Lung metastases from melanoma: when is surgical treatment warranted?

Surgical treatment of lung metastases from melanoma is highly controversial as the expected outcome is much poorer than for other primary tumours and a reliable system for selecting patients is lacking. This study evaluated the long-term results of lung metastasectomy for melanoma, with the aim of defining a subset of patients with better prognosis. By reviewing the data of the International Registry of Lung Metastases (IRLM), we identified 328 patients who underwent lung metastasectomy for melanoma in the period 1945–1995. Survival was calculated by Kaplan–Meier estimate, using log-rank test and Cox regression model for statistical analysis. After complete pulmonary metastasectomy (282 patients) the 5- and 10-year survival was 22% and 16%, respectively. In this group of patients, a time to pulmonary metastases (TPM) shorter than 36 months or the presence of multiple metastases were independent unfavourable prognostic factors. There were no long-term survivors after incomplete resection (46 patients, P< 0.01). Using the IRLM grouping system, patients without risk factors (TPM > 36 months and single lesion) experienced the best survival (29% at 5 years), followed by those with one risk factor only (20% at 5 years). On the other hand, those with two risk factors or incomplete resection showed a significantly poorer survival (7% and 0% at 5 years). Surgery plays an important role in carefully selected cases of pulmonary metastatic melanoma. The prognostic grouping system proposed by the International Registry of Lung Metastases provides a simple and effective method for improving the selection of surgical candidates. © 2000 Cancer Research Campaign

number of pulmonary metastasectomy operations per patient. Differences were tested by the log-rank test. A Cox regression model evaluated all variables with statistically significant impact on survival.
Finally, the prognostic grouping proposed by the International Registry of Lung Metastases in 1997 was tested to assess its value in patients with metastatic melanoma.

Patient details
One hundred and eighty-one patients were male (55.2%) and 147 were female (44.8%) with a mean age of 49 years (range 14-78). Sixty-eight patients (20.7%) had adjuvant therapy in addition to surgical resection of the primary melanoma (chemotherapy with or without radiotherapy in 59 patients, radiotherapy in nine patients). Before metastasectomy, 73 patients (22.2%) required further treatment for recurrence of melanoma at the primary site (51 for single recurrence and 22 for multiple recurrence).
Mean TPM was 53 months for the whole series, 54 months for radical metastasectomies and 42 months for incomplete resections. Fifteen patients had one or more lung metastases synchronous with primary tumour, 40 patients had a TPM < 12 months, 102 had a TPM of 12-35 months, 163 patients (49.7%) had a TPM > 36 months (Table 1). Information about TPM was not available for eight patients (2.4%).
Information about the number of lesions radiologically evident was available in 235 patients: one lesion only was evident in 153 cases (65.1%), multiple lesions were evident in 82 (34.9%).
On pathological assessment, single metastases were present in 197 patients (60%) and multiple metastases in 131 (40%). Fiftytwo patients had four or more metastases. Of the 153 patients with a solitary radiological lesion, 128 had only one pathologically proven metastasis (the accuracy of radiological assessment for single vs multiple lesions was 84%). Evaluation of the hilar or mediastinal nodes resulted positive in 26 patients (8%). Complete resection of all melanoma metastases was performed in 282 patients (86% radical metastasectomy group); 46 patients had a non-radical metastasectomy (14%). Four patients died in the postoperative period (1.2%). Adjuvant chemotherapy was used in 107 cases (32.6%), prior to surgery in 47 patients and post surgery in 60. Overall, 300 patients had one metastasectomy operation, 23 had two metastasectomies, and five had three metastasectomies.
In the R0 group, patients with a TPM of > 36 months had a better long-term survival than patients with a TPM < 36 months (30% vs 15% at 5 years and 22% vs 11% at 10 years, P < 0.01).
The number of pathologically proven metastases also affected survival: patients with a single lesion had better survival than those with multiple lesions (25% vs 19% at 5 years and 22% vs 7% at 10 years, P = 0.03). Prognosis was significantly poorer with four or more metastases (8% at 5 years).
No effect on survival was demonstrated for age, sex, surgical approach, type of lung resection, extrapulmonary resection, chemotherapy before or after metastasectomy, number of metastasectomies, hilar-or mediastinal-node involvement.
Postponing metastasectomy did not aid the selection of patients with better prognosis. In fact the 5-year survival of patients operated within a month of detection (170 patients) was similar to the survival of those operated upon later (158 patients), 20% vs 26% respectively (median survival 30 vs 22 months, P = 0.3).

Pattern of failure
During the follow-up period, 221 patients died (crude mortality rate 67%), 184 in the R0 group (crude mortality rate 65%) and 37 in the R1-2 group (mortality rate 80%). Mean time from metastasectomy to death was 18 months (19 months for R0 patients and 9 months for R1-2 patients). One hundred and eighty patients (63.8%) experienced recurrence of melanoma after radical pulmonary metastasectomy: intrathoracic recurrence in 48 cases, intra-and extrathoracic in 50 cases and extrathoracic relapse alone in 82 cases. The most common sites of extrathoracic disease were brain (58 patients) and liver (15 patients).
Of 48 patients with intrathoracic recurrence, further surgery was possible in 28 (58.3%). Twenty-three such patients had a second metastasectomy, five had three metastasectomies. Their long-term survival was 19% at 5 years.

Prognostic grouping
The three significant prognostic factors from multivariate analysis (radicality, TPM, number of lesions) were used to design a system of classification according to the proposal of the International Registry of Lung Metastases (Pastorino et al, 1997). Following these guidelines, the 328 patients of the study were divided into four groups: • Group I: patients who had radical metastasectomy and no adverse prognostic factors (TPM > 36 months and single metastasis, n = 96) • Group II: radical metastasectomy and one unfavourable factor (TPM < 36 months or multiple metastases, n = 126) • Group III: radical metastasectomy and two unfavourable factors (TPM < 36 months and multiple metastases, n = 52) • Group IV: incomplete metastasectomy (n = 46).
The survival difference among these groups was highly significant (log-rank 2 29.6, df 3, Figure 2). Group I survival was 29% at 5 years and 26% at 10 years. Group II survival was 20% at 5 years and 11% at 10 years. Group III survival was 7% at 5 years. Group IV had no 5-year survivors.

DISCUSSION
Metastatic melanoma has a very poor prognosis, with a median survival of about 9 months (Falkson, 1998). For some of these patients with limited metastatic disease, surgical resection can offer the best chance of cure (Fletcher, 1998). This was confirmed even in the subset of pulmonary metastatic melanoma in two large series previously published (Tafra et al, 1995;Harpole et al, 1992). However, pulmonary metastasectomy for melanoma has the lowest long-term survival after surgery compared to germ cell tumours, carcinomas and sarcomas (Pastorino et al, 1997). This peculiarity emphasizes the importance of prognostic indicators to avoid surgery in patients who will not benefit. When the lung is the only site of recurrence, radicality of pulmonary metastasectomy is the main factor affecting survival (Harpole et al, 1992;Tafra et al, 1995;Pastorino et al, 1997).
We tested many variables on the IRLM series, which represents the largest group of patients with surgically treated pulmonary metastatic melanoma available in the literature, in the search for independent prognostic predictors after lung metastasectomy. In particular, the grouping system proposed by the International Registry of Lung Metastases in 1997 was tested, taking into account the radicality of resection and two more risk factors: number of pathologically proven lesions and time elapsed from primary tumour to pulmonary metastases.
Having a single lesion and a long time to pulmonary metastases means a better survival in surgically treated patients with pulmonary metastases from melanoma (Harpole et al, 1992); it was suggested by Tafra et al (1995) and is confirmed by our study. Applying the IRLM grouping system, long-term results remain satisfactory at 10 years when a radical resection is achieved and only one risk factor is present (Groups I and II). On the contrary, metastasectomy is not effective when multiple metastases are associated with a TPM shorter than 3 years.
Given the fact that the proposed grouping system has a good prognostic value when applied retrospectively, it would seem logical to use a similar system to select patients with pulmonary metastases from melanoma for surgery.
Preoperative evaluation of the number of lesions is crucial. In the Registry experience, the accuracy of radiological assessment in detection of single lesions was good (84%). CT scan was not routinely used in the early period of the series but accuracy remains very similar throughout (considering only the last 10 years, period 1985-1995: accuracy 86%). It is expected that spiral CT scan and possibly PET will improve the diagnostic power of radiological assessment (Holder et al, 1998). Applying the IRLM grouping system prospectively, it can be argued that limiting surgery for pulmonary metastases to Groups I and II would produce better results in melanoma patients.
Other factors indicated in the literature as important prognostic indicators were considered in this study. Chemotherapy, which is reportedly associated with better survival in pulmonary metastatic melanoma (Harpole et al, 1992), did not influence survival after surgery in the IRLM series. This conclusion is greatly limited by the retrospective nature of the IRLM data because information on inclusion criteria, regimen used and response rates are missing. Moreover, the role of immunotherapy, reported to be effective as a post-surgery adjuvant treatment (Tafra et al, 1995), was not verified by this study for the same reason.