Systematic review of systemic adjuvant therapy for patients at high risk for recurrent melanoma



The authors examined the role of systemic adjuvant therapy in patients with high-risk, resected, primary melanoma. Outcomes of interest included overall survival, disease-free survival, adverse effects, and quality of life. A systematic review of the literature was conducted to locate randomized controlled trials, practice guidelines, meta-analyses, and reviews published between 1980 and 2004. Thirty-seven randomized controlled trials, 2 meta-analyses, and 1 systematic review were identified that investigated interferon, levamisole, vaccine, or chemotherapy as adjuvant therapy. For high-dose interferon-α, the results from 3 randomized trials conducted by the Eastern Cooperative Oncology Group were pooled, and a meta-analysis of 2-year death rates yielded a risk ratio of 0.85 (95% confidence interval, 0.73–0.99; P = .03). Five randomized trials comparing low-dose interferon-α with observation only after surgery did not detect a statistically significant improvement in overall survival. A meta-analysis of 4 levamisole trials did not demonstrate a significant survival benefit for levamisole over control; similarly, no survival benefit was demonstrated by data from randomized controlled trials with vaccines (9 trials) or with chemotherapy (10 trials). In this review of the available literature, no systemic adjuvant therapy was identified that conferred a significant overall survival benefit in patients with high-risk, resected, primary melanoma. However, high-dose interferon should be considered in the treatment of these patients, because such therapy is associated with a significant improvement in disease-free survival and a reduction in 2-year mortality. Until the results of ongoing trials are available, the authors could not state with confidence whether such therapy benefits patients with microscopically detected, sentinel lymph node-positive disease. Cancer 2006. © 2006 American Cancer Society.

The incidence and mortality of malignant melanoma continue to rise in Canada.1 Appropriate surgical management remains the mainstay of therapy for patients with early-stage, nonmetastatic disease; however, rising mortality rates clearly indicate that effective therapies desperately are required both as adjuvant treatment and for patients with metastatic melanoma. The observation that melanoma lesions spontaneously regress more frequently than other neoplasms has motivated inquiries into immunomodulatory treatments.2–4 However, the results from trials that have incorporated such treatments have been inconsistent and have generated much debate.5

Similarly, a number of randomized trials exploring the use of high-dose interferon α (HDI-α) in the adjuvant setting have been reported.6–9 Although the initial results had a substantial impact on oncologic practices, the subsequently published, mature data generated considerable controversy. The literature is rich in wide-ranging opinions and reviews on this subject—some regard immunomodulatory treatment as standard therapy, others do not.

Given those findings, the Melanoma Disease Site Group (DSG) decided to examine the available evidence with the express intent of developing an evidence-based clinical practice guideline addressing the question: what systemic therapy should clinicians recommend to patients who have been rendered disease-free after resection of cutaneous melanomas and who are at high risk for subsequent recurrence? Outcomes of interest included overall survival (OS), recurrence-free survival (RFS), adverse effects, and quality of life (QOL).

High-risk patients were defined as those in the following clinical states who had been surgically rendered disease free: primary melanoma with tumor thickness ≥ 4.0 mm or Level V invasion, primary melanoma with in-transit metastases, primary melanoma with regional lymph node metastases that are apparent clinically or are detected at sentinel lymph-node dissection, regional lymph node recurrence, or involved lymph nodes excised but no known primary melanoma. The target population also included patients who currently would be classified as American Joint Committee on Cancer (AJCC) Stage IIB, IIC, and III melanoma.

This report summarizes the available evidence on all therapeutic modalities that have been evaluated in this setting and makes reference to the interpretation and consensus developed for the practice guideline. The DSG has endeavored to place HDI-α in its rightful context. The complete practice guideline report is available at URL: [accessed February 2006] and will be updated when warranted.


The MEDLINE, CANCERLIT, EMBASE, and Cochrane Library data bases were searched from 1980 to 2004, and variations on the following search terms were used: melanoma, clinical trial, random, and adjuvant. The American Society of Clinical Oncology annual meeting proceedings were searched from 1996 to 2004 for reports of new or ongoing trials. A search also was conducted for published practice guidelines, meta-analyses, and reviews.

Inclusion Criteria

Articles were selected for inclusion in this systematic review only if they were randomized controlled trials (RCTs) of systemic therapies (levamisole, interferon, vaccines, or chemotherapy) for the adjuvant treatment of patients with melanoma. Practice guidelines, meta-analyses, and systematic reviews of the adjuvant treatment of malignant melanoma also were eligible for review. We excluded Phase I or Phase II trials, non-English studies, or studies that involved bacillus Calmette–Guerin, Corynebacterium parvum, transfer factor, vitamin A, or megestrol acetate.

Synthesizing the Evidence

The majority of the trials that were selected for the current analysis addressed a common question: does the therapy under investigation, when given as adjuvant treatment, improve survival compared with no treatment? Similar patient groups, albeit with varying risks of recurrence by virtue of entry criteria, participated in the randomized trials. Few trials were restricted to patients at high risk of recurrence (i.e., lesion depth ≥ 4.0 mm or completely resected regional lymph node metastases). Trials that enrolled patients with a range of risks did not report survival results separately. The treatments evaluated comprised four distinct groups of interventions: interferons, levamisole, vaccines, and chemotherapy.

In contrast to a published systematic review by Lens and Dawes,10 we pooled the results from three published Eastern Cooperative Oncology Group (ECOG) trials of HDI-α therapy.6–8 Our decision to pool results both with and without data from the ECOG 1694 trial,8 which used a vaccine as the intervention arm and interferon as the control arm, is addressed below (see Discussion). Results were pooled across studies by using the Review Manager software (RevMan 4.1; provided by the Cochrane Collaboration; Metaview© Update Software available at URL: [accessed February 2006]). Pooled results are expressed as relative risks (also known as risk ratios) for mortality (with 95% confidence intervals [95% CI]), with a relative risk < 1.0 favoring the active treatment group. For data analysis, we used the random effects model.11 All significance tests were two-sided. Ideally, a meta-analysis would be restricted to high-risk patients, as outlined above, but most studies were not limited to that group of patients. We also pooled mortality data for two of the three other therapies (levamisole and chemotherapy) but were unable to do so for trials that evaluated vaccines.



To our knowledge, with the exception of the Southwest Oncology Group (SWOG) study of interferon γ12 and the European Organization for the Research and Treatment of Cancer (EORTC) trial,13 all other trials investigated the use of interferon-α. Although there are 2 types of interferon-α (2a and 2b), which differ slightly in the carbohydrate components of the compound, and to our knowledge only a single report suggested that 1 interferon may be less immunogenic than the other,14 we conclude that there is insufficient evidence to suggest that the interferons should be viewed differently.


The results from 4 RCTs of HDI-α have been reported fully6–9; results from a fifth trial are available only in abstract form15 (Table 1). The North Central Cancer Treatment Group (NCCTG) study compared high-dose, intramuscular interferon-α-2a with observation in 262 patients, including 52 patients who had primary tumors > than 3.5 mm thick and 169 patients who had regional lymph node disease.9 There was no significant difference noted in survival between the 2 treatment groups (hazards ratio [HR] of 0.90; P = .53) at a median follow-up of 6.1 years.

Table 1. Randomized Controlled Trials of Interferon as Adjuvant Treatment for High-Risk Melanoma
Reference/treatment armNo. of patients randomized (evaluated)Median follow-up (yrs)Survival rate (%)
2 yrs5 yrsOther
  • IFN: interferon; NCCTG: North Central Cancer Treatment Group; ECOG: Eastern Cooperative Oncology Group; HDI: high-dose interferon; LDI: low-dose interferon; GMK: GM ganglioside-keyhole limpet hemocyanin vaccine; NR: not reported; WHO: World Health Organization; UKCCCR: United Kingdom Coordinating Committee on Cancer Research; IL-2: interleukin 2; EORTC: European Organization for Research and Treatment of Cancer; SWOG: Southwestern Oncology Group.

  • *

    Data were abstracted from survival curves.

  • P = .023.

  • Survival calculated using the mortality rate (death related to melanoma): 17 patiented died in IFN group (n = 154) vs. 21 patients in the Observation group (n = 157).

Creagan et al., 19959 (NCCTG)264 (262)6.1   
 IFN-α2a  73*73* 
 Observation  5448 
Kirkwood et al., 19966 (ECOG 1684)287 (280)6.9   
 IFN-α2a  63*46 
 Observation  57*37 
Kirkwood et al., 20007 (ECOG 1690)642 (608)4.3  
 HDI-α2b  72*52 
 LDI-α2b  76*53 
 Observation  68*55 
Kirkwood et al., 20018 (ECOG 1694)880 (774)1.3   
 HDI-α2b  78*NR 
 GMK vaccine  73*NR 
Mitchell et al., 200315 (abstract)600 (455)NR   
 HDI-α2b  NRNR 
 LDI-α2b and allogenic melanoma vaccine     
Grob et al., 199820 (France)499 (489)5.0   
 IFN-α2a  92*76 
 Observation  88*68 
Pehamberger et al., 199821 (Austria)311 (311)3.4 (mean)   
 IFN-α2a  NRNR89
 Observation  NRNR87
Cameron et al., 200022 (Scotland)96 (94)6.5   
 IFN-α2a  60*42* 
 Observation  56*29* 
Cascinelli et al., 200123 (WHO)444 (426)7.3   
 IFN-α2a  60*35 
 Observation  53*37 
Hancock et al., 200424 (UKCCCR)674 (674)3.1   
 IFN-α2a  64*46 
 Observation  64*42 
Hauschild et al., 200325225 (223)6.6   
 IFN-α2a plus IL-2  95*78* 
 Observation  92*79* 
Meyskens et al., 199512 (SWOG)285 (202)NR   
 IFN-γ    Stage II, 79 (2.5 yrs); Stage III, 47 (2.5 yrs)
 Observation    Stage II, 89 (2.5 yrs); Stage III, 57 (2.5 yrs)
Kleeberg et al., 200413 (EORTC)830 (830)8.2   
 IFN-α2b    40 (8.0 yrs)*
 IFN-γ    44 (8.0 yrs)*
 Observation    38 (8.0 yrs)*
 Iscador®    37 (8.0 yrs)*
 Observation    43 (8.0 yrs)*

In the landmark ECOG 1684 trial, HDI-α-2b was compared with observation in patients with lesions that measured ≥ 4.0 mm, including patients with resected lymph node disease.6 The majority of the 287 patients who were randomized in that trial had either a recurrence of melanoma with lymph node involvement (61%), or clinically apparent lymph node involvement at presentation (14%), or occult lymph node involvement that was detected through elective lymph node dissection (12%). A median follow-up of 6.9 years demonstrated a statistically significant survival improvement (P = .0237) in the interferon group, with a reduction in mortality at 5 years from 63% to 54%. However, at 12.6 years, an updated analysis reported that the OS difference between the 2 arms was no longer statistically significant (HR of 1.22; P = .18), although RFS for the patients who received HDI continued to demonstrate significant clinical benefit (HR of 1.38; P = .02).

The ECOG 1690 trial was conducted to confirm the results of ECOG 1684 and to evaluate the efficacy of a lower, possibly less toxic dose of interferon.7 Patients were randomized to HDI-α (the ECOG 1684 regimen), low-dose interferon (LDI), or observation. The population that was eligible for the 1690 trial was identical to the 1684 trial population. However, a noteworthy finding was that 75% percent of participants had resected lymph node metastases, in contrast to ECOG 1684, in which 87% of patients were lymph node positive. At a median follow-up of 4.3 years, no survival differences were detected between the 3 arms (HDI vs. observation: HR of 1.0 [95% CI, 0.75–1.33]; and LDI vs. observation: HR of 1.04 [95% CI, 0.78–1.38]). Furthermore, an analysis of RFS by risk group detected no difference in treatment effect between lymph node-negative patients and lymph node-positive patients. Although a benefit of HDI among patients with 2 or 3 positive lymph nodes was suggested, OS by risk group was not reported. A Cox model analysis that compared the 2 ECOG trials (1684 and 1690) demonstrated no difference between the OS curves for the HDI intervention groups; however, it is noteworthy that a significantly higher survival rate was detected in the ECOG 1690 observation group compared with ECOG 1684 (HR of 1.64; P = .0001). Both the preliminary results16 and the most recent ECOG 1690 study update17 (the latter reported at a medium follow-up of 6.6 yrs) concluded that HDI versus observation did not result in an OS benefit (P = .18), although the updated analysis demonstrated a trend toward improvement in RFS (HR of 1.24; P = .09).

A third ECOG-led trial (1694) randomized 880 participants to the intervention (GM ganglioside-keyhole limpet hemocyanin [GM2-KLH] vaccine) or to a control group (HDI). An interim analysis at a 16-month median follow-up detected a statistically significant survival benefit for interferon (HR of 1.52; P = .009; from an analysis of 774 eligible patients). An analysis of the outcomes by lymph node category (0, 1, 2–3, and ≥ 4) indicated that patients without lymph node metastasis (n = 202 patients) had the greatest benefit in terms of RFS (HR of 2.06; P = .012) and OS (HR of 1.88; P value not reported) with interferon. At a median follow-up of 2.1 years, HDI continued to demonstrate superiority to the GM2-KLH vaccine in terms of both OS (HR of 1.33; P = .04) and RFS (HR of 1.33; P = .006).17 Although that study has been interpreted as favoring HDI, the GM2-KLH vaccine possibly may have produced a worse outcome than no treatment.

An abstract that reported interim analysis results from a trial that compared an allogeneic melanoma vaccine (Melacine®; Corixa Corporation, Seattle, WA) combined with LDI-α against HDI (the ECOG 1690 regimen)15 did not indicate a significant difference in RFS: The median OS had not yet been reached at the time of that analysis. A recent pooled analysis of the ECOG and Intergroup trials of HDI adjuvant therapy24 demonstrated a median RFS of 2.3 years (95% CI, 1.9 yrs–2.7 yrs) and a median OS of 6.2 years (95% CI, 5.1 yrs–8.8 yrs) for all randomized patients (n = 1912 patients). By using updated data from the ECOG 1684 and ECOG 1690 trials, a 2-sided univariate log-rank comparison of HDI versus observation (n = 317 patients) revealed a significant benefit for HDI in terms of RFS (HR of 1.30; P = .006) but not in terms of OS (HR of 1.08; P = .42).

Likewise, we pooled the results from the three ECOG trials, excluding the NCCTG study because of large regimen differences. Our meta-analysis of the 2-year death rates abstracted from survival curves in the published reports yielded a risk ratio of 0.85 (95% CI, 0.73–0.99; P = .03) (Fig. 1) that favored HDI. No heterogeneity was detected among the results (P = .91). When the ECOG 1694 trial was excluded from the meta-analysis, the risk ratio changed to 0.87 (95% CI, 0.71–1.07; P = .18).

Figure 1.

Comparison 1: high-dose adjuvant interferon versus control. Outcome 1: mortality risk ratio (RR) at 2 years for interferon versus control. 95% CI: 95% confidence interval; df: degrees of freedom.


The ECOG 1684 trial included a retrospective QOL-adjusted survival analysis (quality-adjusted time without symptoms and toxicity [Q-TWiST]).16 The model constructed included the amount of time patients spent with severe symptomatic toxic events, the amount of time without symptoms of recurrence, and hypothetical (rather than individual patient) utility coefficients. During the 84-month follow-up, patients in the interferon arm spent an average of 5.8 months with at least 1 incident of Grade 3 or 4 toxicity followed by an average of 33.1 months without toxicity before disease recurrence or death. Patients in the control group experienced no severe toxic events and survived for 30.0 months on average before disease recurrence or death.

The findings of a second Q-TWiST analysis that was conducted by using data from both the ECOG 1684 trial and the ECOG 1690 trial18 were almost identical to results from the ECOG 1684 trial: During 84 months of follow-up, patients who received interferon spent an average of 31.5 months without toxicity or tumor recurrence and an average of 7.4 months with at least 1 incident of Grade 3 or 4 toxicity. During 52 months of follow-up, patients who received HDI in the ECOG 1690 trial experienced an average of 23.9 months without toxicity or tumor recurrence and 7.5 months with at least 1 incident of Grade 3 or 4 toxicity, compared with 28.7 months and 0.0 months, respectively, for patients in the observation arm. Although those trials contained a number of methodological flaws, their results demonstrated that, despite its toxicity, HDI is associated with an improvement in QOL-adjusted survival.

Adverse Effects

In the trial of HDI-α-2a by Creagan et al., 44% of patients in the interferon group experienced Grade 3 flu-like symptoms, and 57% had changes in liver function (i.e., a 2-fold increase in aspartate aminotransferase levels).9 Over the course of follow-up (median, 6.1 yrs), 45% of patients in the interferon group had a worsening in ECOG performance score compared with 16% of patients in the observation group (P < .0001).

Sixty-seven percent of patients in the ECOG 1684 trial experienced severe (Grade ≥ 3) toxicity, including constitutional and neurologic symptoms, myelosuppression, and hepatotoxicity, with 9% of patients developing from life-threatening toxicities.6 The dose was reduced or the treatment was delayed in 35% of patients because of toxicity. Consequently, the average daily dose delivered was 20 μu/m2/d for the induction phase and 10 μu/m2/d for the maintenance phase. Two early deaths from hepatotoxicity led to stringent monitoring of liver function in the interferon arm, and that monitoring prevented any further toxic deaths (unpublished data).

In ECOG 1690, similar toxicities were encountered in the HDI group, with the exception that no treatment-related deaths occurred. Dose reductions or delays were experienced in 58% of patients during the HDI therapy induction phase and in 59% during maintenance treatment.7 The average daily interferon dose delivered in the HDI arm was 18.5 U/m2 for the induction phase and 8.2 U/m2 for the maintenance phase. It is noteworthy that the LDI group experienced two toxicity-related deaths.19 The ECOG 1694 trial included Grade 3 or 4 toxicities associated with HDI therapy, such as fatigue (21% of patients), leucopenia (60% of patients), elevation of liver enzymes (27% of patients), and neurologic symptoms. Approximately 10% of patients in that trial discontinued treatment with interferon because of adverse effects. No treatment-related deaths were reported. The most common Grade 3 or 4 adverse effect reported in the vaccine arm was injection-site reactions (2.3%).


Five randomized trials that compared LDI-α with observation20–24 and 1 trial that compared LDI-α plus interleukin-2 with observation25 were located (Table 1). In the French Cooperative Group on Melanoma trial,20 499 patients with lesions that measured > 1.5 mm in depth without clinical evidence of lymph node metastasis were randomized to either LDI-α-2a or no treatment. Fifty-nine deaths were reported in the interferon group, and 76 deaths were reported in the observation group (HR of 0.72; 95% CI, 0.51–1.01). Although 52% of patients had flu-like symptoms in the interferon group, no treatment-related deaths occurred. The Austrian Malignant Melanoma Cooperative Group published results from a trial21 that used the same design and population that were used in the French study20 but employed a different schedule of administration. The Austrian trial reported 17 deaths in the interferon group and 21 deaths in the control group.

A Scottish trial randomized patients who had either primary melanoma that measured ≥ 3.0 mm thick or evidence of regional lymph node involvement to either LDI therapy for 6 months or observation.22 Those authors reported an initial improvement in disease-free survival (DFS) and OS with interferon but noted that the sample size was too small to detect meaningful differences. Although results from the World Health Organization (WHO) adjuvant melanoma trial comparing LDI with observation in 444 patients with resected lymph node melanoma23 demonstrated that the treatment was tolerated well, no differences in either DFS or OS were observed. Similarly, Hancock et al. reported that, at a median follow-up of 3.1 years, no significant differences in OS (P = .6) or RFS (P = .3) were detected in patients with completely resected, high-risk melanoma who were randomized to either LDI-α or observation.24

Finally, Hauschild et al. recently reported results from a trial of LDI plus interleukin-2.25 Patients who had melanomas > 1.5 mm thick and no clinical evidence of lymph node metastases were randomized to treatment with interferon plus interleukin-2 or to observation. After a median observation of 6.6 years, no significant difference in OS between the treatment and control arms has been detected (P = .93).

Published Metaanalyses and Systematic Reviews of HDI-α and LDI-α

Wheatley et al. conducted a literature-based meta-analysis of 12 randomized trials of adjuvant interferon versus observation in patients with melanoma.26 Those authors concluded that RFS was improved with interferon-α (HR for recurrence, 0.83; 95% CI, 0.77–0.90 [P = .000003]), corresponding to a 17% reduction in the odds of recurrence. There was no clear benefit for survival (HR for mortality, 0.93; 95% CI, 0.85–1.02 [P = .1]). That meta-analysis did not include the ECOG 1694 trial. However, Wheatley et al. pooled the vaccine trial data with data from the remaining two ECOG studies separately and reported no evident survival benefit. Further subgroup analyses were conducted to examine evidence for a dose-response relation and indicated a significant trend for an increasing benefit of interferon with increasing dose in terms of RFS. Pirard et al. conducted a second literature-based meta-analysis of nine randomized trials of interferon versus observation.27 Similar to the results reported by Wheatley et al., Pirard et al. detected an improvement in the recurrence rate with interferon (odds ratio of 0.74; 95% CI, 0.64–0.86) but detected no improvement in OS.

Lens and Dawes reported the results of a systematic review of interferon but did not include information from the ECOG 1694 trial in their analysis or pool the results.10 Those authors concluded that interferon does not confer an OS advantage and that larger studies are needed to establish the effect of interferon-α and to identify subgroups of patients who may benefit from adjuvant therapy with interferon.

Interferon γ

Two trials that were identified in our data base search included interferon-γ as an experimental arm (Table 1).12, 13 The SWOG study was closed early, because a planned interim analysis revealed that survival was better in the control group (relative risk of 1.31; 95%CI, 0.88–1.95 [P = .18]), and the likelihood that interferon-γ would reduce the risk of death by ≥ 25% was remote.12

Kleeberg et al. reported the results of a randomized trial that compared LDI-α, interferon-γ, and observation (EORTC 18871).13 At some centers, a fourth treatment arm was included to investigate Viscum album praeparatum mali (Iscador M®, Weleda Inc., Palisades, NY). However, after a median follow-up of 8.2 years, no significant improvements in DFS or OS were detected.


Of the 4 RCTs of levamisole in patients with melanoma (Table 2), 3 were placebo-controlled.28–31 The National Cancer Institute of Canada (NCIC) study5 enrolled a heterogeneous group of patients, and 50% of those patients had a high risk of developing recurrent melanoma (unpublished results).

Table 2. Randomized Controlled Trials of Levamisole as Adjuvant Treatment in High-Risk Melanoma
Reference(s)/treatment armNo. of patients randomized (evaluated)Median follow-up (yrs)2-yr survival (%)5-yr survival (%)
  • EORTC: European Organization for Research and Treatment of Cancer; NCIC-CTG: National Cancer Institute of Canada Clinical Trials Group.

  • a Data were abstracted from survival curves.

  • This trial also included a dacarbazine arm (see Table 4).

  • This trial also included a Bacillus Calmette–Guerin (BCG) arm and a levamisole plus BCG arm.

Spitler, 199128; Spitler and Sagebiel, 198029203 (200)10.5  
 Levamisole  79*62*
 Placebo  78*64*
Loutfi et al., 198730156 (137)5.0  
 Levamisole  92*74*
 Placebo  93*80*
Lejeune et al., 198831 (EORTC)325 (274)4.6  
 Levamisole  85*85*
 Placebo  56*56*
Quirt et al., 1995 (NCIC-CTG)577 (543)8.5  
 Levamisole  78*80*
 Observation  80*62
Table 3. Randomized Controlled Trials of Vaccines as Adjuvant Treatment in High-Risk Melanoma
Reference/treatment armNo. of patients randomized (evaluated)Median follow-up (yrs)Survival rate
  • GM2: GM ganglioside; BCG: Bacillus Calmette–Guerin; VCN: Vibrio cholerae neuraminidase; NCI: National Cancer Institute; SWOG: Southwest Oncology Group.

  • *

    Data were abstracted from survival curves.

  • This trial also included a Bacillus Calmette–Guerin arm and a methyl-lomustine arm (see Table 4).

  • This trial also included a Bacillus Calmette–Guerin arm.

McIllmurray et al., 19773615 (15)2.0 
 Autologous irradiated tumor cells plus BCG  50% (1 yr), 38% (2 yrs)
 Observation  100% (1 yr), 57% (2 yrs)
Aranha et al., 19793731 (31)1.3 
 VCN treated autochthonous tumor cells plus BCG  84% (1 yr),* 34% (2 yrs)*
 Observation  76% (1 yr),* 55% (2 yrs)*
Fisher et al., 198140 (NCI)181 (166)2.4 
 Neuraminidase-treated allogeneic cells plus BCG  65% (2 yrs)*
 Observation  70% (2 yrs)*
Morton et al., 198238149 (140)4.2 (mean) 
 Allogeneic melanoma cells plus BCG  59% (2 yrs), * 51% (4 yrs)*
 Observation  49% (2 yrs),* 46% (4 yrs)*
Livingston et al., 199433123 (122)5.25 
 GM2/BCG vaccine  50% (2 yrs),* 48% (4 yrs)*
 BCG alone  33% (2 yrs),* 29% (4 yrs)*
Wallack et al., 199832250 (217)3.9 
 Vaccinia melanoma oncolysate  70% (2 yrs), 60% (3 yrs), 49% (5 yrs)
 Placebo: live vaccinia airus  66% (2 yrs), 56% (3 yrs), 48% (5 yrs)
Bystryn et al., 20013438 (38)2.5 
 Polyvalent, shed antigen, melanoma vaccine  83% (1 yr), 67% (2 yrs), 53% (3 yrs)
 Placebo vaccine (human albumin)  77% (1 yr), 61% (2 yrs), 33% (3 yrs)
Hersey et al., 200235700 (673)8.0 
 Vaccinia melanoma cell lysates  60% (5 yrs), 53% (10 yrs)
 Observation  550% (5 yrs), 42% (10 yrs)
Sondak et al., 200441 (SWOG)689 (598)7.8 
 Allogeneic melanoma vaccine  82% (5 yrs)
 Observation  76% (5 yrs)

Although the initial report by Spitler and Sagebiel29 described a survival trend in favor of levamisole compared with placebo in the subgroup of patients without lymph node disease (2-sided P = .07), there was no survival difference between treatments for the total study population. This lack of benefit was later confirmed in a long-term follow-up report.28 Loutfi et al.30 and Lejeune et al.31 also concluded that there was no meaningful impact on survival with levamisole compared with placebo. However, in the NCIC trial,5 a statistically significant difference in the survival rate in favor of levamisole was detected when the 5-year point estimates of OS were assessed (78% for the levamisole group vs. 62% for the control group; 2-sided P = .027), representing a risk reduction in mortality of 29% that was observed in all risk groups, including the group discussed in the current systematic review. However, when the OS experience was compared between the groups, the difference was not significant (2-sided P = .08).

A meta-analysis of 5-year death rates (Fig. 2) abstracted from survival curves in the published reports, as expected, disclosed a risk ratio of 0.94 (95% CI, 0.75–1.20; P = .6). No heterogeneity was observed among the results from the studies that were included (P = .19).

Figure 2.

Comparison 2: adjuvant levamisole versus control. Outcome 2: mortality risk ratio (RR) at 5 years for levamisole versus control. 95% CI: 95% confidence interval; df: degrees of freedom.

Although morbidity from levamisole generally is mild, it was severe enough for the discontinuation of therapy in 41% of patients in the NCIC study,5 in 44% of patients in the study by Loutfi et al. (compared with 16% in the placebo group),30 and in 17% of patients in the study by Lejeune et al. (compared with no patients in the placebo group).31 Hematologic abnormalities generally were rare, and no treatment-related mortality was observed. After reviewing the available data, we concluded that, if levamisole has an impact on the clinical course of malignant melanoma when it is administered in the adjuvant setting, then that effect is marginal.


Seven of the nine vaccine trials32–39 were confined to patients with lymph node involvement, and most patients in one of the other trials were lymph node positive40 (Table 3). None of the trials reported a significant improvement in OS for patients who received vaccines, an observation that was confirmed in a recent update of the SWOG 9035 trial.41 However, that study's subset analysis of patients who were positive for human leukemic antigen (HLA)-A2 and/or HLA-C3 demonstrated a significant 5-year OS benefit of 93% for patients in the vaccine group compared with 74% for patients in the observation group (P = .009). This clearly hypothesis-generating observation cannot be used in directing clinical decisions. Given the heterogeneity of the studies and vaccines employed, we elected not to pool those data in our analysis.

Table 4. Randomized Controlled Trials of Chemotherapy as Adjuvant Treatment for High-Risk Melanoma
Reference/treatment armNo. of patients randomized (evaluated)Median follow-up (yrs)SurvivalP value
  • EORTC: European Organization for Research and Treatment of Cancer; COG: Central Oncology Group; FU: follow-up; NR: not reported; BCG: Bacillus Calmette–Guerin; CCNU: lomustine; SWOG: Southwest Oncology Group; NCI: National Cancer Institute; BCNU: carmustine.

  • *

    This trial also included a levamisole arm (for details, see Table 2).

  • Data were abstracted from survival curves.

  • This trial also included a Bacillus Calmette–Guerin (BCG) arm and a dacarbazine plus BCG arm.

  • §

    This trial also included a Bacillus Calmette–Guerin arm and a vaccine arm (for details, see Table 3).

Lejeune et al., 198831 (EORTC)*325 (274)4.0  
 Dacarbazine  68% (3 yrs), 61% (4 yrs) 
 Placebo  74% (3 yrs), 69% (4 yrs)  
Hill et al., 198145 (COG)174 (165)2.5  
 Dacarbazine  43% (last FU) 
 Observation  55% (last FU) 
Veronesi et al., 198242931 (761)3.4 (mean)  
 Dacarbazine  47%(3 yrs) 
 Observation  42% (3 yrs) 
Jacquillat et al., 198246117 (117)NRNR 
Quirt et al., 19835194 (94)6.4  
 Dacarbazine  61% (3 yrs) 
 BCG  47% (3 yrs) 
Hansson et al., 19854326 (26)3.7 < .025
 Dacarbazine or dacarbazine with CCNU and vincristine  95% (3 yrs), 94% (4 yrs) 
 Observation  79% (3 yrs), 52% (4 yrs) 
Karakousis and Emrich, 19875282 (82)6.1  
 Dacarbazine plus Estracyt  59% (3 yrs), 55% (4 yrs), 55% (6 yrs) 
 Observation  68% (3 yrs), 68% (4 yrs), 56% (6 yrs) 
Tranum et al., 198744 (SWOG)123 (121)NR  
 Dacarbazine plus carmustine plus hydroxyurea  65% (6 yrs) 
 Observation  65% (6 yrs) 
Fisher et al., 198140 (NCI)§181 (166)2.4  
 Methyl-CCNU  68% (2 yrs), 51% (3 yrs) 
 Observation  70% (2 yrs), 54% (3 yrs) 
Karakousis and Blumenson, 199353173 (173)4.7 0.59
 BCNU plus actinomycin-D plus vincristine  30% (5 yrs) 
 Observation  25% (5 yrs) 


The 10 trials of adjuvant chemotherapy are summarized in Table 4. In the largest chemotherapy study,42 47% of patients who received dacarbazine were alive after 3 years compared with 42% of patients in the control group (P = .64). Only the small study (n = 26 patients)43 reported a statistically significant survival benefit for patients who received chemotherapy as adjuvant treatment (P < .025). Although data from the two active-treatment arms (dacarbazine alone and dacarbazine in combination with carmustine and vincristine) were combined and compared with results for nine patients in the control group, the trial was far too small to permit any conclusions.

Three-year mortality rates, from the text or from survival curves in the published reports from seven studies, were pooled (Fig. 3). Three studies were not included in the meta-analysis, because the number of deaths at 3 years could not be ascertained44, 45 or because no survival data were reported.46 The mortality risk ratio from the pooled analysis (0.94; 95% CI, 0.84–1.06; P = .3) did not demonstrate any difference between chemotherapy and control. No heterogeneity was found among the results from the studies that were included (P = .52).

Figure 3.

Review of adjuvant therapy of malignant melanoma. Comparison 3: adjuvant chemotherapy versus control. Outcome 1: mortality risk at 3 years for chemotherapy versus control. RR: risk ratio; 95% CI: 95% confidence interval; BCNU: carmustine; df: degrees of freedom.

Chemotherapy plus Interferon

A trial of 3 cycles of single-dose dacarbazine followed by interferon-α-2a compared with observation in 26 patients with melanoma lesions that measured between 1.50 mm and 5.00 mm in depth was closed early.47 At a median follow-up of 4.5 years, there had been 2 deaths in the control group and 8 deaths in the adjuvant therapy group.


In this systematic review, we have addressed a fundamental question from a clinical perspective: what systemic therapy should clinicians recommend to patients who have been rendered disease-free after resection of cutaneous melanomas and who are at high risk for subsequent recurrence? On the surface, the results of this comprehensive review clearly showed that none of the therapeutic modalities identified confers a long-term survival advantage.

Further scrutiny of the potential benefits of adjuvant therapy indicates that treatment with HDI-α consistently produces a significant improvement in DFS. However, the inconsistent results for OS in the three ECOG studies6–8 raise problems. Although a significant survival benefit initially was observed in the first trial, that benefit no longer was apparent with longer follow-up, but the suggestion is that the lack of a survival benefit in ECOG 1690 was due to the fact that the control group in that trial fared better than the control group in the earlier ECOG 1684 trial. One explanation given for that assertion is that a significantly greater proportion of patients in the observation arm were salvaged by an interferon-containing regimen (31% of recurrences vs. 15% of recurrences in the HDI arm; P = .003) or, for patients with more advanced disease, biochemotherapy (17% of recurrences in the observation arm vs. 6% of recurrences in the HDI arm; P = .013)7 There also was a considerable difference in the stage distribution among patients in all three trials. Although the distribution of patients between the treatment regimens in each of those trials appeared to be balanced, the impact of the differences in the characteristics of patients enrolled in those studies on the outcomes observed remains a subject of conjecture that cannot be resolved readily.

A possible confounding issue in the interpretation of the ECOG trial results is how to incorporate information from ECOG 1694, which did not include a no-treatment control arm: The concern that is expressed commonly is that the experimental vaccine arm may have resulted in worse outcomes than may have been observed in a no-treatment control arm. In the absence of a randomized trial comparing the GM2-KLH vaccine with a no-treatment control group, this concern cannot be refuted absolutely. Although the possibility exists that the vaccine may have made the outcome worse than that seen in a no-treatment control group, in our view, the probability that this occurred and that it affected the outcome seen in ECOG 1694 is very low. Therefore, although we have examined the effects of pooling the data from studies of HDI without the results of ECOG 1694, we believe the pooling of data, including the data from ECOG 1694, is justified.

Despite these explanations, the available weight of evidence does not indicate that HDI produces a benefit in terms of long-term survival. However, both RFS and short-term survival are improved significantly, as evidenced in our metaanalysis of 2-year mortality rates. We specifically chose this particular endpoint, because survival data for ECOG 1694 were not available beyond 2 years. In addition, from a practical perspective, we believed that 2-year survival may represent a meaningful benchmark in patients at high risk for recurrence. Indeed, “RFS” itself cannot be discounted, because the relevance and importance of this period has never been evaluated satisfactorily in patients from an emotional, physical, or QOL perspective. Until this has been determined, we recommend considering and discussing HDI as a reasonable option in appropriate patients.

Which patients should be considered for treatment with adjuvant HDI? The ECOG studies on which this recommendation was based examined the role of therapy in patients with AJCC Stage IIB and III melanoma, in which patients with Stage IIB disease had no evidence of lymph node involvement but had lesions that measured > 4.0 mm in depth, and patients with Stage III disease had either evidence of regional lymph node disease or up to 5 in-transit lesions. Staging has evolved since the time of those trials. The latest AJCC staging classification system48 is based on an analysis of 17,600 patients49 and demonstrates the importance of ulceration in the primary lesion. In that system, patients who have lesions with a depth of invasion between 2.0 mm and 4.0 mm and ulcerated primary lesions have the same prognosis as patients who have lesions > 4.0 mm deep without ulceration. The HDI trials provide insufficient information on the benefit of such therapy in those patients. Therefore, a decision to include that group of patients as appropriate candidates for HDI would have to be made on clinical grounds, because there is no information now nor likely will there be in the near future that would enlighten matters. This same rationale also needs to be applied to patients with satellitosis, a group that was underrepresented in the trials mentioned above.

What about patients with known lymph node-metastatic disease, as determined through sentinel lymph node biopsy procedures, in whom the sentinel lymph node is identified as the only lymph node that harbors metastatic disease? This question is particularly relevant to patients with intermediate-risk disease (primary melanoma 1.5–4.0 mm deep), who most commonly undergo sentinel lymph node procedures. We believe that these patients are candidates for adjuvant interferon therapy after a therapeutic lymph node dissection, although the benefit of such therapy currently is being assessed in an ongoing trial.50

The results of the current review are disappointing, because none of the therapeutic modalities examined resulted in meaningful long-term survival advantages in patients with high-risk, primary melanomas. Notwithstanding the results from trials that have incorporated HDI-α, novel therapies should be explored aggressively, and patients should be encouraged to participate in appropriate randomized trials. Indeed, if improvement in overall survival is the primary objective of such trials, then it well may be justified to compare newer experimental therapies with observation only after initial surgical management.