Discrimination between malignant and nonmalignant ascites using serum and ascitic fluid proteins in a multivariate analysis model

Our objectives were to study the value of different proteins in the serum a nd ascitic fluid and assess their potential in discriminating between malig nant and nonmalignant ascites in a model that could be developed to aid cli nical diagnosis. In all, 57 different measurements (30 in serum and 27 in a scitic fluid) including erythrocyte sedimentation rate, number of white blo od cells, cytokines, interleukin-la (IL-1a), IL-lb, IL-2, IL-G! IL-8, tumor necrosis factor-alpha, immunoglobulins (IgG, IgA, IgM), complement factors C3 and C4, acute-phase proteins such as alpha(1)-acid glycoprotein, alpha( 2)-macroglobulin, alpha(1)-antitrypsin, haptoglobin, C-reactive protein, fe rritin, cerulo-plasmin and transferin, were performed in 61 patients with a scites (25 with malignant exudates, 13 with nonmalignant exudates, and 23 w ith transudates). Patients with sepsis were excluded. Correlation tests and one-way ANOVAs were used for comparisons between different groups. Discrim inant analyses were used to assess the significance of each parameter in th e differentiation process. Correct classification of 100% of cases required the use of all 57 ascitic fluid measurements in the model, which was not c onsidered practical in clinical diagnosis. Discriminant analysis showed tha t five ascitic fluid measurements-total protein, LDH, TNF-alpha, C4, and ha ptoglobin-were sufficient for a model to correctly classify 89% of cases. C ross-validation showed that 70% of unknown cases were correctly classified using this model. In conclusion, we have shown that five easily taken prote in measurements in the ascitic fluid can differentiate to a large extent be tween eases with ascites and have proposed a relatively simple statistical model with these parameters that could be developed to be extremely useful in the clinical setting.