Abstract
Background/Aim: Many patients with bone metastases receive palliative radiotherapy. However, treatment personalization tools are needed, due to heterogeneous survival. The aim of this study was to analyze the validity of the prognostic survival model, originally developed by Rades et al., because international variations in clinical practice and survival outcomes may impact on the performance of predictive tools. Patients and Methods: Data from a single institution were retrospectively analyzed. The study included 305 patients managed with palliative radiotherapy for bone metastases. The Rades et al. score was assigned and the resulting 3 prognostic strata were compared. Results: The median overall survival for the 3 strata was 48, 248, and 1065 days, respectively (p<0.001). However, the original break-down (17 points versus 18-25 points versus >25 points) was not in accordance with the overlapping survival curves in some of the subgroups, leading us to propose slight adjustments. The modified model also performed satisfactorily in older patients (age ≥80 years; median survival 26, 192 and 489 days, respectively, p<0.001). Conclusion: The original Rades et al. survival score was a valid prognostic model in our Norwegian validation database. However, inclusion of patients with 18 points into the poor prognosis group is suggested as a modification to enhance the score’s performance.
Population ageing has contributed to increasing numbers of cancer patients and consequently also rising demand for radiotherapy (1, 2). Depending on tumor stage, curative or palliative radiotherapy may be indicated in different settings and phases of the continuum of care (3, 4). Among palliative treatment indications, irradiation for painful uncomplicated bone metastases or complicated bone metastases (sometimes in the post-operative setting) represents a common scenario (5). Given that very convenient and well-tolerable regimens, such as single-fraction radiotherapy (8 Gy total dose for painful uncomplicated metastases), have been established, even frail or geriatric patients may be offered treatment (6-9). Nevertheless, selection of the appropriate fractionation regimen is not always trivial, especially in large or complicated bone metastases. When trying to avoid a mismatch between intense, locally highly effective but more time- and resource-consuming radiotherapy and remaining life span, many institutions have started utilizing prognostic models (10-12). Survival predictions obtained by such models may assist providers who are trying to avoid futile treatment in the final phase of cancer progression.
In order to support decision-making for elderly patients with bone metastases managed with palliative radiotherapy, Rades et al. have recently developed and validated a dedicated survival prediction model (13). They excluded the special setting of metastatic spinal cord compression resulting from bone metastases, because previous research already has resulted in diagnosis-specific models (14, 15). Their study included 348 patients who were ≥65 years of age and had received palliative radiotherapy in the time period 2009-2021, often 10 fractions of 3 Gy (47%). The cohort was divided into equally sized test and validation groups (174 patients each). Based on 4 parameters (sex, cancer type, Eastern Cooperative Oncology Group performance status (ECOG PS) and presence of visceral metastases), 3 prognostic strata were derived (calibrated for 6-months survival rate). Their median survival was 1.5, 7, and 39 months in the test cohort, respectively. The validation confirmed the excellent performance of the model.
To support implementation in other institutions, additional external validation of the model should be performed. Open questions regarding its role relate to the subgroup of very elderly patients and the fact that the 3 strata had very different group sizes [n=10 (6%), 141 (81%), 23 (13%)], resulting in a large group with intermediate prognosis. The present study was performed to answer these questions and validate the Rades et al. score.
Patients and Methods
Our group has previously validated prognostic models developed by other researchers and applied an identical approach in the present study (16, 17). A continuously maintained and updated database was employed, which included data collected from unselected Norwegian patients with bone metastases irradiated in routine clinical practice (Nordland Hospital Trust Bodø, Norway) since 2007. The database that was created for the purpose of regional quality-of-care analyses, has already been utilized and does not require additional approval by the local Ethics Committee (REK Nord). Fractionation regimens are selected by the clinical oncologist in charge at the time of first consultation and treatment planning. Stereotactic ablative radiotherapy was not utilized. To facilitate comparison with the Rades et al. study (13), only patients treated from 2009 onwards were included. Systemic therapy was tailored to disease burden and biology, organ function and patient preferences, and followed the National guidelines. Staging of extra-osseous metastases consisted of computed tomography (CT). If clinically relevant, other modalities, such as ultrasound and magnetic resonance imaging, were added to clarify the overall distribution of metastases. The minimum follow-up was 12 months (median 36 months in patients alive in October 2022 when analyzing the database, n=30). A point sum was calculated as previously described by Rades et al. who had assigned 7 points for ECOG PS 0-1, 4 points for ECOG PS ≥2, 7 points for female sex, 5 points for male sex, 6 points for bone-only metastases, 5 points for visceral metastases, and 8 points for breast cancer (prostate: 7, colorectal/kidney: 5, lung/unknown primary: 3, other primary: 4). Patients with highest point sum, i.e. 26-28, were allocated to the best prognostic group (sum 18-25: intermediate, sum 17: short survival).
Statistical analysis. Overall survival, defined as time from the start of radiotherapy to death, was calculated employing the Kaplan–Meier method. Log-rank tests were employed to compare actuarial survival curves. Cox regression was employed to assess the correlation between survival and the point sum calculated by administering the Rades et al. scoring system (continuous variable). All statistical analyses were performed using IBM SPSS statistical software version 28.0 (IBM Corp., Armonk, NY, USA).
Results
The study included 305 patients whose baseline characteristics are shown in Table I. The largest subgroup consisted of patients with prostate cancer (41%). In 36 patients (12%) radiotherapy was administered in the last month of life. Median overall survival was 8.2 months. The median point sum was 22, range=17-28. Point sum was significantly associated with overall survival in univariate Cox regression analysis, p<0.001. Survival outcomes stratified by point sum are presented in Table II. Based on the Kaplan-Meier survival curves (Figure 1), modification of the final 3-tiered score developed by Rades et al. appears warranted, because the original break-down (17 points versus 18-25 points versus >25 points) is not in accordance with the overlapping survival curves in patients with 17 and 18 points (median survival 48 and 47 days, 6-month rate 7% and 8%, respectively). In addition, survival of patients with 25 points was very close to that of patients with >25 points. The proposed score modification (17-18 points versus 19-24 points versus >24 points, as shown in Figure 2, p<0.001) would result in slightly more balanced group size (unfavorable: 9%, intermediate: 66%, favorable: 25%) and higher chi-square statistics (120 versus 47).
Baseline characteristics of 305 patients.
Survival outcomes of the study population (n=305).
Actuarial overall survival for all different point sums according to the Rades et al. score, p<0.001 (log-rank test pooled over all strata).
Actuarial overall survival for three different strata, according to the modified score (p<0.001, log-rank test pooled over all strata, chi-square 120). 1: Poor prognosis group (n=27) with median survival of 48 days. 2: Intermediate prognosis group (n=201) with median survival of 194 days. 3: Good prognosis group (n=77) with median survival of 822 days. For comparison, the original grouping resulted in median 48, 248 and 1065 days, respectively (p<0.001, chi-square 47).
A subgroup analysis was performed, which included all 74 patients who were at least 80 years old (median=83). Stratification according to the original Rades et al. score was not useful, because 72 patients (97%) were assigned to the intermediate group. The modified score performed satisfactorily (median survival 26, 192, and 489 days, for the groups with 17-18, 19-24, and >24 points, respectively, p<0.001). The Kaplan-Meier curves are displayed in Figure 3.
Actuarial overall survival (age ≥80 years) for three different strata, p<0.001 (log-rank test pooled over all strata). 1: Poor prognosis group (n=5) with median survival of 26 days. 2: Intermediate prognosis group (n=51) with median survival of 192 days. 3: Good prognosis group (n=18) with median survival of 489 days.
Discussion
This study was performed primarily to provide additional validation of the recent prognostic model developed by Rades et al., which predicts survival of patients with bone metastases who receive palliative radiotherapy (13). On one hand, palliative radiotherapy is a well-established and highly efficacious treatment for painful bone metastases (6). On the other hand, not all patients were shown to benefit and several studies have also suggested that measures reducing utilization of futile treatment close to the end-of-life (final month) are needed (18-20). Even if the perfect model predicting short survival has yet to be developed, existing models may be implemented to support decision making. The Rades et al. score focused on a particularly important subgroup of patients, namely elderly patients irradiated for bone metastases (age ≥65 years). Oncology care in the elderly or geriatric population is challenging because of higher risk for toxicity (frailty, reduced organ function, lower PS) and shorter remaining life-span, also due to comorbidity (21, 22). Nevertheless, good symptom palliation after radiotherapy has been reported also in the oldest old patients (8). It is, therefore, tempting to offer these patients well-tolerable, convenient fractionation regimens, such as a single fraction of 8 Gy for uncomplicated painful bone metastases.
Rades et al. successfully validated their score (13), however both test and validation groups included only 174 patients each. The present external validation study was based on 305 patients. The main differences between the two studies relate to primary tumor site (Rades et al.: lung 30%, breast 26%, prostate 20%; present: lung 20%, breast 12%, prostate 41%) and preferred fractionation regimen (Rades et al.: 10 fractions of 3 Gy in 47%; present: 32%). Despite these differences, similar 6-months survival rates were obtained for the unfavorable groups (Rades et al.: test group 0%, validation group 9%; present: 7%) as well as for the favorable groups (100%, 86% and 92%, respectively). Thus, external validation was successful. However, a closer look at the present survival curves (Figure 1), revealed that unfavorable patients defined by a point sum of 17 had survival undistinguishable from those with a point sum of 18. Also, those with a point sum of 25 had relatively similar survival to their counterparts with higher point sum. If one modifies the break-down to adjust for these findings, as reflected in the survival curves shown in Figure 2, excellent discrimination can be maintained. The main advantage of this modification lies in the increased group sizes for both unfavorable and favorable patients, while the intermediate group can be reduced to 66% of patients (still representing a large proportion).
A secondary purpose of our study was to analyze the subgroup of the oldest old patients, defined as age ≥80 years, a particularly vulnerable and potentially frail population. We found that the original Rades et al. model assigned almost all patients to the intermediate group, while the modified model sorted out patients with favorable and unfavorable prognosis. Survival of the latter group (17-18 points) was very short (median 26 days, maximum 47), which is shorter than that of all patients with 17-18 points in the study (median 48 days), and raises concern about the appropriateness of radiotherapy in this age group when adverse prognostic features are present.
Alternative prognostic models have already been published. The most complex one is the so-called Bone Metastases Ensemble Trees for Survival (BMETS) with 27 co-variates (11, 16). Simpler models include Chow’s 3-item (non-breast primary cancer, metastases other than bone only, and Karnofsky PS ≤60) (12) and Westhoff’s 2-item tools (PS, primary tumor) (10). The Rades et al. model is only slightly more complex than the 3- or 2-item models. No head-to-head comparison in a sufficiently large database, ideally with >500 patients to ensure excellent statistical power, and stratified by age, is yet available. In principle, estimation of the remaining life span with a simple, validated model is better than no prognostic assessment at all.
The limitations of our study include its single-institution methodology and predominance of male sex/prostate cancer. Before wide-spread application of our modified score, additional validation in at least one large, external database is needed.
Conclusion
The original Rades et al. survival score was a valid prognostic model in our database of Norwegian patients irradiated for bone metastasis. However, inclusion of patients with 18 points into the poor prognosis group is suggested as a modification to improve performance of the score. In addition, in patients with age ≥80 years, the modified survival score was able to predict those with poorer survival; thus, contributing to the optimization of patient selection for radiotherapy in this age group.
Footnotes
Authors’ Contributions
CN participated in the design of the study and performed the statistical analysis. CN, LS, BM and ECH conceived the study and drafted the article. All Authors read and approved the final article.
Conflicts of Interest
The Authors declare that they have no conflicts of interest.
- Received November 4, 2022.
- Revision received December 9, 2022.
- Accepted December 13, 2022.
- Copyright © 2023 International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved.
