Skip to main content
Download PDF

Prognostic models for survival predictions in advanced cancer patients: a systematic review and meta-analysis

BMC Palliative Care volume 24, Article number: 54 (2025) Cite this article

Abstract

Background

Prognostication of survival among patients with advanced cancer is essential for palliative care (PC) planning. The implementation of a clinical point-of-care prognostic model may inform clinicians and facilitate decision-making. While early PC referral yields better clinical outcomes, actual referral time differs by clinical contexts and accessible. To summarize the various prognostic models that may cater to these needs, we conducted a systematic review and meta-analysis.

Methods

A systematic literature search was conducted in Ovid Medline, Embase, CINAHL Ultimate, and Scopus to identify eligible studies focusing on incurable solid tumors, validation of prognostic models, and measurement of predictive performances. Model characteristics and performances were summarized in tables. Prediction model study Risk Of Bias Assessment Tool (PROBAST) was adopted for risk of bias assessment. Meta-analysis of individual models, where appropriate, was performed by pooling C-index.

Results

35 studies covering 35 types of prognostic models were included. Palliative Prognostic Index (PPI), Palliative Prognostic Score (PaP), and Objective Prognostic Score (OPS) were most frequently identified models. The pooled C-statistic of PPI for 30-day survival prediction was 0.68 (95% CI: 0.62–0.73, n = 6). The pooled C-statistic of PaP for 30-day survival prediction was 0.76 (95% CI: 0.70–0.80, n = 11), while that for 21-day survival prediction was 0.80 (0.71–0.86, n = 4). The pooled C-statistic of OPS for 30-days survival prediction was 0.69 (95% CI: 0.65–0.72, n = 3). All included studies had high risk of bias.

Conclusion

PaP appears to perform better but further validation and implementation studies were needed for confirmation.

Key message

This article describes a systematic review and meta-analysis of prognostic models for patients with advanced cancer and their performance statistics. The results indicate that Palliative Prognostic Score (PaP) has superior predictive accuracy but further validation and implementation studies were needed to confirm its value of clinical utility.

Peer Review reports

Introduction

In 2017, the American Society of Clinical Oncology (ASCO) updated its clinical practice guideline, highlighting the evidence base for early integration of palliative care alongside oncologic care [1]. The clinical benefits of palliative care integration include better quality of life, reduced depression, reduced hospital readmissions, improved satisfaction with care, and potential increase in survival [2, 3]. Moreover, referrals to palliative care services at earlier disease stages would result in greater improvements and overall fewer medical costs [4,5,6,7,8].

There is no universal consensus over the appropriate time for referrals. Data from randomized controlled trials suggest a minimal of 6 months for clinical benefits of palliative care to emerge [4, 5, 9,10,11,12]. ASCO guidelines recommend referrals to be made within 8 weeks of advanced cancer diagnosis [13]. However, such an early referral time may be unrealistic for certain healthcare systems. Referrals are also dependent on patients’ preferences. With the advent of personalized medicine, next-generation sequencing and targeted treatment, the disease trajectory is highly dependent on disease primary site, genotyping and availability of advanced treatments. Identifying the best timing to refer patients to palliative care service, particularly that which is not too late to enable full effect of holistic intervention, and not too early to lack relevance to service user, is therefore extremely challenging.

To facilitate decision-making, a reliable prediction of survival is needed. Multiple studies have suggested the inaccuracy of clinician prediction of survival in advanced cancer population [14, 15]. Patients with wrongly estimated survival may have poorer quality of life and higher symptom burden [14, 16]. While the prognosis of patients with advanced cancers amenable to effective life-prolonging treatment can be estimated using survival data derived from clinical trial reports, they were not immediately generalizable to cancer with no suitable novel treatments [17].

To address this gap, prognostic models have been developed for this particular group of patients, based on their clinical statuses or biologic factors. Many of these models are externally validated, but their implementation into clinical practice could be hindered by two main problems. First, some models involve non-routine biomarkers and complex calculations, rendering them unfeasible for everyday practice [18]. Second, studies on prognostic models differ in terms of patient characteristics, clinical settings, and methodologies. It is not straightforward for clinicians to extrapolate the data on model performance to their own clinical contexts [18].

We aim to make explicit the applicability of prognostic models for advanced cancer patients through two objectives. The primary objective of this study is to identify validated prognostic models and assess their performance. The secondary objective is to explore whether differences in patient characteristics and clinical settings across studies are associated with model performance.

Methods

A systematic search was done on Medline, Embase, CINAHL Ultimate, and Scopus, with complementary reference mining from review articles. The search period was up to August 2022. Search terms were available in appendix I. Studies published in English full texts were eligible. Abstracts and conference articles were excluded. The study population should be targeted at patients with advanced cancer, defined as incurable, who were 18 or above. The study sample should consist of at least 2 cancer types, but not haematological malignancies. The prognostic models examined should include at least 2 factors and were validated internally or externally. The studies should report measures of model performance in terms of discrimination and/or calibration. Two reviewers independently screened out irrelevant articles based on title and abstract. Discrepancies were resolved by a third reviewer. Full texts were then retrieved for the remaining articles, for which detailed screening were done independently by the two reviewers. Disagreements were resolved by the same third reviewer.

Data extraction was done independently by two reviewers in accordance to CHARMS guidelines [19]. Risk of bias were assessed with PROBAST [20]. Where possible, missing data were obtained by contacting authors of original articles. Clinical heterogeneity was assessed in terms of clinical settings, patient characteristics, model types, and prediction timeframe.

Meta analysis was performed if adequate clinical homogeneity was established. Meta-analysis of the C-statistics with logit transformation was conducted using the packages METAFOR in R, to improve validity of its underlying assumptions. Restricted maximum likelihood (REML) estimation method was to calculate 95% confidence intervals for the average performance using the METAFOR package in R.

Test performance characteristics were summarized using a forest plot. Heterogeneity of prognostic model performance across studies was assessed by 95% prediction intervals (PI) and I2 statistic (I2). PI provides an estimated range within which the true effect size of a future study would be expected to fall 95% of the time. Wide PI suggest substantial heterogeneity in model performance. PI was calculated using the METAGEN package in R. I2 was estimated and that I2 > 50% was taken as signifying substantial heterogeneity.

Multilevel analysis and/or meta-regression were performed if more than 10 adequately homogeneous studies could be pooled.

The study was registered on PROSPERO with ID: CRD42023403263. The systematic search in database was carried out between 1st Aug 2022 to 31st Aug 2022. We included all studies publised on or before 31st Aug 2022.

Results

A total of 35 studies covering 35 types of prognostic models were included after the screening process detailed in Fig. 1. Characteristics of the included studies can be found in Table 1. 4 studies tested for survival in one week, 2 studies in one year, and the rest (N = 29) in between. 23 models utilize both clinical and objectively assessed parameters such as physical or laboratory measurements. 8 models adopted only clinical factors while 4 adopted only objective parameters.

Fig. 1
figure 1

Flow Diagram for Study Selection

Full size image
Table 1 Characteristics of Included Studies
Full size table

Out of the 8 clinical-only models, 15 prognostic factors were identified with performance status like ECOG, KPS, etc. being the most included factor (n = 8). Other commonly included factors include distant metastases (n = 4), edema (n = 4) and poor oral intake (n = 4).

Out of the 4 objective models, 9 prognostic factors were identified with hypoalbuminemia being the most included prognostic factor (n = 4). Other commonly included prognostic factors include heart rate (n = 3) and urea (n = 3).

Out of the 23 mixed models, 46 prognostic factors were identified with performance status being the most included prognostic factor (n = 19), followed by WBC count (n = 11), metastases (n = 10), dyspnea (n = 8) and poor oral intake (n = 7).

Table 2 lists the factors involved in each model and their nature. Most studies contain a mixture of clinical and biological factors.

Table 2 Factors of Prognostic Models
Full size table

Table 3 captures performance of models demonstrated in each study. C-indices presented in the included studies ranged from 0.61–0.86. Palliative Prognostic Index (PPI) and Palliative Prognostic Score (PaP) were the most extensively validated models, followed by Objective Prognostic Score (OPS) and Palliative Performance Score (PPS). Details of classification, discrimination, and calibration statistics can be found in Table 3 Characteristics of Included Studies

Table 3 Performance Statistics of Models
Full size table

Palliative Prognostic Index (PPI) was validated in 9 of the included studies. It consists of 5 clinical factors, namely palliative performance scale, oral intake, edema, dyspnea, and delirium. The pooled C-statistic for 30-day survival prediction was 0.68 (95% CI: 0.62–0.73, n = 6) as shown in Fig. 2. The 95% Prediction Interval (PI) was [0.51–0.81]. The I2 statistic was 93.9% (95% CI: 89.4%—96.5%), indicating significant heterogeneity. While PPI was typically compiled at initial assessment in palliative care service, Kao et al. investigated the prognostic value of combining both initial and change in PPI score. The C-statistic for predicting 30-day survival was shown to be significantly higher with the combined initial PPI and ∆score (C-statistic, 0.71; 95% confidence interval (CI), 0.694–0.731) than with the initial PPI (C-statistic, 0.63; 95% CI, 0.61–0.65), week 1 PPI (C-statistic, 0.67; 95% CI, 0.652–0.690), or ∆score (C-statistic, 0.64; 95% CI, 0.62–0.66) alone. [31].

Fig. 2
figure 2

C-statistics of Palliative Prognostic Index (PPI) in 30-days Survival Prediction

Full size image

Palliative Prognostic Score (PaP) was validated in 10 of the included studies. It consists of 4 clinical factors, including dyspnea, anorexia, KPS, and clinical prediction of survival, as well as 2 laboratory factors, including white cell count and lymphocyte percentage. As shown in Figs. 3, 4, the pooled C-statistic for 30-day survival prediction was 0.76 (95% CI: 0.70–0.80, n = 11), while that for 21-day survival prediction was 0.80 (0.71–0.86, n = 4). The 95% PI for 30-day and 21-day survival predictions were [0.54 – 0.90] and [0.57 – 0.92] respectively. The I2 statistics were 95.9% (95% CI: 94.2%—97.1%) and 64.3% (95% CI: 0 – 87.9%) for 30-day and 21-day survival predictions, respectively. Two additional studies assessed the incorporation of delirium to the PaP model (D-PaP). Scarpi et al. presented a marginally higher K statistics for 30-day survival with D-PaP (0.860, 95% CI: 0.817– 0.880) than PaP (0.853, 95% CI: 0.823–0.877) for the PaP score [25]. Maltoni et al. found that D-PaP had a C-statistic of 0.73 (95% CI: 0.71–0.74) compared to PaP which had a C-statistic of 0.72 (95% CI: 0.70–0.73) [28]. We emphasize that due to a relatively small number of studies included in the meta-analysis, the prediction intervals and I2 value presented should not be used to draw strong conclusions about heterogeneity or the range of true effects.

Fig. 3
figure 3

C-statistics of Palliative Prognostic Score (PaP) in 30-days Survival Prediction

Full size image
Fig. 4
figure 4

C-statistics of Palliative Prognostic Score (PaP) in 3-weeks Survival Prediction

Full size image

As shown in Fig. 5, the funnel plot of C-statistics for PaP showed mild asymmetry, with a slight clustering of studies towards higher C-statistics, potentially suggesting a mild publication bias. No extreme outliers were observed.

Fig. 5
figure 5

Funnel Plot: C-statistics of PaP in predicting 30-days survival

Full size image

Objective Prognostic Score (OPS) was validated in 5 of the included studies. It consists of 3 clinical factors, namely Eastern Cooperative Oncology Group (ECOG) Performance Status, dyspnea at rest, and oral intake, along with 3 laboratory factors, namely white cell count, serum bilirubin, and serum creatinine. The C-statistic of OPS for 30-days survival prediction ranged from 0.68 (95% CI: 0.58 – 0.77) [41] to 0.71 (95% CI: 0.64 – 0.77) [50].

As shown in Fig. 6, the pooled C-statistic for 30-day survival prediction was 0.69 (95% CI: 0.65–0.72, n = 3). The 95% PI for 30-day survival predictions was [0.58 – 0.78]. The I2 statistics was 92.1% (95% CI: 87.9%—98.4%) for 30-day survival prediction.

Fig. 6
figure 6

C-statistics of Objective Prognostic Score (OPS) in 30-days Survival Prediction

Full size image

Palliative Performance Score (PPS) was validated in 3 of the included studies. It comprises 5 clinical factors, including ambulation, activity level, self-care, intake, and level of consciousness. Meta-analysis on C-statistics of PPS was not performed due to insufficient number of studies with adequate clinical homogeneity.

Multilevel meta-analysis including 18 datasets from 14 unique studies was conducted to compare the performance of PaP and PPI in predicting 30-days survival, accounting for within-study dependence. The difference in the C-statistic between PaP and PPI was statistically significant (p < 0.0001). The pooled C-statistic of PaP was 0.76 (95% CI: 0.70 – 0.80). In contrast, the C-statistic of PPI was estimated to be 0.0485 (95% CI: 0.0388 – 0.0583) lower than that of PaP. The variance component analysis indicated minimal variability between datasets (σ2 = 0.0046). The nested Study/Model effect showed negligible variance, suggesting that the relative performance of PaP and PPI was consistent across studies. However, these findings should be interpreted with caution due to the relatively small number of studies included.

Risk-of-bias of the studies is summarized in Fig. 7 and Fig. 8. All studies (n = 35) carried high risk of bias due to issues in analysis or its reporting. Specifically, all studies were rated as high risk in the 'Analysis' category due to inadequate reporting of calibration and mishandling of missing data. In many cases, missing data handling was either not mentioned or addressed through complete case analysis, instead of employing multiple imputation or other gold standard approaches. Eleven studies (31.4%) were considered high risk of bias in the 'Participants' domain as they used retrospective data sources rather than prospective ones. Almost all studies (n = 34) carried low risk of bias in the 'Predictors' and 'Outcome' domains, suggesting well-defined measurement of predictors and outcomes. 33 out of the 35 included studies were considered highly applicable for our study question.

Fig. 7
figure 7

Risk of bias (ROB): Traffic Light Report

Full size image
Fig. 8
figure 8

Risk of Bias Summary for Included Studies

Full size image

Discussion

Prior to the completion of this study, three reviews on prognostic models for patients with advanced cancer have been identified. Simmons et al., did not perform meta-analysis to compare the performances of models and the potential underlying factors [18]. Pobar et al. aimed specifically at identifying models suitable for radiation therapy planning, so only two specific timeframes (4-week and 3-month survival) were included for evaluation. [56] Owusuaa et al. included studies that involve stage 1 cancers and that study subjects were not strictly for palliative intent [57]. Our review updated the identification of prognostic models for patients with advanced cancer and presented meta-analysis to add to the literature body. Meta-analyses of specific prognostic models described in this review had been published elsewhere but whose eligibility criteria differed significantly [57]. examined PPS including non-cancer patients [58]. examined PaP involving haematological and non-terminal malignancies [59].

We have systematically searched for validated prognostic models for survival prediction among patients with advanced cancer. The identified models typically combine both clinical and objective biologic factors to estimate survival probability. Whether models solely comprised of biologic factors perform better or worse cannot be ascertained due to limited number of studies available for meta-analysis. Nevertheless, Objective Prognostic Index for Advanced Cancer (OPI-AC), an example of such models, demonstrated possible superiority of objective parameters (C-index > 0.8 for 30-day, 56-day, and 90-day survival) [60]. This represents an important research gap that may determine future directions for prognostic model development.

Our findings were largely in line with the ESMO Clinical Practice Guideline where prognostic models are endorsed for the clinical prediction of survival ranging weeks to months [60]. Further to their recommendations, we added that shorter- (days) and longer-term (months to years) survival predictions have been tested but data were relatively scarce to support clinical incorporation. The superiority of PaP over PPI in terms of discrimination, as reflected in our multi-level regression, also resemble previous cohort studies quoted in the guideline [61, 62]. While the C-index is a widely used measure of predictive accuracy, it represents only one aspect of model performance. The clinical significance of the observed difference in C-statistics between PaP and PPI remains unclear and warrants further investigation.

It is important to note that selecting a suitable model for clinical use or further validation should be guided by considerations beyond performance metrics alone. The clinical settings, prediction timeframe, and patient characteristics underlying the study sample may deviate significantly from the population of interest [63]. PaP incorporates objective laboratory factors, whereas PPI relies solely on clinical parameters. This distinction has practical implications, as PPI might be more convenient and accessible in resource-limited settings where laboratory facilities are scarce [64]. Moreover, the invasiveness of tests, the expertise required from personnel, and the complexity of assessments involved would be potential determinants of what models to be chosen in a certain clinical setting. Even if model performances do not show superiority over clinician predictions, the reproducibility and objectivity of prognostic tools may aid communication and education for less experienced staff as well as patients and their carers [65].

Hence, the choice of prognostic approach in practice may depend on a balance of factors including predictive accuracy, resource availability, ease of use, and the specific clinical context. While this study prioritized models based on their performance, a comprehensive approach considering both statistical performance and practical implementation is necessary for optimal clinical application.

Furthermore, utilizing more accurate prognostic models would theoretically facilitate end-of-life communication with patients and caregivers. However, no studies have yet been conducted to compare the impacts of different prognostication methods (clinician prediction, prognostic models, prognostic factors) in clinical care. As mentioned in the ESMO practice guideline, RCTs on the feasibility and clinical utility of various prognostication methods are warranted [66].

Several methodological limitations across included studies were identified. Cut-offs for assigning patients into prognostic groups varied between studies of the same model. For example, while [26, 66] and [25, 28] categorized those with PPI > = 6 as likely to survive less than 3 weeks, [26, 32] adopted PPI > 5 as the benchmark. Similarly, [28] treated PaP > 9 as unlikely to survive beyond 3 weeks whereas [26] adopted PaP > 10. The lack of standardization across studies obscures the evidence base of predictive performance. However, it also highlights the need to experiment with different cut-offs in a new cohort. The effect of altering cut-offs on model performance was less thoroughly studied and reported in the studies of our review.

Timing of measurement can also affect the accuracy of survival prediction. Some studies included in this review suggest that serial measurements give more reliable prediction. For instance, score changes alone and combination with initial score have been investigated for PPI [31]. Whether the same effect can be appreciated in other models remains under-explored. Regardless of this preliminary finding, timing of measurements needs to be standardized. There were appreciable variations across studies in terms of when model factors were assessed, particularly in relation to previous treatments and palliative care referral. The timing was not always clearly defined in studies either. Understandably, this is heavily dependent on resources and guidelines in localities, but alignment should be sought within one setting and that assessment time for survival prediction should be clearly defined in the journey of care. Theoretically, survival should be counted from the time assessment is done. Use of earlier or later test results for current prediction should be minimized for the purpose of model development and validation.

The body of evidence we have gathered is bound by several problems that may hinder immediate translation into clinical practice. In particular, the wide prediction intervals and high I2 values suggested significant heterogeneity across studies. Palliative care settings and referral criteria differ across localities [8,9,10,11]. Disease and treatment statuses of patients at and before recruitment into studies were therefore heterogeneous and altered the prognostic trajectories. However, certain studies underreport eligibility criteria and/or sample characteristics, rendering pooling and sub-group analyses difficult [20, 25, 33,34,35]. Moreover, dichotomization of otherwise continuous variables (such as blood results), inadequate testing or reporting of model assumptions, a lack of account for missing data, and the absence of calibration plots create uncertainties about the strength of evidence presented, as shown in the high risks of bias of many included studies [41]. With the advent of machine learning and artificial intelligence, it is essential to keep updated with newest guidelines on analysis and reporting [67]. To enable a more precise and directive recommendation from reviews and meta-analyses, future studies may compare performance of different prognostic tools in the same specific patient subgroups or healthcare contexts, such that the discrepancies in performance statistics can be pooled and whose consistencies can be assessed.

Conclusion

Reliable prognostication is essential to inform both patients and clinicians in their planning of palliative care. This review addresses several gaps in existing literature by focusing specifically on patients with advanced solid tumors receiving palliative care, excluding hematological malignancies which have distinct disease trajectories. Through conducting comprehensive meta-analyses of model performance and providing direct comparisons between prognostic tools, our review offers insights specific to this important patient population.

Our study provided preliminary evidence that PaP had a higher discriminative ability than PPI. Yet, definitive conclusions cannot be made as many studies have significant methodological limitations such as the lack of comprehensive statistical testing, failure to report missing data handling, and omission of critical demographic information such as treatment status.

It remains uncertain if accurate prognostication methods would translate into superior clinical care. Future RCTs should investigate the clinical impacts of utilizing different prognostic models on palliative care, advanced care planning, resource allocation, hospice referrals, end-of-life discussions, etc. The feasibility, cost-effectiveness, and patient acceptance of prognostic models should be explored as well.

Data availability

No datasets were generated or analysed during the current study.

References

  1. Ferrell BR, Temel JS, Temin S, Alesi ER, Balboni TA, Basch EM, et al. Integration of palliative care into standard oncology care: American Society of Clinical Oncology clinical practice guideline update. J Clin Oncol. 2017;35(1):96–112.

    Article  PubMed  Google Scholar 

  2. Haun MW, Estel S, Rücker G, Friederich HC, Villalobos M, Thomas M, et al. Early palliative care for adults with advanced cancer. Cochrane Database Syst Rev. 2017;6(6):Cd011129.

  3. O’Connor NR, Moyer ME, Behta M, Casarett DJ. The impact of inpatient palliative care consultations on 30-day hospital readmissions. J Palliat Med. 2015;18(11):956–61.

    Article  PubMed  Google Scholar 

  4. Temel JS, Greer JA, Muzikansky A, Gallagher ER, Admane S, Jackson VA, et al. Early palliative care for patients with metastatic non-small-cell lung cancer. N Engl J Med. 2010;363(8):733–42.

    Article  CAS  PubMed  Google Scholar 

  5. Zimmermann C, Swami N, Krzyzanowska M, Hannon B, Leighl N, Oza A, et al. Early palliative care for patients with advanced cancer: a cluster-randomised controlled trial. Lancet. 2014;383(9930):1721–30.

    Article  PubMed  Google Scholar 

  6. Dionne-Odom JN, Azuero A, Lyons KD, Hull JG, Tosteson T, Li Z, et al. Benefits of early versus delayed palliative care to informal family caregivers of patients with advanced cancer: outcomes from the ENABLE III randomized controlled trial. J Clin Oncol. 2015;33(13):1446–52.

    Article  PubMed  PubMed Central  Google Scholar 

  7. Dyar S, Lesperance M, Shannon R, Sloan J, Colon-Otero G. A nurse practitioner directed intervention improves the quality of life of patients with metastatic cancer: results of a randomized pilot study. J Palliat Med. 2012;15(8):890–5.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Scibetta C, Kerr K, McGuire J, Rabow MW. The costs of waiting: implications of the timing of palliative care consultation among a cohort of decedents at a comprehensive cancer center. J Palliat Med. 2016;19(1):69–75.

    Article  PubMed  Google Scholar 

  9. Bennett MI, Ziegler L, Allsop M, Daniel S, Hurlow A. What determines duration of palliative care before death for patients with advanced disease? A retrospective cohort study of community and hospital palliative care provision in a large UK city. BMJ Open. 2016;6(12): e012576.

    Article  PubMed  PubMed Central  Google Scholar 

  10. Bakitas M, Lyons KD, Hegel MT, Balan S, Barnett KN, Brokaw FC, et al. The project ENABLE II randomized controlled trial to improve palliative care for rural patients with advanced cancer: baseline findings, methodological challenges, and solutions. Palliat Support Care. 2009;7(1):75–86.

    Article  PubMed  PubMed Central  Google Scholar 

  11. Rabow MW, Dibble SL, Pantilat SZ, McPhee SJ. The comprehensive care team: a controlled trial of outpatient palliative medicine consultation. Arch Intern Med. 2004;164(1):83–91.

    Article  PubMed  Google Scholar 

  12. Brumley R, Enguidanos S, Jamison P, Seitz R, Morgenstern N, Saito S, et al. Increased satisfaction with care and lower costs: results of a randomized trial of in-home palliative care. J Am Geriatr Soc. 2007;55(7):993–1000.

    Article  PubMed  Google Scholar 

  13. Osman H, Shrestha S, Temin S, Ali ZV, Corvera RA, Ddungu HD, et al. Palliative care in the global setting: ASCO resource-stratified practice guideline. J Glob Oncol. 2018;4:1–24.

    PubMed  Google Scholar 

  14. Kim YJ, Yoon SJ, Suh S-Y, Hiratsuka Y, Kang B, Lee SW, et al. Performance of clinician prediction of survival in oncology outpatients with advanced cancer. PLoS ONE. 2022;17(4): e0267467.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Cheon S, Agarwal A, Popovic M, Milakovic M, Lam M, Fu W, et al. The accuracy of clinicians’ predictions of survival in advanced cancer: a review. Ann Palliat Med. 2016;5(1):22–9.

    PubMed  Google Scholar 

  16. Gramling R, Gajary-Coots E, Cimino J, Fiscella K, Epstein R, Ladwig S, et al. Palliative care clinician overestimation of survival in advanced cancer: disparities and association with end-of-life care. J Pain Symptom Manage. 2019;57(2):233–40.

    Article  PubMed  Google Scholar 

  17. Elting LS, Cooksley C, Bekele BN, Frumovitz M, Avritscher EB, Sun C, et al. Generalizability of cancer clinical trial results: prognostic differences between participants and nonparticipants. Cancer. 2006;106(11):2452–8.

    Article  PubMed  Google Scholar 

  18. Simmons CPL, McMillan DC, McWilliams K, Sande TA, Fearon KC, Tuck S, et al. Prognostic tools in patients with advanced cancer: a systematic review. J Pain Symptom Manage. 2017;53(5):962-70.e10.

    Article  PubMed  Google Scholar 

  19. Moons KGM, De Groot JAH, Bouwmeester W, Vergouwe Y, Mallett S, Altman DG, et al. Critical appraisal and data extraction for systematic reviews of prediction modelling studies: the CHARMS Checklist. PLoS Med. 2014;11(10):e1001744.

    Article  PubMed  PubMed Central  Google Scholar 

  20. Wolff RF, Moons KGM, Riley RD, Whiting PF, Westwood M, Collins GS, et al. PROBAST: a tool to assess the risk of bias and applicability of prediction model studies. Ann Intern Med. 2019;170(1):51.

    Article  PubMed  Google Scholar 

  21. Chow E, Abdolell M, Panzarella T, Harris K, Bezjak A, Warde P, et al. Predictive model for survival in patients with advanced cancer. J Clin Oncol. 2008;26(36):5863–9.

    Article  PubMed  Google Scholar 

  22. Chow E, Abdolell M, Panzarella T, Harris K, Bezjak A, Warde P, et al. Validation of a predictive model for survival in metastatic cancer patients attending an outpatient palliative radiotherapy clinic. Int J Radiat Oncol Biol Phys. 2009;73(1):280–7.

    Article  PubMed  Google Scholar 

  23. Chiang JK, Lai NS, Wang MH, Chen SC, Kao YH. A proposed prognostic 7-day survival formula for patients with terminal cancer. BMC Public Health. 2009;9: 365.

    Article  PubMed  PubMed Central  Google Scholar 

  24. Chow E, Abdolell M, Panzarella T, Harris K, Bezjak A, Warde P, et al. Recursive partitioning analysis of prognostic factors for survival in patients with advanced cancer. Int J Radiat Oncol Biol Phys. 2009;73(4):1169–76.

    Article  PubMed  Google Scholar 

  25. Scarpi E, Maltoni M, Miceli R, Mariani L, Caraceni A, Amadori D, et al. Survival prediction for terminally ill cancer patients: revision of the palliative prognostic score with incorporation of delirium. Oncologist. 2011;16(12):1793–9.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Cheng WH, Kao CY, Hung YS, Su PJ, Hsieh CH, Chen JS, et al. Validation of a palliative prognostic index to predict life expectancy for terminally ill cancer patients in a hospice consultation setting in Taiwan. Asian Pac J Cancer Prev. 2012;13(6):2861–6.

    Article  PubMed  Google Scholar 

  27. Durand JP, Mir O, Coriat R, Cessot A, Pourchet S, Goldwasser F. Validation of the Cochin Risk Index Score (CRIS) for life expectancy prediction in terminally ill cancer patients. Support Care Cancer. 2012;20(4):857–64.

    Article  PubMed  Google Scholar 

  28. Maltoni M, Scarpi E, Pittureri C, Martini F, Montanari L, Amaducci E, et al. Prospective comparison of prognostic scores in palliative care cancer populations. Oncologist. 2012;17(3):446–54.

    Article  PubMed  PubMed Central  Google Scholar 

  29. Huang Y, Xi Q, Xia S, Wang X, Liu Y, Huang C, et al. Development and validation of a prognostic scale for hospitalized patients with terminally ill cancer in China. Support Care Cancer. 2014;22(1):145–52.

    Article  PubMed  Google Scholar 

  30. Hung CY, Wang HM, Kao CY, Lin YC, Chen JS, Hung YS, et al. Magnitude of score change for the palliative prognostic index for survival prediction in patients with poor prognostic terminal cancer. Support Care Cancer. 2014;22(10):2725–31.

    Article  PubMed  Google Scholar 

  31. Kao CY, Hung YS, Wang HM, Chen JS, Chin TL, Lu CY, et al. Combination of initial palliative prognostic index and score change provides a better prognostic value for terminally ill cancer patients: a six-year observational cohort study. J Pain Symptom Manage. 2014;48(5):804–14.

    Article  PubMed  Google Scholar 

  32. Kim AS, Youn CH, Ko HJ, Kim HM. The survival time of terminal cancer patients: prediction based on clinical parameters and simple prognostic scores. J Palliat Care. 2014;30(1):24–31.

    Article  PubMed  Google Scholar 

  33. Yoon SJ, Jung JG, Kim JS, Kim SS, Kim S. Retrospective assessment of objective prognostic score in terminally ill Korean patients with cancer. Am J Hosp Palliat Care. 2014;31(4):435–40.

    Article  PubMed  Google Scholar 

  34. Chen YT, Ho CT, Hsu HS, Huang PT, Lin CY, Liu CS, et al. Objective palliative prognostic score among patients with advanced cancer. J Pain Symptom Manage. 2015;49(4):690–6.

    Article  PubMed  Google Scholar 

  35. Chiang JK, Koo M, Kao YH. Development of a user-friendly graphic tool to estimate individualized survival curves for advanced cancer patients in hospice care. J Palliat Care. 2015;31(1):29–35.

    Article  PubMed  Google Scholar 

  36. Mendis R, Soo WK, Zannino D, Michael N, Spruyt O. Multidisciplinary prognostication using the palliative prognostic score in an Australian Cancer Center. Palliative Care. 2017;9:7–14.

    Google Scholar 

  37. Bourgeois H, Grudé F, Solal-Céligny P, Dupuis O, Voog E, Ganem G, et al. Clinical validation of a prognostic tool in a population of outpatients treated for incurable cancer undergoing anticancer therapy: PRONOPALL study. Ann Oncol. 2017;28(7):1612–7.

    Article  CAS  PubMed  Google Scholar 

  38. Yoon SJ, Suh SY, Lee YJ, Park J, Hwang S, Lee SS, et al. Prospective validation of objective prognostic score for advanced cancer inpatients in South Korea: a multicenter study. J Palliat Med. 2017;20(1):65–8.

    Article  PubMed  Google Scholar 

  39. Adelson K, Lee DKK, Velji S, Ma J, Lipka SK, Rimar J, et al. Development of Imminent Mortality Predictor for Advanced Cancer (IMPAC), a tool to predict short-term mortality in hospitalized patients with advanced cancer. Journal of Oncology Practice. 2018;14(3):e168–75.

    Article  PubMed  Google Scholar 

  40. Palomar-Muñoz C, Martín-Zamorano M, Mogollo A, Pascual-Pérez S, Rodríguez-Morales I, Girón-González JA. Assessment of the Palliative Prognostic Index in hospitalized oncologic patients treated by a palliative care team: impact of acute concomitant diseases. Oncotarget. 2018;9(27):18908–15.

    Article  PubMed  PubMed Central  Google Scholar 

  41. Ermacora P, Mazzer M, Isola M, Pascoletti G, Gregoraci G, Basile D, et al. Prognostic evaluation in palliative care: final results from a prospective cohort study. Supportive Care Cancer. 2019;27:2095 +.

    Article  Google Scholar 

  42. Hamano J, Takeuchi A, Yamaguchi T, Baba M, Imai K, Ikenaga M, et al. A combination of routine laboratory findings and vital signs can predict survival of advanced cancer patients without physician evaluation: a fractional polynomial model. Eur J Cancer. 2018;105:50–60.

    Article  PubMed  Google Scholar 

  43. Zhao W, He Z, Li Y, Jia H, Chen M, Gu X, et al. Nomogram-based parameters to predict overall survival in a real-world advanced cancer population undergoing palliative care. BMC Palliat Care. 2019;18(1):47.

    Article  PubMed  PubMed Central  Google Scholar 

  44. Arkin FS, Aras G, Dogu E. Comparison of artificial neural networks and logistic regression for 30-days survival prediction of cancer patients. Acta Inform Med. 2020;28(2):108–13.

    Article  PubMed  PubMed Central  Google Scholar 

  45. Hum A, Wong YKY, Yee CM, Lee CS, Wu HY, Koh MYH. PROgnostic Model for Advanced Cancer (PRO-MAC). BMJ Support Palliat Care. 2020;10(4): e34.

    Article  PubMed  Google Scholar 

  46. Miyagi T, Miyata S, Tagami K, Hiratsuka Y, Sato M, Takeda I, et al. Prognostic model for patients with advanced cancer using a combination of routine blood test values. Support Care Cancer. 2021;29(8):4431–7.

    Article  PubMed  Google Scholar 

  47. Yang C, Yu R, Ji H, Jiang H, Yang W, Jiang F. Application of data mining in the provision of in-home medical care for patients with advanced cancer. Transl Cancer Res. 2021;10(6):3013–9.

    Article  PubMed  PubMed Central  Google Scholar 

  48. Chan AS, Rout A, Adamo CRD, Lev I, Yu A, Miller K. Palliative referrals in advanced cancer patients: utilizing the supportive and palliative care indicators tool and rothman index. Am J Hosp Palliat Care. 2022;39(2):164–8.

    Article  PubMed  Google Scholar 

  49. Hiratsuka Y, Yoon SJ, Suh SY, Choi SE, Hui D, Kim SH, et al. Comparison of the accuracy of clinicians’ prediction of survival and palliative prognostic score: an East Asian cross-cultural study. Support Care Cancer. 2022;30(3):2367–74.

    Article  PubMed  Google Scholar 

  50. Hiratsuka Y, Kim D, Suh SY, Kim SH, Yoon SJ, Koh SJ, et al. Comparison of objective prognostic score and palliative prognostic Score performance in inpatients with advanced cancer in Japan and Korea. Palliat Support Care. 2022;20(5):662–70.

    PubMed  Google Scholar 

  51. Hiratsuka Y, Suh SY, Hui D, Morita T, Mori M, Oyamada S, et al. Are prognostic scores better than clinician judgment? A prospective study using three models. J Pain Symptom Manage. 2022;64(4):391–9.

    Article  PubMed  Google Scholar 

  52. Owusuaa C, van der Padt-Pruijsten A, Drooger JC, Heijns JB, Dietvorst AM, Janssens-van Vliet ECJ, et al. Development of a clinical prediction model for 1-year mortality in patients with advanced cancer. JAMA Network Open. 2022;5(11):e2244350-e.

  53. Preto DD, Paiva BSR, Hui D, Bruera E, Paiva CE. HAprog: a new prognostic application to assist oncologists in routine care. J Pain Symptom Manage. 2022;63(6):1014-21.e4.

    Article  PubMed  Google Scholar 

  54. Scarpi E, Nanni O, Maltoni M. Development and validation of the PaP score nomogram for terminally Ill cancer patients. Cancers (Basel). 2022;14(10):2510.

  55. Zachariah FJ, Rossi LA, Roberts LM, Bosserman LD. Prospective comparison of medical oncologists and a machine learning model to predict 3-month mortality in patients with metastatic solid tumors. JAMA Netw Open. 2022;5(5): e2214514.

    Article  PubMed  PubMed Central  Google Scholar 

  56. Pobar I, Job M, Holt T, Hargrave C, Hickey B. Prognostic tools for survival prediction in advanced cancer patients: a systematic review. J Med Imaging Radiat Oncol. 2021;65(6):806–16.

    Article  PubMed  Google Scholar 

  57. Owusuaa C, van der Padt-Pruijsten A, Drooger JC, Heijns JB, Dietvorst AM, Janssens-van Vliet ECJ, et al. Development of a clinical prediction model for 1-year mortality in patients with advanced cancer. JAMA Netw Open. 2022;5(11): e2244350.

    Article  PubMed  PubMed Central  Google Scholar 

  58. Downing M, Lau F, Lesperance M, Karlson N, Shaw J, Kuziemsky C, et al. Meta-analysis of survival prediction with palliative performance scale. J Palliat Care. 2007;23(4):245–54.

    Article  PubMed  Google Scholar 

  59. Yoong SQ, Bhowmik P, Kapparath S, Porock D. Palliative prognostic scores for survival prediction of cancer patients: a systematic review and meta-analysis. J Natl Cancer Inst. 2024;116(6):829–57.

    Article  PubMed  PubMed Central  Google Scholar 

  60. Stone P, Buckle P, Dolan R, Feliu J, Hui D, Laird BJA, et al. Prognostic evaluation in patients with advanced cancer in the last months of life: ESMO Clinical Practice Guideline. ESMO Open. 2023;8(2): 101195.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  61. Baba M, Maeda I, Morita T, Inoue S, Ikenaga M, Matsumoto Y, et al. Survival prediction for advanced cancer patients in the real world: a comparison of the palliative prognostic score, delirium-palliative prognostic score, palliative prognostic index and modified prognosis in palliative care study predictor model. Eur J Cancer. 2015;51(12):1618–29.

    Article  PubMed  Google Scholar 

  62. Stone P, Kalpakidou A, Todd C, Griffiths J, Keeley V, Spencer K, et al. Prognostic models of survival in patients with advanced incurable cancer: the pips2 observational study. Health Technol Assess. 2021;25(28):1.

    Article  PubMed  PubMed Central  Google Scholar 

  63. Hui D, Paiva CE, Del Fabbro EG, Steer C, Naberhuis J, van de Wetering M, et al. Prognostication in advanced cancer: update and directions for future research. Support Care Cancer. 2019;27(6):1973–84.

    Article  PubMed  PubMed Central  Google Scholar 

  64. Zakir D, Mekonnen GK, Negash B, Marami D. Level of health laboratory service quality, service interruptions, and its predictors in public hospitals in Harar town, eastern Ethiopia. Front Health Serv. 2024;4: 1492766.

    Article  PubMed  PubMed Central  Google Scholar 

  65. Stone P, Vickerstaff V, Kalpakidou A, Todd C, Griffiths J, Keeley V, et al. Prognostic tools or clinical predictions: which are better in palliative care? PLoS ONE. 2021;16(4): e0249763.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  66. Steyerberg EW, Moons KGM, Van Der Windt DA, Hayden JA, Perel P, Schroter S, et al. Prognosis Research Strategy (PROGRESS) 3: prognostic model research. PLoS Med. 2013;10(2): e1001381.

    Article  PubMed  PubMed Central  Google Scholar 

  67. Collins GS, Moons KGM, Dhiman P, Riley RD, Beam AL, Van Calster B, et al. TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ. 2024;385:e078378.

    Article  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

None.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Author information

Authors and Affiliations

  1. Faculty of Medicine, The Chinese University of Hong Kong, Shatin, Hong Kong SAR, China

    Mong Yung Fung & Yuen Lung Wong

  2. Department of Clinical Oncology, Queen Elizabeth Hospital, Hong Kong SAR, China

    Ka Man Cheung, King Hei Kelvin Bao & Winnie Wing Yan Sung

Authors
  1. Mong Yung Fung
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Yuen Lung Wong
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Ka Man Cheung
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. King Hei Kelvin Bao
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Winnie Wing Yan Sung
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

All authors contributed to the conception and design. F.M.Y. and W.Y.L. wrote the main manuscript text. F.M.Y. performed the statistical analysis. All authors reviewed the manuscript.

Corresponding author

Correspondence to Mong Yung Fung.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

All authors consent to the publication of this manuscript.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Fung, M., Wong, Y., Cheung, K. et al. Prognostic models for survival predictions in advanced cancer patients: a systematic review and meta-analysis. BMC Palliat Care 24, 54 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12904-025-01696-4

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12904-025-01696-4

Keywords

BMC Palliative Care

ISSN: 1472-684X

Contact us