The review clearly stated the research question to be addressed. However, the inclusion criteria used to define the relevant data set were not clearly defined. The search strategy was limited to two bibliographic databases, one of which (the Cochrane Controlled Trials Register) was unlikely to identify the type of studies apparently being sought (diagnostic cohorts) since these are essentially observational studies. The search was further restricted to studies published in English and it is therefore possible that relevant data were missed. The review methods were incompletely reported, so it is not certain whether adequate efforts were made to reduce reviewer errors and bias.
Relevant details of the included studies and their results, in terms of reported test performance, were reported in the text and a table. Due to the lack of details of the data extraction process and inclusion criteria relating to the outcome measures, it is not possible to assess whether the authors derived the reported results from 2x2 contingency data, or took them directly from reports of the included studies; the inclusion of erroneously calculated estimates of diagnostic performance cannot, therefore, be ruled out. The methodological quality of the included studies was assessed using criteria relevant to studies of test accuracy and the studies were assigned an evidence level (I, II, or II). Level III studies appear to have been excluded from the review, although evidence level was not specified as an inclusion criterion.
The authors' conclusion, that single-field fundus photography can be used as a screening tool for diabetic retinopathy, is very much dependent upon what is considered the lower threshold for satisfactory test performance in a screening programme of this type (an area identified by the authors as needing further research).