A Novel Clinical Prediction Model for Prognosis in Malignant Pleural Mesothelioma Using Decision Tree Analysis

Our plain-English summary of A Novel Clinical Prediction Model for Prognosis in Malignant Pleural Mesothelioma Using Decision Tree Analysis, by F.J. Brims, T.M. Meniawy, I. Duffus, et al (Journal of Thoracic Oncology, 2016).

The study

• Authors: F.J. Brims, T.M. Meniawy, I. Duffus, D. de Fonseka, A. Segal, J. Creaney, N. Maskell, R.A. Lake, N. de Klerk, A.K. Nowak
• Journal: Journal of Thoracic Oncology
• Published: 2016;11(4):573-82
• Cited by: 105 articles, as counted by NIH iCite
• Read it: PubMed · Publisher

Abstract

Introduction: Malignant pleural mesothelioma (MPM) is a rare cancer with a heterogeneous prognosis. Prognostic models are not widely utilized clinically. Classification and regression tree (CART) analysis examines the interaction of multiple variables with a given outcome.

Methods: Between 2005 and 2014, all cases with pathologically confirmed MPM had routinely available histological, clinical, and laboratory characteristics recorded. Classification and regression tree analysis was performed using 29 variables with 18-month survival as the dependent variable. Risk groups were refined according to survival and clinical characteristics. The model was then tested on an external international cohort.

Results: A total of 482 cases were included in the derivation cohort; the median survival was 12.6 months, and the median age was 69 years. The model defined four risk groups with clear survival differences (p < 0.0001). The strongest predictive variable was the presence of weight loss. The group with the best survival at 18 months (86.7% alive, median survival 34.0 months, termed risk group 1) had no weight loss, a hemoglobin level greater than 153 g/L, and a serum albumin level greater than 43 g/L. The group with the worst survival (0% alive, median survival 7.5 months, termed risk group 4d) had weight loss, a performance score of 0 or 1, and sarcomatoid histological characteristics. The C-statistic for the model was 0.761, and the sensitivity was 94.5%. Validation on 174 external cases confirmed the model's ability to discriminate between risk groups in an alternative data set with fair performance (C-statistic 0.68).

Conclusions: We have developed and validated a simple, clinically relevant model to reliably discriminate patients at high and lower risk of death using routinely available variables from the time of diagnosis in unselected populations of patients with MPM.

Copyright © 2016 International Association for the Study of Lung Cancer. Published by Elsevier Inc. All rights reserved.

In plain English

This study developed a survival prediction model for newly diagnosed malignant pleural mesothelioma using information routinely available at diagnosis. The authors analysed 482 confirmed cases from the West Australian State Cancer Centre, diagnosed 2005 to 2014, collected retrospectively and prospectively. Decision tree analysis of 29 clinical and laboratory variables, with 18-month survival as the outcome, sorted patients into risk groups, later refined to four. Weight loss at diagnosis was the strongest predictor. In risk group 1 (no weight loss, haemoglobin above 153 g/L, albumin above 43 g/L or not recorded), 86.7% were alive at 18 months; median survival was 34.0 months. In risk group 4d (weight loss or not recorded, performance score 0 to 1 or not recorded, sarcomatoid histology), none were alive at 18 months; median survival was 7.5 months. The model's C-statistic was 0.761. On 174 UK patients it scored 0.68 (95% confidence interval 0.60 to 0.75). Asbestos exposure was not studied. The authors concluded that the four risk groups can reliably discriminate survival. They noted missing data and a single-centre derivation cohort as limitations.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Prognostic model development study (decision tree / classification and regression tree analysis) using combined retrospective and prospective hospital data, with internal cross-validation and testing on an external cohort
• Who was studied: Derivation cohort: 482 confirmed malignant pleural mesothelioma cases (from 595 mesothelioma cases identified) at a single institution, the West Australian State Cancer Centre, Perth. Of these, 274 were collected retrospectively (2005 to 2010) and 208 prospectively (1 January 2011 to 30 June 2014). The median age was 69 years and 86.9% were male. Validation cohort: 174 of 177 consecutive patients at North Bristol NHS Trust, UK, diagnosed 1 March 2008 to 30 April 2015.
• Exposure: Asbestos exposure was not measured or analysed. The factors studied were 29 routinely available clinical and laboratory variables recorded at diagnosis: symptoms such as weight loss, performance status, histology, blood tests and pleural fluid tests.
• Main results:
  • In the derivation cohort, overall median survival was 12.6 months; 418 of 482 cases (86.7%) had died by the censoring date.
  • Weight loss at diagnosis was the variable most strongly associated with the outcome. Among patients without weight loss, 107/237 (45%) were alive at 18 months. Among those with weight loss, 50/245 (20.4%) were alive (p<0.0001).
  • Risk group 1 (30 patients; no weight loss, haemoglobin >153 g/L, and serum albumin >43 g/L, or albumin missing according to Table 3) had the best predicted survival, with 86.7% alive at 18 months.
  • Risk group 4d had the worst predicted survival, with 0% alive at 18 months. Its members had weight loss or missing weight-loss information, ECOG performance status 0-1 or missing performance status according to Table 3, and sarcomatoid histology.
  • Median survival was 34.0 months (IQR 22.9-47.0) in risk group 1 and 7.5 months (IQR 3.3-10.9) in risk group 4d (p<0.001).
  • In the derivation cohort, the model had a C-statistic of 0.761, sensitivity of 94.5% (95% CI 91.4 to 96.7%) and specificity of 38.2% (95% CI 30.6 to 46.3%).
  • Across the four refined risk groups, median survival ranged from 34.0 months (IQR 22.9-47.0) in Group 1 to 7.4 months (IQR 3.3-11.1) in Group 4 (p<0.0001).
  • By histology alone in the derivation cohort, median survival was 14.8 months for epithelioid, 13.9 for not classified, 9.1 for biphasic and 7.5 for sarcomatoid tumours (p<0.0001).
  • In the UK validation cohort of 174 patients, median survival was shorter: 9.7 months versus 12.6 months in the derivation cohort.
  • In the UK validation cohort, the model's C-statistic was 0.68 (95% CI 0.60 to 0.75). Risk groups 1 and 2 were combined because UK risk group 1 had only two cases.
• The authors' conclusion: The authors concluded that their four risk groups, built from routinely available variables in a large unselected population, can reliably discriminate survival in a simple way. They considered this more useful than single variables and expected it to be more generalizable than earlier groupings from highly selected populations.
• Limitations the authors state:
  • Data were collected through both retrospective and prospective approaches, though the authors described the differences as statistical rather than clinical.
  • Some cases had missing biochemistry and/or clinical variables.
  • The derivation data came from a single centre.
  • The model performed less well on the external dataset, possibly because of smaller numbers and missing weight-loss data.
  • The model has not been validated in a surgical population.

Topics

The National Library of Medicine indexes this study under subjects our site covers: Lung cancer, Mesothelioma.

Medical terms in this study, explained in our Lexicon: Prognosis, Pleural Mesothelioma, Median Survival, Sarcomatoid.

Related studies

• Predictions of mortality from pleural mesothelioma in Italy: a model based on asbestos consumption figures supports results from age-period-cohort models — A. Marinaccio, F. Montanaro, et al. (2005)
• Prediction modelling using routine clinical parameters to stratify survival in Malignant Pleural Mesothelioma patients undergoing cytoreductive surgery — E.J.A. Harris, S. Kao, et al. (2019)
• The use of a next-generation sequencing-derived machine-learning risk-prediction model (OncoCast-MPM) for malignant pleural mesothelioma: a retrospective study — M.G. Zauderer, A. Martin, et al. (2021)
• Development and validation of a lung cancer risk prediction model for African-Americans — C.J. Etzel, S. Kachroo, et al. (2008)
• Comparison of modified Response Evaluation Criteria in Solid Tumors, European Organization for Research and Treatment of Cancer criteria, and PET Response Criteria in Solid Tumors for evaluation of tumor response to chemotherapy and prognosis prediction in patients with unresectable malignant pleural mesothelioma — K. Kitajima, M. Maruyama, et al. (2020)
• Existing models, but not neutrophil-to-lymphocyte ratio, are prognostic in malignant mesothelioma — T.M. Meniawy, J. Creaney, et al. (2013)

Browse more research in Journal of Thoracic Oncology, or return to medical research.