Reply to M.S. Ewer et al

We appreciate Drs Ewer and Swain's¹ interest in our study on cardiovascular disease risks in women with and without breast cancer, which used Kaiser Permanente Northern California (KPNC) electronic health record data.² In their letter, Drs Ewer and Swain¹ expressed concern about the possibility of surveillance bias in our identification of incident heart failure and/or cardiomyopathy events in women with breast cancer treated with trastuzumab who did not receive an anthracycline. Drs Ewer and Swain¹ acknowledged that the higher risks of heart failure and/or cardiomyopathy observed among women with breast cancer exposed to anthracycline with or without trastuzumab is likely real; however, they expressed skepticism regarding the higher risk of heart failure and/or cardiomyopathy observed among women who received trastuzumab without an anthracycline. They questioned whether the excess risk may be due to a surveillance artifact, also known as surveillance bias. They asked whether we had data on the number of ejection fraction measures in cases and controls to assess surveillance bias because of more imaging assessment in the cases. We do not have those data available, but we are responding to their inquiry as this circumstance should not discount the reported results.

The identification of heart failure and/or cardiomyopathy in our study was based on clinical International Classification of Diseases diagnostic codes and not solely on reductions in left ventricular ejection fraction assessed via imaging. We agree that it is possible that an increased frequency of cardiac imaging surveillance in the breast cancer cases may have resulted in a higher likelihood of being diagnosed with cardiomyopathy as cardiomyopathy can be subclinical. It is unlikely that more regular clinical surveillance would result in a substantial increase in clinical heart failure diagnoses as heart failure typically has associated clinical manifestations that would lead to seeking medical care. Accordingly, even in the control group, we anticipate that individuals with heart failure would seek medical attention and be diagnosed with heart failure.

Thus, although one could argue that there could be differential misclassification of the diagnostic codes biasing the risk estimates away from the null, the findings from our small, internal validation study of heart failure events suggest otherwise.² As we previously reported, among both breast cancer cases and controls in our analytic sample, we found that of the 50 heart failure events adjudicated, the positive predictive value of confirmed cases was 91%. Moreover, our analytic approach goes beyond comparing proportion of events and uses robust Cox proportional hazards models to estimate survival by adjusting for factors identified a priori that may have a bidirectional effect on the risk of both cardiovascular disease and breast cancer. These factors include baseline health characteristics (health care utilization, body mass index, menopausal status, smoking history, history of diabetes, history of dyslipidemia, history of hypertension, and history of other comorbidities), socioeconomic status on the basis of neighborhood characteristics, and receipt of other breast cancer therapies (radiation therapy, aromatase inhibitors, and tamoxifen).

The goal of the Pathways Heart Study is to assess the risk of cardiovascular disease in women with and without a history of breast cancer, leveraging data from the KPNC electronic health record to provide risk estimates observed in real-world clinical care while accounting for potential confounding factors. Specific to the results in question, we previously reported that women with a history of breast cancer who received trastuzumab without an anthracycline had an increased hazard of 2.53 (95% CI, 1.33 to 4.81) of developing heart failure and/or cardiomyopathy compared with cases matched on year of birth, race, and ethnicity. In response to the inquiry by Drs Ewer and Swain,¹ we have run that model separately for heart failure and cardiomyopathy. For heart failure, which is less prone to surveillance bias, our results are consistent: Women who received trastuzumab without an anthracycline had a statistically significant increased hazard ratio of 2.28 (95% CI, 1.18 to 4.41) of developing clinical heart failure. For cardiomyopathy, where surveillance bias may be a concern, the risk of cardiomyopathy was not statistically significant with a hazard ratio of 5.54 (95% CI, 0.78 to 39.30); the wide confidence intervals are due to a small number of events. We stand by our conclusion that our analysis identified women with breast cancer who received trastuzumab without an anthracycline as a population at higher risk of developing cardiovascular disease, now with a particular emphasis on heart failure, and that this group of patients should receive appropriate clinical cardiology surveillance.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Reply to M.S. Ewer et al

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).