Breakthrough Study Analyzes Prognostic Significance of Breast Cancer Recurrence

A groundbreaking study conducted at Semmelweis University Budapest delves into the critical factors influencing local recurrence (LR) in breast cancer patients. This research leverages advanced machine learning algorithms to predict outcomes based on comprehensive clinicopathological data, offering valuable insights for healthcare professionals.

Study Overview and Key Objectives

The study began with 448 breast carcinoma cases involving local recurrences. To ensure accuracy, researchers focused on cases where detailed clinicopathological parameters were available for both primary breast cancers and their corresponding local recurrences. As a result, many cases with incomplete data were excluded, leaving a final cohort of 154 primary breast carcinomas and local recurrence pairs diagnosed between 1984 and 2018.

This retrospective study aligns with the Declaration of Helsinki and received ethical approvals from the Hungarian Medical Research Council. Data were sourced from the Department of Pathology, Forensic and Insurance Medicine, and the Health Care Database at Semmelweis University.

Patient and Recurrence Data

Local recurrence was defined using the Maastricht Delphi Consensus criteria. Local recurrence-free survival was calculated as the time from primary breast cancer diagnosis to the occurrence of the first LR diagnosis, while distant metastasis-free survival (DMFS) was measured from the primary diagnosis to the first distant metastasis.

The study’s dataset included detailed patient information, such as age at diagnosis, tumor grade, size, nodal involvement, lymphovascular invasion, resection margins, and treatment regimens. It also classified tumors into subtypes based on four immunohistochemical markers: estrogen receptor (ER), progesterone receptor (PR), Ki67 index, and HER2 status, in accordance with the 2013 St. Gallen Consensus.

PIK3CA Mutation Analysis

Mutations in the PIK3CA gene were assessed in 34 paired primary tumors and local recurrences using the Cobas^® PIK3CA Mutation Test. This test can detect mutations in specific exons of the PIK3CA gene, crucial for understanding tumor behavior and prognosis.

Data Preparation and Machine Learning

To prepare the data for machine learning, categorical variables with three or more distinct values were one-hot encoded, transforming them into binary values. The final dataset comprised 154 patients and 84 features, which were then split into training and testing sets.

The study employed various machine learning algorithms, including XGBoost and Random Forest, optimizing hyperparameters through 5-fold cross-validation. After extensive testing, the best-performing models were selected for each prediction task, including DM after LR.

Model Explanation and Validation

The SHAP (SHapley Additive exPlanations) method was used to interpret model predictions, highlighting which features most significantly impacted the outcome. This approach provides clinicians with a deeper understanding of the model’s decision-making process.

Handling Missing Data

Several clinicopathological variables contained missing values. However, since the machine learning models used, such as XGBoost and Random Forest, can handle missing data as special values, imputation techniques were not necessary. Instead, the models built rules into their decision trees to account for missing data.

Statistical Analysis

The statistical analysis involved various tests, including Fisher’s exact test, t-tests, and Pearson’s correlation coefficients, to assess relationships between variables. Kaplan–Meier survival curves and log-rank tests were used to evaluate distant metastasis-free survival.

Implications and Future Directions

This study underscores the importance of comprehensive clinicopathological data in predicting breast cancer recurrence and distant metastasis. By employing advanced machine learning techniques, researchers can provide valuable insights that could guide treatment decisions and improve patient outcomes.

Future research could expand on this study by including more patients and a wider range of clinicopathological variables. Additionally, further exploration of the role of mutations in key genes, such as PIK3CA, could offer new avenues for understanding tumor biology and improving prognostic accuracy.

Conclusion

The study represents a significant step forward in understanding the complex factors influencing breast cancer recurrence. By integrating comprehensive clinicopathological data with advanced machine learning algorithms, researchers can provide more accurate predictions and ultimately improve patient care.

We encourage readers to share their thoughts and engage with the discussion below. Your insights are valuable to the ongoing dialogue in breast cancer research and treatment.

Subscribe to Archynetys for more expert insights and updates on the latest in medical research and clinical advancements.

Share this article on social media to spread awareness and support ongoing research efforts.