If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Address correspondence to: Koji Amano, MD, Department of Palliative Medicine, Palliative Care Team, Osaka City General Hospital, 2-13-22 Miyakojima-hondori, Miyakojima-ku, Osaka 534-0021, Japan.
Accurate prognoses are needed for patients with advanced cancer.
Objectives
To evaluate the accuracy of physicians' clinical predictions of survival (CPS) and assess the relationship between CPS and actual survival (AS) in patients with advanced cancer in palliative care units, hospital palliative care teams, and home palliative care services, as well as those receiving chemotherapy.
Methods
This was a multicenter prospective cohort study conducted in 58 palliative care service centers in Japan. The palliative care physicians evaluated patients on the first day of admission and followed up all patients to their death or six months after enrollment. We evaluated the accuracy of CPS and assessed the relationship between CPS and AS in the four groups.
Results
We obtained a total of 2036 patients: 470, 764, 404, and 398 in hospital palliative care teams, palliative care units, home palliative care services, and chemotherapy, respectively. The proportion of accurate CPS (0.67–1.33 times AS) was 35% (95% CI 33–37%) in the total sample and ranged from 32% to 39% in each setting. While the proportion of patients living longer than CPS (pessimistic CPS) was 20% (95% CI 18–22%) in the total sample, ranging from 15% to 23% in each setting, the proportion of patients living shorter than CPS (optimistic CPS) was 45% (95% CI 43–47%) in the total sample, ranging from 43% to 49% in each setting.
Conclusion
Physicians tend to overestimate when predicting survival in all palliative care patients, including those receiving chemotherapy.
Accurate prognoses are needed for patients with advanced cancer to determine appropriate end-of-life care, such as medical interventions and timing of referral to hospice care programs.
Previous surveys have revealed that the majority of patients want to know their prognoses, and their prognostic estimates were less accurate than those of health care professionals.
but in reality, physicians' clinical predictions of survival (CPS) for patients with advanced cancer are often inaccurate, which considerably impacts end-of-life care.
Prognostic factors in advanced cancer patients: evidence-based clinical recommendations—a study by the steering committee of the European Association for Palliative Care.
Previous studies of this issue have several limitations regarding clinical implications because they did not survey palliative care physicians' predictions exclusively and they did not involve simultaneous investigations in a variety of clinical settings, including palliative care units, hospital palliative care teams, and home palliative care services.
Prognostic factors in advanced cancer patients: evidence-based clinical recommendations—a study by the steering committee of the European Association for Palliative Care.
More importantly, anticancer treatment is increasingly provided for patients with advanced cancer, and the impact of timing and setting of palliative care referral on quality of end-of-life care has become highlighted.
Therefore, there is a strong need for oncologists and palliative care physicians to accurately predict the survival of patients receiving anticancer treatment in palliative settings.
The objectives of this study were to evaluate the accuracy of CPS and assess the relationship between CPS and actual survival (AS) in patients with advanced cancer in a variety of palliative settings including palliative care units, hospital palliative care teams, and home palliative care services, as well as in patients receiving chemotherapy.
Methods
This is a subanalysis of a multicenter prospective cohort study conducted in 58 palliative care service centers in Japan from September 2012 through April 2014.
The participating units included 16 palliative care units, 19 hospital palliative care teams, and 23 home palliative care services. The palliative care physicians evaluated patients and recorded a point estimate (e.g., 25 days) as the CPS on the first day of admission and followed up all patients to their death or six months after their enrollment. The physicians were palliative care consultants on a palliative care team, attending physicians in palliative care units, and home palliative care physicians primarily responsible for the patients. Patient demographics and clinical characteristics, such as gender, site of primary cancer, metastatic disease, and recent anticancer treatment in the past 30 days (i.e., chemotherapy, hormone therapy, and radiotherapy) were obtained.
Consecutive eligible patients were enrolled in this study if they had been newly referred to the participating institutions during the study period. All institutions were asked to take a sample of data consecutively, up to the designated number of patients of 20, 40, 60, 80, or 100, according to the size of the institution. Inclusion criteria for this study included 1) adult patients, 2) patients diagnosed with locally extensive or metastatic cancer (including hematological neoplasm), 3) patients admitted to palliative care units, receiving care from hospital palliative care teams, or receiving home palliative care services. Patients receiving chemotherapy, hormone therapy, and radiotherapy were not excluded.
This study was conducted in accordance with the ethical standards of the Helsinki Declaration and the ethical guidelines for epidemiological research of the Ministry of Health, Labor and Welfare in Japan, and approved by the local Institutional Review Boards for all participating institutions.
Statistical Analyses
Of all patients enrolled, we analyzed the patients who died during the study period. We then classified four patient groups: 1) patients from hospital palliative care teams (not receiving chemotherapy), 2) patients admitted to palliative care units (not receiving chemotherapy), 3) patients receiving home palliative care services (not receiving chemotherapy), and 4) patients receiving chemotherapy.
To evaluate the accuracy of CPS and the relationship between CPS and AS, we plotted CPS and AS for each setting. Second, the proportion with 95% CIs of “accurate CPS,” the proportion of “pessimistic CPS,” and the proportion of “optimistic CPS” were calculated. For comparability with other studies,
we defined a CPS as accurate if it was within 0.67–1.33 times the AS, that is, CPS was considered accurate if the error was less than ±33%, and the error was calculated as (CPS–AS)/AS × 100. We defined patients living longer than CPS as “pessimistic CPS” and patients living shorter than CPS as “optimistic CPS.” In addition, Spearman's correlation coefficient between CPS and AS was calculated; Spearman's correlation coefficient: <0.2 poor agreement, 0.21–0.4 fair, 0.41–0.6 moderate, 0.61–0.8 good, 0.81–0.99 very good, and 1 perfect.
We analyzed differences in predicting survival for certain primary cancers or location of metastases and the accuracy of CPS in all settings, including patients still alive at the end of the study period. We explored whether the accuracy differed among patients with different survival periods; the accuracy of each AS was assessed. AS was categorized into 1) <7 days, 2) 7–13 days, 3) 14–27 days, 4) 28–41 days, 5) 42–55 days, 6) 56–83 days, and 7) 84 days+.
All results were considered to be statistically significant if the P-value was less than 0.05. All analyses were performed using IBM SPSS v. 22.0 (SPSS Inc., Chicago, IL).
Results
A total of 2426 subjects met the inclusion criteria. Of these, we excluded 352 patients who were still alive at the end of the study period, 36 patients whose CPS was not available and two patients whose information about chemotherapy was not available. Thus, we obtained a total of 2036 patients for this study: 470 from hospital palliative care teams, 764 from the palliative care units, 404 from the home palliative care services, and 398 patients who had received chemotherapy in all settings. Of the 398 patients in the chemotherapy group, 264 patients (66%) were supported by hospital palliative care teams, 71 patients (18%) were admitted to palliative care units, and 63 patients (16%) were supported by home palliative care services.
The characteristics of the study subjects are summarized in Table 1. Mean ages were 68, 71, 74, and 63 years in hospital palliative care teams, palliative care units, home palliative care services, and chemotherapy, respectively. In the three groups (excluding the chemotherapy group), upper and lower gastrointestinal tracts were the most common sites of primary cancer, followed by respiratory/intrathoracic. In the chemotherapy group, upper and lower gastrointestinal tracts were the most common, followed by pancreas/liver/bile duct. The median survival times were 29, 23, 31, and 34 days in hospital palliative care teams, palliative care units, home palliative care services, and chemotherapy, respectively.
Table 1Demographic and Clinical Characteristics of 2036 Patients in Three Palliative Care Settings
Characteristics
Hospital Palliative Care Teams (n = 470)
Palliative Care Units (n = 764)
Home Palliative Care Services (n = 404)
Chemotherapy in All Settings (n = 398)
No. or Mean
% or SD
No. or Mean
% or SD
No. or Mean
% or SD
No. or Mean
% or SD
Age (yrs)
68
13
71
12
74
12
63
12
Male sex
274
58
411
54
265
66
243
61
Site of primary cancer
Upper and lower gastrointestinal tracts
132
29
189
25
120
31
123
31
Respiratory/intrathoracic
120
26
155
21
82
21
76
19
Pancreas/liver/bile duct
83
18
153
21
80
21
92
23
Breast
18
3.9
48
6.4
14
3.6
20
5.1
Bladder/kidney/urinary tracts
15
3.3
35
4.7
18
4.6
21
5.4
Male genital organs (including prostate)
6
1.3
13
1.7
14
3.6
6
1.5
Female genital organs (including ovary, uterus)
35
7.6
53
7.1
15
3.8
11
2.8
Head and neck
13
2.8
40
5.4
11
2.8
8
2.0
Hematological
8
1.7
3
0.4
11
2.8
16
4.1
Others
31
6.7
56
7.5
25
6.4
19
4.8
Metastatic site
Number with metastatic disease
387
82
594
78
305
76
360
90
Liver
193
41
269
35
139
35
174
44
Bone
141
30
213
28
91
23
136
34
Lung
164
35
267
35
127
32
144
36
Central nervous system
53
11
97
13
38
9.5
52
13
Eastern Cooperative Oncology Group Performance Status
0–1
33
7.0
24
3.2
23
5.7
52
13
2
84
18
102
13
62
15
94
24
3
177
38
309
40
155
39
143
36
4
176
37
329
43
163
40
108
27
Anticancer treatment
Chemotherapy
0
0
0
0
0
0
398
100
Hormone therapy
7
1.5
6
0.8
6
1.4
1
0.3
Radiotherapy
43
9.2
3
0.4
11
2.7
30
7.5
Median survival (d, 95% CI)
29 (25–33)
23 (21–25)
31 (27–35)
34 (29–39)
The sums of some percentages do not add up to 100% because of missing values.
The relationship between CPS and AS is shown in Figure 1. Points on the 45-degree line signify patients who lived exactly as long as predicted, points above the line signify patients who lived longer than predicted and points below the line signify patients who lived shorter than predicted.
Fig. 1To evaluate the accuracy of physicians' CPS and the relationship between CPS and AS, we plotted CPS and AS for each setting. Points on the 45-degree line signify patients who lived exactly as long as predicted, points above the line signify patients who lived longer than predicted, and points below the line signify patients who lived shorter than predicted. CPS = clinical predictions of survival; AS = actual survival.
The proportions of CPS meeting the criterion of accuracy (i.e., 0.67–1.33 times AS) were 36% (95% CI 32–40%), 32% (95% CI 29–35%), 34% (95% CI 29–39%), and 39% (95% CI 34–44%) in hospital palliative care teams, palliative care units, home palliative care services, and chemotherapy, respectively. In the total sample, this figure was 35% (95% CI 33–37%) (Table 2). There were no significant differences in accuracy between location of primary cancer or metastases (data not shown).
Table 2The Accuracy of CPS and the Relationship Between CPS and AS in Three Palliative Settings
Variables
Hospital Palliative Care Teams (n = 470)
Palliative Care Units (n = 764)
Home Palliative Care Services (n = 404)
Chemotherapy in All Settings (n = 398)
Total (n = 2036)
n
% (95% CI)
n
% (95% CI)
n
% (95% CI)
n
% (95% CI)
n
% (95% CI)
Accuracy of CPS
169
36 (32–40)
245
32 (29–35)
137
34 (29–39)
156
39 (34–44)
707
35 (33–37)
Proportion of pessimistic CPS
70
15 (12–18)
172
23 (20–25)
92
23 (19–27)
72
18 (14–22)
406
20 (18–22)
Proportion of optimistic CPS
231
49 (45–54)
347
45 (42–49)
175
43 (38–48)
170
43 (38–48)
923
45 (43–47)
P-value
P-value
P-value
P-value
Spearman's correlation coefficient between CPS and AS
0.73
<0.001
0.64
<0.001
0.67
<0.001
0.70
<0.001
0.69
<0.001
CPS = physicians' clinical predictions of survival; AS = actual survival.
We defined patients living longer than CPS as “pessimistic CPS” and patients living shorter than CPS as “optimistic CPS.”
The rate of “optimistic CPS” was more than twice that of “pessimistic CPS” (Table 2). The proportions of patients living longer than CPS (pessimistic CPS) were 15% (95% CI 12–18%), 23% (95% CI 20–25%), 23% (95% CI 19–27%), and 18% (95% CI 14–22%) in hospital palliative care teams, palliative care units, home palliative care services, and chemotherapy, respectively. In the total sample, this figure was 20% (95% CI 18–22%). The proportions of patients living less than the CPS (optimistic CPS) were 49% (95% CI 45–54%), 45% (95% CI 42–49%), 43% (95% CI 38–48%), and 43% (95% CI 38–48%), respectively. In the total sample, this figure was 45% (95% CI 43–47%).
Spearman's correlation coefficients between CPS and AS were 0.73 (P < 0.001), 0.64 (P < 0.001), 0.67 (P < 0.001), and 0.70 (P < 0.001) in hospital palliative care teams, palliative care units, home palliative care services, and chemotherapy, respectively. In the total sample, this figure was 0.69 (P < 0.001) (Table 2).
In comparison with the analysis including patients still alive at the end of the study period, “optimistic CPS” and “pessimistic CPS” were slightly enhanced (within ±5%), whereas the accuracy of CPS in all settings was almost identical (data not shown).
The number and proportion of patients whose survival time was accurately assessed by the physician in each of the seven AS categories are shown in Table 3. The most accurate categories of each group were 84 days+ (54%), 28–41 days (52%), 56–83 days (58%), and 84 days+ (53%) in hospital palliative care teams, palliative care units, home palliative care services, and chemotherapy, respectively. The least accurate categories were <7 days (17%), 7–13 days (15%), <7 days (5.3%), and <7 days (19%), respectively.
Table 3Proportion and Number of Patients Whose Survival Time Was Accurately Assessed by the Physician
Actual Survival
Days
<7
7–13
14–27
28–41
42–55
56–83
>=84
Total
Hospital palliative care teams % (n/N)
17 (12/70)
25 (17/68)
33 (29/88)
45 (30/67)
31 (14/45)
45 (24/53)
54 (43/79)
36 (169/470)
Palliative care units % (n/N)
19 (24/126)
15 (22/145)
38 (62/164)
52 (58/112)
44 (32/73)
49 (35/72)
17 (12/72)
32 (245/764)
Home palliative care services % (n/N)
5.3 (2/38)
28 (16/58)
29 (24/84)
40 (27/68)
45 (19/42)
58 (25/43)
34 (24/71)
34 (137/404)
Chemotherapy in all settings % (n/N)
19 (6/31)
27 (15/55)
43 (38/89)
34 (16/47)
31 (16/51)
50 (23/46)
53 (42/79)
39 (156/398)
Total % (n/N)
17 (44/265)
22 (70/326)
36 (153/425)
45 (131/294)
38 (81/211)
50 (107/214)
40 (121/301)
35 (707/2036)
The number represents the proportion of patients with accurate prognosis estimation (the number of patients with accurate prognosis estimation in the category/the number of patients who died in the category).
This study is one of the largest multicenter prospective cohort studies to evaluate the accuracy of CPS and to assess the relationship between CPS and AS in patients with advanced cancer. This study is unique in that we included patients receiving anticancer treatment, three palliative settings (palliative care units, hospital palliative care teams, and home palliative care services), and palliative care physicians' predictions of survival.
One of the most important findings is that CPS was often inaccurate and “optimistic CPS” was more than twice as common as “pessimistic CPS.” Approximately half of all estimates were optimistic and one-fifth were pessimistic. A systematic review
found that, in all but one of the studies identified, the CPS of physicians referring patients to a hospice care program and oncologists was overestimated. The results of our study confirm that CPS tends to be overestimated. This is true in prediction by palliative care physicians and in all settings from palliative care units to patients receiving chemotherapy. In this study, the accuracy of CPS ranged from 32% to 39% and Spearman's correlation coefficients between CPS and AS ranged from 0.64 to 0.73; these figures were higher than those in previous studies on physicians from other specialties, 4%–29% and 0.18–0.63, respectively.
The potential interpretation is that palliative care physicians actually provide more accurate prediction than referring physicians and oncologists but still have a clear tendency to overestimate. There have been, to the best of our knowledge, no studies claiming that palliative care physicians provide more accurate predictions than others. This may be the first study to suggest the superiority of palliative care physicians' predictions. The Palliative Prognostic Index
is a commonly used measure for survival prediction among palliative care physicians in Japan and routine use of it in daily activity might make their prognoses more accurate.
The second important finding was that overestimation trends of palliative care physicians' predictions were exhibited for patients with advanced cancer. Many global studies have indicated that end-of-life care has become increasingly aggressive over the past decade.
The tendency for “optimistic CPS” observed in patients receiving chemotherapy, therefore, is a focus of attention. It is vital for all palliative care physicians and oncologists to be aware that they systematically overestimate advanced cancer patients' prognoses so they are able to determine appropriate end-of-life care.
The finding of a good correlation between CPS and AS identified in this study suggests that CPS is still an acceptable clinical predictor in general. Clinically, to estimate individual prognosis in patients with advanced cancer, physicians can start with a clinical estimate and modify it according to the specific context, including symptoms, laboratory data, and comorbidities.
Prognostic factors in advanced cancer patients: evidence-based clinical recommendations—a study by the steering committee of the European Association for Palliative Care.
and the combination of CPS with these specific factors would be a way to achieve better prediction of survival.
The third important finding is that the accuracy rates of the seven AS categories tended to be low in the shorter AS categories compared with those in the longer ones. Some studies have reported that CPS is more accurate closer to death,
In the present study, physicians seemed to be more accurate at estimating a long prognosis than they were at estimating a short one. This phenomenon was observed in all palliative settings and in patients receiving chemotherapy. Our data suggest that physicians might have some difficulty in identifying patients with shorter AS.
Limitations
This study has several limitations. First, 352 (15%) patients still alive at the end of the study period were excluded. This is acceptable because dates of death are necessary to calculate accuracy of CPS; previous studies analyzed only patients whose date of death was available.
Explanatory analyses achieved the same results, and the analysis including them does not change the conclusions. Second, 38 (1.6%) patients were excluded because of missing data. But this proportion was sufficiently small so that the conclusions would not be altered. Third, we defined a CPS as accurate if it was within 0.67–1.33 times the AS, in accordance with other studies;
We believe that our arbitrary definition is reasonable because the median survival of all subjects was about one month. However, the extent of prognostic error varied depending on both CPS and AS. With our definition, the shorter AS induced higher error. This might produce lower accuracy of CPS in the shorter AS categories. Fourth, the superiority of palliative care physicians' predictions requires further study of a direct comparison with identical patients between palliative care physicians and others. Fifth, patients receiving chemotherapy in this study were recruited from palliative care services. This population seems to be patients with a later stage of cancer.
Conclusion
All physicians caring for patients with advanced cancer need to be aware of their tendency to overestimate survival, as it may affect end-of-life care. Further research is needed to investigate the reasons why CPS tends to be overestimated, the factors making CPS more accurate, how to incorporate these into existing prognostic scoring systems, whether the accuracy of CPS is beneficial to patients with advanced cancer, and ultimately how to improve their quality of life.
Disclosures and Acknowledgments
This work was supported in part by The National Cancer Center Research and Development Fund (25-A-22). The authors declare no potential conflicts of interest with respect to the research, authorship, and publication of this article.
This study was performed with the Japan ProVal Study Group (Prognostic Scores Validation Study Group). The participating study sites and site investigators may be found in the Appendix, available at jpsmjournal.com.
Appendix.
Japan ProVal Study Group (Prognostic scores Validation Study Group)
The participating study sites and site investigators were Satoshi Inoue, MD, Seirei Hospice, Seirei Mikatahara General Hospital; Masayuki Ikenaga, MD, Hospice Children's Hospice Hospital, Yodogawa Christian Hospital; Yoshihisa Matsumoto, MD, PhD, Department of Palliative Medicine, National Cancer Center Hospital East; Mika Baba, MD, Department of Palliative Care, Saito Yukoukai Hospital; Ryuichi Sekine, MD, Department of Pain and Palliative Care, Kameda Medical Center; Takashi Yamaguchi, MD, PhD, Department of Palliative Medicine, Kobe University Graduate School of Medicine; Takeshi Hirohashi, MD, Department of Palliative Care, Mitui Memorial Hospital; Tsukasa Tajima, MD, Department of Palliative Medicine, Tohoku University Hospital; Ryohei Tatara, MD, Department of Palliative Medicine, Osaka City General Hospital; Hiroaki Watanabe, MD, Komaki City Hospital; Hiroyuki Otani, MD, Department of Palliative Care Team, and Palliative and Supportive Care, National Kyushu Cancer Center; Chizuko Takigawa, MD, Department of Palliative Care, KKR Sapporo Medical Center; Yoshinobu Matsuda, MD, Department of Psychosomatic Medicine, National Hospital; Hiroka Nagaoka, MD, Center for Palliative and Supportive Care, University of Tsukuba Hospital; Masanori Mori, MD, Seirei Hamamatsu General Hospital; Yo Tei, MD, Seirei Hospice, Seirei Mikatahara General Hospital; Shuji Hiramoto, MD, Department of Oncology, Mitsubishi Kyoto Hospital; Akihiko Suga, MD, Department of Palliative Medicine, Shizuoka Saiseikai General Hospital; Takayuki Hisanaga, MD, Tsukuba Medical Center Foundation; Tatsuhiko Ishihara, MD, Palliative Care Department, Okayama Saiseikai General Hospital; Tomoyuki Iwashita, MD, Matsue City Hospital; Keisuke Kaneishi, MD, PhD, Department of Palliative Care Unit, JCHO Tokyo Shinjuku Medical Center; Shohei Kawagoe, MD, Aozora Clinic; Toshiyuki Kuriyama, MD, PhD, Department of Palliative Medicine, Wakayama Medical University Hospital Oncology Center; Takashi Maeda, MD, (Department of Palliative Care, Tokyo Metropolitan Cancer and Infectious Disease Center Komagome Hospital; Ichiro Mori, MD, Gratia Hospital Hospice; Nobuhisa Nakajima, MD, PhD, Department of Palliative Medicine, Graduate School of Medicine, Tohoku University; Tomohiro Nishi, MD, Kawasaki Comprehensive Care Center, Kawasaki Municipal Ida Hospital; Hiroki Sakurai, MD, Department of Palliative Care, St. Luke's International Hospital, Tokyo; Satofumi Shimoyama, MD, PhD, Department of Palliative Care, Aichi Cancer Center Hospital; Takuya Shinjo, MD, Shinjo Clinic; Hiroto Shirayama, MD, Iryouhoujinn Takumikai Osaka Kita Homecare Clinic); Takeshi Yamada, MD, PhD, Department of Gastrointestinal and Hepatobiliary-Pancreatic Surgery, Nippon Medical School; Shigeki Ono, MD, Division of Palliative Medicine, Shizuoka Cancer Center Hospital; Taketoshi Ozawa, MD, PhD, Megumi Zaitaku Clinic; Ryo Yamamoto, MD, Department of Palliative Medicine, Saku Central Hospital Advanced Care Center; Naoki Yamamoto, MD, PhD, Department of Primary Care Service, Shinsei Hospital; Hideki Shishido, MD, Shishido Internal Medicine Clinic; Mie Shimizu, MD, Saiseikai Matsusaka General Hospital; Masanori Kawahara, MD, PhD, Soshukai Okabe Clinic; Shigeru Aoki, MD, Sakanoue Family Clinic; Akira Demizu, MD, Demizu Clinic; Masahiro Goshima, MD, PhD, Homecare-Clinic Kobe; Keiji Goto, MD, Himawari Zaitaku Clinic; Yasuaki Gyoda, MD, PhD, Kanamecyo Home Care Clinic; Jun Hamano, MD, Division of Clinical Medicine, Faculty of Medicine, University of Tsukuba; Kotaro Hashimoto, MD, Fukushima Home Palliative Care Clinic; Sen Otomo, MD, Shonan International Village Clinic; Masako Sekimoto, MD, Sekimoto Clinic; Takemi Shibata, MD, Kanwakeakurinikku Eniwa; Yuka Sugimoto, MD, Sugimoto Homecare Clinic; Mikako Matsunaga, MD, Senri Pain Clinic; Yukihiko Takeda, MD, Hidamari Clinic; Takeshi Sasara, MD, Yuuaikai Nanbu Hospital; Jun Nagayama, MD, Peace Clinic Nakai.
References
Glare P.
Sinclair C.
Downing M.
Stone P.
Predicting survival in patients with advanced disease. Oxford textbook of palliative medicine.
4th ed. Oxford University Press,
Oxford, UK2010: 81-110
Prognostic factors in advanced cancer patients: evidence-based clinical recommendations—a study by the steering committee of the European Association for Palliative Care.
00161-X&id=gr1.jpg){kind=link}