Interventions:
The interventions were clearly described and included standard care in Ontario, which was the study setting. Additional tests not routinely used in Ontario were discussed and their exclusion was justified. These included 14C UBT testing and a faecal antigen test.
Effectiveness/benefits:
Full details of the economic literature review strategy were provided with sufficient detail. But it was unclear whether clinical data used to populate the model (sensitivity, specificity and prevalence) were derived from this evidence. There were no details of a clinical review.
The authors acknowledged that false positives and false negatives were intermediate outcomes and that a better outcome would be related to patient quality of life. Again, it was not clear that any attempt was made to review the appropriate literature to identify relevant quality of life outcomes. The authors also indicated that dyspepsia was a risk factor for a number of cancers so long-term outcomes may also be important. The time horizon of the model was not sufficient to capture more than the intermediate test outcomes.
Costs:
Costs were reported clearly using appropriate sources and methodology. The price year was not reported and this limited both comparisons with other study results and transferability of the results to other settings. The costing was limited to test costs due to the GP payment method in Ontario and the perspective adopted; these issues should be considered when translating the results to other settings. GP costs were assumed to be zero but the number of GP visits required differed between tests.
Analysis and results:
The model structure and inputs were presented clearly. The study results were clearly reported and explained in context. The authors acknowledged limitations related to outcome measures and the model time horizon.
The one-way sensitivity analyses conducted were reported clearly but a full probabilistic analysis that allowed all parameters to vary simultaneously would give a better idea of the effect of uncertainty on model results.
Concluding remarks:
The analysis was generally well reported and methodologically appropriate. There were some limitations: lack of detail for clinical input parameters, an insufficient time horizon for capturing long-term outcomes, use of intermediate outcomes and limited analysis of uncertainty.