br Material and Methods br

Material and Methods
Patients
This is a retrospective case-control study based on elec-tronic chart review of consecutive women with HER2þ breast cancer treated with trastuzumab-based therapy at Princess Margaret Cancer Center (Toronto, ON) between 2002 and 2013. Patients were included if they received a pretherapy multigated acquisition (MUGA) scan and 2 subsequent follow-up scans during the course of their treatment. We identified patients who were receiving any statin (regardless of clinical indication) before and during cancer treatment. Each statin-treated patient was randomly matched with 2 patients of the same age ( 2 years) and anthracycline exposure status but without statin treatment before or during cancer treat-ment. The LVEF data were not available at the time of matching. The study complies with the Declaration of Helsinki; it was approved by the University Health Network (UHN) Research Ethics Board; and, given its retrospective nature, the Research Ethics Board waived the need for informed consent. 
Through electronic patient records, baseline characteristics were collected, including cardiovascular risk factors, cardiac disease history, cardiovascular medications, cancer-related variables (disease stage, estrogen-receptor status, progesterone-receptor status) and cancer-treatment history. For each patient, we calculated the cardiac risk score generated from the National Surgical Adjuvant Breast and Bowel Project (NSABP) B-31 study.10 Trastuzumab exposure time was defined as the Pronase E (months) between the first and the last trastuzumab dose in each patient. Cardiac follow-up was defined as the time (months) between the first and the last MUGA scans.
Outcomes
The primary outcome was the final LVEF, defined as the last LVEF value within follow-up period closest to the last trastuzumab treatment. We also assessed the following secondary outcomes: (1) change in LVEF, defined as the difference between the final LVEF and the LVEF on the pretreatment scans (baseline LVEF); (2) incidence of car-diotoxicity, defined as the proportion of patients meeting the Cardiac Review and Evaluation Committee (CREC) defini-tion for cardiotoxicity (LVEF decline 10% to < 55% without symptoms of heart failure or 5% drop to < 55% with symptoms11) in at least 1 MUGA scan; and (3) the incidence of trastuzumab interruption (at least 1 cycle inter-ruption attributed to LVEF reduction). The median number of MUGA scans between groups was compared to assess for ascertainment bias.
Statistical analysis
Continuous variables are reported as mean standard deviation (SD) or median and interquartile range (IQR) as appropriate. Two sample Students’ t-tests or ManneWhitney U test were used to compare the means of continuous variables between patients and controls based on the data distribution. Categorical variables were expressed as
Calvillo-Argüelles et al. Statins and Trastuzumab Cardiotoxicity
frequencies and percentages and were analyzed using Fishers exact test. Wilcoxon signed-rank test was used to compare the change in LVEF (as defined above) within each group. Analysis of covariance (ANCOVA) was used to assess the relationship between statin exposure status and the primary outcome after adjustment for the baseline LVEF and the following covariates: age, body mass index (BMI), cardiovas-cular risk factors (diabetes, hypertension, coronary artery disease), cardiovascular medications (angiotensin-converting enzyme inhibitors [ACEi], angiotensin II receptor blockers [ARBs], b-blockers), cancer stage (early or metastasic), and anthracycline therapy. Given the established atherosclerosis-related benefits of statins in patients with coronary artery disease or diabetes, we included interaction terms for diabetes and coronary artery disease with statins as covariates in the ANCOVA models. A logistic regression model was con-structed using cardiotoxicity (as defined above) as the outcome, treatment with statins as the predictor, and anthracycline exposure, number of cardiovascular risk factors (diabetes, hypertension, coronary artery disease, smoking), and the NSABP-31 cardiac risk score10 as confounders. An-alyses were performed using SPSS v.20 (IBM Corp, Armonk, NY). Statistical tests were 2-sided, and statistical significance was defined as P < 0.05.