Serum prognostic markers in head and neck cancer.

PURPOSE Recognized prognostic factors do not adequately predict outcomes of head and neck cancer (HNC) patients after their initial treatment. We identified from the literature nine potential serum prognostic markers and assessed whether they improve outcome prediction. EXPERIMENTAL DESIGN A pretreatment serum sample was obtained from 527 of the 540 HNC patients who participated in a randomized controlled trial. During follow-up, 115 had a HNC recurrence, 110 had a second primary cancer (SPC), and 216 died. We measured nine potential serum prognostic markers: prolactin, soluble interleukin-2 (IL-2) receptor-alpha, vascular endothelial growth factor, IL-6, squamous cell carcinoma antigen, free beta-human choriogonadotropin, insulin-like growth factor-I, insulin-like growth factor binding protein-3, and soluble epidermal growth factor receptor. Cox regression was used to identify a reference predictive model for (a) HNC recurrence, (b) SPC incidence, and (c) overall mortality. Each serum marker was added in turn to these reference models to determine by the likelihood ratio test whether it significantly improved outcome prediction. We controlled for the false discovery rate that results from multiple testing. RESULTS IL-6 was the only serum marker that significantly improved outcome prediction. Higher levels of IL-6 were associated with a higher SPC incidence. The hazard ratio comparing the uppermost quartile to the lowest quartile of IL-6 was 2.68 (95% confidence interval, 1.49-4.08). IL-6 was also associated with SPC-specific mortality but not with mortality due to other causes. No marker improved outcome prediction for cancer recurrence or overall mortality. CONCLUSIONS IL-6 significantly improves outcome prediction for SPC in HNC patients.