Design: Prospective interventional diagnostic study, single centre
Population: 42 non-melanoma skin cancers in 39 patients; 261 Mohs margins
Primary outcome: Agreement with histopathology and between readers
Effect: Accuracy 96.4%; inter-reader kappa 0.863 for margins; one false-negative margin

A prospective interventional study in Modern Pathology, published on 9 October, examined 42 non-melanoma skin cancers from 39 patients having Mohs surgery with ex vivo fluorescence confocal microscopy, which produces digitally stained, H&E-like images of fresh tissue in minutes. A dermatologist and a pathologist trained in the method read 261 peripheral and deep margins independently and blind.

Agreement between the two readers was almost perfect for margin status (kappa 0.863, observed agreement 98.6%) and for tumour type (kappa 0.882). Against conventional histopathology, overall accuracy was 96.4%. There were two discrepancies: one false-negative margin and one attributed to processing differences.

The method could replace or supplement frozen sections in Mohs workflows, and the near-identical performance of dermatologist and pathologist readers matters for who reads them. The single false-negative is the number to keep in mind, and time savings were not measured.

Confocal margin reading agreed with histology in about 96% of margins in trained hands.
Train readers formally; agreement here depended on prior training.
Keep a policy for equivocal or discordant margins — re-excise and re-evaluate, as this study did.
Audit false-negative margins locally if adopting the method.

Why it matters

It is a credible alternative to frozen sections for Mohs, and pathologists may be asked to read it.

The statistics, in plain English

A kappa above 0.8 means agreement far beyond chance. The AUC of 0.875 has a wide confidence interval (0.73 to 1.00) because there were few positive margins, so the estimate of how well it discriminates is less certain than the raw accuracy suggests.