Bleeding risk in breast cancer patients during concomitant administration of warfarin and tamoxifen: A population‐based nested case‐control study

We aimed to investigate whether the concomitant use of tamoxifen with warfarin is associated with higher risk for bleeding among patients with early estrogen‐receptor (ER)‐positive breast in a population‐based nested case‐control study. We identified 1787 patients taking warfarin and 92 cases hospitalized for bleeding and found an adjusted odds ratio (OR) of 1.42 (95% confidence interval (CI): 0.84‐2.40) for the risk of bleeding in patients treated with warfarin that initiated tamoxifen within the previous 30 days. As a result, we could not definitively rule out a potential association between tamoxifen use during warfarin and bleeding risk in patients with breast cancer.


| Identification of cases and controls
Within the study cohort, we defined cases as those admitted to hospital with a diagnosis of bleeding as defined above. The date of hospital admission served as the index date for all analyses. From the same cohort, controls were randomly selected using incidence density sampling. For each case, 50 controls exposed to warfarin who had not been hospitalized due to bleeding on the index date of the corresponding case were selected.

| Exposure to Tamoxifen
The primary exposure of interest was the prescription of tamoxifen (ATC code: L02BA01) within 30 days prior to index date because the clinical consequences of this interaction are expected to manifest following the introduction of tamoxifen in patients already receiving warfarin. The primary analysis was restricted to tamoxifen prescriptions initiated within 30 days before the index date. As a secondary analysis, we examined new tamoxifen prescriptions 31-180 days before the index date.

| Statistical analysis
We performed univariate and multivariate conditional logistic regression to estimate the odds ratios (ORs) and 95% confidence interval for the association between bleeding during warfarin therapy and new exposure to tamoxifen (within 30 days) or 31-180 days before index date.
ORs in the multivariable models were adjusted for age, region, year of breast cancer diagnosis, progesterone-receptor status, disease stage, grade, prior history of stroke, prior history of bleeding, comorbidities within 2 years from index date (atrial fibrillation, hypertension, chronic kidney disease, chronic liver disease), education level, and marital status.
Two prespecified sensitivity analyses were performed: separate analyses based on region (Stockholm-Gotland region and Uppsala/ Örebro-Northern regions), and analyses included only patients with bleeding as main diagnosis at hospital admission (surrogate outcome for severe bleeding).

| RE SULTS
We identified 1787 ER-positive breast cancer patients taking warfarin ( Figure 1). Within this group, we identified 159 patients who had been hospitalized for bleeding. Of those, 92 bleeding cases were included in the analysis after the exclusion of patients with prior bleeding or lack of warfarin exposure at index date. The characteristics of the study cohort are presented in Table 1.

| Primary and secondary analyses
In the primary analysis, we found an adjusted OR of 1.42 (95% CI: 0.84-2.40) for the risk of bleeding in patients treated with warfarin that initiated tamoxifen within the previous 30 days ( Table 2).
Adjusted ORs turned to less than 1 with more distant exposures to tamoxifen (adjusted OR: 0.85, 95% CI: 0.20-3.57 for patients initiating tamoxifen 31-180 days before the index date).

| Sensitivity analyses
When analyses were performed separately for different regions, OR for bleeding risk when warfarin and tamoxifen were coprescribed within 30 days from bleeding episode was numerically higher in Stockholm-Gotland region (adjusted OR: 1.72, 95% CI: 0.80-3.72) than in Uppsala/Örebro-Northern regions (OR: 1.05, 95% CI: 0.49-2.26) but not statistically significant in either subanalyses.
A further sensitivity analysis including only patients with bleeding episodes coded as main diagnosis at hospital admission revealed a statistically significant association between bleeding and exposure to tamoxifen within 30 days (adjusted OR: 2.03, 95% CI: 1.05-3.94).
When exposure to tamoxifen was >30 days from bleeding episode, there was no significant association between the two (adjusted OR: 0.85, 95% CI: 0.11-6.58).

| D ISCUSS I ON
This is the first large-scale, population-based study investigating the clinical impact of a potential drug interaction between tamoxifen and warfarin. We found a trend toward increased risk for bleeding in warfarin-treated breast cancer patients that received tamoxifen with a magnitude of OR between 1.42 and 2.03 in different analyses.
Although no statistically significant difference was observed in most of the analyses, the results should be interpreted in the context of the potential clinical relevance. Several results in our analyses imply a potential association between bleeding risk and concomitant administration of warfarin with tamoxifen. First, the direction of a potential association was on the same side (OR > 1) in all analyses using exposure to tamoxifen within 30 days as exposure of interest.
Considering the small number of cases in analyses using tamoxifen exposure 31-180 days from bleeding episode, leading to wide confidence intervals, we could not examine whether the pattern of OR direction was only present in the short "window" of exposure to both medications, namely within 30 days or in exposure within 6 months