Does administration of vaginal estrogens increase the risk of venous thromboembolism: A case-control study.

Abstract

Vaginally administered oestrogen is an effective treatment of vaginal atrophy1 and recurrent urinary tract infections2 in postmenopausal patients. The systemic absorption of vaginal oestrogen is low and thus unlikely to substantially raise the systemic oestrogen level among users.3 However, the United States Food and Drug Administration has issued a black box warning against the use of vaginal oestrogen in patients with current or former venous thromboembolism (VTE) based on an increased risk of VTE among users of systemic estrogens. Whether this increased risk associated with systemic therapy also applies to vaginal oestrogen has not been thoroughly investigated. Of note, this warning is also listed in the Danish package inserts of vaginal oestrogen. Consequently, postmenopausal women with a history of VTE may unnecessarily be withheld a safe and effective treatment for vaginal atrophy and recurrent urinary tract infections. Therefore, we aimed to investigate whether postmenopausal women, with and without a history of VTE, using vaginal oestrogen are at increased risk of VTE. Using the Danish health registers, we conducted a population-based, nested case–control study. This study is reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.4 Eligible individuals were women above 55 years of age residing in Denmark from 1 January 2010 to 31 December 2018. We used a cut-off of 55 years as most women can be assumed to be postmenopausal at this age. Cases were defined by a first hospital diagnosis of venous thrombosis or pulmonary embolism during the study period. We matched up to four controls to each case based on the year of birth and history of prior VTE between 1995 and 2010. Controls were assigned an index date similar to the diagnosis date of the matched case. Cases were eligible to be sampled as controls before the date of receiving a VTE diagnosis. By this control sampling (risk-set sampling), the derived odds ratio (OR) is an unbiased estimate of the incidence rate ratio that would have emerged in a cohort study from the same source population.5 We applied the following exclusion criteria to the cases and controls: dispensing of systemic oestrogen within 12 months prior to index date, dispensing of oral anticoagulants within 12 months prior to index date, a history of cancer or hematologic malignancy, inflammatory bowel disease or coagulopathy. We also required that the subject had not been hospitalised within 3 months prior to index date to avoid confounding by surgical procedures, trauma or critical illness. For a flowchart over the selection of study cases, see Figure S1. We considered a case or control exposed if they redeemed a prescription for vaginal oestrogen during the 3 months prior to the index date. This choice of exposure window was validated by analysing the waiting time distribution for local oestrogen dispensing (Figure S2).6 The OR associating local oestrogen with VTE was estimated using multivariable conditional logistic regression. Potential confounders and risk factors of VTE were included as covariates in the regression model, for example, frequency of recurrent UTI, a hospital diagnosis of obesity or venous insufficiency and use of drugs to treat cardiovascular disease. Confounding by age, calendar time and history of VTE was controlled for by employing conditional logistic regression stratified on the individual risk-set. For a detailed list of covariates, prescription and diagnosis codes used to define exposures, outcomes and covariates, see Table S1. Out of 1 223 474 women aged 55 years or older, we identified 13 748 incident cases of VTE and 54 948 age-matched controls. The median age in the study population was 74 years (interquartile range 66 to 82). Compared with controls, cases had a higher prevalence of most comorbidities and were more often users of prescription drugs used in the treatment of chronic conditions (Table 1). Among cases, 1296 (9.4%) filled a prescription for vaginal oestrogen within 3 months before receiving a VTE diagnosis, compared with 5482 (10%) controls. We did not find an increased risk of VTE (adjusted OR 0.87, 95% confidence interval: 0.81 to 0.93), deep venous thrombosis (adjusted OR 0.90, 0.83 to 0.98) or pulmonary embolism (adjusted OR 0.81, 0.73 to 0.91) when comparing users of vaginal estrogens to unexposed individuals (Table 2). We mostly found null associations in stratified analyses (Table 2). Most importantly, we found no association between vaginal oestrogen therapy and VTE among women with a history of VTE (adjusted OR 1.13, 0.91 to 1.42). We performed several post hoc analyses. To investigate the effect of exposure misclassification, we shortened the exposure assessment window to 30 days, which yielded unchanged risk estimates (adjusted OR 0.89, 0.81 to 0.99). To investigate whether the risk of VTE increases with the cumulative dose, we estimated ORs for the following dose strata (0, 1–39, 40–119, 120–199, 200–399, 400 + DDD) compared with never use of vaginal estrogens. The cumulative dose analysis yielded a slight tendency towards an inverse association between a higher cumulative dose and the risk of VTE (Table S2). Finally, we investigated whether the slightly reduced risk of VTE in users of vaginal oestrogen could be related to a healthy-user bias. We performed a case-time-control analysis7 among the previously sampled cases and controls. The case-time-control design is unaffected by confounders that are stable over time, such as a healthy lifestyle. We used one focal window and four referent windows, all with a duration of 90 days spaced equally apart. Using conditional logistic regression, we obtained an OR of 0.99 (0.87 to 1.11) among cases. In this population-based case–control study, we found no increased risk of VTE among women exposed to vaginal estrogens. Furthermore, we found reassuring results regarding patients with prior VTE, as vaginal oestrogen therapy was not associated with an increased risk of VTE in this group either. Our study had the following strengths: It was based on the highly valid Danish nationwide registers and Denmark's universal tax-funded healthcare system. The risk of selection bias in our study was minimal, as health care is free for all residents, and all hospital contacts are registered in the Danish National Patient Registry. The codes used to identify patients with VTE have been validated, with a reported positive predictive value of 88% and 72% for first-time and recurrent VTE, respectively.8 The main limitation of our study was the lack of information on potential confounders such as smoking, alcohol consumption and body weight. We sought to adjust for these using hospital diagnoses as proxies (Table S1). Furthermore, we saw no reason to suspect that any of these factors are strongly associated with the use of vaginal oestrogen, thus limiting their potential to confound our associations. Another limitation of our study was the potential presence of nondifferential exposure misclassification, that is, a redeemed prescription of vaginal oestrogen not necessarily leading to continuous use of the drug, which may bias results towards the null. We addressed this bias by shortening the exposure assessment window in post hoc analyses and found comparable results. The supplementary analysis on the effect of the cumulative dose revealed a dose-related fall in risk of VTE. Slightly reduced risks of VTE in the main and cumulative dose analysis may be explained by a healthy-user bias. This bias was eliminated in the case-time-control analysis. Our findings are compatible with the results from two recent cohort studies, which found no increased risk of pulmonary embolism or venous thrombosis among users of vaginal oestrogen.9, 10 However, both studies lacked statistical power to rule out a nontrivial risk, having upper limits of confidence intervals of 1.28 and 1.93. Neither of these studies addressed the risk among women with a history of VTE. Our study may reassure clinicians that vaginal oestrogen use is safe and that it should not be withheld with reference to VTE risk, not even in women with a history of VTE. This work was funded by the University of Southern Denmark. None. Table S1. Prescription and diagnosis codes used to define exposures, outcomes, and covariates Table S2. Number of cases and controls, crude, and adjusted odds ratios for each cumulative dose stratum Figure S1. Flowchart over the selection of study cases.* = including combined oral contraceptives and selective oestrogen receptor modulators Figure S2. Waiting time distribution for vaginal oestrogen. Number of individual users of vaginal oestrogen redeeming their first prescription per month since January first, 2010. The majority of users appear early in the period, implying that regular users tend to redeem their prescription with short intervals. After this short period, the graph is dominated by new, incident users. Judged from the graph the majority of prevalent users renew their prescription within the first three months of the window, which suggest that a three-month window offers a reasonable tradeoff between sensitivity and specificity Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.