Mild pediatric traumatic brain injury: adjusting significance levels for multiple comparisons

The outcome following mild traumatic brain injury (TBI) is controversial. We addressed this topic in a study of neurobehavioural and 'real-world' functioning among 53 children with mild TBI and their matched controls, using statistical methods recently developed for multiple comparisons. Because the study involved calculation of 414 p-values, four methods of adjustment for multiple comparisons, including the Bonferroni method, were used to avoid 'false-positive' statistical significance. The additional three methods allowed greater insight into the data than provided by the standard Bonferroni adjustment. Results showed that at initial testing, three areas of 'real-world' functioning (eating, domestic and home/community living skills) had weak but statistically significant associations with mild injury. No other specific areas of neuro-behavioural or 'real-world' functioning had plausible associations with the injury either initially, at 1 year, or when changes over the year were considered. However, the adjustment for multiple comparisons provided additional results. There were statistical significant associations of the injury with the entire domain of neurobehavioural variables both initially and at 1 year. These results suggest that the injury affects a spectrum of neurobehavioural skills weakly, rather than a single area substantially. The gain from using multiple comparison methods is discussed.