Retrospective outcome measurement after circumferential reconstructive surgery with lumbar fusion in patients with chronic discogenic low back pain. To examine the effect of the number of fusion levels on surgical outcomes in patients with chronic discogenic low back pain using provocative pressure-controlled diskography as a primary diagnostic tool. Although there is general agreement that construct length adversely affects arthrodesis success rates, the effect of the number of levels on lumbar fusion surgery outcome has not been reported. Previous fusion outcome studies have generally relied on magnetic resonance imaging or conventional diskography for diagnosis. From 1994 through 2000, prospectively collected medical records of patients who underwent reconstructive lumbar spine surgery with confirmation of the pain generator by pressure-controlled diskography were retrospectively analyzed. Data were subdivided into 2 groups of patients. The first group, designated the short segment group, contained patients who underwent fusion at 1 or 2 levels. The second group, designated the long segment group, contained patients who underwent fusion at 3-5 levels. Surgical methods included circumferential reconstruction of the lumbar spine by either posterior or combined anterior and posterior approach. Surgeries included posterior decompression necessary to relieve documented regions of neural compression, combined with interbody arthrodesis at selective levels, augmented by posterior segmental spinal instrumentation and posterolateral arthrodesis. All patients completed a preoperative aquatic-conditioning program. Whenever possible, coexisting medical conditions were corrected or stabilized before surgery. A preoperative Short Form RAND 36-Item Health Survey (SF-36) was completed, and repeated at 1 and 2 years after surgery. The short and long segment groups contained 142 and 82 patients, respectively, who completed the preoperative SF-36 questionnaire completely. One hundred patients in the short segment group (vide infra) were available for 1-year follow-up, and68 were available for 2-year follow-up. In the long segment group, 81 patients were available for 1-year follow-up, and 49 were available for 2-year follow-up. Mean ages were 41.0 and 47.6 years in the short and the long segment groups, respectively. The 2 groups did not differ significantly in gender, smoking habits, workers' compensation, or litigation (P > 0.05). In the short segment group, postoperative 1-year mean Physical Component Summary (PCS) and Mental Component Summary scores significantly improved (P < 0.001 and P = 0.002, respectively). Domains other than general health perceptions showed significantly improved 1-year follow-up scores (P <or= 0.001). Two-year follow-up scores showed significant improvement (P < 0.001 for physical function [PF], role function as limited by physical problems [RP], bodily pain [BP], social function [SF], and PCS). The vitality (VT) and role function as limited by emotional problems (RE) also improved (P = 0.005 and P < 0.05, respectively). In the long segment group, postoperative 1-year mean PCS scores improved significantly (P < 0.001), with some improvement in Mental Component Summary score (P < 0.05). The long segment group also showed significantly improved PF, RP, BP, and SF scores (P < 0.001). The VT and RE scores gave P = 0.002 and P < 0.05, respectively. Comparing preoperative and 2-year follow-up scores, PCS, PF, RP, BP, and SF showed significant improvement (P < 0.001), and the VT score gave P < 0.01. Mean difference in postoperative and preoperative scores for both groups did not show significant differences (P > 0.05), although the PF score showed differences in 1 and 2-year follow-up scores (P = 0.048 and P = 0.068, respectively). When using strict patient selection criteria that include independent determination of pain generators via pressure-controlled diskography and completion of a preoperative conditioning program for improving general health status, the number of levels in reconstructive lumbar surgery may not significantly impact overall clinical outcome.