The Current Status of Evaluation and Treatment of High-Risk Cutaneous Melanoma: Title and subTitle BreakTherapeutic Breakthroughs Remain Elusive

In this month's ARCHIVES, González¹ and Tsai² provide evidence-based critiques of 2 recently published articles regarding cutaneous melanoma. These articles evaluate the clinical utility of sentinel node biopsies (SNBs) and systemic therapy for patients with high-risk melanoma. Both topics have generated much discourse over the past decade, as until recently, “standards of care” have been dictated by theories and consensus owing to the lack of good evidence.

González¹ provides an excellent commentary on the results of the Multicenter Selective Lymphadenectomy Trial (MSLT-I),³ which evaluated the effectiveness of SNB in patients with cutaneous melanoma. The MSLT-I is an excellent example of a “negative” study given a positive “spin” by the authors. In fact, the negative results of the primary aim of this study are nowhere to be found in the conclusions of the article's abstract.

While the MSLT-I was generally well designed and executed, inappropriate use and interpretation of subset analyses compromised the integrity of the reported results. Examining the original protocol as registered in ClinicalTrials.gov (Identifier NCT00275496), the inclusion criteria for this trial were patients with primary cutaneous melanoma with a Breslow thickness of 1.00 mm or greater or patients with Clark level IV or V tumors, of any Breslow thickness. However, the final publication presented results only for a nonconventional subset of patients, defined by the authors as those with primary melanomas with a Breslow thickness of 1.2 to 3.5 mm. Why were the results from a specific subset of the original study reported instead of the entire study population? As pointed out in a postpublication letter to the editor,⁴ the exclusion of 654 prospectively randomized patients from the reported results is highly unorthodox. Despite this irregularity, the MSLT-I failed in its primary aim to show a benefit in overall survival for the SNB procedure.³

The authors did find that the mean ± SE 5-year disease-free survival (DFS) was statistically greater for the SNB group vs the observation group (78.3% ± 1.6% vs 73.1% ± 2.1%).³ However, DFS is a problematic end point in melanoma trials. The implications of local and nodal recurrences are different from those of recurrences at distant sites. While all of these types of recurrence would qualify as a DFS failure, the prognostic implications are significantly different. In fact, the MSLT-I trial results showed that while patients in the SNB group had fewer nodal recurrences compared with the observation group (not an unexpected finding, since the regional nodes were entirely removed in 16% of the patients in the SNB group because of microscopic involvement), the prevalence of distant metastases were actually higher in the SNB group (11.0% vs 7.8%; reported in oral presentation of the MSLT-I data). This significant difference in the site of recurrences may explain why the SNB group showed an improvement in DFS compared with the observation group but no improvement in overall survival.

The most dangerous error in the MSLT-I article is the conclusion made by the authors that the SNB procedure “identifies patients with [microscopic] nodal metastases whose survival can be prolonged by immediate lymphadenectomy.”^{3 (p1307)} The MSLT-I authors imply that if not removed, microscopically involved nodes will progress to macroscopic disease. If such an assumption were correct, 19.4% of the patients randomized to the SNB intervention arm of the study would have been expected (if they had been followed by observation only) to develop clinically palpable nodes by the end of the study (the sum of the 16.0% of patients with positive SNB results plus the 3.4% of patients who initially had a negative SNB result but who later developed nodal disease during follow-up). This incidence is 24% higher than the actual incidence found during follow-up for patients assigned to the observation arm (15.6%). The large number of patients prospectively enrolled in this trial assures that an equal percentage of patients with aggressive disease were randomized to each arm of the study. Therefore, the discrepancy can only be explained if a quarter of patients with positive sentinel nodes in the SNB arm would never have developed palpable nodes if followed by observation. Thus, it is not surprising that survival of the subgroup of patients with positive SNB results (a quarter of whom had disease that was not destined to progress) was superior to a subgroup of the observation arm that had been selected to include only those patients whose tumors had already displayed biologically aggressive behavior.⁵ As pointed out by González¹ in his review, a subset analysis of the MSLT-I data comparing survival data for select portions of previously randomized groups, especially when the subsets were selected at different follow-up time intervals (lead-time bias), is inappropriate, and results from such comparisons are not statistically sound.

Unfortunately, the MSLT-I authors inappropriately implied a cause-and-effect relationship between the SNB procedure itself and prognostic information of select subsets of patients identified by the procedure. There is no debate that SNB offers excellent prognostic information. However, Breslow thickness and ulceration status have also been used for years as gold standards of prognosis, and this prognostic information is included within the cost of the routine histopathology fees. Performing SNB procedures adds approximately $10 000 to $15 000 to the standard cost of performing melanoma excisions without the procedure.⁶ While multivariate analysis of the MSLT-I data confirm that SNB status is the most “statistically significant” prognostic tool, its clinical superiority to Breslow thickness is less clear.

In an editorial on the MSLT-I trial, Balch and Cascinelli^{7 (p1371)} state that

the 80% of patients with pathologically negative results on sentinel-node biopsy can be spared some of the cost and anxiety incurred by frequent visits to an oncologist for clinical and radiologic examinations to detect nodal metastases.

Unfortunately, this statement is not true. One cannot reassure patients with 4-mm-thick melanomas that they have a favorable prognosis just because their SNB result was negative. The sensitivity and specificity of SNB procedures is poor. When used as a prognostic tool, the most appropriate definition for a “false-negative” SNB result would be the failure of a negative SNB result to predict long-term overall survival. Used in this clinically applicable fashion, 9.7% of patients with negative SNB results would be falsely reassured that they would not die from their disease.³ Perhaps more concerning is the corollary: patients found to have positive SNB results may live in fear that their cancers have already spread, when in fact melanoma-specific mortality rates were only 26.2% even for patients with positive SNB results.³

With regard to sparing the cost of radiological examinations to detect nodal metastases, no study has shown that such tests are effective at detecting early disease. The National Comprehensive Cancer Network's Clinical Practice Guidelines for Melanoma^{8 (p10)} state the following:

The consensus of the panel was that routine follow-up CT [computed tomographic] scans were unnecessary; however, they certainly should be performed as clinically indicated. The recommendations recognize the extremely low yield of routine screening chest x-rays and screening blood work in this population.

Any discussion of a diagnostic or prognostic test would be incomplete without considering potential complications of the procedure. Unfortunately, the authors did not discuss complications of the SNB procedure in their article, stating that the data “have been published elsewhere.”^{3 (p1308)} In fact, the data from the MSLT-I trial showed a 10.1% complication rate (most of which were mild) after SNB.⁹ However, complications increased to 37.2% with the addition of selective lymphadenectomy in those patients requiring the procedure.⁹ Of note, these complication rates are for the experienced investigators and therefore most likely underestimate the true incidence of complications expected in practice.

No one would argue that the risks and cost of SNB procedures would be inconsequential had the procedure proven to be therapeutically effective, or even if it identified a group of patients whose clinical course could be altered by adjuvant therapy. In this month's Archives, Tsai² critiqued a recently published systematic review of systemic adjuvant therapy for melanoma.¹⁰ In summary, to date, no systemic adjuvant therapy (including interferon, vaccines, and chemotherapy) has had a significant effect on overall survival for these patients. The best data available is an updated and pooled analysis of the Eastern Cooperative Oncology Group (ECOG) and Intergroup trials of adjuvant high-dose interferon for melanoma published by Kirkwood et al.¹¹ After a median follow-up of 12.6 years, the overall survival benefit of interferon initially found in ECOG 1684 had been lost. An analysis of pooled data from ECOG 1684 and ECOG 1690 (high-dose data) was also performed, which, like the individual studies, showed improved DFS but no benefit in overall survival.¹¹

As pointed out by Tsai² in his critique, most patients treated with interferon experience at least grade 3 toxic effects (prevents daily activity or repeated use of narcotic pain reliever). The toxic effects from the year-long therapy must be weighed against the benefit of obtaining months of improved DFS, especially in light of the fact that ultimately, overall survival is not affected. With regard to SNB procedures identifying appropriate patients for further treatment, the rationale that effective adjuvant therapies will benefit only patients with positive sentinel nodes is unsubstantiated and not intuitive.

Quality-of-life analyses are frequently undertaken for therapies such as interferon, in which no improvement in overall survival is found, but DFS is improved at the expense of notable toxic effects. Unfortunately, published quality-of-life studies regarding the use of interferon in patients with melanoma have suffered from considerable methodological flaws.¹⁰ However, important practical information can be extracted from one of these studies¹² in which the authors found that they could predict which patients would experience improved quality of life with adjuvant interferon and those who would not by applying the following decision rule: Treat a patient with interferon if he or she agrees with the statement “If I had a serious disease, I would gladly accept feeling lousy for a year if it improved my chances of living longer” and if he or she answered “I would put up with the side effects of interferon treatment only if it decreased the chance of the melanoma returning by at least 10%.”^{12 (p1315)} Unfortunately, these 2 statements cannot be answered in the positive with regard to interferon therapy for melanoma.

Patients with melanoma are justifiably frightened and often seek the “newest” innovations. Patients may request unnecessary tests, therapies, and procedures, having obtained inaccurate data from sources such as the Internet prior to seeking consultation from their physician. All too often physicians feel pressured into submitting to these demands. (How many of us are guilty of prescribing antibiotics for those nagging viral respiratory tract infections, despite evidence that such therapies are ineffective?) Levitt¹³ describes these subliminal pressures to which physicians succumb in his book Freakonomics: A Rogue Economist Explores the Hidden Side of Everything. Patients come to us expecting to be healed, not to be told that they have a life-threatening disease for which we have no effective treatment other than early local excision. However, it is our duty as physicians to discuss all of the current data with these patients in a nonbiased manner, helping them sort out fact from hearsay.

Any discussion of treatment options with patients with high-risk melanoma should include the following facts: (1) Undergoing SNB will not improve the patients' overall survival. While the MSLT-I data has been released as “interim results,” median follow-up time is already at 5 years. It is highly unlikely that events encountered after this time will alter the overall results. (2) While SNB might be “statistically” more accurate as a prognostic tool than Breslow measurements, they are both excellent prognostic tools, and the clinical difference between the two in an individual case is less clear. (3) The false-negative rate of SNB procedures is approximately 10%. One must also consider that there is a 20% chance of “positive” results. Would the patient be able to live with this “bad” prognostic information, even though most patients with positive SNB results do not die from their disease? (4) Complications occur in at least 10% of patients undergoing the procedure. If the SNB result is positive, subsequent lymphadenectomy increases the complication rate to approximately 40%. (5) Although SNB may be able to identify patients who might benefit from adjuvant therapy, currently, no therapies (including interferon) have been shown to have an effect on overall survival in patients with high-risk melanoma. Patients have to weigh the months of improved DFS afforded by interferon against the notable toxic effects associated with the treatment. And (6), perhaps the best course of action for patients with high-risk melanomas is to enroll in ongoing clinical trials. In addition to satisfying both the patients' and physicians' need to “do something,” participation in ongoing research gives the patient a potential for a cure not available from current therapies, while helping advance science, bringing us closer to finding an eventual cure for this difficult disease.

Correspondence: Dr Kanzler, Division of Dermatology, Santa Clara Valley Medical Center, 751 S Bascom Ave, San Jose, CA 95128 (matthew.kanzler@stanford.edu).

Financial Disclosure: None reported.