Article type
Year
Abstract
Background: In a Cochrane systematic review of Intravenous immunoglobin (IVIG) for preventing infection in preterm and/or low-birth-weight infants, the major outcome of any serious infection (one or more episodes) has previously been analyzed using standard meta-analysis techniques for binary outcome as implemented in the RevMan software. This review synthesized results of sixteen clinical trials published over a period of fourteen years from 1986 to 2000 and enrolling approximately 5000 neonates. Despite the accumulation of so much data, the evidence in favour of IVIG treatment has not been quantified analytically.
Objective: The aim of this study was to apply Bayesian methodology to quantify the strength of the evidence in support (or lack thereof) of the prophylactic use of IVIG treatment for preventing serious infection in preterm and/or low-birth-weight infants.
Methods: We applied a Bayesian logit model to the data from the sixteen clinical trials and obtained posterior distribution for the log odds of the probability of infection. Non-informative priors were used for the hyper-parameters and the model was implemented using WinBUGS. We also compared the Bayesian results with conclusions from cumulative meta-analysis.
Results: The area under the posterior distribution for the log odds corresponding to the region favouring the control group that received a placebo or no intervention to the IVIG group was only 0.005. This figure quantifies the strength of the evidence from the accumulated data to date, and leaves little doubt about the prophylactic benefit of IVIG in preventing serious infection in the population under study. In addition to the Bayesian analysis, a cumulative meta-analysis also showed that IVIG group has been statistically significantly associated with better outcome than the control group as of 1991.
Conclusions: Our Bayesian analysis supports the conclusion from a cumulative meta-analysis favouring IVIG and that there is no justification for conducting new studies to test the prophylactic efficacy of IVIG to prevent serious infection in preterm and/or low birth weight infants.
Objective: The aim of this study was to apply Bayesian methodology to quantify the strength of the evidence in support (or lack thereof) of the prophylactic use of IVIG treatment for preventing serious infection in preterm and/or low-birth-weight infants.
Methods: We applied a Bayesian logit model to the data from the sixteen clinical trials and obtained posterior distribution for the log odds of the probability of infection. Non-informative priors were used for the hyper-parameters and the model was implemented using WinBUGS. We also compared the Bayesian results with conclusions from cumulative meta-analysis.
Results: The area under the posterior distribution for the log odds corresponding to the region favouring the control group that received a placebo or no intervention to the IVIG group was only 0.005. This figure quantifies the strength of the evidence from the accumulated data to date, and leaves little doubt about the prophylactic benefit of IVIG in preventing serious infection in the population under study. In addition to the Bayesian analysis, a cumulative meta-analysis also showed that IVIG group has been statistically significantly associated with better outcome than the control group as of 1991.
Conclusions: Our Bayesian analysis supports the conclusion from a cumulative meta-analysis favouring IVIG and that there is no justification for conducting new studies to test the prophylactic efficacy of IVIG to prevent serious infection in preterm and/or low birth weight infants.