ABSTRACT
OBJECTIVES: Epidemiological data have shown that individuals with advanced fibrosis are at greatest risk of premature morbidity in non-alcoholic fatty liver disease (NAFLD). Individuals included in clinical trials are often highly selected to remove confounding factors, but selection can introduce bias and limit external validity. We examined the external validity of trials in NAFLD by examining characteristics of participants in observational studies (OS) and randomized controlled trials (RCT) in NAFLD. DESIGN: A systematic review was performed with structured literature searches for relevant OS and RCT using PubMed and Ovid Embase (1948-2016). Identified studies were screened for inclusion by the authors and data extracted. Study populations were compared using t-tests to compare means and variances, in each case weighted by the size of individual studies. Dichotomous data were compared by Chi-squared test. RESULTS: In total, 148 studies were included: 67 RCT and 81 OS including data from 44 860 individuals. Fifteen RCT participants differed from individuals in OS with regard to age, body mass index, prevalence of Diabetes mellitus, and gender (P < 0.001 in each case). The most pronounced differences were seen between RCT participants and patients with advanced fibrosis. Comorbid conditions prevalent among individuals with NAFLD were frequent exclusion criteria in RCT. CONCLUSIONS: The characteristics of participants in randomized controlled trials differ to those of the wider population of individuals with NAFLD. These differences may reduce the utility of trial data to individuals with NAFLD at greatest risk of death.
Subject(s)
Liver Cirrhosis , Non-alcoholic Fatty Liver Disease/therapy , Patients , Age Factors , Body Mass Index , Chi-Square Distribution , Comorbidity , Databases, Bibliographic , Diabetes Mellitus , Humans , Liver Cirrhosis/epidemiology , Non-alcoholic Fatty Liver Disease/epidemiology , Observational Studies as Topic , Randomized Controlled Trials as Topic , RiskABSTRACT
BACKGROUND: Prognostic scores are used to assess the likelihood of mortality in cirrhosis and the necessity of liver transplantation. These models are imperfect and refinement would allow more accurate prognostication and selection of patients for transplant. This study investigated association of red cell parameters and mortality in liver transplant candidates. METHODS: Data from patients with cirrhosis assessed for transplantation from 2008 to 2010 at Queen Elizabeth Hospital Birmingham, UK were reviewed retrospectively. Kaplan-Meier analysis and Cox regression models were used to generate indices predicting mortality. Accuracy of existing and updated models was tested by calculation of c-statistics. Results were validated in a cohort of patients assessed for liver transplant in Jena, Germany. RESULTS: Data were collected from 386 patients in the study cohort. Median follow-up was 15 months (0-45). During follow-up, 151 patients (39%) were transplanted, 138 (36%) died, and 97 (25%) survived without transplant. Abnormal reticulocyte count (P<0.001, c-statistic 0.623) and hemoglobin concentration (P<0.001, c-statistic 0.609) predicted mortality in Cox regression analysis. Abnormal reticulocyte count was also found to predict mortality in competing risk analysis. Refining the Model for End-Stage Liver Disease (MELD) to incorporate reticulocyte count and hemoglobin concentration (MELD-red) improved predictive power from 0.701 to 0.731 (c-statistics). This was confirmed in an independent validation cohort of 157 patients with c-statistics of 0.787 and 0.816, respectively, for MELD and MELD-red. CONCLUSIONS: Abnormal red cell indices, in particular increased reticulocyte count and decreased hemoglobin concentration, are associated with increased risk of death in liver transplant candidates. Refining MELD to incorporate these indices improves prediction of mortality.
