Critical Care
Use of L’Abbé and pooled calibration plots to assess the relationship between severity of illness and effectiveness in studies of corticosteroids for severe sepsis

https://doi.org/10.1093/bja/aeq417Get rights and content
Under an Elsevier user license
open archive

Background

The role of corticosteroids in severe sepsis and pneumonia remains controversial. This study described the use of L’Abbé and pooled calibration plots to assess the relationship between severity of illness and effectiveness of corticosteroids for severe sepsis.

Methods

Randomized controlled trials (RCTs) comparing corticosteroids and placebo from Cochrane Controlled Trial Register, MEDLINE, and EMBASE databases were retrieved. The observed and predicted mortality rates of the placebo groups were used as a measure of the severity of illness of the patients in L’Abbé and calibration plots.

Results

A total of 1089 patients from 10 RCTs fulfilled the inclusion criteria and were subject to further analysis. L’Abbé and calibration plots did not suggest significant interactions between the effectiveness of corticosteroids and severity of illness. The pooled calibration plot suggested that the mortality rates of the placebo groups from three studies were higher than predicted. After excluding these studies in the meta-analysis, there was a reduction in the point estimate of benefit of corticosteroids on mortality [odds ratio (OR) 0.97, 95% confidence interval (CI): 0.71–1.33, P=0.87 by a fixed-effect model, P=0.59 by a random-effects model vs OR 0.85, 95% CI: 0.66–1.10%].

Conclusions

The pooled calibration plot suggested that there were excessive deaths in the placebo groups of some RCTs that could explain the apparent benefit of corticosteroids on mortality of patients with severe sepsis. L’Abbé and pooled calibration plots might be useful as adjuncts to assess interactions between severity of illness and effectiveness of an intervention.

Key words

clinical trials
predicted mortality
quality of study
standardized mortality ratio

Cited by (0)