To better understand the patterns and predictors of chemotherapy use for women with HR-positive, LN-negative breast cancer, we analyzed data for 3,190 women treated at eight comprehensive cancer centers from 1997 to 2004. Although a majority of women received chemotherapy, almost half (45%) did not. The rate of chemotherapy use across all institutions increased only modestly over time (from 52% in 1997 to 1998 to 58% in 2003 to 2004). That older patients and women with larger tumors were significantly more likely to receive chemotherapy over time suggests that there may have been a transition away from age-based determinants toward tumor-specific determinants of chemotherapy use. On multivariable analysis, patient variables (eg, younger age, premenopausal status, and fewer comorbidities) and tumor characteristics (eg, larger primary cancer, absence of estrogen or progesterone receptor, high grade, lymphovascular invasion, and HER-2 overexpression) were associated with a higher likelihood of receiving chemotherapy. Institutions exhibited significantly different odds of chemotherapy use and divergent trends in use over time.
That almost half of the cohort did not receive chemotherapy is remarkable for several reasons. First, clinical trial data demonstrated that chemotherapy led to superior outcomes for women with HR-positive, LN-negative breast cancer before the patients we analyzed were treated.3
Second, practice guidelines recommended chemotherapy for HR-positive, LN-negative breast cancer during the time period included in our analysis.8,29
Third, studies questioning the benefits of chemotherapy came out largely after our analysis concluded.11,30
Fourth, our cohort excluded women unlikely to benefit from chemotherapy (ie, those > age 70 years or with tumors < 1 cm). Considering that no institution administered chemotherapy to more than 78% of women, there seemed to be some agreement that not all women with localized HR-positive breast cancer should receive adjuvant chemotherapy. One possible explanation for this willingness to selectively forego chemotherapy is expressed in a footnote to the NCCN guidelines, which states that chemotherapy confers only a small absolute benefit that some may feel is not clinically significant enough to warrant the potential risks in all patients.10
However, this does not explain how providers decided when to give and when to forego chemotherapy. Our analysis showed that institutions demonstrated significant variability with regard to which women got chemotherapy, the proportion of patients who received it, and the trend in chemotherapy use over time. These interinstitutional differences were not explained by variations in established prognostic markers. Furthermore, all institutions had access to the same randomized controlled trial data and clinical practice guidelines, which, during the period analyzed in our study, did not offer clear guidance regarding which subgroups should and should not receive chemotherapy. Perhaps non–evidence-based differences in providers’ opinions and preferences about chemotherapy contributed to the interinstitutional variability that we observed. One could argue that not administering chemotherapy to a significant subset of women with HR-positive, LN-negative breast cancer measuring more than 1 cm reflects poor quality care. However, we do not believe these data can be used to characterize institutions’ quality of care because there is no clear expert consensus regarding what should be considered optimal care. This lack of agreement makes it all the more difficult to derive clinical practice guidelines that provide straightforward evidence and consensus-based recommendations.
Because our analysis did not include data from community-based cancer centers or office-based practices, it is not possible to address the care provided outside comprehensive cancer centers. However, the rates of chemotherapy use observed in our study were notably higher than those from previous population-based studies. During the early 1990s, a Surveillance, Epidemiology, and End Results–Medicare analysis of women age 65 to 69 years with HR-positive, LN-negative breast cancer found that 4.8% received chemotherapy,31
and a study of women with HR-positive or -negative breast cancer treated in Quebec found that 37% of women with stage II breast cancer received chemotherapy.32
A more recent analysis of Surveillance, Epidemiology, and End Results data found that only 26% of women with HR-positive, LN-negative breast cancer treated in 2000 received chemotherapy.33
On the basis of these studies, community-based providers seem to be less likely to administer chemotherapy. However, some community-based providers could be more likely to administer chemotherapy if they tend to follow clinical practice guidelines or are influenced by the financial implications of administering chemotherapy. Detailed data on patterns of care in the community setting are needed but are hard to come by.
Our analysis has several limitations. First, although we did not demonstrate an association between receipt of chemotherapy and sociodemographic variables such as race/ethnicity, we had limited power to detect such a relationship because the cohort included a relatively small number of minority patients. Second, our analysis did not consider patient preferences. It is unlikely that patient preferences could account for the observed patterns of care for several reasons. First, much higher rates of chemotherapy use were observed in women with HR-negative, LN-negative breast cancer measuring more than 1 cm (91%) and in women with HR-positive, LN-positive breast cancer (92%).34
Second, patient preferences are unlikely to vary significantly from institution to institution, so they could not explain the significant interinstitutional variability in chemotherapy use that we observed. Finally, we cannot assess the degree to which individual providers influenced patterns of chemotherapy use because the NCCN does not collect provider-specific data.
Deciding whether to recommend and receive adjuvant chemotherapy for HR-positive, LN-negative breast cancer is a frequent and important decision faced by many patients and nearly every oncologist in the United States. Although several studies demonstrate that chemotherapy improves survival,3-6
others suggest that postmenopausal women and women with lower risk breast cancer may not benefit from chemotherapy.5,7,11,12
Moreover, chemotherapy may confer only a small absolute benefit and can cause serious adverse events.10
Given the modest benefits and nontrivial risks of chemotherapy, together with the lack of evidence defining which subgroups actually benefit, the patterns of care that we observed are not surprising. Interestingly, these patterns of care suggest that institutions were willing to embrace risk-adapted decision making before high-quality data supported this approach.
Recent gene expression profiling studies have identified subsets of women who have a lower risk of recurrence35-37
and do not seem to benefit from chemotherapy.30
These new tests help validate the concept of risk-adapted decision making endorsed by several practice guidelines7,8,13-15
and offer evidence-based guidance regarding who should and should not receive chemotherapy. However, the recommendations that come from these gene expression profiles do not always agree with the direction provided by clinical practice guidelines and other decision tools (eg, Adjuvant Online). However, only prospective randomized controlled trials, such as Trial Assigning Individualized Options for Treatment38
and Microarray in Node-Negative Disease trial,39
will conclusively establish the benefits and risks of chemotherapy within the different subgroups of HR-positive, LN-negative breast cancer.