Results

A total of 281 children enrolled in the trial, with a median age of 2.6 years. Forty-three percent were female.

Key Result: A five-day course of antibiotics was inferior to a 10-day course of antibiotics in children with CAP.

• Primary Outcome: Clinical cure at 14–21 days after enrollment
  • Per-protocol (PP) analysis: 88.6 percent in the intervention group, 90.8 percent in the control group; risk difference was −0.016 (97.5 percent confidence limit −0.087) and cannot claim noninferiority
  • Intention-to-treat (ITT) analysis: 85.7 percent in the intervention group, 84.1 percent in the control group; risk difference was 0.023 (97.5 percent confidence limit −0.061)
• Secondary Outcomes: Caregivers were off work two days instead of three in the intervention group. All other secondary outcomes were the same.

Evidence-Based Medicine Commentary

1. Representative Cohort: There is a question of whether this cohort represents children with CAP presenting to the emergency department. Only 281 (5 percent) of the 5,406 children diagnosed with CAP were randomized. The study flow diagram shows researchers missed 3,215 possible children to include, suggesting they were not recruited consecutively. This also could have introduced some selection bias.

2. Chest X-Ray: This is not needed to make the diagnosis of CAP in children, and it is actively discouraged by the IDSA guidelines.3

3. Clinical Cure: Their definition of clinical cure included some subjective criteria. Different physicians could have different interpretations on what a “significant improvement” looked like clinically and if the child required additional antibiotics or hospital admission. This could have introduced uncertainty into the data.

4. Statistical Versus Clinical Outcome: This was a noninferiority trial, and they correctly performed a per-protocol analysis. The noninferiority margin was based on several assumptions. Because the one-sided 97.5 percent confidence limit of the point estimate of 7.5 percent was exceeded, a formal conclusion of noninferiority could not be made.

However, this is a statistical outcome and may not be a clinically important difference. Physicians will need to interpret the finding for themselves and think about how to apply the data. Both groups had about a 90 percent clinical cure rate, with only a 1.6 percent absolute risk difference between the five- and 10-day course of antibiotics. Will crossing a one-sided, and seemingly arbitrary, statistical barrier by 1.2 percentage points (7.5 versus 8.7 percent) make a difference in clinically applying this data?

5. External Validity: This trial was conducted at two pediatric emergency departments in Canada. It is unclear if these represent similar patients presenting to community emergency departments, rural emergency departments, or facilities in other countries.