Abstract
Patients with advanced hematological malignancies are less likely to be referred to specialist palliative care services compared with patients having solid tumors. It has been reported that one of the most important reasons for the lack of referral is difficulties in the prognostication of terminally ill patients with hematologic malignancies. The study objective was to evaluate the predictive accuracy of the Palliative Prognostic Index (PPI) and the prognostic model developed by Kripp et al in hospitalized patients under the care of a hematologist. Using clinical charts, we retrospectively calculated the above scores. We reviewed the records of 114 patients admitted to the hematology ward. The inclusion criterion was patient with disease considered incurable using standard treatments. The prognostic models were assessed according to the original reports. Using PPI cutoff points of 2 and 4, we divided the patients into 3 groups of significantly different survival times (P < .01). Moreover, we confirmed the usefulness of predicting survival <3 and <6 weeks using PPI scores of 6 and 4 as cutoff points, respectively. When we classified patients according to the prognostic model of Kripp et al, the high-risk group survived significantly shorter times than the intermediate- and low-risk groups (P < .001). However, there was no significant difference in survival between the intermediate- and low-risk groups. Use of these models might enable physicians to provide more appropriate end-of-life care and to refer patients to palliative care earlier.
Keywords
Introduction
Major therapeutic advances have contributed to improved survival in patients with hematological malignancies over the past decades. However, a substantial proportion of patients eventually die from their disease. End-of-life care is extremely important for these patients. However, in reality, hematologists provide the bulk of palliation in acute care hospitals. Conventional research has indicated that there has been little integration of hematology with palliative care services, 1 -3 despite clinical guideline recommendations. 4 Patients with hematological malignancies less frequently access specialist palliative care services (SPCS). 5,6 Moreover, the time between referral to SPCS and death is shorter than in patients with solid tumors, 7 and such late referral limits the benefits of palliative care.
One of the most important reasons for the lack of referral to SPCS is likely the lack of good prognostic prediction models. 8 Patients and their families often underestimate the possibility of imminent death. 9 In addition, it is known that physicians’ estimates are often unreliable. 10 -12 Moreover, a lack of consensus among hematologists on when they should refer patients for SPCS has been reported. 8 To improve the integrated palliative care of hematological patients, prognostic assessment should be carried out by hematologists before the decision to refer is made. However, the absence of reliable prognostic tools for patients with advanced and incurable hematological malignancies who are hospitalized in hematology wards can make prognostication extremely difficult. 8,13
Several prognostic prediction tools for patients with advanced cancer have been proposed, for example, the Palliative Prognostic Index (PPI), 14 Palliative Performance Scale (PPS), 15 Palliative Prognostic Score, 16 and Prognosis in Palliative Care Study model. 17 These models have been reported mainly in patients with solid tumors, and hematologic malignancies have only been nominally included. Furthermore, these models are developed for use in assessing palliative care populations, and the appropriateness of their application to patients still under the care of hematologists is uncertain. Recently, Chou et al 18 reported that the PPI might be useful in patients with hematological malignancies receiving palliative care consultation services. Kripp et al 19 also developed prognostic models in a palliative care unit for patients with hematological malignancies. However, both of these have been verified in palliative care settings. Accurate survival prediction in hematological care settings may be useful for patients with advanced hematological malignancies to avoid unnecessary anticancer therapies and provide appropriate palliative care.
The purpose of this study was thus to clarify the prognostic accuracy of the PPI and the model by Kripp et al in hospitalized patients with advanced hematological malignancies who were being managed by hematologists but who had not yet been referred to SPCS.
Methods
We retrospectively reviewed the records of all consecutive patients admitted to the hematology ward of Almeida Memorial Hospital in Oita, Japan. The study was conducted from January 2012 to December 2014. The inclusion criteria were patients aged 18 years or older, hospitalized for hematological malignancies at an advanced stage, and with disease considered incurable using standard treatments. We determined the PPI and risk factors according to Kripp et al, based on information recorded in patient clinical charts at the time of admission. Survival was calculated as the number of weeks from the date of this assessment until the date of death.
Morita et al 14 developed the PPI in 1999 to predict the prognosis of patients with advanced cancers in palliative care units. The index was calculated by summing the scores for factors of the PPS, together with scores for oral intake, delirium, dyspnea at rest, and edema. The range of PPI scores is from 0 to 15: 0 indicates the best performance status and absence of symptoms, whereas 15 indicates the worst performance status and presence of all symptoms. Validated cutoff points have been established that categorize patients into 3 prognostic groups, based on the total PPI score: group A, comprising patients with a PPI score of ≤2, group B with PPI score of >2 and ≤4, and group C with a PPI score of >4. A PPI score of more than 6 predicts survival of less than 3 weeks, and a score of more than 4 predicts survival of less than 6 weeks.
Using 5 clinical and laboratory parameters, Kripp et al 19 recently developed a prognostic model to estimate the survival time in patients with advanced hematological malignancies treated in a palliative care unit. A poor prognosis was associated with 5 parameters: low performance status, low platelet count, opioid-based pain therapy, high lactate dehydrogenase levels, and low albumin levels. The sum score of all items, ranging from 0 to 5 points, falls into one of 3 “risk” categories: low-risk (0-1 points), intermediate-risk (2-3), and high-risk (4-5) groups.
Kaplan-Meier survival curves were constructed for each of these 3 groups, and survival analysis was performed by log-rank test, with a post hoc multiple comparison test using the Holm method. Applying PPI cutoff points of 4 and 6 divided the sample into 2 groups based on PPI score, and the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy with 95% confidence intervals (CIs) of predictions of survival of less than 3 and 6 weeks were calculated. All statistical analyses were conducted using EZR, a graphical user interface for R version 3.0.2 (The R Foundation for Statistical Computing, Vienna, Austria). The institutional review board of our hospital approved this study for ethical and scientific validity.
Results
Patient Characteristics
A total of 124 patients were included in the study. Ten were lost to follow-up as a result of being discharged. We therefore finally included the data of 114 patients. At the end of the follow-up period, all patients had died. The median survival time was 6.4 weeks. Patient mean age was 74.3 ± 10.6 years, and men accounted for 63.1% of the sample. Disease characteristics are detailed in Table 1. The most common diseases were malignant lymphoma in 42 (36.8%) patients, acute leukemia in 32 (28.1%), multiple myeloma in 21 (18.4%), and myelodysplastic syndromes in 19 (16.7%). Of the patients with myelodysplastic syndrome, 8 cases eventually progressed to acute leukemia. None of the patients had undergone allogeneic stem cell transplantation. Table 2 shows the reason for admission. The most frequent reasons for admission were infection (26.3%), palliative chemotherapy and/or radiation (25.4%), appetite loss (17.5%), general fatigue (14.9%), tumor fever (11.4%), and pain (10.5%). In many cases, more than one reason for admission was recorded.
Patient Characteristics.
Abbreviations: ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; ATL, adult T-cell leukemia/lymphoma; B-NHL, B-cell non-Hodgkin’s lymphoma; HL, Hodgkin’s lymphoma; MDS, myelodysplastic syndromes; MM, multiple myeloma; PTCL, peripheral T-cell lymphoma (ex ATL).
Reason for Admission.
Palliative Prognostic Index Analysis
Figure 1 shows the Kaplan-Meier survival curve for the 3 groups divided according to the sum of PPI scores: (1) patients with PPI sum score ≤2.0 (group A; n = 23), (2) patients with PPI sum score >2.0 and ≤4.0 (group B; n = 27), and (3) patients with PPI sum score >4.0 (group C; n = 64). Patients in group C had a median survival of 3.1 weeks (95% CI: 2.6-3.9), whereas patients in groups B and A had respective survival times of 6.4 weeks (95% CI: 4.7-8.4) and 11.1 weeks (95% CI: 6.6-13.0). Patients in group C had significantly shorter survival times than those in the other groups (P < .01), and patients in group B survived shorter times than those in group A (P < .01).
Kaplan-Meier survival curve for each Palliative Prognostic Index (PPI) group.
Palliative Prognostic Index cutoff points of 4 and 6 divided the sample into 2 groups. A total 33 of the 51 patients with PPI score >6 and 3 of the 63 patients with PPI score ≤6 survived for less than 3 weeks. A total 55 of the 64 patients with PPI score >4 and 17 of the 50 patients with PPI score ≤4 survived for less than 6 weeks. The sensitivity, specificity, PPV, NPV, and overall accuracy of PPI in predicting survival with the 2 cutoff values are shown in Table 3. Our data were similar to those of the original work. 14
Validation of Predictions Using PPI.
Abbreviation: PPI, Palliative Prognostic Index.
Prognostic Model by the Analysis of Kripp et al
We established 3 so-called risk groups according to the sum of the 5 risk factors of Kripp et al: (1) total number of risk factors is 0 or 1 (low-risk group, n = 15), (2) presence of 2 or 3 factors (intermediate-risk group, n = 70), and (3) presence of 4 or 5 factors (high-risk group, n = 29). Patients in the high-risk group had a median survival of 3.0 weeks (95% CI: 1.7-3.6), whereas patients in the intermediate- and low-risk groups survived 5.4 weeks (95% CI: 4.1-6.4) and 8.0 weeks (95% CI: 2.7-13.0), respectively. The difference in survival between the 3 groups was statistically significant (P < .001). Patients in the high-risk group survived significantly shorter times than those in the other groups (P < .001). However, there was no significant difference between the intermediate- and low-risk groups (P = .11). Kaplan-Meier survival curves of the 3 groups are shown in Figure 2.
Kaplan-Meier survival curve for the 3 risk groups according to the sum of risk factors by Kripp et al.
Discussion
To the best of our knowledge, this is the first study in which the PPI 14 and the prognostic model developed by Kripp et al, 19 originally developed for palliative care populations, have been validated in patients with advanced hematological malignancies who are still under the care of a hematologist. The PPI was first designed for patients in palliative care units 14 ; however, the usefulness of the PPI has been subsequently verified in different settings. 20 -23 However, patients with hematological malignancies accounted for only 1.0% to 17.6% of the study populations in these studies. For patients with early disease at diagnosis of hematological malignancies, prognostic prediction models such as the International Prognostic Index for malignant lymphoma, 24 European LeukemiaNet genetic risk classification for acute myeloid leukemia, 25 International Staging System for multiple myeloma, 26 and International Prognostic Scoring System for myelodysplastic syndromes 27 are available to provide prognostic information. However, no prognostic model existed for incurable patients with advanced hematologic malignancies. Recently, Chou et al 18 confirmed that PPI has significant predictive value for the life expectancy of terminally ill patients with hematologic malignancies in palliative care settings. Further, Kripp et al 19 developed a prognostic scoring system in 2014 to predict the prognosis of patients with hematologic malignancies in a palliative care unit. However, both of these have been validated in SPCS.
The present study revealed a close association between PPI and survival in patients with advanced incurable hematological malignancies receiving the care of a hematologist. The patients in the present study had not yet been referred to SPCS, and they had a median survival of 6.4 weeks. The patients in the study by Chou et al 18 were in palliative care settings and survived only 2.3 weeks. In this way, although the patients were clinically different compared with a palliative care population, the PPI could be used to categorize them into 3 different prognosis groups on the basis of survival probability.
Moreover, whether patients with hematological malignancies lived less than 3 or 6 weeks was capably predicted by the PPI. The sensitivity, specificity, PPV, NPV, and overall accuracy determined in the present study were consistent with prior findings. 20 -22 Unfortunately, physicians generally tend to provide an optimistic prognosis for patients with advanced cancer. 11 An accurate survival estimate is useful to determine when to stop anticancer therapies and to make decisions about end-of-life care. Our results indicate that the PPI may assist hematologists with selecting patients who will benefit from SPCS.
On the basis of the model developed by Kripp et al, we showed that the high-risk group demonstrated significant predictive value for the life expectancy of patients with advanced hematological malignancies under the care of a hematologist. However, when comparing the intermediate- and low-risk groups, there was no significant difference in survival between these 2 patient groups. This is likely because patients in our low-risk group had a median survival of 8.0 weeks, whereas those in the original study had a reported survival of 62.9 weeks. 19 Hematological malignancies are a heterogeneous group of diseases. There are differences in disease trajectory, with some malignancies being extremely aggressive, whereas others follow a more indolent course. Chronic lymphocytic leukemia (CLL) is a type of slow-growing hematological malignancies. When patients with multidrug refractory CLL had received ofatumumab, the median survival time was 9.8 to 10.2 months in nonresponders. 28 No patients in our study had CLL, whereas in the original study, 9.3% of the patients were diagnosed with CLL. Because CLL is quite rare in Japan, 29 this result may reflect recognized variances in the demographics of hematological malignancies between Japan and the Western world.
In conclusion, our findings revealed the existence of a close association between PPI and survival in patients with advanced hematologic malignancies who were under the care of a hematologist. Furthermore, we found that the prognostic model by Kripp et al was suited to predict which patients with advanced hematologic malignancies had the poorest survival. These tools are quick and easy to use and can be applied to patients with hematological malignancies who are being managed in a hematology ward. Accurate survival prediction is essential for decision making about cancer therapies and for end-of-life care planning. These prognostic models could be useful to help patients and hematologists determine when to refer patients to SPCS. This study was carried out retrospectively and included only participants from a single institution. Further studies including larger number of patients and an improved design may be needed to clarify the predictive value of these prognostic models in hematological care settings other than palliative care.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.