End-of-life care in a population-based cohort of cancer patients: clinical trial participation versus standard of care
- Carrie A Thompson1,
- Sarah E Hugo2,
- Keith Mark Swetz3,
- Paul J Novotny4,
- Jeff A Sloan4,
- Charles L Loprinzi5,
- Timothy J Moynihan5 and
- Tait D Shanafelt1
- 1Division of Hematology, Department of Medicine, Mayo Clinic, Rochester, Minnesota, USA
- 2Division of Primary Care Internal Medicine, Department of Medicine, Mayo Clinic, Rochester, Minnesota, USA
- 3Division of General Internal Medicine, Department of Medicine, Mayo Clinic, Rochester, Minnesota, USA
- 4Department of Health Sciences Research, Mayo Clinic, Rochester, Minnesota, USA
- 5Department of Medical Oncology, Mayo Clinic, Rochester, Minnesota, USA
- Correspondence to Dr Carrie A Thompson, Division of Hematology, Department of Medicine, Mayo Clinic, 200 1st Street SW, Rochester, MN 55905, USA;
Objectives To evaluate end-of-life care in a cohort of oncology patients in Olmsted County, Minnesota, USA, and compare differences between patients participating in clinical trials and those not in clinical trials.
Methods A population-based cohort of subjects with active oncological disease who died between 2000 and 2002 was constructed retrospectively using institutional databases. Clinical trial participation and care during the last 2 months of life were analysed.
Results A total of 395 eligible patients were identified. In the 2 months prior to death, 94 (24%) patients received chemotherapy, 232 (59%) were hospitalised, 249 (63%) were in hospice and 315 (80%) had a do not resuscitate (DNR) code status. Only 8 (2%) patients received cardiopulmonary resuscitation (CPR) and 26 (7%) patients participated in a clinical trial. Patients in clinical trials were more likely to receive chemotherapy (69.2% vs 20.6%; p<0.001), undergo intubation/mechanical ventilation (15.4% vs 5.4%; p=0.040) and less likely to have DNR code status (50.0% vs 81.8%; p<0.001) when compared with patients not in clinical trials. However, no differences in hospice enrolment, days in hospice, days in the hospital, CPR or location of death were noted.
Conclusions Although opportunities for improvement exist, high quality end-of-life care was found in this study of patients with active malignancy. A majority (over 60%) of patients enrolled in hospice prior to death, 80% had a DNR status and only 2% received CPR. Although clinical trial participants received more aggressive treatments during the last 2 months of life, they did not appear to have lower quality end-of-life care.
Although the process of dying is a personal and subjective experience, indicators suggesting good end-of-life care exist for patients with cancer. Favourable indicators include avoiding inappropriate chemotherapy use near the time of death, limiting emergency room visits and hospitalisations, appropriate utilisation of hospice services and advanced directives, preferred location of death, open communication, shared decision making, and multidisciplinary care including pain and symptom management.1 ,2 The American Society of Clinical Oncology Quality Practice Initiative program has recently defined specific measures for quality end-of-life care, which include comprehensive assessment and treatment of pain and dyspnoea, hospice enrolment, and low rates of chemotherapy in the last 2 weeks of life.3 In addition, patients generally do not desire more aggressive treatment in the end-of-life, suggesting that less intensive practices at the end-of-life provide greater quality of care.4
Clinical trials are the key mechanism to identify future oncological advances and improve quality of care for patients with cancer. For patients with advanced oncological disease, clinical trials provide patients the opportunity to receive new and innovative treatments that can improve treatment for future cancer patients. Despite this, participation in clinical trials is low. According to the National Cancer Institute, only 3%–5% of all cancer patients in the USA participate in a clinical trial.5 One barrier to patient participation in clinical trials is concern that quality of life might be reduced, particularly for patients near the end-of-life.6 Despite this perception, it is not well understood how participation in a clinical trial affects end-of-life care.
This study was designed to evaluate care received in the last 2 months of life in a population-based cohort of oncology patients residing in Olmsted County, Minnesota, USA (population 145 769 individuals), including clinical trial participation and differences in end-of-life care between patients participating in a clinical trial and those not in a clinical trial.
A population-based cohort of 395 subjects with active oncological disease who died between 2000 and 2002 was constructed retrospectively. Eligible patients were identified using institutional databases, including the Rochester Epidemiology Project, an infrastructure that provides access to comprehensive and complete medical records of all inhabitants of Olmsted County.7 All patients were ≥18-years-old, resided in Olmsted County, were seen at least once by a member of the Mayo Clinic Department of Oncology (to ensure that the patient had the opportunity to participate in a clinical trial if eligible) and died in the years 2000–2002. Patients were excluded if they did not have active oncological disease in the last 2 months of life (ie, if they had previous diagnosis of oncological disease but were in complete remission at the time of death) or if they had a haematological malignancy (eg, leukaemia, lymphoma) rather than a solid tumour. The Mayo Institutional Review Board reviewed the study and determined it to be exempt from review, in accordance with the Code of Federal Regulations, 45 CFR 46.
A total of 395 study subjects fulfilling the eligibility criteria were identified. Clinical data from the last 2 months of life were abstracted from medical records including age at death, gender, marital status, cancer subtype, date of diagnosis and clinical trial participation. Therapies used in the last 2 months of life were also reviewed including: chemotherapy, radiation treatment, mechanical ventilation, haemodialysis, antibiotics and blood transfusions. Procedures reviewed included invasive procedures such as surgery, number of days spent in the hospital, number of days spent in an intensive care unit (ICU) and cardiopulmonary resuscitation (CPR). CPR was defined as the act of CPR being performed and documented in the medical record. Advanced care planning, including most recent documented resuscitative wishes (code status), hospice enrolment and number of days spent on hospice, was reviewed. Do not resuscitate (DNR) code status was defined as documentation of patient wishes for no resuscitation. Medical records were searched in entirety for code status as this was documented in a variety of locations, including physician notes, social work notes and hospice enrolment documentation. Minnesota state death certificates were also reviewed to determine the cause (cancer-related death or other) and location (home, hospital, nursing home, inpatient hospice) of death.
Statistical analyses were performed using SAS V.9.2 (Cary, North Carolina, USA; SAS Institute Inc.). Differences in demographics between patients in clinical trials versus patients not in clinical trials were tested using Fisher's exact tests for categorical variables and Wilcoxon tests for continuous variables. Survival time differences were also tested using Wilcoxon tests instead of standard survival tests because there were no censored survival times. All tests were two sided and done at 5% significance levels with no adjustments for multiple testing. Fisher's exact tests had about 80% power to detect about a 40% difference in proportions. The Wilcoxon tests had about 80% power to detect a difference if there was about a 75% probability that an observation in one group is less than an observation in the second group. Multivariate models were fit to adjust for the confounding factors of gender, age at diagnosis, marital status and cancer site. Separate logistic models were used for modelling whether or not patients had DNR codes, used hospice and were on a clinical trial. Linear regression models were used for survival time from diagnosis, days in the ICU, days in the hospital and days in hospice.
According to institutional databases (the Rochester Epidemiology Project), in the year 2000, 272 adults died from malignancies in Olmsted County. In the year 2001, 280 individuals died from malignancy, and in 2002, 321 individuals died from malignancies. After excluding patients with haematological malignancies, those in complete remission at the time of death, those who were not seen at least once by a member of the Mayo Clinic Department of Oncology and those whose mailing address at the end-of-life was outside of Olmsted County, there were 395 eligible individuals for this cohort. Table 1 depicts the patient characteristics. The median age at death was 72.0 (21.0–95.0) years, and the proportion of male subjects was 55%. Lung cancer (n=123, 31%) and gastrointestinal malignancies (n=115, 29%) were the most common histological subtypes. Similar to the national average of 3%–5%,5 26 (7%) patients participated in a clinical trial during the last 2 months of life.
Treatments received during the last 2 months of life and end-of-life care are detailed in table 2. In the 2 months prior to death, 94 (24%) patients received chemotherapy and 59 (15%) received radiation therapy. Over 60% of the patients were enrolled in hospice prior to death. Among patients enrolled in hospice, the median number of days in hospice prior to death was 26.5 (1–704) days. Of the 233 (59%) patients who were hospitalised during the 2 months prior to death, the median number of days in the hospital was 8.0 (1–55) days. While the largest percentage of patients died at home (46%), 20% died in the hospital, 17% died in inpatient hospice and 14% of patients died in a nursing home. In all, 80% of patients had advance directives indicating DNR code status within the last 2 months of life. A total of 24 (6.1%) patients were intubated prior to death; however, only 8 (2%) patients received CPR.
Collectively, 26 (7%) participated in a clinical trial during the last 2 months of life. Of these, 11 (42%) participated in a phase I trial, 11 (42%) in a phase II trial, 2 (8%) in a phase III trial and 2 (8%) were in a trial of unknown phase. Patients in a clinical trial were younger at diagnosis (mean age 60.0 vs 66.7 years, p=0.02), younger at death (mean age 61.9 vs 69.7 years, p<0.01) and were more likely to be married (81% vs 60%, p=0.03). In multivariate analysis, female subjects (OR=2.89, p=0.02), younger patients (OR=1.04, p=0.03) and married patients (OR=3.28, p=0.03) were more likely to participate in clinical trials. Cancer type did not relate to clinical trial participation.
A comparison of treatments received during the last 2 months of life and end-of-life care between patients participating in clinical trials and those receiving standard care is shown in table 3. Patients enrolled in a clinical trial were more likely to receive chemotherapy (69% vs 21%, p<0.01), undergo endotracheal intubation (15% vs 5%, p=0.04) and have a full code resuscitation status (30.8% vs 9.8%, p<0.01). Phase I clinical trial patients were almost twice as likely to be DNR as phase II or III trial patients (63.6% vs 33.3%, p=0.22), although this was not statistically significant due to the small numbers of patients in each group. No differences were noted in number of days spent in the hospital or ICU, hospice admission rates or days on hospice between those participating in clinical trials and those receiving standard care. Patients enrolled in a clinical trial had fewer numbers of days between date of diagnosis and death, but this finding was not statistically significant (median days 308 vs 436, p=0.33). After multivariate adjustment for age, marital status, gender and cancer type, patients in clinical trials remained about a fourth as likely to have an active DNR code status documented in their chart as compared with those treated with standard of care (OR 0.23, p<0.01). On multivariate analysis, married patients were more likely to be admitted to hospice (OR 2.27, p<0.01), spend less days in the ICU (regression estimate −1.13, p<0.01) and less days in the hospital (regression estimate −3.66, p<0.01).
This study provides valuable information regarding the experience of dying in a population-based cohort of cancer patients. Overall, we found that the quality of end-of-life care was good. Most patients died at home, which is generally the preference of cancer patients at the end-of-life.8 In addition, the majority of patients participated in hospice programmes, with a median duration of enrolment of almost 1 month. Most patients had a DNR code status documented at the time of death, indicating that discussions regarding care wishes at the end-of-life were likely held between the patients and their medical providers. Only small percentages of the patients received invasive procedures in the last 2 months of life. This study highlights opportunities for improvement in quality of end-of-life care. For example, 24% of all patients received chemotherapy within 2 months of death, although the use of chemotherapy is not a definitive marker for aggressive end-of-life care as palliative chemotherapy can improve quality of life.
Enrolling patients in cancer clinical trials is a challenge. The current study cohort had slightly higher rates of clinical trial involvement when compared with the national average, 7% versus 3%–5%. In this cohort, younger patients were more likely to enrol in clinical trials, which has been reported in other studies.9–11 We found a positive association between marital status and clinical trial enrolment, supporting that a lack of adequate social support may be a barrier to trial enrolment, which is consistent with previously reported findings.6 ,12
There were significant differences in end-of-life care for patients enrolled in a clinical trial, as compared with patients treated with standard of care. Patients in clinical trials were over three times more likely to receive chemotherapy. This finding is expected based on patient selection, as most clinical trials for cancer patients involve chemotherapy. However, there was a concurrent trend towards other medical interventions for patients in clinical trials, including endotracheal intubation, which is suggestive of more aggressive care overall. In addition, clinical trial treatments may have toxic side effects, and some of these medical interventions may have been directed at alleviating the complications of therapy. However, interventions may have been performed for diagnostic or palliative reasons, and therefore conclusions regarding differences in these should be limited. Previous studies have concluded that patients may elect for chemotherapy near the end-of-life regardless of the information discussed with them, as it may be difficult for the patients to conceptualise their own mortality, or they may accept high degrees of toxicity with minimal chances of benefit as a personal preference.13
In this study, patients in clinical trials were significantly less likely to have a DNR code status. These patients may have tended to be more aggressive regarding their care and chose to be full code over DNR, although maximal medical therapy may still be consistent with a DNR status as long as the plan is consistent with the patient's goals of care.14 ,15 Alternatively, providers may not have been initiating code status discussions as frequently with patients on clinical trials. Importantly, differences were not found between rates of hospice admission, the number of days in the hospital or the number of days on hospice in those on clinical trials versus those treated with standard therapy. This suggests that providers are appropriately introducing hospice discussions with patients on clinical trials at similar rates to patients not on clinical trials.
The strengths of this study include its population-based design. As Mayo Clinic is the only provider of oncological services in Olmsted County, the patients studied can be considered truly representative of the Olmsted County community as a whole. While Olmsted County cancer patients may travel elsewhere to enrol in clinical trials, we believe that the number of patients who seek care at other institutions is low, given that the rate of clinical trial enrolment in our population is similar to national rates. Clinical patient data, rather than administrative data, were used, which allowed the ability to study patients enrolled in clinical trials and to capture variables such as resuscitation status, regarded as a critical component of quality end-of-life care. However, studying end-of-life care is not without limitations, including the retrospective nature of this study. This study is limited by small numbers of patients enrolled in clinical trials and the infrequent nature of some of the interventions assessed (eg, dialysis, surgery). Therefore, due to small numbers in the clinical trial group, results should be interpreted with caution. In addition, the results might not be extrapolated to other communities, as Olmsted County is primarily made up of Caucasian individuals (90.3%),16 and every community has different resources for end-of-life care.
Future studies to study quality of end-of-life care for cancer patients are necessary, including care for those who enrol in clinical trials. At the time that these data were collected and analysed, quality oncology care was not well defined and, since then, organisations such as the American Society for Clinical Oncology have undertaken initiatives to define and measure quality care.3 In addition, the practice of palliative care and clinical trials may have evolved and, therefore, these data must be interpreted with some caution. However, these data are valuable to inform about the care delivered to those who choose to participate in clinical trials. In the future, prospective studies starting at the time of enrolment into a clinical trial would be helpful to obtain a better understanding of how care is given at the end-of-life for clinical trial participants. In addition, the oncology community needs to continue to encourage enrolment into clinical trials.
In conclusion, high quality end-of-life care was seen in this population-based study of oncological patients during the last 2 months of life in which the majority of patients participated in hospice services and had a DNR code status. Few cancer patients underwent intubation, invasive surgery, haemodialysis or endotracheal intubation in the last 2 months of life. Patients participating in clinical trials showed a trend towards more aggressive treatment with more chemotherapy use and less frequent DNR code status compared with those not on trial, but overall did not appear to have a lower quality end-of-life experience.
The authors would like to thank Charles Erlichman, MD, for his thoughtful review and comments.
Funding This study was funded by CA90628-08 Paul Calabresi Award for Clinical Oncology (K12) and was made possible by the Rochester Epidemiology Project (Grant Number R01 AR30582 from the National Institute of Arthritis and Musculoskeletal and Skin Diseases).
Competing interests None.
Ethics approval The Mayo Institutional Review Board reviewed the study and determined it to be exempt from review, in accordance with the Code of Federal Regulations, 45 CFR 46.
Provenance and peer review Not commissioned; externally peer reviewed.
- Received 31 May 2012.
- Revision received 15 November 2012.
- Accepted 21 January 2013.
- Published Online First 8 March 2013
- Published by the BMJ Publishing Group Limited. For permission to use (where not already granted under a licence) please go to http://group.bmj.com/group/rights-licensing/permissions