Prognostic Factors of Marjolin Ulcers: A Meta-Analysis and Systematic Review Assisted by Machine Learning Techniques. Issue 4 (9th December 2022)

• Record Type: Journal Article
• Authors: Cheng, Kai-Yuan; Yu, Jiaxin; Liu, En-Wei; Hu, Kai-Chieh; Lee, Jian-Jr
• Is Part Of: Plastic and reconstructive surgery. Volume 151:Issue 4(2023)
• Page Start: 875
• Page End: 884
• Publication Date: 2022-12-09
• Subjects: Surgery, Plastic -- Periodicals; 617.95205
• DOI: 10.1097/PRS.0000000000010012
• Languages: English
• ISSNs: 0032-1052

Abstract

Abstract : Background: Marjolin ulcers (MUs) are malignant tumors arising from previously injured skin, including burn wounds, scars, chronic ulcers, and other chronic nonhealing inflammatory conditions. They have a potentially long latent period. The authors aimed to establish the prognostic factors for recurrence, metastasis, and disease-specific death related to MU. Methods: The authors performed a comprehensive search of PubMed, Embase, and the Cochrane Library. After assessing the methodologic quality of case series, they performed a meta-analysis and systematic review. Furthermore, the authors used machine learning to predict patient survival time. Results: MUs on the upper limbs, head, and neck had a higher risk of recurrence. Contrastingly, lower grade lesions, absence of lymph node metastasis, and a tumor diameter of less than 10 cm were associated with lower recurrence risk. The risks were unrelated to age and latent period. In addition, patients without lymph node metastasis had a lower risk of developing distant metastasis. Furthermore, the risk of disease-specific death was lower in patients with a lower tumor grade, absent lymph node metastasis, small tumor diameter (<10 cm), and tumors located in regions other than the head and neck. Correlation analysis showed that the age at initial injury was negatively correlated with the latent period of MU. Conclusions: The authors found that tumor grade, tumor site, lymph node status, and tumor size are important predictors of a worse prognosis. To integrate these predictors, the authors created an equation to predict the survival time for individual patients by means of machine learning processes. Moreover, the authors found that MU developed more quickly in older individuals with injuries.