Sentinel Lymph Node Biopsy Improves Survival Among SEER Patients with Melanoma
Reviewer: Eric Shinohara MD, MSCI
Abramson Cancer Center of the University of Pennsylvania
Last Modified: June 1, 2008
University of Pennsylvania
Type of Session:
- Regional lymph node metastases increase the risk of melanoma-specific death.
- Sentinel lymph node biopsy (SLNB) provides prognostic information in patients with malignant melanoma (MM); however, its therapeutic role is not well defined.
- SLNB is of modest cost and limited toxicity
- Prior studies have not shown a benefit to prophylactic lymph node dissection or SLNB.
- The largest prior randomized trial (MSLT 1 trail, Morton et al.) of patients 18-75 years old with intermediate-risk primary melanoma (INT), defined as MM with a thickness of between 1.2-3.5 mm (level III-V), found no significant melanoma-specific survival (MSS) in patients randomized to SLNB. However, this study may have been underpowered to detect a difference between the two arms.
- In order to validate the results form the Morton study, a larger group of patients was needed, and hence the Surveillance Epidemiology and End Results (SEER) database was used.
- The SEER database of the National Cancer Institute is a national cancer surveillance program that collects information about the incidence and survival of cancer cases from 13 cancer registries. These registries cover approximately 26% of the population in the United States, and are representative of national demographics.
- The purpose of this study was to validate the results from the Morton study as well as to extend the analysis to include patients with MM that were 0.75-1.2 mm in thickness (level II-V) (THIN)
Materials and Methods
- This is a retrospective study using data from the SEER database.
- Data for patients who were treated during a similar time period as the Morton study was collected (from 1998-2002).
- Data was collected in patients with INT and THIN MM.
- MM patients who did and did not undergo SLNB were matched using seven previously described prognostic factors as well as the year of diagnosis (diagnosis prior to 2000). These prognostic factors include: sex, age, thickness, level, anatomical site, ulceration, and histology. Propensity score analysis was used to match patients.
- Kaplan-Meier curves were used to estimate 5-year MSS and overall survival (OS).
- The Prentice-Wilcoxon test was used to compare paired survival times.
Rates of SNLB in patients with MM between 1998-2002 per the SEER data was:
- INT (n=2309): 71%
- THIN (n=2634): 43%
Based on the results of matching, 674 INT and 1126 THIN patient pairs were found. Characteristics of the 674 INT patients were as follows:
- The INT control arm was compared with two other control arms based on SEER data. Control patients from the time period of 1988-1992 were used as a historical control group, and patients from 1993-1997 were used as a “transition” control group, since the use of SLNB was increasing during this time period. There was no significant difference between these other control groups and the 1998-2002 SEER control group.
- INT patients from the SEER database were compared with patients from the Morton study. For INT patients, tumor factors were similar between the SEER patients and those in the Morton study, aside from the rate of ulceration and the level. The Morton study had lower levels and more ulceration( ~30%).
- For SEER INT MM patients:
- 5-year MSS rate was 91.9% (95% CI=89.5-94.4%) for patients who underwent SLNB
- 5-year MSS rate was 82.6% (95% CI=79.4-86%) without SLNB
- SEER data was also matched to control data from the Morton study to add an additional control group:
- 5-year MSS rate was 90.8% (95% CI=87.8-93.8%) for patients who underwent SLNB when the SEER data was matched to controls from the Morton trial
- 5-year MSS rate was 83.7 % (95% CI=78.8-88.5%) for patients without SLNB when the SEER data was matched to controls from the Morton trial
- For SEER THIN MM patients:
- 5-year MSS rate was 97.2% (95% CI=96-98.5%) for patients who underwent SLNB
- 5-year MSS rate was 95.6% (95% CI=94.2-97.1%) without SLNB
- Patients who underwent SLNB had a significantly longer MSS with a hazard ratio of 0.53 (p=0.034)
- Control of potential Biases:
- Other prognostic factors may be present which were not accounted for in the present study which could affect the outcomes of the study.
- In order to try to account for this, a sensitivity analysis was performed which suggested that only a very strong hidden prognostic factor could be biasing the results (a factor with a hazard ration of 1.7 or greater)
- Non-melanoma related comorbidities could affect treatment decisions and outcomes.
- To account for this, the investigators looked at the cause of death and found there was not a difference between patients who had SLNB and those who did not
- Post-operative adjuvant therapy could also affect outcomes
- Difficult to account for as SEER does not collect data on most adjuvant therapies
- The data from this study suggest that SLNB improves survival in patients with both INT and THIN MM, although there was less of an effect in THIN MM.
- These results need to be validated in other datasets.
- Longer follow up is needed to see if these results continue to be valid over time.
- Models based on prognostic factors need to be created to help physicians define how best to treat patients with MM.
- In any retrospective database, it is difficult to account for all of the potential prognostic factors. The present study has done an excellent job of trying to account for variables which sould affect outcome ,as well as trying to account for variables which are not recorded in the SEER database. Using propensity scores to match patients is a valid way to attempt to match patients in a retrospective database, and has been used in prior studies which have analyzed data from the SEER database. However, as the authors note, data on comorbidities, performance status, and adjuvant therapy were not available. Sensitivity analysis can be used to try to account for these variables, but would only be an estimate of the size of the effect (from a hidden prognostic factor) needed to invalidate the results.
- Systematic biases are also difficult to account for in this study. For example, a surgeon is less likely to do a SLNB on a patient who is less likely to survive long-term, as compared to a patient who is otherwise healthy. There are also issues with the SEER database coding itself. There have been changes in the coding over the years, for example when a patient who is having a SLNB is found to have a positive node and thus proceeds to a lymph node dissection. These patients were coded as undergoing a lymph node dissection rather than a SLNB prior to 2002. Because of this, certain patients may have been excluded from this study, and this could have affected the results. Hence, while the results from these investigators are certainly interesting, due to some of the above limitations, the findings should ultimately be looked at as hypothesis-generating.
- Unfortunately, perhaps the best way to resolve the discrepancy between the findings of this study and of the prior study by Morton is with another larger randomized trial. However, given that the Morton trial was already the largest trial of this sort undertaken, it is unlikely that another larger trial comparing SLNB to no SLNB will ever be feasible.