Concerns relating to the conduct and statistical analysis of the Multicenter Selective Lymphadenectomy Trial (MSLT-1) in patients with melanoma

Abstract

The first Multicenter Selective Lymphadenectomy Trial (MSLT-I) was designed to test for a survival difference following wide excision of primary melanoma between patients randomised to sentinel lymph node biopsy (SLNB) and early lymphadenectomy when metastatic disease was identified (the biopsy arm) versus observation alone and delayed lymphadenectomy when regional lymph nodes became palpable (the observation arm). Contrary to that stated in the protocol, almost half the patients entered to the observation arm of MSLT-I were investigated by lymphoscintigraphy and regular targeted high-resolution ultrasound which detected nodal metastasis in some patients before it became palpable, thus influencing the primary end-point of the trial. The method of calculating disease-free survival (DFS) in MSLT-1 has been successfully challenged and to avoid bias caused by trial design, recent guidance from the National Cancer Institute states that this end-point should in future be calculated either by excluding nodal recurrence as an event or by expressing the end-point as distant disease-free survival. Patients with melanoma die of distant metastatic spread and currently there is no evidence that the SLNB procedure influences distant disease-free survival. The provisional results of the fourth interim analysis of MSLT-I support the hypothesis that prognostic false-positivity is the explanation for the large survival advantage claimed for patients having early lymphadenectomy versus delayed lymphadenectomy. This survival difference is best explained by a prognostic difference in the two sub-groups of patients compared. In turn that suggests that removing minimally involved sentinel nodes in a proportion of patients offers no therapeutic benefit.