Exploratory classification of breast cancer status using anthropometric and physiological parameters: A logistic regression pilot study

Abstract

Context: Breast cancer is a major global health problem, and early identification of individuals requiring further clinical assessment remains important. Anthropometric and physiological measurements are inexpensive and readily obtainable, but their ability to distinguish malignant from benign breast conditions is uncertain.

Objective: This pilot study evaluated the exploratory classification performance of anthropometric and physiological variables in distinguishing breast-cancer status using logistic regression.

Materials and Methods: A hospital-based observational pilot study was conducted at the University of Benin Teaching Hospital. Data from 20 participants were analysed. The recorded variables were age, weight, height, systolic blood pressure, diastolic blood pressure, and body-surface temperature. Missing values were replaced using column means, and the predictors were standardized before model fitting. The dataset was divided into training and testing subsets using an 80:20 split. Classification performance was assessed using accuracy, precision, recall, F1 score, confusion-matrix analysis, and the area under the receiver-operating-characteristic curve.

Results: The test set contained four participants. The logistic-regression model achieved an accuracy of 75.0%. For the benign class, precision was 1.00, recall was 0.50, and the F1 score was 0.67. For the malignant class, precision was 0.67, recall was 1.00, and the F1 score was 0.80. The ROC-AUC was 0.50, indicating no demonstrated discrimination beyond chance in this test set. Because the model was evaluated on only four participants, these performance estimates are highly unstable and should not be interpreted as definitive.

Conclusion: In this small pilot dataset, age, weight, height, blood pressure, and body-surface temperature alone did not demonstrate adequate discriminatory performance for breast-cancer classification. Larger, independently validated datasets incorporating clinical, imaging, pathological, and possibly molecular variables are required before such a model can be considered for clinical risk stratification.