Hematoxylin and eosin or double stain for CD34/SOX10: Which is better for the detection of lymphovascular invasion in cutaneous melanoma?

BACKGROUND

Lymphovascular invasion (LVI) is considered an unfavorable prognostic factor in cutaneous melanoma (CM). However, its detection by hematoxylin and eosin (H&E) is challenging, with discordant data about its association with clinical-pathological features and no previous studies investigating the inter- (IrOA) and intra-observer (IaOA) agreement. Herein, we tested H&E and double staining (DS) for CD34/SOX10 to detect the LVI in a cohort of 92 CMs, evaluating the IrOA, the IaOA, and the association with the other clinical-pathological features.

METHODS

Five authors independently evaluated 92 consecutive and retrospectively enrolled cases of CMs. We assessed the IrOA (Fleiss's Kappa/FK and intraclass correlation coefficient/ICC) and the IaOA (Cohen's Kappa/CK) with both H&E and CD34/SOX10. Furthermore, we compared the LVI assessment with the two stains and analyzed the association with other clinical-pathological features [χ tests for dichotomous and categorical data; Student t-test (normal distribution) and Mann-Whitney U-test (non-normal distribution) for continuous data].

RESULTS

The IrOA was almost identical with H&E (FK=0.446; ICC=0.805) and CD34/SOX10 (FK=0.454; ICC=0.810); by contrast, the IaOA was higher with H&E for one pathologist (CK: 0.809) and with CD34/SOX10 for the other one (CK: 0.563). Applying previously defined criteria, LVI was detected in 10 (9.2%) and 11 (10.1%) cases with H&E and CD34/SOX10, respectively (p = 1.000). Both H&E and CD34/SOX10 were significantly associated with vertical growth phase (H&E, p: 0.014; CD34/SOX10, p: 0.010), mitosis ≥ 1/mm2 (H&E, p: 0.000; CD34/SOX10, p: 0.004), pT (H&E, p: 0.000; CD34/SOX10, p: 0.001), Breslow thickness (H&E, p: 0.000; CD34/SOX10, p: 0.001), and lymph node and/or distant metastasis (H&E, p: 0.005; CD34/SOX10, p: 0.000); only H&E was associated with ulceration (p: 0.002) and distant metastasis (p: 0.000), conversely, only CD34/SOX10 was associated with lymph node metastasis (p: 0.003).

CONCLUSIONS

CD34/SOX10 does not improve the IrOA and the IaOA of the LVI assessment in CM; furthermore, H&E and CD34/SOX10 show a similar profile of association with the other unfavorable clinical-pathological features of CM. As result, CD34/SOX10 could be a redundant diagnostic tool if applied for the prognostic characterization of not-selected CM in a routine diagnostic scenario.