Inter‐observer variation in the histopathological diagnosis of clinically suspicious pigmented skin lesions

When a biopsy is taken of a suspicious pigmented skin lesion, histological examination is expected to establish the definitive diagnosis. This study evaluated the inter-observer variation of 20 pathologists in the histological diagnosis of a randomly selected set of suspicious pigmented skin lesions (PSLs), by comparing their diagnoses to a reference diagnosis. Overall sensitivity for melanoma was 87%, ranging from 55% to 100% between the observers. Sensitivity was significantly lower for thin (Breslow thickness <1 mm) than for thick melanomas (83% versus 97%, p=0.005). Overall melanoma specificity was 94%, ranging from 83% to 100% between observers. Dysplastic naevus was the most important source of false-positive diagnoses, mainly in situ melanomas. Positive and negative predictive values in the given test set were 75% and 97%, respectively. In the case of melanoma, there was quite some variation in measured Breslow thickness. This would have led to a different therapeutic approach in 12% of the readings. Some of the variation seemed to be due to a different interpretation of the presence of a co-existent naevus. In 9% (3/35) of the readings, participants did not agree on the presence of ulceration. These results reflect a tendency to overdiagnose mainly thin melanomas in general histopathological practice. They also demonstrate variation in the assessment of major prognostic factors of melanoma.