Psallidas I, et al. - Researchers aimed to create, validate, and prospectively evaluate biomarkers of survival and pleurodesis response in malignant pleural effusion. Also, they wanted to create a score for survival prediction. Biological and clinical parameters were integrated to create the PROMISE score (the first prospectively validated prognostic model for malignant pleural effusion as per researcher's knowledge) to accurately estimate 3-month mortality. In addition to yielding important information on patient prognosis and serving as a guide to select appropriate management strategies, this robust, clinically relevant prognostic score could be applied immediately.

Methods

• Researchers performed this multicohort study, where they used five separate and independent datasets from randomized controlled trials to investigate potential biomarkers of survival and pleurodesis.
• They used mass spectrometry-based discovery to investigate pleural fluid samples for differential protein expression in patients from the discovery group with different survival and pleurodesis outcomes.
• The development of a model to predict 3-month mortality was attempted by entering clinical, radiological, and biological variables into least absolute shrinkage and selection operator regression.
• They used internal and external validation to assess the model.

Results

• In the discovery dataset, they identified 17 biomarker candidates for survival and seven for pleurodesis.
• Biomarker validation was accomplished using 3 independent datasets (n=502).
• They found that all pleurodesis biomarkers failed, and accurate predictors of survival included gelsolin, macrophage migration inhibitory factor, versican, and tissue inhibitor of metalloproteinases 1 (TIMP1).
• Validation of 8 variables (haemoglobin, C-reactive protein, white blood cell count, Eastern Cooperative Oncology Group performance status, cancer type, pleural fluid TIMP1 concentrations, and previous chemotherapy or radiotherapy) was done and these were used to develop a survival score.
• Good discrimination was demonstrated by internal validation with bootstrap resampling and external validation with 162 patients from two independent datasets (C statistic values of 0.78 [95% CI 0.72–0.83] for internal validation and 0.89 [0.84–0.93] for external validation of the clinical PROMISE score).