Evaluation of a new multiple regression model based on biochemical parameters for the distinction of canine exudates...

The classification of effusions in human medicine currently uses biochemical parameters of verified analytical accuracy, while veterinary medicine is traditionally guided by protein content (TP) and total nucleated cell count (TNCC) in the effusion, without solid scientific support.

We aimed to assess the accuracy of the current veterinary classification system to distinguish transudates from exudates and create new tools involving biochemical parameters that better classify canine cavitary effusions.

Clinical, laboratory, and imaging data from 250 canine pleural and peritoneal effusions were retrospectively and prospectively collected, organized, and statistically evaluated. Multiple logistic regression analysis was performed using biochemical and cellular parameters.

For identifying exudates, the accuracy (87.7%, n = 204) of the best traditional classification system (TNCC > 3000 cells/μL) was similar to that of the individual biochemical cutoff values with the greatest accuracy in the abdominal cavity (eg, cholesterol, CHO-E > 40.1 mg/dL, 87.3%, n = 55). The accuracy of albumin (ALB-E > 0.8 g/dL) in the pleural cavity was nonetheless higher (100%, n = 23). The best multiple predictive models for any cavity used the percentage of neutrophils and CHO-E (n = 72), presenting an accuracy, sensitivity, and specificity for the diagnosis of exudate of 88%, 96%, and 67%, respectively.

Biochemical classification of pleural effusions has a higher accuracy than the traditional system (based on TP and TNCC). Utility and cutoff of analytes are different for each cavity. Implementing a multiple regression model or establishing ratios or gradients with concurrent serum values adds no significant improvement in the diagnostic potential of distinguishing transudate and exudates in dogs.