On developing a pragmatic strategy for clinical trials: A case study of hepatocellular carcinoma

Randomised controlled trials (RCTs) with sufficiently high statistical power are not always feasible for patients when the administration of the treatment is burdensome. Nevertheless, useful information concerning the relative effectiveness of the Test and Standard therapies, may be gleaned from under powered trials, non-randomised comparative studies and/or clinician's beliefs: the latter possibly additionally providing some suggestion of the strength of evidence required in order to adopt the Test therapy into clinical practice. In such circumstances, a Bayesian synthesis may be useful in quantifying the evidence of treatment effectiveness. In this article, we aim to present a Bayesian approach for synthesizing the cumulative evidence of the use of adjuvant hepatic intra-arterial iodine-131-lipiodol (I131L) following curative resection in hepatocellular carcinoma (HCC) patients. We constructed a posterior distribution using the information from two small RCTs, three non-randomised comparative studies, three single arm studies and the views of investigators on the use of I131L. This distribution enables calculation of the probability that the Test therapy is more effective than the Standard by a pre-stipulated amount. If this is very high, then for example, one may conclude the Test may replace the Standard therapy. If it is not, then the Standard would be retained for clinical use. Despite a strong early indication of the effectiveness of I131L, the evolving evidence over a 10-year period became more sceptical of its value. Although highly recommended, difficulties of implementing a Bayesian approach in this context are highlighted.