Association Between Complementary and Alternative Medicine Use and Breast Cancer Chemotherapy Initiation: The Breast Cancer Quality of Care (BQUAL) Study

Importance  Not all women initiate clinically indicated breast cancer adjuvant treatment. It is important for clinicians to identify women at risk for noninitiation.

Objective  To determine whether complementary and alternative medicine (CAM) use is associated with decreased breast cancer chemotherapy initiation.

Design, Setting, and Participants  In this multisite prospective cohort study (the Breast Cancer Quality of Care [BQUAL] study) designed to examine predictors of breast cancer treatment initiation and adherence, 685 women younger than 70 years with nonmetastatic invasive breast cancer were recruited from Columbia University Medical Center, Kaiser Permanente Northern California, and Henry Ford Health System and enrolled between May 2006 and July 31, 2010. Overall, 306 patients (45%) were clinically indicated to receive chemotherapy per National Comprehensive Cancer Network guidelines. Participants were followed for up to 12 months.

Exposures  Baseline interviews assessed current use of 5 CAM modalities (vitamins and/or minerals, herbs and/or botanicals, other natural products, mind-body self-practice, mind-body practitioner-based practice). CAM use definitions included any use, dietary supplement use, mind-body use, and a CAM index summing the 5 modalities.

Main Outcomes and Measures  Chemotherapy initiation was assessed via self-report up to 12 months after baseline. Multivariable logistic regression models examined a priori hypotheses testing whether CAM use was associated with chemotherapy initiation, adjusting for demographic and clinical covariates, and delineating groups by age and chemotherapy indication.

Results  A cohort of 685 women younger than 70 years (mean age, 59 years; median age, 59 years) with nonmetastatic invasive breast cancer were recruited and followed for up to 12 months to examine predictors of breast cancer treatment initiation. Baseline CAM use was reported by 598 women (87%). Chemotherapy was initiated by 272 women (89%) for whom chemotherapy was indicated, compared with 135 women (36%) for whom chemotherapy was discretionary. Among women for whom chemotherapy was indicated, dietary supplement users and women with high CAM index scores were less likely than nonusers to initiate chemotherapy (odds ratio [OR], 0.16; 95% CI, 0.03-0.51; and OR per unit, 0.64; 95% CI, 0.46-0.87, respectively). Use of mind-body practices was not related to chemotherapy initiation (OR, 1.45; 95% CI, 0.57-3.59). There was no association between CAM use and chemotherapy initiation among women for whom chemotherapy was discretionary.

Conclusions and Relevance  CAM use was high among patients with early-stage breast cancer enrolled in a multisite prospective cohort study. Current dietary supplement use and higher number of CAM modalities used but not mind-body practices were associated with decreased initiation of clinically indicated chemotherapy. Oncologists should consider discussing CAM with their patients during the chemotherapy decision-making process.

Despite the survival benefits associated with adjuvant treatment for breast cancer, not all women initiate treatment.¹ Psychosocial factors, belief systems, and clinical, demographic, and health care provider characteristics are associated with chemotherapy initiation among patients with breast cancer.^2-5 High cancer-related anxiety, optimism, negative experiences with physicians, belief in the value of alternative medicine, concerns about risks associated with chemotherapy, and lack of knowledge of breast cancer treatment protocols have been reported as characteristics of patients with breast cancer who do not receive indicated chemotherapy.^6-8 The same set of patient characteristics also tends to be associated with high complementary and alternative medicine (CAM) use. CAM use among patients with breast cancer has been increasing over the past 2 decades^9-11 but few studies have evaluated how CAM use affects decisions regarding chemotherapy.^8,12

Studies often define CAM using a broad, nonspecific definition that includes any therapies or approaches that are a complement or alternative to conventional medical therapies.^13,14 However, when defining clinical populations using CAM, it may be more useful to consider CAM using specific subcategories, such as dietary supplements and mind-body practices. Dietary supplements include vitamins, minerals, botanicals, and other natural products. Mind-body practices include practices such as yoga, meditation, qi gong, acupuncture, and massage. In the oncology setting, patients may use both dietary supplements and mind-body practices to relieve symptoms, for general health and wellness promotion and to increase their sense of hope and control, though the evidence base for mind-body practices is stronger than for dietary supplement use.^15-17 Patient motivations and the benefits they hope to obtain may also differ between these 2 categories of CAM. It has been hypothesized that CAM use may be a risk factor for not initiating standard oncology treatments because patients may be exploring other alternative treatment approaches or because conventional oncology treatments are not congruent with patient belief systems.^8,12 However, prior studies on CAM use among patients with breast cancer have not differentiated whether standard oncology treatments were indicated based on clinical characteristics of patients.

The Breast Cancer Quality of Care Study (BQUAL) is a prospective cohort study examining demographic, behavioral, biomedical, and psychosocial predictors of initiation and adherence to adjuvant therapy among early stage breast cancer patients. Prior to initiation of study enrollment, we inserted a baseline questionnaire assessing current CAM use to test the a priori hypothesis that CAM use is associated with chemotherapy initiation after controlling for demographic and clinical covariates, and herein we present those results.

Box Section Ref ID

The BQUAL study was a large, multicenter prospective cohort study of women with nonmetastatic invasive breast cancer.¹⁸ Participants were recruited from May 2006 to July 31, 2010, from 3 sites: Columbia University Medical Center (CUMC), Kaiser Permanente Northern California (KPNC), and Henry Ford Health System (HFHS). Women were eligible if they were 20 years or older, recently diagnosed with invasive stage I to III breast cancer, and either 3 or more months from diagnosis or had not begun their third cycle of chemotherapy. The goal was to enroll women prior to hormone therapy initiation. Women were excluded if they did not speak English, were previously diagnosed with cancer excluding nonmelanoma skin cancer, did not have a telephone, and/or had documented or self-reported serious memory deficits. All participants provided written informed consent. All study procedures were approved by the institutional review boards at CUMC, KPNC, and HFHS, and the US Army Medical Research and Material Command Office of Research Protections and Human Research Protection Office.

A total of 1158 women (mean age, 59 years; median age, 59 years) enrolled in the BQUAL study and completed a baseline interview. For this analysis, we excluded women for whom chemotherapy was contraindicated based upon National Comprehensive Cancer Network (NCCN) criteria¹⁹ (n = 89), who had missing medical records (n = 2) or chemotherapy eligibility data (n = 41), had stage IV recurrent or metastatic disease (n = 10), did not provide data on CAM use at baseline (n = 123), or had unknown or missing demographic and clinical characteristics (n = 109). Women 70 years and older (n = 191) were excluded as this group was not indicated to receive chemotherapy according to NCCN criteria and other competing factors may influence chemotherapy decision making in this groups that cannot be accounted for in analysis. The final sample size was 685.

Subject sociodemographic, comorbidities, and CAM use data were collected via self-report in baseline interviews. Breast cancer treatment data were collected via self-report during telephone interviews at baseline, 4 to 8 weeks and up to 12 months after baseline, and via medical record reviews. Medical record reviews were used to define tumor characteristics.

At baseline, participants were asked whether they were currently using 7 different types of CAM modalities. Dichotomous (yes/no) variables were created to indicate current use of dietary supplements (including vitamins, minerals, herbal and/or botanical supplements, and other over-the-counter natural products) and mind-body practices (including self-practices such as yoga and practitioner-based approaches such as acupuncture). We computed a CAM use index summarizing the number of CAM types used (range 0 to 5) including vitamin and/or mineral supplements, herbal and/or botanical supplements, other over-the-counter natural products (eg, fish oil, glucosamine, melatonin), mind-body self-practices (eg, yoga and meditation) and practitioner-based mind-body practices (eg, acupuncture, massage, reiki). For this analysis, special diets and multivitamins were excluded from the definition of CAM, as these are very commonly used and often are not categorized as CAM.²⁰

Chemotherapy initiation (yes/no) up to 12 months after baseline was ascertained via self-report and medical record review. Prior publications on chemotherapy initiation have reported 98% to 100% concordance between self-report and medical record abstraction.^21-24 The 2006 NCCN breast cancer treatment guidelines¹⁹ (eTable 1 in the Supplement) were used to determine whether chemotherapy was clinically indicated or discretionary based on age, ER and PR status, ERBB2 (formerly HER2 or HER2/neu) status, tumor size, and nodal involvement.

Sociodemographic and clinical characteristics were examined as covariates and potential confounders. Sociodemographic characteristics included age, race, education, household income, employment status, and marital status. Breast cancer and other clinical characteristics include year of diagnosis, American Joint Committee on Cancer (AJCC) stage, grade, nodal status, tumor size, ER and PR status, ERBB2 status, and Charlson comorbidity score.²⁵

Separate unadjusted and multivariable logistic regression models evaluated associations between current use of any CAM (yes/no), current dietary supplement use (yes/no), current mind-body use (yes/no), and number of CAM modalities currently used (range 0-5) and subsequent chemotherapy initiation. All statistical tests were 2-sided with a significance level of .05. Demographic and clinical factors were tested as potential confounders using the change-in-estimate approach,²⁶ in which confounders were identified as factors that modify the β coefficient of the exposure-outcome association by 10%. Race/ethnicity and Charlson comorbidity index were identified a priori as potential confounders. The final covariates included in models were age, race/ethnicity, education, grade, tumor size, ER and PR status, ERBB2 status, and Charlson comorbidity index.

Because one-third of participants joined the study after initiating chemotherapy, we conducted sensitivity analyses to examine whether observed associations differed when analyses were restricted to women who did not initiate chemotherapy prior to the baseline interview. To evaluate the association between vitamin and/or mineral use and chemotherapy initiation, sensitivity analyses were conducted by restricting the definition of dietary supplement use to vitamin and/or mineral use. To account for potential bias introduced by the clustering of data by recruitment site, we performed additional sensitivity analyses using generalized estimating equation models clustering by site²⁷ with the exchangeable working correlation matrix and logit link function.

All analyses were conducted using R statistical software, version 3.2.2 (R Foundation).²⁸

Summaries of participant characteristics by chemotherapy indication group and chemotherapy initiation can be found in eTable 2 of the Supplement. Overall, 306 women (45%) were clinically indicated to receive chemotherapy based on NCCN guidelines; the remaining women were considered to have a discretionary recommendation for chemotherapy. By 12 months following baseline, 272 women (89%) for whom chemotherapy was clinically indicated initiated treatment, while the chemotherapy discretionary group had a much lower initiation rate (n = 135 [36%]).

At baseline, the prevalence of current CAM use was high (Table 1). Among all participants, 598 women (87%) reported using any CAM at baseline, and use of dietary supplements and mind-body practices were common (n = 482 [70%] and n = 483 [71%], respectively). The median number of CAM modalities used was 2 with 261 women (38%) reporting currently using 3 or more modalities. Compared with nonusers, dietary supplement and mind-body users tended to have higher education and been recruited from KPNC (eTable 3 in the Supplement); there were few other differences across groups.

In unadjusted analyses, among women for whom chemotherapy was indicated, women who did not initiate chemotherapy were more likely to report current dietary supplement use compared with women who did initiate chemotherapy (30 of 34 [88%] vs 169 of 272 [62%]; P = .002) (Table 1). Of the specific categories of dietary supplements, vitamin and/or mineral use was the strongest correlate of not initiating chemotherapy, with 24 of 34 (71%) of the noninitiators using vitamins and/or minerals, compared with 75 of 272 (28%) of the initiators (P < .001). In addition, women who used more CAM modalities, as indicated by a higher CAM index score, were significantly less likely to initiate chemotherapy (P = .001). There was no association between mind-body practices and chemotherapy initiation in unadjusted analyses (Table 1).

In multivariable models, among women for whom chemotherapy was indicated, current dietary supplement use (adjusted OR, 0.16; 95% CI, 0.03-0.51) and a higher score on the CAM index (adjusted OR per unit, 0.64; 95% CI, 0.46-0.87) were associated with decreased likelihood of chemotherapy initiation, after controlling for demographic and clinical characteristics (Table 2). However, mind-body practices were not associated with decreased chemotherapy (adjusted OR, 1.45; 95% CI, 0.57-3.59).

In unadjusted analyses, among women for whom chemotherapy was clinically discretionary, there were no associations between CAM use and chemotherapy initiation, except for other natural products (Table 1). Similarly, analyses using multivariable models showed no statistically significant associations between current CAM use and chemotherapy (Table 2). Of note, among women for whom chemotherapy was discretionary, there was a trend toward an association between a higher CAM use index and chemotherapy initiation (adjusted OR, 0.83; 95% CI, 0.67-1.01).

At baseline (mean [SD], 89 [33] days) postdiagnosis, 265 women (39%) had initiated chemotherapy. To control for temporality, sensitivity analyses restricted participants to women who did not initiate chemotherapy prior to baseline. Results were consistent with our primary analyses, suggesting that reverse causation could not explain our results (data not shown). Similarly, sensitivity analyses restricting dietary supplement use to vitamin and/or mineral use yielded results consistent with the results for all dietary supplements (data not shown). Generalized estimating equation analyses accounting for clustering by recruitment site also did not affect the results (data not shown). Sensitivity analyses were run using a data set that included women with missing and unknown data. The results were consistent with the primary analyses (data not shown), suggesting that missingness was not a source of bias.

In a multicenter cohort of patients with nonmetastatic breast cancer, CAM use was high. Dietary supplement use and higher number of CAM modalities used, but not mind-body practices, were associated with failure to initiate chemotherapy among women for whom chemotherapy was clinically indicated based upon NCCN guidelines. This result was robust after adjusting for demographic and clinical covariates, such as age, race/ethnicity, education, tumor characteristics, and comorbidity. Women for whom chemotherapy is discretionary will likely have different reasons for chemotherapy initiation compared with women for whom chemotherapy is indicated. Here, we found that CAM use in the discretionary group was not associated with chemotherapy initiation. We hypothesized that CAM use would be associated with chemotherapy initiation among the discretionary group because there is more leeway in the decision making process, but we did not find that this was true. Instead, in the group with a clear clinical indication for chemotherapy, users of CAM were significantly overrepresented among the relatively small group that did not receive this indicated treatment. We previously showed that in the same cohort, chemotherapy noninitiation was associated with older age, more negative beliefs about chemotherapy effectiveness, and more concern about adverse effects.² This analysis identifies dietary supplement use as a potential new risk factor for chemotherapy noninitiation.

Among patients with cancer, patients with breast cancer are the highest users of CAM.^9,10,29,30 The prevalence of CAM use observed in the BQUAL study population is similar to what has been observed in other populations of patients with breast cancer, ranging from 62% in the Nurses’ Health Study,³¹ to 80% in Women’s Healthy Eating and Living,³² to 95% in the Long Island Breast Cancer Study Project.³⁰ Estimates of use in different cohorts likely vary depending on the definition of CAM use, the population, and the year, as patterns of CAM use in adults have varied over time.³³

Among CAM modalities, dietary supplement use was associated with a lower likelihood of chemotherapy initiation, contrasting with mind-body practices, which were not. By nature, dietary supplement use differs from mind-body practices. Dietary supplements are ingested substances, whereas mind-body practices include behaviors and body-work. Presumably, patients with cancer who use either type of CAM hope to achieve a health-related benefit.¹⁵ In the general population, users of dietary supplements often use these products to improve or maintain overall health or for organ-specific health reasons (eg, heart, joint, skin).³⁴ A population-based study of patients with cancer³⁵ reported that patients with cancer took dietary supplements because “it was something they could do to help themselves”^35(p677) to boost immune function and to improve energy. In contrast, users of mind-body practices often do so for psychological benefits (eg, stress, anxiety, depression) and pain management.³⁶ Distinguishing these motivations for use may explain our findings. To date, in the breast oncology setting the majority of dietary supplement interventions have not proven beneficial, whereas mind-body practices have a stronger evidence base for improving treatment-related adverse effects, psychological conditions, and quality of life.¹⁶

Our results contrast with an observational analysis from the multicenter Women’s Healthy Eating and Living (WHEL) study,³² that reported that women who reported regular or occasional CAM use with consultation of a CAM practitioner were more likely to receive chemotherapy compared with individuals who did not use CAM.³² However, the WHEL study ascertained CAM use approximately 5 years after breast cancer treatment, potentially leading to recall bias. Moreover, the WHEL study excluded dietary supplements from their definition of CAM use. As we found that chemotherapy noninitiation was most likely to occur among users of dietary supplements as compared with other CAM modalities, the exclusion of dietary supplements from the WHEL study CAM definition may have led to conclusions different from ours.

It is important to consider possible alternative explanations for our findings. It is unclear whether the association between CAM use and chemotherapy noninitiation reflects longstanding decision-making patterns among participants. It is possible that CAM users who did not initiate treatment were longtime CAM users and chose CAM as an alternative to chemotherapy. However, the current study did not collect data on CAM use history and therefore cannot rule out this possibility. Nevertheless, regardless of a woman’s prior CAM use, our findings suggest that current CAM use, which in a clinical encounter is easier to assess than past use, was associated with chemotherapy noninitiation.

The difference in the findings between women with and without clinically indicated chemotherapy warrants exploration. We previously reported that treatment, decision making, and psychosocial factors related to chemotherapy initiation were largely consistent between these 2 groups.² In contrast, our findings here show that CAM use is a predictor of chemotherapy noninitiation among women with clinically indicated treatment but not among those with discretionary chemotherapy. In the face of the uncertainty of the benefit of chemotherapy, it is possible that biomedical factors, such as recommendation by a physician, may trump other health beliefs and practices, such as CAM use.

Our study’s major strength is the prospective examination of CAM use and chemotherapy initiation. Though causation cannot be inferred from an observational study, the time lag between baseline data collection and follow-up for chemotherapy initiation for the majority of participants established a temporal relationship between CAM use and chemotherapy initiation. In addition, we used NCCN guidelines to ascertain clinical indication of chemotherapy.

However, there are important limitations to acknowledge. Among women for whom chemotherapy was clinically indicated, the majority of women (89%) initiated chemotherapy, and thus noninitiation (11%) was a rare event in a relatively small population of women (n = 34). Despite the small number of events, we were able to detect statistically significant differences that were robust after adjusting for clinical and demographic factors. Both chemotherapy initiation and CAM data were self-reported, possibly resulting in misclassification and reporting bias. In addition, we did not collect data from the physicians on the specific chemotherapy recommendation that was given to each patient, nor did we collect data from the patients on the reason(s) for declining chemotherapy. Also, we do not have data on genomic risk of recurrence, which is now standard of care and may be an important driver of chemotherapy initiation but was not commonly considered in clinical settings while BQUAL was enrolling participants. The BQUAL CAM data lack detailed information on individual CAM modalities used, history of use prior to breast cancer diagnosis, frequency and duration of use, and specific reason(s) for use. Similarly, we do not have data on institutional and provider-level practices regarding recommendations on CAM therapy use prior to or during chemotherapy. The sample is largely from integrated health care systems, so the results may not be representative of chemotherapy initiation determinants in other populations.

In a prospective cohort study testing an a priori hypothesis, current dietary supplement use and number of CAM therapies used were associated with lower chemotherapy initiation after controlling for demographic and clinical factors. Though the majority of women with clinically indicated chemotherapy initiated treatment, 34 of 306 (11%) did not. A cautious interpretation of results may suggest to oncologists that it is beneficial to ascertain CAM use among their patients, especially dietary supplement use, and to consider CAM use as a potential marker of patients at risk of not initiating clinically indicated chemotherapy. This information can prompt further discussions between physicians and patients on chemotherapy expectations, with the common goal of addressing concerns and enhancing shared and informed decision making about clinically indicated chemotherapy. Future studies are needed to further distinguish the role of CAM in cancer management.

Corresponding Author: Heather Greenlee, ND, PhD, Herbert Irving Comprehensive Cancer Center and Mailman School of Public Health, Columbia University, 722 W 168th St, Rm 733, New York, NY 10032 (hg2120@columbia.edu).

Accepted for Publication: February 24, 2016.

Published Online: May 12, 2016. doi:10.1001/jamaoncol.2016.0685.

Author Contributions: Dr Greenlee and Mr Shi had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Greenlee, Mandelblatt, Lee, Kushi, Hershman.

Acquisition, analysis, or interpretation of data: Greenlee, Neugut, Falci, Hillyer, Buono, Mandelblatt, Roh, Ergas, Kwan, Tsai, Shi, Lamerato, Kushi, Hershman.

Drafting of the manuscript: Greenlee, Falci, Mandelblatt, Roh, Lee, Shi, Hershman.

Critical revision of the manuscript for important intellectual content: Greenlee, Neugut, Hillyer, Buono, Mandelblatt, Ergas, Kwan, Lee, Tsai, Shi, Lamerato, Kushi, Hershman.

Statistical analysis: Greenlee, Falci, Buono, Mandelblatt, Ergas, Tsai, Shi.

Obtained funding: Neugut, Mandelblatt, Lee, Kushi, Hershman.

Administrative, technical, or material support: Greenlee, Hillyer, Roh, Kwan, Lee, Shi, Lamerato, Kushi.

Study supervision: Greenlee, Kushi, Hershman.

Conflict of Interest Disclosures: Dr Kushi is on the Board of Directors of the Kushi Foundation, a nonprofit organization that promotes educational activities regarding macrobiotics. No other conflicts are reported.

Funding/Support: This research was supported by an NCI K23 (grant No. CA141052) from the National Cancer Institute (Dr Greenlee); a Department of Defense Breast Cancer Center of Excellence Award (award No. BC043120; Dr Neugut); an NCI R01 (grant No. CA105274; Dr Kushi); and NCI R01s (grants CA124924 and 127617), U10 (grant No. CA84131), and K05 (grant No. CA96940) (Dr Mandelblatt).

Role of the Funder/Sponsor: The funders/sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Previous Presentations: An abstract from a previous version of these analyses was presented in an oral poster discussion at the 2015 San Antonio Breast Cancer Symposium; December 10, 2015; San Antonio, Texas.