In recent years, sweat has gained increasing attention from the scientific community as a new analyte for health monitoring. The main advantage with respect to the "Gold Standard" for laboratory analysis, i.e. blood, is of course the possibility of performing non-invasive assays, granting the maximum comfort for the user. Sweat contains a wide collection of relevant biomarkers, such as ions (Na+, K+, Mg+, Ca++...) and biomolecules (lactate, cortisol, cholesterol...). The most widespread technology for wearable sweat sensors, as of now, consists in the employ of big electrodes for potentiometric detection of target markers. This strategy, on top of not being CMOS-compatible and therefore harder to integrate in a complex wearable system (e. g. a smartwatch), requires a huge amount of sweat, which is produced only under intense heath or physical activity. Ion-Sensitive Field-Effect Transistors (ISFETs), on the other hand, can be made extremely small without reducing their sensing performances, allowing monitoring of sweat composition at very low sweating rates, compatible with at-rest values. In this work, such advantage is pursued via the integration of a low-volume capillary microfluidics, with a capacity of just a few hundreds of nanoliters, on a chip containing a set of Fully-Depleted Silicon-On-Insulator FETs (FD SOI FETs) and miniaturized Ag/AgCl Reference Electrodes (mREs). This system, able to host up to four different functionalizations, each for monitoring a specific sweat biomarker, is capable of collecting sweat from the skin of an user at rest, passively driving it in contact with the sensors, and readily transducing the chemical parameters measured by the functionalized gates in an electric signal, which is then analyzed by the read-out system, providing data on the composition of the analyte and, therefore, on the health condition of the wearer. The fabricated devices showed excellent performances in terms of both electrical characterization and biomarker sensitivity: the FETs characterized with a metal gate have shown an ION/IOFF ratio of 106, a nearly-ideal Subthreshold Swing (SS) of just 65 mV/dec, a hysteresis free characteristics and fully ohmic contacts on ultra-thin silicon. The threshold voltage (Vth) shifts linked to a tenfold variation in the biomarkers concentrations also reached the nearly-ideal values of 55 to 60 mV/dec, and extremely linear dependence over a wide range of dilutions. One of the strengths of the ISFET technology is its ability to exploit technological progresses of MOSFETs for computing. For this reason, in the final part of this thesis we have investigated the application to sensors of one of the latest strategies for the improvement of MOSFET performances: the Negative Capacitance (NC). Addition of a ferroelectric capacitor (Fe-Cap) to the gate stack of our sensors reduced their SS by 21%, therefore improving their current sensitivity by 26%. The measured Current Efficiency also improved. Further experiments employing different sets of Fe-Caps and pulsed mode measurements, moreover, demonstrated a SS six times lower than the one of the baseline ISFETs and three times lower than the thermionic limit, paving the way to a new class of sensors with ideality factor larger than 1.