Randomized controlled trial. To determine the validity of a previously suggested clinical prediction rule (CPR) for identifying patients most likely to experience short-term success following lumbar stabilization exercise (LSE). Background Although LSE is commonly used by physical therapists in the management of low back pain, it does not seem to be more effective than other interventions. A 4-item CPR for identifying patients most likely to benefit from LSE has been previously suggested but has yet to be validated. One hundred five patients with low back pain underwent a baseline examination to determine their status on the CPR (positive or negative). Patients were stratified by CPR status and then randomized to receive an LSE program or an intervention consisting of manual therapy (MT) and range-of-motion/flexibility exercises. Both interventions included 11 treatment sessions delivered over 8 weeks. Low back pain-related disability was measured by the modified version of the Oswestry Disability Index at baseline and upon completion of treatment. The statistical significance for the 2-way interaction between treatment group and CPR status for the level of disability at the end of the intervention was P = .17. However, among patients receiving LSE, those with a positive CPR status experienced less disability by the end of treatment compared with those with a negative CPR status (P = .02). Also, among patients with a positive CPR status, those receiving LSE experienced less disability by the end of treatment compared with those receiving MT (P = .03). In addition, there were main effects for treatment and CPR status. Patients receiving LSE experienced less disability by the end of treatment compared to patients receiving MT (P = .05), and patients with a positive CPR status experienced less disability by the end of treatment compared to patients with a negative CPR status, regardless of the treatment received (P = .04). When a modified version of the CPR (mCPR) containing only the presence of aberrant movement and a positive prone instability test was used, a significant interaction with treatment was found for final disability (P = .02). Of the patients who received LSE, those with a positive mCPR status experienced less disability by the end of treatment compared to those with a negative mCPR status (P = .02), and among patients with a positive mCPR status, those who received LSE experienced less disability by the end of treatment compared to those who received MT (P = .005). The previously suggested CPR for identifying patients likely to benefit from LSE could not be validated in this study. However, due to its relatively low level of power, this study could not invalidate the CPR, either. A modified version of the CPR that contains only 2 items may possess a better predictive validity to identify those most likely to succeed with an LSE program. Because this modified version was established through post hoc testing, an additional study is recommended to prospectively test its predictive validity. Prognosis, level 1b-.