Receiver Autonomous Integrity Monitoring (RAIM) is currently a simple and efficient solution for civil aviation applications to check the integrity of GNSS down to Non Precision Approaches. The future introduction of new satellite constellations such as the European satellite navigation system Galileo or modernized Global Positioning System (GPS) will imply great improvements in the number as well as the quality of available measurements. More demanding phases of flight such as approaches with vertical guidance could be targeted using RAIM to provide integrity monitoring. The targeted probability of missed detection constitutes a major input of RAIM algorithm. This parameter derives from the integrity risk but also depends on the probability of satellite failure. Thus it refers to the threat model and particularly needs to be detailed. Up to now, most of RAIM algorithms assumed that only one satellite failure could occur at the same time and that the rate of occurrence of such a failure was the one of major service failure. But those assumptions have to be reconsidered for multi constellation RAIM designed for approaches with vertical guidance. Indeed, a larger number of available measurements also implies a larger number of potential faulty measurements for the receiver. Moreover, the targeted phases of flight are characterized by smaller horizontal and vertical tolerable position errors compared to NPA. Therefore, the threatening range errors that need to be detected by the fault detection algorithm have to be reconsidered, since they could have smaller amplitude, and a probability of occurrence that is not clearly defined currently. The aim of the proposed paper is to present some assumptions to be adopted for the design and the evaluation of RAIM algorithms for vertically guided approaches. The way RAIM algorithms can be implemented in order to take into account both civil aviation requirement and threat model is addressed and the way the required probability of missed detection can be set is particularly investigated. In the first part of the paper, the User Equivalent Range Error variance computation is detailed. The concept of critical bias which is the smallest bias on a single pseudorange measurement that leads to a positioning failure is then developed. It is based on the fact that integrity monitoring requires that the navigation system detects the presence of an unacceptably large position error for a given mode of flight. Then we demonstrate that, for the single failure case using GPS + Galileo constellations, the amplitude of pseudo range additional biases that lead to a positioning failure systematically belongs to the major service failure category for both APV I and LPV 200 (VAL=35 m) operations. Therefore even if the targeted phases of flight are characterized by smaller horizontal and vertical tolerable position errors compared to NPA, this effect is mitigated by the great number of available measurements that reduces the impact a of single satellite bias on the global positioning error. Thus only Major Service Failures need be taken into account for single failure case consideration using GPS + Galileo constellations. This assumption is more questionable if an effective monitor threshold is set to 10 meters or 15 meters for LPV200 operations. Some biases, smaller than major service failures, could lead in some worst case situations to dangerous positioning failure. Unfortunately, the rate of occurrence of such biases is not currently known due to a lack of monitoring. This constitutes a major issue for the use of RAIM for approaches with vertical guidance. Multiple failure case is also addressed in this paper. A method that benefits from the fact that multiple failures are very rare is used. It consists in not trying to detect these multiple failures and setting the probability of detecting an integrity failure caused by multiple faults to zero. This operation leads to more stringent required probability of missed detection for single failure but allows the use of various detection algorithms that have been designed assuming only one pseudorange failure at the same time. Finally, a review of major RAIM assumption for approaches with vertical guidance operations is proposed.