Article Text

Download PDFPDF

Thromboembolic risk among Danish children and adults with inflammatory bowel diseases: a population-based nationwide study
  1. Michael D Kappelman1,
  2. Erzsebet Horvath-Puho2,
  3. Robert S Sandler3,
  4. David T Rubin4,
  5. Thomas A Ullman5,
  6. Lars Pedersen2,
  7. John A Baron3,6,
  8. Henrik T Sørensen2
  1. 1Department of Pediatrics, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA
  2. 2Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark
  3. 3Department of Medicine, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA
  4. 4University of Chicago, Chicago, Illinois, USA
  5. 5Mount Sinai School of Medicine, New York, USA
  6. 6Department of Medicine, Dartmouth Medical School, Hanover, New Hampshire, USA
  1. Correspondence to Dr Michael Kappelman, University of North Carolina Chapel Hill, Department of Pediatrics, Division of Paediatric Gastroenterology, 130 Mason Farm Road, campus box 7229, Chapel Hill, NC 27599, USA; michael_kappelman{at}med.unc.edu

Abstract

Background Recommendations for venous thromboembolism (VTE) prophylaxis in patients with inflammatory bowel disease (IBD) can be refined by incorporating patient-specific risk factors.

Objectives To determine the risk of deep venous thrombosis (DVT) and pulmonary embolism (PE) in children and adults with Crohn's disease and ulcerative colitis and evaluate whether this risk varies by age and/or presence of other risk factors.

Methods We performed a cohort study using Danish administrative data. Incidence rates of DVT and PE were calculated among patients with IBD and an age- and gender-matched comparison population and compared using Cox proportional hazards regression. We performed additional analyses stratifying by age, gender and disease type and restricting outcomes to unprovoked events (occurring without known malignancy, surgery, fracture/trauma or pregnancy). We next performed a nested case–control study to adjust for additional co-morbidities (congestive heart failure, diabetes, myocardial infarction and stroke) and the use of hormone replacement and antipsychotic medications.

Results The study included 49 799 patients with IBD (14 211 Crohn's disease, 35 229 ulcerative colitis) and 477 504 members of the general population. VTE risk was elevated in patients with IBD (HR=2.0 (95% CI 1.8 to 2.1) for total events, HR=1.6 (95% CI 1.5 to 1.8) for unprovoked events). Although the incidence of VTE increased with age, the RR was higher in younger patients. Among those ≤20 years old, HRs were 6.0 (95% CI 2.5 to 14.7) for DVT and 6.4 (95% CI 2.0 to 20.3) for PE. After further adjusting for co-morbidity and medication use in the case-control analysis, ORs for all events remained in the 1.5–1.8 range.

Discussion Patients with IBD have twice the incidence of PE or DVT as does the general population. This risk persisted after taking into account other VTE risk factors. Relative risks were particularly high at young ages, though actual incidence increased with age. These findings can further inform risk–benefit analysis of VTE prophylaxis.

  • Crohn's disease
  • ulcerative colitis
  • inflammatory bowel disease
  • deep venous thrombosis
  • pulmonary embolus
  • venous thromboembolism
  • IBD
View Full Text

Statistics from Altmetric.com

Significance of this study

What is already known about this subject?

  • The risk of venous thromboembolism is increased in patients with inflammatory bowel disease (IBD), as compared with the general population.

  • The risk of venous thromboembolism in IBD may be highest during periods of active disease and/or hospitalisation.

  • Other risk factors for venous thromboembolism include patient age, hypercoagulable states, and the use of some medications.

What are the new findings?

  • This study confirms that patients with IBD have twice the risk of developing venous thromboembolism as compared with the general population.

  • This study demonstrates the risk of venous thromboembolism associated with IBD is dependent upon patient age. Younger patients have higher RR but lower absolute risk as compared with older patients.

  • The risk of venous thromboembolism associated with IBD is higher in patients with other known thromboembolic risk factors, but remains after adjusting for these risk factors.

How might it impact on clinical practice in the foreseeable future?

  • Recommendations for thromboembolic prophylaxis could be refined by taking into account patient age and the presence of other known thromboembolic risk factors.

  • The risk of venous thromboembolism in patients with IBD persists after adjusting for many of the other known thromboembolic risk factors suggesting that IBD is, itself, a hypercoagulable state.

Background

Venous thromboembolism (VTE), including deep venous thrombosis of the lower limb (DVT), pulmonary embolism (PE), and superior sagittal sinus thrombosis (SSST), are common and often serious events affecting 2 per 1000 persons per year in Western populations.1 All types of VTE are associated with high mortality. Well-established risk factors include surgery, cancer, fractures, immobilisation, pregnancy and the use of oestrogens.2 3

Over the past 70 years, numerous case reports and case series have described an association between VTE and inflammatory bowel diseases (IBDs), comprised of Crohn's disease and ulcerative colitis.4–7 This risk has also been observed in several population-based8–10 and hospital-based epidemiological studies11 and a study of hospitalised patients,12 and has been associated with increased mortality associated with ulcerative colitis.13 A single study failed to demonstrate an association between IBD and VTE.14

Based on these epidemiological studies, current guidelines recommend VTE prophylaxis or consideration of prophylaxis for hospitalised patients with IBD,15–17 and the American Gastroenterological Association has proposed VTE screening and prophylaxis as quality indicators in draft IBD quality measures submitted for public comment in early 2010. However, these recommendations have been based on overall risk, and do not stratify based on patient-specific risk factors. Available evidence suggests that they could be refined by taking into account other known thromboembolic risk factors including patient age, prior malignancy, recent fracture, surgery, pregnancy, other co-morbidities (heart failure, diabetes, myocardial infarction and stroke2 3), and the use of prothrombotic medications, such as hormone replacement therapy18 and antipsychotic medications.19 Indeed, it is well-recognised that patients with IBD have increased autoimmune disorders and other co-morbidities,20–22 experience complications such as osteoporosis and fracture,23–25 and frequently undergo surgery to treat their IBD. Whether the increased risk of VTE in patients with IBD is independent of these additional factors remains unknown.

The mechanism of a prothrombotic state in IBD is not known. Although it is widely thought that IBD is itself a prothrombotic state and many haemostatic abnormalities have been reported in patients with IBD,26 27 the prevalence of inherited causes of thrombophilia has been found to be no higher in patients with IBD than controls.26 28 Epidemiological studies that effectively control for acquired risk factors (ie, co-morbidity, need for surgery, etc.) are needed to quantify the degree to which IBD is an independent risk factor for thrombosis and to inform future prevention strategies.

In addition, although several studies have reported decreased rates of VTE in young persons with IBD compared with older adults with this condition,8 12 14 the risk of VTE in the paediatric IBD population has yet to be systematically evaluated. This is likely due to insufficient sample size in prior studies, given the low prevalence of both IBD and VTE in younger patients.

In order to determine age-, gender- and disease-specific estimates of both the absolute and RR of thromboembolic events in individuals with IBD and more fully evaluate whether IBD is an independent risk factor for such events, we performed a series of nationwide, population-based cohort and case–control analyses in Denmark. These results could provide strong evidence on which to base the development of screening and prophylaxis recommendations.

Methods

Description of data sources

The Danish Civil Registration System (DCRS) assigns a personal identifier to each Danish resident at the time of his or her birth or immigration to Denmark, and also monitors the occurrence of death or emigration from the country.29 The DCRS can be used as a population registry in epidemiological studies, facilitating the sampling of population-based controls and/or providing denominator data for determining event rates. The civil registration identifier can be used to link patients throughout the numerous health registries and socioeconomic datasets maintained in Denmark.

The Danish National Patient Registry (DNPR), which can be linked to the DCRS, contains records on 99.4% of all hospital discharges occurring after 1 January 1977 in Denmark. Data elements include civil registration number, dates of hospital admission and discharge, surgical procedures, and up to 20 discharge diagnoses for each hospitalisation. Since 1995, the database has also included hospital outpatient visits, accounting for essentially all specialist medical care in Denmark. All inpatient and outpatient hospital visits are assigned diagnosis codes according to the International Classification of Diseases (8th revision, (ICD-8) until the end of 1993 and 10th revision (ICD-10) thereafter).

The Prescription Databases of the Central Denmark Region and the North Denmark Region (covering one-third of the Danish population) contain records of all filled prescriptions, including date of dispensing, name of medication, strength and quantity. Data are complete starting in 1989 for the former North Jutland county, 1996 for the former Aarhus county, and 1998 for the former Viborg and Ringkjobing counties. (The Danish county system was reorganised in 2007).

Cohort analysis

We first performed a cohort analysis to determine the absolute and RR of venous thromboembolic events in patients with IBD and a matched, non-IBD, comparison population.

Study population and IBD exposure classification

The source population consisted of all Danish residents during the time period spanning 1 January 1980 through 31 December 2007. We used the DNPR to identify patients with Crohn's disease and patients with ulcerative colitis. We considered patients with ICD-8 code 563.01 and/or ICD-10 K50 to have Crohn's disease, and patients with ICD-8 code 563.19 and/or ICD-8 code 569.04, and/or ICD-10 K51 to have ulcerative colitis. Patients with codes for both conditions were considered to have unspecified IBD. Prior work has validated the completeness and accuracy of the DNPR for identifying patients with IBD30 and other chronic illnesses.31 For IBD, a validation study compared diagnoses of Crohn's disease and ulcerative colitis recorded in the database with hospital and pathology records and found the sensitivity to be 94% for both Crohn's disease and ulcerative colitis, and the specificity to be 97% for Crohn's disease and 90% for ulcerative colitis, respectively. For each patient with IBD, the follow-up period began on the date of the first IBD-related encounter occurring after 1 January 1980 (3 years after the Danish National Patient Registry began recording discharge data) or 3 years after his/her immigration to Denmark. The 3 year window prior to the start of exposure time was selected in order to identify both incident and prevalent cases of IBD prior to the assignment of exposure status. Follow-up continued until the occurrence of a study outcome (DVT, PE or SSST), emigration from Denmark, a 100th birthday, or death, whichever came first.

For each patient with Crohn's disease, ulcerative colitis or unspecified IBD, 10 comparison cohort (ie, without an IBD diagnosis) members were randomly chosen from within the entire Danish Population using the DCRS registry, matched on sex and age at the time of IBD diagnosis of the corresponding IBD patient. Follow-up time for unexposed persons began on the date of the first IBD code of the corresponding IBD patient, and continued until occurrence of a study outcome (DVT, PE or SSST), emigration from Denmark, a 100th birthday, or death, whichever came first.

Classification of venous thromboembolic outcomes

We searched the primary and secondary diagnosis codes in the DNPR to identify incident cases of DVT in the lower limb (code 451.00 in ICD-8 and codes I80.1-I80.3 in ICD-10) and PE (code 450.99 in ICD-8 and codes I26.0 and I26.9 in ICD-10). This approach to identifying diagnosed cases of DVT and PE has been described previously32 33 and recently validated against chart review with a positive predictive value of 58.5% for VTE diagnosed in any setting, and 75% for VTE diagnosed during an inpatient admission.34 We also identified SSST events using code 321.xx in ICD-8 and codes I67.6 and I63.6 in ICD-10. For patients with multiple thromboembolic events, the date of the first event was used in the analysis. Patients with DVT and PE occurring concomitantly were included in the PE group.

A secondary analysis was conducted on ‘unprovoked’ thromboembolic events, defined as outcomes occurring in the absence of a diagnosis of fracture, trauma, surgery, or pregnancy within 90 days prior to a thromboembolic event or a diagnosis of cancer at any time prior to or within 90 days following the event.35

We also performed a sub-analysis focussing only on the period from 1 January 1995 until 31 December 2007, when both inpatient and outpatient visits were recorded in the DNPR.

Statistical analysis

Incidence rates of DVT, PE and VTE (combined DVT, PE and SSST) were calculated in the IBD population and the matched unexposed population, and expressed as rates per 10 000 person-years. Because SSST events are so uncommon and its diagnosis codes have not been validated, separate analyses of SSST were not performed. Incidence rates were also stratified by disease type (Crohn's disease, ulcerative colitis, and unspecified), age group (0–20 years, 21–40 years, 41–60 years, and ≥61 years), and gender. HRs and 95% CIs, as a measure of RR, were then calculated using Cox proportional hazards regression, with adjustment for age and gender.

Case–control analysis

We performed a nested case–control analysis to determine whether IBD itself (rather than co-morbidies that may serve as potential confounders) is independently associated with venous thromboembolism.

Identification of VTE cases

Cases of DVT, PE, and SSST occurring between 1 January 1980 and 31 December 2007 were identified using the diagnosis codes specified above. The date of the first diagnosis code was assigned as the index date for the event. Events occurring in individuals 100 years of age or older were considered outliers and excluded from the analysis. We also performed secondary analyses of ‘unprovoked’ events as described above for the cohort analysis.

Identification of population controls

For each case of DVT, PE or SSST, we randomly selected five controls from the general population, from the DCR registry, matched by age and gender. Each control was assigned an index date identical to the date of VTE diagnosis for his/her matched case. Controls had to be alive with no recorded VTE prior to this date. This technique of risk set sampling allows determination of unbiased estimates of corresponding rate ratios in a similar cohort study. As required by this study design, all controls thus had the opportunity to become cases at a future date.

Identification of inflammatory bowel disease

Diagnoses of Crohn's disease, ulcerative colitis and unspecified IBD were extracted through the DNPR as described above. Exposure status was ascertained during the period from 1 January 1977 (or the date the case or control was entered into the DCRS) until the diagnosis date of the VTE for the case (or index date for the control).

Measurement of co-variates

We used the DNPR to ascertain the presence of one or more diagnosis codes for each of the following conditions that, based on prior literature, may confound the relationship between IBD and thromboembolic events: congestive heart failure,36 diabetes,21 37 myocardial infarction22 38 39 and stroke (appendix 1).22 38 39

A sub-analysis focused on residents of several regions of Denmark where outpatient prescription medications have been recorded for over 10 years (since 1989 for the former North Jutland county, since 1996 for the former Aarhus county, and since 1998 for the former Viborg and Ringkjobing counties. (These regions are now known as the Central Denmark Region and the North Denmark Region.) In the sub-analysis, we identified cases occurring 1 or more years following the inclusion of prescription claims in the database. This allowed us to adjust for potential confounding through the use of post-menopausal hormone replacement therapy and antipsychotic medications which are known to be associated with VTE.18 19 Each case was then matched to 10 population controls using risk set sampling as described above. The use of medications was coded ‘recent’ if at least two prescriptions were filled in the year prior to the event, ‘remote’ if 2 or more prescriptions were filled more than 1 year prior to the event, and ‘persistent’ if 2 or more prescriptions were filled in both time periods.

As with the cohort analysis, we also performed a sub-analysis covering the period between 1 January 1995 and 31 December 2007, when both inpatient and outpatient visits were recorded in the DNPR.

To evaluate the association between IBD disease activity and risk of thromboembolysis, we performed a final case control analysis involving patients with IBD residing in the Central Denmark Region and the North Denmark Region, for which outpatient prescription data have been recorded. In this analysis, patients with IBD with VTE were matched to patients with IBD without VTE on the basis of age, gender and county of residence. Exposures of interest in this analysis included (1) an outpatient prescription for oral corticosteroids at or within 30 days of VTE event and/or index date for controls, (2) a hospitalisation with a primary discharge diagnosis of Crohn's disease or ulcerative colitis at or within 30 days of a VTE event and/or index date for controls, and (3) either of these exposures.

Statistical analyses

We first constructed contingency tables for the each thromboembolic outcome separately, as well as for the combined outcome of all thromboembolic events. As in the cohort analysis, separate analyses of SSST were not performed because these events are so uncommon and their diagnosis codes have not been validated. Crohn's disease and ulcerative colitis exposures were analysed separately. The combined exposure of all types of IBD (Crohn's disease, ulcerative colitis or unspecified IBD) was also analysed. Conditional logistic regression was used to determine crude and adjusted ORs (and 95% CIs) after adjusting for potential confounders including presence of events or conditions that provoke VTE, co-morbidities, and, in a sub-analysis, medication use. In this design, ORs are an unbiased estimate of the incidence rate ratio. Analyses of ‘unprovoked events’ were adjusted only for co-morbidities and, in the sub-analysis, for medication use. As each of the potential confounder was considered of interest, all were retained in final models.

All statistical analyses were performed using SAS version 9.1, and the study protocol was approved by the Danish Ethical Review Board. The protocol was granted exemption from review by the Institutional Review Board at the University of North Carolina at Chapel Hill because it involved use of existing, de-identified data.

Results

Cohort analysis

The study population included 49 799 patients with IBD (14 211 Crohn's disease, 35 229 ulcerative colitis, and 359 unspecified IBD) and 477 504 members of the matched general population comparison cohort. The demographics of IBD population (exposed) and the matched, non-IBD comparator population (unexposed) are shown in table 1. The cohorts were followed for a total of 5 448 511 person-years (492 344 exposed person-years and 4 956 167 unexposed-person years). A total of 7827 venous thromboembolic events occurred (1181 in exposed persons and 6646 in unexposed persons).

Table 1

Demographic characteristics of the inflammatory bowel disease (IBD) population and a matched non-IBD cohort, Denmark, 1980–2007

The number of events and incidence rates of VTE, DVT and PE in patients with IBD and the matched comparison cohort are shown in table 2, stratified by age, gender and IBD type. Overall, the incidence of each event increased with age but was relatively similar across gender and disease type. The HRs for the association between IBD and each thromboembolic complication are summarised in table 3, with Kaplan–Meier curves displayed in figure 1. Overall, the risk of VTE in patients with IBD was twice that of the general population (HR=2.0 (95% CI 1.8 to 2.1)). Risks were slightly higher for patients with Crohn's disease, compared to those with ulcerative colitis (HR=2.2 (95% CI 2.0 to 2.5) vs HR=1.9 (95% CI 1.8 to 2.0)). In analyses restricted to ‘unprovoked events’ (defined as events occurring without a diagnosis of fracture, trauma, surgery or pregnancy within 90 days prior to the event or a diagnosis of cancer at any time prior to or within 90 days following the event), the RR estimates for each thromboembolic event remained elevated, but somewhat attenuated (overall HR=1.6 (95% CI 1.5 to 1.8)).

Table 2

Number of cases and incidence rates (IR) per 10 000 person-years of venous thromboembolism, deep venous thrombosis, and pulmonary embolism in the Danish inflammatory bowel disease (IBD) population and a matched, non-IBD cohort

Table 3

HRs associated with venous thrombombolic events in patients with inflammatory bowel disease (IBD)

Figure 1

The risk of deep venous thrombosis (DVT) and pulmonary embolism (PE) among ulcerative colitis (UC) and Crohn's disease (CD) patients. Kaplan–Meier curves for the associations between (A) DVT and ulcerative colitis, (B) PE and ulcerative colitis, (C) DVT and Crohn's disease, and (D) PE and Crohn's disease.

When we stratified the analyses by age group, we found that the HRs for both overall and ‘unprovoked’ DVT and PE were highest in the youngest group (age 0–20 years) and decreased markedly with increasing age (table 4). Finally, in the sub-analysis using data only from the 1 January 1995 to 31 December 2007 period when both inpatient and outpatient visits were recorded in the DNPR, we found that the HRs for each thromboembolic event were essentially unchanged from those derived from the complete data set (data not shown).

Table 4

Association of inflammatory bowel disease (IBD) and venous thromboembolism (VTE) by age (cohort study)

Case–control analysis

The study population consisted of 125 963 cases of VTE (71 367 cases of DVT and 54 018 cases of PE) and their population controls. Demographic characteristics of the cases and controls are shown in table 5. As expected, based on prior literature, the number of cases for each event increased with age. Slightly more than half of cases were female. Additionally, as expected, each of the known risk factors (cancer, surgery, fracture and pregnancy) occurred at a higher frequency in cases than in controls.

Table 5

Demographic characteristics of cases of venous thromboembolism (VTE), deep vein thrombosis (DVT) and pulmonary embolism (PE) and matched population controls in the Danish Population, 1980–2007

Crude (age- and gender-matched) and adjusted ORs for the associations between IBD and VTE events are shown in table 6. Overall, in the unadjusted analysis, patients with a VTE were twice as likely as controls to have had a prior diagnosis of IBD (OR=2.0, 95% CI 1.9 to 2.2). As expected, the magnitude of this association, as well as that of the associations between IBD and ‘unprovoked’ VTE, DVT, ‘unprovoked’ DVT, PE, and ‘unprovoked’ PE were all nearly identical to those found in the cohort analysis. The similarities between the cohort and unadjusted case–control analyses persisted after stratifying by IBD type (Crohn's disease vs ulcerative colitis, data not shown). As in the cohort analysis, stratifying thrombotic risk estimates by age group yielded the highest RR in the population aged 0–20 years, with decreasing risk at older ages (data not shown).

Table 6

ORs (95% CI) for the association between inflammatory bowel disease (IBD) and venous thromboembolism in Denmark, 1980–2007

After adjusting for co-morbidities associated with venous thrombotic events (heart failure, diabetes, MI, and stroke), the association between IBD and each type of thrombotic event were somewhat attenuated, but remained significant. For the analysis of IBD and VTE, adjusted ORs for each of the provoking factors and co-morbidities were as follows: malignancy (OR 2.3, 95% CI 2.3 to 2.4), surgery (OR 7.3, 95% CI 7.2 to 7.5), fracture (OR 3.4, 95% CI 3.3 to 3.5), pregnancy (OR 1.7, 95% CI 1.5 to 1.9), congestive heart failure (OR 2.6, 95% CI 2.5 to 2.7), diabetes (OR 1.3, 95% CI 1.3 to 1.4), myocardial infarction (OR 1.2, 95% CI 1.1 to 1.3), and stroke (OR 1.5, 95% CI 1.5 to 1.6).

In the sub-analysis that included utilisation of prescription medication, adjusting for the use of hormone replacement therapy and antipsychotics reduced the effect estimates either minimally or not at all; after adjusting for co-morbidity and medication use ORs for all events remained in the 1.5–1.8 range. For the analysis of IBD and VTE, adjusted ORs for recent, remote and persistent HRT were close to unity, and 95% CIs crossed 1. Adjusted ORs for recent, remote and persistent antipsychotic use were 2.0 (95% CI 1.7 to 2.5), 1.2 (95% CI 1.1 to 1.4), and 1.6 (95% CI 1.5 to 1.8) respectively.

In the final case–control analysis, 203 patients with IBD experiencing VTE events were matched to 1917 cases of IBD without VTE events on the basis of age, gender and county of residence. Use of oral steroids 0–30 days prior to VTE (or index date for controls) was positively associated with VTE (OR 8.4, 95% CI 5.1 to 13.8), as was hospitalisation with a primary IBD diagnosis occurring 0–30 days prior to VTE (OR 22.8, 95% CI 10.9 to 47.9). The OR for occurrence of either exposure was 11.3 (95% CI 7.3 to 17.6).

Discussion

In this large, population-based nationwide study, we have confirmed an increased risk of VTE in patients with inflammatory bowel diseases. Overall, we found that patients with IBD have twice the incidence of PE or DVT as the general population. The relative risks were particularly high at young ages, though the absolute risks increased with age. Substantial increases in risk remained, even after taking into account the factors that are known to predispose to VTE and common risk factors for it.

This work expands upon prior research in several ways. First, we found that this increased risk persists even in the absence of common provoking factors such as fracture, malignancy, surgery or pregnancy. After adjusting for additional thromboembolic risk factors including diabetes, heart failure, MI, stroke, and use of hormone replacement therapy and antipsychotic medications, patients with IBD had a 50% higher risk of VTE than controls, strongly suggesting that IBD is an independent prothrombotic risk factor.

A key finding in this study was the strong effect of patient age on thromboembolic risk. In patients with IBD age 20 or younger, the annual incidence of VTE (absolute risk) was very low: 8.9 events per 10 000 person-years, approximately 85% less than that of persons 60 years of age or older. Yet, the RR of VTE in patients with IBD (vs controls) was approximately four times higher in children than in the elderly. These findings underscore the need to consider both absolute risk as well as RR in recommendations for screening and prophylaxis regimens. Of course, both the efficacy and potential for harm of VTE prophylaxis in different populations are also crucial elements of this equation.

A recent report by Grainge et al10 has provided strong evidence that IBD disease activity is another factor associated with thromboembolic risk. Hospitalised patients with IBD and those experiencing active disease as indicated by corticosteroid use had a higher risk of VTE than patients with well-controlled disease. Our study confirms the strong association between IBD disease activity and VTE, and indicates that additional factors including patient age and the presence of other known risk factors for VTE can be used to further refine patient-specific risk estimates.

The absolute risk of VTE in patients with IBD found in our study was nearly identical to that reported by Grainge et al10 (24 vs 26 per 10 000 person-years); however, the relative risks observed in our study were somewhat lower (HR=2.0 vs 3.4). This is likely due to the surprisingly low incidence of events in the general population observed in the Grainge study (6 events per 10 000 person years, versus 13.4 events per 10 000 person years in our study). Because the background incidence of VTE in our study resembles that reported for other Western populations,1 we believe that the 2-fold increase observed in RR (1.5-fold after adjustment) more closely resembles the actual risk associated with IBD.

The strengths of our study include its nationwide, population-based design which virtually eliminates selection and information bias. Additionally, the very large sample size allowed us to calculate precise estimates of both the absolute risk and RR of DVT and PE, even after stratifying by age and the presence of provoking factors, as well as adjusting for other potential confounders. We used previously validated definitions of both the exposures30 and outcomes34 of interest. The sensitivity and specificity of IBD diagnoses in the DNPR were extremely high, thus minimising the misclassification inherent to any study utilising administrative data. Diagnoses of thromboembolic events were less reliable, though any misclassification of exposures, outcomes, or other variables ascertained through administrative data would likely be non-differential and result in bias towards the null. An additional limitation is the possibility of detection bias; that is, detection of subclinical events more often in patients with IBD than controls due to more rigorous medical follow-up of patients with IBD. We believe that this bias is unlikely, as a sub-analysis of data obtained prior to 1995 (before the association between IBD and VTE was first published8) yielded similar results. Another limitation is that identification of both exposures and outcomes relied heavily on inpatient encounters, and thus milder events would not have been captured. However, a sub-analysis restricted to data collected after 1995 when outpatient data were added to the DNPR yielded nearly identical results, suggesting that significant underestimation of these events is unlikely. Finally, we were unable to adjust for potential unmeasured confounders not included in Danish healthcare registries including BMI, tobacco, oral contraceptive use, and indwelling central lines. Nevertheless, when we controlled for other known VTE risk factors, the risk estimates were only mildly attenuated. Thus, even if we could have adjusted for all possible confounders, it is unlikely that the risk estimates reported here would change substantially.

In conclusion, this study provides further robust evidence of an association between IBD and VTE. The results from this study and the prior study reported by Grainge et al provide strong evidence that the absolute risk of VTE varies markedly by patient age, IBD disease activity and, to a lesser degree, co-morbidity. Taken together, these recent studies suggest the need to modify the currently recommended ‘one size fits all’ approach to recommendations regarding VTE prophylaxis We suggest that guidelines and individual patient decisions about VTE prophylaxis in patients with IBD should take into account patient-specific estimates of absolute risk as well as both the safety and efficacy of different prophylaxis regimens. Based on the reported safety, clinical efficacy, and cost-effectiveness of VTE prophylaxis, this and other recent epidemiological studies of VTE risk in patients with IBD add further support for the recommendation to pursue prophylaxis for most hospitalised adult patients with IBD. However, in the paediatric population and, to a lesser degree, adults less than 40 years of age, the absolute risks of VTE are considerably lower than in older adults. In these younger patients, and specifically those without other VTE risk factors or limited mobility, the benefits of prophylaxis may no longer outweigh the risks. Similarly, older, non-hospitalised patients with IBD (>60 years of age) experiencing disease flares might benefit from outpatient VTE prophylaxis. In addition to the above considerations regarding VTE prophylaxis, we also urge physicians to be vigilant and have a low threshold to carefully evaluate patients with IBD who develop the signs or symptoms of possible VTE.

Appendix 1 Additional diagnosis and pharmaceutical codes used in this manuscript

ICD codes defining co-morbidities

  • Congestive heart failure: ICD-8: 427.09–427.11, 427.19, 428.99, 782.49; ICD-10: I50.

  • Diabetes: ICD-8: 249, 259; ICD-10: E10, E11.

  • Myocardial infarction: ICD-8: 419; ICD-10: I21-I23.

  • Stroke: ICD-8: 431, 432, 434, 435; ICD-10: I61, I63, I64, I65, I66 (except I63.6).

ICD codes defining provoking factors

  • Cancer: ICD-8: 140–209; ICD-10: C00–C99.

  • Trauma or fracture: ICD-8: 800–929, 950-959; ICD-10: S00–T14.

  • Pregnancy: ICD-8: 630–680; ICD-10: O00–O99.

  • Surgery: All surgery codes.

ATC codes defining prescription medications

  • Hormone replacement therapy: G03C.

  • Antipsychotics: N05A.

References

View Abstract

Footnotes

  • Funding This research was supported, in part, by grants from the National Center for Research Resources (NCRR) Grant KL2 RR025746 (MDK) and the National Institute for Diabetes and Digestive and Kidney Diseases Grant P30 DK034987 (RSS).

  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the Danish Ethical Review Board.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.