BACKGROUND: The National Institute for Health and Clinical Excellence (NICE) was reviewing its previous guidance on continuous subcutaneous insulin infusion (CSII). The review provided an assessment of evidence which had been published since the previous NICE appraisal (TA 151) in 2007. OBJECTIVES: To examine the clinical effectiveness and cost-effectiveness of using CSII to treat diabetes. To update the previous assessment report by reviewing evidence that has emerged since the last appraisal, and to take account of developments in alternative therapies, in particular the long-acting analogue insulins, which cause fewer problems with hypoglycaemia. DATA SOURCES: A systematic review of the literature and an economic evaluation were carried out. The bibliographic databases used were MEDLINE and EMBASE, 2002 to June 2007. The Cochrane Library (all sections), the Science Citation Index (for meeting abstracts only) and the website of the 2007 American Diabetes Association were also searched. REVIEW METHODS: The primary focus for type 1 diabetes mellitus (T1DM) was the comparison of CSII with multiple daily injection (MDI), based on the newer insulin analogues, but trials of neutral protamine Hagedorn (NPH)-based MDI that had been published since the last assessment were identified and described in brief. For type 2 diabetes mellitus (T2DM), all trials of MDI versus CSII were included, whether the long-acting insulin was analogue or not, because there was no evidence that analogue-based MDI was better than NPH-based MDI. Trials that were shorter than 12 weeks were excluded. Information on the patients' perspectives was obtained from four sources: the submission from the pump users group--Insulin Pump Therapy (INPUT); interviews with parents of young children who were members of INPUT; some recent studies; and from a summary of findings from the previous assessment report. Economic modelling used the Center for Outcomes Research (CORE) model, through an arrangement with the NICE and the pump manufacturers, whose submission also used the CORE model. RESULTS: The 74 studies used for analysis included eight randomised controlled trials (RCTs) of CSII versus analogue-based MDI in either T1DM or T2DM, eight new (since the last NICE appraisal) RCTs of CSII versus NPH-based MDI in T1DM, 48 observational studies of CSII, six studies of CSII in pregnancy, and four systematic reviews. The following benefits of CSII were highlighted: better control of blood glucose levels, as reflected by glycated haemoglobin (HbA1c) levels, with the size of improvement depending on the level before starting CSII; reduction in swings in blood glucose levels, and in problems due to the dawn phenomenon; fewer problems with hypoglycaemic episodes; reduction in insulin dose per day, thereby partly off-setting the cost of CSII; improved quality of life, including a reduction in the chronic fear of severe hypoglycaemia; more flexibility of lifestyle--no need to eat at fixed intervals, more freedom of lifestyle and easier participation in social and physical activity; and benefits for the patients' family. The submission from INPUT emphasised the quality of life gains from CSII, as well as improved control and fewer hypoglycaemic episodes. Also, there was a marked discrepancy between the improvement in social quality of life reported by successful pump users, and the lack of convincing health-related quality of life gains reported in the trials. With regard to economic evaluation, the main cost of CSII is for consumables, such as tubing and cannulas, and is about 1800-2000 pounds per year. The cost of the pump, assuming 4-year life, adds another 430-720 pounds per annum. The extra cost compared with analogue-based MDI averages 1700 pounds. Most studies, assuming a reduction in HbA1c level of 1.2%, found CSII to be cost-effective. LIMITATIONS: The most important weakness of the evidence was the very small number of randomised trials of CSII against the most modern forms of MDI, using analogue insulins. CONCLUSIONS: Based on the totality of evidence, using observational studies to supplement the limited data from randomised trials against best MDI, CSII provides some advantages over MDI in T1DM for both children and adults. However, there was no evidence that CSII is better than analogue-based MDI in T2DM or in pregnancy. Further trials with larger numbers and longer durations comparing CSII and optimised MDI in adults, adolescents and children are needed. In addition, there should be a trial of CSII versus MDI with similar provision of structured education in both arms. A trial is also needed for pregnant women with pre-existing diabetes, to investigate using CSII to the best effect.