Therapeutic Apheresis and Dialysis 10(2):175–179, Blackwell Publishing Asia Pty Ltd © 2006 International Society for Apheresis 175 Blackwell Publishing AsiaMelbourne, AustraliaTAPTherapeutic Apheresis and Dialysis1774-99792006 International Society for Apheresis? 2006 102 175179 Original Article Design of Appropriate Clinical StudiesW Bernal and J Wendon Received July 2005. Address correspondence and reprint requests to Dr William Bernal, Liver Intensive Therapy Unit, Institute of Liver Studies, Kings College Hospital, Denmark Hill, London SE5 9RS, UK. Email: william.bernal@kcl.ac.uk Cell-Free Artificial Liver Support: Design of Appropriate Clinical Studies William Bernal and Julia Wendon Liver Intensive Therapy Unit, Institute of Liver Studies, Kings College Hospital, Denmark Hill, London, UK Abstract: The clinical use of cell-free liver support devices is dependent upon proof of safety and efficacy in clinical trials. The current published data to support their use is limited in both quantity and quality. In most studies, the methodology is such that the effects of these devices are difficult to establish with certainty. Bias might be intro- duced by the use of uncontrolled studies, and in random- ized controlled trials limited size, poorly matched control groups or medical therapy or the use of clinically inappro- priate endpoints might be important. Guidelines are now available to provide a framework for the design and exe- cution of such trials, specifically to minimize the systematic errors that are frequently present and that might result in biased estimates of treatment effects. In the present review, the limitations of current studies are discussed and sugges- tions made as to the design and conduct of future clinical trials. Key Words: Clinical trials, Critical care, Life support care, Liver failure, Methodology. BACKGROUND There can be little doubt as to the importance of the development of forms of extracorporeal liver support devices. The outcomes of standard medical care for patients with liver failure (either acute liver failure (ALF) or acute on chronic liver failure (AoCLF)) requiring intensive care therapy are at best poor, with a mortality range of 40–90% (1–3). It seems logical that if some of the functions of the failing liver can, even temporarily, be supported though the use of such devices, the possibility of hepatic regeneration and a restoration of premorbid levels of hepatic function will be increased and clin- ical outcomes improved. This clinical need has resulted in the development of a wide variety of devices for liver support in this setting, and the commercial availability of many for clinical use. The ready availability of these devices does not however, necessarily equate with clinical benefits for the patients to whom they are applied. Caution must be applied in the interpretation of the data presented in the current literature to support their use. From an evidence-based standpoint, the quality of this literature is low (Table 1). In the present article, the limitations of these studies will be analyzed in this regard; sources of potential bias will be discussed and suggestions made as to the design and conduct of future trials. CURRENT LITERATURE ON LIVER SUPPORT A recent comprehensive analysis of the literature on artificial and bioartificial liver support systems was undertaken by members of the Cochrane Col- laboration for an evidence-based meta-analysis of evidence from randomized controlled trials of bene- ficial or harmful effects of these devices (4). The authors identified 528 references in the field from 1966 to 2002. Of these, 505 (96%) were excluded as they were duplicates, non-clinical studies or did not meet prespecified inclusion criteria. Of the remaining 23 studies, nine referred to non or quasi-randomized studies and 14 to full randomized controlled trials. Of these 14, two appeared only in abstract and 12 trials were published as full randomized controlled trials (RCT). Of the 12 published trials, one referred to a 30-year-old trial of whole blood exchange and the other 11 to six different machine based therapies. Thus, of the total of 528 reports, only 11 (2%, with a median number of 42 subjects per study) can be con-