Univariate Cox regression analyses including treatment as a covariable were performed, and identified a number of factors associated with reduced TTP. These Carfilzomib chemical structure included poor Karnofsky performance status (KPS; 70 vs 100%), liver metastases, multiple metastatic sites and younger age, with prognostic significance at a level of 15% (Table 4 ). Previous adjuvant treatment and gender did not appear to impact on disease progression. A multivariate Cox regression analysis using backward elimination of factors, with the first step including all factors identified in the univariate analyses, was then performed to assess the impact of independent prognostic factors on the treatment effect at a significance level of 5%. This analysis confirmed an increased risk of progression with lower age, liver metastases, poor KPS (70 or 80% at baseline) and multiple metastatic sites (Table 5 ).

Of note, the analyses confirmed that risk of disease progression was independent of treatment, as indicated by the HR of 1.02 (95% CI 0.90�C1.15; P=0.79). Table 4 Results of univariate Cox regression analysis: TTP Table 5 Results of multivariate Cox regression analysis: TTP Overall survival Overall survival data, updated in June 2002 after 1147 events, confirm the integrated analysis results reported by Twelves (2002). In patients receiving capecitabine, overall survival was equivalent to that in patients treated with 5-FU/LV (HR 0.95, 95% CI 0.84?1.06, P=0.48). The median survival was 12.9 months (95% CI 11.8?14.0) in the capecitabine group and 12.8 months (95% CI 11.7?14.

0) in the 5-FU/LV group after 583 and 564 events, respectively (Figure 3). Univariate analyses identified nine prognostic factors for survival with a significance level of 15%. Multivariate Cox regression analysis confirmed that lower KPS (70 or 80 vs 100%), multiple metastatic sites and the presence of liver metastases were independent prognostic indicators for poor survival (Table 6 ). Patients with well-differentiated tumours had a slightly reduced risk of death compared with patients having less well-differentiated tumours. The data also suggest that previous surgery may have a prognostic effect, but the number of patients who did not have prior surgery was extremely small, as approximately 90% of patients had received prior surgery.

Similarly, multivariate Cox regression analysis suggested that risk of death was markedly reduced in patients defined as race ��other�� compared with the rest of the population (HR=0.64). However, this subgroup, comprising multiple ethnic groups (including Orientals and GSK-3 Hispanics), included only 35 patients in the capecitabine arm and 34 patients in the 5-FU/LV arm, and thus the results are difficult to interpret. Of note, the analyses confirmed that survival was independent of treatment.