Purpose: To evaluate the overall survival prognostic value of preoperative 18F-fluorodeoxyglucose positron emission tomography (PET) in breast cancer, as compared with the lymph node ratio (LNR).
Methods: Data were abstracted at a median follow-up 14.7 years from a retrospective cohort of 104 patients who underwent PET imaging before curative surgery. PET-Axillary|Sternal was classified as PET-positive if hypermetabolism was visualized in ipsilateral nodal axillary and/or sternal region, else as PET-negative. The differences of 15 years restricted mean survival time ∆RMST according to PET and LNR were computed from Kaplan-Meier overall survival. The effect of PET and other patients' characteristics was analyzed through rankit normalization, which provides with Cox regression the Royston-Sauerbrei D measure of separation to compare the characteristics (0 indicating no prognostic value). Multivariate analysis of the normalized characteristics used stepwise selection with the Akaike information criterion.
Results: In Kaplan-Meier analysis, LNR > 0.20 versus ≤ 0.20 showed ∆RMST = 3.4 years, P = 0.003. PET-Axillary|Sternal positivity versus PET-negative showed a ∆RMST = 2.6 years, P = 0.008. In Cox univariate analyses, LNR appeared as topmost prognostic separator, D = 1.50, P < 0.001. PET ranked below but was also highly significant, D = 1.02, P = 0.009. In multivariate analyses, LNR and PET-Axillary|Sternal were colinear and mutually exclusive. PET-Axillary|Sternal improved as prognosticator in a model excluding lymph nodes, yielding a normalized hazard ratio of 2.44, P = 0.062.
Conclusion: Pathological lymph node assessment remains the gold standard of prognosis. However, PET appears as a valuable surrogate in univariate analysis at 15-year follow-up. There was a trend towards significance in multivariate analysis that warrants further investigation.
Keywords: 18F-fluorodeoxyglucose; D measure of prognostic separation; Lymph node ratio; Normalized hazard ratio; Positron emission tomography; Predictive factor; Prognostic ranking method.