"Point-of-care Analysis for Non-invasive Diagnosis of Oral cancer (PANDORA): a technology-development proof of concept diagnostic accuracy study of dielectrophoresis in patients with oral squamous cell carcinoma and dysplasia," by Fedele is an interesting paper that updates the concept of using diffentail movenet of purified cells in variable frequency alternating current electric field to differentiate transformed cells from non-transformed cells. The authors take advantage advances in measurement of cellular changes during dielectrophoresis which has been automated in a commercially available instrument. A crucial question is how the cells were stored. It would be nice if there was a little information about this as it seems as if the cells were originally collected prior to 2016 and were presumably were assayed much more recently.
It is gratifying that the location of the lesion are considered but is not absolutely required for fairly high accuracy of classification for large shallow lesions location is not always clear. The inclusion of dysplasia with OSCC makes sense as the authors state identifying lesions that are potentially precancerous, and not just cancers, is a major clinical goal. It would be nice to have a report on the levels of dysplasia present. Also need to report on types of benign lesions. One might speculate that it would be fairly easy to separate OSCC from some types of benign lesions but not others based on dielectrophoresis properties of the dominant cells.

The authors use reciever operator characteristic curves to determine optimal cutoff for positive and negative classification of samples. Of course it needs to be validated with an external data set but this paper is an important start toward that task.

As the authors state themselves it may be misleading to report on positive and negative predictive clues when dealing with a study group that has an artifical distribution of different sample types. Screening a population with a 36% rate of dysplasia/OSCC does not represent a real high risk population that might be screened for OSCC. The authors might be better off leaving out PPV and NPV. I’m not sure but I think leaving it in might detract from an otherwise interesting study that is extremely well presented and written up.

Is there anything known about the effect of current tobacco usage on oral cell dielectrophoresis properties – that is always a a quetiton when OSCCs are compared to normals in any assay.