INTRODUCTION

Surgical excision is the recommended treatment for head and neck lentigo maligna (melanoma) (LM/LMM).^{1, 2} However, lesion delineation remains challenging because of the underestimation of lesion extent by clinical examination using dermoscopy and Wood's lamp examination.^3-5 This can result in incomplete excisions and contributes to an increased risk of local recurrence.^6-8 To address this issue, various staged excisional techniques have successfully reduced the local recurrence risk.⁹ Even so, these techniques do not predict the definitive size of the surgical defect, which is particularly crucial in the head and neck owing to cosmetic and functional considerations. Nonsurgical options such as topical imiquimod (IMQ) and radiotherapy can be considered in cases where surgical excision is not feasible.^10-13

Reflectance confocal microscopy (RCM) is a noninvasive diagnostic device that has been shown to improve the detection of subclinical LM, leading to reduced rates of incomplete excisions and local recurrence following wide local excision (WLE) and staged surgical techniques.^{9, 14-17} A pilot study in our centre has demonstrated the reliability of handheld reflectance confocal microscopy (HH-RCM) for presurgical mapping.¹⁸ In this study, we aim to evaluate the impact of HH-RCM following its introduction as the standard of care in our centre and assess its effect on management decisions and long-term follow-up outcomes.

MATERIALS AND METHODS

This research paper expands on a pilot study (2015–2017) conducted in our centre¹⁸ and includes consecutive head and neck LM/LMM patients referred to the Department of Dermatology of the Netherlands Cancer Institute (NKI) with no contra-indication for surgical excision (2018–2023). A historical cohort of consecutive head and neck LM/LMM treated by WLE (n = 92) before HH-RCM was introduced was used as the control group (2003–2014).¹⁹ The researchers followed ethical guidelines and received approval from the institutional review board. A single investigator (Y.S. Elshot) with 33 months of experience with HH-RCM at the start of the inclusion period performed all imaging/analyses. Study data were recorded using a standardized form for the extension cohort. Staging was performed according to the eighth edition of the American Joint Committee on Cancer (AJCC).

The HH-RCM mapping procedure was followed by a multidisciplinary consultation with a dermatologist and a head and neck surgeon with the patient present and used in the management decision-making process. The standard of care in NKI for LM/LMM consists of HH-RCM-guided wide local excision (WLE) with a 5–10 mm surgical margin. Excised tissue was cut perpendicular to the long axis (i.e. bread loafing), and immunohistochemical staining was used when necessary for Melan-A, sox10, S100 or PRAME. Diagnostic sampling errors were defined as LM being reclassified as LMM after surgery.

The lesions were analysed for morphological characteristics and invasive LMM missed at initial diagnosis. The clinical border was determined using polarized dermoscopy (DermLite DL4, 3Gen, Inc., San Juan Capistrano, California, USA). The planned surgical margin was delineated using a surgical pen. Adhesive rings were placed around the entire circumference of the margin and examined for subclinical LM. The margin was extended in the case of atypical dendritic cells, and a new ring was applied. This process was repeated until the entire lesion was mapped, and no subclinical LM was found beyond the proposed surgical margins (Figure 1). The mapping procedure is described in detail elsewhere.¹⁸ Deviation from the standard of care was discussed with the patient based on the extent of the subclinical component with potential functional or cosmetic consequences and/or suspicion of invasive LMM. The diagnostic accuracy of the HH-RCM in detecting subclinical LM and invasive components was compared to the histological outcome (reference standard). The HH-RCM predicted defect size and the duration of the mapping procedure were recorded.

The study's primary endpoint was the surgical outcome or changes in management based on HH-RCM results. Modified management options consisted of imiquimod (IMQ) monotherapy or limited surgery, followed by adjuvant IMQ. Topical IMQ was applied daily for 12 weeks, with intensity reduction (up to a minimum of 5 days per week) in case of side effects. Six months after completing topical IMQ treatment, the entire lesion area was assessed with HH-RCM, and a biopsy was performed when residual LM was suspected. The secondary outcome was the recurrence rate following surgical excision for patients with at least 6 months of follow-up. The pilot study's follow-up outcomes were retrospectively updated. The surgical and follow-up outcomes of the HH-RCM-mapped patients were compared to the historical cohort.

RESULTS

Effect of HH-RCM mapping procedure on LM/LMM management

The diagnostic accuracy outcomes are summarized in Table 2. The median mapping duration was 14 min (range 4–50). Subclinical atypical cells beyond the initial surgical margin were detected in 60% of cases (n = 70). For cases with subclinical LM, the median diameter increased from 20 mm (range 4-55 mm) to 33 mm (range 16-60 mm) for the longitudinal axis and from 14 mm (range 2–40 mm) to 25 mm (range 10–50 mm) for the transverse axis. Subclinical LM was significantly associated with increased lesion size (long-axis, p = 0.04; short-axis, p < 0.001). No significant differences were found in lesion type, subtype, localization, demarcation, pigmentation or Breslow thickness.

TABLE 2. Diagnostic accuracy of the HH-RCM procedure for detecting subclinical LM and invasive components, compared to the histological outcome.

	Sensitivity (95% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)	Accuracy (95% CI)
Subclinical, presence	97.0 (84.2–99.9)	84.6 (71.9–93.1)	80.0 (67.8–88.4)	97.8 (86.4–99.7)	89.4 (80.9–95.0)
Subclinical, extenta	94.3 (80.8–99.3)	82.0 (68.6–91.4)	78.6 (66.9–86.9)	95.3 (84.1–98.7)	87.1 (78.0–93.4)
Detection invasive component	80.0 (52.0–95.7)	88.2 (76.1–95.6)	66.7 (47.5–81.5)	93.7 (84.4–97.7)	86.4 (75.7–93.6)

• Abbreviations: CI, confidence interval; HH-RCM, handheld reflectance confocal microscopy; NPV, negative predictive value; PPV, positive predictive value.
• ^a In cases with detected subclinical LM, histological margins >5 mm for LM and >10 mm for LMM were considered false-positive (overtreatment); positive histological margins were considered false-negative.

Sixteen of 51 LMM (31%) were initially missed at diagnosis, of which 12/16 (75%) were identified during the HH-RCM procedure. The suspicion of the invasive component was based on the presence of cerebriform nests (n = 2), atypical (i.e. sparse) melanocytic nests at the dermo-epidermal junction/papillary dermis (n = 3), epidermal/junctional disarray (n = 4), large melanocyte size (n = 6), pleomorphic pagetoid/atypical cells (n = 1) and widespread atypical cobblestone pattern (n = 1). The five false-positive cases were based on large melanocyte size (n = 2), disarranged pattern (n = 2) and the single case with widespread atypical cobblestone pattern (n = 1).

Modifications in lesion management occurred in 27% (n = 32) of cases. The remaining patients (n = 85) underwent HH-RCM-guided WLE (Figure 2). The primary reasons for modified management were the extent of the subclinical component (n = 23; 72%), suspicion of an invasive component (n = 1, 3%), suspicion of an invasive component combined with the extent of the subclinical component (n = 5, 16%), or refusal of (further) surgery (n = 3, 9%). In the modified management group, the longitudinal (p = 0.049) and transverse (p = 0.008) diameters were notably larger at baseline. There was no significant difference in the median age (p = 0.748) or Breslow thickness (p = 0.95). The median (IQR) age of patients treated by topical IMQ mono- or adjuvant therapy (n = 26) was 70.5 (65.0–76.0). The majority of lesions (n = 10; 39%) were localized on the cheek, whereas perinasal and orbital lesions represented 35% (n = 9) and 15% (n = 4), respectively.

DISCUSSION

Our study demonstrates that implementation of HH-RCM-guided surgery improved the histological clearance rate from 81% to 96.5% (p = 0.001) while maintaining long-term local recurrence-free survival in the pilot cohort. The mapping procedure also changed management strategies in over 25% of patients and identified 75% of LMM missed at initial diagnosis.

Our study's 96.5% histological clearance rate of HH-RCM-guided WLE is notably higher than the reported 78%–83% range for LM/LMM in a recent systematic review when using guideline-recommended WLE surgical margins.⁹ Negative histological margins do not always correlate with recurrence-free survival, however. Only a limited portion of the surgical margin is evaluated due to transversal tissue processing (‘bread-loafing’).²⁰ In this regard, the extent of the histological margin is the strongest predictor of local recurrence.²¹ A 3 mm cut-off point has been proposed as a histological margin <3 mm is associated with a 27% risk of local recurrence, compared to 2.6% for ≥3 mm.^{21, 22} Although the extension cohort's follow-up period was limited, our study's median histological margin of 3 mm aligns with prior data with a limited risk of local recurrence.^{21, 22} In support of our data, Yélamos et al.²³ have shown that the HH-RCM-estimated surgical defect using video mapping correlates well with the eventual surgical outcome.

Approximately one-fourth of surgically treated LM were upstaged to LMM. As HH-RCM has a penetrative depth limited to the level of the papillary dermis in the horizontal plane, it can be challenging to determine the invasion depth of atypical cells.²⁴ Nonetheless, HH-RCM successfully identified 75% of invasive cases that were initially missed. To our knowledge, the study by Melhoranse Gouveia et al.²⁵ is the only study evaluating the diagnostic accuracy of RCM in detecting subclinical invasion components in LMM. A total of 229 LM/LMM were evaluated, and the invasive component was detected in 89% of the cases with a sensitivity of 63% (95% CI 52%–78%) and a specificity of 79% (95% CI 74%–88%). The three features most predictive of an invasive LMM component were epidermal/junctional disarray, melanocytic nests and large melanocyte size. Only including these three predictive criteria would remove two cases from our analysis and result in a modest change in sensitivity from 80% to 79% and specificity from 88% to 90%. These proposed features align with an observational RCM study showing that melanocytic nests presenting as lentiginous perifollicular (i.e. medusa-head-like structures) and dermal nests were significant RCM features distinguishing LM from LMM.^{26, 27} A possible explanation for false-positive outcomes could be pagetoid cells. While they have been shown to predict invasive LMM in histology,²⁸ they are not a predictive RCM feature for LMM, possibly because intraepidermal Langerhans cells and melanocytic hyperplasia also have dendritic morphology.^{25, 29} False-negative outcomes are likely the result of the horizontal orientation of the imaging and limited penetration depth. Finally, similar to our data, the NPV (99%) was higher than the PPV (46%). Negative predictive value is a critical diagnostic outcome when selecting LM cases to be treated using nonsurgical modalities.

Over a quarter of the cases had their management strategy changed due to the outcome of the HH-RCM mapping procedure. Most patients underwent limited surgical excision, followed by adjuvant IMQ or IMQ monotherapy. A recent Delphi Consensus paper endorsed topical IMQ as the primary alternative to surgical intervention in monotherapeutic and adjuvant contexts.¹ The clinical clearance rates based on data from systematic reviews of IMQ monotherapy range from 63% to 79%. Still, the heterogeneity of the data should be considered when interpreting the reported efficacy of topical IMQ.^{30, 31} The treatment effectiveness of IMQ highly depends on the intensity, duration and presence of local inflammation. In line with the standard of care in our centre, the greatest odds for clearance are found with 6–7 applications per week with at least 60 applications, with a significant decrease in efficacy when used less than 5 days per week.^{10, 30}

Clinical clearance rates following IMQ treatment may overestimate histological clearance rates due to limited partial sampling by single-punch biopsies. To facilitate early detection of treatment failure, all patients undergoing topical IMQ treatment were evaluated using HH-RCM 6 months after completing treatment. Of the 28 patients, only one case of residual LM was histologically confirmed by HH-RCM-guided biopsy. Previous studies have indicated that RCM is more accurate in detecting treatment failure than clinical-dermoscopic evaluation, with 100% sensitivity and specificity exceeding 92%.^{32, 33} More than half of the IMQ-treated patients received adjuvant treatment, which appears to increase clearance rates compared to monotherapy and has recurrence rates comparable to micrographically controlled surgical techniques, supporting our low residual LM rate.^{34, 35}

CONCLUSIONS

Our findings support the efficacy of HH-RCM for precisely identifying surgical margins, reducing recurrence rates, and detecting subclinical LMM. This allows for an informed selection of patients suitable for nonsurgical treatment. Although staged micrographically controlled surgical techniques have demonstrated a clear advantage in minimizing local recurrence compared to WLE, they fall short in predicting the final defect size.⁹ Topical IMQ or radiotherapy may be considered when surgical excision is not feasible.^{10, 12} However, nonsurgical treatment requires caution, given that clinical features are unreliable in predicting invasive LMM. Implementing HH-RCM as the standard of care could help resolve diagnostic challenges, enabling more tailored treatment approaches for LM/LMM patients.⁴⁰ To confirm our findings, a prospective, preferably multi-centre, matched-controlled study comparing HH-RCM-assisted excision to micrographically controlled surgery for LM/LMM with long-term follow-up and survival outcomes is needed.

FUNDING INFORMATION

None.

CONFLICT OF INTEREST STATEMENT

None.

ETHICS STATEMENT

Approval from the ethics committee was waived following a review by the NKI Institutional Review Board. The patients in this manuscript have given written informed consent to the publication of their case details.