USE OF MORPHOMETRY AS AN AID IN THE DIFFERENTIAL-DIAGNOSIS OF LARGE-CELL CARCINOMA OF THE LUNG

In a previous cytopathologic study a regression equation was generated that aided in the differential diagnosis of large cell carcinoma (LCC ) of the lung from poorly differentiated adenocarcinoma (AC). This stu dy was undertaken to validate and extend those findings. Cytopathologi c and histopathologic specimens from 20 new patients were examined usi ng computerized morphometry. There were 28 specimens from 10 cases of LCC and 25 specimens from 10 cases of AC (10-25 cells measured from ea ch specimen). Histopathologic review by an expert (S.D.G.) provided th e ''gold standard'' diagnosis. Morphometric parameters studied were (1 ) nucleolar/nuclear area ratio, (2) cell area, (3) cell form factor, a nd (4) nuclear form factor. Sensitivity analysis validated the diagnos tic utility of our prior equation. A logistic regression with these pa rameters determined the probability that each sample was LCC. Relative operating characteristic curve analysis determined an optimum logisti c cutoff point of 0.83. At this decision level the equation had a sens itivity of 72% for distinguishing LCC. Specificity was 46%, positive p redictive value was 65%, negative predictive value was 55%, and classi fication efficiency was 61%. Morphometry of histopathologic material c ontributed no important additional information. Therefore, our logisti c regression and sensitivity analysis supported the clinical utility o f certain morphometric measurements in the cytopathologic, but not his topathologic, diagnosis of LCC.