M. Rodero, A. Wright, M. Malt, M. Smithers, A. Green, K. Khosrotehrani
Introduction: Despite advances in staging of melanoma
through the sentinel node procedure it is diffi cult to predict
whether patients with local invasive high risk melanoma
are at risk of recurrence or death or whether they are cured.
Our objective was to establish a prognostic multivariable
assay on primary melanomas to predict outcome independently
of classical clinic-histological covariates.
Methods and patients: High risk melanoma patients
selected to undergo a sentinel node biopsy (Stage Ib and II)
at the Princess Alexandra Hospital in Brisbane were recruited
from 2000 to 2007. Primary melanoma tissue was obtained
and analysed by immunohistochemistry. Follow-up information
on recurrence and death was prospectively recorded.
Results: 545 melanoma patients were included. To date
analysis has been performed on 117 patients for three
markers: p16, Ki67 and P-STAT5. P-STAT5 expression was
present mainly in the infl ammatory infi ltrate. A combined
variable (CV) representing presence of p16, presence of
P-STAT5 and Ki67 staining < 25% was used. When tested in
multivariate cox proportional hazard models preliminary
analyses on the 117 patients identifi ed the CV as a strong
independent predictor of disease free survival (HR = 0.14
[0.03–0.64], p = 0.01) in addition to age and sentinel node
status (HR = 6.4 [2.5–16.7], p = 0.0001) in models where
Breslow and ulceration were included and were no longer
predictors. Similar fi ndings were obtained for overall survival
(HR for death = 0.11 [0.01–0.8], p = 0.03).
Conclusions: Combinations of molecular markers on
primary melanomas have prognostic information of higher
value than classical clinicopathological co variates such as
Breslow index or Ulceration.