br We extracted variables about
We extracted variables about sociodemographic characteristics (including age, race/ethnicity, marital status, income quartile), comorbidities, medication prescriptions, chemotherapy, radio-therapy and surgical treatment for breast cancer, and survival. We used the Deyo AUY922 of Charlson’s index to assess the burden of comorbidities [13,14]. Bisphosphonates included were alendro-nate, pamidronate, ibandronate, zoledronic acid, risedronate, and etidronate. Bisphosphonate use was defined as having at least one Medicare part D claim in the period from 6 month prior to breast cancer diagnosis to the first CVD event. CVD and other comorbid-ities were identified by searching the hospital discharge summary and outpatient claims for diagnostic (International Classification of Diseases, Ninth Revision [ICD-9]) codes. Specific comorbidities that were abstracted included arterial hypertension (ICD-9 codes: 401.0e402.0), type 2 diabetes mellitus (ICD-9 codes: 250.0e250.9), hyperlipidemia (ICD-9 codes: 272.4, 272.2, 272.0), hyperthyroidism (ICD-9 codes: 242.0e242.40, 242.80e242.90), and osteoporosis (ICD-9 codes: 733.0, 733.03, 733.09) Hypertension, diabetes, hyperlipidemia and hyperthyroidism were included in the adjusted analyses as these are known risk factors for CVD. The Health Care Procedure Coding System (HCPCS) was used to identify receipt of chemotherapy, radiation therapy, and surgical treatment any time from cancer diagnosis to 36 months after diagnosis. 2.3. Study outcome
2.4. Statistical analysis
Student’s t-test, Wilcoxon signed-rank test, and chi-square were used as appropriate to compare baseline characteristics between those who were users vs. nonusers of bisphosphonates. We used propensity score methods to control for potential allocation bias since differences in patient characteristics and comorbidities may have influenced bisphosphonate prescribing. The propensity score represents the probability that a patient will receive a bisphosph-onate based on their known baseline characteristics. We calculated propensity scores using a logistic model that included patients’ sociodemographics characteristics, income quartiles, diagnosis of osteoporosis, Charlson comorbidity score, concurrent use of car-diovascular medications (e.g., angiotensin converting enzyme in-hibitor (ACE-i), aldosterone antagonist, angiotensin receptor blocker (ARB), beta-blocker and statins), cancer stage, and anti-cancer treatment (chemotherapy, surgery and radiotherapy). We used a greedy 1:1 matching algorithm to match bisphosphonate users and non-users by their propensity score. Nonusers were selected from patients with primary breast cancer who had not received any bisphosphonate therapy between January 1, 2007 and December 31, 2013. We used McNemar and matched-paired t-test to evaluate whether covariates were balanced across bisphospho-nate users vs. non-users after propensity matching. Unadjusted Kaplan-Meier curves for incident CVD events were plotted for matched patients treated with or without bisphosphonates and compared using the log-rank test.
The association between bisphosphonate use and incident CVD events was evaluated using a competing risks Cox proportional hazard regression model that censored for death from any cause. The results are presented as hazard ratios (HR) with corresponding two-sided 95% confidence intervals (CI). All data analyses were conducted using SAS software 9.4 (Cary, NC). Our study was deemed exempt by the Institutional Review Board at Icahn School of Medicine at Mount Sinai.
A total of 48,182 women with stage 0-III primary breast cancer diagnosed between 2007 and 2010 were identified from the SEER-Medicare database. Of these, 10,115 women did not have a history of CVD prior to breast cancer diagnosis, 2314 (22.8%) patients were bisphosphonate users and 7801 (77.1%) had never used bisphosphonate. After propensity score matching, 2178 bisphosphonate non-users were matched with 2178 bisphospho-nate users. The flowchart for our cohort selection is presented in Fig. 1. Bisphosphonate-exposed women had an average use of 15 months. The median age (interquartile range [IQR]) of the patients was 72 (68e78) years, in both groups of patients. After propensity score matching, there were no significant differences between the two groups by race, breast cancer stage at diagnosis, comorbidities,
Patients diagnosed with stage 0-III breast
CVD before breast cancer diagnoses
Missing cancer stage data n=839
Breast cancer survivors without