The impact of preexisting comorbidities on receipt of cancer therapy among women with Stage I–III breast cancer in the Detroit Research on Cancer Survivors cohort

Abstract Purpose Pre‐existing comorbidities play an important role in choice of cancer treatment. We retrospectively evaluated the relationship between pre‐existing comorbidities and receipt of local and systemic therapy in a cohort of Black women with Stage I–III breast cancer. Methods The study population for analysis included 1169 women with Stage I–III disease enrolled in the Detroit Research on Cancer Survivors (ROCS) cohort. Information on comorbidities, socio‐demographic, and clinical variables were obtained from self‐reported questionnaires and the cancer registry. Comorbidities were analyzed individually, and comorbidity burden was categorized as low (0–1), moderate (2–3) or high (≥4). We used logistic regression analysis to evaluate factors associated with receipt of local treatment (surgery ± radiation; N = 1156), hormonal (N = 848), and chemotherapy (N = 680). Adjusted models included variables selected a priori that were significant predictors in univariate analysis. Results Receipt of treatment was categorized into local (82.6%), hormonal (73.7%), and/or chemotherapy (79.9%). Prior history of arthritis and depression were both associated with a lower likelihood to receive local treatment, [odds ratio (OR), 95% confidence interval (CI), 0.66, 0.47–0.93, and 0.53, 0.36–0.78], respectively. Obesity was associated with higher likelihood of receiving hormonal therapy (OR: 1.64, 95% CI: 1.19, 2.26), and heart failure a lower likelihood (OR: 0.46, 95% CI: 0.23, 0.90). Older age (Ptrend <0.01) and increasing co‐morbidity burden (Ptrend = 0.02) were associated with lower likelihood of receiving chemotherapy. Conclusion History of prior co‐morbidities has a potentially detrimental influence on receipt of recommended cancer‐directed treatment among women with Stage I–III breast cancer.


| INTRODUCTION
Breast cancer is the most common cancer diagnosed among women in United States. 1 In recent years, breast cancer incidence has been rising by 0.3% per year, however mortality rates have decreased by 1.3% per year. 1,24][5] Potential implications associated with altered treatment recommendations include lower and potentially less effective chemotherapy dosing in the setting of multiple comorbidities, concern about enhanced side effects, and the potential biological interplay between other diseases and cancer which may diminish the effectiveness of cancer directed treatment.][5] We evaluated the association between self-reported, preexisting comorbidities and receipt of recommended local and systemic therapy in a cohort of Black women with Stage I-III breast cancer enrolled in the Detroit Research on Cancer Survivors (ROCS) study.With the extensive collection of data on sociodemographic, clinical and treatment information, the ROCS cohort provides a unique opportunity to study factors associated with receipt of treatment among Black cancer survivors. 6With increasing incidence and higher mortality from breast cancer in the Black community and a high comorbidity burden, 1,2,24 a better understanding of factors that may influence choice and receipt of cancer-directed treatment is crucial to understanding ongoing racial disparities in breast cancer outcomes.Other studies looked at mainly White women, such as that by Klepin et al. 13 (87% White vs. 11% Black), Parise et al. 3 with more than 50% White women, or Minicozzi et al. 4 looking at cancer registries across nine European countries, which likely has few Black participants-hence this study provides valuable insight into an exclusively Black cohort.We expect these results to be generalizable to other urban Black populations in the United States.

| Study population
The population-based ROCS study is the largest cohort study conducted exclusively among Black cancer survivors addressing multilevel determinants of poorer outcomes in this high-risk population. 6,7Briefly, Black men and women with invasive breast, lung, prostate, or colorectal cancer diagnosed since 2013, as well as endometrial cancer and other cancer sites diagnosed at less than age 50, since 2016, were identified through the Metropolitan Detroit Cancer Surveillance System (MDCSS).Survivors were sent letters inviting them to complete an enrollment survey by one of three methods; either online, by phone with a trained interviewer, or by written questionnaire.
As of November 20, 2020, 1408 breast cancer survivors were enrolled in the study.The current study is based on the first 1280 female breast cancer survivors who had complete data available for analysis.We excluded women diagnosed with American Joint Committee on Cancer (AJCC) 8th Edition stage 0, IV, or unknown stage (N = 105), and those who did not complete the medical history section of the questionnaire (N = 6), leaving 1169 women with AJCC Stage I-III breast cancer that were included in the final analysis.In regards to the methodology of survey completion, the survey was completed by 266 (22.8%) women online, 476 (40.7%) were interviewer-assisted, and 427 (36.5%) were completed by written questionnaire.The median time from cancer diagnosis to study enrollment in the ROCS cohort was 21 months (Interquartile range: 13-35 months).
We evaluated three categories of cancer treatment identified through the MDCSS including localized treatment (surgery +/− radiation therapy), hormonal therapy, and chemotherapy.The study population for each category of treatment was based on the availability of treatment data and on NCCN recommendations for treatment.For localized treatment we included 1156 women after excluding 13 who had unknown information on receipt of radiation.For hormone therapy, we excluded 278 women for which hormone therapy was not recommended based on National Comprehensive Cancer Network (NCCN) guidelines 8 as well as those with unknown hormone receptor status (HR) or missing information on receipt of hormone therapy (N = 43), resulting in 848 women included in this analysis.For chemotherapy, we excluded survivors with unknown HR/HER2 status and/or unknown lymph node status (N = 44), HR-positive, human epidermal growth factor receptor 2 (HER2)-negative breast cancer with unknown Oncotype DX scores (N = 270), and survivors who were not recommended chemotherapy as per NCCN criteria (mentioned in detail below) (HR-positive, HER2negative breast cancer; N = 175), leaving a sample of 680 women for the chemotherapy analysis.(Figure S1).

| Study measures
Self-reported characteristics at enrollment included gender, education, marital status, annual household income, health insurance coverage, smoking status, and participation in moderate or vigorous physical activity in the past 4 weeks.Participants were asked if a doctor ever told them that they had any of the following medical conditions including their corresponding age at onset: arthritis, emphysema/COPD, depression, diabetes, hepatitis, high cholesterol, hypertension, stroke, thyroid problem, heart attack, congestive heart failure (CHF), atrial fibrillation, or coronary artery disease.Participants were considered to have a pre-existing comorbid condition at the time of cancer diagnosis if they answered in the affirmative to the specified condition, and the age at onset was less than the age at cancer diagnosis.In addition, if the age of onset was unknown, it was assumed that the diagnosis occurred prior to the cancer diagnosis as the patients were likely unable to recall a diagnosis made several years ago.Obesity was defined as a body mass index (BMI; body weight in kg/height in meters squared) of 30 or more calculated from self-reported height and weight 1 year before cancer diagnosis.Our data set did not include any information on Eastern Cooperative Oncology Group (ECOG) performance status which is derived by clinicians in the office.Instead we hypothesized that self-reported ability to perform moderate or vigorous physical activity (listed in Table 1) would potentially be associated with whether or not physicians were willing to recommend standard of care cancer treatment.
Comorbidity burden was based on count of co-morbid conditions, with 0-1 conditions categorized as "low," 2-3 as "moderate," and 4 or higher as "high."The categorization of conditions as outlined was based on the available numbers in each category.Other cancer-specific data were obtained from the MDCSS registry database and included age at diagnosis, diagnosis year, AJCC stage, HR/HER2 status, Oncotype DX scores, and first-course treatment: surgery (none, lumpectomy, mastectomy), radiation, hormone therapy, and chemotherapy.Receipt of radiation, hormonal therapy, or chemotherapy were classified as yes versus no.
Localized treatment, hormonal therapy, and chemotherapy were considered recommended if they met standard treatment guidelines including NCCN guidelines. 8ocalized treatment with surgery +/− radiation was considered "recommended" for women who had lumpectomy and radiation for any stage of disease, mastectomy with or without radiation for T1-T2 stage and lymph node-negative disease, and mastectomy with radiation for T3-T4 stage or lymph node-positive disease.Hormone therapy was considered "recommended" for survivors with HR-positive, HER2-positive, or HR-positive, HER2-negative disease.Chemotherapy was considered recommended for survivors who had the following tumor characteristics: HRpositive, HER2-positive or HR-negative, HER2-positive or

| Statistical analysis
All analyses were conducted using SAS software (version 9.4; SAS, Cary, NC) and an alpha value of 0.05 was set to determine statistical significance.The distribution of selected participant characteristics and cancer-related factors were summarized using counts and percentages in Tables 1 and 2 and distributions by receipt of recommended treatments using counts and percentages in Tables 3-5.Logistic regression models were used to estimate odds ratios (OR) and 95% confidence intervals (CI) to quantify associations for age at diagnosis, comorbidity burden, pre-existing individual comorbidities, and receipt of recommended treatment.Adjusted models included variables that were selected a priori and were significant predictors in univariate models using categories presented in Tables 1 and 2. The analysis for receipt of localized treatment was adjusted for marital status, annual household income, AJCC stage, and diagnosis year; the analysis for receipt of hormone therapy adjusted for survey method at enrollment and diagnosis year, and the analysis for receipt of chemotherapy was adjusted for AJCC stage, diagnosis year, and HR status.We assessed trends by including ordinal variables in the models for age and comorbidity.Table 2 shows the distribution of first-line local treatment received (derived from the MDCSS) showing that 96% of the women had breast surgery and 71% had radiation therapy.The proportion of women overall who received cancer-directed treatment included 82.6% who received localized treatment, 73.7% hormonal treatment, and 79.9% chemotherapy.

| RESULTS
Table 3 shows the un-adjusted and adjusted models evaluating the association between age at diagnosis and self-reported history of prior co-morbidities and receipt of recommended localized treatment.In the multivariable adjusted models, there was no significant relationship between age (P trend = 0.17) or comorbidity burden (P trend = 0.29) and receipt of localized treatment, however history of arthritis and history of depression were both associated with a lower chance of receiving localized treatment (OR: 0.66, 95% CI: 0.47, 0.93 and OR: 0.53, 95% CI: 0.36, 0.78), respectively; there was no significant relationship between any of the other co-morbid conditions and receipt of localized therapy.
Table 4 shows the model for receipt of recommended hormone therapy, and Table 5 for receipt of chemotherapy.After multivariable analyses, history of obesity was associated with a greater likelihood for receipt of hormonal therapy (OR: 1.64, 95% CI: 1.19, 2.26) and history of CHF was associated with a lower likelihood of receipt of hormonal therapy (OR: 0.46, 95% CI: 0.23, 0.90).For chemotherapy, after multivariable analysis, increasing age and increasing comorbidity burden were both associated with a lower likelihood of receipt of chemotherapy, (P trend <0.01) and (P trend = 0.02), respectively.In addition, women with a history of arthritis (OR: 0.66, 95% CI: 0.44, 0.99) and a history of hypercholesterolemia (OR: 0.62, 95% CI: 0.41, 0.95) were also less likely to receive chemotherapy.None of the other self-reported co-morbid conditions were associated with receipt of hormonal therapy or chemotherapy.

| DISCUSSION
We evaluated the relationship between pre-existing comorbidities and receipt of recommended cancer-directed treatment in a cohort of Black survivors with early-stage invasive breast cancer in an attempt to better understand factors associated with receipt of less than adequate cancer-directed therapy.As expected, the majority of women in the ROCS cohort received recommended standard treatments, however there was a mixed relationship regarding receipt of different therapy modalities based on prior medical history.In ROCS, women with a history of either depression or arthritis were less likely to received standard localized therapy, women with obesity were more likely to receive hormonal therapy, and women with CHF were less likely to receive hormonal therapy.Negative predictors for receipt of chemotherapy included increasing age and increasing number of comorbid conditions, as well as a personal history of arthritis or hypercholesterolemia.4][5] In one report of 20,177 women with triple negative breast cancer from the California Cancer registry (2000-2015), 3 women with high co-morbidity burden defined by a Charlson comorbidity index (CCI) of 2 or greater versus CCI-0 9 had higher breast-cancer specific mortality.In addition, women with high CCI with any stage at diagnosis, were less likely to receive chemotherapy, had reduced use of lumpectomy + radiation (Stages 2-4) and reduced rate of mastectomy for Stage 3 disease. 3lso noted in the California Cancer Registry study, the proportion of Black women with CCI 2+ was almost twice as high as those with CCI of 0 (20.4% vs. 10.7%) as opposed to a similar proportion of White women with CCI 0 or CCI of 2 (51.2% vs. 57.6%). 3n the ROCS cohort, as data needed to construct the CCI variable was not collected, we used a count of prior co-morbid conditions as a proxy for CCI.Despite the lack of data on CCI, our results are consistent with those reported by other studies which used CCI. 3,4In a cohort of elderly breast cancer survivors, Parise et al., 3 demonstrated lower receipt of chemotherapy among older patients and those with a higher co-morbidity burden.Garg et al. also reported that older age and increased co-morbidity burden was associated with an increased likelihood for treatment stoppage. 10Others also reported less use of axillary lymph node dissection and radiation therapy in older women. 11In a similar analysis from nine European countries, older women were more likely to receive delayed and non-standard treatments for their breast cancer and less likely to be offered breast conservation surgery and radiation. 4Similar findings showing negative impact of age and increasing comorbidity burden on receipt of standard cancer treatments, and consequently on outcomes, were replicated in many other analyses [14][15][16][17][18][19][20][21][22][23] including in a metaanalysis of 60 studies. 5otential explanations for the reported lower use of cancer-directed therapy among individuals who had more co-morbid conditions include the potential for physician concern or perception that older age and greater co-morbidity burden is more likely to result in potentially dangerous and/or life-threatening side effects from chemotherapy. 12It is also possible that patients who have multiple co-morbid medical conditions have a higher likelihood to refuse treatment.Our data provides no specific explanations for why any particular treatment such as history of arthritis or history of hypercholesteremia might influence the recommendation or completion of chemotherapy, although again history of any co-morbid medical condition might raise both physician and patient related concerns for added toxicity on top of the usual side effects associated with cancer treatment and thereby influence treatment decision making. 13Choice of cancer treatment is a complicated process taking into account numerous clinical factors, physician perception of a patient's ability to tolerate treatment and other aspects including perceived patient compliance and willingness to take treatment.The consistency of our findings with those of other studies however, suggests that the presence of co-morbidities among women with early stage breast cancer might influence recommendations for cancer-directed treatment, and emphasizes the importance of collaboration with primary care to optimize treatment of other medical conditions among women with breast cancer.Better treatment and control of co-morbid medical conditions could potentially positively impact the optimal care of women with early stage breast cancer.
15][16][17][18][19][20][21][22][23] An analysis in a Black survivorship cohort is an important addition to the literature as other studies only included 10%-20% of Black women. 3,13The importance of ROCS is also exemplified by the fact that Black cancer survivors have been shown to have worse outcomes after cancer diagnosis than other groups. 24It is also important to study the relationship between prior co-morbid medical conditions and treatment in a Black survivorship cohort given that Blacks generally have higher rates of co-morbid conditions than other groups. 24Lastly, the ROCS data set provides a comprehensive listing of socio-demographic and clinical information derived through linkage with the MDCSS cancer registry database.
Limitations of our analysis include the retrospective data collection and possibility for recall bias particularly in regard to reports of co-morbid medical conditions.In addition, we did not have information on performance status T A B L E 5 (Continued)  from a patient survey or a cancer registry which is commonly used by physicians to assess whether or not patients should get systemic chemotherapy.We felt however that the methods used to collect clinical information in ROCS can be justified in that it roughly approximates data collected from patients in the medical office, and we also used information on cancer diagnoses and treatment available in a population based cancer registry.There may however be misclassification and possibly under-reporting of hormonal and chemotherapy in the MDCSS as that treatment is largely administered in the outpatient setting and not completely captured in the registry.Despite the fact that we did not collect all of the information necessary to construct a CCI, our results mimic those of other studies where CCI was used as a measure of the impact of co-morbid medical conditions, 3 justifying our analytic decisions.As indicated above, choice of cancer-directed treatment is a complicated process involving medical providers, patients and care-givers and a direct cause-and effect relationship between any of the studied variables including medical history and rationale for treatment recommendation or whether treatment was completed cannot be made.Other limitations include a possible survivor bias in that healthier cancer survivors are more likely to participate in ROCS, although the focus of our analysis on women with early stage breast cancer lessens this possibility.In summary, in the ROCS cohort, age and medical history were associated with receipt of various forms of cancer-directed therapy.These findings suggest that collaboration with primary care providers to optimize the medical care of co-morbidities in women with early breast cancer could favorably impact the completion of cancerdirected treatment.

T A B L E 1
Socio-demographic and clinical characteristics of women with early stage breast cancer in the Detroit Research on Cancer Survivors cohort.
triple negative disease; HR-positive, HER2-negative, and lymph node-positive disease, 50 years of age or younger at diagnosis with HR-positive, HER2-negative, lymph nodenegative disease and Oncotype DX score of 26 or higher; and women older than 50 years at diagnosis with HRpositive, HER2-negative, lymph node-negative disease and an Oncotype DX score of 31 or higher.
Cancer directed therapy received by women with early stage breast cancer in the Detroit Research on Cancer Survivors cohort.The relationship between age at diagnosis and history of co-morbid conditions and the receipt of recommended localized treatment in the Detroit Research on Cancer Survivors early stage breast cancer cohort.
Table1shows the demographic and clinical characteristics of women in the study cohort that was derived from participant self-report.The mean age at diagnosis was 57 years, Standard deviation (SD) 11.2 and interquartileN %Note: Values were not reported or were unknown for the following: education (n = 17), marital status (n = 6), income (n = 89), insurance (n = 33), physical activity (n = 7), smoking status (n = 8).Abbreviation: SD, standard deviation.TA B L E 1 (Continued) T A B L E 2Note: Values were not reported or were unknown for the following: radiation (n = 13), hormone therapy (n = 1), local/regional treatment per recommendation (n = 13), hormone therapy per recommendation (n = 321), chemotherapy treatment per recommendation (n = 489).All treatment recommendations per NCCN guidelines and best clinical practice.Abbreviations: AJCC, American Joint Committee on Cancer, 8th Edition; SD, standard deviation.a Recommended local/regional treatment (Surgery ± radiation) = lumpectomy + radiation for any stage of disease, mastectomy ± radiation for T1-T2 stage and lymph node-negative disease, and mastectomy + radiation for T3-T4 stage or lymph node-positive disease.b Recommended hormonal therapy = HR-positive, HER2-positive, or HRpositive, HER2-negative disease.c Recommended chemotherapy = HR-positive, HER2-positive or HR-T A B L E 3

Receipt of recommended localized treatment a
Associations between age at diagnosis and co-morbid conditions and recommended hormone therapy for participants in the ROCS early stage breast cancer cohort.
Note: Values were not reported or were unknown for the following: obesity (n = 26), myocardial infarction (n = 12), congestive heart failure (n = 12), atrial fibrillation (n = 12), coronary artery disease (n = 12).Abbreviations: CI, confidence interval; OR, odds radio.aCases with known radiation status.bAdjusted models control for marital status, annual household income, AJCC stage, and diagnosis year.T A B L E 3 (Continued) T A B L E 4

Receipt of recommended hormone therapy a
Values were not reported or were unknown for the following: obesity (n = 19), myocardial infarction (n = 10), congestive heart failure (n = 10), atrial fibrillation (n = 10), coronary artery disease (n = 10).Adjusted models control for survey method at enrollment and diagnosis year.Associations between age at diagnosis and co-morbid conditions and recommended chemotherapy for participants in the ROCS early stage invasive breast cancer cohort.
Note:a Cases who are recommended to receive hormone therapy with known hormone receptor and hormone therapy status.b
Note:Abbreviations: CI, confidence interval; OR, odds radio.a Cases who are recommended chemotherapy with known hormone receptor status, lymph node status, and Oncotype DX score.b Adjusted models control for AJCC stage, diagnosis year, and hormone receptor status.