The tumour invasion depth according to Breslow [1] is the single most important prognostic factor in cutaneous malignant melanoma (MM) [2], and it is well established as the main determinant in recent staging systems for cutaneous MM. High resolution ultrasound imaging can be used for determination of melanoma invasion depth in vivo for preoperative staging purposes [3,4]. This may become especially useful for the planning of procedures like sentinel lymph node dissection (SLND), which is mostly performed in melanomas of more than 1.0 mm invasion depth.

Although the correlation between sonographical and histological measurements is high, it is subject to inaccuracies and several sources of error [5, 6]. Sonometric measurement of MM invasion depth in vivo generally reveals somewhat higher values than histometric measurement [3, 5, 7]. Various factors including shrinkage of the fixed tissue [8] and misclassification of subtumoral inflammatory cells as part of the tumour [4] have been made responsible for this effect.

We investigated the ultrasound velocity in human melanoma tissue, since its overestimation would be an alternative explanation for measurement bias in sonometric measurement of MM invasion depth.

Biopsy material of four patients with very large primary melanomas was used in order to obtain enough tissue for experimental measurements but still preserve enough of the specimen for routine histological evaluation. The specimen was processed immediately after surgery, cleaned from all subtumoral or subcutaneous tissue and placed in 0.7% saline.

Ultrasonographic measurements were performed using a DBR300 ultrasound device (Sonometrics Inc., New York, NY, USA) equipped with a highly damped 20 MHz transducer.

Ultrasound velocity was calculated by measuring signal runtimes along a constant distance (s₀) of 12.0 mm between the transducer and a plane plexiglas bottom either through water alone or with the melanoma specimen placed onto the plexiglas (Fig. 1). The runtimes were calculated using electronic cursors placed on the appropriate flanks of the scope signal. If to is the time for the signal to pass the distance so through water alone, t_M the time to pass the tissue distance (S_M), and t_w the time to pass the remaining water distance (s_w), then the velocity of the signal propagation in melanoma tissue can be calculated as:

V_M = V_w (t₀-t_w)/t_M

with the ultrasound velocity in water v_w given as 1496 m/s at 25°C.[11] For theoretical reasons, the maximal resolution of the runtime measurement of a 20 MHz signal is about 0.05 μs. Technical reasons, i.e. the steepness of the signal flanks, lead to additional inaccuracies. With tenfold measurements of the empty system a standardised coefficient of variance of about 0.42 percent was reached (data not shown).

Depending on the available amount of tissue, measurements were performed repeatedly at different sites of the specimen. Mean values were calculated from tenfold measurements and compared between the different specimens using simple analysis of variance. The results were applied to a set of 402 melanoma patients using simple regression analysis of log-normalised values of Breslow's and sonographic invasion depth measurements. All calculations were done with a statistical software package (SPSS 5.0, SPSS Inc, Chicago, IL).

The recent proposal for a new staging system for malignant melanoma makes pathological staging of regional lymph nodes obligatory for all melanomas of more than 1.0 mm invasion depth [12]. Since most institutions have abandoned radical elective lymph node dissection for primary malignant melanoma, sentinel lymph node dissection (SLND) will be the preferred method for the assessment of the nodal status. SLND has been shown to give the most reliable results when the primary has not been removed at the time of lymphatic mapping. Therefore, preoperative determination of tumour thickness using high resolution ultrasound would be of great value in the staging process of primary malignant melanoma.

Ultrasonographic measurements of melanoma invasion depth in vivo are usually made under the two assumptions. First, that ultrasound velocity in primary melanoma tissue equals that of normal skin. Second, that ultrasound velocity is constant in different individual tumours. As we could show, both assumptions can only partially be confirmed.

Although we investigated only material from 4 different tumour specimens, our measurements indicate that in some melanomas the ultrasound velocity may be lower than the commonly accepted value of 1580 m/s. As a consequence, the estimated distances will be biased towards greater values in these tumours. Since the ultrasound velocity in biological tissues depends on certain characteristics of the particular tissue, it is not surprising that the velocity varies between individual tumors and even between different sites of normal skin [10].

It is a consistent phenomenon that ultrasound derived estimates of melanoma invasion depth are statistically higher than the histological figures [5, 3, 7]. Several factors have been implied in the overestimation of tumour thickness including tissue shrinkage [8] during formalin fixation, or the presence of nevus cells or subtumoral inflammatory cells, both of which cannot be distinguished from the melanoma sonographically [4]. In contrast to this, the loss of skin tension after excision of the tumour leads to an increase in tumour thickness, partly eliminating the effect of shrinkage during fixation of the tissue [8]. Histological measurement of tumour invasion depth underestimates the 'true' invasion depth when step sections have not been performed. In their study of melanoma specimens less than 0.75 mm thick, Solomon et al.[13] found the maximum tumour thickness to be 25 percent larger when serial sequences were investigated as compared to the routine processing method. Therefore, the overestimated tumour thickness values may also reflect the more precise detection of the thickest part of the tumour by real-time sonographic imaging.

While systematic overestimation is a statistical problem which can be compensated for by simple regression methods, accuracy is yet another problem in determining melanoma invasion depth sonographically. It has been correctly stated that even high correlation coefficients as given by several studies are not an appropriate measure of the usefulness of this method [5]. The strength of correlation between sonometry and histometry is influenced by different factors. First, all of the factors mentioned above not only have systematic influence but also contribute to the variation of sonometric or histometric measurements. Second, accuracy of both methods is limited. In histological measurement of invasion depth significant intra- and interobserver variation exists, revealing kappa statistics between only 0.49 and 0.76 among different pathologists [14, 15, 16].

Corresponding data for sonographic measurements of melanomas are lacking, but based on studies of normal skin[17] and sonographic distance measurements in other clinical fields [18], a significant interobserver variability has to be assumed. Additionally, as we could show, the variances of ultrasound velocities between different tumours are significantly higher than variances within different parts of an individual tumour. This means that there are differences in ultrasound velocity among individual tumours. These differences constitute another source of error in sonographic distance measurements, although the relative amount of variation seems to be quite low.

In summary, we could demonstrate that overestimation of the ultrasound velocity in primary melanoma tissue may be an additional reason for measurement bias in the prediction of melanoma thickness by high resolution sonography. Furthermore, our data show that tissue specific differences of ultrasound velocity may be a source of error in ultrasound determination of melanoma invasion depth, but with a maximal difference of about 1.5 percent from the mean value, the impact of this type of error seems to be limited.

As we also illustrated, both the overestimation of the mean ultrasound velocity as well as its variation in different tissue specimens may only explain a small part of the deviation from histological measurements. For a better quantification of tumour specific variation of ultrasound velocity in malignant melanoma further investigations of larger samples are needed. For clinical purposes however, it will still be appropriate to use a setting of 1580 m/s for the ultrasonography of human skin including the imaging of malignant melanoma. Based on these findings, the use of preoperative thickness measurement in malignant melanoma can be advocated, especially when the performance of SLND in dependence on tumor thickness is intended.