gms | German Medical Science

G-I-N Conference 2012

Guidelines International Network

22.08 - 25.08.2012, Berlin

Association between risk of bias and treatment effects in randomized controlled trials in emergency and critical care medicine

Meeting Abstract

  • S. Unverzagt - Martin-Luther University Halle/Wittenberg, Halle, Germany
  • R. Prondzinsky - Carl-von Basedow-Klinikum, Merseburg, Germany
  • F. Peinemann - University of Cologne, Cologne, Germany

Guidelines International Network. G-I-N Conference 2012. Berlin, 22.-25.08.2012. Düsseldorf: German Medical Science GMS Publishing House; 2012. DocO19

doi: 10.3205/12gin051, urn:nbn:de:0183-12gin0518

Published: July 10, 2012

© 2012 Unverzagt et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License ( You are free: to Share – to copy, distribute and transmit the work, provided the original author and source are credited.



Methodological deficiencies are known to affect the results of randomized controlled trials (RCTs) and systematic reviews. This study investigates the degree to which specific quality components are associated with treatment effects on mortality within the emergency and critical care medicine. Risk of bias was measured by the Cochrane Collaboration's assessment tool and six additional components. Detailed instructions on judgments were defined on the basis of the nature of selected indications and outcome. Eleven meta-analyses of Cochrane-reviews that involved 77 RCTs on patients with sepsis, septic or cardiogenic shock were selected. Risk of bias is currently judged by a second author. As biases in different quality components often co-occur, suspected biases in eleven quality components are evaluated simultaneously. Hierarchic logistic models are used to investigate the influence of risk of bias in eleven quality components on the estimated treatment effect with adjustments for three sampling levels. Initial results from single judgments seem to indicate an overestimation of treatment effects in trials with contra-active or supporting interventions before randomization, inadequate selective outcome reporting or inadequate use of early stopping. Diminished estimates of treatment effects were found in trials with incomplete reporting of mortality, inadequate double blinding or cross-over. Biases in different quality components which affect treatment estimates are unlikely to operate independently, confounding between quality components may have modified these associations. Investigated sources of bias can lead to exaggerated or diminished estimates of the treatment benefit and alter the conclusions about intervention effects in intensive care and emergency medicine.