Nickel–Titanium (Ni-Ti) shape memory alloys (SMAs) are extensively employed in self-expanding cardiovascular devices for minimally invasive procedures, due to their superelastic behavior induced by stress-driven phase transformations. However, guaranteeing long-term durability remains a major challenge, especially for peripheral stents and heart valves exposed to millions of loading cycles associated with physiological motion. Device safety is currently assessed through a combination of in vitro tests on representative specimens and numerical simulations. Nonetheless, fatigue evaluation often relies on simplified indicators that fail to capture the material’s inherent complexity, neglecting the presence of defects and the role of phase transformation at the crack tip in damage evolution. This work investigates complementary predictive strategies to improve fatigue life assessment of Ni-Ti biomedical devices, focusing on surrogate multi-wire specimens. Uniaxial fatigue tests were conducted under different mean and alternating strain conditions to characterize fatigue life. A first approach based on fracture mechanics was implemented, assuming fatigue failure controlled by crack propagation from pre-existing defects. Fatigue crack growth experiments were performed on notched Ni-Ti samples, using the cyclic energetic J-integral as the driving force to calibrate an appropriate crack growth law [1]. This methodology yielded conservative fatigue life predictions for the multi-wire specimens, in good agreement with experimental trends, while neglecting the crack initiation phase. In parallel, a more advanced phase-field fracture model was developed by coupling a SMA constitutive model with a gradient-enhanced damage formulation within a variational framework [2]. After calibration, fatigue behavior of the multi-wire samples was numerically simulated, exploiting the model’s inherent capability to account for fatigue effects. The resulting fatigue life predictions closely matched experimental observations and provided insight into the localized material response in the damaged region. Ongoing work aims to integrate these approaches into a unified predictive framework to further enhance the reliability assessment of SMA-based medical devices.