Refusal of Recommended Chemotherapy for Ovarian Cancer: Risk Factors and Outcomes; a National Cancer Data Base Study

NCCN: Continuing Education

Background

Ovarian cancer remains the deadliest gynecologic malignancy, with relative 5-year survival rates for patients diagnosed with epithelial ovarian cancer (EOC) ranging from 17% to 90% depending on stage at diagnosis.¹ Treatment of ovarian cancer has evolved over several decades; the standard treatment for advanced-stage disease is either debulking surgery followed by chemotherapy or neoadjuvant chemotherapy with (or without) surgery.^2–4 Multiple studies have defined the role and efficacy of chemotherapy in ovarian cancer, showing increased progression-free survival (PFS) and overall survival (OS) in patients receiving the standard treatment of combination platinum- and taxane-based chemotherapy, as recommended in the NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) for Ovarian Cancer.^6–8

As with any treatment, patient autonomy is the center of decision-making. As such, after discussing and recommending chemotherapy to patients with ovarian cancer, not all patients will elect to proceed with this treatment. The decision to proceed with known, effective cancer therapy is not always clearcut, and various demographic, socioeconomic, and psychological influences have been implicated in the decision to refuse therapy. Racial disparities, medical comorbidities, and physical access to care have been shown to be barriers in health systems with universal insurance coverage, and other studies have highlighted subjective reasons patients may give for refusing recommended chemotherapy.^9,10 Patient education of risks/benefits of therapy, goals of care, and quality-of-life (QOL) considerations have also been shown to be of utmost importance in the decision to refuse recommended therapy. A study by Havrilesky et al¹¹ suggests that women with ovarian cancer would exchange improved PFS for reduced treatment side effects resulting in improved QOL.

Although patient autonomy is a cornerstone of modern medical ethics and the current health care delivery paradigm, it seems that for some patients, refusal of recommended therapy may sometimes be reflective of incompletely informed decisions. We are unaware of prior studies that have examined the possible interactions of such externalities on treatment decisions among patients with ovarian cancer, whose socioeconomic circumstances and concentration of care with gynecologic oncology subspecialists may be different from those of other cancer populations.

Furthermore, a recent article in Time magazine highlighted patients' decision to refuse chemotherapy, and explained motives for such decisions, thus bringing national attention to this important and relevant topic. The article reminded physicians that respecting patient decisions and wishes remains integral to a healthy physician-patient relationship.¹² Given the possibility of chemotherapy refusal among patients with ovarian cancer, and the importance of maintaining patient rapport, we sought to understand the parameters of refusal of treatment in the ovarian cancer patient population. As such, the objective of this study was to identify risk factors associated with refusal of recommended chemotherapy and its impact on OS using multivariable models in an attempt to control for confounders in patients with EOC using data from the National Cancer Data Base (NCDB).

Methods

We identified patients in the used the NCDB diagnosed with EOC from January 1998 to December 2011. The NCDB is an extensive database established jointly by the American Cancer Society and American College of Surgeons Commission on Cancer. The database collects approximately 70% of new cancer diagnoses from more than 1,500 accredited hospitals. The data are used to track new diagnoses and outcomes, in addition to quality improvement measures and surveillance.¹³ Initial data selection was as previously described by Cliby et al¹⁴ in their recently published NCDB-based article. Patients who refused chemotherapy were identified and compared with those who received recommended, multiagent chemotherapy; prior surgery was not requisite to be included in this group. We compared specific data points collected regarding chemotherapy administration. These data points were defined in the NCDB as, “Multiagent chemotherapy administered as first course therapy,” which was compared with, “Chemotherapy was not administered. It was recommended by the patient's physician, but this treatment was refused by the patient, a patient's family member, or the patient's guardian. The refusal was noted in patient record.” Our primary end point was to determine what risk factors identifiable in the NCDB are associated with refusal of recommended chemotherapy. Secondary end points include effects on OS and associated risk of death. Univariate analyses were performed using Chi-square tests with Bonferroni correction and log-rank test. Multivariable analyses were performed using Cox proportional hazards modeling and binary logistic regression. OS was determined using the Kaplan-Meier method. Threshold for statistical significance was set at a P value of less than 0.05. Odds ratios (OR) and hazard ratios (HR) with 95% CIs were created, and an OR greater than 1.0 implied an increased likelihood of association with chemotherapy refusal and an HR of greater than 1.0 implied increased likelihood of death. Data were analyzed using SPSS statistical software (IBM, Armonk, NY). The University of Pittsburgh Medical Center Institutional Review Board granted this study exempt status. The data used in this study were derived from a de-identified NCDB file. The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methodology used nor the conclusions drawn from these data by the investigator.

Results

A total of 147,713 patients were identified, of whom 2,707 (1.8%) refused chemotherapy. The patients who were defined as refusing chemotherapy were identified in the NCDB by the entry, “Chemotherapy was not administered. It was recommended by the patient's physician, but this treatment was refused by the patient, a patient's family member, or the patient's guardian. The refusal was noted in patient record.” In general, most patients in our cohort were younger than 60 years (44.6%), followed closely by those aged 60 to 75 years (40.3%), and least commonly by those older than 75 years (15.1%). When evaluating stage, 12.8% were diagnosed as stage I, 0.09% as stage II, 50.1% as stage III, and 28.2% as stage IV. Charlson-Deyo comorbidity scores of 0 were recorded in 81.5%, a score of 1 was recorded in 14.9%, and scores of 2 or greater were recorded in 3.6% of cases. Most facilities that administered chemotherapy reported treating 16.5 to 27.7 (26.3%) or more than 27.7 (27.0%) cases per year. The median OS of patients who received multiagent chemotherapy was 43 months, whereas the median survival time was 4.8 months for those who refused chemotherapy (P<.0005).

When comparing those who received chemotherapy and those who refused, several factors were found to be associated with chemotherapy refusal. On exploratory univariate analysis, older age, African American race, more medical comorbidities, having Medicare/Medicaid or no insurance, lower income, lower education level, residence near the treating facility, facilities in the Northeast, low-volume facilities, facilities with lower-than-expected adherence rate to the NCCN Guidelines for Ovarian Cancer, later year of diagnosis, lower grade, and higher stage of disease were all significantly associated with chemotherapy refusal (all P<.0005), whereas unplanned surgical readmission and residential setting were not (Tables 1 and 2).

On multivariable logistic regression, older age (>70 vs ≤50 years; OR, 4.2), more medical comorbidities (≥2 vs 0; OR, 1.8), not having insurance (OR, 1.4–2.9, depending on type), later year of diagnosis (2009–2011 vs 2002–2004; OR, 1.3), lower-than-expected facility adherence to NCCN Guidelines (OR, 1.2), treatment at a low-volume center (lowest vs highest; OR, 1.6), lower grade (1 vs 3; OR, 2.0), and stage (I vs IV; OR, 2.2) were all significantly and independently associated with chemotherapy refusal, whereas patient race, residential urban status/income/education level, unplanned surgical readmission, and treatment facility academic status/location/distance were not. After controlling for known patient, facility, and disease prognostic factors, chemotherapy refusal

Table 1.

Comparison of Variables in Univariate Analyses in Those Who Received and Refused Chemotherapy

View Table

View Table

Discussion

Standard treatment for patients with EOC follows the recommendations in the NCCN Guidelines for Ovarian Cancer.⁷ Following these treatment guidelines can result in significantly improved outcomes, which translates into increased PFS and OS for these patients.¹⁵ Cliby et al¹⁴ showed that adherence to NCCN Guidelines ranged from 30.8% in community cancer centers to 49.1% in academic comprehensive cancer centers. This finding was associated with a significantly improved 5-year OS in those treated at academic comprehensive cancer centers with higher adherence to recommendations in the NCCN Guidelines.¹⁴ In addition, patients are more

Table 2.

Comparison of Variables in Multivariable Analyses in Those Who Received and Refused Chemotherapy

View Table

View Table

Given lack of good screening measures, it remains difficult to diagnose ovarian cancer at an early stage. Bristow et al¹⁵ reported on racial and socioeconomic disparities in patients diagnosed with ovarian cancer. Their study showed that insurance status and household income less than $35,000 translated into worse OS, with an approximately 30% increased risk of death. Investigation of patients with colorectal cancer found that 8.1% refused chemotherapy due to economic difficulty, and 11.9% declined chemotherapy due to cost or lack of family support.⁹ Similarly, our study showed that having insurance coverage decreased the rate of chemotherapy refusal. Offering comprehensive care, complete with financial counseling

Table 3.

Comparison of Variables in Univariate Analyses With Respect to Effect on Survival

View Table

Table 4.

Comparison of Variables in Multivariable Analyses With Respect to Effect on Survival

View Table

On occasion, regardless of all efforts, patients will continue to decline chemotherapy. Shared decision-making and patient autonomy remain extremely important aspects of care. Collaborating with the patient and her family to discuss chemotherapy and options or alternatives is essential to this shared decision-making.¹⁶ Previous studies have evaluated patient QOL and preferences in relation to receiving treatment for EOC. Emotional and psychological aspects of treatment, such as patient outlook toward treatment, body image, and fatigue, can often play a significant role in a patient's treatment experience.^17,18 Havrilesky et al¹¹ queried patients regarding symptoms, side effects, and PFS preferences when choosing chemotherapy. Their data showed that patients prefer improved QOL with reduced symptoms over PFS. That study showed that participating patients would choose to decrease nausea and vomiting from severe to mild during chemotherapy even if this decision resulted in PFS decreasing by 6.7 months. Similar results were observed when evaluating peripheral neuropathy and abdominal symptoms, with patients preferring fewer symptoms and/or adverse effects despite decreasing PFS.¹¹ In patients who ultimately refuse chemotherapy, early referral to and institution of palliative care has been shown to improve QOL.¹⁹

This study demonstrates that, despite standard surgical and chemotherapeutic treatments with known efficacy, approximately 1.8% of patients diagnosed with EOC will refuse recommended chemotherapy, which appears to be in line with, or slightly better than, the percentages for other cancer sites.^9,10 This is limited by the relative proportion of unavailable data in our study, which is a limitation of the ongoing nature of the NCDB data collection process; although demographic data from recent cases may be captured, treatment and survival data do not mature for at least 5 years.

Our study reports multiple factors, both modifiable and not, associated with refusal of recommended chemotherapy in patients with EOC. In addition, our data show a significant reduction in unadjusted OS of 38.2 months, with an adjusted 189% increased risk of death, in patients who refuse chemotherapy (Figure 1). Given this, one might consider improving patient and practitioner education and discussion regarding the risks and benefits of chemotherapy in the treatment of ovarian cancer, and focus on goals of care. This is borne out in the data showing that patients being treated at high-volume centers seem to refuse chemotherapy less frequently, perhaps due to interactions with other cancer survivors or frequent interactions of health care providers and patients. Interestingly, being in an academic setting does not necessarily confer this advantage.

Unadjusted (A) and adjusted (B) overall survival according to chemotherapy refusal.

Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 14, 5; 10.6004/jnccn.2016.0062

View Full Size

Unadjusted (A) and adjusted (B) overall survival according to chemotherapy refusal.

Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 14, 5; 10.6004/jnccn.2016.0062

Unadjusted (A) and adjusted (B) overall survival according to chemotherapy refusal.

Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 14, 5; 10.6004/jnccn.2016.0062

One encouraging finding from this study is the lack of geographic or facility distance factor on the patient refusal, suggesting that physical parameters may be less of a barrier to care for patients with ovarian cancer compared with other cancers.

Overall, our study suggests opportunities to reduce the rate of chemotherapy refusal, including improved education, counseling, and access to high-volume centers.

This study is limited by the retrospective nature, significant proportion of unavailable data points, and the inherent selection bias and dependence on reporting accuracy that comes with such study design. As mentioned previously, our study is limited by the proportion of unavailable data in the NCDB. For example, although we were able to identify patients who decided to refuse chemotherapy, we do not have information regarding why they made such a decision. In addition, 36.2% of patients did not receive chemotherapy due to contraindication or for unknown reasons, or died before treatment initiation; however, we do not have an explanation or further information for these patients. In our cohort, performance status data were not reported, and we used the previously validated Charlson-Deyo comorbidity score²⁰ for patient frailty, which is not as immediate and proximate as the Gynecologic Oncology Group/ECOG/Karnofsky performance status scores, none of which are available in the NCDB.

Strengths include the use of the NCDB—a large comprehensive database that covers most new cancer diagnoses in the United States, with more variables than previously available in other national databases, such as SEER²¹—to allow for a big data analytical approach to this question regarding which patients with ovarian cancer are refusing therapy and what happens to them. Because of the large cohort size, with many collected data points, we were able to control for confounding factors on multivariable analysis. This study is the first to evaluate risk factors for and associations with chemotherapy refusal in patients with ovarian cancer, and the findings high-light the need for offering a broad approach when recommending chemotherapy.

Conclusions

The goal of this study was to identify patients who are at increased risk of refusing chemotherapy, a standard treatment with known efficacy in patients with ovarian cancer. Further research could focus on what therapies, if any, are received by patients who refuse chemotherapy, and the efficacy and tolerance of those treatments. Further research is also needed to identify individualized best practices for this high-risk subgroup. Consideration should be given for low-volume centers to refer patients refusing chemotherapy for second opinion at a higher-volume facility. The study results suggest that improved health care access and individualized counseling may improve care delivery, thereby providing patients with all treatment options and improved outcomes.