Cancer Risk Prediction Using Machine Learning for Supporting Early Cancer Diagnosis in Symptomatic Patients: A Systematic Review of Model Types

Our plain-English summary of Cancer Risk Prediction Using Machine Learning for Supporting Early Cancer Diagnosis in Symptomatic Patients: A Systematic Review of Model Types, by F. Pennisi, S. Borlini, H. Harrison, et al (Cancer Medicine, 2025).

The study

• Authors: F. Pennisi, S. Borlini, H. Harrison, R. Cuciniello, A.C. D'Amelio, M. Barclay, G.E. Ricciardi, G. Lyratzopoulos, C. Renzi
• Journal: Cancer Medicine
• Published: 2025;14(24):e71463
• Type of article: Systematic review
• Cited by: 15 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Introduction: Predictive models could support clinicians in identifying patients who may benefit from cancer investigations. We aimed to examine published evidence on machine learning models (ML) developed to estimate cancer risk based on symptoms and other patient characteristics.

Methods: Using MEDLINE, Scopus, and EMBASE, we performed a systematic review of studies published in 2014-2024, which included data on signs/symptoms for cancer risk prediction. We used the QUADAS-AI tools to assess study quality. We performed a quantitative synthesis of diagnostic performance, including accuracy, sensitivity, specificity, area under the curve (AUC). Adherence to TRIPOD guidelines was assessed.

Results: Among the 5646 initially identified articles, 34 met inclusion criteria. Included studies most frequently examined lung (n = 9 studies), mesothelioma (n = 7), and gastrointestinal cancers (n = 4) and used hospital electronic health records (n = 8) or publicly available online datasets (n = 13). In addition to signs/symptoms (n = 34), most models included sociodemographic characteristics (n = 27) and lifestyle factors (n = 20). In 70% of studies, internal validation was performed. ML models demonstrated variable performance, with AUC values ranging from 0.60 to 1 during validation. Random Forest, Support Vector Machine, Decision Tree, and Multilayer Perceptron showed the best predictive performance. Most of the studies (94.1%) had a high risk of bias for the index test.

Conclusion: ML models have been reported to demonstrate potential in managing complex data for cancer risk prediction. However, the current evidence is heterogeneous and frequently limited by bias and incomplete reporting. Further validation and thorough assessments of real-world performance are necessary before these models can be considered reliable for clinical use.

Trial Registration: International Prospective Register of Systematic Reviews (PROSPERO) registration number: CRD42024548088.

© 2025 The Author(s). Cancer Medicine published by John Wiley & Sons Ltd.

In plain English

This was a systematic review, not a new study of patients. The authors searched for studies published from 2014 to May 2024 that built machine learning (computer) models to estimate cancer risk in people who had symptoms. Of 5646 records found, 34 studies were included. The most common cancers were lung (9 studies), mesothelioma (7) and gastrointestinal cancer (4). The review reported that the seven mesothelioma studies considered asbestos, C-reactive protein and pleural protein exposure as a key risk factor. It does not name asbestos fiber types. Model performance varied, with AUC values from 0.60 to 1 during validation. AUC measures how well a model separates people with and without cancer, and 1 is perfect. Only two studies tested their models on outside data, and only two checked calibration. The authors rated 94.1% of studies at high risk of bias for the index test. The authors concluded that the evidence was heterogeneous and often limited by bias and incomplete reporting, and that further validation is needed before these models can be considered reliable for clinical use. They noted that only English-language, peer-reviewed studies were included.

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: Systematic review of published studies that developed machine learning models for cancer risk prediction including signs/symptoms (PROSPERO CRD42024548088)
• Who was studied: 34 published studies (from 5646 records identified) of machine learning models predicting cancer in symptomatic patients, published January 1, 2014 to May 17, 2024; studies came from Europe (7), North America (7) and Asia (17)
• Exposure: The review examined machine learning models using signs/symptoms plus other patient characteristics. For the seven mesothelioma studies, asbestos (together with C-reactive protein and pleural protein exposure) was considered as a key risk factor. No fiber types are named.
• Main results:
  • Of 5646 records identified, 34 studies were included
  • Mesothelioma was the second most common cancer studied (7 studies), after lung cancer (9) and before gastrointestinal cancer (4)
  • Across included studies, models showed variable performance, with AUC values from 0.60 to 1 during validation
  • Only two of the included studies performed external validation
  • Only two included studies evaluated model calibration
  • 94.1% of included studies were rated at high risk of bias for the index test
  • None of the 34 included studies explicitly acknowledged adherence to the TRIPOD reporting guidelines
  • In one mesothelioma study using a feature-selection algorithm (Harris Hawk Optimization), SVM and ANN models both achieved 100% accuracy with 17 selected variables. The review states these models were not validated.
• The authors' conclusion: The authors concluded that ML models have been reported to show potential, but the evidence is heterogeneous and frequently limited by bias and incomplete reporting. They stated that further validation and real-world assessment are needed before the models can be considered reliable for clinical use.
• Limitations the authors state:
  • Only English-language studies were included
  • Gray literature, preprints, theses and conference proceedings were not included
  • Sparse reporting in many included studies limited evaluation of calibration and reproducibility
  • Included studies were highly heterogeneous, limiting direct comparison and meta-analysis
  • Geographic distribution was skewed toward Asian countries, which may limit generalizability
  • Statistical comparators in the included studies may not have been optimal, possibly biasing comparisons in favor of ML

Topics

Medical terms in this study, explained in our Lexicon: Systematic Review, Mesothelioma.

Related studies

• 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)
• Prediction of Radiation Pneumonitis With Machine Learning in Stage III Lung Cancer: A Pilot Study — M. Yakar, D. Etiz, et al. (2021)
• Cancer-Associated Fibroblasts Influence Survival in Pleural Mesothelioma: Digital Gene Expression Analysis and Supervised Machine Learning Model — S. Borchert, A. Mathilakathu, et al. (2023)
• Distant Metastatic Pattern and Its Prognostic Significance in Malignant Pleural Mesothelioma: A Population-Based Study Based on a Machine Learning Model — J. Yu, C. Peng, et al. (2025)
• Near-infrared spectroscopy combined with machine learning for plasma-based discriminant diagnosis of malignant mesothelioma: a retrospective study — Y. Gu, R. Wang, et al. (2025)
• Evaluation of imaging techniques for early detection of intrathoracic cancers in symptomatic patients in primary care: a systematic review — B. Grigore, J.L. Peters, et al. (2025)

Browse more research in Cancer Medicine, or return to medical research.