Desley van Zoggel

Chapter 3 44 When comparing complete responders with noncomplete responders (Mandard 2–5), multivariable analyses demonstrated a significant improved effect of complete response for OS with a 3-year OS of 92 percent versus 57 percent respectively (HR 4.706; P = 0.045, 95 percent CI 1.063–20.833; Figure 1a, b). For disease free survival, only response remainedsignificant aftermultivariableanalysis (HR 2.070; 95 percent CI 0.904–4.736 for ‘good’ responders and HR 3.525; 95 percent CI 1.388–8.951 for ‘poor’ responders, P =0.024). MedianDFS for complete responders, good responders, and poor responders was 35, 14, and 11 months, respectively. For local recurrence free survival, response (HR 2.232 95 percent CI 0.787–6.328 for ‘good’ responders and HR 5.684; 95 percent CI 1.776–18.191 for ‘poor’ responders, P = 0.004), resection margin (HR 2.104; 95 percent CI 1.125–3.933; P = 0.020), and lymph node status of the primary tumour (HR 0.816; 95 percent CI 0.388–1.719 for N1 and HR 2.156; 95 percent CI 1.001–4.640 for N2, P = 0.036) remained significant after multivariable analysis (Figure 1a, b). For metastasis free survival, only response was significant after univariable analyses (P =0.003). Hazard ratios were 2.0 and 5.0 for good and poor responders comparedwith complete responders. Median MFS for complete responders was not reached (3-year MFS 75 percent): 45 months for good responders, and 11 months for poor responders (Figure 1a, b). For all oncological outcome parameters, both response and complete resection were the most independent predictive variables. By definition, all complete responders have an R0 resection. However, there also is a linear relation between increasing response and the rate of R0 resections (P = 0.026, Figure 2). Therefore, both variables were combined to see whether outcome prediction can be fine-tuned and may help to make treatment decisions. Five new categories were analysed: pCR (automatically R0), good response and R0 resection, poor response and R0 resection, good response and R1/R2 (R+) resection, and finally, poor response and R+ resection. In Figure 1c, the outcome of combining these variables is shown. Complete responders consistently demonstrate an excellent outcome for OS, MFS, and LRFS. In addition, patients with a good response and R0 resection (n = 45) show an outcome almost as good as patients with a pCR for all outcome parameters. Poor responders with an R0 resection (n = 16) have a less favourable metastasis free survival compared with good responders with an R+ resection (n = 29). Patients with a poor response and R0 resection have a similar LRFS compared with patients with a good response but R+ resection.