The objective of this study was to analyze systematically neck responses and associated injury predictors under complex loading conditions similar to real rollover scenarios using an updated finite element (FE) human head-neck model. The effect of changing the coefficient of friction (COF), impact velocity, padding material thickness and stiffness, and muscle force was analyzed along 16 different impact orientations. The maximum principal strain in the vertebrae was considered as injury predictor to evaluate the injury risk. The results showed that impact velocity is the most important factor in determining the risk of neck injury. Decrease the COF between the head and impact surface can effectively reduce the risk of neck injury. For frictionless head contact onto a rigid surface, the maximum injury risk occurred when the impact surface was approximately perpendicular to the longitudinal axis of the cervical spine. When the COF was 0.5, lateral contacts generated much higher injury risks than those without a lateral component due to the asymmetric loading distribution in the vertebrae. Muscles, especially the active muscles, increased the risk of neck injury significantly. Interestingly, a layer of padding with a proper stiffness and reduced COF between the padding and head could decrease the neck injury risk. This was in contrast to some experimental data which showed that all padding could increase the risk of neck injury. More simulations further demonstrated that if padding material was installed onto the roof with a frictionless sliding interface, the risk of neck injury could be reduced significantly. In real-world rollovers, it is likely that both COF and lateral forces are present. Results from this series of simulations suggest that a careful selection of proper padding stiffness along with a very low COF between the padding and its supporting structure may decrease head and neck injuries simultaneously. This new design concept should be further validated before it is implemented.