%0 Journal Article
%T Risk Prediction Model and Early Recommendation for Postoperative Recurrence of Breast Cancer Patients in Tanzania
%A Asha Khatib Iddi
%A Siyu Zhang
%A Bongani Mbambara
%A Alfred Jubilate
%A Mundhir Semm Ally Basinda
%A Paul Michael Malugulu
%A Chen Chen
%A Nazima Dharsee
%A Jianjun Zou
%J Open Access Library Journal
%V 12
%N 7
%P 1-22
%@ 2333-9721
%D 2025
%I Open Access Library
%R 10.4236/oalib.1113728
%X <p>
	<strong>Introduction</strong>: Breast Cancer (BC) remains a significant health concern worldwide, and accurate prediction of its recurrence after surgery is vital for patient management and treatment decisions. This study aimed to develop a predictive model for assessing the risk of breast cancer recurrence (BCR) after surgery in the Tanzanian population. <strong>Methods:</strong> This study retrospectively analyzed data collected from BC patients at multiple centers in Tanzania. The outcome was BCR within 2 years after surgery. Six different ML models were established, and their performances were compared. The SHapley Additive explanations (SHAP) method was utilized to interpret the importance of variables. Finally, a web-based risk calculator was developed to facilitate its clinical application. <strong>Results</strong>: 199 BC patients were included, of which 139 were used for model development and 60 for model evaluation. The BCR incidence was 72.9%. Key predictors of BCR included lymph node metastasis, tumor grade, number of lymph nodes, tumor size at diagnosis, and marital status. In the testing set, the multilayer perceptron (MLP) model demonstrated the highest performance: AUROC (0.935; 95% CI: 0.878 - 0.992), AUPRC (0.969; 95% CI: 0.839 - 0.995), accuracy (0.850), sensitivity (0.875), specificity (0.800), and F1-score (0.886). The MLP model demonstrated superior calibration performance (Brier score: 0.115) and showed greater clinical utility than other models in decision curve analysis across the threshold probability range of 0.2 - 0.8. <strong>Conclusion</strong>: We successfully developed a risk prediction model and constructed it as a dynamic web-based risk calculator available online. These findings support the implementation of early intervention strategies to reduce postoperative BCR rates in Tanzania.
</p>
%K Breast Cancer Recurrence
%K Risk Prediction Model
%K Machine Learning
%K Personalized Treatment
%K Zanzibar-Tanzania
%U http://www.oalib.com/paper/6864870