Does the type of hormone replacement therapy influence the risk of deep vein thrombosis? A prospective case–control study
J. D. Douketis, St Joseph's Hospital, Room F-541, 50 Charlton Ave. East, Hamilton, Ontario, Canada, L8N 4A6.
Tel.: +1 905 521 6178; fax: +1 905 521 6068; e-mail: email@example.com
Summary. Background: Although hormone replacement therapy (HRT) is associated with an increased risk of deep vein thrombosis (DVT), it is not clear if the risk differs in users of combined estrogen-progestin HRT and estrogen-only HRT. Methods: We prospectively studied postmenopausal women with suspected DVT in whom HRT use status was ascertained and who subsequently had objective diagnostic testing to confirm or exclude DVT. Cases were patients with idiopathic DVT, in whom there were no DVT risk factors, and controls were patients without DVT, in whom there were also no DVT risk factors. The risk of DVT was determined in users of estrogen-progestin HRT and estrogen-only HRT by comparing the prevalence of current HRT use in cases with idiopathic DVT and controls without DVT (reference group). Multivariable regression analysis was done to adjust for factors that might confound an association between HRT use and the risk of DVT. Results: One thousand one hundred and sixty-eight postmenopausal women with suspected DVT were assessed, from whom 95 cases of idiopathic DVT and 610 controls without DVT and no DVT risk factors were identified. Estrogen-only HRT was associated with an increased risk for DVT that was not statistically significant [odds ratio (OR) = 1.22; 95% confidence interval (CI) 0.57, 2.61]. Estrogen-progestin HRT was associated with a greater than 2-fold increased risk for DVT (OR = 2.70; 95% CI 1.44, 5.07). Conclusion: The risk of developing DVT may be higher in users of combined estrogen-progestin HRT than in users of estrogen-only HRT.
An association between oral hormone replacement therapy (HRT) and deep vein thrombosis (DVT) is established, as HRT use confers a 2–3-fold increased risk of developing DVT [1–8]. One issue that requires further consideration is the effect of the type of HRT on the risk of DVT, specifically whether the risk of DVT differs in users of combined estrogen-progestin HRT and unopposed estrogen-only HRT. Previous observational studies provided conflicting results: the risk of DVT was higher with estrogen-progestin HRT in two studies [1,4]; the risk of DVT was higher with estrogen-only HRT in one study ; the risk of DVT was essentially the same with estrogen-progestin and estrogen-only HRT in one study . Furthermore, the validity of these findings might be questioned because ascertainment of the type of HRT used was based on a retrospective review of patient records, with the potential to misclassify the type of HRT used and, therefore, misrepresent the risk of DVT with estrogen-progestin and estrogen-only HRT [1–4]. Randomized controlled trials have not been able to address this issue, as these studies assessed the risk of DVT only with estrogen-progestin HRT [6–8].
Investigating the risk of DVT with different types of HRT is clinically relevant for two reasons. First, because a considerable proportion of postmenopausal women who have had a hysterectomy are only eligible to receive estrogen-only HRT , it is important to provide reliable estimates of the risk of DVT for such patients. Second, prior knowledge of whether the progestin component of HRT influences the risk of DVT is relevant when considering the safety of progestin-only hormonal therapy in women with previous DVT in whom estrogen-containing hormonal therapy is avoided . Although progestin-only hormonal therapy is considered safe in women with previous DVT [11–13], the absence of a prothrombotic effect of progestins, which would support this practice, is not established. We therefore performed a prospective case–control study to investigate the risk of DVT in users of estrogen-progestin HRT and estrogen-only HRT.
Patients and methods
We prospectively assembled a group of postmenopausal women with suspected DVT. Patients were recruited from outpatient, emergency room, or in-hospital settings from 12 clinical centres (eight in Canada, two in Italy, two in the Netherlands). Patients were enrolled between 1998 and 2001, prior to the publication of the Women's Health Initiative Study . After a clinical assessment, during which information was obtained about HRT use and DVT risk factors, objective diagnostic testing for DVT was performed to confirm or exclude DVT reliably. Cases were defined as patients with idiopathic DVT who had no known DVT risk factors, and controls were defined as patients without DVT who, similarly, had no known DVT risk factors. This study design has two advantages over previous case–control studies that investigated HRT use and DVT risk. First, cases and controls were phenotypically similar, as they were derived from the same group of postmenopausal women with suspected DVT . Second, the exposure of interest (HRT use) was ascertained prior to and independent of the outcome of interest (DVT).
Consecutive postmenopausal women with suspected lower limb DVT were eligible for this study. Postmenopausal status was defined by one of the following criteria [15,16]: spontaneous menopause, with no menses for at least 6 months; surgical menopause due to bilateral oopherectomy; or the presence of menopausal symptoms, such as vasomotor flushing, for at least 6 months. Patients were excluded if they had one or more of the following characteristics: suspected pulmonary embolism; amenorrhea due to primary or secondary ovarian failure; cognitive impairment or language barrier. Patients with suspected pulmonary embolism were excluded because the accuracy of diagnostic testing is suboptimal, as up to 20% of patients will not have pulmonary embolism reliably diagnosed or excluded .
Eligible and consenting patients had a clinical assessment that included documentation of presenting clinical features, risk factors for DVT, and factors that might influence HRT use. HRT use status was ascertained by direct patient questioning, and was aided by showing patients photographs of current HRT preparations. The clinical likelihood of DVT was subjectively classified as low, moderate or high, based on an assessment of patients' presenting clinical features. After the clinical assessment, patients underwent objective diagnostic testing for DVT, with test interpretation by observers who had no prior knowledge of patients' HRT use status.
Determination of HRT use status
HRT was defined as an oral or transdermal formulation consisting of an estrogen with or without a progestin. Patients who were current users of selective estrogen receptor modulators, estrogen antagonists, or phytoestrogens were excluded from the study during a post hoc analysis of study data. Patients were classified as HRT current users or HRT non-users, with the latter group consisting of previous and never users. Current users were HRT users who had been receiving HRT for at least 1 month prior to the study assessment. Previous users were patients who had received HRT prior to 1 month of the study assessment but not within the month before the assessment. Never users were patients who had never received HRT. In current users, the HRT type was classified as estrogen-progestin or estrogen-only, and as oral or transdermal.
Determination of DVT status
Diagnostic testing for DVT was based on center-specific diagnostic algorithms. However, all patients required objective confirmation or exclusion of DVT based on validated criteria to be included in the study. DVT was confirmed by a non-compressible vein segment on venous ultrasound , or a constant intraluminal filling defect on two or more views on venography . DVT was excluded by one or more of the following: normal venography ; normal compression ultrasound and a normal SimpliREDTMD-dimer test (Agen Inc., Brisbane, Australia) [20,21]; normal compression ultrasound and a low clinical likelihood for DVT [21–23]; normal serial compression ultrasound testing on day 1 and day 7 ± 3 after presentation ; or normal impedance plethysmography and a normal SimpliREDTMD-dimer test .
Identification and definition of cases and controls
From the assembled study population, cases and controls were identified centrally, by the Study Coordinating and Methods Center, based on prespecified criteria. Cases were identified from patients in whom DVT was confirmed and controls were identified from patients in whom DVT was excluded. Cases were defined as patients with idiopathic DVT, in whom DVT occurred in the absence of the following DVT risk factors: recent (within 1 month) surgery, immobility or trauma; thrombophilia; previous venous thromboembolism; active cancer (treated within 6 months or palliative) . Controls were defined as patients in whom DVT was excluded, and who did not have any of the aforementioned DVT risk factors.
To assess the effect of HRT type on the risk of DVT, we compared the proportion of cases with idiopathic DVT who were current users of estrogen-progestin or estrogen-only HRT with the proportion of controls without DVT risk factors who were current users of estrogen-progestin or estrogen-only HRT. In this way, an effect of HRT on the risk of DVT would not be confounded by additional DVT risk factors such as recent surgery or active cancer. Reported risk estimates were based on the multivariable modeling and expressed as odds ratios (ORs), with corresponding 95% confidence intervals (CIs). All P-values were two-sided.
Multivariable logistic regression analysis (SAS version 8.1, Cary, NC, USA) was done to adjust for factors (unrelated to DVT risk) that might confound an association between HRT use and DVT risk. The potential confounders considered were: clinical center; patient age (deciles); education level (primary, secondary, postsecondary); body mass index (≤ 19.9, 20–29.9, or ≥ 30 kg m−2); smoking status (yes/no); previous hysterectomy (yes/no); bilateral oopherectomy (yes/no); vertebral or long bone fracture (yes/no); osteoporosis (yes/no); breast cancer (yes/no); and clinical likelihood for DVT (low, moderate, high). Variables in the final multivariable models were selected based on their achieved level of statistical significance, defined as P < 0.10 in the simple single-factor univariate model. To determine if the risk estimates were consistent across clinical centers, interaction terms between the HRT variable and center were added to the model and assessed using the likelihood ratio test.
There were 1178 postmenopausal women with suspected DVT who were assessed for this study. After a post hoc review of patient data, we excluded 10 patients with thrombophilia because this small sample would preclude meaningful analyses of such patients. From the remaining group of 1168 patients, we identified 95 cases with idiopathic DVT and no antecedent DVT risk factors (other than possible HRT use), and 610 controls without DVT and no DVT risk factors. The characteristics of the cases and controls and the source group of 1168 patients are presented in Table 1.
Table 1. Patient characteristics: cases, controls, and all patients
|Age (years): < 50||3 (3)||37 (6)||70 (6)|
| 50–59||26 (27)||151 (25)||261 (22)|
| 60–69||26 (27)||147 (24)||292 (25)|
| 70–79||24 (25)||193 (32)||374 (32)|
| > 79||16 (17)||82 (13)||171 (15)|
|Body mass index* < 20 kg m−2||9 (9)||25 (4)||63 (6)|
| 20–29 kg m−2||57 (60)||368 (60)||715 (62)|
| > 29 kg m−2||26 (27)||206 (34)||367 (32)|
|Highest level of education attained† primary (grades 1–6)||27 (28)||161 (26)||313 (27)|
| secondary (grades 7–12)||47 (49)||342 (56)||651 (57)|
| postsecondary (college or university)||18 (19)||100 (16)||187 (16)|
|Previous bilateral oopherectomy (with or without hysterectomy)||15 (16)||145 (24)||277 (24)|
|Previous hysterectomy only (without bilateral oopherectomy)||22 (23)||272 (45)||502 (43)|
|Osteoporosis||12 (13)||143 (23)||267 (23)|
|Current smoker||16 (17)||90 (15)||178 (15)|
|Previous vertebral or long bone fracture||6 (6)||41 (7)||101 (9)|
|Previous breast cancer||8 (8)||29 (5)||83 (7)|
|Hormone replacement therapy use current users||36 (38)||183 (30)||314 (27)|
| previous users||10 (11)||71 (12)||130 (11)|
| never users||49 (51)||356 (58)||724 (62)|
|Current hormone replacement therapy users (duration) < 1 year||9 (9)||32 (5)||62 (5)|
| 1–4 years||5 (5)||16 (3)||31 (3)|
| > 4 years||22 (23)||135 (22)||221 (19)|
|Current hormone replacement therapy users (type) oral estrogen-progestin||24 (25)||56 (9)||105 (9)|
| oral estrogen-only||12 (13)||127 (21)||209 (18)|
| transdermal||3 (3)||26 (4)||48 (4)|
|Clinical likelihood for DVT low||15 (16)||472 (77)||649 (56)|
| moderate||48 (50)||128 (21)||356 (31)|
| high||32 (34)||10 (2)||152 (13)|
|Risk factors for deep vein thrombosis:|
| Previous venous thromboembolism||0||0||159 (14)|
| Active cancer (palliative or treated within 6 months)||0||0||127 (11)|
| Recent (within 4 weeks) confinement to bed for ≥ 3 days||0||0||118 (10)|
| Recent (within 4 weeks) lower limb trauma||0||0||102 (9)|
| Recent (within 4 weeks) surgery requiring an anesthetic||0||0||94 (8)|
| Lower limb paralysis||0||0||27 (2)|
| Currently receiving tamoxifen therapy||0||0||20 (2)|
Effect of HRT type and risk of DVT
As shown in Table 2, use of any type of HRT was associated with about a 2-fold increased risk of DVT compared with no HRT use (OR = 1.93; 95% CI 1.16, 3.22). Users of estrogen-progestin HRT had a greater than 2-fold increased risk of DVT compared with non-users (OR = 2.70; 95% CI 1.44, 5.07). Users of estrogen-only HRT had a small increased risk of DVT compared with non-users that was not statistically significant (OR = 1.22; 95% CI 0.57, 2.61). Despite the discrepant point estimates of DVT risk with estrogen-progestin and estrogen-only HRT (2.70 vs. 1.22), a direct comparison of risks based on the likelihood ratio test indicated that this difference in risks did not attain statistical significance (P = 0.096). The exclusion of 26 cases and controls who were current users of transdermal HRT did not change these findings.
Table 2. Risk of deep vein thrombosis (DVT) related to hormone replacement therapy (HRT) use
|HRT non-users (reference group)||59||427||1.0||1.0|| |
|Current HRT users (any type)†||36||183||1.42 (0.91, 2.23)||1.93 (1.16, 3.22)||0.011|
|HRT type†|| || || || || |
| Estrogen-only||12||127||0.68 (0.36, 1.31)||1.22 (0.57, 2.61)||0.61|
| Estrogen-progestin||24||56||3.10 (1.79, 5.38)||2.70 (1.44, 5.07)||0.002|
In terms of results across clinical centers, there was no statistically significant difference in center-specific point estimates for DVT risk with HRT use based on the likelihood ratio tests (P = 0.24), thereby indicating no significant across-center heterogeneity of results and supporting the calculation of summary effect measures across centers.
Potential confounders included in the final multivariable models were previous hysterectomy and clinical center. A previous hysterectomy is likely to be a confounder with regard to the HRT type used because patients with an intact uterus typically receive estrogen-progestin HRT, whereas those with a hysterectomy receive estrogen-only HRT . Clinical center was included as a potential confounder because of variability in HRT use rates and confirmed DVT rates across centers, which probably reflect country-specific HRT use patterns and center-specific patient referral patterns, respectively. Other potential confounders that were considered and rejected (P > 0.10) for the final multivariable models were: age, education level, body mass index, smoking status, bilateral oopherectomy, vertebral or long bone fracture, osteoporosis, breast cancer, and the clinical likelihood of DVT.
We found that the association between HRT use and the risk of DVT appeared stronger in users of estrogen-progestin HRT (OR = 2.70) than in users of estrogen-only HRT (OR = 1.22). This finding is unexpected because progestins have not been considered to have prothrombotic properties [11–13,27]. However, there is emerging evidence that progestin-only hormonal therapy may be associated with an increased risk for DVT. High-dose medroxyprogesterone, when used for the treatment of menstrual disorders, is associated with a 5-fold increased risk of DVT [28,29]. Furthermore, the progestin component of the oral contraceptive appears to influence the risk of DVT [30–32]. Finally, observational studies involving women who were receiving progestin-only oral contraceptives suggest that preparations such as intramuscular medroxyprogesterone may confer an increased risk of DVT [28,29]. Despite such data that support a progestin-associated prothrombotic effect, the biological mechanism to account for these findings is not known. Although it has been postulated that progestins might augment a prothrombotic effect of estrogen, due to conversion of progestins into estrogenic compounds [33,34], this process appears to be marginal and unlikely to be of clinical importance . Furthermore, studies investigating the effect of HRT on coagulation parameters have not found differences in coagulation activation between estrogen-progestin and estrogen-only HRT preparations [35,36].
The validity of our findings is contingent on demonstrating that the potential for bias, which can occur in any observational study, would not have influenced the study results. In this study, the main source of bias is diagnostic suspicion bias—that HRT users with DVT were preferentially referred for diagnostic assessment over HRT users without DVT, which would artificially strengthen an association between HRT use and DVT. We believe this potential source of bias was minimized and did not influence the validity of our findings based on the following considerations. First, cases with DVT and controls without DVT were derived from a single population of postmenopausal women with clinically suspected DVT, thereby ensuring that all patients were subjected to the same referral process. This would minimize preferential referral of HRT users with DVT over HRT users without DVT. Second, DVT was diagnosed by objective criteria, and diagnostic testing was performed without prior knowledge of HRT use, thereby minimizing the likelihood of preferential diagnosis of DVT in HRT users over non-users. Third, ascertainment of HRT exposure occurred prior to and independent of diagnostic testing for DVT, thereby minimizing the potential that HRT use would be preferentially ascribed to patients with DVT than without DVT. Furthermore, it is likely that the ascertainment of HRT exposure was reliable because this information was obtained prospectively during a face-to-face patient interview, and with the use of pictorial aids to assist in patients' recall of the HRT preparations they might have been receiving. Fourth, when we performed a sensitivity analysis of results by adjusting for patients with a low clinical probability for DVT who, in the presence of minimal symptoms, might have been preferentially referred for diagnostic DVT testing because of HRT use, our findings remained unchanged. Finally, diagnostic suspicion bias would not have influenced the risk of DVT in users of estrogen-progestin HRT and estrogen-only HRT, because it is unlikely that there was preferential referral of patients based on the type of HRT they were receiving.
Our study has potential limitations that warrant discussion. First, we could not assess the risk of DVT in users of transdermal HRT as only 26 patients were users of this type of HRT. This issue is important, because of evidence suggesting that transdermal HRT does not confer an increased risk of DVT . Second, we acknowledge that when we compared the risk of DVT in users of estrogen-progestin and estrogen-only HRT, the difference in the risk ratio point estimates (2.70 vs. 1.22) was not statistically significant (P = 0.096). However, our findings appear consistent with data from the recently published Women's Health Initiative randomized placebo-controlled trial investigating estrogen-only HRT . Whereas, estrogen-progestin HRT was associated with a greater than 2-fold increased risk for venous thromboembolism (risk ratio = 2.11; 95% CI 1.26, 3.55) , estrogen-only HRT was associated with a non-significant increased risk of venous thromboembolism (risk ratio = 1.33; 95% CI 0.99, 1.79) .
The principal clinical implication of our study is that the recommended use of progestin-only hormonal therapy in women with previous DVT and women with other DVT risk factors, based on its purported lack of a prothrombotic effect, should be reconsidered [11–13]. Although our study showed that a higher risk of DVT occurred when a progestin was coadministered with estrogen, the possibility that progestin-only therapy confers an increased of DVT should be considered. Additional studies are needed to assess the prothrombotic effects of progestin-only therapy and to determine if such treatment can be safely administered in women with previous DVT or other risk factors for DVT.
In summary, our study suggests that the risk of developing DVT may be higher in users of combined estrogen-progestin HRT than in users of estrogen-only HRT.
This study was supported by funding from the Physician's Services Incorporated Foundation of Ontario, Canada, and the Hamilton Civic Hospitals Research Center. The authors thank the following colleagues for their valued assistance in completing this study: M. Frulla, P. Bagatella, T. Schnurr, K. Kinnon, P. Stevens, J. Johnson and S. Smail.
Contribution of authors
Study conception and design: J.D.D., J.A.J., J.S.G., C.K., S.M.B. Acquisition of data: J.D.D., J.A.J., M.A.C., C.K., S.M.B., M.B., F.P., A.G.T., S.M., P.v.N., P.P., D.R.A., P.S.W., M.J.K., M.R.M-G., L.C., J.S.G. Analysis and interpretation of data: J.D.D., J.A.J., C.K., D.R.A., J.S.G. Drafting of the manuscript: J.D.D., J.A.J., C.K., D.R.A., J.S.G. Critical revision of the manuscript for important intellectual content: J.D.D., J.A.J., M.A.C., C.K., S.M.B., M.B., F.P., A.G.T., S.M., P.v.N., P.P., D.R.A., P.S.W., M.J.K., M.R.M-G., L.C., J.S.G. Statistical expertise: J.A.J. Obtained funding: J.D.D., J.A.J., J.S.G. Administrative, technical, or material support: J.D.D., J.A.J., M.A.C., L.C., J.S.G., M.A.C., C.K., S.M.B., M.B., F.P., A.G.T., S.M., P.v.N., P.P., D.R.A., P.S.W., M.J.K., M.R.M-G. Study supervision: J.D.D., L.C., J.S.G.