Meta-analysis: testing for reporting bias

Researchers undertook a meta-analysis to evaluate the effectiveness of oral decontamination on the incidence of ventilator associated pneumonia in mechanically ventilated adults. Randomised controlled trials that recruited mechanically ventilated adults and compared the effects of a daily oral application of antibiotics or antiseptics with no prophylaxis were included.1

In total, 11 trials involving 3242 patients met the inclusion criteria. When the results of these trials were pooled, rates of ventilator associated pneumonia were significantly lower in patients who received oral decontamination than in those who received no prophylaxis (relative risk 0.61, 95% confidence interval 0.45 to 0.82). The researchers constructed the funnel plot for the primary outcome of ventilator associated pneumonia, and visual inspection suggested that it was asymmetrical, indicating the existence of reporting bias. A formal statistical test did not, however, support the presence of reporting bias (Egger’s test, P=0.111). Subgroup analysis showed that the incidence of ventilator associated pneumonia was significantly reduced with a daily oral application of antiseptics (0.56, 0.39 to 0.81) but not with antibiotics (0.69, 0.41 to 1.18).

The researchers concluded that oral decontamination of mechanically ventilated adults with a daily oral application of antiseptics is associated with a lower risk of ventilator associated pneumonia.

Which of the following types of bias, if any, would have contributed to reporting bias for the results of the meta-analysis?

• a) Citation bias

• b) Language bias

• c) Publication bias

• d) Time lag bias