The delineation of valid and practical treatment failure criteria is a major challenge in childhood pneumonia research. Composite outcomes may obscure important differences in the nature of treatment effects. Therefore, it would be very helpful if the authors displayed a table showing the proportions of failures in each treatment group who fulfilled each of the 5 specific failure criteria. Since the individual criteria differed greatly in their relative severity and almost certainly represent heterogeneous underlying pathology, it is difficult to assess the nature of the actual clinical benefit provided by amoxicillin. For example, the development of severe or very disease would be almost universally considered an adverse short-term outcome (i.e., likely predictive of prolonged hospitalization and/or death), whereas persistence of wheeze may have entirely different health implications (e.g., suggestive of underlying asthma).
