Study Design. A retrospective design comparing three matched groups was used to evaluate the application of a multidimensional approach to outcomes analysis using a variety of disease-specific and generic outcome measures to assess three treatments for failed back surgery syndrome. Objective. The objective of this study was to explore the use of a multidimensional analysis of outcomes to compare and contrast the effects of three different treatments: 1) intrathecal therapy using an implantable drug administration system (DAS), 2) standard medical therapy emphasizing the use of oral opioids (OO), and 3) residential pain and rehabilitation program (RPRP) for the treatment of chronic low back pain. Summary of Background Data. The incidence of low back pain in patients with prior back surgery remains significant. Treatments for failed back surgery syndrome include interventional, pharmacologic, and functional restorative therapies. Outcome studies have rarely compared these three treatments over an extended follow-up period using both disease-specific and generic outcome measures. This study examined three groups of patients completing treatment in clinical setting with a 4-year follow-up. Methods. Patients were selected from three different clinical treatment populations: DAS, OO, and RPRP. The three groups consisted of 40-50 patients each, matched on a number of demographic and pain variables. The average duration of follow-up for each group was about 4 years. Data were collected on disease-specific outcome variables using Numerical Pain Ratings, the McGill Pain Questionnaire, the Beck Depression Inventory, the Oswestry Disability Scale, medication utilization, and other generic outcomes such as SF-36, Quality of Well-Being, Overall Improvement in Pain, and Patient Satisfaction. Information was obtained via chart review and telephone interviews with patients by a disinterested third party. Analysis of variance, chi-square test, and t-test were performed on the various dependent measures. Results. The DAS group appeared to be superior to the OO and RPRP groups on 8 of 11 dependent measures, but only one dependent measure, the posttreatment numerical pain rating, was statistically significant. The RPRP group reported the highest Quality of Well-Being score, while the OO group had the highest level of patient satisfaction. Overall, each of the groups fell well below the "normal" range on generic measures, even when compared to other clinical populations. Measures of depression, subjective disability, and pain remained in the "moderate" range with little evidence of improved functioning or quality of life (QoL). Despite these findings, patients rated their overall improvement at 50-60% and reported high levels of satisfaction with their treatment. Conclusions. The interpretation as to the most effective treatment depended on the particular outcome measure emphasized. There appeared to be a "disconnect" between ratings of pain, disability, mood, and QoL. The use of a multidimensional outcomes approach revealed a number of inconsistencies in the data, which could have been overlooked using only pain ratings and patient satisfaction data. No one treatment emerged as the most effective across all of the disease-specific and generic measures. However, patients in the DAS group tended to report greater improvement. Overall, although generally "satisfied" with treatment, they were generally "satisfied" with treatment despite continuing to report significant levels of pain, disability, and impaired QoL. © 2006 International Neuromodulation Society.