Accurate prediction of vertebral fracture risk in patients with metastatic bone disease is a major clinical and computational challenge. Current clinical assessment relies largely on qualitative interpretation of imaging and clinical experience, limiting its ability to deliver patient-specific and quantitative fracture risk estimates. While recent advances in computational biomechanics have demonstrated the potential for structural assessment of bone, robust and clinically interpretable modelling frameworks for metastatic vertebrae are still lacking. In particular, existing models often fail to consistently integrate patient-specific vertebral anatomy, lesion characteristics, and physiological loading, leading to large variability in predicted fracture and fatigue behaviour. This study presents a patient-specific phase-field finite element framework to model fracture initiation, propagation, and fatigue damage in vertebrae affected by metastatic lesions. The framework integrates medical imaging–based geometry reconstruction with spatially varying material properties to represent tumour-induced bone degradation, together with loading conditions informed by spinal biomechanics. A continuum phase-field formulation for brittle fracture is coupled with linear elasticity and extended via a history-dependent degradation law to capture damage accumulation and progressive stiffness loss under cyclic loading representative of daily activities. Both patient-specific and synthetic vertebral models are employed to enable controlled and reproducible comparisons. This allows investigation of common modelling assumptions, including geometric simplifications, cortical shell representations, and constitutive formulations with and without plasticity, and their influence on predicted mechanical response and fracture behaviour. Numerical simulations demonstrate a strong sensitivity of vertebral failure and fatigue response to lesion size, patient-specific geometry, and material modelling choices. The findings identify which modelling assumptions reliably capture the global mechanical response and fracture mechanisms, and which may lead to non-negligible deviations. By addressing key gaps between qualitative clinical assessment and quantitative mechanical prediction, this work contributes to the development of mechanically informed tools for vertebral fracture risk assessment, with potential impact on surgical decision-making and fracture prevention in spinal oncology.