929P - Advanced laryngeal squamous cell carcinoma prognosis and machine learning insights

Background

The incidence of Laryngeal Squamous Cell Carcinoma (LSCC) has increased over the past 30 years, presenting new challenges in the management of this disease. In response to these evolving needs, we developed a machine learning (ML) prognostic model to optimize the treatment strategies for patients diagnosed with advanced LSCC.

Methods

Data from 2000 to 2020 were obtained from the SEER database. Patients who were not diagnosed based on histology, previous history of cancer or other concurrent malignancies, or unknown data were excluded. The chi-square test was used to compare variables. The Kaplan-Meier estimator, log-rank test, and Cox regression analysis identified prognostic factors for overall survival (OS) and cancer-specific survival (CSS). We constructed prognostic models using ML algorithms to predict 5-year survival. The data were randomly divided into training (70 %) and validation (30 %) sets. A validation method incorporating the area under the curve (AUC) of the receiver operating characteristic curve was used to validate the accuracy and reliability of the ML models.

Results

A total of 7,350 patients were included: 2,689 with glottic (GC), 4,349 with supraglottic (SuGC), and 312 with subglottic (SG) cancers. The median patient age was 60 years. A total of 88.8% of the patients had a T3 and T4. The largest racial group was white, comprising 74.8% of the cases, and 21.1% of the cases were black. Most cases were stage IV AJCC (89.5%), and 41.4% of the patients had N2. Of the patients, 52% underwent total laryngectomy, which was the most common surgical procedure (39.1%). ML was used to identify the most important features of GC, including age, sex, and AJCC staging. Age and sex were considered as key factors for SuGC metastasis. For SC, the most important features identified were surgery combined with radiotherapy, sex, and surgery combined with chemoradiotherapy.

Conclusions

This study presents ML models tailored to three anatomical sites of LSCC. Key factors such as age and sex for GC and SuGC and treatment combinations for SG. These models underscore their utility in personalized medicine, enabling clinicians to customize treatment strategies based on individual characteristics and improve clinical outcomes.

Clinical trial identification

Editorial acknowledgement

Funding

Has not received any funding.

Disclosure

All authors have declared no conflicts of interest.