Accuracy of clinical diagnosis of skin lesions

Skin cancer is an increasing problem in fair-skinned populations worldwide. It is important that doctors are able to diagnose skin lesions accurately.

To compare the clinical with the histological diagnosis of excised skin lesions from a set of epidemiological data. We analysed diagnostic accuracy stratified by histological subtype and body site and examined the histological nature of misclassified diagnosis.

All excised and histologically confirmed skin cancers in Townsville/Thuringowa, Australia from December 1996 to October 1999 were recorded. Positive predictive values (PPVs) and sensitivities were calculated for the clinical diagnoses and stratified by histological subtype and body site.

Skin excisions in 8694 patients were examined. PPVs for the clinical diagnoses were: basal cell carcinoma (BCC) 72.7%; squamous cell carcinoma (SCC) 49.4%; cutaneous melanoma (CM) 33.3%. Sensitivities for the clinical diagnosis were: BCC 63.9%; SCC 41.1%; CM 33.8%. For BCC, PPVs and sensitivities were higher for the trunk, the shoulders and the face and lower for the extremities. The reverse pattern was seen for SCCs.

Diagnostic accuracy was highest for BCC, the most prevalent lesion. Most excisions were correctly diagnosed or resulted in the removal of malignant lesions. With nonmelanocytic lesions, doctors tended to misclassify benign lesions as malignant, but were less likely to do the reverse. Although a small number of clinically diagnosed common naevi subsequently proved to be melanoma (6.3%), a higher proportion of all melanomas had been classified as common naevi (20.9%). Accuracy of diagnosis was dependent on body site.