Magnetic beads (MB) have now proven to be a powerful tool in both research and biomedical applications. They are available in a wide range of sizes (from nm to several µm) and their surface can be modified with molecules having biological specificities and functions. The large choice of functionalization developed over the last few years covers non-specific interactions such as ionic or hydrophobic ones, as well as group-specific interactions, like immobilized metal affinity chromatography (IMAC) and, finally, specific interactions such as antigen-antibody recognition. In microfluidics, where the goals are faster reaction time and reduced sample consumption, MB offer many advantages. Firstly, compared to an open microchannel or an empty capillary, a packed bed of beads increases the specific surface available for molecule binding. The diffusion pathway is thus significantly reduced, improving interactions between molecules. Moreover, in comparison to classical beads, they are easily manipulated by electromagnets or permanent magnets. For these reasons, in this work they were chosen to act as a solid support in a view of performing immunoassays. As the literature often focuses mainly either on the applications or on very advanced studies that are not accessible to a non-specialist, it was difficult to find information on the magnetic aspects such as the kind of magnets, their number, their size or their arrangement. First, a background study was dedicated to the understanding of some basic magnetic aspects. The effect of the magnet shape or the size on the magnetic induction and the magnetic force was explored theoretically. Numerical simulations showed the strength and location of the magnetic forces versus three simple magnet configurations: two magnets in attraction/repulsion or a single magnet and the results were corroborated by microscopical visualizations. With this knowledge, it was then possible to enhance the magnetic force. Indeed, if the magnetic force produced is too weak, magnetic beads are unpredictably lost, resulting in poorly reproducible results. Concretely, a new magnet configuration using ring magnets, which are disks drilled along their magnetization axis, was studied to increase the magnetic force in a capillary, giving the opportunity to work at higher flow rates and consequently decrease the experimental time. This configuration also makes possible the formation of a chain of magnets alternating with non-magnetic spacers like a string of pearls, increasing in a controllable manner the surface available for molecule binding. Adsorption is a fundamental process in immunoassays. As antibodies are expensive reagents and as diagnostic tests should be as least invasive as possible for the patient, volumes should be reduced to their minimum. It is thus essential to optimize adsorption by choosing the right experimental parameters. In order to help the experimenter, an abacus representation was proposed showing, for a given surface coverage, the required binding time versus the flow velocity in a microchannel with known dimensions. Then, the question of knowing if the amount of analyte adsorbed is increased with a surface divided into small patches relative to a continuous surface for an equal total length was raised. Then, the question of the amount of beads trapped was handled as it also relates to the sensitivity of the technique. In a capillary, the number of magnetic beads possibly trapped is limited, particularly in small diameter capillaries, e.g. 25 µm, as the capillary is rapidly blocked as the plug size increases. In this work, a simple bubble cell was used as a very convenient magnetic bead trap where a large amount of beads can be spatially immobilized without inducing a strong pressure drop, as it is the case when magnetic beads are trapped in a standard capillary. Finally, a sandwich-type immunoassay using magnetic beads was developed for total IgE quantification in serum using the Gravi™-Cell device from DiagnoSwiss. After the determination of the optimal conditions, the method was successfully applied to the measurement of total IgE in a patient serum sample with a concentration in the same range as those determined previously by two other methods.