Individual therapy-dependent prognosis based on data from controlled clinical trials in chronic liver disease

In a given disease the manifestations and course of disease may vary markedly between the patients. This complicates prediction of the prognosis and treatment effect in individual patients. Most controlled clinical trials present only the "average" effect e.g. the therapy-dependent survival in the studied patient group. To estimate the therapy-dependent prognosis in individual patients it is necessary to utilize the covariation between survival time and variables characterizing each patient including the therapy given. This paper describes current methods for identification of variables which covary with survival time (prognostic variables) or the effect of therapy (therapeutic variables). Analyzing data from two large controlled clinical trials in patients with chronic liver disease: 1) the multicenter trial of prednisone versus placebo in cirrhosis conducted by the Copenhagen Study group for Liver diseases (CSL-1) and 2) the multinational trial of azathioprine versus placebo in primary biliary cirrhosis (PBC-1) we have developed indices for prediction of prognosis and therapeutic effect using Cox's multiple regression model for censored survival data. Using the indices one can estimate the therapy-dependent prognosis in new patients from their base-line data. Furthermore, a time-dependent index by which the risk of a given patient can be estimated repeatedly to update prognosis during the course of the disease is presented. To simplify application "pocket charts" have been devised by which a prognostic index for a patient can easily be obtained at the bedside. By simple graphs, a prognostic index can be translated to estimates of the probability of surviving a given time or the median survival time predicted for the patient. The indices have been validated by comparing the survival predicted by the indices with the observed survival in new patients or using data splitting. The results allow a more differentiated treatment strategy based on the characteristics of the individual patient. Even if the results apply to chronic liver disease, the general principles are valid for study of the individual therapy-dependent prognosis in other diseases.