A statistical view of clinical trials in chronic hepatitis B

From a review of thirty published papers it is clear that the variety of options for treatment of hepatitis B have led to a large number of small exploratory trials. However, small groups lead to imprecise estimates of seroconversion rates; for example a result of 4/10 gives an estimate of 40% with 80% confidence limits of 20% and 60%. A number of possible prognostic factors have been proposed and, whilst their importance is generally unproven, their estimated effects are considerable. Comparison between studies can therefore be misleading; for example comparing two treatments A and B of identical efficacy in two common populations could give a 1 in 4 chance of declaring B to give double the seroconversion rate of A. Larger, randomised, controlled trials on homogeneous patient groups are needed to give conclusive evidence of the efficacy or otherwise of various treatment regimes.