Next Article in Journal
The Assessment of Race, the 21-Gene Recurrence Score, and Breast Cancer Outcomes at Kaiser Permanente
Previous Article in Journal
The Plant Steroid Hormone, Epibrassinolide, Reverses Drug Resistance in Small-Cell Lung Carcinoma Cells by Modulating Ferroptosis
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Cancers
Volume 18
Issue 16
10.3390/cancers18162660
cancers-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Incorporating Metastatic Disease Burden into Prognostic Assessment in Metastatic Castration-Resistant Prostate Cancer: The MIRA Score

by
Mario Uccello
1,
Panagiotis J. Vlachostergios
2,3,4,5,
Aruni Ghose
6,7 and
Stergios Boussios
3,4,7,8,9,10,*
1
Royal United Hospitals Bath NHS Foundation Trust, Bath BA1 3NG, UK
2
Department of Medical Oncology, IASO Thessalias Hospital, 41500 Larissa, Greece
3
Department of Medical Oncology, Ioannina University Hospital, 45500 Ioannina, Greece
4
Faculty of Medicine, School of Health Sciences, University of Ioannina, 45110 Ioannina, Greece
5
Division of Hematology & Medical Oncology, Department of Medicine, Weill Cornell Medicine, New York, NY 10065, USA
6
Cancer Division, University College London Hospitals NHS Foundation Trust, London NW1 2BU, UK
7
Department of Research and Innovation, Medway NHS Foundation Trust, Gillingham ME7 5NY, UK
8
School of Cancer and Pharmaceutical Sciences, Faculty of Life Sciences and Medicine, King’s College London, London SE1 1UL, UK
9
Faculty of Medicine, Health and Social Care, Canterbury Christ Church University, Canterbury CT1 1QU, UK
10
AELIA Organisation, 9Th Km Thessaloniki-Thermi, 57001 Thessaloniki, Greece
*
Author to whom correspondence should be addressed.
Cancers 2026, 18(16), 2660; https://doi.org/10.3390/cancers18162660
Submission received: 16 July 2026 / Revised: 9 August 2026 / Accepted: 17 August 2026 / Published: 18 August 2026
(This article belongs to the Section Tumor Microenvironment)
Browse Figure
Versions Notes

Simple Summary

People with advanced prostate cancer can have very different survival outcomes, even when their disease appears similar. This exploratory study developed a tool, called MIRA, to estimate survival in men whose prostate cancer had spread and had become resistant to standard hormone treatment. Researchers analysed anonymised data from 266 men in the placebo group of a historical clinical trial. The score combined the amount of cancer in the bones and other measurable sites, with information including age, pain, speed of disease progression, inflammation and blood test results. Patients were divided into three different prognostic groups. In this dataset, the MIRA groups performed better than an established tool, called the Halabi classification. The findings suggest that measuring the overall amount of cancer may improve prognostic assessment. However, MIRA must be refined and tested in contemporary patient populations before its use in clinical practice.

Abstract

Background: Prognostic assessment in metastatic castration-resistant prostate cancer (mCRPC) remains important, but metastatic burden is not routinely captured as a simple combined measure. This exploratory study developed the Metastatic Integrated Risk Assessment (MIRA) score using disease burden and routine variables. Methods: Anonymised placebo-arm individual patient data from D4320C00014/ENTHUSE-M1 were analysed. Baseline variables were assessed for association with overall survival (OS) using Cox regression. A hierarchical composite tumour-burden variable combined bone-metastasis categories with RECIST target-lesion presence and sum of longest diameters (SLD). Selected variables formed an additive score, stratified into three risk groups, and were compared with the reconstructed Halabi classification using Harrell’s C-index. The final multivariate Cox model underwent internal validation using 1000 bootstrap resamples. Results: Among 266 placebo-treated patients, 133 deaths occurred, and median OS was 22.21 months. Composite tumour burden showed the strongest prognostic association and remained independently associated with OS. MIRA risk groups showed clear OS separation. In this derivation cohort, MIRA showed higher apparent discrimination than the reconstructed Halabi classification, with C-indices of 0.755 and 0.646, respectively; the paired difference was statistically significant (p < 0.001). For the final multivariate Cox model, the apparent C-index was 0.779, and the optimism-corrected C-index was 0.760; the optimism-corrected calibration slope was 0.841. Conclusions: These exploratory findings support metastatic tumour burden as a relevant prognostic factor in mCRPC and suggest that a simple combined measure of skeletal and measurable soft-tissue disease may have value for risk stratification. Refinement and independent external validation of MIRA are required before clinical use.

1. Introduction

Survival in metastatic castration-resistant prostate cancer (mCRPC) is highly variable. Patients with the same broad disease state may differ substantially in metastatic extent, symptom burden, prior treatment history, systemic inflammation, laboratory profile and physiological reserve. Prognostic assessment remains important for counselling, trial stratification and interpretation of survival outcomes [1,2]. Several prognostic models have been developed in advanced prostate cancer, including the Smaletz–Kattan nomogram [3], the TAX327 nomogram [4], Halabi prognostic models [1,5,6], Templeton et al.’s [7] simple prognostic score, incorporating the neutrophil-to-lymphocyte ratio (NLR), Chi et al.’s [8] prognostic index model for abiraterone-treated mCRPC, the PREVAIL Model for Prostate Cancer Survival [2] and the Prostate Cancer DREAM Challenge model [9]. Across these approaches, recurrent adverse prognostic domains include performance status, pain, anaemia, alkaline phosphatase (ALP), lactate dehydrogenase (LDH), prostate-specific antigen (PSA), metastatic site, inflammatory status and opioid use. However, some models require weighted calculations, are anchored to specific treatment settings, or depend on more complex analytic methods. This may limit their use as simple, transparent risk classifications across heterogeneous clinical-trial and real-world datasets [1,2,8]. Metastatic disease burden is one of the most clinically apparent features of advanced cancer, but it is not always captured in a practical or consistent form in mCRPC prognostic models [1,8,10]. This is particularly relevant in prostate cancer, where metastatic involvement is commonly skeletal and may not be fully represented by conventional measurable-disease criteria. A prognostic score that incorporates both skeletal and measurable soft-tissue disease burden, together with routinely available clinical and laboratory variables, may provide a reproducible approach to risk stratification without requiring specialised modelling infrastructure or molecular data. The Metastatic Integrated Risk Assessment for the mCRPC (MIRA) score was developed using anonymised placebo-arm patient-level data from the ENTHUSE-M1 trial [11]. The aim was to construct a simple prognostic score based on routinely available baseline variables, with explicit incorporation of metastatic disease burden, and to compare its discriminatory performance with the established Halabi classification.

2. Materials and Methods

2.1. Data Source and Study Population

This was a secondary analysis of anonymised individual patient-level data accessed through Project Data Sphere [12]. The analysis used the available placebo-arm dataset from D4320C00014/ENTHUSE-M1 [11], a phase III, randomised, double-blind, placebo-controlled trial evaluating ZD4054 versus placebo in men with hormone-resistant prostate cancer and bone metastases who were pain-free or mildly symptomatic. Only patients allocated to placebo were included, as these were the only patient-level data available for the present analysis. Available study documentation, including the clinical study protocol, case report form, data files and accompanying format files, was reviewed to identify relevant baseline clinical, laboratory, patient-reported and imaging variables. Overall survival (OS) was defined as the time from randomisation to death from any cause. Patients alive at last follow-up were censored.

2.2. Variables and Data Preparation

Baseline demographic, clinical, laboratory, patient-reported and imaging variables were extracted and harmonised at patient level. Laboratory variables were based on baseline values. In rare cases where a baseline laboratory value was missing, the next available value was used; this was considered acceptable in the context of a placebo-arm prognostic analysis. The frequency and timing of these substitutions were quantified for each laboratory variable. Candidate predictors were identified from the available baseline variables on the basis of established or plausible prognostic relevance in mCRPC, routine clinical availability and reproducibility. Clinically established or dataset-prespecified cut-offs were used rather than thresholds optimised against OS in the present cohort. For the present exploratory analysis, a hierarchical composite tumour-burden variable was developed using both the number of bone metastases and Response Evaluation Criteria in Solid Tumours (RECIST) target-lesion burden. The bone-metastasis categories already recorded in the dataset, including <5, 5–20 and ≥21 lesions, were used, while the RECIST sum of the longest diameters (SLD) threshold of >3 cm was selected because 3 cm corresponded to the first tertile of the observed SLD distribution among patients with RECIST target lesions. Composite tumour-burden level was assigned hierarchically. Level 2 was assigned to patients with ≥21 bone metastases or RECIST SLD > 3 cm. Among the remaining patients, level 1 was assigned to those with 5–20 bone metastases or any RECIST target lesion. Level 0 was assigned to those with <5 bone metastases and no RECIST target lesion. Other evaluated variables included age, race, baseline World Health Organization (WHO)/Eastern Cooperative Oncology Group (ECOG) performance status, time from first oncological therapy, NLR, ALP, haemoglobin (Hb), PSA, LDH, albumin and platelet count. Baseline symptom burden was assessed using the Brief Pain Inventory (BPI) and the EQ-5D pain/discomfort dimension; the selected thresholds were BPI ≥ 1 and EQ-5D pain/discomfort ≥ 2. The composite was considered positive if either threshold was met. For patients with one of the two measures unavailable, the other was used; no patient had both measures missing, so the composite BPI/EQ-5D variable was available for all 266 patients. When one or more MIRA components were missing, the exact total score was considered unavailable. Risk-group assignment was retained when the available components were sufficient to determine the category unequivocally.

2.3. Statistical Analysis

Associations between candidate variables and OS were first assessed using univariate Cox proportional-hazards analysis. Variables considered for multivariate modelling and score construction were selected manually, rather than through a formal automated procedure, according to statistical evidence, clinical interpretability, established prognostic relevance and ease of reproducibility. Variables independently associated with OS were prioritised, while variables with borderline adjusted associations or strong univariate associations were also considered when routinely available and clinically meaningful. Particular attention was given to tumour burden because of its expected prognostic relevance. Opioid use at baseline was available and was used for reconstruction of the Halabi score, but it was not evaluated as a candidate prognostic variable for the MIRA score because it was reported in only five patients. The MIRA score for mCRPC was constructed as an additive point-based prognostic score. The resulting score included age, symptom burden assessed by BPI/EQ-5D, composite tumour-burden level, NLR and time since first prior oncological therapy, together with ALP, Hb and PSA, which were retained because of their established prognostic relevance and strong univariate or borderline multivariate associations in this cohort. Among laboratory variables not independently associated with OS, PSA was retained because of its stronger univariate association, disease-specific prognostic relevance and greater likelihood of routine availability across future clinical and research datasets, whereas LDH showed a weaker univariate association in this cohort and was excluded to preserve parsimony and reproducibility. Point allocation was pragmatic rather than derived by formal scaling or rounding of the Cox regression coefficients. The hierarchical tumour-burden component was weighted progressively according to increasing disease burden, while time since first prior oncological therapy <2 years was assigned two points to reflect its adverse prognostic association and its interpretation as a marker of aggressive disease tempo. Age ≥ 80 years was deliberately limited to one point despite the strength of its association with OS, because chronological age was considered a non-specific host factor and a potential surrogate for more direct measures of comorbidity, frailty, physiological reserve and functional status. All other predictors were assigned one point to preserve the simplicity of the score. OS according to MIRA risk group was estimated using the Kaplan–Meier method and compared using the log-rank test. Cox proportional-hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs), with the low-risk group used as the reference category. For benchmarking, the updated Halabi prognostic score for metastatic castration-resistant prostate cancer was reconstructed using available baseline variables from the placebo dataset, including opioid analgesic use, in accordance with the published 2014 model [1]. Patients were classified according to the published three-tier Halabi risk-group classification, allowing direct comparison with MIRA. Discrimination was assessed using Harrell’s C-index. C-indices for the two classifications were compared using the non-parametric method for two correlated C-indices with right-censored survival outcomes proposed by Kang et al. [13]. This method accounts for the paired nature of the comparison, because both classifications were evaluated in the same patients, and estimates the variance of the difference between C-indices using an asymptotic normal approximation. As a sensitivity analysis, the MIRA risk-group Cox model and Harrell’s C-index were recalculated after excluding patients for whom a post-baseline laboratory measurement had been used for at least one score component. The final multivariate Cox model included 10 predictors. Because the three-level composite tumour-burden variable required two coefficients, 11 regression parameters were estimated in total. With 131 deaths, this corresponded to approximately 11.9 events per parameter. The number of regression parameters was limited relative to the number of observed events to reduce the risk of overfitting. The proportional-hazards assumption was assessed for the final multivariate Cox model and the MIRA risk-group Cox model using scaled Schoenfeld residuals with the Kaplan–Meier transformation of time. Where covariate-specific evidence of non-proportionality was observed, a sensitivity analysis incorporating time-dependent effects was performed. Internal validation of the final multivariate Cox model was performed using 1000 bootstrap resamples. The complete-case analysis dataset used to fit the final multivariate model was resampled with replacement, and the same final set of covariates was refitted in each bootstrap sample. Optimism was estimated by comparing model discrimination within each bootstrap sample with its performance when applied to the original dataset. Mean optimism was then subtracted from the apparent C-index to obtain the optimism-corrected C-index. An optimism-corrected calibration slope and bootstrap 95% CIs were also estimated. Calibration of predicted survival probabilities was assessed at 12 and 24 months using bootstrap-corrected calibration plots comparing predicted with observed survival. Risk groups were coded so that higher values indicated higher mortality risk. Two-sided p values were reported, with p < 0.05 considered statistically significant. C-indices and their confidence intervals were reported to three decimal places. All statistical analyses were performed using MedCalc Statistical Software version 23.6.1, R version 4.6.1 and RStudio version 2026.04.0 + 526, with the survival, compareC and rms packages used where appropriate.

3. Results

3.1. Patient Characteristics and Baseline Tumour Burden

The cohort included 266 placebo-assigned patients with hormone-resistant prostate cancer and bone metastases. At data cut-off, 133 deaths had occurred and 133 patients were censored. Median OS was 22.21 months (95% CI, 18.76–23.23). Baseline demographic, clinical, laboratory and disease-burden characteristics are summarised in Table 1. Missing baseline laboratory values were infrequent. The earliest post-baseline measurement was used for one to seven patients per variable (0.4–2.6%), while one platelet value remained unavailable. The frequency and timing of these substitutions are reported in Supplementary Table S1. Overall, 43 patients (16.2%) were aged ≥80 years and 70 (26.3%) had a WHO/ECOG performance status 1. Although patients were pain-free or mildly symptomatic at trial entry, baseline patient-reported measures identified a measurable symptom burden: BPI worst pain ≥1 was recorded in 132 of 258 evaluable patients (51.2%), and EQ-5D pain/discomfort level 2 in 103 of 258 (39.9%). The composite BPI/EQ-5D variable was positive in 153 of 266 patients (57.5%). Overall, 189 patients (71.1%) had ≥5 bone metastases and 53 (19.9%) had ≥21 bone metastases. A total of 265 patients were evaluable for visceral disease, nodal disease, RECIST target lesions and composite tumour-burden level. Visceral metastases were documented in 45 patients (17.0%) and nodal metastases in 80 (30.2%). RECIST target lesions were present in 94 patients (35.5%), with a RECIST SLD > 3 cm in 58 (21.9%). Composite tumour-burden level was classified as level 0 in 52 patients (19.6%), level 1 in 120 (45.3%) and level 2 in 93 (35.1%).

3.2. Univariate and Multivariate Associations with OS

In univariate Cox regression analysis, several demographic, clinical, laboratory and disease-burden variables were associated with OS (Table 2). These included age ≥80 years, a positive BPI/EQ-5D composite, time since first prior oncological therapy <2 years, Hb < 120 g/L, NLR > 2.5, albumin ≤ 40 g/L, ALP > 130 U/L, LDH > 240 U/L, PSA ≥ 100 ng/mL, higher bone-metastatic burden, nodal metastases, presence of RECIST target lesions, RECIST SLD > 3 cm and higher composite tumour-burden level. The strongest univariate associations were observed for composite tumour-burden level, with HRs of 4.35 (95% CI, 2.04–9.27) for level 1 and 10.38 (95% CI, 4.88–22.09) for level 2, compared with level 0. A multivariate Cox model including composite tumour-burden level and candidate clinical and laboratory variables was fitted in 264 patients with complete covariate data, including 131 deaths (Table 3). The overall model was statistically significant (χ2 = 115.806, df = 11, p < 0.001), with a Harrell’s C-index of 0.779 (95% CI, 0.741–0.818). In this model, composite tumour-burden level remained independently associated with OS: compared with level 0, HRs were 2.62 (95% CI, 1.20–5.69; p = 0.015) for level 1 and 5.52 (95% CI, 2.52–12.11; p < 0.001) for level 2. Other variables independently associated with OS were age ≥80 years, a positive BPI/EQ-5D composite, time since first prior oncological therapy <2 years and NLR > 2.5. Hb < 120 g/L and ALP > 130 U/L showed borderline associations, whereas albumin ≤ 40 g/L, LDH > 240 U/L and PSA ≥ 100 ng/mL were not independently associated with overall survival after adjustment. The proportional-hazards assumption was not violated globally for the final multivariate Cox model (global Schoenfeld test, χ2 = 10.90, df = 11, p = 0.453). Covariate-specific tests indicated possible non-proportionality for Hb < 120 g/L (p = 0.041) and PSA ≥ 100 ng/mL (p = 0.021). In a sensitivity analysis incorporating time-dependent effects for Hb and PSA, model fit was not significantly improved (likelihood-ratio χ2 = 5.84, df = 2, p = 0.054), and discrimination was essentially unchanged (Harrell’s C-index, 0.779 versus 0.779). Internal validation of the final multivariate Cox model using 1000 bootstrap resamples estimated a mean optimism of 0.020 in Harrell’s C-index. The apparent C-index of 0.779 was reduced to an optimism-corrected C-index of 0.760 (bootstrap 95% CI, 0.720–0.800). The optimism-corrected calibration slope was 0.841 (95% CI, 0.555–1.138), indicating modest evidence of overfitting. Bootstrap-corrected calibration plots at 12 and 24 months are shown in Figure S1. At 12 months, predicted and observed survival probabilities were closely aligned over most of the range, with some deviations at intermediate predicted probabilities; 194 patients remained at risk. At 24 months, greater deviations between predicted and observed survival were observed, and 37 patients remained at risk.

3.3. Development and Performance of the MIRA Score

The MIRA score included eight variables selected as described in the Methods. The components and point allocation are presented in Table 4. The resulting score had a theoretical range of 0–10 points and an observed range of 0–9 points. In the 265 patients evaluable for risk-group classification, Kaplan–Meier analysis showed clear separation of OS across the low-, intermediate- and high-risk MIRA groups (Figure 1; log-rank p < 0.001). Median OS was not reached in the low-risk group, was 22.97 months (95% CI, 22.21–23.23) in the intermediate-risk group and 11.56 months (95% CI, 10.09–14.16) in the high-risk group (Table 5). The proportional-hazards assumption was not significantly violated for the MIRA risk-group Cox model (global Schoenfeld test, χ2 = 5.27, df = 2, p = 0.072). The MIRA score was compared with the Halabi score using the three-tier risk-group classification as the comparator. Survival also differed significantly across Halabi risk groups (log-rank p < 0.001), but discrimination was lower than with MIRA. Harrell’s C-index was 0.755 (95% CI, 0.722–0.788) for the MIRA classification and 0.646 (95% CI, 0.603–0.690) for the Halabi classification (Table 5). The absolute C-index improvement with MIRA was 0.110, and paired comparison confirmed that this difference was statistically significant (z = 4.30, p < 0.001). In a sensitivity analysis excluding the six patients with post-baseline substitution for at least one MIRA component, results were materially unchanged. Harrell’s C-index was 0.752 (95% CI, 0.719–0.785), compared with 0.755 in the primary analysis, and both the intermediate- and high-risk groups remained significantly associated with shorter OS (HR, 5.51 and 20.36, respectively; both p < 0.001).

4. Discussion

In this exploratory secondary analysis of the placebo arm of the ENTHUSE-M1 trial [11], we developed the MIRA score for mCRPC, a simple three-tier prognostic classification that showed clear separation of OS. Patients classified as low, intermediate and high risk showed markedly different survival trajectories, with a strong gradient in mortality across groups. The score was based entirely on baseline variables that are routinely available in clinical-trial or real-world datasets: tumour burden, age, symptom burden, prior treatment interval, inflammatory status and standard laboratory parameters. Importantly, the strongest signal was observed for baseline tumour burden, supporting the concept that the extent of metastatic disease remains a major determinant of the natural history of mCRPC, even in a population that was pain-free or only mildly symptomatic at study entry. The comparison with the Halabi classification [1] provides a clinically relevant benchmark. In this dataset, the Halabi three-tier classification remained prognostic, confirming its value as an established reference model in mCRPC. Although its C-index was lower than that of MIRA (0.646 versus 0.755), the Halabi classification was retrospectively reconstructed using the variables available in ENTHUSE-M1, with opioid use recorded in only five patients. In addition, MIRA was developed and evaluated in the same cohort. The observed difference in C-index should therefore be regarded as exploratory and not as evidence of superiority. In contrast to weighted nomograms or algorithmically derived models, the MIRA score was constructed as a transparent point-based score. Its purpose is not to replace more detailed prognostic tools in all settings, but to test whether a simple score incorporating tumour burden can retain prognostic relevance. As a prognostic rather than predictive model, MIRA should not currently be used to select treatment or determine surveillance intensity; following independent validation, it could potentially support prognostic counselling and clinical-trial stratification. The strong prognostic signal associated with tumour burden observed in this study is central to the interpretation of the MIRA score. The composite tumour-burden variable integrated bone-metastasis category and RECIST target-lesion burden, thereby capturing both skeletal disease extent and measurable soft-tissue disease. It was the dominant predictor in univariate analysis and remained independently associated with OS after adjustment for clinical and laboratory factors. Tumour burden is a recognised prognostic determinant across solid tumours [10,14], reflecting both extent of disease and underlying tumour biology. In prostate cancer, this concept has been formalised in the hormone-sensitive setting by classifications such as CHAARTED volume [15] and LATITUDE risk [16], which use metastatic extent and distribution to define clinically distinct risk groups. The present findings suggest that the same principle remains relevant after transition to castration resistance, even in patients who are pain-free or only mildly symptomatic. The interval from first oncological therapy to castration resistance also contributed substantially to risk stratification. A short interval to castration resistance is an established adverse prognostic marker and reflects rapid disease progression and more aggressive tumour biology [8,17]. It provides information complementary to baseline tumour burden, capturing disease tempo rather than disease extent at enrolment. In MIRA, the allocation of two points to this variable reflected both its independent association with OS and its clinical interpretation as a marker of aggressive disease tempo, rather than formal scaling of the regression coefficient. The historical context of ENTHUSE-M1 is an important limitation but also helps define how the findings should be interpreted. The trial was conducted before the current therapeutic era in mCRPC; therefore, the cohort cannot be considered fully representative of contemporary practice. However, patients were pain-free or only mildly symptomatic at study entry, making the population closer to chemotherapy-naïve mCRPC cohorts such as COU-AA-302 [18] and PREVAIL [19] than to later-line, highly symptomatic populations. This restricted clinical spectrum may explain why WHO/ECOG performance status, opioid analgesic use and the derived Halabi score had limited discriminatory value in the present dataset. Their prognostic contribution is expected to be greater in cohorts including patients with more advanced disease. The inflammatory component of MIRA was based on NLR, which has the advantage of being simple, inexpensive and widely available, while reflecting the balance between systemic inflammation and host immune status [7,20]. Alternative indices such as platelet-to-lymphocyte ratio or systemic immune-inflammation index may also be prognostic, but their reliance on platelet count can complicate interpretation [21,22]. In prostate cancer, both thrombocytopenia and thrombocytosis may reflect adverse biology or clinical vulnerability, including marrow involvement, systemic inflammation, frailty or treatment-related risk [23,24]. For a simple point-based score, NLR may therefore be preferable to platelet-based composite indices. The association between age and survival also deserves consideration. Age ≥ 80 years may reflect not only reduced background life expectancy, but also lower physiological reserve, competing comorbidity, vulnerability to intercurrent illness and reduced eligibility for subsequent systemic therapy, including chemotherapy. This distinction is important because chronological age is mainly an indirect surrogate for fitness [25]. Future versions of MIRA could assess whether age should be complemented or replaced by validated measures of comorbidity and frailty, such as the Charlson Comorbidity Index or the G8 geriatric screening tool, which may capture clinical vulnerability more accurately than age alone [26,27]. Several limitations should be acknowledged. First, this was a secondary analysis of a historical clinical-trial dataset and the score was developed and evaluated within the same cohort. Although bootstrap internal validation of the final multivariate Cox model suggested only modest optimism, the procedure did not reproduce the complete process of variable selection, point allocation and risk-group definition. The apparent performance of the MIRA classification may therefore retain residual optimism. Calibration was generally acceptable at 12 months. The greater deviations observed at 24 months should be interpreted cautiously because only 37 patients remained at risk, resulting in reduced precision of the longer-term calibration assessment. MIRA should consequently be regarded as a hypothesis-generating prognostic score within this cohort, and independent external validation and potential recalibration are essential before clinical use. Second, only placebo-arm patient-level data were available, limiting sample size and precluding assessment of whether the score behaves similarly in patients receiving active treatment. Third, ENTHUSE-M1 enrolled a selected, relatively favourable population with bone-metastatic, pain-free or mildly symptomatic mCRPC and WHO/ECOG performance status 0–1; visceral metastases were present in only 17% of evaluable patients. Consequently, the generalisability of the findings to broader contemporary mCRPC populations is uncertain, particularly for patients with poorer performance status, clinically significant pain, opioid requirements, frailty or more extensive visceral disease. Fourth, the dataset predates contemporary imaging and treatment pathways; androgen receptor pathway inhibitor sequencing, poly(ADP-ribose) polymerase (PARP) inhibitor eligibility, radioligand therapy, PSMA PET staging, genomic testing and circulating tumour DNA assessment were not represented. These advances may improve survival outcomes and the prognostic relevance of individual variables; therefore, MIRA should not be assumed to apply directly to patients treated in current practice. Following independent validation, MIRA could potentially support prognostic counselling and clinical-trial stratification in mCRPC.

5. Conclusions

In this exploratory secondary analysis, metastatic tumour burden emerged as a major prognostic determinant in mCRPC, and its integration with routinely available clinical and laboratory variables provided a pragmatic three-tier MIRA classification with clear separation of OS. MIRA should be regarded as a hypothesis-generating prognostic score because it was developed and evaluated within the same historical, selected placebo-arm cohort. Its performance therefore needs to be confirmed in contemporary mCRPC populations before clinical use. Future studies should also assess whether the prognostic performance of MIRA can be improved by integrating quantitative disease-burden measures with genomic and molecular biomarkers, advanced imaging and machine-learning approaches, while preserving transparency, clinical usability and responsible implementation [28,29,30].

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/cancers18162660/s1. Figure S1: Bootstrap-corrected calibration of the final multivariate Cox model at (A) 12 months and (B) 24 months. Predicted survival probabilities are plotted against observed survival probabilities. The diagonal line represents perfect calibration. Calibration estimates were obtained using 1000 bootstrap resamples; Table S1: Frequency and handling of missing baseline data for candidate prognostic predictors with incomplete observations.

Author Contributions

Conceptualization, M.U.; data curation, M.U.; formal analysis, M.U.; investigation, M.U. and P.J.V.; methodology, M.U.; project administration, M.U.; resources, M.U.; supervision, M.U., S.B. and A.G.; validation, S.B. and A.G.; visualization, M.U.; writing—original draft preparation, M.U.; writing—review and editing, M.U. and P.J.V. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

This study was a secondary analysis of de-identified patient-level data obtained through Project Data Sphere. No participants were recruited or contacted, and the investigators had no access to directly identifiable information. Consistent with Section 4.2.8 of the UK Research Ethics Committee Policy Document, separate ethical review was not required. Data were analysed in accordance with Project Data Sphere terms of use and applicable local data-governance requirements.

Informed Consent Statement

Only anonymised, non-identifiable patient-level data made available through Project Data Sphere were analysed. No directly identifiable patient information was accessed. Data were handled in accordance with Project Data Sphere terms of use and local data governance requirements. Patient consent was waived.

Data Availability Statement

The individual patient-level data analysed in this study were obtained from the publicly accessible Project Data Sphere repository and are available through the platform subject to registration and applicable data-access requirements. The dataset corresponds to D4320C00014/ENTHUSE-M1 (NCT00554229).

Acknowledgments

This publication is based on research using information obtained from data.projectdatasphere.org, which is maintained by Project Data Sphere. Neither Project Data Sphere nor the owner(s) of any information from the website have contributed to, approved or are in any way responsible for the contents of this publication.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Halabi, S.; Lin, C.-Y.; Kelly, W.K.; Fizazi, K.S.; Moul, J.W.; Kaplan, E.B.; Morris, M.J.; Small, E.J. Updated prognostic model for predicting overall survival in first-line chemotherapy for patients with metastatic castration-resistant prostate cancer. J. Clin. Oncol. 2014, 32, 671–677, Erratum in J. Clin. Oncol. 2014, 32, 1387. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  2. Armstrong, A.J.; Lin, P.; Higano, C.S.; Sternberg, C.N.; Sonpavde, G.; Tombal, B.; Templeton, A.J.; Fizazi, K.; Phung, D.; Wong, E.K.; et al. Development and validation of a prognostic model for overall survival in chemotherapy-naïve men with metastatic castration-resistant prostate cancer. Ann. Oncol. 2018, 29, 2200–2207. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  3. Smaletz, O.; Scher, H.I.; Small, E.J.; Verbel, D.A.; McMillan, A.; Regan, K.; Kelly, W.K.; Kattan, M.W. Nomogram for overall survival of patients with progressive metastatic prostate cancer after castration. J. Clin. Oncol. 2002, 20, 3972–3982. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  4. Armstrong, A.J.; Garrett-Mayer, E.S.; Yang, Y.C.; de Wit, R.; Tannock, I.F.; Eisenberger, M. A Contemporary prognostic nomogram for men with hormone-refractory metastatic prostate cancer: A TAX327 study analysis. Clin. Cancer Res. 2007, 13, 6396–6403. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  5. Halabi, S.; Small, E.J.; Kantoff, P.W.; Kattan, M.W.; Kaplan, E.B.; Dawson, N.A.; Levine, E.G.; Blumenstein, B.A.; Vogelzang, N.J. Prognostic model for predicting survival in men with hormone-refractory metastatic prostate cancer. J. Clin. Oncol. 2003, 21, 1232–1237. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  6. Halabi, S.; Lin, C.Y.; Small, E.J.; Armstrong, A.J.; Kaplan, E.B.; Petrylak, D.; Sternberg, C.N.; Shen, L.; Oudard, S.; de Bono, J.; et al. Prognostic model predicting metastatic castration-resistant prostate cancer survival in men treated with second-line chemotherapy. J. Natl. Cancer Inst. 2013, 105, 1729–1737. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  7. Templeton, A.J.; Pezaro, C.; Omlin, A.; McNamara, M.G.; Leibowitz-Amit, R.; Vera-Badillo, F.E.; Attard, G.; de Bono, J.S.; Tannock, I.F.; Amir, E. Simple prognostic score for metastatic castration-resistant prostate cancer with incorporation of neutrophil-to-lymphocyte ratio. Cancer 2014, 120, 3346–3352. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  8. Chi, K.N.; Kheoh, T.; Ryan, C.J.; Molina, A.; Bellmunt, J.; Vogelzang, N.J.; Rathkopf, D.E.; Fizazi, K.; Kantoff, P.W.; Li, J.; et al. A prognostic index model for predicting overall survival in patients with metastatic castration-resistant prostate cancer treated with abiraterone acetate after docetaxel. Ann. Oncol. 2016, 27, 454–460. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  9. Guinney, J.; Wang, T.; Laajala, T.D.; Winner, K.K.; Bare, J.C.; Chaibub Neto, E.; Khan, S.A.; Peddinti, G.; Airola, A.; Pahikkala, T.; et al. Prediction of overall survival for patients with metastatic castration-resistant prostate cancer: Development of a prognostic model through a crowdsourced challenge with open clinical trial data. Lancet Oncol. 2017, 18, 132–142. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  10. Santorsola, M.; Di Lauro, V.; Nasti, G.; Caraglia, M.; Capuozzo, M.; Perri, F.; Cascella, M.; Misso, G.; Ottaiano, A. Tumour Burden Reporting in Phase III Clinical Trials of Metastatic Lung, Breast, and Colorectal Cancers: A Systematic Review. Cancers 2022, 14, 3262. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  11. Nelson, J.B.; Fizazi, K.; Miller, K.; Higano, C.; Moul, J.W.; Akaza, H.; Morris, T.; McIntosh, S.; Pemberton, K.; Gleave, M. Phase 3, randomized, placebo-controlled study of zibotentan (ZD4054) in patients with castration-resistant prostate cancer metastatic to bone. Cancer 2012, 118, 5709–5718. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  12. Green, A.K.; Reeder-Hayes, K.E.; Corty, R.W.; Basch, E.; Milowsky, M.I.; Dusetzina, S.B.; Bennett, A.V.; Wood, W.A. The project data sphere initiative: Accelerating cancer research by sharing data. Oncologist 2015, 20, 464-e20. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  13. Kang, L.; Chen, W.; Petrick, N.A.; Gallas, B.D. Comparing two correlated C indices with right-censored survival outcome: A one-shot nonparametric approach. Stat. Med. 2015, 34, 685–703. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  14. Uccello, M.; Gee, A.L.; Bennett, J.A.; Hazell, H.L.; Ruiz-Echarri Rueda, M.; Beresford, M.J. Metastatic Tumour Burden as Prognostic Factor in Nivolumab-Treated Metastatic Renal Cell Carcinoma: An Exploratory Study. Curr. Oncol. 2026, 33, 320. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  15. Sweeney, C.J.; Chen, Y.-H.; Carducci, M.; Liu, G.; Jarrard, D.F.; Eisenberger, M.; Wong, Y.-N.; Hahn, N.; Kohli, M.; Cooney, M.M.; et al. Chemohormonal Therapy in Metastatic Hormone-Sensitive Prostate Cancer. N. Engl. J. Med. 2015, 373, 737–746. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  16. Fizazi, K.; Tran, N.; Fein, L.; Matsubara, N.; Rodriguez-Antolin, A.; Alekseev, B.Y.; Özgüroğlu, M.; Ye, D.; Feyerabend, S.; Protheroe, A.; et al. Abiraterone plus Prednisone in Metastatic, Castration-Sensitive Prostate Cancer. N. Engl. J. Med. 2017, 377, 352–360. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  17. Bournakis, E.; Efstathiou, E.; Varkaris, A.; Wen, S.; Chrisofos, M.; Deliveliotis, C.; Alamanis, C.; Anastasiou, I.; Constantinides, C.; Bamias, A.; et al. Time to castration resistance is an independent predictor of castration-resistant prostate cancer survival. Anticancer Res. 2011, 31, 1475–1482. [Google Scholar] [PubMed]
  18. Ryan, C.J.; Smith, M.R.; de Bono, J.S.; Molina, A.; Logothetis, C.J.; de Souza, P.; Fizazi, K.; Mainwaring, P.; Piulats, J.M.; Ng, S.; et al. Abiraterone in metastatic prostate cancer without previous chemotherapy. N. Engl. J. Med. 2013, 368, 138–148, Erratum in N. Engl. J. Med. 2013, 368, 584. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  19. Beer, T.M.; Armstrong, A.J.; Rathkopf, D.E.; Loriot, Y.; Sternberg, C.N.; Higano, C.S.; Iversen, P.; Bhattacharya, S.; Carles, J.; Chowdhury, S.; et al. Enzalutamide in metastatic prostate cancer before chemotherapy. N. Engl. J. Med. 2014, 371, 424–433. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  20. Moschetta, M.; Uccello, M.; Kasenda, B.; Mak, G.; McClelland, A.; Boussios, S.; Forster, M.; Arkenau, H.T. Dynamics of Neutrophils-to-Lymphocyte Ratio Predict Outcomes of PD-1/PD-L1 Blockade. Biomed. Res. Int. 2017, 2017, 1506824. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  21. Lolli, C.; Caffo, O.; Scarpi, E.; Aieta, M.; Conteduca, V.; Maines, F.; Bianchi, E.; Massari, F.; Veccia, A.; Chiuri, V.E.; et al. Systemic Immune-Inflammation Index Predicts the Clinical Outcome in Patients with mCRPC Treated with Abiraterone. Front. Pharmacol. 2016, 7, 376. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  22. Li, D.Y.; Hao, X.Y.; Ma, T.M.; Dai, H.X.; Song, Y.S. The Prognostic Value of Platelet-to-Lymphocyte Ratio in Urological Cancers: A Meta-Analysis. Sci. Rep. 2017, 7, 15387. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  23. Nieder, C.; Haukland, E.; Pawinski, A.; Dalhaug, A. Anaemia and thrombocytopenia in patients with prostate cancer and bone metastases. BMC Cancer 2010, 10, 284. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  24. Giannakeas, V.; Kotsopoulos, J.; Brooks, J.D.; Cheung, M.C.; Rosella, L.; Lipscombe, L.; Akbari, M.R.; Austin, P.C.; Narod, S.A. Platelet Count and Survival after Cancer. Cancers 2022, 14, 549. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  25. Mohile, S.G.; Dale, W.; Somerfield, M.R.; Schonberg, M.A.; Boyd, C.M.; Burhenn, P.S.; Canin, B.; Cohen, H.J.; Holmes, H.M.; Hopkins, J.O.; et al. Practical Assessment and Management of Vulnerabilities in Older Patients Receiving Chemotherapy: ASCO Guideline for Geriatric Oncology. J. Clin. Oncol. 2018, 36, 2326–2347. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  26. Uccello, M.; Alam, T.; Abbas, H.; Nair, A.; Paskins, J.; Faust, G. Assessing Outcomes and Prognostic Factors for First-Line Therapy in Elderly Patients With Metastatic Renal Cell Carcinoma: Real-Life Data From a Single United Kingdom Institution. Clin. Genitourin. Cancer 2019, 17, e658–e663. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  27. Bellera, C.A.; Rainfray, M.; Mathoulin-Pélissier, S.; Mertens, C.; Delva, F.; Fonck, M.; Soubeyran, P.L. Screening older cancer patients: First evaluation of the G-8 geriatric screening tool. Ann. Oncol. 2012, 23, 2166–2172. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  28. Beltran, H.; Calais, J.; Emmett, L.; Kuo, P.H.; Logothetis, C.J. Biomarkers to optimize PSMA-targeted radioligand therapy for metastatic castration-resistant prostate cancer. Front. Oncol. 2025, 15, 1583168. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  29. Bang, S.; Ahn, Y.J.; Koo, K.C. Harnessing machine learning to predict prostate cancer survival: A review. Front. Oncol. 2025, 14, 1502629. [Google Scholar] [CrossRef] [Scilit] [PubMed]
  30. Malerba, S.; Vladimirov, M.; Goyal, A.; Dulskas, A.; Baušys, A.; Cwalinski, T.; Girnyi, S.; Skokowski, J.; Duka, R.; Molchanov, R.; et al. Artificial Intelligence Applications in Gastric Cancer Surgery: Bridging Early Diagnosis and Responsible Precision Medicine. J. Clin. Med. 2026, 15, 2208. [Google Scholar] [CrossRef] [Scilit] [PubMed]
Cancers 18 02660 g001
Figure 1. Overall survival (OS) according to the Metastatic Integrated Risk Assessment (MIRA) risk group. The Kaplan–Meier curves show OS according to MIRA risk classification. Median OS was not reached in the low-risk group, 22.97 months in the intermediate-risk group and 11.56 months in the high-risk group. Tick marks indicate censored observations, and the number of patients at risk is shown below the plot. OS distributions differed significantly across the three groups by the log-rank test (p < 0.001).
Figure 1. Overall survival (OS) according to the Metastatic Integrated Risk Assessment (MIRA) risk group. The Kaplan–Meier curves show OS according to MIRA risk classification. Median OS was not reached in the low-risk group, 22.97 months in the intermediate-risk group and 11.56 months in the high-risk group. Tick marks indicate censored observations, and the number of patients at risk is shown below the plot. OS distributions differed significantly across the three groups by the log-rank test (p < 0.001).
Cancers 18 02660 g001
Table 1. Baseline characteristics and candidate prognostic variables.
Table 1. Baseline characteristics and candidate prognostic variables.
VariableCategoryn/N (%)VariableCategoryn/N (%)
Demographic and clinical characteristicsNLR≤2.5110/266 (41.4)
RaceWhite160/266 (60.2) >2.5156/266 (58.6)
Asian99/266 (37.2)Albumin>40 g/L202/266 (75.9)
Other/unknown7/266 (2.6)≤40 g/L64/266 (24.1)
Age45–59 years29/266 (10.9)ALP≤130 U/L135/266 (50.8)
60–64 years29/266 (10.9)>130 U/L131/266 (49.2)
65–69 years54/266 (20.3)LDH≤240 U/L226/266 (85.0)
70–74 years57/266 (21.4)>240 U/L40/266 (15.0)
75–79 years54/266 (20.3)PSA<100 ng/mL176/266 (66.2)
≥80 years43/266 (16.2)≥100 ng/mL90/266 (33.8)
BMI≥25 kg/m2162/266 (60.9)Disease-burden variables
<25 kg/m2104/266 (39.1)Bone metastases<577/266 (28.9)
WHO/ECOG PS0196/266 (73.7)5–20136/266 (51.1)
170/266 (26.3)≥2153/266 (19.9)
Opioid use at baselineNo261/266 (98.1)Visceral metastasesAbsent220/265 (83.0)
Yes5/266 (1.9)Present45/265 (17.0)
BPI worst pain0126/258 (48.8)Nodal metastasesAbsent185/265 (69.8)
≥1132/258 (51.2)Present80/265 (30.2)
EQ-5D pain/discomfortLevel 1155/258 (60.1)RECIST target lesionAbsent171/265 (64.5)
Level 2103/258 (39.9)Present94/265 (35.5)
BPI/EQ-5D compositeNegative113/266 (42.5)RECIST SLD≤3 cm207/265 (78.1)
Positive153/266 (57.5)>3 cm58/265 (21.9)
Time since first prior oncological therapy≥2 years191/265 (72.1)Composite tumour-burden levelLevel 052/265 (19.6)
<2 years74/265 (27.9) Level 1120/265 (45.3)
Laboratory variables Level 293/265 (35.1)
Hb≥120 g/L208/266 (78.2)
<120 g/L58/266 (21.8)
PLT150–450 × 109/L236/264 (89.4)
<150 or >450 × 109/L28/264 (10.6)
Note: Percentages were calculated using evaluable patients for each variable. Abbreviations: ALP, alkaline phosphatase; BMI, body mass index; BPI, Brief Pain Inventory; ECOG, Eastern Cooperative Oncology Group; EQ-5D, EuroQol 5-Dimension questionnaire; Hb, haemoglobin; LDH, lactate dehydrogenase; NLR, neutrophil-to-lymphocyte ratio; PLT, platelet count; PSA, prostate-specific antigen; RECIST, Response Evaluation Criteria in Solid Tumours; SLD, sum of longest diameters; WHO, World Health Organization.
Table 2. Univariate Cox regression analysis of candidate prognostic variables for survival.
Table 2. Univariate Cox regression analysis of candidate prognostic variables for survival.
VariableCategoryEvents/TotalHR (95% CI)p Value
Demographic and clinical variables
RaceAsian51/991.17 (0.82–1.67)0.38
White78/160
Age≥80 years30/432.11 (1.40–3.18)<0.001
<80 years103/223
BMI<25 kg/m259/1041.40 (0.99–1.97)0.06
≥25 kg/m274/162
WHO/ECOG PS138/701.34 (0.92–1.95)0.13
095/196
BPI/EQ-5D compositePositive87/1531.74 (1.21–2.48)0.003
Negative46/113
Time since first prior oncological therapy<2 years53/742.45 (1.73–3.47)<0.001
≥2 years79/191
Laboratory variables
Hb<120 g/L42/582.42 (1.67–3.49)<0.001
≥120 g/L91/208
PLT<150 or >450 × 109/L15/281.50 (0.87–2.58)0.14
150–450 × 109/L116/236
NLR>2.591/1561.79 (1.24–2.58)0.002
≤2.542/110
Albumin≤40 g/L41/641.71 (1.18–2.47)0.005
>40 g/L92/202
ALP>130 U/L84/1312.03 (1.42–2.89)<0.001
≤130 U/L49/135
LDH>240 U/L26/401.72 (1.12–2.64)0.014
≤240 U/L107/226
PSA≥100 ng/mL63/902.53 (1.80–3.56)<0.001
<100 ng/mL70/176
Disease-burden variables
Bone metastases≥2142/535.03 (2.93–8.63)<0.001
5–2071/1362.52 (1.52–4.16)<0.001
<520/77
Visceral metastasesPresent27/451.49 (0.97–2.27)0.07
Absent105/220
Nodal metastasesPresent50/801.78 (1.25–2.53)0.001
Absent82/185
RECIST target lesionPresent60/942.24 (1.58–3.17)<0.001
Absent72/171
RECIST SLD>3 cm42/582.69 (1.86–3.90)<0.001
≤3 cm90/207
Composite tumour-burden levelLevel 269/9310.38 (4.88–22.09)<0.001
Level 155/1204.35 (2.04–9.27)<0.001
Level 08/52
HRs were estimated using univariate Cox proportional-hazards models. For bone-metastasis category and composite tumour-burden level, the reference categories were <5 bone metastases and level 0, respectively. Race was analysed as Asian versus White; patients recorded as other/unknown race were excluded from this comparison. Abbreviations: ALP, alkaline phosphatase; BMI, body mass index; BPI, Brief Pain Inventory; CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; EQ-5D, EuroQol 5-Dimension questionnaire; Hb, haemoglobin; HR, hazard ratio; LDH, lactate dehydrogenase; NLR, neutrophil-to-lymphocyte ratio; PLT, platelet count; PSA, prostate-specific antigen; PS, performance status; RECIST, Response Evaluation Criteria in Solid Tumours; SLD, sum of longest diameters; WHO, World Health Organization.
Table 3. Multivariate Cox regression analysis for overall survival.
Table 3. Multivariate Cox regression analysis for overall survival.
VariableHR (95% CI)p Value
Age ≥ 80 years1.77 (1.15–2.74)0.010
Albumin ≤ 40 g/L1.11 (0.74–1.69)0.609
ALP > 130 U/L1.43 (0.97–2.11)0.070
BPI/EQ-5D composite positive1.54 (1.06–2.25)0.025
Composite tumour-burden level 12.62 (1.20–5.69)0.015
Composite tumour-burden level 25.52 (2.52–12.11)<0.001
Hb < 120 g/L1.52 (1.00–2.32)0.052
LDH > 240 U/L1.39 (0.88–2.20)0.155
NLR > 2.51.50 (1.03–2.20)0.036
PSA ≥ 100 ng/mL1.34 (0.92–1.94)0.128
Time since first prior oncological therapy <2 years1.91 (1.33–2.75)<0.001
HRs were estimated using a multivariate Cox proportional-hazards model. Reference categories were the complementary lower-risk categories for each binary variable; level 0 was the reference category for composite tumour-burden level. The model included 264 patients with complete covariate data and 131 deaths. Overall model fit: chi-square = 115.806, df = 11, p < 0.001. Harrell’s C-index was 0.779 (95% CI, 0.741–0.818). Abbreviations: ALP, alkaline phosphatase; BPI, Brief Pain Inventory; CI, confidence interval; EQ-5D, EuroQol 5-Dimension questionnaire; Hb, haemoglobin; HR, hazard ratio; LDH, lactate dehydrogenase; NLR, neutrophil-to-lymphocyte ratio; PSA, prostate-specific antigen.
Table 4. Prognostic components and risk-group classification of the MIRA score.
Table 4. Prognostic components and risk-group classification of the MIRA score.
Score ComponentCriterionPoints
Composite tumour-burden levelLevel 00
Level 11
Level 22
Age≥80 years1
BPI/EQ-5D compositePositive1
Time since first prior oncological therapy<2 years2
NLR>2.51
ALP>130 U/L1
Hb<120 g/L1
PSA≥100 ng/mL1
Composite tumour-burden level was assigned hierarchically. Level 2 was assigned to patients with ≥21 bone metastases or RECIST SLD >3 cm. Among the remaining patients, level 1 was assigned to those with 5–20 bone metastases or any RECIST target lesion. Level 0 was assigned to those with <5 bone metastases and no RECIST target lesion. Low risk was defined as 0–2 points, intermediate risk as 3–4 points and high risk as ≥5 points. Abbreviations: ALP, alkaline phosphatase; BPI, Brief Pain Inventory; CI, confidence interval; EQ-5D, EuroQol 5-Dimension questionnaire; Hb, haemoglobin; MIRA, Metastatic Integrated Risk Assessment; NLR, neutrophil-to-lymphocyte ratio; OS, overall survival; PSA, prostate-specific antigen; RECIST, Response Evaluation Criteria in Solid Tumours; SLD, sum of longest diameters.
Table 5. Comparison of MIRA and Halabi three-tier risk classifications.
Table 5. Comparison of MIRA and Halabi three-tier risk classifications.
Risk ClassificationRisk Group and Cut-Off, PointsEvents/Total (%)Median OS (95% CI), MonthsHarrell’s C-Index (95% CI)
MIRA scoreLow risk, 0–26/69 (8.7)Not reached0.755 (0.722–0.788)
Intermediate risk, 3–436/87 (41.4)22.97 (22.21–23.23)
High risk, 5–990/109 (82.6)11.56 (10.09–14.16)
Halabi scoreLow risk, <14052/143 (36.4)27.14 (22.57–not reached)0.646 (0.603–0.690)
Intermediate risk, 140–194.9659/96 (61.5)18.17 (14.92–22.54)
High risk, >194.9622/27 (81.5)9.72 (6.60–14.16)
Abbreviations: CI, confidence interval; MIRA, Metastatic Integrated Risk Assessment for mCRPC; OS, overall survival. Survival distributions differed significantly across risk groups for both classifications by the log-rank test (p < 0.001). Harrell’s C-index values apply to the corresponding three-tier risk classification. Halabi risk groups were assigned using the published nomogram point cut-offs [1].
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Uccello, M.; Vlachostergios, P.J.; Ghose, A.; Boussios, S. Incorporating Metastatic Disease Burden into Prognostic Assessment in Metastatic Castration-Resistant Prostate Cancer: The MIRA Score. Cancers 2026, 18, 2660. https://doi.org/10.3390/cancers18162660

AMA Style

Uccello M, Vlachostergios PJ, Ghose A, Boussios S. Incorporating Metastatic Disease Burden into Prognostic Assessment in Metastatic Castration-Resistant Prostate Cancer: The MIRA Score. Cancers. 2026; 18(16):2660. https://doi.org/10.3390/cancers18162660

Chicago/Turabian Style

Uccello, Mario, Panagiotis J. Vlachostergios, Aruni Ghose, and Stergios Boussios. 2026. "Incorporating Metastatic Disease Burden into Prognostic Assessment in Metastatic Castration-Resistant Prostate Cancer: The MIRA Score" Cancers 18, no. 16: 2660. https://doi.org/10.3390/cancers18162660

APA Style

Uccello, M., Vlachostergios, P. J., Ghose, A., & Boussios, S. (2026). Incorporating Metastatic Disease Burden into Prognostic Assessment in Metastatic Castration-Resistant Prostate Cancer: The MIRA Score. Cancers, 18(16), 2660. https://doi.org/10.3390/cancers18162660

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop