The prevalence of chemotherapy use in this cohort has decreased over time, with a marked increase in neoadjuvant treatment, consistent with other studies conducted in US populations. (13–14) Previous studies have reported that the use of 21-gene recurrence scores that inform decisions to forgo chemotherapy in lower risk patients may be driving decreases in chemotherapy use in breast cancer patients. (15–16) The average time spent undergoing chemotherapy and the average number of visits has increased. While one may assume this is due to the increasing use of anti-HER2 therapies (which have a substantially longer treatment duration than other drugs used to treat EBC), it should be noted that trastuzumab and/or pertuzumab-containing regimens have decreased in length and administration visits over time, while the number of visits has increased among those receiving cytotoxic drugs. The decrease in treatment duration for those receiving anti-HER2 therapies may reflect more treatment discontinuation secondary to toxicity or patient preference, as well as changing perceptions regarding the ideal duration of anti-HER2 therapies.

The overall increases in infusion visits and the average duration of chemotherapy represent an increasing time commitment for patients and providers, as well as a greater burden on the healthcare system for treatment delivery. These trends may also reflect improvements in symptom management that facilitate longer duration of chemotherapy, or administrative challenges in chemotherapy delivery, resulting in longer gaps between cycles. Further research is needed to better understand the drivers behind these trends.

This study provides an overview of broad chemotherapy trends in a sizeable cohort. It has several strengths, most notably a large sample size over a longitudinal 14-year period in a diverse cohort of breast cancer patients. Patients treated in integrated healthcare delivery settings have been shown to be representative of such underlying populations; therefore results are likely to be more generalizable and inform broader health services research [15]. Given this setting, results are not influenced by changes in insurance coverage. There are several limitations to consider; although this study considers chemotherapy administration visits, it does not incorporate data on laboratory services or other health services utilization which could result from toxicities or monitoring. We were unable to comment on 21-gene recurrence score uptake throughout the study period, which may have driven the decline in chemotherapy over time. Our results reflect real-world data, which has some limitations in use. For example, we identified a very small number of individuals in this large cohort classified as HER2-,who received trastuzumab and/or pertuzumab. HER2 status is largely obtained from cancer registries and extracted from diagnostic biopsy specimens. We abstracted medical chart data to address/correct any misclassified data, and thus any discordance that remains, even minimally, may reflect the complexity of obtaining and classifying HER2 status for research (e.g., HER2 may be reclassified with subsequent surgical assessment, or may be classified by the primary tumor in the setting of bilateral disease). It may also reflect research around HER2 expression in breast cancer stem cells and subsequent treatment of HER2- patients with anti-HER2 therapy [17], or the heterogeneity expected when examining a large treatment dataset.

In summary, the trends reported provide a systems-level overview of changing patterns of chemotherapy over time, showing a reduction in the prevalence of chemotherapy and increased chemotherapy visits. These data are useful to show that the treatment burden of routine chemotherapy care is changing over time, which has implications for cost, and resource allocation and administrative planning for health systems. This may inform future research to understand treatment uptake and changing patterns of care including regimen selection and how regimens are administered/received, and highlights the need to understand how these factors impact patient outcomes.