The New England Journal of Medicine has just published an open access paper, "Fibulin-3 as a Blood and Effusion Biomarker for Pleural Mesothelioma", describing a study of the use of plasma fibulin-3 in detecting mesothelioma early and without generating a lot of false positives among those with significant prior asbestos exposure but who do not have mesothelioma. The results were impressive – sensitivity was 100% and specificity was 94.1%. Fibulin-3 was also found to be particularly useful in discriminating between pleural effusions driven by mesothelioma and those driven by other malignant or benign processes. The latter finding will be particularly useful in asbestos litigation as we still routinely fight over the diagnosis even when immunohistochemical staining at first blush appears definitive. Why?
We’re not sure but an answer is suggested in another new paper, this one published in The American Journal of Surgical Pathology. It’s "Utility of a CEA, CD15, Calretinin, and CK5/6 Panel for Distinguishing Between Mesotheliomas and Pulmonary Adenocarcinomas in Clinical Practice". If you had four biomarkers repeatedly demonstrated to be useful in deciding the mesothelioma versus adenocarcinoma question you’d think your accuracy would go up as the number of biomarkers confirming your decision went up. But apparently you’d be wrong – at least if the staining was done and interpreted by clinical rather than medical laboratories specialized in immunohistochemical staining. Of the four biomakers commonly used to either rule-in (it’s mesothelioma) or rule-out (it’s really adenocarcinoma) the disease, errors were so common in clinical medicine laboratories that determinations requiring all four factors to line up as expected were only about half as accurate as those determinations based on which way a majority of the factors pointed.
One answer to our question about why we’re still fighting over diagnoses would thus be that a surprisingly high percentage of mesothelioma versus adenocarcinoma decisions based on tissue preparation and staining done in typical clinical laboratories are simply wrong. Another, and in many ways similar answer, is that none of these biomarkers are perfectly predictive so that the more of them you consider the less likely a particular conflicting result is truly dispositive.