Since 1996, a number of investigators have carried out phase I-II studies of high-dose immunosuppression with autologous hematopoietic stem cell transplantation (HSCT) in autoimmune diseases. Most of this activity has been in studies of multiple sclerosis (MS), systemic sclerosis (SSc), systemic lupus erythematosus (SLE), rheumatoid arthritis (RA), and juvenile idiopathic arthritis (JIA). Supported by animal models of antigen-induced autoimmunity, the rationale of HSCT is to time-shift the clinical autoimmunity to an earlier period, restoring self-tolerance. Even with the considerable experience of more than 200 transplantations since 1996, it is difficult to judge the optimal approach. This difficulty is in part because of the multiplicity of centers and protocols and the variability in patient eligibility and assessment, the extent of T-cell depletion, and the intensity of the preparatory regimens used. Other than that found in RA, treatment-related mortality has been higher than expected: 17% in SSc (with an additional 10% mortality from progressive disease), 13% in SLE, 13% in JIA, and 8% in MS. Protocol changes to improve safety have been instituted. These changes include the avoidance of high-dose rabbit antithymocyte serum in patients who received T-cell-depleted grafts, use of corticosteroids with granulocyte colony-stimulating factor during stem cell mobilization and as prophylaxis for the engraftment syndrome in MS, lung radiation shielding in SSc, and multiple precautions against the macrophage activation syndrome in JIA. Responses to primary and secondary endpoints have been seen, and there is a consensus among investigators and regulatory bodies that the time has come for randomized phase II-III studies. Each disease presents distinct difficulties: in MS, restriction of eligibility to patients with active inflammatory disease; in SSc, formulation of cardiopulmonary eligibility criteria to decrease risk; in SLE, judgment of whether HSCT adds any advantage to high-dose nonmyeloablative immunosuppressive treatment alone; and in RA, enhancement of response durability. All prospective randomized studies in these diseases must address problems in selection of the comparison nontransplantation treatment and appropriate stopping rules, particularly with treatment arms of unequal risk. Parallel trials in Europe and in the United States are in the late stages of design.