|
|
Background Controversy exists concerning the necessary margin of excision for cutaneous melanoma 2 mm or greater in thickness.
Methods We conducted a randomized clinical trial comparing 1-cm and 3-cm margins.
Results Of the 900 patients who were enrolled, 453 were randomly assigned to undergo surgery with a 1-cm margin of excision and 447 with a 3-cm margin of excision; the median follow-up was 60 months. A 1-cm margin of excision was associated with a significantly increased risk of locoregional recurrence. There were 168 locoregional recurrences (as first events) in the group with 1-cm margins of excision, as compared with 142 in the group with 3-cm margins (hazard ratio, 1.26; 95 percent confidence interval, 1.00 to 1.59; P=0.05). There were 128 deaths attributable to melanoma in the group with 1-cm margins, as compared with 105 in the group with 3-cm margins (hazard ratio, 1.24; 95 percent confidence interval, 0.96 to 1.61; P=0.1); overall survival was similar in the two groups (hazard ratio for death, 1.07; 95 percent confidence interval, 0.85 to 1.36; P=0.6).
Conclusions A 1-cm margin of excision for melanoma with a poor prognosis (as defined by a tumor thickness of at least 2 mm) is associated with a significantly greater risk of regional recurrence than is a 3-cm margin, but with a similar overall survival rate.
The width of the excision margin has been investigated in three randomized trials of patients with predominantly8 or exclusively9,10 thin tumors with a good prognosis. Wide and narrow margins were associated with similar rates of recurrence and survival, but these trials may have been too small to detect an adverse effect of a narrow margin owing to an insufficient number of events. The Intergroup Melanoma Surgical Trial8 found no difference in results between 2-cm margins and 4-cm margins for melanomas that were 1 to 4 mm thick, but the mean thickness of the melanomas was 1.96 mm, and only 207 patients (44 percent) had melanomas that were more than 2 mm thick. We conducted a multicenter clinical trial to investigate the effect of the margin of excision on the outcome in patients with high-risk malignant melanoma.
Methods
Study Design
The randomized trial was performed under the auspices of the United Kingdom Melanoma Study Group, the British Association of Plastic Surgeons, and the Scottish Cancer Therapy Network. It was coordinated by the Clinical Trials and Statistics Unit at the Institute of Cancer Research, Sutton, United Kingdom. Recruitment began in January 1993 and ended in July 2001. Eligible patients had a single, primary, localized cutaneous melanoma 2 mm or greater in thickness on the trunk or limbs (excluding the palms of the hands or the soles of the feet), where a 3-cm excision margin was technically possible. Patients had to be at least 18 years old and could not be pregnant. Patients who had a history of cancer (other than basal-cell carcinoma) or who were receiving immunosuppressive therapy were ineligible. Elective lymph-node dissection, sentinel-node biopsy, or adjuvant therapy was not allowed. Investigations other than chest radiography to determine the stage of disease were deemed unnecessary.
Participating surgeons chose one of two primary treatment approaches. The primary tumor could be excised before randomization, with either a 1-mm or a 1-cm margin to confirm the diagnosis and determine the thickness of the lesion. The patients were then randomly assigned to receive a 1-cm or 3-cm margin after the 1-mm primary excision or to receive no further treatment or an additional 2-cm margin after the 1-cm primary excision. The trial surgery was to be performed within 45 days after the primary excision, and all excisions were to extend to or include the deep fascia. The wound-closure techniques used were at the discretion of the surgeon.
Local recurrence was defined as a recurrence within 2 cm of the scar or graft. In-transit recurrence was defined as a recurrence from beyond the first 2 cm of the scar or graft to the regional nodes. All locoregional recurrences were detected clinically and confirmed by biopsy. The thickness of the tumor was measured at the referral centers before randomization. The presence or absence of ulceration was assessed by means of a centralized review. Data on the quality of life were collected from a sample of 426 patients and are reported separately.
Statistical Analysis
Randomization was performed by telephone at a central site and was stratified according to the center and the extent of primary surgery (1-mm initial margin of excision or 1-cm initial margin of excision). Permuted blocks of random size were used for randomization, and the allocation ratio was 1:1. All analyses were conducted according to the intention-to-treat principle. An independent data-monitoring committee regularly reviewed the trial results and reported its conclusions in a blinded fashion to the trial management group. The time to an event was measured from the date of randomization. The originally specified primary end points were the time to the first local or in-transit recurrence (the study had 90 percent power to detect an absolute increase in the rate from 10 percent to 20 percent at three years; 82 events were required) and regional lymph-node or distant metastasis (90 percent power to detect an absolute increase in the incidence from 40 percent to 55 percent at three years; 238 events required). A secondary end point was overall survival (90 percent power to detect an absolute difference of 15 percent in the mortality rate corresponding to a 40 percent rate, as compared with a 55 percent rate; 238 events required). The target enrollment was originally intended to be 600 patients.
We anticipated that the three-year rate of local or in-transit recurrence in the groups as a whole would be 15 percent, but it was found to be approximately half this figure. The data-monitoring committee and the trial management group therefore agreed that the sample size should be increased to 900 patients and the end points altered by combining the rates of local or in-transit recurrence with the rate of nodal recurrence into the single primary end point of locoregional recurrence. This decision coincided with the publication of the results of the Swedish Melanoma Study Group trial,9 which had combined local recurrence, in-transit recurrence, and regional-node metastases into the single end point of locoregional recurrence. The trial management group was unaware of the emerging results of our trial when the end points and the study size were changed. Locoregional recurrence and disease-free survival therefore became the primary end points. Since most first events in clinical practice are locoregional recurrences, these two primary end points are not independent events, and consequently no adjustment was made for multiple comparisons.
Life-table curves were calculated according to the Kaplan–Meier method,11 and the effect of prognostic factors was assessed with the use of Cox regression analysis.12 Tests for heterogeneity were used in all cases except those that examined the effect of tumor thickness, which was fitted as a linear trend. In the analysis of the time to locoregional recurrence, follow-up was truncated at the time of the first local, in-transit, or nodal recurrence (event) or distant metastasis (the censoring event) owing to the fact that only the first "nodal or distant" recurrence was recorded on the case-report form. Death without recurrence was counted as an event in the calculation of disease-free survival (for which the event was recurrence or death). Two-sided significance tests were used throughout.
The independent data-monitoring committee regularly reviewed the progress of the trial; the alpha spending function used a P value of less than 0.001 for interim analyses and a P value of 0.05 for the final analysis. Ten patients (1.1 percent) were lost to follow-up immediately after randomization. A P value for interaction of 0.01 was predefined as the level to be achieved before a subgroup result would be considered to override the overall treatment result.
The ethics committee at each center approved the study. All patients gave written or oral informed consent.
Results
Table 1 summarizes the features of the trial and characteristics of the patients. A total of 900 patients were enrolled, 453 of whom were assigned to a 1-cm margin of excision and 447 to a 3-cm margin of excision. Of these patients, 742 initially had a 1-mm margin of excision and 158 a 1-cm margin of excision. There were protocol deviations in 14.0 percent of the patients (Table 1); the majority were minor. The median tumor thickness was 3.0 mm in the group with 1-cm margins of excision and 3.1 mm in the group with 3-cm margins of excision. The site of the melanoma had the expected distribution within the two sexes. Among the men, 48.1 percent had tumors on the back and 18.1 percent had lower-limb tumors, whereas 18.4 percent of the women had tumors on the back and 52.9 percent had lower-limb tumors.
|
Table 2 lists the distribution of events, and Table 3 summarizes the end points. Univariate analysis showed that the overall relative rate of locoregional recurrence was 26 percent higher in the group with 1-cm margins of excision than in the group with 3-cm margins (hazard ratio, 1.26; 95 percent confidence interval, 1.00 to 1.59; P=0.05) (Table 3 and Table 4 and Figure 1A). The protocol-defined end points refer to event rates at three years, because most recurrences occur within this period; the hazard ratio for locoregional recurrence up to three years was 1.34 for the group with 1-cm margins of excision, as compared with the group with 3-cm margins of excision (95 percent confidence interval, 1.06 to 1.71; P=0.02); beyond three years it was 0.69 (95 percent confidence interval, 0.36 to 1.37; P=0.30), on the basis of only 36 events. The difference in disease-free survival approached statistical significance, but the largest component of the combined end point was locoregional recurrence (hazard ratio for death or recurrence for the group with 1-cm margins as compared with the group with 3-cm margins, 1.21; 95 percent confidence interval, 0.99 to 1.46; P=0.06) (Table 3 and Figure 1B).
|
|
|
|
A total of 769 patients had complete data on all the factors included in the multivariate analysis; evaluation of ulceration was possible in 780 patients, the tumor site was unclassifiable in 5 patients, 4 patients lacked follow-up data, and in 2 patients the tumor thickness at the time of randomization was unknown. The effect of a 1-cm margin width, as compared with a 3-cm margin width, was increased by multivariate analysis adjusting for known prognostic factors (hazard ratio for locoregional recurrence, 1.34; 95 percent confidence interval, 1.06 to 1.70; P=0.02) (Table 5). Similar hazard ratios were obtained when all patients with protocol deviations (Table 1) were excluded from the analysis (data not shown). Table 5 shows that male sex, increasing tumor thickness, and ulceration were each independent risk factors for all end points and that a distal-limb site was an indicator of a good prognosis with respect to melanoma-specific survival and, possibly, overall survival.
|
We anticipated that patients with a recurrence after the more extensive surgery would have more aggressive disease than those with a recurrence after less extensive surgery. Examination of overall survival after locoregional recurrence in the two groups supported this hypothesis: the overall survival rate after locoregional recurrence was higher in the group with 1-cm margins of excision than in the group with 3-cm margins of excision (hazard ratio for death, 0.75; 95 percent confidence interval, 0.56 to 1.00; P=0.05). Median survival after locoregional recurrence was 27.9 months in the group with 1-cm margins of excision and 18.5 months in the group with 3-cm margins of excision.
Discussion
In this randomized trial of the surgical treatment of high-risk melanoma, we found a greater risk of locoregional recurrence when melanomas that were at least 2 mm thick were excised with a 1-cm margin, rather than a 3-cm margin. A result of similar magnitude was reported by the Swedish Melanoma Study Group,9 which compared 2-cm and 5-cm margins of excision for tumors with a depth of 0.8 to 2.0 mm. The hazard ratio for locoregional recurrence in our trial — 1.26 (95 percent confidence interval, 1.00 to 1.59; P=0.05) — was marginally larger when we used multivariate analysis to adjust for known prognostic factors; this was also the case when the analysis was restricted to the first three years of follow-up. We have presented two-sided P values, but it is implausible that a narrower margin would decrease the risk of locoregional recurrence. Taking all these factors together, we do not believe that the outcome of this trial is simply a chance finding. An interaction of margin width with sex just failed to reach statistical significance, but to our knowledge, such an interaction has never been reported before and should therefore be regarded as a hypothesis requiring further investigation.
Since approximately 40 percent of patients with tumors of this thickness ultimately die of melanoma,2 the expected total number of deaths from melanoma in our trial should be 360. With this number of deaths, it would be possible to detect an increase in the death rate by a factor of 1.4 with 90 percent statistical power. Since margin width most likely has less of an effect on melanoma-specific deaths than on locoregional recurrence (i.e., less than our estimated hazard ratio of 1.26), it may never be possible to reach a clear conclusion about the effect of the width of the excision margin on survival on the basis of our results alone. Although we did not find a statistically significant effect of margin width on melanoma-specific survival, the hazard ratio for death from melanoma was 1.24 for a 1-cm margin of excision, as compared with a 3-cm margin of excision (95 percent confidence interval, 0.96 to 1.61; P=0.1).
The effect on melanoma-specific survival in the Swedish Melanoma Study Group trial9 was similar, with a hazard ratio of 1.22 for a 2-cm margin of excision, as compared with a 5-cm margin of excision (95 percent confidence interval, 0.88 to 1.69; P=0.24). In addition, the hazard ratio for death from melanoma in the Intergroup Melanoma Surgical Trial8 was approximately 1.36 in the group with narrow margins of excision, as compared with the group with wide margins (95 percent confidence interval, 0.97 to 1.92; P=0.07). We based this estimate on 10-year survival rates of 70 percent in the narrow-margin group and 77 percent in the wide-margin group and on the P value for the difference between the two groups. Our review of these three trials suggests that there is a significant increase in the risk of death from melanoma associated with a narrow margin of excision, as compared with a wide margin (hazard ratio, 1.26; 95 percent confidence interval, 1.06 to 1.50; P=0.008). In both our study and that of the Swedish Melanoma Study Group,9 there were fewer deaths from other causes in the group with narrow margins of excision, but this difference was not statistically significant.
The necessary margin of excision is controversial,13 but it is clear that inadequate margins can increase the risk of locoregional recurrence and, hence, may be associated with increased mortality.8,14 This issue is crucial, because doubts remain about the efficacy of adjuvant treatment with interferon.15,16 Interferon increases the disease-free interval but probably not overall survival.17 If we had allowed sentinel–lymph-node biopsy in our trial, the pattern of recurrence would have differed, but the extent of the difference between the two groups and its effect on the outcome of the trial are a matter of conjecture. The most likely outcome would have been an increased incidence of local or in-transit recurrence, as recently reported in patients with a positive sentinel node18,19 and as was found after elective lymph-node dissection.20,21,22,23 Our findings confirm that metastasis is an ongoing phenomenon over time. Melanoma cells may not have reached the sentinel node by the time sentinel–lymph-node biopsy is performed, which may partly explain the false negative results that are obtained.
Which margin should be recommended? Many national and international guidelines7,24 support the use of 2-cm margins as a result of the Intergroup Melanoma Surgical Trial,8 whereas the Australian guidelines25 allow the use of a minimal 1-cm margin for tumors between 2 and 4 mm in thickness. Our findings suggest that, in a small number of patients, the melanoma cells that remain after excision with a 1-cm margin will prove fatal. We therefore recommend that the use of a 1-cm margin should be avoided in patients with melanomas that are at least 2 mm thick. Our study provides no data to support the preferred use of either a 2-cm or 3-cm margin, and it would seem reasonable, pending further data, for the patient to make the choice after an informed discussion of the surgical options. It may be prudent to use a 3-cm margin of excision in patients with deeper tumors (those more than 4 mm thick), because such patients have a higher risk of locoregional recurrence.
Funded by North Thames National Health Service Executive Research and Development, the Northern and Yorkshire National Health Service Executive; the Imperial Cancer Research Fund (now Cancer Research UK), the Cancer Research Campaign (now Cancer Research UK), which provides funding to the Institute of Cancer Research Clinical Trials and Statistics Unit, the British United Provident Association Foundation; the British Association of Plastic Surgeons, and the Meirion Thomas Cancer Research Trust.
We are indebted to Professor Julian Peto at the Institute of Cancer Research, Sutton, Surrey, United Kingdom; to the members of the data-monitoring committee for giving their time to ensure proper oversight of the accumulating data; and to the 900 patients who kindly agreed to enter the trial.
Source Information
From the Royal Marsden Hospital National Health Service Trust, London (J.M.T., R.A.); the Division of Genetic Epidemiology, Cancer Research UK, Clinical Center, Leeds, Yorkshire (J.N.-B.); the Institute of Cancer Research, Sutton, Surrey (G.C., J.M.B.); Bradford Royal Infirmary, Bradford, Yorkshire (M.T.); Nuffield Hospital, Plymouth, Devon (J.E.); Royal Surrey Hospital, Guildford, Surrey (M.C.); Southampton General Hospital, Southampton, Hampshire (J.T.); Broomfield Hospital, Colchester, Essex (M.F.); and Norfolk and Norwich Hospital, Norwich, Norfolk (T.O.) — all in the United Kingdom; and the Maria Sklodowska-Curie Memorial Cancer Center, Warsaw, Poland (W.R.).
Address reprint requests to Mr. Thomas at the Royal Marsden NHS Trust, Fulham Rd., London SW3 7JJ, United Kingdom.
References
The following institutions and investigators (all from the United Kingdom unless otherwise specified) were the main participants in the trial (the number of randomized patients from each institution is given in parentheses): Marie Sklodowska-Curìe Memorial Cancer Center, Warsaw, Poland — W. Ruka, Z. Nowecki (108); Norfolk and Norwich University Hospitals, Norwich, Norfolk — T. O'Neill, A. Bardsley, A. Logan, M. Meyer (86); Derriford Hospital, Plymouth, Devon — J. Evans, D. Hanley, D. Harris (75); University Hospital, Sunderland, Durham — R.B. Berry, M. Erdmann, J. James, J. Langtry, G. Rao (54); Mount Vernon Hospital, Northwood, Middlesex — B. Morgan, P. Cussons, D. Gault, A. Grobellaar, D. Harrison, D. Ross, P. Sanders, P. Smith (46); Royal Victoria Infirmary, Newcastle upon Tyne — M. Black, P. Hodgkinson, C. Lawrence, M. Dahl, R. Milner, S. Pape (39); Radcliffe Infirmary, Oxford — D. Coleman, T. Goodacre, H. Giele, S. Wall (39); Frenchay Hospital, Bristol — J. Kenealey, N. Mercer, P. Townsend, D.A. Burd (34); Whiston Hospital, Prescot, Merseyside — R. Curley, R. Green, K. Graham, J. Bryson, K. Hancock, M. James (33); Addenbrooke's Hospital, Cambridge — P. Hall, B. Lamberty, G. Cormack (29); Nottingham University Hospital (Queens Medical Centre), Nottingham — A. Perks, J. Daly (29); Royal Devon and Exeter Hospital, Exeter, Devon — J. Palmer, V. Deveraj (29); Royal Marsden Hospital, London — J.M. Thomas (20); Chelsea and Westminster Hospital, London — J.M. Thomas (20); Brighton General Hospital, Brighton, Sussex — C. Darley, M. Price, P. Hale (19); Glan Clwyd, District General Hospital, Rhyl, Wales — C. Davies (15); Southampton University Hospitals, Southampton, Hampshire — J. Smallwood (14); Ulster Hospital, Dundonald, Belfast — M. Brennan, D. Gordon, A. Leonard, A. Small (13); Salisbury District Hospital, Salisbury, Wiltshire — R. McDowall, M. Cadier, J. Hobby, D. McNeill, L. Rossi (12); Stoke Mandeville Hospital, Aylesbury, Buckinghamshire — A. Heywood, P. Budny (12); Frimley Park Hospital, Frimley, Surrey — R. Lallemand, I. Laidlaw (11); Wrexham Maelor Hospital, Wrexham — K. Crumplin, J. Sowden, J. Pye, P. Richards, M. Rosenberg (9); St. George's Hospital, Tooting, London — B. Powell (9); Kingston Hospital, Kingston on Thames, Surrey — C. Cahill, R. Leach (8); the Scottish Cancer Therapy Network — U. Chetty, M. Davies, J. Holmes, A.M. Morris, A. Nassan, J.H. Stevenson, A.D. Wilmhurst (44).
| |||||||||||||||||||||||||||||||||||||||||||||||||||||
Related Letters:
Excision Margins in High-Risk Malignant Melanoma
Hellman S., Hurt M. A., Thomas J. M., A'Hern R., Newton-Bishop J.
Extract |
Full Text |
PDF
N Engl J Med 2004;
350:2305-2306, May 27, 2004.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
