterial TE among IBD patients versus controls, and the proportion of venous TE in IBD patients and controls. We calculated a CV-disease-specific standardized mortality ratio (SMR) in IBD patients. When data were available we evaluated the RRs or ORs of ischemic heart disease (IHD), stroke, mesenteric ischemia, peripheral artery disease, DVT and PE in IBD patients and controls.

Statistical analysis
We calculated weighted-pooled summary estimates of RR (pooled RR) for all thrombotic events combined, arterial and venous thromboembolism, as well as for each outcome individually (arterial TE, IHD, Stroke, mesenteric ischemia, peripheral artery disease, venous TE, DVT, and PE). Similarly we calculated weighted-pooled summary estimates of the SMR for studies specifically evaluating CV mortality. Analyses were performed if at least two studies evaluating the same outcome could be combined.
For each meta-analysis, the method of Der Simonian and Laird 9 was used. According to this method, studies were considered as a random sample from a population of studies. Statistical heterogeneity was tested for each analysis. 10,11 Due to heterogeneity among studies a random effect model was used to analyze data. The overall effect was estimated by a weighted average of individual effects, with weights inversely proportional to the variance in observed effects. The effect measures estimated were the relative risks between the IBD and control groups, with 95% confidence intervals (CI). 12 Metaregression was performed to assess the modulation effect of pathology (UC or CD). All analyses were performed using R software 13 and metaphor package. 14

Pulmonary embolism
We identified 3 studies assessing the risk of PE in IBD patients 23

Stroke
We identified 3 studies assessing the risk of stroke in IBD patients 15 (Fig. 3), with no difference between CD and UC (p = 0.62).

Cardiovascular mortality
We identified 15 studies evaluating CV mortality among IBD patients 33-47 (

Heterogenity and publication bias
A statistical heterogeneity was observed for all analysis. The first analysis evaluating the risk of overall risk of thrombosis was associated with significant heterogeneity among studies (I2 98%, Peth b 0.001). A funnel plot ( Supplementary Fig. 1) showed some asymmetry and suggested publication bias, as there were few studies with high precision (large sample size) and large RR. Nevertheless, the result of Egger's regression test for asymmetry was not significant (egger = 0.93). For the pooled analysis of cardiovascular mortality the funnel plot ( Supplementary Fig. 2) showed some asymmetry (egger = 0.47) and there was a significant heterogeneity among studies (I2 70%, p b 0.05, Phet b 0.001).

Discussion
This is the first meta-analysis of observational studies designed to assess incidence of both venous and arterial thromboembolic events as well as cardiovascular mortality in patients with IBD. Despite great heterogeneity in design and clinical setting, our findings demonstrate that patients with IBD are at major risk for venous thromboembolism and mesenteric ischemia and, to a lesser degree, arterial thromboembolism and ischemic heart disease. Importantly, we did not find an increase in the risk of cardiovascular mortality in IBD patients. Venous thromboembolic complications, including DVT and PE 1 have been shown to be associated with IBD, but the magnitude of the thromboembolic risk is somewhat disputed. 48 When analyzing a total of 5,982,712 patients and 939,664,862 hospitalizations, we found a 60% increase in the risk of TE in patients with IBD compared to the general population. This increase was largely attributed to venous TE events including both DVT and PE. Similar to the study by Grainge et al., 24 the risk of venous TE was lower in studies that included only hospitalized IBD patients. While this observation may initially appear counterintuitive given that hospitalized patients typically have the most severe disease, it may be explained by the use of prophylactic heparin in hospitalized patients in accordance with international recommendations. 49,50 Reduction of venous thromboembolism by low molecular weight heparin or the new orally bioavailable anticoagulants, 51-54 may be further highlighted in future guidelines for both hospitalized and ambulatory patients. In studies including only inpatients, VTE increased risk was greater in UC than in CD patients (p = 0.0029). However, prophylaxis should be recommended both in UC and in CD.
Several studies have shown an increased carotid intima-media thickness in IBD patients, including pediatric cases, suggesting that atherosclerosis may be an early complication of these diseases. [55][56][57][58] When compared to the general population, the overall risk of arterial thrombosis was not significantly increased in IBD patients. However, there was a significant increased risk of ischemic heart disease and mesenteric ischemia. This increased risk was observed despite the absence of the traditional cardiovascular risk factors such as arterial hypertension, dyslipidemia, diabetes mellitus, age, male and family history of cardiovascular disease. 16,18 The magnitude of the risk was similar in patients with CD who are more often smokers than the general population and in patients with UC who are often non-smokers. 59 Chronic inflammation may be the most important driver of cardiovascular complications in IBD. C-reactive protein and interleukin-6, often elevated during IBD flares, have been associated with an increased risk of coronary artery disease and mesenteric ischemia independent of other cardiovascular risk factors. 6,[60][61][62] In IBD, the increased expression of CD40L by platelets appears to contribute to the pro-inflammatory response 1,63 while in atherosclerosis, the inflammation caused by CD40L-CD40 interaction leads to unstable plaque resulting in thrombosis. 64 An increase in CV mortality compared to the general population is well documented in rheumatoid arthritis (RA) with higher death rates resulting from ischemic heart disease and cerebrovascular occlusion. 7,8 As a result, aggressive cardiovascular disease primary prevention has become standard of care in RA. 65 In contrast, a meta-analysis of 11 observational studies published 5 years ago found no increased risk of CV mortality in IBD patients. 66 We confirmed and reinforced these findings by pooling 69,383 patients including those from the most recent European studies. [33][34][35]47 In RA part of this risk appears to be mediated by long-term inflammation. Traditional CV risk factors as smoking or metabolic syndrome also play an important role. 67 In IBD, the impact of systemic inflammation on the cardiovascular risk may be different. Moreover, except smoking, more common in patients with Crohn's disease, 68 the prevalence and impact of traditional cardiovascular risk factors remains unknown. This discrepancy between IBD and RA may also be explained by the inclusion in most IBD studies of a majority of young patients with a follow-up of less than 5 years.
Our study is not without limitations. We included cohorts from a variety of clinical settings with differences in diagnostic criteria, age at enrollment, period at risk and study design. In addition, the methodology, inclusion criteria and outcome measures differed between studies. Some studies are population-based cohorts, some are from referral centers, and others look only at hospitalized patients through discharge databases. As expected we found statistically significant heterogeneity among the studies included in this meta-analysis. Therefore we used a random-effect model to give an estimate of variability. 69 Statistical heterogeneity was identified, because the magnitude of the association was different between studies. However the risk was consistent despite different study designs, populations and methods. Therefore our results confirmed higher risk of venous TE in IBD. The high heterogeneity observed suggests that relative risk may differ according to clinical setting therefore precise relative risk is hard to assess. In addition there were not enough studies to statified analysis by clinical setting. We determined that cohort source (patients or hospitalizations) explained some of the observed heterogeneity. Studies from large administrative databases of hospital admissions were included in the analysis. These studies report opposite results than other included studies. These different results may be explained by methodological defect. In addition, these studies have an important weight in the overall analysis. For this reason, sub-analysis including these studies was separately performed. Of note, most of the studies we included were retrospective with a relatively short average median duration of follow-up. Furthermore, well-established risk factors for TE such as smoking, personal or family history of cardiovascular events or BMI were not available in all studies and our results therefore could not be adjusted for them. Also, as a result of missing data, we were also unable to take into account age, sex, disease extension, disease severity and disease duration. Finally, we could not analyze the impact of IBD-related Figure 3 Meta-analysis of studies on arterial thromboembolic events in IBD patients, including stroke, peripheral artery disease (Periph. Art. Dis.), mesenteric ischemia (Mesent. Isch.) and ischemic heart disease (Isch. Heart Dis.). TE, thromboembolism; CD, Crohn's disease; UC, ulcerative colitis. medications on our findings. This underlines the fact that new prospective high quality studies evaluating the risk of thrombosis in IBD are required.
In conclusion, IBD is associated with an overall increased risk of thrombovascular events. As a result, the prevention of DVT and PE in particular should be at the forefront of gastroenterologists' concerns. Unfortunately this is not the case: in a recent survey involving 591 US physicians, 29% were unaware of any recommendations addressing pharmacologic prophylaxis included in American College of Gastroenterology IBD guidelines. Furthermore only 35% reported that they would give pharmacologic VTE prophylaxis to a hospitalized patient with severe ulcerative colitis. 70 In clinical practice, one study found that only 50% of patients hospitalized for severe ulcerative colitis at a referral center received pharmacologic venous thromboembolism prophylaxis. 71 With respect to the risk of ischemic heart disease and mesenteric ischemia in IBD patients, more studies are needed to further characterize risk factors such as whether a tight control of inflammation could ultimately prevent these potentially devastating events.

Disclosures
No conflicts of interest exist in this manuscript for any author.

Financial support
None.