The use of a next-generation sequencing-derived machine-learning risk-prediction model (OncoCast-MPM) for malignant pleural mesothelioma: a retrospective study

Our plain-English summary of The use of a next-generation sequencing-derived machine-learning risk-prediction model (OncoCast-MPM) for malignant pleural mesothelioma: a retrospective study, by M.G. Zauderer, A. Martin, J. Egger, et al (The Lancet. Digital Health, 2021).

The study

• Authors: M.G. Zauderer, A. Martin, J. Egger, H. Rizvi, M. Offin, A. Rimner, P.S. Adusumilli, V.W. Rusch, M.G. Kris, J.L. Sauter, M. Ladanyi, R. Shen
• Journal: The Lancet. Digital Health
• Published: 2021;3(9):e565-e576
• Cited by: 36 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Background: Current risk stratification for patients with malignant pleural mesothelioma based on disease stage and histology is inadequate. For some individuals with early-stage epithelioid tumours, a good prognosis by current guidelines can progress rapidly; for others with advanced sarcomatoid cancers, a poor prognosis can progress slowly. Therefore, we aimed to develop and validate a machine-learning tool-known as OncoCast-MPM-that could create a model for patient prognosis.

Methods: We did a retrospective study looking at malignant pleural mesothelioma tumours using next-generation sequencing from the Memorial Sloan Kettering Cancer Center-Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT). We collected clinical, pathological, and routine next-generation sequencing data from consecutive patients with malignant pleural mesothelioma treated at the Memorial Sloan Kettering Cancer Center (New York, NY, USA), as well as the MSK-IMPACT data. Together, these data comprised the MSK-IMPACT cohort. Using OncoCast-MPM, an open-source, web-accessible, machine-learning risk-prediction model, we integrated available data to create risk scores that stratified patients into low-risk and high-risk groups. Risk stratification of the MSK-IMPACT cohort was then validated using publicly available malignant pleural mesothelioma data from The Cancer Genome Atlas (ie, the TCGA cohort).

Findings: Between Feb 15, 2014, and Jan 28, 2019, we collected MSK-IMPACT data from the tumour tissue of 194 patients in the MSK-IMPACT cohort. The median overall survival was higher in the low-risk group than in the high-risk group as determined by OncoCast-MPM (30·8 months [95% CI 22·7-36·2] vs 13·9 months [10·7-18·0]; hazard ratio [HR] 3·0 [95% CI 2·0-4·5]; p<0·0001). No single factor or gene alteration drove risk differentiation. OncoCast-MPM was validated against the TCGA cohort, which consisted of 74 patients. The median overall survival was higher in the low-risk group than in the high-risk group (23·6 months [95% CI 15·1-28·4] vs 13·6 months [9·8-17·9]; HR 2·3 [95% CI 1·3-3·8]; p=0·0019). Although stage-based risk stratification was unable to differentiate survival among risk groups at 3 years in the MSK-IMPACT cohort (31% for early-stage disease vs 30% for advanced-stage disease; p=0·90), the OncoCast-MPM-derived 3-year survival was significantly higher in the low-risk group than in the high-risk group (40% vs 7%; p=0·0052).

Interpretation: OncoCast-MPM generated accurate, individual patient-level risk assessment scores. After prospective validation with the TCGA cohort, OncoCast-MPM might offer new opportunities for enhanced risk stratification of patients with malignant pleural mesothelioma in clinical trials and drug development.

Funding: US National Institutes of Health/National Cancer Institute.

Copyright © 2021 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY-NC-ND 4.0 license. Published by Elsevier Ltd.. All rights reserved.

In plain English

This retrospective study built and tested OncoCast-MPM, a computer risk model for malignant pleural mesothelioma. It combined tumour gene-sequencing results with age, sex, smoking, stage, histology and self-reported classic occupational asbestos exposure. It was developed on 194 patients treated at Memorial Sloan Kettering Cancer Center, New York, between February 2014 and January 2019. It was then checked on 74 patients from The Cancer Genome Atlas. Among the New York patients, median overall survival was 30.8 months (95% CI 22.7 to 36.2) for patients rated low risk versus 13.9 months (10.7 to 18.0) for high-risk patients (hazard ratio 3.0, 95% CI 2.0 to 4.5). The check group showed 23.6 months (15.1 to 28.4) versus 13.6 months (9.8 to 17.9), hazard ratio 2.3 (95% CI 1.3 to 3.8). Grouping the New York patients by early versus advanced stage gave no statistically significant survival difference (hazard ratio 1.4, 95% CI 1.0 to 2.1, p=0.064). The model counted self-reported asbestos exposure as favourable. Fiber types and doses were not reported. The authors concluded the model gave accurate individual risk scores but noted it has not been prospectively validated.

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: Retrospective study developing a machine-learning survival risk-prediction model (OncoCast-MPM) in one hospital cohort and validating it in a separate public dataset (TCGA)
• Who was studied: Development cohort: 194 consecutive patients with malignant pleural mesothelioma treated at Memorial Sloan Kettering Cancer Center, New York, USA, between Feb 15, 2014, and Jan 28, 2019, with tumour sequencing (MSK-IMPACT) data. Median age 70; 148 (76%) of 194 male; 163 (84%) epithelioid histology. Validation cohort: 74 patients with malignant pleural mesothelioma from The Cancer Genome Atlas (TCGA).
• Exposure: Not an exposure study. The model used tumour gene alterations (341 genes common to all sequencing panels) plus age, histology, sex, smoking status, stage and self-reported classic occupational asbestos exposure. Asbestos exposure was reported as yes for 105 (54%) of 194 MSK-IMPACT patients and 46 (62%) of 74 TCGA patients. The article does not report fiber types or exposure doses.
• Main results:
  • MSK-IMPACT cohort: 52 (27%) of 194 patients were classed high risk and 142 (73%) low risk.
  • MSK-IMPACT cohort: median overall survival was 30.8 months (95% CI 22.7 to 36.2) in the low-risk group vs 13.9 months (10.7 to 18.0) in the high-risk group; HR 3.0 (95% CI 2.0 to 4.5); p less than 0.0001.
  • TCGA validation cohort (74 patients): median overall survival was 23.6 months (95% CI 15.1 to 28.4) in the low-risk group vs 13.6 months (9.8 to 17.9) in the high-risk group; HR 2.3 (95% CI 1.3 to 3.8); p=0.0019.
  • MSK-IMPACT cohort grouped by stage (early IA to IIIA vs advanced IIIB to IV): median overall survival was 27.0 months for early stage vs 18.3 months for advanced stage. This difference was not statistically significant (HR 1.4, 95% CI 1.0 to 2.1; p=0.064).
  • MSK-IMPACT cohort, 3-year survival: 40% in the model's low-risk group vs 7% in its high-risk group (p=0.0052). By stage it was 31% for early-stage vs 30% for advanced-stage disease (p=0.90).
  • Cross-validated concordance probability estimate of the model (how well the score ranked survival) in the MSK-IMPACT cohort: 0.67 (IQR 0.64 to 0.69).
  • MSK-IMPACT cohort: 27 (21%) of 129 early-stage patients were rated high risk by the model, and 40 (62%) of 65 advanced-stage patients were rated low risk. In total, 67 (35%) of 194 were described as potentially mischaracterised by stage.
  • In the model's multivariate feature-importance analysis, reported classic occupational asbestos exposure and a history of smoking were listed among highly favourable features. Male sex, CDKN2A and CDKN2B deletion, TP53 and TERT mutations, age, advanced stage and biphasic histology were listed among highly unfavourable features.
  • MSK-IMPACT early-stage patients: median overall survival was 33.2 months for the model's low-risk group vs 15.0 months for its high-risk group (HR 3.9, 95% CI 2.3 to 6.9; p less than 0.0001).
  • MSK-IMPACT advanced-stage patients: median overall survival was 21.1 months for the model's low-risk group vs 13.5 months for its high-risk group (HR 2.1, 95% CI 1.1 to 4.0; p=0.022).
• The authors' conclusion: The authors concluded that OncoCast-MPM generated accurate individual patient-level risk scores. They said that, after prospective validation, it might improve risk stratification of patients with malignant pleural mesothelioma in clinical trials and drug development.
• Limitations the authors state:
  • The model has not been prospectively validated, so the retrospective analysis could be confounded by variables outside the model.
  • Gene-testing panels vary, and the 341 genes used may not be available for all patients.
  • Next-generation sequencing may not be available everywhere.
  • The model does not take treatment regimens into account.
  • The patients used to build and validate the model may not reflect most patients with mesothelioma worldwide.

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: Next-Generation Sequencing (NGS), Pleural Mesothelioma, Histology, Epithelioid, Prognosis, Sarcomatoid, Memorial Sloan Kettering, Median Overall Survival.

Related studies

• Targeted next-generation sequencing of cancer genes in advanced stage malignant pleural mesothelioma: a retrospective study — M. Lo Iacono, V. Monica, et al. (2015)
• Prediction of Radiation Pneumonitis With Machine Learning in Stage III Lung Cancer: A Pilot Study — M. Yakar, D. Etiz, et al. (2021)
• Next-generation sequencing informs diagnosis and identifies unexpected therapeutic targets in lung squamous cell carcinomas — J.M. Sands, T. Nguyen, et al. (2020)
• Next Generation Sequencing for miRNA Detection on the Exhaled Breath Condensate: A Pilot Study — R. Cherchi, R. Cusano, et al. (2023)
• Clinical Next Generation Sequencing Application in Mesothelioma: Finding a Golden Needle in the Haystack — L. Cerbone, S. Orecchia, et al. (2023)
• Computational prediction of diagnosis and feature selection on mesothelioma patient health records — D. Chicco, C. Rovelli (2019)

Browse return to medical research.