Atrophic non-union is a major complication following fracture of a bone. It represents a biological failure of the fracture healing process and occurs in 5-10% of cases. A number of factors predispose to atrophic non-union including high energy injuries, open fractures, diabetes, and smoking. Atrophic non-unions cause immense patient morbidity and consume large amount of health care resources. Bone grafts taken from the iliac crest contain biologic components required for fracture healing and are considered as the gold standard treatment of aseptic atrophic non-union. However, harvesting bone grafts from the iliac crest is associated with significant patient morbidity which can reduce quality of life. Mesenchymal stem cells (MSCs) have the ability to proliferate and undergo multilineage differentiation. The emergence of MSC therapy provides an alternative strategy for treating impaired fracture healing. MSCs contribute to normal fracture healing both directly as bone progenitor cells and indirectly as mediator secreting cells. Although a number of studies have shown that MSCs can promote bone regeneration in small animal fresh critical size defects, this is not analogous to most clinical aseptic atrophic non-unions which do not have a significant bone gap. There remains therefore a clinical need for an appropriate strategy for using stem cells in atrophic non-unions. Thus, the aim of this study aim was to develop a clinically relevant strategy to promote fracture healing in an atrophic non-union model using the percutaneous injection of MSCs as a minimally invasive technique. An atrophic non-union model was established and validated. A small (1 mm) non-critical size defect was created at the mid shaft tibia and the fracture site was stabilised using an external fixator. Atrophic non-union was induced by stripping the periosteum for one bone diameter either side of the osteotomy site and curettage of the intramedullary canal over the same distance. The procedure reliably created an atrophic non-union. Fracture healing was evaluated using (1) serial radiography, (2) micro-computed tomography, (3) histomorphology and (5) biomechanical testing. Fracture scoring systems including the radiographic union scale in tibia (RUST) and the Lane & Sandhu score were validated in a preclinical model. A simple sample preparation technique for evaluating bone mechanical properties was developed and used to assess the stiffness and strength of the fracture repair. Percutaneous injection of MSCs locally into the fracture site in the early ‘post-injury’ period at three weeks after induction of atrophic non-union was found to improve the fracture healing process significantly (83% of cases), while MSCs implantation in the late ‘post-injury’ period at eight weeks after induction of atrophic non-union showed no significant improvement of fracture healing (20% of cases). Percutaneous local implantation of MSCs rescued the fracture healing process in cases destined to progress to atrophic non-union. In clinical practice, there may be an advantage using MSCs from a universal donor as the processes of MSC isolation and preparation are expensive and time consuming. To investigate the feasibility of using non-autologous cells, the atrophic non-union was used to determine the bone regenerative potential of using xenogeneic donor hMSCs in an atrophic non-union. The results demonstrated that the therapeutic effect of using hMSCs in a xenogeneic manner to promote fracture healing in the rat atrophic non-union model was comparable with rMSCs (88% of cases in both hMSCs and rMSCs) and there were neither significant clinical adverse effects nor adverse immune responses with the xenogeneic transplantation. However, MSCs did not persist at the fracture following injection. Perivascular stem cells (PSCs) taken from adipose tissue, which is an expendable source, have advantages over conventional MSCs as they are a defined and homogenous population and can be used without culture expansion. The administration of PSC using percutaneous injection improved the fracture healing process in atrophic non-union (60% of cases). This suggested that PSCs may present an appropriate choice for use in cell therapies to promote fracture healing in atrophic non-union. The results from this thesis can be applied to the development of a clinically relevant strategy using MSCs as a minimally invasive technique to promote fracture healing in atrophic non-union, in particular (1) the effectiveness of a cell therapy is likely to be highly dependent of the timing of injection relative to the stage of fracture healing, (2) hMSCs were as effective as rMSCs in promoting fracture healing, suggesting that it may be feasible to use an allogeneic strategy in humans, (3) the injected MSCs were not detectable even in case of successful repair, suggesting that they may act through a paracrine effect and (4) PSCs isolated from adipose tissue contributed to fracture healing in the atrophic non-union model, suggesting that adipose tissues can be used as an alternative cell sources for bone repair.