Both of the interventions were well reported. Both brinzolamide and brimonidine were widely used in contemporary primary care practice in the UK, as shown by the data from the UK GPRD.
The efficacy data were derived from a large retrospective cohort study, the UK GPRD. This design is associated with many potential biases, such as inclusion bias and differences between the study groups at baseline. In this study, although the patients were comparable in terms of age and gender, those in the brimonidine group had longer follow-up (owing, according to the authors, to an earlier market access authorisation) and significantly lower number of general and eye co-morbidities. In order to control for these biases, the authors used a series of regression analyses, adjusted using the propensity scores. These methods were reported in detail, and both the adjusted and non-adjusted efficacy results were reported.
The perspective was well reported. Given this NHS perspective, it would appear that all the major cost categories and costs were included. The authors appropriately reported their methods for obtaining the resource quantities and unit costs. As with the efficacy estimates, given the differences between the groups at baseline, the cost differences were adjusted using regression analyses and both the adjusted and non-adjusted results were reported. The time horizon of the analysis was four years, therefore, discounting was relevant. However, no discounting was reported.
Analysis and results:
The costs and benefits were not combined as brinzolamide was found to be dominant. All the differences between the groups were adequately tested using appropriate statistical methods and adequate adjustments were made for the differences in patients’ baseline characteristics. The authors discussed at length the limitations of their analysis, citing the retrospective design as a major limitation. Overall the analysis was well reported.
Although the authors used a retrospective study design, the quality of the methodology was good, with the authors using appropriate statistical methods to control for biases. Both the methods and results were well reported. The qualified conclusions of the authors appear to be appropriate, given their study design.