large, randomized, controlled clinical trials. The observa- tion in the current study of S aureus, K pneumonia, and S pneumonia also suggests that the gastrointestinal tract may not be the only source of systemic circulating bacteria in CD. Currently, clinicians are unable to accurately predict the behavior of CD in an individual. Previous studies have focused on identifying clinical parameters, biomarkers, and environmental factors as predictors of CD relapse. These include young age at diagnosis, an early need for steroids, perianal involvement (Gastroenterology 2006; 130:650–665), use of nonsteroidal anti-inflammatory drugs (Clin Gastroenterol Hepatol 2006;4:196–202), fecal calprotectin (Gastroenterology 2000;119:15–22), and es- trogen use and smoking (Gastroenterology 1998;114: 1143–50), all of which are associated with the relapse of CD. Unfortunately, these lack specificity, limiting clin- ical usefulness. Similarly, with the large number of genetic risk loci for CD, susceptibility genes associated with the development of CD have also been investigated as factors predicting the relapse of CD. Some such as the NOD2 polymorphism have been associated with a more aggressive clinical course of CD (Gastroenterology 2002; 122:2091–2092), but none have been linked with the relapse of CD. It is unlikely that genetic markers in isolation will be able to fully predict the relapsing nature of CD given the major influence of nongenetic variables. However, in combination with other variables it may be possible to develop such predictive tools and models. In support of this the multivariate analysis by Gutiérrez et al observed that bactDNA translocation was an independent risk factor for the relapse of CD and that NOD2 and ATG16L1 variants in a subgroup of patients appeared to facilitate bactDNA translocation. In randomized, controlled trials between 30% and 40% of responders to anti–TNF-a experience a relapse of their symptoms and are unable to maintain clinical response to anti–TNF-a treatment after 1 year (Gastroenterology 2007;132:52–65). The mechanisms behind this are poorly understood but increasing the frequency and/or dose of anti–TNF-a agents (dose escalation) can restore clinical response (J Clin Gastroenterol 2011;45:113–118; Gut 2007;56:1232–1239). In an effort to determine predictors of dose escalation, studies have identified age, male gender, noncolonic CD, extraintestinal manifestations, family history of CD, and smoking as clinical parameters that positively predict the need for anti–TNF-a dose escalation (Inflamm Bowel Dis 2012;18:10–16, Inflamm Bowel Dis 2012;18:685–690). The use of these parameters in a clinical setting has been limited by conflicting results from other studies and low specificity. In some patients loss of response to anti–TNF-a has been attributed to the development of antibodies against anti–TNF-a agents (N Engl J Med 2003;348:601–608) or autoantibodies (Gastroenterology 2003;125:32–39). These may exert their effects via neutralization and enhanced clearance of anti–TNF-a, although the underlying mechanisms are poorly understood and the predictive value requires large- scale, clinical studies. Gutiérrez et al showed that those with CD and NOD2/ATG16L1 genotype variants secrete higher levels of TNF-a and have lower free anti–TNF-a levels after anti–TNF-a treatment. This increased TNF-a seems to be in response to bacterial DNA stimuli, as sug- gested by in vitro experiments. Excess secretion of TNF-a leading to consumption of anti–TNF-a agents abrogating the effects of anti–TNF-a is supported by other studies reporting that undetectable serum trough concentrations of anti–TNF- a agents (Clin Gastroenterol Hepatol 2006;4:1248–1254) and decreased free TNF-a binding capacity of anti–TNF-a agents (Am J Gastroenterol 2008;103:944–948) predict poorer responses to anti–TNF-a agents. In conclusion, this important study links NOD2 and ATG16L variants, loss of barrier function, and circu- lating systemic bacteria with response to anti–TNF-a agents. It highlights the important pharmacologic point that high circulating TNF-a reduces the efficacy of therapeutic anti–TNF-a antibodies by binding and reducing levels of free anti–TNF-a antibodies. Further understanding of the relationship between genotype, bacterial entry, and immune response will enable a more rational approach to the therapy of CD with anti–TNF-a and other biological therapies. SIMON S. M. CHAN ALASTAIR J. M. WATSON Norwich Medical School University of East Anglia Norwich, England LIVER REGENERATION DURING ACUTE- ON-CHRONIC LIVER FAILURE USING GROWTH FACTORS: IN VIVO OR EX VIVO INDULGENCE OF BONE MARROW? Duan XZ, Liu FF, Tong JJ, et al. Granulocyte colony stimulating factor therapy improves survival in patients with hepatitis B virus-associated acute-on-chronic liver failure. World J Gastroenterol 2013;19:1104–1110. In the current prospective, double-blind, randomized, controlled trial, Duan et al (World J Gastroenterol 2013;19:1104–1110) aimed to evaluate the safety and ef- ficacy of granulocyte-colony stimulating factor (G-CSF) therapy in patients with hepatitis B virus (HBV)-associated acute-on-chronic liver failure (ACLF). Fifty-five treatment- naive patients with HBV-associated ACLF were randomized into 2 groups; the treatment group (n ¼ 27) received G-CSF (5 mg/kg per day; 6 doses) treatment plus standard medical therapy, and patients in the control group (n ¼ 28) received standard medical therapy only. All patients received entecavir (0.5 mg/d). The peripheral CD34 þ cell count was measured consecutively by flow cytometry in both the treatment and the control groups. Sera from patients were also collected and direct polymerase chain reaction sequencing was used to screen for HBV reverse transcriptase domain if serum HBV DNA tests were positive after patients received entecavir therapy. A total of 55 patients had HBV-associated ACLF as October 2013 SELECTED SUMMARIES 901