Biosensors have been developed for diagnostics and other medical applications for decades due to their many beneficial qualities such as cost-effectiveness, ease of use and high selectivity. Biosensors are used to detect disease biomarkers in blood in low concentrations. This allows for early diagnosis of diseases and real-time monitoring of disease evolution to enable more effective treatment. To achieve these goals, it is important for biosensors to reach low limits of detection (LOD) to be able to accurately sense low concentrations. A current critical bottleneck for reaching this goal in diagnostic biosensors is non-specific binding (NSB) due to the presence of a large amount of background molecules in biological samples. The aim of this thesis is to describe the negative impact that NSB has on the performance of a biosensor and present ways to reduce this impact. The thesis describes the basic function of biosensors and how their performance is evaluated. Ways to characterize NSB and differentiate it from desirable specific binding (SB) are presented. For this, the concept of affinity, the Langmuir Model of molecular adsorption and the energy landscape theory are used. The direct effect NSB has on the LOD is also explained. Based on these, methods for reducing the impact of NSB on the biosensor performance are presented. Force- and time-based discrimination between SB and NSB were found to be promising state of the art methods. In addition, referencing is an attractive method, but there are currently still challenges to its implementation. These methods still need further development but could potentially enable for biosensors to reach significantly lower LODs. Consequently, this could lead to much wider use of biosensors in diagnostic applications.