Recent advances in sensor technology allow for investigating emotional and cognitive states of learners. However, making use of sensor data is a complex endeavor, even more so when considering physiological data to support learning. In the BMBF-funded project Learning Analytics for sensor-based adaptive learning (LISA), we developed a comprehensive solution for adaptive learning using sensor data for acquiring skin conductance, heart rate, as well as environmental factors (e.g. CO2). In particular, we developed, (i) a sensor wristband acquiring physiological and environmental data, (ii) a tablet application (SmartMonitor) for monitoring and visualizing sensor data, (iii) a learning analytics backend, which processes and stores sensor data obtained from SmartMonitor, and (iv) learning applications utilizing these features. In an ongoing study, we applied our solution to a serious game to adaptively control its difficulty. Post-hoc interviews indicated that learners became aware of the adaptation and rated the adaptive version better and more exciting. Although potentials of utilizing physiological data for learning analytics are very promising, more interdisciplinary research is necessary to exploit these for real-world educational settings.