Study objectives: To assess the accuracy of pleural fluid (PF) pH in predicting duration of survival of patients with malignant pleural effusions.
Design: Analysis of patient-level data from nine sources retrieved from a MEDLINE search and correspondence with primary investigators.
Study selection: Published and unpublished studies that report PF pH values and duration of survival of patients with malignant pleural effusions.
Data collection and analysis: Primary investigators supplied patient-level data (n = 417), which was examined by receiver operating characteristic (ROC) analysis, logistic regression, and survival time modeling to determine the utility of PF pH for predicting survival compared with other clinical factors. The primary investigations were graded for study design.
Measurements and results: Median survival (n = 417) was 4.0 months: PF pH (p < 0.0039) was an independent predictor of survival duration. A PF pH test threshold < or = 7.28 had the highest accuracy for identifying poor 1-, 2-, and 3-month survivals. The predictive accuracies of PF pH (area under the ROC curve range, 0.571 to 0.662) and a PF pH-high-risk tumor (lung, soft tissues, renal, ovary, gastrointestinal, prostate, and oropharynx) model (odds ratio range, 2.91 to 6.67), however, were modest for predicting 1-, 2-, and 3-month survival. Only 54.4% and 62.7% of patients identified by PF pH < or = 7.28 or the PF pH-high-risk tumor model to die within 3 months were correctly classified. Weaknesses of the primary data were identified.
Conclusions: PF pH has insufficient predictive accuracy for selecting patients for pleurodesis on the basis of estimated survival.