Rate of deep vein thrombosis (DVT) among men and women according to agent.
Ray JG, Mamdani M, Tsuyuki RT, Anderson DR, Yeo EL, Laupacis A. Use of Statins and the Subsequent Development of Deep Vein Thrombosis. Arch Intern Med. 2001;161(11):1405-1410. doi:10.1001/archinte.161.11.1405
Some of the benefit of statins for the prevention of cardiovascular disease may be due to their antithrombotic properties. Little is known about the effect of these drugs on the development of deep vein thrombosis.
Materials and Methods
We conducted a retrospective cohort study over an 8-year period by linking Ontario provincial health care administrative databases covering more than 1.4 million Ontario residents aged 65 years or older. We excluded those with a documented history of atherosclerosis, venous thromboembolism, or cancer within 36 months prior to study enrollment, as well as those prescribed warfarin sodium within 12 months before enrollment. In the primary cohort, we evaluated the subsequent risk of deep vein thrombosis (DVT) among men and women prescribed thyroid replacement therapy, nonstatin lipid-lowering agents, or statins. A second cohort of women only was evaluated in a similar fashion, but estrogen use was added as a third comparison drug group.
There were 125 862 men and women in the primary cohort. After adjusting for age; sex; prior hospitalization; newly diagnosed cancer; or prescribed aspirin, warfarin, or estrogen, statin users (n = 77 993) had an associated decreased risk of DVT relative to those prescribed thyroid replacement therapy (n = 35 978) (adjusted hazard ratio [HR], 0.78; 95% confidence interval [CI], 0.69-0.87). Compared with thyroid replacement therapy, users of nonstatin lipid-lowering agents (n = 11 891) did not seem to be at lower risk for deep vein thrombosis (HR, 0.97; 95% CI, 0.79-1.18). In the secondary cohort of 89 508 women, after adjusting for age, prior hospitalization, newly diagnosed cancer, or prescribed aspirin or warfarin, estrogen users (n = 29 165) had an associated increased risk for DVT compared with those receiving thyroid replacement therapy (n = 22 118) (HR, 1.16; 95% CI, 1.01-1.33), while statin users had an associated decreased risk (HR, 0.68; 95% CI, 0.59-0.79). Nonstatin lipid-lowering agents (n = 5155) were not associated with a reduced risk of DVT compared with thyroid replacement therapy (HR, 0.84; 95% CI, 0.63-1.12).
Among selected individuals aged 65 years or older, statins were associated with a 22% relative risk reduction in the risk of DVT. A randomized clinical trial is needed to evaluate the efficacy of statins for the primary and secondary prevention of DVT.
RANDOMIZED clinical trials have conclusively demonstrated that 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors (statins) are efficacious in the secondary prevention of cardiovascular disease among individuals with both normal and elevated lipid levels.1- 4 In addition to their effect on lipid levels, these drugs seem to have antithrombotic properties.5 In a retrospective, subgroup analysis by the Heart Estrogen Replacement Study (HERS)6 investigators, statin use was associated with a 50% risk reduction of venous thromboembolism (VTE). However, because statin use was one of many VTE risk factors studied and because its protective effect was not explored a priori, only limited conclusions can be drawn from these findings. Therefore, we further investigated whether statin use was associated with a reduction in the risk for deep vein thrombosis (DVT) in a large cohort of men and women.
We conducted a retrospective cohort study by linking Ontario provincial health care administrative databases covering more than 1.4 million senior residents of Ontario over an 8-year period. All persons aged 65 years or older and enrolled with the Ontario Health Insurance Plan for the province of Ontario were considered eligible. The Ontario Health Insurance Plan covers all medical care and prescription drug costs for every Ontario senior citizen. For the primary cohort of men and women, the following mutually exclusive drug groups were studied: (1) statins, the exposure of interest; (2) thyroid replacement hormones, the referent control group, selected because of their lack of known association with VTE; and (3) nonstatin lipid-lowering agents (ie, fibrates, niacin, or bile acid sequestrants), a biological comparison group.
We also studied a second cohort of women only, derived from the same Ontario population as the primary cohort, but included estrogen users as a third, positive comparison drug group.6 This secondary cohort of women was added to test our study design, in that we expected there to be a positive association between estrogen use and DVT.6 Because the primary cohort consisted of both men and women, and since men are not prescribed estrogens, an estrogen drug comparison group would have been impossible. Finally, since estrogen use in the secondary cohort had to be "mutually exclusive" (ie, no concomitant use of statins, thyroid replacement hormones, or nonstatin lipid-lowering agents), secondary cohort participants may have differed slightly from women in the primary cohort.
We studied all individuals who met the following inclusion criteria: (1) had a prescription filled for a study drug between January 1, 1994, and January 31, 1999; (2) had not filled a prescription for another study drug within 365 days of the current study drug (ie. mutually exclusive drug groups); and (3) had filled at least 2 prescriptions for the current study drug within 180 days of its initiation. Excluded were all individuals who (1) had a history of angina, myocardial infarction, coronary revascularization, stroke, carotid endarterectomy, peripheral vascular disease, or peripheral artery revascularization within 36 months prior to study enrollment; (2) had a diagnosis of cancer within 36 months prior to study enrollment; (3) had a diagnosis of DVT or pulmonary embolism within 36 months prior to study enrollment; or (4) had been prescribed warfarin sodium within 12 months prior to study enrollment. Those who were newly prescribed warfarin or diagnosed with cancer after study enrollment remained in the cohort, but the presence of either factor was adjusted for, as outlined below.
The primary outcome of the study was the development of DVT during the follow-up period, after initiation of a statin or control agent. The period of observation was January 1, 1994, to March 31, 1999, but databases were used from January 1, 1991, to March 31, 1999, to account for the presence of antecedent inclusion and exclusion criteria. Individuals who began receiving one of the study drugs were followed up over time to assess the outcome of interest. Participants were censored (ie, determined not to have had a DVT event up to that point) if they temporarily or permanently discontinued taking the study medication. This was defined as having greater than 180 days between subsequent prescriptions, died during the study period, or having reached the end of follow-up to March 31, 1999. An individual was also censored if he or she was prescribed a drug from another study drug group (ie, contamination effect). At least 2 consecutive study drug prescriptions were required to select individuals who were more likely to continue receiving their medications. It has been demonstrated that 60% of patients discontinue their lipid-lowering medications over a 12-month period, most within 3 months of starting treatment.7,8 Individuals older than 65 years are more likely to continue receiving their medications,7 including statins,8 further increasing the likelihood that our defined sample would continue receiving their prescribed agent throughout the period of analysis.
The Ontario Drug Benefits database was used to identify the medications each elderly participant was prescribed during the observation period. This database is maintained by the Ontario Ministry of Health, Ottawa, and includes encrypted patient identifiers, prescription dates, and drug information for all residents of Ontario 65 years or older. Participants who received medications from any of the study drug groups during the year prior to cohort enrollment were excluded, allowing only new users of the medications to be included. There is little missing information in the Ontario databases (<1%), while there is a high degree of coding accuracy.9 Hospitalizations, identified using the Canadian Institute for Health Information Discharge Abstract Database, were used to characterize subsequent events and comorbid illnesses. The discharge abstracts contain the unique health care number, age, and sex of the participant, date of admission, and up to 16 diagnoses, as coded by the International Classification of Diseases, Ninth Revision (ICD-9). The Ontario Health Insurance Plan database, which contains records of all outpatient visits by Ontario residents, including a service date and diagnosis field, was used to capture outpatient diagnoses of VTE. Participant age and sex were retrieved from the Registered Persons Database. This database contains demographic information and health care numbers for all individuals eligible for the Ontario Health Insurance Plan. All data at Institute for Clinical Evaluative Sciences are maintained in an anonymous fashion to ensure confidentiality.
Time-to-event analyses were conducted using the Cox proportional hazards regression model. For the analysis of the primary cohort, adjustments were made for age, sex, hospitalization within 1 year prior to study enrollment,10 newly diagnosed cancer,11 or concurrent prescription of aspirin,12 warfarin,13,14 or estrogen6 during the observation period. For the analysis of the secondary cohort of women, since conjugated estrogen was added as a third comparison agent, no adjustment was made for its use. All participants were identified as being exposed to a potential confounder if the confounding event occurred at any time between 100 days prior to study enrollment and the end of follow-up. The comparative risk for DVT between drug classes was expressed as the adjusted hazard ratio (HR) along with its 95% confidence interval (CI).
Baseline characteristics between men and women were compared within each drug group using either an analysis of variance for continuous variables or the χ2 test for categorical data. All P values were 2-sided and a statistical significance level of .05 was set a priori. All statistical analyses were performed using SAS for UNIX, Version 6.12 (SAS Institute, Cary, NC).
There were 125 862 adults in the primary cohort, for a total of 190 601 person-years of drug use (Table 1). The mean age of all participants was 72.9 years. The average duration of observation varied from 1.1 years for nonstatin lipid-lowering agents to 1.4 years for both statins and thyroid replacement therapy. Within each drug class, men were more likely to be hospitalized or diagnosed as having cancer during the period of observation (P<.001), and had a higher rate of both concurrent aspirin and warfarin use compared with women (P<.001) (Table 1).
During the period of observation, statin users experienced a lower rate of DVT (7.4 per 1000 person-years) than those prescribed thyroid replacement therapy (10.9 per 1000 person-years) (HR, 0.78; 95% CI, 0.69-0.87) (Table 1 and Figure 1). No significant difference was observed between those prescribed nonstatin lipid-lowering agents vs thyroid replacement drugs (HR, 0.97; 95% CI, 0.79-1.18).
Women experienced a higher rate of DVT than men in the thyroid replacement group (12.0 per 1000 person-years vs 7.8 per 1000 person-years), nonstatin lipid-lowering group (10.1 per 1000 person-years vs 8.4 per 1000 person-years), and statin group (8.1 per 1000 person-years vs 6.7 per 1000 person-years) (Table 1). Furthermore, statins were associated with a lower risk of DVT to a greater extent among women (HR, 0.72; 95% CI, 0.63-0.82) than men (HR, 0.97; 95% CI, 0.78-1.22).
There were 89 508 women included in the secondary cohort, for a total of 124 568 person-years of drug use (Table 2). Their overall mean age was 73.5 years. The rate of concurrent aspirin use was highest among those prescribed statins (35.6%) and lowest in the estrogen replacement therapy group (18.9%), while warfarin use was highest among those prescribed thyroid replacement therapy (8.3%) and lowest among estrogen recipients (3.9%).
Women prescribed estrogen replacement therapy were at increased risk for DVT compared with those receiving thyroid replacement therapy (HR, 1.16; 95% CI, 1.01-1.33). Relative to thyroid replacement therapy, statins were associated with a statistically significant reduced risk for DVT (HR, 0.68; 95% CI, 0.59-0.79), while nonstatin lipid-lowering agents were not (HR, 0.84; 95% CI 0.63-1.12).
In a large retrospective cohort of individuals aged 65 years or older, prescribed statins were associated with a 22% relative reduction in the risk for DVT compared with control. This benefit seemed to be only significant in women. Prescription of nonstatin lipid-lowering agents was not associated with a decrease in DVT.
We excluded all individuals with a history of a malignant neoplasm within 36 months prior to study enrollment and further adjusted for prior hospital admission or a concurrent diagnosis of cancer during the observation period.6,10 Although controlling for such VTE risk factors may have produced less biased estimates, we acknowledge the potential for misclassification of either DVT or confounding disease event by depending on physician diagnostic billing codes, which may be unreliable for some hospital diagnoses.15 Such misclassification, if nondifferential, may have lead to an underestimation of the association between either statin or estrogen use and the risk of DVT. Furthermore, we did not account or control for the possibility that some individuals may have been taking over-the-counter aspirin preparations, which could potentially lower their risk for DVT and affect our results.12
This study was composed of a large sample of men and women derived from a population of senior citizens whose health care and medications were covered by the Ontario Health Insurance Plan. By omitting individuals with a previous diagnosis of cardiovascular disease, we hoped to reduce the likelihood of confounding by indication,16 because statin use could merely reflect the presence of atherosclerosis and, accordingly, the use of antithrombotic drugs and lifestyle changes that might reduce an individual's risk of VTE. We also applied strict criteria to define drug use, in that individuals could not receive a drug belonging to another group, and had at least 2 prescriptions filled within 180 days of initiation of study enrollment.8 Finally, these data were based on more than 110 000 person-years of statin use, thus providing reasonable precision for our estimates.
The incidence of DVT in our secondary cohort of women prescribed estrogens was higher than in other studies.6,10,17 In the HERS trial, during 10 985 person-years of follow-up, the rate of combined DVT and pulmonary embolism (PE) in the hormone-treated group was 6.2 per 1000 person-years,6 half that observed in the current study (12.6 per 1000 person-years). Such differences may be explained by the fact that participants in randomized clinical trials tend to be healthier than the general population, and that women in the HERS trial were younger (mean age, 67 years), had established coronary heart disease, and were more likely to receive aspirin (79%) or a lipid-lowering drug (45%) therapy.6 Nevertheless, our observation that estrogens increase the risk for DVT is consistent with that of other studies,6,18 further increasing the likelihood that our data were valid.
It is puzzling that we observed a protective effect from statins in the subgroup of women but not men, since these drugs seemed to be equally effective in both sexes for the treatment of established coronary artery disease.1,3,4 Exploring differences between sexes was not a primary research question, and the detected variance may have been due to chance. Furthermore, the higher rate of aspirin and warfarin use among the men in our study (Table 1) may have resulted in a diminution of any potential protective effect from statins. Our overall findings are supported by those of the HERS investigators, who observed an even greater reduction in DVT or PE with statin use (adjusted HR, 0.5; 95% CI, 0.2-0.9).6 We chose to not evaluate the effect of statins on the risk for PE since testing strategies for PE are often less accurate than for DVT.19,20 Thus, even though they are part of the same spectrum of disease, it remains to be established whether these drugs appear protective against PE.
There is some biological basis to explain how statins may attenuate the risk for VTE. Data from 2 case-control studies have demonstrated a strong association between combined hypercholesterolemia and hypertriglyceridemia and DVT in middle-aged adults (crude odds ratio, 5.1; 95% CI, 2.0-13.0),21 as well as elevated lipoprotein(a) levels and VTE in children (adjusted odds ratio, 7.2; 95% CI, 3.7-14.5).22 Circulating lipids seem to have both prothrombotic and endothelial altering properties.23 For example, ingestion of a fatty meal seems to cause venous endothelial dysfunction, manifested by reduced acetylcholine-mediated venodilatation in healthy adults.24 Thus, it is conceivable that specific circulating lipoproteins heighten the risk for VTE, and that their suppression might be protective. However, because we and the HERS investigators6 failed to observe any beneficial effect against DVT with use of nonstatin lipid-lowering agents, we speculate that statins may also possess other protective properties.
In the Post Coronary Artery Bypass Graft Trial,25 aggressive lipid-lowering with lovastatin effectively reduced both the rate of progression of atherosclerosis in saphenous vein coronary artery bypass grafts, as well as the need for coronary revascularization. Beyond their lipid-reducing ability, statins also seem to alter elements within the vascular endothelium and coagulation cascade, consistent with an antithrombotic effect. For example, 6 months of pravastatin treatment significantly lowered levels of prothrombin fragment 1 + 2 in women,26 indicating reduced thrombin generation, while simvastatin produced the same effect in men.27 Others hypothesize that these drugs may inhibit platelet-derived protease-activated receptor 1 and tissue factor up-regulation that leads to thrombin generation,28 in addition to reducing the levels of both factor VIIc29 and soluble thrombomodulin.30 Collaborative research among those who study lipid-related disorders and arterial and venous thrombotic disease could help to clarify whether it is their lipid-reducing or antithrombotic properties, or both, that best explain the beneficial role of statins in the treatment of atherosclerosis,5 and their potential to do the same for VTE.
We believe that these preliminary data provide the rationale for a randomized clinical trial of statins for the secondary prevention of recurrent VTE. Consenting individuals who have completed 3 months of anticoagulant therapy following a first idiopathic VTE event could be randomized to either continue warfarin therapy for another 12 to 18 months,13,14 or to stop warfarin therapy and begin taking a statin drug for the same duration. The major end point in this trial would be the development of recurrent VTE, balanced against major bleeding. Current or future clinical trials designed to examine the effect of statins on cardiovascular disease might consider collecting secondary data on VTE events as well, which could be combined in a systematic overview. Future observational studies may provide greater insight as to which statins and what dose is there a maximal reduction in primary or secondary VTE events, including PE.
Accepted for publication March 3, 2001.
This investigation was supported in part by the Thrombosis Research Fund, University Health Network, Toronto, Ontario, and the Institute for Health Economics, Edmonton, Alberta.
This study is in memory of Christa Bos, MD, whose encouragement to pursue new ideas was not forgotten.
Corresponding author: Joel G. Ray, MD, FRCPC (e-mail: firstname.lastname@example.org).