A logit model to evaluate the performance of diagnosis of solid palpable breast tumours

The applicability of log-linear (logit) analysis has been tested with preoperative data of 303 consecutive patients with a solid palpable breast tumour. The log-linear routine of the SPSS-package provides for the selection of appropriate diagnostic tests, which can be suitably judged in a given situation. The histologic diagnosis (benign or malignant) has been chosen as dependent variable. The independent variables are: (1) the patient's age (less than 50 years or greater than or equal to 50 years), (2) the size of the lesion (diameter less than 2 cm or greater than or equal to 2 cm) and the test results (benign or malignant) of (3) physical examination, (4) mammography, (5) ultrasound scanning and (6) thermography. Of the tested models with one independent variable, the best model uses ultrasonography test results (H = 0.35). Physical examination and mammography also have good predictive power (H, respectively: 0.23 and 0.25). The combination of two independent variables augments accuracy. Models combining ultrasonography with age, physical examination or mammography are shown to be most successful (H respectively: 0.41, 0.42 and 0.39).