- Email: [email protected]
Outcomes, preferences for resuscitation, and physician-patient communication among patients with metastatic colorectal cancer
Outcomes, Preferences for Resuscitation, and Physician-Patient Communication among Patients with Metastatic Colorectal Cancer Paul Haidet, MD, MPH, Mary Beth Hamel, MD, MPH, Roger B. Davis, ScD, Neil Wenger, MD, Douglas Reding, MD, MPH, Peter S. Kussin, MD, Alfred F. Connors, Jr, MD, Joanne Lynn, MD, Jane C. Weeks, MD, Russell S. Phillips, MD, for the SUPPORT Investigators PURPOSE: To describe characteristics, outcomes, and decision making in patients with colorectal cancer metastatic to the liver, and to examine the relationship of doctor-patient communication with patient understanding of prognosis and physician understanding of patients’ treatment preferences. PATIENTS AND METHODS: The Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments (SUPPORT) was a prospective cohort study conducted at five teaching hospitals in the United States between 1989 and 1994. Participants in this study were hospitalized patients 18 years of age or older with known liver metastases who had been diagnosed with colorectal cancer at least 1 month earlier. Data were collected by patient interview and chart review at study entry; patients were inteviewed again at 2 and 6 months. Data collected by physician interview included estimates of survival and impressions of patients’ preferences for cardiopulmonary resuscitation (CPR). Patients and physicians were also asked about discussions about prognosis and resuscitation preferences. RESULTS: We studied 520 patients with metastatic colorectal cancer (median age 64, 56% male, 80% white, 2-month survival 78%, 6-month survival 56%). Quality of life (62% “good” to “excellent”) and functional status (median number of disabili-
ties 5 0) were high at study entry and remained so among interviewed survivors at 2 and 6 months. Of 339 patients with available information, 212 (63%) of 339 wanted CPR in the event of a cardiopulmonary arrest. Factors independently associated with preference for resuscitation included younger age, better quality of life, absence of lung metastases, and greater patient estimate of 2-month prognosis. Of the patients who preferred not to receive CPR, less than half had a do-not-resuscitate note or order written. Patients’ self-assessed prognoses were less accurate than those of their physicians. Physicians incorrectly identified patient CPR preferences in 30% of cases. Neither patient prognostication nor physician understanding of preferences were significantly better when discussions were reported between doctors and patients. CONCLUSIONS: A majority of patients with colorectal cancer have preferences regarding end of life care. The substantial misunderstanding between patients and their physicians about prognosis and treatment preferences appears not to be improved by direct communication. Future research focused on enhancing the effectiveness of communication between patients and physicians about end of life issues is needed. Am J Med. 1998;105:222–229. q1998 by Excerpta Medica, Inc.
C
55,000 persons died from colorectal cancer in the US in 1995 (1). Recent research on colorectal cancer has focused on screening and treatment of the disease (2– 6). While new treatments have been developed for persons with metastatic disease (7–9), few studies have focused on outcomes and end of life preferences among these patients (10). Anecdotal evidence suggests that patients with metastatic colorectal cancer often enjoy relatively good health while contemplating the poor prognosis associated with the diagnosis (11–14). Studies of other patient populations demonstrate that physicians misunderstand patient preferences for aggressive treatments at the end of life (15–17). In the case of colorectal cancer, the combination of an intact mental capacity with a known prognosis may provide the optimal setting in which to discuss end of life preferences (18,19), suggesting that physicians should have good understanding of end of life preferences among these patients. The purpose of this study was to characterize hospitalized patients diagnosed with metastatic colorectal cancer, to describe clinical outcomes at 2 and 6 months, and to
olorectal cancer is the fifth most common cause of cancer and the second leading cause of cancer death in the United States. Despite an increase in early detection and screening in recent years, more than
From the Division of General Medicine and Primary Care (PH, MBH, RD, RSP), Beth Israel Deaconess Medical Center, Boston, Massachusetts; the UCLA Medical Center (NW), Los Angeles, California; the George Washington University Medical Center (JL), Washington, DC; the University of Virginia School of Medicine (AFC), Charlottesville, Virginia; Marshfield Clinic (DR), Marshfield, Wisconsin; Duke University Medical Center (PSK), Durham, North Carolina; and The Dana Farber Cancer Institute (JCW), Boston, Massachusetts. Supported by the Robert Wood Johnson Foundation. The opinions and findings contained herein are those of the authors and do not necessarily represent the views of the Robert Wood Johnson Foundation or their Board of Trustees. Dr. Haidet was supported in part by National Research Service Award No. PE11001-09. An earlier version of this report was presented at the meeting of the Society of General Internal Medicine, May 2– 4, 1996, Washington, DC. Requests for reprints should be addressed to Russell S. Phillips, MD, Division of General Medicine and Primary Care, Beth Israel Deaconess Medical Center, 330 Brookline Avenue, Boston, Massachusetts 02215. Manuscript submitted October 26, 1997 and accepted in revised form June 1, 1998. 222
q1998 by Excerpta Medica, Inc. All rights reserved.
0002-9343/98/$19.00 PII S0002-9343(98)00242-3
Metastatic Colorectal Cancer/Haidet et al
examine perceptions of prognosis and preferences for end of life treatment. In addition, we sought to examine physician understanding of patient preferences and the effect of communication on physician and patient understanding of treatment preferences and prognosis.
METHODS Patient Population The data in this study were collected as part of SUPPORT, the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments. SUPPORT investigators collected information on patient, doctor, and surrogate decision making in seriously ill patients at five academic acute care teaching hospitals: Beth Israel Hospital in Boston, Massachusetts; MetroHealth Medical Center in Cleveland, Ohio; Duke University Medical Center in Durham, North Carolina; St. Joseph’s Hospital in Marshfield, Wisconsin; and the University of California Medical Center at Los Angeles (20). Enrollment occurred between 1989 and 1994 in two phases. Phase I was observational and enrolled patients between June 1989 and June 1991. Phase II involved an intervention and enrolled patients between January 1992 and January 1994. Clinicians randomly assigned to the intervention group were given information about their patients’ prognoses and preferences for care. Since no differences were observed in targeted outcomes between phase I and phase II control patients or between phase II control patients and intervention patients, all patients were combined into a single group (11). Hospitalized patients were eligible for SUPPORT if they were 18 years of age or older and met criteria specific to one of nine diagnostic categories. In the colorectal cancer category, patients were considered for enrollment if there was known metastasis to the liver. Diagnostic criteria were designed to identify patients at advanced stages of the illness with a 6-month survival estimated to be approximately 50% based on pilot studies performed by the SUPPORT investigators (21). In addition, at least 1 month’s time was required to elapse between the time of colorectal cancer diagnosis and study enrollment to allow patients sufficient time to formulate preferences for their care.
Data Collection Data were obtained through chart review by trained abstractors and by interviews with patients, physicians, and surrogates. Data abstraction and interview supervisors were trained centrally and quarterly site visits were conducted to monitor compliance with protocols and to audit interviews. Ten percent of all interviews were audited by a supervisor and 10% of all charts were reabstracted by a second abstractor (22). Overall reliability (defined as complete agreement between the two abstractors) was
89% for the acute physiology score of the APACHE III severity index abstracted from chart data. Information obtained by chart review included previous surgery, radiation and chemotherapy, number of liver metastases (single or multiple), date of initial diagnosis of colorectal cancer, date of initial diagnosis of liver metastasis, presence of lung or pleural metastasis, and physiologic variables used in the computation of the APACHE III acute physiology score (23,24). Data collected by patient interview included demographic information (level of education, income, age, gender, and race), functional status, whether any recent weight loss had occurred, anxiety, depression, quality of life, and the patient’s estimate for their probability of 2- and 6-month survival. Information on patients’ preferences for cardiopulmonary resuscitation in the event of cardiac arrest and whether they had a discussion with their doctor about CPR preferences was also collected (24 –27). When patients were not interviewed, surrogate responses for CPR preferences were substituted, as would be done in clinical practice. Surrogates were defined as the persons who would speak or participate in medical decision making on behalf of the patient if the patient was unable to do so. Physicians were asked whether they thought the patient would want to be resuscitated in the event of cardiac arrest. The wording of questions about preferences for CPR has been published previously (28). Functional status was measured using a modified version of the activities of daily living scale developed by Katz (29 –31), reflecting independence in seven self-care functions. Higher scores indicate a greater number of dependencies. Quality of life was reported on a scale of excellent, very good, good, fair or poor (32). Anxiety and depression were measured using shortened subscales of the Profile of Mood States (POMS) (33–35). Six anxiety items and eight depression items were used to construct average anxiety and depression scores on a scale ranging from 0 (no anxiety or depression) to 4 (worst anxiety or depression). Patients were asked to predict their probability of being alive at 2 and 6 months after the time of the interview by choosing one of the following responses: .90%, about 75%, about 50-50, about 25%, and ,10 % (27). Physicians were asked to estimate 2- and 6-month patient survival on a continuous scale of 0% to 100% (36) An overall SUPPORT prognostic model estimate for survival was developed and validated (37). For each study patient, model-based estimates for survival at 2 and 6 months were calculated. Survival was determined by contacting all patients or surrogates 180 days after study entry and by searching the National Death Index for survival beyond 6 months. Date of death, or condition on day 180, was confirmed for all patients. Quality of life, functional status, depression, and anxiety were obtained at 2 and 6 months by patient or
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
Volume 105 223
Metastatic Colorectal Cancer/Haidet et al
surrogate interview. Information on intensive care unit stay, mechanical ventilation, and CPR during the index admission was obtained by chart review (38).
Table 1. Characteristics of Patients with Metastatic Colorectal Cancer at Study Entry Characteristic
Statistical Methods Analyses were performed using a commercial statistical package (SAS Institute, Inc., Cary, North Carolina). Median and the interquartile range (25th to 75th percentile) were used to describe continuous variables, except in the case of anxiety and depression scores, where mean values are reported to facilitate comparison with published norms. Factors independently associated with CPR preferences were identified using multivariable logistic regression, and adjusted odds ratios are reported. Variables were included in the model based on clinical criteria and the results of prior studies. Subjects without a patient or surrogate CPR preference were not included in multivariable analyses. Patient survival was estimated using the product limit method (39). A proportional hazards model with a time varying covariate was used to assess the association of do-not-resuscitate status with survival. Physician and patient estimates of prognosis were compared by dividing the continuous physician estimates into five categories (.90%, 61% to 89%, 40% to 60%, 11% to 39%, and ,10%) designed to coincide with patient prognostic categories. Discrimination in predicting survival was estimated by calculating the area under receiver operating characteristic curves for both physicians and patients (40). Calibration was estimated by direct comparison of patient and physician estimates of 2-month survival to observed survival for the three highest response categories. Data from the two lowest response categories were collapsed into a single category due to small numbers. The effect of discussions between physicians and patients on agreement about prognosis was estimated by computing the absolute value of the difference between patient and physician estimates of prognosis; the patient’s (categorical) estimate was subtracted from the physician’s (continuous) estimate. The mean difference in pairs who had a discussion was then compared with the mean difference in those who did not have a discussion. Similarly, the effect of discussions between patients and physicians on physician understanding of CPR preferences was estimated by comparing the percentages of physicians with an incorrect understanding between pairs who did and did not have such a discussion. The effect of differences in timing of doctor and patient interviews was examined by comparing percentages of physicians with an incorrect understanding between pairs who were interviewed within 1 day of one another and pairs who were interviewed more than 1 day apart. The differences in prognostic estimates were compared using Wilcoxon rank sum tests, and differences in percentages of 224
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
Percent of Interviewed Patients with Factor*
Age 50 to 75 years White race Male gender Income #$25,000 per annum Education #12 years Previous treatment for colorectal cancer Surgery Radiation Chemotherapy Multiple liver metastases Diagnosis duration .6 months Liver metastasis .6 months Lung or pleural metastasis Weight loss Independent in activities of daily living† Acute physiology score #30, day 3 Quality of life good to excellent† Anxiety score ,1.38‡ Depression score ,.687‡
72% (374/520) 85% (440/520) 56% (291/520) 58% (303/520) 60% (311/520) 88% (460/520) 25% (129/520) 77% (400/520) 46% (233/503) 76% (394/519) 48% (250/520) 22% (116/520) 39% (164/425) 83% (268/322) 86% (446/520) 62% (188/304) 83% (230/277) 69% (189/276)
* Fraction indicates the number of patients with factor/number of patients with available data. † During the 2 weeks before admission. ‡ Anxiety and depression scores are compared with previously published means for patients with cancer (35).
physicians incorrect were compared using the chi-square test.
RESULTS Description of Study Population A total of 520 patients with colorectal cancer were enrolled in SUPPORT (Table 1). The median age was 64 at study entry (interquartile range 56 to 71) and the median years of education was 12 (interquartile range 11 to 14). Eighty-five percent were white, 9% black, and 4% were Hispanic. Most of the patients (96%) had received at least one cancer therapy. While the selection criteria required only the presence of a single liver metastasis, 57% had additional liver, lung, or pleural metastases. Among those who reported weight loss, the median loss was 12 pounds (interquartile range 7 to 22 pounds). The majority of the patients (60%) had an oncologist as their attending physician. The remaining physicians were surgeons (32%), general medicine physicians (6%), pulmonary/intensive care unit physicians (0.5%), and cardiologists (1.5%). Despite the stage of disease, the cohort had a relatively high level of functioning. The median number of dependencies at 2 weeks before study entry was 0, and 62%
Volume 105
Metastatic Colorectal Cancer/Haidet et al
Table 2. End of Life Preferences and Clinical Decisions among Patients with Metastatic Colorectal Cancer Preference/Decision Patient estimates for survival Two-month survival $90% Six-month survival $90% Physician estimates for survival Two-month survival $90% Six-month survival $90% Patient preferences for CPR Patient (or surrogate) preferred CPR Physician thinks patient (or surrogate) prefers CPR Do-not-resuscitate note or order in chart Advance directives Have power of attorney Have living will
Percentage* 82% (207/252) 75% (176/235) 22% (89/406) 5% (20/406) 63% (214/339) 55% (173/314) 22% (114/520) 43% (83/192) 35% (97/276)
* Fraction indicates the number of patients with factor/number of patients with available data. CPR 5 cardiopulmonary resuscitation.
(hazard ratio 5 4.0, 95% confidence interval 3.2 to 5.0). Based on the SUPPORT prognostic model, the mean 2-month predicted survival for the 113 patients with donot-resuscitate orders was .61, significantly less than the .78 prediction for those without such orders (P ,.001). Similar results were obtained using 6-month predictions. The survivors who were interviewed at 2 (n 5 210) and 6 (n 5 142) months had relatively good functional status, with a median number of disabilities of 0 at both times. Among living patients who were not available for interview but had surrogate responses at 2 (n 5 65) and 6 (n 5 38) months, the median number of disabilities was 1 at 2 and 6 months (interquartile range 5 0 to 5 at 2 months and 0 to 3 at 6 months). Quality of life was rated by interviewed patients as good to excellent in 73% at 2 months, and 68% at 6 months. Anxiety and depression scores remained low at follow-up with mean (ISD) anxiety scores of .67 6 .70 at 2 months, .66 6 .74 at 6 months, and mean depression scores of .46 6 .66 at 2 months and .44 6 .68 at 6 months.
Use of Intensive Care reported quality of life as good to excellent. In addition, mean scores on the Profile of Mood States subscales were low as compared with other cancer patients (35), indicating low levels of psychologic distress due to either anxiety or depression.
Patient Preferences and Decision Making Patients had high estimates of 2- and 6-month survival as compared to their physicians (Table 2). More than 4 in 5 patients estimated that their likelihood of being alive in 2 months was 90% or more; the median physician estimate of 2-month survival was 80%. Three quarters of the patients estimated that their likelihood of 6-month survival was at least 90%, whereas the median physician estimate of 6-month survival was 50%. Overall, 113 patients (22%) had a do-not-resuscitate note or order written during their index hospital stay. Of these, the median time from study entry to the time that the order was written was 1 day (interquartile range 1 to 5 days). The median time from do-not-resuscitate order to death was 32 days (interquartile range 11 to 76 days). When asked about preferences concerning cardiopulmonary resuscitation, 263 of 288 patients interviewed were able to state a preference. Among the 232 patients unable to be interviewed about CPR, 76 of 79 surrogates stated a preference. Thus, of the 339 patients or surrogates who expressed preferences about CPR, 127 (37%) preferred not to have CPR; of these, 69 did not have a do-notresuscitate note.
Patient Outcomes Overall survival for the cohort was 78% at 2 months and 56% at 6 months (Table 3). Patients who had a do-notresuscitate order had a significantly increased mortality
Fifteen percent (n 5 78) of the cohort was treated in an intensive care unit (ICU), and 7% (n 5 35) required at least 1 day of mechanical ventilation. Patients who received ICU care had a 2-month survival of 87% and 6-month survival of 83%, whereas the 2-month survival was 77% and the 6-month survival was 52% for those who did not receive ICU care. Eighty-one percent of patients who received ICU care were treated at the two SUPPORT sites that offered partial liver resection for colon cancer metastatic to the liver. At the other three sites, only Table 3. Outcomes at 2 and 6 Months among Patients with Metastatic Colorectal Cancer 2 Months*
6 Months*
Outcome
Percentage
Alive Functional status No dependencies One dependency Two or more dependencies Quality of life Excellent Very good Good Fair or poor Anxiety and depression Anxiety score ,1.38† Depression score ,.687†
78% (406/520) 56% (291/520) 82% (172/210) 88% (125/142) 8% (17/210) 2% (3/142) 10% (21/210) 10% (14/142)
14% (30/209) 26% (55/209) 32% (67/209) 27% (57/209)
16% (22/140) 23% (32/140) 29% (41/140) 32% (45/140)
87% (180/207) 86% (120/140) 74% (153/207) 77% (108/140)
* Fraction indicates the number of patients with the factor/number of patients with available data. † Anxiety and depression scores are compared with previously published means for patients with various forms of cancer (35).
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
Volume 105 225
Metastatic Colorectal Cancer/Haidet et al
Table 4. Factors Independently Associated with Preferences for Cardiopulmonary Resuscitation among 310 Patients with Metastatic Colorectal Cancer*
Characteristic Younger age (per decade) Quality of life (good to excellent versus fair or poor) Absence of lung metastasis Patient estimate 2-month survival $90%
Odds Ratio*
95% Confidence Interval
1.6 2.2
1.4, 1.8 1.6, 2.9
1.9 2.2
1.4, 2.6 1.6, 3.0
* Model includes race, functional status, gender, income, weight loss, APACHE III acute physiology score, and SUPPORT model 2-month survival prediction.
15 of 207 patients received ICU care. The group that received ICU care did not differ significantly from the rest of the cohort in terms of age, gender, income, education, or patient estimate of prognosis.
Factors Associated with Preferences for Cardiopulmonary Resuscitation In multivariable logistic regression models that included race, functional status, gender, income, weight loss, illness severity as measured by the acute physiology score, and SUPPORT model predictions, four factors were found to be independently associated with patient preference for CPR (Table 4). Younger patients were more likely to want CPR, as were those who considered their quality of life to be good to excellent, those who did not have lung metastases, and those who estimated their 2-month survival to be 90% or greater. These results did not change when the model was adjusted for SUPPORT phase and intervention status. All four of these factors were associated with lower mortality rates at 2 months (P ,.05).
of the ability to discriminate nonsurvivors from survivors at 2 months. The area under the physician curve is significantly higher (0.80) than the area under the patient curve (0.66; P ,0.005), demonstrating that physicians are better than patients in predicting survival.
Effect of Patient-Physician Communication We examined the association of discussions between physician and patient (or surrogate) on their estimates of prognosis. Of 262 patients and physicians who indicated whether or not a discussion about prognosis had taken place, 155 (59%) reported doing so. The average absolute difference between 2-month survival estimates for physicians and patients was 20% for those who reported having a discussion, compared with 21% for those who reported no discussion (P 5 .35). Discussions between patients (or surrogates) and physicians also did not appear to improve physicians’ understanding of patients’ preferences for CPR. Overall, physicians incorrectly identified patients’ CPR preferences 30% of the time. These rates were 25% in the 64 pairs that had a discussion and 32% in the 136 pairs that did not have a discussion (P 5 .29). When we adjusted for the small differences in timing between the patient and physician interviews, our results were unchanged. The rates of physician misunderstanding of patient CPR preferences were 26% for surgeons, 35% for oncologists, and 12% for general medicine attending physicians (P 5 .10). Cardiologists and pulmonary/intensive care unit attending physicians were not analyzed owing to small numbers of patients.
Prognostic Estimates Physician estimates of survival were categorized into five categories corresponding to patient responses. In each category, the observed survival is somewhat closer to the physician estimate than the patient estimate (Figure). For example, the physicians of 123 subjects estimated their survival to be approximately 75%; these subjects had an observed survival of 86%. For the 60 patients who estimated their own survival to be approximately 75%, the observed survival was only 53%. Most patients placed themselves into the two best prognostic categories; the actual survival for these two groups was lower than they had predicted. We constructed receiver operating characteristic curves for patients’ and physicians’ 2-month survival estimates. The area under the curves represents a measure 226
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
Figure. Actual 2-month survival compared with physician and patient estimates. Example: In the 75% survival estimate category, 106 of the 123 patients whose physicians placed them in that category survived (actual survival 86%) whereas 32 of the 60 patients who placed themselves in this category survived (actual survival 53%).
Volume 105
Metastatic Colorectal Cancer/Haidet et al
DISCUSSION Among patients with metastatic colorectal cancer, almost 2 in 5 of those with a preference did not want CPR. Factors independently associated with this choice included older age, worse quality of life, presence of lung metastases, and a perception of a worse prognosis. All four of these factors were associated with higher mortality at 2 months, and may indicate that patient prognosis is the chief influence on decisions about CPR. Of the patients who preferred not to have CPR, more than half did not have a do-not-resuscitate order or note in their charts during the index hospital admission. Physicians incorrectly understood patients’ resuscitation preferences 30% of the time. A simple explanation for the mismatch between patient preferences and do-not-resuscitate orders or physician understanding of their preferences would be a lack of communication between physicians and patients. Our findings suggest otherwise. Where a conversation between physician and patient was reported, physician understanding of patients’ preferences was not better than in those instances when such a conversation had not occurred. These findings suggest that communication about resuscitation does not necessarily improve physician understanding of patient preferences. Similarly, patient-physician communication about prognosis did not appear to enhance patients’ understanding of their prognosis. Patients with metastatic colon cancer tend to overestimate their prognosis, whereas their physicians have a more accurate understanding of their survival probability. While the reasons for over-optimistic prognostic estimates by patients are understandable (41– 43), it may lead to a technological, less comfortable death (44,45) and unnecessary resource use. While the goal may be to optimize the patient’s mood and quality of life and to retain elements of hope, more effective communication about prognosis may facilitate decision making at the end of life. In a survey of 883 physicians and outpatients, Johnston et al (46) found that patients preferred to have discussions about advance directives at a younger age, and earlier in the course of disease and in the physician-patient relationship than did physicians. Both groups thought that it was the physician’s responsibility to initiate the discussion (46). That patients prefer earlier discussions, but do not prefer to initiate them, underscores the difficulties to effective communication between physicians and patients about end of life preferences and treatments. Patients with metastatic cancer tend to rate their quality of life higher than might be expected. Among patients with colorectal cancer and liver metastases, Allen-Mersh and colleagues (47) found that those treated with palliation reported spending 95% of their survival time with normal quality of life scores. Other studies have produced
similar results (48 –51). This suggests that, unlike other life-threatening conditions that may have more rapid courses or leave patients in a state of depressed health or mind, patients with metastatic colorectal cancer may enjoy a period of relatively normal health before dying or becoming incapacitated. In addition, the findings of shorter survival among patients with do-not-resuscitate orders, combined with relatively long periods of time from the writing of the do-not-resuscitate note or order to death—an average of 32 days in these patients compared with 3 days for the overall SUPPORT cohort (52)— implies that the do-not-resuscitate order represents earlier advance care planning, rather than a decision made just before death. It is in this period of relatively normal health when patients retain decision making capacity that improved patient-physician communication could be most effective in facilitating optimal end of life care. Our study has several limitations. First, SUPPORT data were collected from patients hospitalized in five tertiary care centers. Patients with metastatic cancer who had decided to forgo hospitalization were not included; such patients may make up a substantial proportion of patients with metastatic colorectal cancer. Also, since the five institutions are all major referral centers, patients with colorectal cancer may have been admitted to receive experimental therapies, and the doctors taking care of them may not have been their usual primary care doctors. Other physicians might have a better understanding of these patients’ preferences. Another factor that may have contributed to the observed misunderstanding of patients’ preferences by their doctors is the time difference between the doctor and the patient interview. However, patient preferences tend to be stable over time (53), and adjustment for time differences had no effect upon our findings. Although surrogate responses are imperfect (54), we decided to include surrogate responses, as surrogates often make CPR decisions in clinical practice. Our analysis of the effect of patient-physician communication on physician understanding of patient CPR preferences was based on 200 patient-physician pairs. We found a nonsignificant 7% difference between pairs that had discussed these preferences and those who had not. Although a larger study may have found a statistically significant difference, 25% of physicians misunderstood patient CPR preferences even when a discussion did take place. Function, mood, and quality of life were preserved among survivors at 2 and 6 months. However, these data were not collected among survivors who were not available for interview. These patients may have had lower function and quality of life; for example, surrogates reported slightly lower functioning than did patients who were interviewed. Nevertheless, the relatively long period of time between do-not-resuscitate orders and subsequent death suggests that some patients with metastatic
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
Volume 105 227
Metastatic Colorectal Cancer/Haidet et al
colorectal cancer and their physicians do consider end of life planning issues when function and quality of life are intact. However, the infrequent use of do-not-resuscitate orders among patients who do not want CPR suggests that effective advance planning is needed for more patients. In summary, despite the poor prognosis of patients with metastatic colorectal cancer, they often have preserved function and quality of life. Even during a hospitalization, a majority of these patients are able to express preferences about resuscitation. However, there is substantial misunderstanding between them and their physicians about prognosis and treatment. Both the quantity and quality of communication between physicians and patients may need improvement.
ACKNOWLEDGEMENTS The authors wish to thank Jane Soukup, MS, for assistance in data management, Erin Hartman, MS, for editorial support, and Candance Fifer for support in preparation of the manuscript.
REFERENCES 1. Parker SL, Tong L, Bolden S, Wingo PA. Cancer statistics. CA Cancer J Clin. 1996;46:5–27. 2. Toribara NW, Sleisenger MH. Screening for colorectal cancer. NEJM. 1995;332:861– 867. 3. Ferrante JM. Colorectal cancer screening. Med Clin North Am. 1996;80:27– 43. 4. Steele G Jr. Accomplishment and promise in the understanding and treatment of colorectal cancer. Lancet. 1993;342:1092–1096. 5. Sprangers MA, Taal BG, Aaronson NK, te Velde A. Quality of life in colorectal cancer stoma vs nonstoma patients. Dis Colon Rectum. 1995;38:361–369. 6. van Triest B, van Groeningen CJ, Pinedo HM. Current chemotherapeutic possibilities in the treatment of colorectal cancer. Eur J Cancer. 1995;31A(7-8):1193–1197. 7. Meropol NJ, Creaven PJ, Petrelli NJ. Metastatic colorectal cancer: advances in biochemical modulation and new drug development. Semin Oncol. 1995;22:509 –524. 8. Kemeny N. Current approaches to metastatic colorectal cancer. Semin Oncol. 1994;21(suppl)7:67–75. 9. Mandel JS, Bond JH, Church TR, et al. Reducing mortality from colorectal cancer by screening for fecal occult blood. Minnesota Colon Cancer Control Study. NEJM. 1993;326:653– 657. 10. Wade TP, Virgo KS, Li MJ, et al. Outcomes after detection of metastatic carcinoma of the colon and rectum in a national hospital system. J Am Coll Surg. 1996;182:353–361. 11. Saenz NC, Cady B, McDremott WV Jr, Steele GD Jr. Experience with colorectal carcinoma metastatic to the liver. Surg Clin North Am. 1989;69:361–370. 12. Patanaphan V, Salazar OM. Colorectal cancer: metastatic patterns and prognosis. South Med J. 1993;86:38 – 41. 13. Zalcberg JR, Friedlander ML, Collopy BT, et al. Treatment principles in advanced colorectal cancer. Aust N Z J Surg. 1996;66:202– 205. 14. The SUPPORT Principal Investigators. A controlled trial to improve care for seriously ill hospitalized patients; the study to understand prognoses and preferences for outcomes and risks of treatments. JAMA. 1995;274(20);1591–1598. 228
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
15. Bedell SE, Delbanco TL. Choices about cardiopulmonary resuscitation in the hospital. When do physicians talk with patients? NEJM. 1984;310:1089 –1093. 16. Uhlmann RF, Pearlman RA, Cain KC. Physicians’ and spouses’ predictions of elderly patients’ resuscitation preferences. J Gerontol. 1988;43:M115–M121. 17. Uhlmann RF, Pearlman RA, Cain KC. Understanding of elderly patients’ resuscitation preferences by physicians and nurses. West J Med. 1989;150:705–707. 18. Reilly BM, Magnussen CR, Ross J, et al. Can we talk? Inpatient discussions about advance directives in a community hospital; attending physicians’ attitudes, their inpatients’ wishes, and reported experience. Arch Intern Med. 1994;154:2299 –2308. 19. Murphy DJ, Burrows D, Santilli S, et al. The influence of the probability of survival on patients’ preferences regarding cardiopulmonary resuscitation. NEJM. 1994;330:545–549. 20. Murphy DJ, Cluff LE. SUPPORT: study to understand prognoses and preferences for outcomes and risks of treatments. Study design. J Clin Epidemiol. 1990;43(suppl). 21. Murphy DJ, Knaus WA, Lynn J. Study population in SUPPORT: patients (as defined by disease categories and mortality projections), surrogates, and physicians. J Clin Epidemiol. 1990;43(suppl): 11S–28S. 22. Kreling B, Robinson DK, Bergner M. Data collection strategies in SUPPORT. J Clin Epidemiol. 1990;43(suppl):5S–10S. 23. Phillips RS, Murphy DJ, Goldman L, Knaus WA. Patient characteristics in SUPPORT: disease specific clinical data. J Clin Epidemiol. 1990;43(suppl):41S– 46S. 24. Phillips RS, Knaus WA. Patient characteristics in SUPPORT: sociodemographics, admission diagnosis, co-morbidities and acute physiology score. J Clin Epidemiol. 1990;43(suppl):29S–32S. 25. Phillips RS, Goldman L, Bergner M. Patient characteristics in SUPPORT: activity status and cognitive function.J Clin Epidemiol. 1990;43(suppl):33S–36S. 26. Landefeld CS, Phillips RS, Bergner M. Patient characteristics in SUPPORT: functional status. J Clin Epidemiol. 1990;43(suppl)37S– 40S. 27. Coulton CJ. Decision making in SUPPORT: patient perceptions and preferences. J Clin Epidemiol. 1990;43(suppl):51S–54S. 28. Phillips RS, Wenger NS, Teno J, et al. Choices of seriously ill patients about cardiopulmonary resuscitation: correlates and outcomes. Am J Med. 1996;100:128 –137. 29. Katz S, Ford AB, Moskowitz RW, et al. Studies of illness in the aged. The index of ADL: A standardized measure of biological and psychosocial function. JAMA. 1963;185:914 –919. 30. Borsson B, Asberg KH. Katz index of independence in ADL: reliability and validity in short term care. Scand J Rehabil Med. 1984; 16:125–132. 31. Katz S, Apkom CA. A measure of primary sociobiological functions. Int J Health Serv. 1976;6:493–507. 32. Tsevat J, Dawson NV, Matchar DB. Assessing quality of life and preferences in the seriously ill using utility theory. J Clin Epidemiol. 1990;43(suppl):73S–78S. 33. McNair DM, Lorr M, Droppleman LF. EITS Manual For the Profile of Mood States. San Diego: Educational and Industrial Testing Service; 1971. 34. Shacham S. A shortened version of the Profile of Mood States. J Pers Assess. 1983;47:305–306. 35. Cella DF, Tross S, Orav EJ, et al. Mood states of patients after the diagnosis of cancer. J Psychosocial Oncol. 1989;7:45–53. 36. Connors AF Jr, Dawson NV, Arkes HR, Roach MJ. Decision making in SUPPORT: physician perceptions and preferences. J Clin Epidemiol. 1990;43(suppl):59S– 62S. 37. Knaus WA, Harrell FE Jr, Lynn J, et al. The SUPPORT prognostic model. objective estimates of survival for seriously ill hospital-
Volume 105
Metastatic Colorectal Cancer/Haidet et al
38. 39. 40.
41.
42. 43.
44.
45.
46.
ized adults. Study to understand prognoses and preferences for outcomes and risks of treatments. Ann Intern Med. 1995;122: 191–203. Oye RK, Landefeld CS, Jayes RL. Outcomes in SUPPORT. J Clin Epidemiol. 1990;43(suppl): 83S– 88S. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Statist Assoc. 1958;53:457– 481. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982;143: 29 –36. Weinstein ND. Unrealistic optimism about susceptibility to health problems: conclusions from a community-wide sample. J Behav Med. 1987;10:481–500. Weinstein ND. The precaution adoption process. Health Psychol. 1988;7:355–386. Weinstein ND, Klotz ML, Sandman PM. Optimistic biases in public perceptions of the risk from radon. Am J Public Health. 1988;78: 796 – 800. O’Day SJ, Weeks JC, Cook EF, et al. Relationship between cancer patients’ predictions of prognosis and their treatment preferences. Clin Res. 1993;41:579A. Weeks JC, Cook EF, O’Day SJ, et al. Relationship between cancer patients’ predictions of prognosis and their treatment preferences. JAMA. 1998;279:1709 –1714. Johnston SC, Pfeifer MP, McNutt R. The discussion about advance directives. patient and physician opinions regarding when and how
47.
48. 49.
50.
51.
52.
53.
54.
it should be conducted. End of life study group. Arch Intern Med. 1995;155:1025–1030. Allen-Mersh TG, Earlam S, Fordy C, et al. Quality of life and survival with continuous hepatic-artery floxuridine infusion for colorectal liver metastases. Lancet. 1994;344:1255–1260. Padilla GV, Grant MM, Lipsett J, et al. Health quality of life and colorectal cancer. Cancer. 1992;70(suppl 5):1450 –1456. Glimelius B, Hoffman K, Graf W, et al. Quality of life during chemotherapy in patients with symptomatic advanced colorectal cancer. The nordic gastrointestinal tumor adjuvant therapy group. Cancer. 1994;73:556 –562. Poon MA, O’Connell MJ, Moertel CG, et al. Biochemical modulation of fluorouracil: evidence of significant improvement of survival and quality of life in patients with advanced colorectal carcinoma. J Clin Oncol. 1989;7:1407–1418. Glimelius B, Graf W, Hoffmann K, et al. General condition of asymptomatic patients with advanced colorectal cancer receiving palliative chemotherapy. A longitudinal study. Acta Oncol. 1992;31: 645– 651. Hakim RB, Teno JM, Harrell FE, et al. Factors associated with donot-resuscitate orders: patients’ preferences, prognoses, and physicians’ judgements. Ann Intern Med. 1996;25:284 –293. Layde PM, Beam CA, Broste SK, et al. Surrogates predictions of seriously ill patients’ resuscitation preferences. Arch Fam Med. 1995;4:518 –523. Rosenfeld KE, Wenger NS, Phillips RS, et al. Factors associated with change in resuscitation preference of seriously ill patients. Arch Intern Med. 1996:156:1558 –1564.
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
Volume 105 229
ties 5 0) were high at study entry and remained so among interviewed survivors at 2 and 6 months. Of 339 patients with available information, 212 (63%) of 339 wanted CPR in the event of a cardiopulmonary arrest. Factors independently associated with preference for resuscitation included younger age, better quality of life, absence of lung metastases, and greater patient estimate of 2-month prognosis. Of the patients who preferred not to receive CPR, less than half had a do-not-resuscitate note or order written. Patients’ self-assessed prognoses were less accurate than those of their physicians. Physicians incorrectly identified patient CPR preferences in 30% of cases. Neither patient prognostication nor physician understanding of preferences were significantly better when discussions were reported between doctors and patients. CONCLUSIONS: A majority of patients with colorectal cancer have preferences regarding end of life care. The substantial misunderstanding between patients and their physicians about prognosis and treatment preferences appears not to be improved by direct communication. Future research focused on enhancing the effectiveness of communication between patients and physicians about end of life issues is needed. Am J Med. 1998;105:222–229. q1998 by Excerpta Medica, Inc.
C
55,000 persons died from colorectal cancer in the US in 1995 (1). Recent research on colorectal cancer has focused on screening and treatment of the disease (2– 6). While new treatments have been developed for persons with metastatic disease (7–9), few studies have focused on outcomes and end of life preferences among these patients (10). Anecdotal evidence suggests that patients with metastatic colorectal cancer often enjoy relatively good health while contemplating the poor prognosis associated with the diagnosis (11–14). Studies of other patient populations demonstrate that physicians misunderstand patient preferences for aggressive treatments at the end of life (15–17). In the case of colorectal cancer, the combination of an intact mental capacity with a known prognosis may provide the optimal setting in which to discuss end of life preferences (18,19), suggesting that physicians should have good understanding of end of life preferences among these patients. The purpose of this study was to characterize hospitalized patients diagnosed with metastatic colorectal cancer, to describe clinical outcomes at 2 and 6 months, and to
olorectal cancer is the fifth most common cause of cancer and the second leading cause of cancer death in the United States. Despite an increase in early detection and screening in recent years, more than
From the Division of General Medicine and Primary Care (PH, MBH, RD, RSP), Beth Israel Deaconess Medical Center, Boston, Massachusetts; the UCLA Medical Center (NW), Los Angeles, California; the George Washington University Medical Center (JL), Washington, DC; the University of Virginia School of Medicine (AFC), Charlottesville, Virginia; Marshfield Clinic (DR), Marshfield, Wisconsin; Duke University Medical Center (PSK), Durham, North Carolina; and The Dana Farber Cancer Institute (JCW), Boston, Massachusetts. Supported by the Robert Wood Johnson Foundation. The opinions and findings contained herein are those of the authors and do not necessarily represent the views of the Robert Wood Johnson Foundation or their Board of Trustees. Dr. Haidet was supported in part by National Research Service Award No. PE11001-09. An earlier version of this report was presented at the meeting of the Society of General Internal Medicine, May 2– 4, 1996, Washington, DC. Requests for reprints should be addressed to Russell S. Phillips, MD, Division of General Medicine and Primary Care, Beth Israel Deaconess Medical Center, 330 Brookline Avenue, Boston, Massachusetts 02215. Manuscript submitted October 26, 1997 and accepted in revised form June 1, 1998. 222
q1998 by Excerpta Medica, Inc. All rights reserved.
0002-9343/98/$19.00 PII S0002-9343(98)00242-3
Metastatic Colorectal Cancer/Haidet et al
examine perceptions of prognosis and preferences for end of life treatment. In addition, we sought to examine physician understanding of patient preferences and the effect of communication on physician and patient understanding of treatment preferences and prognosis.
METHODS Patient Population The data in this study were collected as part of SUPPORT, the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments. SUPPORT investigators collected information on patient, doctor, and surrogate decision making in seriously ill patients at five academic acute care teaching hospitals: Beth Israel Hospital in Boston, Massachusetts; MetroHealth Medical Center in Cleveland, Ohio; Duke University Medical Center in Durham, North Carolina; St. Joseph’s Hospital in Marshfield, Wisconsin; and the University of California Medical Center at Los Angeles (20). Enrollment occurred between 1989 and 1994 in two phases. Phase I was observational and enrolled patients between June 1989 and June 1991. Phase II involved an intervention and enrolled patients between January 1992 and January 1994. Clinicians randomly assigned to the intervention group were given information about their patients’ prognoses and preferences for care. Since no differences were observed in targeted outcomes between phase I and phase II control patients or between phase II control patients and intervention patients, all patients were combined into a single group (11). Hospitalized patients were eligible for SUPPORT if they were 18 years of age or older and met criteria specific to one of nine diagnostic categories. In the colorectal cancer category, patients were considered for enrollment if there was known metastasis to the liver. Diagnostic criteria were designed to identify patients at advanced stages of the illness with a 6-month survival estimated to be approximately 50% based on pilot studies performed by the SUPPORT investigators (21). In addition, at least 1 month’s time was required to elapse between the time of colorectal cancer diagnosis and study enrollment to allow patients sufficient time to formulate preferences for their care.
Data Collection Data were obtained through chart review by trained abstractors and by interviews with patients, physicians, and surrogates. Data abstraction and interview supervisors were trained centrally and quarterly site visits were conducted to monitor compliance with protocols and to audit interviews. Ten percent of all interviews were audited by a supervisor and 10% of all charts were reabstracted by a second abstractor (22). Overall reliability (defined as complete agreement between the two abstractors) was
89% for the acute physiology score of the APACHE III severity index abstracted from chart data. Information obtained by chart review included previous surgery, radiation and chemotherapy, number of liver metastases (single or multiple), date of initial diagnosis of colorectal cancer, date of initial diagnosis of liver metastasis, presence of lung or pleural metastasis, and physiologic variables used in the computation of the APACHE III acute physiology score (23,24). Data collected by patient interview included demographic information (level of education, income, age, gender, and race), functional status, whether any recent weight loss had occurred, anxiety, depression, quality of life, and the patient’s estimate for their probability of 2- and 6-month survival. Information on patients’ preferences for cardiopulmonary resuscitation in the event of cardiac arrest and whether they had a discussion with their doctor about CPR preferences was also collected (24 –27). When patients were not interviewed, surrogate responses for CPR preferences were substituted, as would be done in clinical practice. Surrogates were defined as the persons who would speak or participate in medical decision making on behalf of the patient if the patient was unable to do so. Physicians were asked whether they thought the patient would want to be resuscitated in the event of cardiac arrest. The wording of questions about preferences for CPR has been published previously (28). Functional status was measured using a modified version of the activities of daily living scale developed by Katz (29 –31), reflecting independence in seven self-care functions. Higher scores indicate a greater number of dependencies. Quality of life was reported on a scale of excellent, very good, good, fair or poor (32). Anxiety and depression were measured using shortened subscales of the Profile of Mood States (POMS) (33–35). Six anxiety items and eight depression items were used to construct average anxiety and depression scores on a scale ranging from 0 (no anxiety or depression) to 4 (worst anxiety or depression). Patients were asked to predict their probability of being alive at 2 and 6 months after the time of the interview by choosing one of the following responses: .90%, about 75%, about 50-50, about 25%, and ,10 % (27). Physicians were asked to estimate 2- and 6-month patient survival on a continuous scale of 0% to 100% (36) An overall SUPPORT prognostic model estimate for survival was developed and validated (37). For each study patient, model-based estimates for survival at 2 and 6 months were calculated. Survival was determined by contacting all patients or surrogates 180 days after study entry and by searching the National Death Index for survival beyond 6 months. Date of death, or condition on day 180, was confirmed for all patients. Quality of life, functional status, depression, and anxiety were obtained at 2 and 6 months by patient or
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
Volume 105 223
Metastatic Colorectal Cancer/Haidet et al
surrogate interview. Information on intensive care unit stay, mechanical ventilation, and CPR during the index admission was obtained by chart review (38).
Table 1. Characteristics of Patients with Metastatic Colorectal Cancer at Study Entry Characteristic
Statistical Methods Analyses were performed using a commercial statistical package (SAS Institute, Inc., Cary, North Carolina). Median and the interquartile range (25th to 75th percentile) were used to describe continuous variables, except in the case of anxiety and depression scores, where mean values are reported to facilitate comparison with published norms. Factors independently associated with CPR preferences were identified using multivariable logistic regression, and adjusted odds ratios are reported. Variables were included in the model based on clinical criteria and the results of prior studies. Subjects without a patient or surrogate CPR preference were not included in multivariable analyses. Patient survival was estimated using the product limit method (39). A proportional hazards model with a time varying covariate was used to assess the association of do-not-resuscitate status with survival. Physician and patient estimates of prognosis were compared by dividing the continuous physician estimates into five categories (.90%, 61% to 89%, 40% to 60%, 11% to 39%, and ,10%) designed to coincide with patient prognostic categories. Discrimination in predicting survival was estimated by calculating the area under receiver operating characteristic curves for both physicians and patients (40). Calibration was estimated by direct comparison of patient and physician estimates of 2-month survival to observed survival for the three highest response categories. Data from the two lowest response categories were collapsed into a single category due to small numbers. The effect of discussions between physicians and patients on agreement about prognosis was estimated by computing the absolute value of the difference between patient and physician estimates of prognosis; the patient’s (categorical) estimate was subtracted from the physician’s (continuous) estimate. The mean difference in pairs who had a discussion was then compared with the mean difference in those who did not have a discussion. Similarly, the effect of discussions between patients and physicians on physician understanding of CPR preferences was estimated by comparing the percentages of physicians with an incorrect understanding between pairs who did and did not have such a discussion. The effect of differences in timing of doctor and patient interviews was examined by comparing percentages of physicians with an incorrect understanding between pairs who were interviewed within 1 day of one another and pairs who were interviewed more than 1 day apart. The differences in prognostic estimates were compared using Wilcoxon rank sum tests, and differences in percentages of 224
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
Percent of Interviewed Patients with Factor*
Age 50 to 75 years White race Male gender Income #$25,000 per annum Education #12 years Previous treatment for colorectal cancer Surgery Radiation Chemotherapy Multiple liver metastases Diagnosis duration .6 months Liver metastasis .6 months Lung or pleural metastasis Weight loss Independent in activities of daily living† Acute physiology score #30, day 3 Quality of life good to excellent† Anxiety score ,1.38‡ Depression score ,.687‡
72% (374/520) 85% (440/520) 56% (291/520) 58% (303/520) 60% (311/520) 88% (460/520) 25% (129/520) 77% (400/520) 46% (233/503) 76% (394/519) 48% (250/520) 22% (116/520) 39% (164/425) 83% (268/322) 86% (446/520) 62% (188/304) 83% (230/277) 69% (189/276)
* Fraction indicates the number of patients with factor/number of patients with available data. † During the 2 weeks before admission. ‡ Anxiety and depression scores are compared with previously published means for patients with cancer (35).
physicians incorrect were compared using the chi-square test.
RESULTS Description of Study Population A total of 520 patients with colorectal cancer were enrolled in SUPPORT (Table 1). The median age was 64 at study entry (interquartile range 56 to 71) and the median years of education was 12 (interquartile range 11 to 14). Eighty-five percent were white, 9% black, and 4% were Hispanic. Most of the patients (96%) had received at least one cancer therapy. While the selection criteria required only the presence of a single liver metastasis, 57% had additional liver, lung, or pleural metastases. Among those who reported weight loss, the median loss was 12 pounds (interquartile range 7 to 22 pounds). The majority of the patients (60%) had an oncologist as their attending physician. The remaining physicians were surgeons (32%), general medicine physicians (6%), pulmonary/intensive care unit physicians (0.5%), and cardiologists (1.5%). Despite the stage of disease, the cohort had a relatively high level of functioning. The median number of dependencies at 2 weeks before study entry was 0, and 62%
Volume 105
Metastatic Colorectal Cancer/Haidet et al
Table 2. End of Life Preferences and Clinical Decisions among Patients with Metastatic Colorectal Cancer Preference/Decision Patient estimates for survival Two-month survival $90% Six-month survival $90% Physician estimates for survival Two-month survival $90% Six-month survival $90% Patient preferences for CPR Patient (or surrogate) preferred CPR Physician thinks patient (or surrogate) prefers CPR Do-not-resuscitate note or order in chart Advance directives Have power of attorney Have living will
Percentage* 82% (207/252) 75% (176/235) 22% (89/406) 5% (20/406) 63% (214/339) 55% (173/314) 22% (114/520) 43% (83/192) 35% (97/276)
* Fraction indicates the number of patients with factor/number of patients with available data. CPR 5 cardiopulmonary resuscitation.
(hazard ratio 5 4.0, 95% confidence interval 3.2 to 5.0). Based on the SUPPORT prognostic model, the mean 2-month predicted survival for the 113 patients with donot-resuscitate orders was .61, significantly less than the .78 prediction for those without such orders (P ,.001). Similar results were obtained using 6-month predictions. The survivors who were interviewed at 2 (n 5 210) and 6 (n 5 142) months had relatively good functional status, with a median number of disabilities of 0 at both times. Among living patients who were not available for interview but had surrogate responses at 2 (n 5 65) and 6 (n 5 38) months, the median number of disabilities was 1 at 2 and 6 months (interquartile range 5 0 to 5 at 2 months and 0 to 3 at 6 months). Quality of life was rated by interviewed patients as good to excellent in 73% at 2 months, and 68% at 6 months. Anxiety and depression scores remained low at follow-up with mean (ISD) anxiety scores of .67 6 .70 at 2 months, .66 6 .74 at 6 months, and mean depression scores of .46 6 .66 at 2 months and .44 6 .68 at 6 months.
Use of Intensive Care reported quality of life as good to excellent. In addition, mean scores on the Profile of Mood States subscales were low as compared with other cancer patients (35), indicating low levels of psychologic distress due to either anxiety or depression.
Patient Preferences and Decision Making Patients had high estimates of 2- and 6-month survival as compared to their physicians (Table 2). More than 4 in 5 patients estimated that their likelihood of being alive in 2 months was 90% or more; the median physician estimate of 2-month survival was 80%. Three quarters of the patients estimated that their likelihood of 6-month survival was at least 90%, whereas the median physician estimate of 6-month survival was 50%. Overall, 113 patients (22%) had a do-not-resuscitate note or order written during their index hospital stay. Of these, the median time from study entry to the time that the order was written was 1 day (interquartile range 1 to 5 days). The median time from do-not-resuscitate order to death was 32 days (interquartile range 11 to 76 days). When asked about preferences concerning cardiopulmonary resuscitation, 263 of 288 patients interviewed were able to state a preference. Among the 232 patients unable to be interviewed about CPR, 76 of 79 surrogates stated a preference. Thus, of the 339 patients or surrogates who expressed preferences about CPR, 127 (37%) preferred not to have CPR; of these, 69 did not have a do-notresuscitate note.
Patient Outcomes Overall survival for the cohort was 78% at 2 months and 56% at 6 months (Table 3). Patients who had a do-notresuscitate order had a significantly increased mortality
Fifteen percent (n 5 78) of the cohort was treated in an intensive care unit (ICU), and 7% (n 5 35) required at least 1 day of mechanical ventilation. Patients who received ICU care had a 2-month survival of 87% and 6-month survival of 83%, whereas the 2-month survival was 77% and the 6-month survival was 52% for those who did not receive ICU care. Eighty-one percent of patients who received ICU care were treated at the two SUPPORT sites that offered partial liver resection for colon cancer metastatic to the liver. At the other three sites, only Table 3. Outcomes at 2 and 6 Months among Patients with Metastatic Colorectal Cancer 2 Months*
6 Months*
Outcome
Percentage
Alive Functional status No dependencies One dependency Two or more dependencies Quality of life Excellent Very good Good Fair or poor Anxiety and depression Anxiety score ,1.38† Depression score ,.687†
78% (406/520) 56% (291/520) 82% (172/210) 88% (125/142) 8% (17/210) 2% (3/142) 10% (21/210) 10% (14/142)
14% (30/209) 26% (55/209) 32% (67/209) 27% (57/209)
16% (22/140) 23% (32/140) 29% (41/140) 32% (45/140)
87% (180/207) 86% (120/140) 74% (153/207) 77% (108/140)
* Fraction indicates the number of patients with the factor/number of patients with available data. † Anxiety and depression scores are compared with previously published means for patients with various forms of cancer (35).
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
Volume 105 225
Metastatic Colorectal Cancer/Haidet et al
Table 4. Factors Independently Associated with Preferences for Cardiopulmonary Resuscitation among 310 Patients with Metastatic Colorectal Cancer*
Characteristic Younger age (per decade) Quality of life (good to excellent versus fair or poor) Absence of lung metastasis Patient estimate 2-month survival $90%
Odds Ratio*
95% Confidence Interval
1.6 2.2
1.4, 1.8 1.6, 2.9
1.9 2.2
1.4, 2.6 1.6, 3.0
* Model includes race, functional status, gender, income, weight loss, APACHE III acute physiology score, and SUPPORT model 2-month survival prediction.
15 of 207 patients received ICU care. The group that received ICU care did not differ significantly from the rest of the cohort in terms of age, gender, income, education, or patient estimate of prognosis.
Factors Associated with Preferences for Cardiopulmonary Resuscitation In multivariable logistic regression models that included race, functional status, gender, income, weight loss, illness severity as measured by the acute physiology score, and SUPPORT model predictions, four factors were found to be independently associated with patient preference for CPR (Table 4). Younger patients were more likely to want CPR, as were those who considered their quality of life to be good to excellent, those who did not have lung metastases, and those who estimated their 2-month survival to be 90% or greater. These results did not change when the model was adjusted for SUPPORT phase and intervention status. All four of these factors were associated with lower mortality rates at 2 months (P ,.05).
of the ability to discriminate nonsurvivors from survivors at 2 months. The area under the physician curve is significantly higher (0.80) than the area under the patient curve (0.66; P ,0.005), demonstrating that physicians are better than patients in predicting survival.
Effect of Patient-Physician Communication We examined the association of discussions between physician and patient (or surrogate) on their estimates of prognosis. Of 262 patients and physicians who indicated whether or not a discussion about prognosis had taken place, 155 (59%) reported doing so. The average absolute difference between 2-month survival estimates for physicians and patients was 20% for those who reported having a discussion, compared with 21% for those who reported no discussion (P 5 .35). Discussions between patients (or surrogates) and physicians also did not appear to improve physicians’ understanding of patients’ preferences for CPR. Overall, physicians incorrectly identified patients’ CPR preferences 30% of the time. These rates were 25% in the 64 pairs that had a discussion and 32% in the 136 pairs that did not have a discussion (P 5 .29). When we adjusted for the small differences in timing between the patient and physician interviews, our results were unchanged. The rates of physician misunderstanding of patient CPR preferences were 26% for surgeons, 35% for oncologists, and 12% for general medicine attending physicians (P 5 .10). Cardiologists and pulmonary/intensive care unit attending physicians were not analyzed owing to small numbers of patients.
Prognostic Estimates Physician estimates of survival were categorized into five categories corresponding to patient responses. In each category, the observed survival is somewhat closer to the physician estimate than the patient estimate (Figure). For example, the physicians of 123 subjects estimated their survival to be approximately 75%; these subjects had an observed survival of 86%. For the 60 patients who estimated their own survival to be approximately 75%, the observed survival was only 53%. Most patients placed themselves into the two best prognostic categories; the actual survival for these two groups was lower than they had predicted. We constructed receiver operating characteristic curves for patients’ and physicians’ 2-month survival estimates. The area under the curves represents a measure 226
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
Figure. Actual 2-month survival compared with physician and patient estimates. Example: In the 75% survival estimate category, 106 of the 123 patients whose physicians placed them in that category survived (actual survival 86%) whereas 32 of the 60 patients who placed themselves in this category survived (actual survival 53%).
Volume 105
Metastatic Colorectal Cancer/Haidet et al
DISCUSSION Among patients with metastatic colorectal cancer, almost 2 in 5 of those with a preference did not want CPR. Factors independently associated with this choice included older age, worse quality of life, presence of lung metastases, and a perception of a worse prognosis. All four of these factors were associated with higher mortality at 2 months, and may indicate that patient prognosis is the chief influence on decisions about CPR. Of the patients who preferred not to have CPR, more than half did not have a do-not-resuscitate order or note in their charts during the index hospital admission. Physicians incorrectly understood patients’ resuscitation preferences 30% of the time. A simple explanation for the mismatch between patient preferences and do-not-resuscitate orders or physician understanding of their preferences would be a lack of communication between physicians and patients. Our findings suggest otherwise. Where a conversation between physician and patient was reported, physician understanding of patients’ preferences was not better than in those instances when such a conversation had not occurred. These findings suggest that communication about resuscitation does not necessarily improve physician understanding of patient preferences. Similarly, patient-physician communication about prognosis did not appear to enhance patients’ understanding of their prognosis. Patients with metastatic colon cancer tend to overestimate their prognosis, whereas their physicians have a more accurate understanding of their survival probability. While the reasons for over-optimistic prognostic estimates by patients are understandable (41– 43), it may lead to a technological, less comfortable death (44,45) and unnecessary resource use. While the goal may be to optimize the patient’s mood and quality of life and to retain elements of hope, more effective communication about prognosis may facilitate decision making at the end of life. In a survey of 883 physicians and outpatients, Johnston et al (46) found that patients preferred to have discussions about advance directives at a younger age, and earlier in the course of disease and in the physician-patient relationship than did physicians. Both groups thought that it was the physician’s responsibility to initiate the discussion (46). That patients prefer earlier discussions, but do not prefer to initiate them, underscores the difficulties to effective communication between physicians and patients about end of life preferences and treatments. Patients with metastatic cancer tend to rate their quality of life higher than might be expected. Among patients with colorectal cancer and liver metastases, Allen-Mersh and colleagues (47) found that those treated with palliation reported spending 95% of their survival time with normal quality of life scores. Other studies have produced
similar results (48 –51). This suggests that, unlike other life-threatening conditions that may have more rapid courses or leave patients in a state of depressed health or mind, patients with metastatic colorectal cancer may enjoy a period of relatively normal health before dying or becoming incapacitated. In addition, the findings of shorter survival among patients with do-not-resuscitate orders, combined with relatively long periods of time from the writing of the do-not-resuscitate note or order to death—an average of 32 days in these patients compared with 3 days for the overall SUPPORT cohort (52)— implies that the do-not-resuscitate order represents earlier advance care planning, rather than a decision made just before death. It is in this period of relatively normal health when patients retain decision making capacity that improved patient-physician communication could be most effective in facilitating optimal end of life care. Our study has several limitations. First, SUPPORT data were collected from patients hospitalized in five tertiary care centers. Patients with metastatic cancer who had decided to forgo hospitalization were not included; such patients may make up a substantial proportion of patients with metastatic colorectal cancer. Also, since the five institutions are all major referral centers, patients with colorectal cancer may have been admitted to receive experimental therapies, and the doctors taking care of them may not have been their usual primary care doctors. Other physicians might have a better understanding of these patients’ preferences. Another factor that may have contributed to the observed misunderstanding of patients’ preferences by their doctors is the time difference between the doctor and the patient interview. However, patient preferences tend to be stable over time (53), and adjustment for time differences had no effect upon our findings. Although surrogate responses are imperfect (54), we decided to include surrogate responses, as surrogates often make CPR decisions in clinical practice. Our analysis of the effect of patient-physician communication on physician understanding of patient CPR preferences was based on 200 patient-physician pairs. We found a nonsignificant 7% difference between pairs that had discussed these preferences and those who had not. Although a larger study may have found a statistically significant difference, 25% of physicians misunderstood patient CPR preferences even when a discussion did take place. Function, mood, and quality of life were preserved among survivors at 2 and 6 months. However, these data were not collected among survivors who were not available for interview. These patients may have had lower function and quality of life; for example, surrogates reported slightly lower functioning than did patients who were interviewed. Nevertheless, the relatively long period of time between do-not-resuscitate orders and subsequent death suggests that some patients with metastatic
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
Volume 105 227
Metastatic Colorectal Cancer/Haidet et al
colorectal cancer and their physicians do consider end of life planning issues when function and quality of life are intact. However, the infrequent use of do-not-resuscitate orders among patients who do not want CPR suggests that effective advance planning is needed for more patients. In summary, despite the poor prognosis of patients with metastatic colorectal cancer, they often have preserved function and quality of life. Even during a hospitalization, a majority of these patients are able to express preferences about resuscitation. However, there is substantial misunderstanding between them and their physicians about prognosis and treatment. Both the quantity and quality of communication between physicians and patients may need improvement.
ACKNOWLEDGEMENTS The authors wish to thank Jane Soukup, MS, for assistance in data management, Erin Hartman, MS, for editorial support, and Candance Fifer for support in preparation of the manuscript.
REFERENCES 1. Parker SL, Tong L, Bolden S, Wingo PA. Cancer statistics. CA Cancer J Clin. 1996;46:5–27. 2. Toribara NW, Sleisenger MH. Screening for colorectal cancer. NEJM. 1995;332:861– 867. 3. Ferrante JM. Colorectal cancer screening. Med Clin North Am. 1996;80:27– 43. 4. Steele G Jr. Accomplishment and promise in the understanding and treatment of colorectal cancer. Lancet. 1993;342:1092–1096. 5. Sprangers MA, Taal BG, Aaronson NK, te Velde A. Quality of life in colorectal cancer stoma vs nonstoma patients. Dis Colon Rectum. 1995;38:361–369. 6. van Triest B, van Groeningen CJ, Pinedo HM. Current chemotherapeutic possibilities in the treatment of colorectal cancer. Eur J Cancer. 1995;31A(7-8):1193–1197. 7. Meropol NJ, Creaven PJ, Petrelli NJ. Metastatic colorectal cancer: advances in biochemical modulation and new drug development. Semin Oncol. 1995;22:509 –524. 8. Kemeny N. Current approaches to metastatic colorectal cancer. Semin Oncol. 1994;21(suppl)7:67–75. 9. Mandel JS, Bond JH, Church TR, et al. Reducing mortality from colorectal cancer by screening for fecal occult blood. Minnesota Colon Cancer Control Study. NEJM. 1993;326:653– 657. 10. Wade TP, Virgo KS, Li MJ, et al. Outcomes after detection of metastatic carcinoma of the colon and rectum in a national hospital system. J Am Coll Surg. 1996;182:353–361. 11. Saenz NC, Cady B, McDremott WV Jr, Steele GD Jr. Experience with colorectal carcinoma metastatic to the liver. Surg Clin North Am. 1989;69:361–370. 12. Patanaphan V, Salazar OM. Colorectal cancer: metastatic patterns and prognosis. South Med J. 1993;86:38 – 41. 13. Zalcberg JR, Friedlander ML, Collopy BT, et al. Treatment principles in advanced colorectal cancer. Aust N Z J Surg. 1996;66:202– 205. 14. The SUPPORT Principal Investigators. A controlled trial to improve care for seriously ill hospitalized patients; the study to understand prognoses and preferences for outcomes and risks of treatments. JAMA. 1995;274(20);1591–1598. 228
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
15. Bedell SE, Delbanco TL. Choices about cardiopulmonary resuscitation in the hospital. When do physicians talk with patients? NEJM. 1984;310:1089 –1093. 16. Uhlmann RF, Pearlman RA, Cain KC. Physicians’ and spouses’ predictions of elderly patients’ resuscitation preferences. J Gerontol. 1988;43:M115–M121. 17. Uhlmann RF, Pearlman RA, Cain KC. Understanding of elderly patients’ resuscitation preferences by physicians and nurses. West J Med. 1989;150:705–707. 18. Reilly BM, Magnussen CR, Ross J, et al. Can we talk? Inpatient discussions about advance directives in a community hospital; attending physicians’ attitudes, their inpatients’ wishes, and reported experience. Arch Intern Med. 1994;154:2299 –2308. 19. Murphy DJ, Burrows D, Santilli S, et al. The influence of the probability of survival on patients’ preferences regarding cardiopulmonary resuscitation. NEJM. 1994;330:545–549. 20. Murphy DJ, Cluff LE. SUPPORT: study to understand prognoses and preferences for outcomes and risks of treatments. Study design. J Clin Epidemiol. 1990;43(suppl). 21. Murphy DJ, Knaus WA, Lynn J. Study population in SUPPORT: patients (as defined by disease categories and mortality projections), surrogates, and physicians. J Clin Epidemiol. 1990;43(suppl): 11S–28S. 22. Kreling B, Robinson DK, Bergner M. Data collection strategies in SUPPORT. J Clin Epidemiol. 1990;43(suppl):5S–10S. 23. Phillips RS, Murphy DJ, Goldman L, Knaus WA. Patient characteristics in SUPPORT: disease specific clinical data. J Clin Epidemiol. 1990;43(suppl):41S– 46S. 24. Phillips RS, Knaus WA. Patient characteristics in SUPPORT: sociodemographics, admission diagnosis, co-morbidities and acute physiology score. J Clin Epidemiol. 1990;43(suppl):29S–32S. 25. Phillips RS, Goldman L, Bergner M. Patient characteristics in SUPPORT: activity status and cognitive function.J Clin Epidemiol. 1990;43(suppl):33S–36S. 26. Landefeld CS, Phillips RS, Bergner M. Patient characteristics in SUPPORT: functional status. J Clin Epidemiol. 1990;43(suppl)37S– 40S. 27. Coulton CJ. Decision making in SUPPORT: patient perceptions and preferences. J Clin Epidemiol. 1990;43(suppl):51S–54S. 28. Phillips RS, Wenger NS, Teno J, et al. Choices of seriously ill patients about cardiopulmonary resuscitation: correlates and outcomes. Am J Med. 1996;100:128 –137. 29. Katz S, Ford AB, Moskowitz RW, et al. Studies of illness in the aged. The index of ADL: A standardized measure of biological and psychosocial function. JAMA. 1963;185:914 –919. 30. Borsson B, Asberg KH. Katz index of independence in ADL: reliability and validity in short term care. Scand J Rehabil Med. 1984; 16:125–132. 31. Katz S, Apkom CA. A measure of primary sociobiological functions. Int J Health Serv. 1976;6:493–507. 32. Tsevat J, Dawson NV, Matchar DB. Assessing quality of life and preferences in the seriously ill using utility theory. J Clin Epidemiol. 1990;43(suppl):73S–78S. 33. McNair DM, Lorr M, Droppleman LF. EITS Manual For the Profile of Mood States. San Diego: Educational and Industrial Testing Service; 1971. 34. Shacham S. A shortened version of the Profile of Mood States. J Pers Assess. 1983;47:305–306. 35. Cella DF, Tross S, Orav EJ, et al. Mood states of patients after the diagnosis of cancer. J Psychosocial Oncol. 1989;7:45–53. 36. Connors AF Jr, Dawson NV, Arkes HR, Roach MJ. Decision making in SUPPORT: physician perceptions and preferences. J Clin Epidemiol. 1990;43(suppl):59S– 62S. 37. Knaus WA, Harrell FE Jr, Lynn J, et al. The SUPPORT prognostic model. objective estimates of survival for seriously ill hospital-
Volume 105
Metastatic Colorectal Cancer/Haidet et al
38. 39. 40.
41.
42. 43.
44.
45.
46.
ized adults. Study to understand prognoses and preferences for outcomes and risks of treatments. Ann Intern Med. 1995;122: 191–203. Oye RK, Landefeld CS, Jayes RL. Outcomes in SUPPORT. J Clin Epidemiol. 1990;43(suppl): 83S– 88S. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Statist Assoc. 1958;53:457– 481. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982;143: 29 –36. Weinstein ND. Unrealistic optimism about susceptibility to health problems: conclusions from a community-wide sample. J Behav Med. 1987;10:481–500. Weinstein ND. The precaution adoption process. Health Psychol. 1988;7:355–386. Weinstein ND, Klotz ML, Sandman PM. Optimistic biases in public perceptions of the risk from radon. Am J Public Health. 1988;78: 796 – 800. O’Day SJ, Weeks JC, Cook EF, et al. Relationship between cancer patients’ predictions of prognosis and their treatment preferences. Clin Res. 1993;41:579A. Weeks JC, Cook EF, O’Day SJ, et al. Relationship between cancer patients’ predictions of prognosis and their treatment preferences. JAMA. 1998;279:1709 –1714. Johnston SC, Pfeifer MP, McNutt R. The discussion about advance directives. patient and physician opinions regarding when and how
47.
48. 49.
50.
51.
52.
53.
54.
it should be conducted. End of life study group. Arch Intern Med. 1995;155:1025–1030. Allen-Mersh TG, Earlam S, Fordy C, et al. Quality of life and survival with continuous hepatic-artery floxuridine infusion for colorectal liver metastases. Lancet. 1994;344:1255–1260. Padilla GV, Grant MM, Lipsett J, et al. Health quality of life and colorectal cancer. Cancer. 1992;70(suppl 5):1450 –1456. Glimelius B, Hoffman K, Graf W, et al. Quality of life during chemotherapy in patients with symptomatic advanced colorectal cancer. The nordic gastrointestinal tumor adjuvant therapy group. Cancer. 1994;73:556 –562. Poon MA, O’Connell MJ, Moertel CG, et al. Biochemical modulation of fluorouracil: evidence of significant improvement of survival and quality of life in patients with advanced colorectal carcinoma. J Clin Oncol. 1989;7:1407–1418. Glimelius B, Graf W, Hoffmann K, et al. General condition of asymptomatic patients with advanced colorectal cancer receiving palliative chemotherapy. A longitudinal study. Acta Oncol. 1992;31: 645– 651. Hakim RB, Teno JM, Harrell FE, et al. Factors associated with donot-resuscitate orders: patients’ preferences, prognoses, and physicians’ judgements. Ann Intern Med. 1996;25:284 –293. Layde PM, Beam CA, Broste SK, et al. Surrogates predictions of seriously ill patients’ resuscitation preferences. Arch Fam Med. 1995;4:518 –523. Rosenfeld KE, Wenger NS, Phillips RS, et al. Factors associated with change in resuscitation preference of seriously ill patients. Arch Intern Med. 1996:156:1558 –1564.
September 1998
THE AMERICAN JOURNAL OF MEDICINEt
Volume 105 229