In postmenopausal women, oral hormone therapy (HT) is responsible for more than two venous thrombosis (VT) events per 1000 person-years of treatment (Rossouw et al, 2002). Recent evidence supports a synergistic effect of smoking and oral contraceptives on the risk of VT (Pomp et al, 2008). Because HT has similar components as oral contraceptives – oestrogen ± progestogen, albeit in lower doses – smokers could represent a group in whom the VT risk associated with HT is considerably increased. Although no interaction between smoking and HT was found in the Women's Health Initiative (WHI) clinical trial (Cushman et al, 2004), the small number of smoking VT cases and the selection of healthy participants in that trial may limit the validity and generalizability of this null finding. Our aim was to investigate the presence of this interaction in a large population-based study.
Among participants of the Heart and Vascular Health Study (HVH), a case–control study of cardiovascular outcomes among Group Health Cooperative (GHC) members in Washington State (Smith et al, 2004), we identified all post-menopausal women who suffered a first deep-venous thrombosis (DVT) and/or pulmonary embolism (PE), from 1 January 1995 to 31 December 2009. More than 90% of them agreed to participate (n = 1967), of whom >95% had positive diagnostic imaging test results. Women with no history of VT (n = 5146) were selected from a pool of randomly-chosen HVH controls. We excluded subjects who smoked pipes/cigars (n = 3), subjects with missing smoking data (n = 25), users of oral contraceptives during menopause (=31) and users of oestrogen patches or oral progestogen without oestrogen (n = 62).
The use of HT was determined from the GHC pharmacy database, from which we assumed 80% compliance with prescribing instructions (Smith et al, 2004). Information on smoking status came from a telephone interview for 58% of subjects and from the medical record review otherwise, with an agreement of 92% when both were available (Kappa = 0·85). Demographic characteristics, past medical history and details of the VT were abstracted from the entire GHC ambulatory medical record and the telephone interview.
Logistic regression models evaluated the association between HT use and the risk of VT, adjusting for matching factors (age, index years, hypertension), body mass index (BMI), recent diagnoses of cancer (within 5 years before to 3 months after index) and recent hospitalizations (within 6 months before index). Effect modification by smoking status was assessed by adding a multiplicative interaction term. Missing values for BMI (1%) and recent hospitalization (16%) were multiply imputed with a multivariate normal regression model generating 20 imputations (Schafer, 1999).
During the study period, 1926 cases suffered a DVT (n = 949), a PE (n = 690) or both (n = 287). Mean age and BMI of cases were 71 years and 29·5 kg/m2. Among 5066 controls, 90% were White, 9·4% were current smokers and 25·1% were using oral HT (9·9% with progestogen, almost exclusively medroxyprogesterone acetate). More than 85% of oestrogen preparations were conjugated equine oestrogen or esterified oestrogens. The median duration of use of oral HT after enrollment in GHC was 8·6 years for cases and 10·2 years for controls.
Overall, compared with non-users, women using oral HT had an increased risk of VT (odds ratio [OR] 1·35, 95% confidence interval [CI] 1·16–1·57). The smoking status of the participants did not significantly modify this association (Table 1). The OR for VT associated with HT was 1·42 (95%CI 1·16–1·74) among never smokers and 1·60 (95%CI 1·00–2·62) among current smokers, with no significant difference between these (OR ratio: 1·13, 95%CI 0·68–1·90, P = 0·55). Similarly, no interaction was found when comparing former with never smokers: ratio of ORs 0·84, 95%CI 0·61–1·14, P = 0·29.
|No hormone therapy||Any hormone therapy|
|N [cases/controls]||OR (95%CI)a||N [cases/controls]||OR (95%CI)a||Ratio of ORs (Interaction term)|
|Never smokers||767/2062||1·0 (reference)||222/709||1·42 (1·16–1·74)||Reference|
|Current smokers||140/371||1·0 (reference)||38/107||1·60 (1·00–2·62)||1·13 (0·68–1·90)b|
|Former smokers||599/1361||1·0 (reference)||160/456||1·16 (0·91–1·48)||0·84 (0·61–1·14)c|
We explored effect modification by dose, oestrogen type, progestogen use and smoking dose. In analyses restricting HT users to those taking the modal dose of oestrogen, the OR associated with HT was similar for both never and current smokers: 1·41 (95%CI 1·12–1·77) and 1·45 (95%CI 0·83–2·54), respectively (OR ratio 1·03, 95%CI 0·57–1·87). When oestrogen types were restricted to conjugated equine oestrogen, the ORs were not different between never and current smokers: OR 1·94 (95%CI 1·51–2·49) and OR 1·67 (95%CI 0·86–3·24), respectively, OR ratio 0·86 (0·43–1·74). Further, we found no interaction when stratifying HT use into unopposed and opposed HT and when stratifying current smoking into light and heavy smoking (<20 and >20 cigarettes/d).
Compared with never-smokers not using HT, we observed a 20% higher risk of VT in current smokers not using HT, a 40% higher risk of VT in never-smokers using the modal dose of HT and a 70% higher risk of VT in current smokers using HT (Table 2). There was no suggestion of a material additive interaction between smoking and HT.
|Smoking status||Hormone therapya||Cases||Controls||OR (95%CI)b|
In this large, population-based, case–control study, smoking status was not found to modify the association between HT and VT. Smoking itself is a weak risk factor for incident VT (Blondon et al, 2013). Therefore, our results suggest that smoking is not an important factor for the development VT in users of oral HT, similarly to those of the WHI clinical trial (Cushman et al, 2004; Curb et al, 2006).
Our findings contrast with the observed synergistic effect of smoking and OC on the risk of VT (Pomp et al, 2008). This difference may be explained by the much lower potency of oestrogen found in HT preparations than in OC preparations, if a true biological interaction between oestrogen and smoking exists.
The increased risk of VT associated with oral HT in our data is not as strong as reported elsewhere (Sweetland et al, 2012). Reasons for this may include the use of esterified oestrogens (Smith et al, 2004), low oestrogen dose and the high proportion of prevalent users.
Oral oestrogens also increase the incidence of stroke and coronary heart disease in postmenopausal women (Rossouw et al, 2007), and smoking increases the risk of arterial cardiovascular events more than that of VT (Brækkan et al, 2012). For these reasons, caution is needed when prescribing oral HT to current smokers.
The strengths of our study include its population-based design, the ascertainment of VT events and the high-quality measures of HT use through pharmacy records. Its limitations include the low generalizability to non-White races and new HT users.
In conclusion, our data do not support the hypothesis that smoking modifies the association between HT and the risk of VT.