BackgroundDiabetic retinopathy is an important cause of visual loss. Laser photocoagulation preserves vision in diabetic retinopathy but is currently used at the stage of proliferative diabetic retinopathy (PDR).ObjectivesThe primary aim was to assess the clinical effectiveness and cost-effectiveness of pan-retinal photocoagulation (PRP) given at the non-proliferative stage of diabetic retinopathy (NPDR) compared with waiting until the high-risk PDR (HR-PDR) stage was reached. There have been recent advances in laser photocoagulation techniques, and in the use of laser treatments combined with anti-vascular endothelial growth factor (VEGF) drugs or injected steroids. Our secondary questions were: (1) If PRP were to be used in NPDR, which form of laser treatment should be used? and (2) Is adjuvant therapy with intravitreal drugs clinically effective and cost-effective in PRP?Eligibility criteriaRandomised controlled trials (RCTs) for efficacy but other designs also used.Data sourcesMEDLINE and EMBASE to February 2014, Web of Science.Review methodsSystematic review and economic modelling.ResultsThe Early Treatment Diabetic Retinopathy Study (ETDRS), published in 1991, was the only trial designed to determine the best time to initiate PRP. It randomised one eye of 3711 patients with mild-to-severe NPDR or early PDR to early photocoagulation, and the other to deferral of PRP until HR-PDR developed. The risk of severe visual loss after 5 years for eyes assigned to PRP for NPDR or early PDR compared with deferral of PRP was reduced by 23% (relative risk 0.77, 99% confidence interval 0.56 to 1.06). However, the ETDRS did not provide results separately for NPDR and early PDR. In economic modelling, the base case found that early PRP could be more effective and less costly than deferred PRP. Sensitivity analyses gave similar results, with early PRP continuing to dominate or having low incremental cost-effectiveness ratio. However, there are substantial uncertainties. For our secondary aims we found 12 trials of lasers in DR, with 982 patients in total, ranging from 40 to 150. Most were in PDR but five included some patients with severe NPDR. Three compared multi-spot pattern lasers against argon laser. RCTs comparing laser applied in a lighter manner (less-intensive burns) with conventional methods (more intense burns) reported little difference in efficacy but fewer adverse effects. One RCT suggested that selective laser treatment targeting only ischaemic areas was effective. Observational studies showed that the most important adverse effect of PRP was macular oedema (MO), which can cause visual impairment, usually temporary. Ten trials of laser and anti-VEGF or steroid drug combinations were consistent in reporting a reduction in risk of PRP-induced MO.LimitationThe current evidence is insufficient to recommend PRP for severe NPDR.ConclusionsThere is, as yet, no convincing evidence that modern laser systems are more effective than the argon laser used in ETDRS, but they appear to have fewer adverse effects. We recommend a trial of PRP for severe NPDR and early PDR compared with deferring PRP till the HR-PDR stage. The trial would use modern laser technologies, and investigate the value adjuvant prophylactic anti-VEGF or steroid drugs.Study registrationThis study is registered as PROSPERO CRD42013005408.FundingThe National Institute for Health Research Health Technology Assessment programme.