Scholarlink

Background and purpose <p>This study aimed to predict the treatment outcomes and survival of patients with locally advanced cervical cancer (LACC) receiving chemoradiotherapy (CRT) using an unsupervised clustering machine learning method.</p> Materials and methods <p>This retrospective study was based on a&#xa0;cohort of 152 consecutive patients. Treatment consisted of definitive CRT, combining external beam radiotherapy to the pelvis with intracavitary brachytherapy to achieve a&#xa0;total equivalent dose of 85–90 Gy at the tumor site. Patient-related data including age, body mass index, standard blood tests and complete blood count were recorded before CRT. Various inflammatory indices were analyzed, including the neutrophil–lymphocyte ratio (NLR), platelet-lymphocyte ratio (PLR), leukocyte–lymphocyte ratio (LLR), systemic immune inflammation index (SII), and aspartate aminotransferase (AST) to neutrophil ratio index (ANRI). Based on these covariates, an unsupervised clustering method based on the agglomerative hierarchical clustering (AHC) algorithm was used to identify clusters of patients. The groups of patients were compared in terms of local control (LC), disease-free survival (DFS), distant metastases-free survival (DMFS), and overall survival (OS). A&#xa0;Cox proportional hazard regression analysis was performed to evaluate the relationship between the clusters and the clinical outcomes.</p> Results <p>Clustering analysis reported an optimal number of clusters equal to two. Analysis of variance indicated that the variables contributing most to the separation of the clusters were SII, LLR, ANRI, PLR, NLR, hemoglobin, and white cells count. Significant differences were found between the two groups of lesions in terms of LC (<i>p</i> &lt; 0.001), DFS (<i>p</i> = 0.019), and OS (<i>p</i> = 0.017). At 2&#xa0;years, LC, DFS, and OS were 93.5%, 72.0%, and 93.1%, and 92.7%, 72.0%, and 70.8% for clusters&#xa0;1 and&#xa0;2, respectively. In the unadjusted Cox model, patients in cluster&#xa0;1 were significantly more likely to experience higher local control (HR 3.88 [95% CI 1.80–8.37]; <i>p</i> = 0.001), disease-free survival (HR 1.97 [95% CI 1.10–3.51]; <i>p</i> = 0.022), and overall survival (HR 2.16 [95% CI 1.13–4.14]; <i>p</i> = 0.021).</p> Conclusion <p>This study highlights the predictive value of blood parameters and inflammatory indexes for risk stratification in LACC. An unsupervised clustering approach is able to stratify the treatment outcomes with significant performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Unsupervised clustering analysis unravels the role of systemic inflammatory indices in the prognosis of patients with locally advanced cervical cancer treated with chemoradiation

  • 期刊论文

  • 发表日期：

    2026年3月17日

查看全文
  • Savino Cilla,
  • Federica Medici,
  • Martina Ferioli,
  • Alessandra Arcelli,
  • Anna Myriam Perrone,
  • Milly Buwenge,
  • Pierandrea De Iaco,
  • Alessio Giuseppe Morganti

摘要

Background and purpose

This study aimed to predict the treatment outcomes and survival of patients with locally advanced cervical cancer (LACC) receiving chemoradiotherapy (CRT) using an unsupervised clustering machine learning method.

Materials and methods

This retrospective study was based on a cohort of 152 consecutive patients. Treatment consisted of definitive CRT, combining external beam radiotherapy to the pelvis with intracavitary brachytherapy to achieve a total equivalent dose of 85–90 Gy at the tumor site. Patient-related data including age, body mass index, standard blood tests and complete blood count were recorded before CRT. Various inflammatory indices were analyzed, including the neutrophil–lymphocyte ratio (NLR), platelet-lymphocyte ratio (PLR), leukocyte–lymphocyte ratio (LLR), systemic immune inflammation index (SII), and aspartate aminotransferase (AST) to neutrophil ratio index (ANRI). Based on these covariates, an unsupervised clustering method based on the agglomerative hierarchical clustering (AHC) algorithm was used to identify clusters of patients. The groups of patients were compared in terms of local control (LC), disease-free survival (DFS), distant metastases-free survival (DMFS), and overall survival (OS). A Cox proportional hazard regression analysis was performed to evaluate the relationship between the clusters and the clinical outcomes.

Results

Clustering analysis reported an optimal number of clusters equal to two. Analysis of variance indicated that the variables contributing most to the separation of the clusters were SII, LLR, ANRI, PLR, NLR, hemoglobin, and white cells count. Significant differences were found between the two groups of lesions in terms of LC (p < 0.001), DFS (p = 0.019), and OS (p = 0.017). At 2 years, LC, DFS, and OS were 93.5%, 72.0%, and 93.1%, and 92.7%, 72.0%, and 70.8% for clusters 1 and 2, respectively. In the unadjusted Cox model, patients in cluster 1 were significantly more likely to experience higher local control (HR 3.88 [95% CI 1.80–8.37]; p = 0.001), disease-free survival (HR 1.97 [95% CI 1.10–3.51]; p = 0.022), and overall survival (HR 2.16 [95% CI 1.13–4.14]; p = 0.021).

Conclusion

This study highlights the predictive value of blood parameters and inflammatory indexes for risk stratification in LACC. An unsupervised clustering approach is able to stratify the treatment outcomes with significant performance.