A, Incidence rate of deep venous thrombosis (DVT) and pulmonary embolism (PE) stratified by age group; B, incidence rate of venous thromboembolism by sex and age group.
Huerta C, Johansson S, Wallander M, García Rodríguez LA. Risk Factors and Short-term Mortality of Venous Thromboembolism Diagnosed in the Primary Care Setting in the United Kingdom. Arch Intern Med. 2007;167(9):935-943. doi:10.1001/archinte.167.9.935
Venous thromboembolism (VTE) manifesting as deep vein thrombosis (DVT) and pulmonary embolism (PE) remains a common vascular disease with high mortality and morbidity. Our aim was to study the clinical spectrum of VTE, assess its incidence in the general population, and evaluate potential risk factors.
Prospective cohort study with nested case-control analysis using the General Practice Research Database (1994-2000). Venous thromboembolism was newly diagnosed in 6550 patients. Cases were compared with a random sample of 10 000 controls and frequency-matched by age, sex, and year.
The incidence rate of VTE was 74.5 per 100 000 person-years. Overweight, varicose veins, inflammatory bowel disease, cancer, and oral corticosteroid use were associated with a greater risk of VTE. Ischemic heart disease, heart failure, and cerebrovascular diseases were associated with an increased risk of PE but not with DVT. Venous thromboembolism was strongly associated with fractures (odds ratio [OR], 21.3; 95% confidence interval [CI], 15.7-28.9) and surgery (OR, 25.0; 95% CI, 14.4-43.5). In women, the risk of VTE was 1.9 (95% CI, 1.5-2.3) among those receiving opposed hormone therapy (in which the woman takes estrogen throughout the month and progesterone for 10-14 days later in the month) and 1.9 (95% CI, 1.4-2.5) among those taking oral contraceptives. Cancer and cerebrovascular diseases presented a greater relative risk of fatal PE compared with nonfatal PE.
Overweight, varicose veins, cancer, inflammatory bowel disease, fractures, surgery, and use of oral corticosteroids, oral contraceptives, and opposed hormone therapy were independent risk factors for both DVT and PE. The magnitude of the association with some risk factors varied between DVT and PE, as well as between fatal and nonfatal PE.
Deep vein thrombosis (DVT) and pulmonary embolism (PE) represent different manifestations of the same clinical entity, which is referred to as venous thromboembolism (VTE). Venous thromboembolism can start with the formation of a venous clot (thrombosis), often occurring in the deep veins of the legs, thighs, or pelvis (ie, DVT) and can often be asymptomatic. Pulmonary embolism occurs if part or all of a thrombus is dislodged from a vein wall, travels to the lungs, and lodges within the pulmonary arteries, and there is a substantial chance of fatality.1
Venous thromboembolism occurs for the first time in about 100 people per 100 000 every year in the United States, with about one third of cases manifesting as PE, while two thirds of patients have DVT alone.1 In addition, approximately 6% of DVT and 12% of PE cases result in the patient's death within 1 month of diagnosis,1 and the survivors often experience serious and costly long-term complications.2 However, much remains to be learned regarding the risk of VTE to prevent its occurrence and complications and to improve patients' survival. Although risk factors have been identified3,4 and anticoagulant prophylaxis is recommended in selected groups,5 mortality and morbidity associated with VTE remains high. Malignancy, surgery, pregnancy, and heart failure are well-known risk factors for VTE, but results from epidemiological studies have produced varying results regarding the magnitude of these and other risk factors.3,4
We have conducted a large, prospective cohort study with a nested case-control analysis using data from the General Practice Research Database (GPRD) in the United Kingdom (UK). Our aim was to study the complete clinical spectrum of major VTE, measure its incidence in the general population, identify risk factors associated with the occurrence of new diagnoses of VTE, DVT, and PE, and estimate the magnitude of risk.
The GPRD contains computerized information entered by family physicians (FPs) in the UK. Most of the UK population is registered with an FP. At present, approximately 1500 FPs participated in the GPRD, covering a population of about 3 million individuals, who are broadly representative of the population of the UK as a whole. The GPRD holds the complete medical records of registered individuals, including demographic data, all medical diagnoses (OXMIS [Oxford Medical Information System] and Read clinical classification codes), consultant and hospital referrals, and a record of all prescriptions issued. Prescriptions are generated directly from the FP's computer and entered into the patient's computerized file. All the information is recorded by FPs during consultations in a standard fashion, and practices regularly send these data (made anonymous) to the Medicines and Healthcare Products Regulatory Agency, which is in charge of the quality control and data management for use in research projects. Several validation studies have shown the accuracy and completeness of data in the GPRD.6,7 Previous studies have also confirmed the validity of using the GPRD for epidemiological research in the field of DVT and PE.8
Individuals aged 20 to 79 years who had been enrolled with a participating FP for more than 2 years starting January 1, 1994, were identified. Each individual's start date was set as the first day that they met the eligibility criteria. People with a specific code of DVT or PE before the start date were excluded to remove prevalent cases. The study protocol was approved by the GPRD Scientific Ethical and Advisory Group.
All members of the study cohort were followed from the start date until the earliest occurrence of any of the following: a specific or nonspecific recorded code suggesting PE or DVT, death, or December 31, 2000. The resulting study cohort consisted of 1 814 669 patients contributing 8 799 714 person-years of follow-up.
A patient was considered as a DVT or PE case when he or she had a recorded code of DVT or PE and received anticoagulant therapy (heparin, warfarin sodium, or a similar agent), independently of whether a diagnostic procedure (ventilation-perfusion scan, pulmonary angiography or venography, or ultrasonography) was recorded in the database. Based on a previous validation study,8 evidence of anticoagulation from the database was not restricted solely to a prescription for an anticoagulant, but a wider search strategy was used such as international normalized ratio testing or attendance at anticoagulation clinics. Since VTE is a potentially fatal disease, we also included patients with a fatal outcome because a patient may have died before they could receive anticoagulation treatment. Under this assumption, patients with a recorded code of DVT or PE were included in the absence of recorded anticoagulant therapy if there was a fatal outcome within 1 month of diagnosis regardless of whether autopsy results were available.
While blinded to personal identifiers, we manually reviewed the computerized patient profiles including all comorbid conditions, use of anticoagulants, referrals to consultants, and hospitalizations. At this stage, we excluded patients with a history of DVT or PE, those with thrombophlebitis and upper extremity VTE, those with no anticoagulant prescription, and those with a computer profile clearly not confirming the initial recorded diagnosis of DVT or PE. Finally, 6550 patients fulfilled these conditions, of whom 3544 (54%) had DVT, 2588 (39%) had PE, and 418 (6%) had evidence of both at the time of the diagnosis. The date of DVT or PE diagnosis was assigned as the index date.
Even though in a recent validation study of DVT using the GPRD, 84% of events were confirmed by the FP and documented by hospital records,7 we chose to send a questionnaire to the FPs of a random sample of 5% of patients (n = 304) with DVT or PE to gauge the validity of our case ascertainment strategy for DVT and PE. Patient confidentiality was always maintained. We requested the following information from the FPs: (1) confirmation of the diagnosis of DVT or PE; (2) confirmation that this was the first episode of DVT or PE; (3) confirmation of anticoagulant therapy, and (4) information on performed diagnostic procedures. The response rate was 99% (300 patients). After reviewing the questionnaires received with complete information (n = 283), we confirmed the diagnosis of DVT or PE in 266 patients (94%).
To select controls, a random date during the study period was assigned to all members of the study cohort from which cases of DVT or PE were ascertained. If the random date was included in a patient’s eligible person-time, we used this random date as the index date and marked that patient as an eligible control. This mechanism, known as incidence density sampling, means that the likelihood of being selected as a control is proportional to the person-time at risk. Under this design, the odds ratio (OR) is an unbiased estimator of the incidence rate ratio (RR). Of the pool of eligible controls, 10 000 individuals were randomly sampled and frequency matched to DVT and PE cases by age (within 1 year), sex, and calendar year.
Information on comorbid conditions and drug prescriptions was obtained from the database. We ascertained information prior to the index date on cardiovascular disease (heart failure, ischemic heart disease, myocardial infarction, cerebrovascular disease, hypertension, and varices), asthma, chronic obstructive pulmonary disease, gastrointestinal diseases (inflammatory bowel disease [IBD] and chronic liver disease), cancer, renal disease, diabetes, anemia, osteoarthritis, rheumatoid arthritis, pregnancy, and surgery. Alcohol intake and smoking status was used as directly registered by the FP. We analyzed prescription drug use before the index date including cardiovascular drugs, aspirin, nonsteroidal anti-inflammatory drugs (NSAIDs), acetaminophen, disease-modifying antirheumatic drugs, oral corticosteroids, and gastrointestinal drugs. We defined 3 time windows of exposure for each class of drugs as follows: current use when the supply of the most recent prescription lasted until the index date or ended in the 30 days before the index date, past use when it ended more than a month before the index date, and nonuse when there was no recorded use in the database. Duration of use (time since beginning of therapy, defined by consecutive supplies [refill gap <3 months]) was calculated among current users. We ascertained the use of hormone therapy (HT) in women 50 years and older and prescription of oral contraceptives in women younger than 50 years. We also analyzed outcomes among new users of HT. A new user was defined as a woman with more than 6 months between the most recent HT prescription before the start date and the most recent HT after the start date (with the start date being the date when an individual enters into the study follow-up).9 Healthcare utilization variables such as the number of FP visits, referrals to specialists, and admissions to hospital in the year prior to index date were also evaluated.
We estimated incidence rates and 95% confidence intervals (CIs) of DVT alone and PE with or without DVT (as PE is a complication of DVT), dividing the number of incident cases by the total follow-up period in our study cohort. We also calculated the age- and sex-specific incidence rate of DVT and PE. Adjusted ORs were calculated for DVT and PE for several risk factors by unconditional logistic regression, using stepwise techniques to identify the most parsimonious model. Matching factors were always included in the model.
Of 6550 patients newly diagnosed as having VTE, 53% were female and the median age was 63 years. The overall incidence rate of VTE in the cohort was 74.5 per 100 000 person-years. The incidence of DVT alone and the incidence of PE (with or without DVT) was 40.3 and 34.2 per 100 000 person-years, respectively. Panel A of the Figure shows the incidence rate of DVT and PE stratified by age. The incidence increased markedly with age. The incidence of VTE according to sex and age are shown in panel B of the Figure.
Of the VTE cases, 29% presented with 1 of the following 3 risk factors: surgery, fracture in the month before the diagnosis, or cancer. The distribution of risk factors and comorbidity among overall VTE, DVT, and PE cases compared with controls is given in Table 1 and Table 2.
We found that patients with a fracture had a greatly increased risk of developing VTE in the following month (OR, 25.05; 95% CI, 14.41-43.53). The highest OR was for hip fractures (69.38; 95% CI, 9.34-515.28). Patients with fractures of the lower extremities, excluding the hip, also had a marked increased risk of VTE (OR, 33.16; 95% CI, 15.39-71.67), and they were more strongly associated with occurrence of PE (OR, 45.02; 95% CI, 20.46-99.06) than DVT (OR, 21.52; 95% CI, 9.54-48.57). The risk, though based on few cases, was greatest during the first week for both hip and lower extremities (data not shown).
Surgery was also a strong predictor of developing VTE (Table 2). The likelihood of developing VTE following surgery was particularly high during the first month (OR, 34.64; 95% CI, 25.14-47.73) and the first 6 months (OR, 9.39; 95% CI, 8.02-10.99) after surgery. The 3 surgical procedures associated with the highest risk of VTE (within 6 months) were musculoskeletal surgery (OR, 21.31; 95% CI, 15.70-28.91), neurosurgery (OR, 9.79; 95% CI, 3.38-28.36), and vascular surgery (OR, 9.34; 95% CI, 5.32-16.38). The risk of VTE returned to the background rate 6 months after surgery (OR, 1.10; 95% CI, 1.02-1.19), with a small increased risk remaining in patients 6 months after vascular surgery (OR, 1.48; 95% CI, 1.23-1.79).
As well as fractures and surgery, the risk of VTE was significantly increased in patients with recent health care contact, those who were overweight, those with comorbid disorders (eg, varicose veins, IBD, and cancer), and in woman who were pregnant in the year prior to the index date (Table 1).
Interestingly, we found a number of risk factors to be clearly associated with PE but to a lesser extent with DVT. Heart failure was associated with an increased risk of PE (OR, 3.03; 95% CI, 2.50-3.68), but with only a minor increase in the risk of DVT (OR, 1.35; 95% CI, 1.07-1.69). Cerebrovascular disease presented an OR for PE of 1.54 (95% CI, 1.28-1.86), while corresponding results for DVT were 1.10 (95% CI, 0.91-1.33). Myocardial infarction gave an increased risk of PE (OR, 1.88; 95% CI, 1.55-2.28) but not DVT (OR, 1.10; 95% CI, 0.89-1.36). Estimates for atrial fibrillation were 1.94 (95% CI, 1.47-2.57) and 1.46 (95% CI, 1.08-1.96) for PE and DVT, respectively.
There was little association between asthma and either DVT or PE, while chronic obstructive pulmonary disease was clearly associated with an increased risk of PE and to a lesser extent with DVT. We found that diabetes was associated with a slightly reduced risk of VTE compared with controls. The reduced risk was similar for both DVT and PE and was present irrespective of treatment. Smoking, alcohol, hypertension, hyperlipidemia, osteoarthritis, rheumatoid arthritis, chronic liver disease, and coagulation disorders were not associated with VTE.
Associations with specific drug treatment are given in Table 3. Oral corticosteroids and NSAIDS were associated with an increased risk of VTE. When we looked at this indication among users of NSAIDs with treatment duration greater than a month, we saw no increased risk among patients taking NSAIDs for osteoarthritis, the main indication (data not shown). Unlike NSAIDs, long-term prescription of oral corticosteroids was associated with an increased risk of VTE. Statins were associated with an OR of DVT of 0.70 (95% CI, 0.50-0.97).
Women receiving HT were at a slightly greater risk of developing VTE (OR, 1.32; 95% CI, 1.09-1.59). However, as shown in Table 4, the risk of VTE was observed among women taking opposed HT (OR, 1.85; 95% CI, 1.46-2.34) but not among women taking unopposed therapy (OR, 1.03; 95% CI, 0.80-1.33). (In opposed HT, the woman takes estrogen throughout the month and progesterone for 10-14 days later in the month, whereas in unopposed HT, estrogen alone is taken.) When we studied the effect of HT duration, the risk was slightly greater at the beginning of therapy. Results were similar when we studied DVT and PE separately. It is noteworthy that when we restricted the analysis to new users of opposed HT, the OR increased to 2.38 (95% CI, 1.68-3.38). Women taking oral contraceptives were at double the risk of developing VTE (OR, 1.85; 95% CI, 1.38-2.49). The risk increased slightly with treatment duration and was more marked for DVT than for PE.
Of the 730 fatal VTE cases, death occurred in 667 at the index date (97% of these were PE cases) and the remainder (63 cases) had a fatal outcome within the first month after the occurrence of VTE (52% of these were PE cases). More than 50% of fatal cases occurred in those at least 70 years old. The 1-month fatality rate was 1.4% after an episode of DVT and 22.6% after PE (with or without DVT). When we restricted the analysis to fatal and nonfatal PE cases (Table 5), we observed that cancer particularly increased the risk of fatal PE (OR, 6.28; 95% CI, 5.06-7.80). The same pattern was observed with heart failure, cerebrovascular disease, and atrial fibrillation. Surgery, however, was more strongly associated with nonfatal PE (OR, 12.70; 95% CI, 10.59-15.24) than with fatal PE (OR, 2.48; 95% CI, 1.76-3.48). The ORs for oral corticosteroid use were 4.27 (95% CI, 3.14-5.80) for fatal PE cases and 3.69 (95% CI, 2.91-4.68) for nonfatal cases.
Our results agree broadly with previous epidemiology studies. Venous thromboembolism is still a common disease, occurring predominantly in elderly individuals,4,10 and the present study shows that among individuals aged 20 to 39 years, there is a trend toward a greater incidence of VTE in women than in men, with this trend reversing in those older than 60 years.10 The incidence of symptomatic VTE in other epidemiological studies has ranged from 71 to 117 cases per 100 000 of the population (standardized for age and sex).1 Several clinical studies including DVT and PE have reported the incidence of a first DVT to be approximately twice that of PE.11,12 However, in studies including many cases diagnosed by autopsy, the incidence of PE is often higher than DVT.10,13 Autopsy data may overestimate the incidence of PE through the detection of asymptomatic cases, whereas studies reliant on clinical diagnoses probably underestimate the incidence of PE.1 We found the incidence of a first DVT to be only slightly greater than the incidence of a first PE. This could be owing to the inclusion of both fatal and nonfatal cases in our study, with most of the fatal cases involving PE.
Our findings that surgery, fractures (in particular, hip fractures), and cancer were major risk factors for VTE are fully consistent with the literature.4,14,15 It is notable that fractures of the lower extremities other than hip fractures also carried an increased risk of VTE. We found a 10-fold increased risk of developing an episode of DVT within 6 months after surgery, with musculoskeletal surgery carrying the greatest risk—close to 37% of VTE occurring in patients
Among cardiovascular diseases, the magnitude of risk of VTE associated with heart failure varies between studies, from an independent risk factor for DVT15 to only being a risk factor for autopsy-discovered VTE that was not the cause of death.4 In our study, heart failure was a strong risk factor for PE, while the impact on DVT was much lower. The increased risk associated with heart failure was observed both for fatal and nonfatal cases of PE. To our knowledge, myocardial infarction has not been clearly established as an independent risk factor of VTE. In our study, we found that myocardial infarction carried a small increased risk of PE but with no effect on the risk of DVT. The risk of VTE reported for varicose veins has been found to either vary with age4 or be nonexistant.15,16 In our study, varicose veins were a risk factor for both DVT and PE.
Additional factors that increased the risk of VTE in our study were obesity and IBD. In most studies, obesity has been found to be a risk factor,16- 19 although one study found no association.4 The association with IBD has not been so well studied. One study did not identify it as a risk factor,14 while 2 studies found an association between IBD and VTE.20,21 Based on the pathological mechanism of DVT, this association is plausible, and it has been suggested that IBD could contribute to producing thrombosis-inducing hypercoagulability.22 Our data did not demonstrate a reduced risk of VTE in patients with liver diseases as has been found in other studies,4 although this may be because of different definitions of liver disorders.
We found an increased risk of VTE in users of NSAIDs and oral corticosteroids. Use of NSAIDs was associated with an elevated risk of VTE, especially when starting therapy, and the risk tended to disappear with extended treatment. However, the increased risk with oral corticosteroids, although smaller with extended treatment duration, was still present. We tried to exclude the possibility that the association with NSAIDs was primarily due to confounding by indication, a potential explanation for effects observed with drug use. Patients starting NSAID therapy (first month of use) were in many instances taking NSAIDs for prodromal symptoms related to the episode of VTE (sometimes referred to as protopathic bias). However, when we looked at the effect of indications in patients already taking NSAIDs for at least 1 month or more, they did not present an increased risk of VTE when the indication was osteoarthritis. Our data, based on a small number of exposed individuals, suggest a reduced risk of VTE in users of statins, as reported in a previous study.23 The influence of hyperlipidemia and/or lipid-lowering therapy on DVT remains controversial, but hyperlipidemia has been suggested as a risk factor for DVT,24 and statins can have antithrombotic effects.
Evidence from epidemiological studies consistently shows an increased risk of DVT in women receiving HT and taking oral contraceptives and that this risk is more pronounced at the beginning of therapy with both types of hormonal preparations.25 Our results also showed an increased risk of DVT and PE in women taking opposed HT, but we did not find such an increased risk with unopposed HT. We also confirmed the association between VTE and oral contraceptives. Unlike previous studies, we did not find the risk associated with unopposed HT or oral contraceptives to vary noticeably with the duration of treatment. New users of HT, the designation of whom eliminates the possible bias of “healthy” women receiving HT for some time before the study follow-up began, carried a slightly greater risk than that estimated for all durations of HT.9 We also found an increased risk of VTE in women who were pregnant in the year prior to diagnosis, which is consistent with the literature.4
One strength of the present study is its large size, involving a cohort of 1 856 206 patients, thus allowing a relatively short study period (maximum of 7 years). As such, these results are more contemporary than similar but smaller studies conducted over longer periods4,10 and thus should more closely reflect current populations in terms of age demographics and the prevalence of certain conditions (eg, type 2 diabetes mellitus and obesity) that have become more common. In addition, a validation study allowed us to confirm the diagnosis of VTE in 94% of the patients. This rules out a major misclassification bias in our study. Moreover, a study specifically designed to validate the diagnosis in the GPRD showed that this database provides information of sufficiently high quality to allow valid epidemiological research of VTE events.8 Since the confirmation rate was extremely high, we did not calculate the incidence of VTE by weighting the number of newly diagnosed VTE cases identified by the confirmation rate obtained in the validation study.
In summary, we found that VTE is still a common condition with a substantial mortality among patients with PE. We confirmed known risk factors and presented estimates of the magnitude of risk for other more debatable risk factors such as heart failure or varicose veins. Our study provides important data on the risk following fracture in different locations, such as those of the lower extremities. Our study also presents risk estimates for DVT and PE separately, and we found that the magnitude of risk often varies between DVT and PE. Although the use of anticoagulant therapy is recommended in patients at high risk of VTE,5 the remaining high mortality and morbidity means that there may be other new risk factors. We have identified potential new risk factors for further investigation. Population-based studies with incident cases are needed to further investigate these new potential risk factors and to identify whether they could be candidates for prophylaxis of VTE.
Correspondence: Consuelo Huerta, MD, MPH, Spanish Centre for Pharmacoepidemiologic Research (CEIFE), Almirante 28 (2°), Madrid 28004, Spain (firstname.lastname@example.org).
Accepted for Publication: January 5, 2007
Author Contributions: Dr Huerta had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Huerta, Johansson, Wallander, and García Rodríguez. Acquisition of data: Huerta and García Rodríguez. Analysis and interpretation of data: Huerta and García Rodríguez. Drafting of the manuscript: Huerta, and García Rodríguez. Critical revision of the manuscript for important intellectual content: Huerta, Johansson, Wallander, and García Rodríguez. Statistical analysis: Huerta, and García Rodríguez. Obtained funding: Johansson, Wallander, and García Rodríguez.
Financial Disclosure: Drs Johansson and Wallander are employees at AstraZeneca, R&D.
Funding/Support: This study was funded by an unrestricted grant from AstraZeneca.
Acknowledgment: We thank the family physicians for their excellent collaboration. We also thank Christopher Winchester, DPhil, from Oxford PharmaGenesis Ltd, who provided editorial assistance.