Follow-up information was collected until the first of the ensuing events occurred: death of the patient, loss to follow-up, transplantation or hepatectomy. Primary and secondary issues The occurrence of cytolysis following chemoembolization was our main variable of interest. We used the definition by Paye et al. for cytolysis that is an elevation

of AST above 100 UI/L with at least a doubling of the baseline value for AST (12) occurring within the first 5 days following treatment. Our primary issue was to evaluate if cytolysis was associated with a favourable radiological response Inhibitors,research,lifescience,medical two months after treatment. Our secondary issues were to investigate if cytolysis was associated with the development of hepatobiliary complications and overall survival. Liver failure was defined as the development of hepatic encephalopathy, doubling of baseline bilirubin, 25% increase in INR or appearance of ascites during the hospitalisation post TACE. Statistical analysis The statistical analysis was twofold: Inhibitors,research,lifescience,medical first, we considered the treatment outcomes using the treatment

as the unit of interest, as patients could undergo several sequential treatments. Second, we analysed the survival-related outcomes, this time using the patient as the unit of interest. When the unit of analysis was the treatment, generalized estimating equations (GEE) with an exchangeable correlation Inhibitors,research,lifescience,medical structure were used to account for the correlation between Inhibitors,research,lifescience,medical multiple treatments from the same patients. Continuous variables expressed as mean (standard deviation, sd) were compared with the Student t test when the unit of analysis was the patient. Similarly, categorical variables were compared using GEE at the treatment level and with the chi-square test at the patient level. We constructed Kaplan-Meier curves for the time to death according to the presence or absence of cytolysis at the time of the first treatment. Cases were censored in case of transplantation or loss to follow-up. Administrative censoring was set at 18 months after the Inhibitors,research,lifescience,medical first treatment. Association among demographic, biochemical and

prognostic Etomidate score variables was estimated using multivariable Cox’s proportional hazards regression model. Variables selected were those that are known to be associated with survival from liver cancer and liver disease and included the alphafetoprotein (AFP) levels (16,17), Okuda score (18,19) or CLIP score (18,20,21), MELD score and patient’s age. The Doxorubicin order natural logarithm of the AFP was used for the analysis to improve the fit of the regression model. To account for the impact of tumour differentiation on the response to chemotherapy, radiological response was adjusted according to the log of the alphafetoprotein levels as high AFP levels are associated with poorer tumour differentiation (22). Analyses were performed using R version 2.13.1 (The R Foundation for Statistical Computing, Vienna, Austria) statistical software and Stata v. 11.