This is the first study to describe patterns of chemotherapy refusal among healthy older women with high-RS breast cancer, and the association of chemotherapy refusal with survival. Among women with high genomic risk ER-positive breast cancer, chemotherapy refusal increased with increasing age. Chemotherapy refusal was associated with decreased 5-year OS in women aged 65–79, but did not impact 5-year OS in women aged ≥ 80 years.
Previous retrospective studies have evaluated the use of RS in older women, finding that for all RS (low, intermediate, or high), older patients (≥ 50 years) were less likely to be recommended or receive chemotherapy than younger patients with similar RS (< 50 years) [Williams27]; that older women (> 70 years) with high RS who received chemotherapy did not have a significant benefit in OS [Kizy28]; and that most deaths in older women with high RS breast cancer were due to the presence of competing risks factors [Zhou29]. However, each of these studies included women with comorbidities, which, in addition to age, impact the choice to use chemotherapy.
We expand this body of knowledge by illustrating the characteristics of healthy older women with high RS who refused chemotherapy, and by describing the association of chemotherapy refusal in this population with survival. Our study is unique in that it in part eliminates the confounding variable of poor health from the survival equation, creating a cohort of patients who, other than their age, should be fair candidates for adjuvant chemotherapy. To our knowledge, these data have not been previously reported elsewhere.
Chemotherapy refusal stems from multifactorial causes. Among breast cancer patients, those who refuse chemotherapy tend to be older in age [Schwedhelm19], to distrust the healthcare system [Dean16], to lack social support, [Puts30], and to not have access to transportation [Goodwin15]. Additionally, breast cancer patients who underwent genomic RS testing were more likely to refuse chemotherapy when cared for by oncologists with high-volume practices [Schwedhelm19]. While most oncologists understand that social determinants of health negatively impact patient outcomes, they often do not have time to deconstruct barriers to care during visits [Zettler31]. This doubtless impacts patients across cancer subtypes, as patients who are more likely to refuse treatment are non-White, unmarried, or have government insurance/are uninsured [Dias10]. Furthermore, unaccounted for social factors may contribute to poor patient outcomes among those who refuse chemotherapy [Coughlin32]. Other reasons cited by older patients when declining cancer treatment included fear of side effects, self-perception of being a high-risk patient due to age or comorbidities, and concern that treatment would result in becoming a burden to others [Puts29].
The chemotherapy-related concerns cited by older patients are borne out in the literature. The risk of chemotherapy toxicity is elevated in older adults, relative to younger adults [Muss33, Jones34]. In an observational study of patients older than 65 years of age, low-grade toxicities (grades 1 and 2) were found to be associated with modifications to chemotherapy regimen, or early discontinuation of chemotherapy 55% of the time [Kalsi35]. Frailty, distinct from comorbidities, is present in up to 40% of patients older than 80 years of age [Fried36], with many components of frailty associated with susceptibility to high-grade toxicity (grades 3 and 4) in older patients [Hurria37]. Unfortunately, studies examining chemotherapy in older adults tend to disproportionately include fit patients who may not represent the average older patient [Kalsi35, Wildiers38].
Competing causes of death in the older adult population are significant. In a matched cohort study using the Surveillance Epidemiology and End Results Program (SEER) database, women older than 80 with stage I or II breast cancer were more likely to die from cardiovascular disease than they were to die from breast cancer [Schonberg39]. At the same time, the 5-year breast cancer-specific mortality of women older than 80 ranges from 11% for stage I disease to 63% for stage III/IV disease [Schonberg39]. Older patients are as willing as younger patients to accept the toxicities of therapy, yet they differ from younger patients in their willingness to trade survival for quality of life [Yellen40]. The decision to pursue therapy, therefore, requires prudent shared decision-making to mitigate the risks of toxicity, while bearing in mind competing causes of mortality and patient goals.
If a patient would prefer to avoid chemotherapy toxicity in favor of preserved quality of life, RS testing in such a scenario may be superfluous. In previous work by our group, 8% of patients older than 70 years old underwent RS testing at facilities captured in SEER [Kizy28] While this is not a high percentage, this test could be avoided in some patients by a thorough discussion of risks, benefits, and patient preferences prior to genomic testing. Furthermore, by using frailty indices to evaluate appropriateness for chemotherapy, physicians could eliminate unnecessary genomic testing in the oldest of the old [Mandelblatt41]. The use of comprehensive geriatric assessments could bolster treatment decisions, support referrals to “prehabilitation,” and inform RS testing, with the goal of tailoring workup and treatment to the needs and preferences of our oldest patients [O’Connor14].
The Choosing Wisely Campaign advocates similar decision-making strategies [Choosing Wisely42]. They suggest that specialized tumor genome testing should not be done if the test results would not change a patient’s choice of treatment. However, even if prospective studies demonstrate that genomic risk testing in older women has limited benefit, de-implementation of an accepted practice is complex [Niven43]. Moving away from engrained practices is subject to the vagaries of human decision-making, which inconsistently prioritize scientific rationale, patient autonomy, and social pressures [Smith44].
Our work is not without limitations. Regarding our data, we lack granular detail on chemotherapy agents used, patient frailty not captured by CCI, the specific reasons why patients refused chemotherapy, and the use of OS rather than cancer-specific survival as our endpoint. These are limitations inherent to the use of the NCDB. Changes in clinical practice occurred immediately following our study time frame – TAILORx was published, adjusting the cutoff for high-risk RS from the original Oncotype DX manufacturer cutoff of 31 to the currently cutoff of 26 [Sparano45]. We chose to analyze data based on the validated high-risk RS cutoff of 26, given that this is used daily in clinical practice. From an analysis standpoint, there was a relatively high prevalence of chemotherapy refusal in our study population. Our logistic regression model is therefore likely to overstate strengths of association on a relative scale; to evaluate this, we used a Poisson model with robust variance estimation as part of our sensitivity analysis. The magnitude and direction of our results were unchanged. Additionally, small sample size in the cohort of women ≥ 80 years old could introduce flaws [Peterson46]. Finally, the NCDB is inherently biased toward standards set by the CoC, since centers contributing to the NCDB are CoC accredited. Therefore, findings from the NCDB may not exactly represent practice at non-CoC centers across the United States. Despite these limitations, our study is sound. This large national dataset captures a wide range of clinical practice, underscoring the real-life behavior of older patients with high genomic risk breast cancer.