Application of a time-varying covariate model to the analysis of CA 19–9 as a biomarker for time-to-progression (TTP) and overall survival (OS) in patients with advanced pancreatic cancer
e15545 Background: It remains unclear whether baseline CA 19–9 or CA 19–9 kinetics during chemotherapy may serve as predictive biomarker in patients (pts) with pancreatic cancer (PC). Methods: Main inclusion criteria for this retrospective multicenter analysis: histologically confirmed diagnosis of PC, treatment with first-line therapy, pre-treatment CA 19–9 level of > 5.2 U/ml. Analysis of CA 19–9 was exclusively performed using the Elecsys® assay (Roche Diagnostics). The effect of the pre- treatment CA 19–9 level on TTP and OS was modelled by Cox proportional hazards regression. The effect of CA 19–9 kinetics was also modelled by Cox proportional hazards regression where CA 19–9 was treated as time-varying covariate. When modelling CA 19–9 we developed univariate and multivariate Cox models where we selected additional predictors (e.g. performance status) using backward elimination performing likelihood ratio tests on a significance level of 0.05. Results: One-hundred and fifteen pts from 5 German centers were included. Median age was 63 years, 12% had locally advanced and 88% metastatic disease; 73 % of the pts were treated within prospective clinical trials. Median baseline CA 19–9 was 1059 U/ml (range 9.5–100000), median pre- treatment bilirubin 0.6 mg/dl. The median TTP in the study population was 4.4 months, median OS 9.4 months. Univariate analysis showed that the pre-treatment CA 19–9 level (as continuous variable, log [CA 19–9]) was significantly associated with TTP (HR 1.24, 95% CI 1.12–1.37, p<0.001) and OS (HR 1.16, 95% CI 1.06–1.28, p=0.002). These associations remained significant also within a multivariate analysis. For CA 19–9 kinetics during chemotherapy, data from 69 pts (TTP) and 84 pts (OS) were available, respectively; log [CA 19–9] kinetics were found to be a significant predictor for TTP in univariate (HR 1.44, 95% CI 1.25–1.67, p<0.001) and multivariate (HR 1.39, 95% CI 1.19–1.62, p<0.001) analyses, and also for OS (univariate: HR 1.34, 95% CI 1.20–1.49, p<0.001; multivariate: HR 1.39, 95% CI 1.23–1.57, p<0.001). Conclusions: According to this new statistical model, CA 19–9 may serve as a useful predictive biomarker in advanced PC. [Table: see text]