[Skip to Navigation]
Oct 2011

Visualizing Bayesian Analysis Using a Spreadsheet Geometry —Reply

Author Affiliations

Author Affiliations: Department of Dermatology, Rush Medical College, Rush University, Chicago, Illinois.

Arch Dermatol. 2011;147(10):1225-1227. doi:10.1001/archdermatol.2011.296

In reply

Melski provides a geometric representation of derived Bayesian analysis data1 according to sensitivity and specificity of SLNB status for cutaneous melanoma (CM) related death, segregated by tumor thickness. Given these data, the prognostic usefulness of SLNB for CM-related death must be questioned.

Wong et al2 provide a thoughtful commentary immediately following the Bayesian analysis,1 highlighting the paucity of informative reports with raw numbers available for patients with CM who have thick tumors ( ≥4 mm). Most SLNB reports are univariate and/or multivariate analyses and include tumors of any thickness, a broad range of patient age, and short follow-up intervals. Wong et al2 quote their own univariate and multivariate analysis3 for patients with thick tumors who routinely undergo SLNB, showing a markedly improved 5-year overall survival rate when the SLNB result is negative for CM. However, statistical significance does not always equate with clinical usefulness.

Add or change institution