Assessment of prognosis in hepatocellular carcinoma (HCC) using biomarkers and serum measurements.
e15093 Background: Factors that influence prognosis in HCC are well-recognised. They include tumour size and number, the severity of liver dysfunction and the symptomology. Several staging systems combining a number of these factors offer an estimate of prognosis. Some concerned that these staging systems encompass factors that are inherently subjective, e.g. the extent of HCC in CLIP or the use of performance status in BCLC. This have led to the development BALAD score by Toyoda et al, using serum Bilirubin and Albumin and three biomarkers: alpha-fetoprotein-L3 (L3), Alpha-fetoprotein (AFP) and Des-gamma-carboxy prothrombin (DCP). This system was developed in a Japanese population which has mainly HCV-related HCC. Our aim was to develop an objective staging system using the three biomarkers for a western population and compared it to BALAD to assess its performance. Methods: 317 subjects with HCC were recruited from Birmingham, UK between 2007 and 2011. Blood samples were collected at the time of diagnosis. Liver, renal functions plus AFP, L3 and DCP were measured. L3 and DCP were measured by microchip capillary electrophoresis and liquid-phase binding assay on a uTASWako i30 auto analyzer (Wako Pure Chemical Industries Ltd, Japan). Results: Univariable and multivariable regression analyses were performed. Our scoring system (CRCTU score) was developed based on a flexible baseline hazard functions fitted using restricted cubic splines. Suitable fractional polynomials (FP) were used to explain relationship between the outcome and covariates. The level of discrimination achieved using the risk groups was assess by Harrell’s–C statistic. The CRCTU score permitted identification of 4 clearly distinct subgroups as indicated both graphically and by Harrell’s-C values. Both the CRCTU and BALAD scores return Harrell’s-C values of 0.7. Conclusions: We have shown that it is possible to use entirely objective markers to predict prognosis of HCC in a western population. Our model is comparable to the BALAD with the same variables included. However, our treatment of variables with FP is much more robust statistical technique for continuous variable as compared to having artificial categories that were seen in the BALAD system.