Absence of these effects results in a low background and very low limits of detection for BRET assays. Due to these characteristics, BRET has been used for a variety of applications including RNA detection, investigating protein-protein interactions, drug screening, imaging, and general biosensing. So far, however, uses of BRET have largely been restricted to fundamental research, often using sophisticated imaging equipment. Despite its lower sensitivity, FRET has been used more broadly than BRET for screening and biosensing applications. Recently, however, we demonstrated that a form of BRET is 50 times more sensitive than FRET for measuring thrombin-catalysed proteolytic cleavage of a target peptide sequence in a microplate assay. A microfluidic format offer several advantages over static microplate assays but the generally low luminance of some BRET systems makes it quite challenging to detect a signal in the small volumes typical of a microfluidic device. Therefore, although substantial effort has been invested in developing FRET detection systems and bioluminescence detection for microfluidic systems, there have been very few studies combining BRET-based sensors with microfluidics. The aim of this study was therefore to test the feasibility, sensitivity and limits of detection of a homogenous BRET-based biosensor in a microfluidic format, compared with the same biosensor in a conventional assay. We selected a thrombin bioprobe for this work because it is well characterised, highly specific and clinically relevant. In earlier work we used the BRET2 system because of its long Fo¨rster distance, sensitivity and its potential application to measuring intramolecular rearrangements as well as dissociations. Unfortunately, however, the BRET2 system was not bright enough to be detected in the microfluidic system used here, due to its well-documented low quantum yield and the small volume sampled optically. For comparison, the standard volume optically sampled in the 96 well microplate format is 100 mL. We therefore adopted a novel BRET variant, which we named BRETH, combining the donor and acceptor domains of BRET2: i.e. GFP2 and RLuc, with the original BRET1 substrate, native coelenterazine. This provides much greater luminosity with limits of detection intermediate between BRET1 and BRET2, at the expense of an undefined, presumably shorter, Fo¨rster distance. For an assay involving complete molecular dissociation, such as the one here, a short Forster distance is of lesser concern. This compromise allowed us to compare the static and microfluidic versions fo the assay without the inconvenience of having to substitute the BRET2 donor and acceptor proteins. To the best of our knowledge, this is the first realisation of a BRET based biosensor in the fluid phase of any microfluidic system. The limit of detection for thrombin in the microfluidic format was 27 pM, which was more than tenfold lower than when measured using the same sensor in a microwell plate and two log units lower than comparable FRET-based microfluidic assays.