Limitations of high frequency ultrasound in determining the invasiveness of cutaneous malignant melanoma.

Preoperative assessment of melanoma invasion by sonometry has been reported to exhibit a high correlation with postoperative histometry based on the determination of a regression coefficient. We believe that this approach is not adequate and that it can be misleading. We therefore applied an appropriate statistical approach to our data to re-examine this issue. Preoperative measurements by sonometry were compared with postoperative histometry in 71 consecutive patients with the clinical diagnosis of malignant melanoma. Patients with the histological diagnosis of melanoma in situ (n = 13) were excluded from this comparison, as sonometry cannot distinguish structures restricted to the epidermis and Breslow thickness is not routinely determined in Clark level I melanomas. The agreement between sonometry and histometry was analysed by plotting the mean of the two methods against their difference. By linear regression analysis the correlation coefficient (r) in the invasive melanomas (n = 58) was found to be 0.92. The median thickness of the invasive tumours by sonometry was 1.36 mm, and by histometry 0.89 mm. In 95% of the cases sonometry differed by 37% above to 48% below the values obtained by histometry. Thus, we conclude that the accuracy of sonometry in predicting preoperative tumour thickness- and tumour invasiveness-is far lower than previously reported.