Study design: Randomized trial with a concurrent observational cohort study.
Objective: To compare 8-year outcomes of surgery with nonoperative care for symptomatic lumbar spinal stenosis.
Summary of background data: Surgery for spinal stenosis has been shown to be more effective than nonoperative treatment during 4 years, but longer-term data are less clear.
Methods: Surgical candidates from 13 centers in 11 US states with at least 12 weeks of symptoms and confirmatory imaging were enrolled in a randomized cohort or observational cohort. Treatment was standard, decompressive laminectomy versus standard nonoperative care. Primary outcomes were SF-36 (MOS 36-Item Short-Form Health Survey) Bodily Pain and Physical Function scales and the modified Oswestry Disability Index assessed at 6 weeks, 3 months, 6 months, and yearly up to 8 years.
Results: Data were obtained for 55% of participants in the randomized group and 52% of participants in the observational group at the 8-year follow-up. Intent-to-treat analyses showed no differences between randomized cohorts; however, 70% of those randomized to surgery and 52% of those randomized to nonoperative had undergone surgery by 8 years. As-treated analyses in the randomized group showed that the early benefit for surgery out to 4 years converged over time, with no significant treatment effect of surgery seen in years 6 to 8 for any of the primary outcomes. In contrast, the observational group showed a stable advantage for surgery in all outcomes between years 5 and 8. Patients who were lost to follow-up were older, less well-educated, sicker, and had worse outcomes during the first 2 years in both surgical and nonoperative arms.
Conclusion: Patients with symptomatic spinal stenosis show diminishing benefits of surgery in as-treated analyses of the randomized group between 4 and 8 years, whereas outcomes in the observational group remained stable. Loss to follow-up of patients with worse early outcomes in both treatment groups could lead to overestimates of long-term outcomes but likely not bias treatment effect estimates.