SEARCH

SEARCH BY CITATION

Keywords:

  • epidemiology;
  • risk;
  • subarachnoid hemorrhage;
  • venous thromboembolism

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Disclosure of Conflict of Interests
  9. References

Summary.  Background: Venous thromboembolism is a predictor of subsequent risk of ischemic stroke and intracerebral hemorrhage, but no data are available regarding its association with risk of subarachnoid hemorrhage. Objectives: To examine this issue, we conducted a nationwide cohort study in Denmark. Patients and methods: Between 1977 and 2007, we identified 97 558 patients with a hospital diagnosis of venous thromboembolism and obtained information on risk of subsequent subarachnoid hemorrhage during follow-up in the Danish Registry of Patients. The incidence of subarachnoid hemorrhage in the venous thromboembolism cohort was compared with that of 453 406 population control cohort members. Results: For patients with pulmonary embolism (PE), there was clearly an increased risk of subarachnoid hemorrhage, both during the first year of follow-up [relative risk 2.69; 95% confidence interval (CI), 1.32–5.48] and during later follow-up of 2–20 years (relative risk 1.40; 95% CI, 1.05–1.87). For patients with deep venous thrombosis (DVT) the risk was likewise clearly increased during the first year of follow-up (relative risk 1.91; 95% CI, 1.13–3.22), but not during later follow-up (relative risk 1.04; 95% CI, 0.81–1.32). Conclusions: We found evidence that PE is associated with an increased long-term risk of subarachnoid hemorrhage. The two diseases might share etiologic pathways affecting the vessel wall or share unknown risk factors.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Disclosure of Conflict of Interests
  9. References

Venous thrombosis is a common and serious disorder in Western countries, with an estimated overall incidence of one per 1000 persons per year, increasing with age [1]. Over the past 5 years, increasing evidence has been found of an association between venous thromboembolism and atherosclerotic diseases [2–6]. Thus, previous studies have shown a clearly increased risk of myocardial infarction, ischemic stroke, atherosclerosis and cardiovascular events in patients with venous thromboembolism [6].

Subarachnoid hemorrhage is relatively uncommon, comprising 1–7% of all strokes, and is mainly caused by rupture of an intracranial aneurysm [7]. In contrast to other types of stroke, relatively little is known regarding its epidemiology. It is not known if venous thromboembolism is also a risk factor for this type of stroke as it is for other types [8], although it is a well-known complication of subarachnoid hemorrhage [9]. The two disorders have somewhat different risk factor profiles. Smoking and hypertension are well-established risk factors for subarachnoid hemorrhage, but obesity, diabetes and factor (F)V Leiden may be inversely associated with risk [10]. Hypertension may be a risk factor for venous thromboembolism, but for smoking the data are inconclusive [11–15]. However, a systematic review has reported a positive association between atherosclerosis and the occurrence of an aneurysm [16]. Both venous thromboembolism and subarachnoid hemorrhage are characterized by a female preponderance, a contrast to other types of stroke [8,11,12]. Data regarding the relationship between venous thromboembolism and subarachnoid hemorrhage are important, as they could foster the understanding of both disorders. We therefore undertook a large population-based assessment of the risk of hospitalization as a result of subarachnoid hemorrhage after a diagnosis of venous thromboembolism, using data from Danish medical databases.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Disclosure of Conflict of Interests
  9. References

With approval of the Danish Registry Board we obtained data from the National Registry of Patients which, since 1977, has recorded 99.4% of all discharges from Danish acute care and non-psychiatric hospitals [17]. The recorded information includes: dates of hospital admission and discharge, surgical procedures performed, up to 20 discharge diagnoses classified according to the International Classification of Diseases (ICD), 8th revision until the 31 December, 1993 and the 10th revision thereafter. In all Danish medical registries, patients are identified through the Civil Registration Number [18,19]. These are unique identifiers, assigned at birth, and stored in the Danish Civil Registration System along with the date of birth, residency status and date of immigration/emigration and death (if any). The entire Danish population is provided tax-supported health care by the National Health Service, allowing free access to health care. To form a cohort of individuals with venous thromboembolism, we identified the first recorded in-patient hospital discharge diagnosis of lower limb deep venous thrombosis (DVT) or pulmonary embolism (PE), or both, between 1 January 1980 and 31 December 2007 in all Danish residents.

We defined ‘provoked’ venous thromboembolism cases as those with a diagnosed malignancy before and within 90 days after a venous thrombotic event in the Hospital Registry, and those with a discharge diagnosis of fracture, surgery, trauma and pregnancy within 90 days before hospitalization for venous thromboembolism. The remaining cases of venous thromboembolism were classified as unprovoked [6,11].

We formed a population-based control cohort using the Danish Civil Registration System [6,18,19]. For each patient in the venous thromboembolism cohort, approximately five population controls were randomly chosen for the entire registry, matched for gender and age. Each control was required to be alive on the date of hospitalization of the corresponding case, the index date for the matched set. We excluded all cohort members with a hospital diagnosis of cerebrovascular disease before the venous thromboembolism event (in cases) or the index date (among population controls).

We also collected available data on potential confounding factors from the hospital registry: hypertension, alcoholism, acute myocardial infarction, renal failure, diabetes, chronic pulmonary obstructive lung disease, obesity, atrial fibrillation and valvular heart disease [8,11–14]. The last conditions are the main indications for anticoagulation therapy in Denmark, in addition to venous thromboembolism.

Using the Civil Registration Number, all members of the study cohorts were linked to the Civil Registration System and to the National Registry of Patients so as to identify all in-patient hospitalizations after the index date for subarachnoid hemorrhage.

Statistical analysis

We assessed the association between venous thromboembolism and later subarachnoid hemorrhage both overall and separately for unprovoked thrombotic episodes. We followed the cohorts from the index date to the occurrence of a hospitalization for subarachnoid hemorrhage, or to immigration, death, end of follow-up on 31 December 2007 or 20 years of follow-up, whichever came first. We used Kaplan–Meier life-table techniques to compute cumulative risks of the outcomes and proportional hazard progression to estimate hazards ratios and 95% confidence intervals (CIs) as measures of relative risk. In all the models, we adjusted for age, gender, index calendar year and for hypertension, alcoholism, acute myocardial infarction, renal failure, chronic obstructive pulmonary disease, diabetes and obesity (all potential risk factors for subarachnoid hemorrhage) and atrial fibrillation and valvular heart disease [8]. We used Wald statistics to compute P-values for the difference in risk of subarachnoid hemorrhage between patients with DVT and PE. We used a χ2-test to compute P-values for differences in proportions. We also redid the analysis using only subarachnoid hemorrhage patients admitted to the departments of neurology and neurosurgery because of the particularly high validity of the diagnoses at these departments [20].

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Disclosure of Conflict of Interests
  9. References

We identified 97 558 individuals with a hospital diagnosis of venous thromboembolism and 453 406 population controls. 54 134 (55.5%) patients had a DVT diagnosis and 43 424 (44.5%) a PE diagnosis with or without venous thrombosis. The majority of venous thromboembolism episodes were characterized as unprovoked (n = 57 184, 58.6%).

Table 1 shows characteristics of the 54 134 patients with DVT and 43 424 patients with PE and their population controls. In both cohorts there were slightly more women than men. Between a third and a half of the two venous thromboembolism cohorts were older than 70 years (Table 1). As expected, in comparison to population controls, more venous thromboembolism patients had cancer, recent surgery and trauma/fractures. Likewise, venous thromboembolism cohort members had a higher prevalence of a history of cardiovascular diseases, chronic pulmonary diseases, renal failure and alcoholism.

Table 1.   Descriptive data for patients with venous thromboembolism and their population controls
 Venous thrombo-embolism cohort (n = 97 558)Venous thrombo-embolism control cohort (n = 453 406)Pulmonary embolism cohort(n = 43 424)Pulmonary embolism control cohort (n = 201 261)Deep venous thrombosis cohort (n = 54 134)Deep venous thrombosis control cohort (n = 252 145)
  1. The values given in parenthesis are expressed in percentage.

Age group
 –5526 270 (26.9)%126 279 (27.9)%8907 (20.5)%42 950 (21.3)%17 363 (32.1)%83 329 (33.1)%
 56–7028 936 (29.7)%137 909 (30.4)%12 669 (29.2)%60 434 (30.0)%16 267 (30.0)%77 475 (30.7)%
 71–42 352 (43.4)%189 218 (41.7)%21 848 (50.3)%97 877 (48.6)%20 504 (37.9)%91 341 (36.2)%
Female51 564 (52.8)%240 098 (52.9)%23 350 (53.8)%108 376 (53.8)%28 214 (52.1)%131 722 (52.2)%
Male45 994 (47.2)%213 308 (47.1)%20 074 (46.2)%92 885 (46.2)%25 920 (47.9)%120 423 (47.8)%
Provoked venous thromboembolism
 Surgery27 677 (28.4)15 210 (3.4)13 869 (31.9)6756 (3.4)13 808 (25.5)8454 (3.4)
 Cancer17 219 (17.7)27 639 (6.1)8315 (19.2)12 987 (6.5)8904 (16.5)14 652 (5.8)
 Pregnancy902 (0.9)810 (0.2)261 (0.6)233 (0.1)641 (1.2)577 (0.2)
 Trauma-fracture8607 (8.8)5766 (1.3)3950 (9.1)2351 (1.2)4657 (8.6)3415 (1.4)
Hypertension9769 (10.0)24 482 (5.4)4521 (10.4)11 437 (5.7)5248 (9.7)13 045 (5.2)
Alcoholism3442 (3.5)5759 (1.3)1144 (2.6)2167 (1.1)2298 (4.3)3592 (1.4)
Myocardial infarction8125 (8.3)14 428 (3.2)5474 (12.6)6808 (3.4)2651 (4.9)7620 (3.0)
Renal failure2038 (2.1)2548 (0.6)864 (2.0)1021 (0.5)1174 (2.2)1527 (0.6)
Chronic obstructive pulmonary disease8937 (9.2)13 080 (2.9)5214 (12.0)6109 (3.0)3723 (6.9)6971 (2.8)
Diabetes6500 (6.7)12 071 (2.7)3099 (7.1)5535 (2.8)3401 (6.3)6536 (2.6)
Obesity4300 (4.4)5418 (1.2)1799 (4.1)2397 (1.2)2501 (4.6)3021 (1.2)
Atrial fibrillation6550 (6.7)10 364 (2.3)3940 (9.1)4899 (2.4)2610 (4.8)5465 (2.2)
Valvular heart disease1375 (1.4)2905 (0.6)876 (2.0)1317 (0.7)499 (0.9)1588 (0.6)

Overall, venous thromboembolism was a marker of subsequent risk of subarachnoid hemorrhage (Figs 1 and 2). During the first year after DVT, 0.04% of patients had subarachnoid hemorrhage vs. 0.02% in the control cohort (relative risk 1.91; 95% CI, 1.13–3.22). The corresponding risks for PE were 0.03% and 0.01% (relative risk 2.69; 95% CI, 1.32–5.48) (P-value for difference = 0.452) (Table 2). Restricting the analysis to cohort members with unprovoked venous thromboembolism resulted in almost identical relative risks (Table 3).

image

Figure 1.  Relative risk of subarachnoid hemorrhage (SAH) during follow-up in patients with deep venous thrombosis.

Download figure to PowerPoint

image

Figure 2.  Relative risk of subarachnoid hemorrhage (SAH) during follow-up in patients with pulmonary embolism.

Download figure to PowerPoint

Table 2.   Risk (in brackets) and relative risk of subarachnoid hemorrhage in patients with venous thromboembolism compared with population controls
 Venous thrombo-embolism cohort (n = 97 558)Venous thrombo-embolism control cohort (n = 453 406)Adjusted relative risk* (95% CI)Pulmonary embolism cohort (n = 43 424)Pulmonary embolism control cohort (n = 201 261)Adjusted relative risk* (95% CI)Deep venous thrombosis cohort (n = 54 134)Deep venous thrombosis control cohort (n = 252 145)Adjusted relative risk* (95% CI)
  1. *Adjusted for age group, gender, index calendar year, hypertension, alcoholism, myocardial infarction, renal failure, chronic obstructive pulmonary disease, diabetes, obesity, atrial fibrillation and valvular heart disease.

Overall165 (0.17%)847 (0.19%)1.28 (1.08–1.51)68 (0.16%)375 (0.19%)1.51 (1.16–1.98)97 (0.18%)472 (0.19%)1.14 (0.91–1.42)
1-year follow-up31 (0.03%)75 (0.02%)2.20 (1.45–3.36)11 (0.03%)26 (0.01%)2.69 (1.32–5.48)20 (0.04%)49 (0.02%)1.91 (1.13–3.22)
2-20 year follow-up134 (0.20%)772 (0.18%)1.17 (0.97–1.41)57 (0.25%)349 (0.19%)1.40 (1.05–1.87)77 (0.17%)423 (0.18%)1.04 (0.81–1.32)
Table 3.   Risk (in brackets) and relative risk of subarachnoid hemorrhage in patients with unprovoked venous thromboembolism compared with population controls
 Venous thrombo-embolism cohort (n = 57 184)Venous thrombo-embolism control cohort (n = 241 708)Adjusted relative risk (95% CI)Pulmonary embolism cohort (n = 24 030)Pulmonary embolism control cohort (n = 100 988)Adjusted relative risk (95% CI)Deep venous thrombosis cohort (n = 33 154)Deep venous thrombosis control cohort (n = 140 720)Adjusted relative risk (95% CI)
  1. *Adjusted for age group, gender, index calendar year, hypertension, alcoholism, myocardial infarction, renal failure, chronic obstructive pulmonary disease, diabetes, obesity, atrial fibrillation and valvular heart disease.

Overall116 (0.20%)474 (0.20%)1.39 (1.13–1.71)45 (0.19%)192 (0.19%)1.69 (1.20–2.37)71 (0.21%)282 (0.20%)1.22 (0.94-1.60)
1-year follow-up21 (0.04%)44 (0.02%)2.24 (1.33–3.78)7 (0.03%)13 (0.01%)2.97 (1.17–7.53)14 (0.04%)31 (0.02%)1.90 (1.01–3.57)
2–20 year follow-up95 (0.22%)430 (0.19%)1.29 (1.03–1.61)38 (0.27%)179 (0.19%)1.58 (1.10–2.26)57 (0.20%)251 (0.19%)1.13 (0.84–1.51)

Tables 2 and 3 also summarize the relative risk estimates for 2–20 years of follow-up. Elevated risk was clearly extended for the PE cohort (relative risk 1.40; 95% CI, 1.05–1.87) in contrast to the venous thrombosis cohort (relative risk 1.04; 95% CI, 0.81–1.32) (P-value for difference = 0.115). Restriction to unprovoked venous thromboembolism patients or to subarachnoid hemorrhage cases admitted to departments of neurology and neurosurgery did not change the risk estimates substantially. In general, the risk estimates were slightly higher for men than for women. The absolute risk of subarachnoid hemorrhage was in general low in all groups.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Disclosure of Conflict of Interests
  9. References

Our large nationwide population-based study provides evidence that patients with PE have an increased risk of subsequent subarachnoid hemorrhage, compared with population controls. The risk was also increased during the first year of follow-up for venous thrombosis patients. Our relative risk estimates describing the association between PE and subarachnoid haemorrhage are slightly higher than those reported for PE and risk of acute myocardial infarction and stroke [6]. However, the absolute risk is low, < 0.25% during long-term follow-up.

Our population-based data are thus largely consistent with our hypothesis that there is a link between PE and subararchnoid hemorrhage. There are several potential explanations why venous thromboembolism might be associated with subarachnoid hemorrhage, though the mechanisms underlying the association are not clear. In contrast to venous thromboembolism little epidemiology exists on subarachnoid hemorrhage [8]. It is not plausible that venous thromboembolism in itself causes subarachnoid hemorrhage. Rather, the association we found must be as a result of shared risk factors or etiological pathways, or both. However, there is only weak and inconsistent evidence that venous thromboembolism and subarachnoid hemorrhage share common risk factors [8,12]. As we found the strongest associations for males, it is unlikely that use of external hormones explains the association and there is inconsistent evidence of external hormones being a risk factor for subarachnoid haemorrhage. Cigarette smoking is a risk factor for subarachnoid hemorrhage [8], while the data are not entirely consistent for venous thromboembolism [11,21].

We did not have data on use of anticoagulation therapy, only information regarding the indications for anticoagulation. Major bleeding occurs in 2% of the patients during the first 3 months of oral anticoagulation treatment with vitamin K-antagonist and 1% per year thereafter [22]. Thus an association with anticoagulation therapy could explain the elevated risk of subarachnoid hemorrhage over the short term. However, the increased risk many years after the embolic episode is very unlikely to be explained by anticoagulation therapy, which is typically given for 3–12 months after venous thrombosis or PE [23]. Moreover, there is actually no evidence that anticoagulation therapy with vitamin K-antagonists is a risk factor for subarachnoid hemorrhage, at least as it is diagnosed at specialized departments [24]. Moreover, we controlled for the occurrence of atrial fibrillation and valvular heart disease which are the main indications for anticoagulation therapy in addition to venous thromboembolism.

Shared genetic factors could underlie the associations between venous thromboembolism and subarachnoid hemorrhage. There is a familial tendency for both disorders, and for venous thromboembolism several genetic risk factors have been identified [25,26]. Most of these are moderately strong, with reported relative risks of two to five. In contrast, very little is known about a potential genetic component behind the familial clustering of subarachnoid hemorrhage, but the magnitude of the relative risk estimates is around two to four [20,27,28]. Whether the two diseases occur in the same families remains to be investigated.

The mechanisms behind our findings are not clear. The vascular endothelium and inflammation are believed to play an important role in the development of both venous thromboembolism and subarachnoid bleeding. There is increasing evidence from both a randomized trial and observational studies that use of statins is inversely associated with the risk of venous thromboembolism [29–31]. A recent study has also shown a protective association of current use of statins on the risk of subarachnoid bleeding and that withdrawal increased the risk 2-fold [32]. As there is weak evidence of a link between hypercholesterolemia and subarachnoid bleeding, the mechanisms that might underlie a preventive effect of statins on both venous thromboembolism and subarachnoid bleeding are not clear. However, there are suggestions that these drugs have anti-inflammatory effects [30–36] that could affect both disorders.

Hemodynamic stress and poor vascular structure have been suggested to be the main mechanisms of pathogenesis of subarachnoid hemorrhage [25]. However, the reasons for a difference in subarachnoid hemorrhage risk between PE and DVT are not clear. It is unlikely that the well-known complications of PE – chronic pulmonary hypertension [14], venous thrombosis and post-thrombotic syndrome [13] – could lead to differences in risk between the two types of venous thromboembolism.

Our study has both strengths and limitations. Our risk estimates are obtained from a population-based cohort study [6,19] in a setting with a national health service with free access to health care that largely removes referral and diagnostic bias. The population we studied was well defined and the follow-up was complete because our design relied on computerized registries with complete nation wide coverage. We had access to the entire hospital history and, since 1994, out-patient clinic data as well.

The validity of our findings depends ultimately on the accurate coding of venous thromboembolism and of subarachnoid hemorrhage. In administrative databases, the predictive value of the discharge in-hospital diagnoses of PE, venous thromboembolism and subarachnoid hemorrhage have been reported to be 75–90% [37–39]. In the most recent decade, the validity of the subarachnoid bleeding diagnosis is expected to be higher as more than 90% of all patients with stroke admitted to Danish hospitals receive at least one CT/MR scan in connection with their hospitalization. Moreover, the intracerebral hemorrhage diagnosis in the registry has a high predictive value when the diagnosis is made in hospitals with access to CT and MI scans [40,41]. However, lack of specificity of the outcome diagnosis would bias our risk estimates towards the null if the misclassification was similar in patients with, and without, venous thromboembolism. The cancer and procedure data we used to define provoked venous thromboembolism are of high validity and make the specificity of this classification high [42]. As a result of the low incidence of subarachnoid hemorrhage, it is difficult to study the association in studies based on primary data collection. Therefore, we have relied on a population-based longitudinal registry. This use of routine data might actually be a strength, as the study in itself could not have affected the diagnostic process [42].

In conclusion, we found evidence that PE is associated with an increased long-term risk of subarachnoid hemorrhage. Common undetected risk factors or pathways affecting the vessel wall are most likely responsible for the association. Further studies are needed to clarify the association and evaluate its implications for clinical practice.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Disclosure of Conflict of Interests
  9. References

The study obtained support from the Danish Medical Research Council grant 271-05-0511 and Department of Clinical Epidemiology’s Research Foundation.

Disclosure of Conflict of Interests

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Disclosure of Conflict of Interests
  9. References

The authors state that they have no conflict of interest.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Disclosure of Conflict of Interests
  9. References