Predictive models of malignant transudative pleural effusions

Area of Science:

• Pulmonology
• Oncology
• Diagnostic Medicine

Background:

• Malignant pleural effusions (MPEs) can present biochemically as transudates, lacking clear guidelines for diagnostic cytology.
• Determining when to perform cytology on transudative pleural effusions is crucial for accurate diagnosis.

Purpose of the Study:

• To assess the justification for performing cytology in cases of transudative pleural effusion.
• To develop and evaluate predictive models for identifying neoplastic origin in transudative pleural effusions.

Main Methods:

• Logistic regression analysis was used to estimate malignancy probability in transudative pleural effusions.
• Two prognostic models were developed: (I) clinical-radiological variables, and (II) clinical-radiological plus analytical variables.
• Model performance was assessed using receiver operating characteristic (ROC) curves and area under the curve (AUC).

Main Results:

• A total of 281 transudative pleural effusions were analyzed (26 malignant, 255 non-malignant).
• Model 1 (clinical-radiological) achieved an AUC of 0.973, and Model 2 (including CEA) achieved an AUC of 0.995.
• Both models demonstrated high discrimination with no false negatives at specific probability thresholds, though adjusted thresholds for 95% confidence required lower cut-offs.

Conclusions:

• Predictive models effectively distinguish transudative pleural effusions of neoplastic origin.
• Combining clinical-radiological and analytical variables (Model 2) showed superior predictive ability over clinical-radiological variables alone.
• These models aid in determining the necessity of cytology for transudative pleural effusions.