The interventions were reasonably described. The authors indicated that standard practice had changed since the studies that informed the model were conducted, so the results may not be generalisable to current practice. Most of the sources included comparisons to the proton pump inhibitor omeprazole, it is unclear why this intervention was not considered.
There was no description of how the studies were identified or selected; it is unclear whether the best available evidence was used. The derivation of the clinical parameters, and the design of the studies, were not reported, so it is hard to assess the clinical evidence. The authors indicated that the trials were conducted before the use of intravenous proton pump inhibitors, which could have exaggerated the absolute risk difference between second endoscopy and no second endoscopy. They acknowledged that if the absolute risk of bleeding was lower after the first endoscopy, which is likely given that proton pump inhibitors reduce bleeding, then there could be little or no difference in bleeding; the only outcome that favoured second endoscopy. Also, bleeding could have other outcomes and comorbidities. The analysis did not consider outcomes such as aspiration pneumonia, stroke, myocardial infarction, and renal failure; nor any exacerbation of current comorbidities.
The cost analysis had a narrow perspective, as it only included the direct hospital costs for procedures and hospitalisation. This did not include any costs for comorbid events and exacerbations, nor any measure of hospital stay. These limitations were acknowledged by the authors, but it was not clear how they identified and selected their costs. CMS costs are based on reimbursement, which may not be the same as the cost to the provider. The cost year was not reported, nor were any methods used to normalise the prices from different sources.
Analysis and results:
The analysis was clearly reported. The limitations were acknowledged, and appropriate conclusions were reached. Appropriate comparisons were made with other work, including a study of proton pump inhibitors and second-look endoscopy for a high-risk patients (see Other Publications of Related Interest). Given the age of many of the sources, the results may not be relevant to current practice, as acknowledged by the authors.
The study was limited as it did not include current practice, and had a narrow costing perspective. The limitations were acknowledged and the conclusions seem reasonable.