OBJECTIVES: To determine the best photographic surrogate markers for detecting sight-threatening macular oedema (MO) in people with diabetes attending UK national screening programmes. DESIGN: A multicentre, prospective, observational cohort study of 3170 patients with photographic signs of diabetic retinopathy visible within the macular region [exudates within two disc diameters, microaneurysms/dot haemorrhages (M/DHs) and blot haemorrhages (BHs)] who were recruited from seven study centres. SETTING: All patients were recruited and imaged at one of seven study centres in Aberdeen, Birmingham, Dundee, Dunfermline, Edinburgh, Liverpool and Oxford. PARTICIPANTS: Subjects with features of diabetic retinopathy visible within the macular region attending one of seven diabetic retinal screening programmes. INTERVENTIONS: Alternative referral criteria for suspected MO based on photographic surrogate markers; an optical coherence tomographic examination in addition to the standard digital retinal photograph. MAIN OUTCOME MEASURES: (1) To determine the best method to detect sight-threatening MO in people with diabetes using photographic surrogate markers. (2) Sensitivity and specificity estimates to assess the costs and consequences of using alternative strategies. (3) Modelled long-term costs and quality-adjusted life-years (QALYs). RESULTS: Prevalence of MO was strongly related to the presence of lesions and was roughly five times higher in subjects with exudates or BHs or more than two M/DHs within one disc diameter. Having worse visual acuity was associated with about a fivefold higher prevalence of MO. Current manual screening grading schemes that ignore visual acuity or the presence of M/DHs could be improved by taking these into account. Health service costs increase substantially with more sensitive/less specific strategies. A fully automated strategy, using the automated detection of patterns of photographic surrogate markers, is superior to all current manual grading schemes for detecting MO in people with diabetes. The addition of optical coherence tomography (OCT) to each strategy, prior to referral, results in a reduction in costs to the health service with no decrement in the number of MO cases detected. CONCLUSIONS: Compared with all current manual grading schemes, for the same sensitivity, a fully automated strategy, using the automated detection of patterns of photographic surrogate markers, achieves a higher specificity for detecting MO in people with diabetes, especially if visual acuity is included in the automated strategy. Overall, costs to the health service are likely to increase if more sensitive referral strategies are adopted over more specific screening strategies for MO, for only very small gains in QALYs. The addition of OCT to each screening strategy, prior to referral, results in a reduction in costs to the health service with no decrement in the number of MO cases detected. STUDY REGISTRATION: This study has been registered as REC/IRAS 07/S0801/107, UKCRN ID 9063 and NIHR HTA 06/402/49. SOURCE OF FUNDING: This project was funded by the NIHR Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 17, No. 51. See the HTA programme website for further project information.