A detailed study of the performance of helical screw-thread inserts in tubular membranes has been carried out at the University of Oxford. The design combines a predominantly helical flow, in which Dean vortices are generated, with an approximately axial how, which modifies the Dean vortices into a continuous 'corkscrew vortex'. This promotes good mixing of the feed fluid and minimises concentration polarisation effects. Computational fluid dynamics (CFD) modelling of the flow has been carried out. Comparisons of the performance of plain membranes with membranes fitted with various inserts are reported. The behaviour of the screw thread inserts was modelled using the commercial CFD code 'Fluent'. Good agreement was found between the CFD predictions and experimental measurements. Laminar flow conditions appear to be maintained at high cross-flow rates, suggesting that the helical insert geometry could improve the filtration of shear-sensitive fluids such as blood. Measurements using bovine blood and suspensions of marine algae supported this conclusion. Testing of tubular membranes fitted with one-start and three-start helical inserts has been completed. The inserts were tested under microfiltration conditions using yeast solutions, under ultrafiltration conditions using reconstituted powdered milk solutions, and under nano filtration conditions using synthetic dyes. Two pumped loops were built for these experiments. The helical inserts produced much higher fluxes at low cross-flow rates than membranes without inserts (up to a factor of 6 higher), and also performed better than a plain cylindrical insert which had a similar pressure drop versus flow characteristic. Thus helical inserts improve the filtration of shear-sensitive fluids at low cross-flows; they also produce high Auras with low cross-flow rates, which is beneficial if high value products are being separated. By suitably modifying the geometry of the insert, lower fluxes and pressure drops can be obtained: this offers the opportunity of tailoring the insert performance to the requirements of a particular process.