Rational choice of a diagnostic test for the detection of bone metastases in breast cancer as a function of prevalence (based on the Bayes theorem)

• Published in: Bulletin du cancer Vol. 74; no. 1; p. 82
• Main Authors: Robillard, J, Ly Van Hoa, Switzers, O, Couette, J E, Brune, D
• Format: Journal Article
• Language: French
• Published: France 1987
• Subjects: Analysis of Variance; Bayes Theorem; Bone Neoplasms - diagnosis; Bone Neoplasms - secondary; Breast Neoplasms - diagnosis; Breast Neoplasms - pathology; Female; Humans; Probability; Tomography, X-Ray Computed
• ISSN: 0007-4551

Since 1976 to 1978 inclusive, 465 patients with primary breast cancers underwent whole body analogical scintiscans, 456 numerical scans, and 453 were X-rayed. Ninety per cent received axillary node ablation, 62% presented adenopathies of which a third were capsular ruptures and 54% were T3 and T4. In 53 patients, the presence of metastasis was estimated by clinical evaluation. In 38 patients, metastasis were confirmed. Probability for the patient to present metastasis when the test is positive, is: P (M/T +) = Se.p/(Se.p) + (1-p) + (1-p) (1-Spe) Probability when the test is negative is: P (M/T -) = p(1-Spe)/p(1-Se) + (1-p)Spe p = prevalence Se = sensibility Spe = specificity Performance of the test is the probability of obtaining a correct result is: P = (f1) - (f2) = (f3) The (f3) curves were plotted for each of the three tests. For a prevalence of 11% (clinical estimate) the X-ray or analogical scintiscans tests achieve the same performance, but that for a prevalence of 8% (exact number) the radiological examination yields the best. In two cases, numerical scan is better.