Study Design. Meta-analysis methodology was extended to derive comparative effectiveness information on spinal manipulation for low back pain.Objective. Determine relative effectiveness of spinal manipulation therapies (SMT), medical management, physical therapies, and exercise for acute and chronic non-surgical low back pain.Summary of Background Data. Results of spinal manipulation treatments of non-surgical low back pain are equivocal. Nearly 40 years of SMT studies were not informative.Methods. Studies were chosen based on inclusion in prior evidence syntheses. Effect sizes were converted to standardized mean effect sizes and probabilities of recovery. Nested model comparisons isolated non-specific from treatment effects. Aggregate data were tested for evidential support as compared to shams.Results. Of 84% acute pain variance, 81% was from non-specific factors and 3% from treatment. Compared to sham treatments, no treatment was better than sham results. Most acute results were within 95% confidence bands of that predicted by natural history alone. For chronic pain, 66% out of 98% was non-specific, but treatments influenced 32% of outcomes. Chronic pain treatments also fit within 95% confidence bands as predicted by natural history. Though the evidential support for treating chronic back pain as compared to sham groups was weak, chronic pain appeared to respond to SMT, while whole systems of chiropractic management did not.Conclusion. Meta-analyses can extract comparative effectiveness information from existing literature. The relatively small portion of outcomes attributable to treatment explains why past research results fail to converge on stable estimates. The probability of treatment superiority between treatment arms was equivalent to that expected by random selection.Treatments serve to motivate, reassure, and calibrate patient expectations – features that might reduce medicalization and augment self-care. Exercise with authoritative support is an effective strategy for acute and chronic low back pain.