Melanoma is the most lethal of all skin cancers. More than 232,000/year people are diagnosed with melanoma worldwide and more than 65,000 die.1 Fortunately, the majority of patients have a thin (≤1.00 mm) lesion and a resultant excellent long-term survival.2–5

Sentinel lymph node biopsy (SNB) is recommended in patients with intermediate-thickness melanoma and impacts staging, prognostication, therapeutic decision-making, and entry into clinical trials. However, the role of SNB in thin melanoma is controversial. Although SNB may help to stratify patients with a heightened risk of progression or recurrence, critics argue that the proportion of patients with sentinel lymph node (SN) positivity in thin melanoma is low, and therefore, procedural risks, including infection (1.1 %), lymphedema (0.7 %), and a 3/1000 chance of an anaphylactic reaction to the blue dye are unwarranted. In addition, there is a false-negative rate of 5–12 %.6–8 Defining groups of patients with thin melanoma where performance of SNB will affect management and improve outcomes is crucial.

Previous studies have identified high-risk features of thin melanomas linked to SN positivity; however, variability in the literature and predominant retrospective design limit widespread change in clinical practice. International published guidelines recommend discussion and consideration of SNB if any of the following high-risk features are present in patients with thin melanoma: Breslow thickness ≥0.75 mm, mitoses, ulceration, lymphovascular invasion, and microsatellites.5,9–12 Other postulated high-risk features include Clark level IV or V, regression, and young patient age (<40 years).13–17

Recently, several landmark advances have occurred in the systemic treatment of melanoma, spurring a renewed interest in determining predictors of SN metastases in thin melanoma to maximize accrual to clinical trials and potential benefit for patients.18–20 The objective of this systematic review and meta-analysis is to quantify the overall proportion of SN positivity in patients with thin melanoma and to determine the effect of postulated high-risk features of the primary lesion on SN positivity.

Methods

Research Question

  • What is the proportion of SN metastases in patients with thin cutaneous melanoma?

  • What is the association of high-risk features of the primary lesion and SN metastases?

Method of Preparation and Reporting

We conducted and reported this review following standard guidelines.21

Types of Included Studies

Studies of any language and research methodology were included. Narrative reviews, letters, editorials, conference abstracts, and commentaries were excluded but were read to identify potential studies for inclusion. Studies that included a single isolated high-risk feature of the primary lesion and/or SN positivity were excluded, because they were believed to be biased.

To be included in the secondary-outcome analysis (effect of high-risk features of the primary melanoma on SN positivity), studies had to include information about at least one high-risk feature and its relationship with SN positivity. This included all postulated high-risk features: thickness ≥0.75 mm, Clark level IV/V, ulceration, mitoses, regression, microsatellites, lymphovascular invasion (LVI), and the absence of tumor infiltrating lymphocytes (TILs). When multiple reports on the same patient cohort existed, the article with the largest population and/or most comprehensive information on high-risk features was selected.

Types of Participants

The study population included patients with (1) pathologically verified thin melanoma (≤1.00 mm), and (2) completed SNB. Patients with a positive or unknown deep margin were excluded to ensure a uniform population of patients with thin melanoma.

Outcomes

Studies were included if they reported at least one of the following outcomes:

  1. 1.

    Proportion of SN positivity within patients with thin melanoma;

  2. 2.

    Proportion of thin melanoma patients with any high-risk feature of the primary, and proportion of SN positivity grouped by the targeted high-risk feature.

Review Methods

Search Strategy

A search of published literature from 1980 until May 31, 2015 was undertaken using MEDLINE and EMBASE. Search terms included “melanoma” and “sentinel lymph node” (“Appendix” section). The search was supplemented by examining references of retrieved articles and examining the last 5 years of abstracts presented at the American Society of Clinical Oncology (ASCO) and the Society of Surgical Oncology (SSO) annual meetings. Two authors (EC, MKG) reviewed all retrieved citations, selected articles for full review, applied inclusion criteria, and collected data. If required, consensus was reached by discussion and involvement of a third author (FW).

Data Extraction and Quality Assessment

Data extraction parameters were defined a priori. Two reviewers (EC, MKG) assessed the risk of bias among included studies using the Newcastle Ottawa Scale (NOS).22 The NOS assesses risk of bias in three domains: selection of study groups (4 points), comparability of the groups (2 points), and ascertainment of outcome (3 points) of interest. The maximum score is 9, and a higher score indicates a lower risk of bias.

Data Synthesis and Statistical Analysis

Outcomes The primary outcome was proportion of SN positivity in patients with thin melanoma, calculated as the number of patients with a positive SN divided by the total number of patients with thin melanoma who underwent a SNB. SN positivity was defined based on criteria used in each individual study. The majority of studies utilized H&E staining, with immunohistochemistry (S100, HMB45, Melan-A) used either when H&E was negative or as a confirmation of H&E findings. Several studies did not specifically describe their protocol.15 , 23–31 SN detection was either by intradermal injection of technetium-99m (99mTc)-labeled sulfur colloid or both 99mTc and blue dye, with or without lymphoscintigraphy. Studies using other experimental SN techniques were excluded.

The secondary outcome used individual predictors to determine the proportion of patients with and without each predictor with a positive SNB. From this, an unadjusted odds ratio (OR) was calculated. If a study provided an adjusted OR (AOR), this was recorded with its adjusted covariates and pooled with other AORs in the meta-analysis, as described later in this article. The adjusted analysis has the advantage of accounting for confounding (typically other high-risk features of the primary and patient factors), albeit with differing combinations of factors in each reported study. The likelihood of having a positive SN given the presence of an individual high-risk predictor (likelihood = number with a high-risk feature and a positive SN/total number with the high-risk feature) was calculated for each feature and pooled.

Meta-analysis A study-level meta-analysis was performed. The pooled effect measure for the primary outcome was proportion of positive SNs and 95 % confidence intervals (CI). For secondary outcomes, pooled unadjusted OR and AORs were calculated with 95 % CI from studies that provided available data using Mantel–Haenszel method. Included variables in the reported AORs differ from study to study and thus may not be directly comparable; however, the purpose of the pooled statistic was to ascertain the direction of impact and generate hypotheses regarding possible association between factors and SN positivity. Heterogeneity between studies was assessed both clinically and statistically, the latter using the I 2 statistic.32 Random effects model was utilized to generate the weighted summary estimate for each model. Statistical analysis was performed with Review Manager (RevMan), version 5.333 and OpenMetaAnalyst.34

Results

Literature Search and Study Characteristics

Of the initially identified 6224 studies, 60 (10,928 patients) met criteria for inclusion in this review (Fig. 1) and are summarized in Table 1. The number of patients in each study ranged from 4 to 1286 (median 75). The majority of studies had minimal risk of bias with mean NOS of 7 (range 6–8). The nonrandom selection of thin melanoma patients for SNB was the most common limiting feature in these studies. No study was excluded based on risk of bias measures.

Fig. 1
figure 1

Flow diagram of included studies

Full size image
Table 1 Characteristics of the 60 studies included in the systematic review and meta-analysis
Full size table

Proportion of Positive SNs

All 60 studies were included in generation of the primary outcome. The pooled estimate for the proportion of positive SNs within thin melanoma patients was 4.5 % (95 % CI 3.8–5.2 %; I 2 = 61 %; Fig. 2). The overall proportion in each individual study varied [mean 5.0 %, standard deviation (SD) 3.9 %; median 4.4 %, IQR 4.0 %].

Fig. 2
figure 2

Summary of proportion of positive sentinel lymph nodes amongst thin melanoma patients

Full size image

Predictors of a Positive SN

Twenty-seven of the 60 included studies had data regarding high-risk features of the primary on SN positivity. This included information for eight high-risk features (Table 2). Patient age could not be included, because there was wide variability on reporting thresholds of this predictor.

Table 2 Summary results for predictors of sentinel lymph node positivity
Full size table

Significant predictors of SN positivity in both adjusted and unadjusted analyses were thickness ≥0.75 mm, Clark level IV/V, and ≥1 mitoses/mm2. The presence of ulceration was a significant predictor of SN metastases on unadjusted, but not adjusted analysis. The presence of microsatellites was significantly predictive of SN metastases on unadjusted analysis; unfortunately, no studies reported AORs for this predictor.

Results of pooled analyses on the presence or absence of regression, LVI, and TILs as predictors of a positive SN also are reported in Table 2. None were significant predictors of SN positivity.

Discussion

In this comprehensive systematic review and meta-analyses of 60 studies and 10,928 patients, we identified a pooled proportion of SN metastases in patients with melanomas ≤1.00 mm in depth of 4.5 % (95 % CI 3.8–5.2 %). Pathologic features of the primary melanoma significant for predicting SN metastases included: thickness ≥0.75 mm, Clark level IV/V, ≥1 mitoses/mm2, and microsatellites. Ulceration is not considered an independent predictor, because with adjusted analysis its presence did not significantly predict SN metastases. To our knowledge, this represents the largest and most comprehensive systematic review of the available literature on this topic. This is the first study to demonstrate weighted summary estimates of predictors of SN metastases in patients with thin melanoma; however, the majority of the estimates were unadjusted and the number of studies reporting adjusted estimates remained small.

Identifying the subgroup of patients with thin melanomas who have a worse prognosis is critical, because these patients may benefit from adjuvant therapy, clinical trials, and/or more vigilant follow-up. It also is important to identify which patients would not benefit from having a SNB to eliminate morbidity associated with a potentially low-yield procedure. In this study, the overall risk of SN positivity was 4.5 % for patients with thin cutaneous melanoma selected to undergo a SNB. However, the overall positivity rate of 4.5 % was for patients who were highly selected, not the overall thin melanoma population. We noted in this analysis that the proportion of patients with a thickness ≥0.75 mm was more than 50 %, and thus the SNB positivity rate of 4.5 % is likely an overestimation of the true incidence of occult metastases in the thin melanoma population.

Many surgeons use an estimated 5 % rate of SNB positivity rate as an indication for offering SNB in light of the low associated procedural morbidity rates.7 , 23 From our findings, we suggest that patients with melanomas ≥0.75 mm should have a discussion regarding SNB and be offered this procedure, because their likelihood of a positive SNB is 8.8 %. If other high-risk features also are present in addition to depth ≥0.75 mm (Clark level IV/V or microsatellites), patients should be informed that their likelihood of a positive SNB may be higher than 8.8 % and that strong consideration should be given to SNB, given appropriate medical risk stratification. In our analysis, the presence of microsatellites was the strongest predictor on unadjusted analysis, but the presence of ≥1 mitoses/mm2 (vs. no mitoses) was the strongest predictor of SN positivity on adjusted analysis (although no studies reported AORs for microsatellites). Because the combined influence of high-risk factors cannot be mathematically represented from our data, presence of either factor, regardless of the depth of the lesion should prompt strong recommendation of a SNB. Interestingly, the presence of ulceration did not predict SN metastases in adjusted analysis. This could be because few studies performed this analysis.

The false-negative rate of SNB also must be considered when offering a procedure with low likelihood of positivity. The false-negative rate for SNB has been reported to be as high as 12 %, but as low as 5 % when more than 25 cases have been completed, thereby suggesting a relationship between case volume and procedural accuracy.6 , 7 , 35 Other studies also have suggested that patients with melanoma have better outcomes if treated in high-volume centres. Rossi et al. demonstrated that high-volume centres (>300 cases per year) had a higher number of lymph nodes retrieved, which correlated with improved melanoma-specific survival and improved staging.36 , 37 We suggest that patients with thin melanoma have SNB performed in high-volume centres.

Our current review has some limitations. This was a study-level meta-analysis. A patient-level meta-analysis could be performed to help clarify: the variability in reporting of some predictors (e.g., mitosis), as well as a meta-regression could be performed to define the independent effect of high-risk features of the primary on SN positivity. Another limitation is that our weighted estimate on the proportion of SN metastases within thin melanoma patients had statistically significant heterogeneity (I 2 = 61 %; p < 0.001), likely due to one study, which demonstrated a significantly higher proportion of SN metastases than the others.38 This particular study had 210 patients with thin melanomas and high-risk features (ulceration 4 %, Clark level IV 44 %, mitoses 59 %) who underwent SNB. This selection of patients with multiple high-risk features could explain why the SNB positivity rate was 18 % and suggests that multiple high-risk features puts the patient at higher risk of SN metastases. All of the included studies were at risk of selection bias, because they only included patients who were selected by the surgeon to undergo a SNB typically, due to the presence of one or more high-risk features of the primary lesion.

The main strength of this review is that it includes multiple studies examining thousands of patients with thin cutaneous melanoma. As well, multiple predictors of SN positivity were examined in both unadjusted and adjusted analyses. There has only been one previous meta-analysis on this topic. Warycha et al. pooled 34 studies with 3651 participants and found an overall SN positivity rate of 5.6 % (95 % CI 4.5–6.8 %). Our analysis includes 22 of the same studies.39 Our study has more than three times the number of patients, and the richness of the analysis is enhanced by the inclusion of quantitative meta-analysis on high-risk features of the primary lesion.

Future research on thin melanoma should focus on combining and weighting high-risk factors to create validated and dependable decision-making tools. This may be accomplished with patient-level data and creation of a validated nomogram; however, it cannot be extrapolated from the current study-level meta-analysis. In addition, evaluation of patient preferences given various risk estimates needs to be studied.

Conclusions

Our study demonstrated an overall rate of 4.5 % for SNB metastases in patients with a melanoma ≤1.00 mm. Pooled analysis demonstrated an 8.8 % SN positivity rate for melanoma ≥0.75 mm, 7.3 % for Clark level IV/V, 8.8 % for mitoses ≥1/mm2, and 26.6 % for presence of microsatellites. We suggest that all patients with melanomas ≥0.75 mm treated in a high-volume centre should be offered SNB. In addition to a depth of ≥0.75 mm, if other high-risk features also are present (Clark level IV/V or presence of microsatellites), patients should be advised that their chance of a positive SNB may be higher than 8.8 %, and they should be advised to undergo SNB. Further refinement of patient risk stratification may be accomplished with more granular patient-level analysis.