Regression is known to be a progressive inflammatory process in many melanomas.1 2 It is characterised by a lymphocytic infiltrate among the tumour cells separating the melanoma into small groups with individual tumour cell necrosis. This is followed by obvious scarring in the dermis with proliferation of fibroblasts and neo-vascularisation, which cannot be explained by a previous biopsy. Subsequently the lymphocytic infiltrate gradually diminishes and the tumour cells become sparse or disappear completely. Often the only sign of previous dermal tumour in one segment of a melanoma is residual pigment within an area of fibrosis.

Regression in primary melanoma may be segmental or total. While segmental spontaneous regression is a relatively common finding,3 complete spontaneous regression of the whole lesion is rare, with an estimated incidence of 0.22–0.27%.4 The prognostic significance of regression in primary melanoma has been in dispute for many years. Some studies indicate that regression in primary melanomas does not increase the risk of metastasis,5^–7 while other reports have suggested that regression is associated with a relatively poor prognosis.8^–10 Despite this uncertainty, it has been used as a component of the AFIP prognostic tables,11 and has been proposed as an explanation for metastasis in some thin metastasising melanomas.12 13 The presence of regression in thin melanomas has been quoted as justification for a sentinel lymph node biopsy (SLNB).14^–17

The variation in interpretation of the significance of regression has been attributed to lack of uniformity in its definition.18 We have therefore attempted to further understand the significance of regression in primary malignant melanoma by analysing a consecutive series of patients with a histological diagnosis of primary malignant melanoma who subsequently underwent SLNB. In order to overcome the possible confusion of definition we have created a definition of regression in three stages. We have not included those cases with a lymphocytic host response and individual cell necrosis, which is probably the earliest form of regression but which is subject to variable interpretation. The system also avoids potential confusion between the stage of the regression and the proportion of the tumour which is involved, by not using terms such as partial and complete, early and late, that are not always clearly understood.

We assessed the stage and extent of regression, and correlated this with the presence or absence of metastasis in the SLNB. The aim was to provide further information on whether patients should be subjected to SLNB on the basis of regression. In addition we used sentinel lymph node (SLN) status, which is known as a powerful surrogate for prognosis,16 19^–21 to assess the effect of regression in primary melanoma on the likely survival of the patients.

METHODS

We reviewed 146 consecutive melanoma cases with SLNB, for which the histology of the primary melanoma was available for review, submitted to the histopathology Departments at the Royal Surrey County Hospital, Guildford, and St George’s Hospital, London, during 1999–2005. Prior to this time SLNB had been offered to patients with melanoma ⩾1 mm in thickness or Clark’s level IV. For the duration of the study, the procedure was also offered to patients with melanoma <1 mm who had regression. The melanomas were assessed for age, gender, anatomic site, growth phase, Breslow thickness, Clark’s level, presence or absence of ulceration, mitotic count, tumour infiltrating lymphocytes (TIL), and regression. All cases were first assessed by one of the first two authors, and then reviewed by the senior author (MGC). Those melanomas without regression acted as a control group.

Regression was said to be present only if a segment of melanoma appeared to be lost and replaced by fibrosis, with increased vascularity and melanophages, and a lymphocytic infiltrate of variable intensity.22 All slides of each case were reviewed. The total horizontal extent of the melanoma and the total horizontal extent of the regression as represented on the slides were measured. The percentage of regression with respect to the total length of melanoma was then estimated to the nearest 5%. The thickness of regression was measured from the granular layer to the deepest part of the dense fibrosis. This is included as an attempt to reflect the possible thickness of the component of the melanoma that has regressed. This is problematic since it requires distinction from dermal collagen and may also undergo shrinkage so that the estimate may be inaccurate.

Regression was classified into three stages in order to cover the possibility of differences of definition being responsible for the known variability in interpretation of significance of regression.18

• Stage 1 included cases where regression was noted, when some of the dermal component of the melanoma was lost, but leaving viable groups of melanoma cells and all of the epidermal melanocytic component still present (fig 1).

• Stage 2 included cases where all the dermal component of the melanoma was lost in the segment of regression, but there was a residual junctional component (fig 2).

• Stage 3 included cases where both the junctional and dermal components of the melanoma were lost in the involved segment, leaving only dermal scarring (fig 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Regression stage 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Regression stage 2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 Regression stage 3.

The criteria for offering SLNB were that the primary melanomas were either ⩾1 mm in thickness, and/or extending to Clark’s level IV, or for the duration of the study, that regression was present regardless of the thickness of the primary lesion or the growth phase. The SLN were processed according to the standard EORTC protocol23 for the presence or absence of metastasis. The micro-anatomic site and size of the nodal metastases were also noted.

The data collected for each of the parameters were then analysed by the Pearson χ² test, the likelihood ratio χ² test, and the t-test in order to ascertain the statistical significance of the results.

Patients were followed up to determine disease free and overall survival and to correlate this with sentinel node biopsy results.

RESULTS

Of the 146 cases analysed (table 1), 129 did not have nodal metastasis and 17 had positive sentinel nodes. Of the node negative cases, 63 cases did not show regression while 66 cases had regression in the primary lesion. There was a statistically significantly greater proportion (p = 0.028) of individuals without regression (table 1) who are sentinel lymph node positive (13/76) compared with those showing regression (4/70). A p value around 0.03 in a two-sided test is evidence of a difference, but is not strong evidence; it is just above the range that would be considered borderline.

Metastatic disease was found to correlate with growth phase of the primary lesion. All the 17 node positive cases were in the vertical growth phase, while none of the cases in radial growth phase and showing regression were associated with nodal metastasis (p = 0.029).

Table 2 shows the correlation of the regression stage in the primary lesion with the node status. None of the stage 1 and 2 regression cases showed metastasis, but there were 4 stage 3 cases who were SLN positive (4/41). These four node positive cases with stage 3 regression also showed residual melanoma in VGP with thickness >1.5 mm. It seems likely this residual melanoma is the source of the metastases, although some value for staging of regression with p = 0.026 cannot be entirely dismissed.

Other parameters of regression which did not correlate with sentinel lymph node status were the thickness of regression (p = 0.028) and the proportion of primary lesion involved with regression (p = 0.145).

The Breslow thickness of the primary melanoma in the cases with regression ranged from 0.26 mm to 9.0 mm (mean 2.02 mm); the Breslow thickness of the primary lesions without regression ranged from 0.20 mm to 5.3 mm (mean 1.23 mm).

Of our total of 146 cases, 62 had melanomas with thickness <1 mm or were in radial growth phase, yet were offered sentinel node biopsy because of regression. Of these 62 cases, 44 showed features of regression and all were node negative. The remaining 16/62 cases which did not show regression comprised 9 melanomas extending to level IV, and 7 thin melanomas which were originally reported to have regression, but on review by us were reclassified as not showing features of regression. Two of these 16 cases of thin (<1.0 mm) melanomas without regression had sentinel node metastasis (table 3).

The correlation of age, sex, site and presence of tumour infiltrating lymphocytes (TIL) with regression and sentinel node status was not statistically significant in our study.

Follow-up data were available for 116 of our 146 cases (79%), 30 patients being lost to clinical follow-up. The follow-up period ranged from 1 year to 6 years. At the end of follow-up, 94% of the patients (109/116) were alive and well. Two (1.7%) patients with thick melanomas without features of regression had died during follow-up period because of the disease. Two patients had died due to unrelated causes.

Three of 116 patients (2.5%) had developed additional metastatic disease. None of these showed regression in the primary lesion; two of these three cases were associated with thick melanomas and nodal metastasis. Only one patient with negative SLNB had developed subsequent metastasis.

DISCUSSION

The presence of regression in melanomas has been investigated in the past by several authors, but its effect on prognosis is disputed. A controversy exists regarding the favourable effect versus a negligible versus a detrimental effect of regression on patient prognosis. It has been presumed that regression in a melanoma which is in radial growth phase might have destroyed a vertical growth phase component. On this basis the regression could cause an underestimate of the metastatic potential of the melanoma as a whole, and has been at least part of the explanation of why regression in melanoma might have an adverse effect on prognosis. The regressive process might lead to the destruction of all the dermal component of the melanoma so that the lesion appears as melanoma in situ with evidence of dermal regression. No examples of this were included in the study.

Sentinel lymph node biopsy is a surrogate marker of patient prognosis and has been established as the most reliable prognostic factor, superior to thickness.16 17 24 25 We have used sentinel node biopsy as a measure of likely patient prognosis.

We did not assess the sentinel lymph nodes in this series of cases using RT-PCR for tyrosinase or other probes. The findings are based on histological observations of H&E and immunohistochemistry stained slides only. Occasionally scattered melanophages were seen, but in none of this series were melanophages focally arranged with fibrosis, which would raise the possibility of tumour destruction by an immunological response within the lymph node. This seems not likely to have occurred in the lymph nodes in this series, although that possibility cannot be completely excluded.

In our study, there is a trend (p = 0.028) for cases with regression to be associated with fewer nodal metastases compared with melanomas which did not show regression. This suggests that regression is if anything a favourable feature and certainly not unfavourable. There were no cases where regression was the only explanation for metastasis. The four primary melanomas with regression which were node positive had a residual thick component; since thickness is a known strong prognostic factor,26 27 it is this part of those melanomas that is likely to be the primary source of the nodal metastasis.

None of the thin melanomas with regression, including the melanomas in radial growth phase, who were offered sentinel node biopsy because of associated regression, were found to have metastasis in SLN. The two cases of thin melanomas (2/62) which had sentinel node metastases did not show regression. In both these cases the melanoma extended to Clark’s level IV and the mitotic counts were 2 and 3/mm2. Thus our findings suggest that regression did not contribute towards metastatic potential in these melanomas.

These observations were supported by our patient follow-up studies, which revealed that all the cases which had metastatic disease or had died due to melanoma, did not show features of regression in the primary lesion.

On the other hand, previous studies point to regression as a possible explanation of metastasis in 50% of those rare cases of thin melanomas, which have metastasised,2 12 13 28^–30 suggesting that despite the overall favourable effect of regression on the metastatic , occasional melanomas will metastasise before or in spite of the process of regression. This appears to be very rare and not sufficiently common to justify performing SLNB on thin regressed melanomas.

SUMMARY

Regression is an immune phenomenon associated with spontaneous disappearance of cancer in many malignancies including melanomas. Its significance in melanomas is not yet established. Regression has been used as a justification for sentinel lymph node biopsy in thin melanomas because of the presumption of adverse effect on prognosis. We, in contrast, have concluded that regression is usually a favourable process, particularly in thin melanomas. This supports what might be expected intuitively considering its biological effect on the tumour cells. The underlying reasons for the discrepancy between our results and previous studies in thin melanomas could be further investigated. We have suggested that metastasis in “thin melanomas showing regression” is real but rare. Variant vertical growth phase and mitoses13 may be more important predictors of metastatic potential in thin melanomas. In the vast majority of melanomas, particularly thin ones, the effect of regression appears to be favourable.