The review question was broadly defined and inclusion criteria reported. Several relevant sources were searched to 2009 with no language restriction. It was unclear whether attempts were made to locate unpublished studies. Methods to reduce reviewer error and bias were only used for data extraction.
Study quality was not assessed systematically, which made it difficult to determine the reliability of the evidence presented. Over half the studies were case series which were liable to multiple biases. A narrative synthesis appeared appropriate given the variability between the studies in dosage, treatment durations, assessment tools, patient setting, sample sizes and study designs. Details of the included patients were sparse. Results were based on statements by study authors, and some without corresponding levels of statistical significance, which made it difficult to verify the findings of the review.
The authors stated that, as no side effects were reported in the studies, it was reasonable to assume that melatonin was safe and well-tolerated at the dosages studied. However, this may be due to a lack of reporting in the original studies and the statement may not be reliable. The authors also offered some hypotheses on the effectiveness of melatonin on sundowning/agitated behaviour in patients with delirium, but these were not derived directly from the evidence in the included studies.
The authors’ conclusions appear overly strong given the limitations of the evidence presented.