A Bayesian network meta-analysis of 133 analytic units compared spinal manipulation, exercise, medical care, physical modalities, self-care and sham against an independent natural-history estimate. In acute low back pain, 75.2% were predicted to recover by 26 weeks without treatment; spinal manipulation yielded 63.2%, no better than sham (+3.1 per 100, interval spanning zero). In chronic pain, natural-history recovery was 18.1% at 54 weeks, versus 31.8% with manipulation (+13.8 per 100) and a similar gain with exercise (+12.5 versus sham).
The key move is the comparator. Most trials compare treatment with sham or usual care, which blends natural recovery with the comfort of being treated. Anchoring to separate prognosis cohorts shows how much apparent benefit is simply time passing. The acute result is sobering but fits earlier reviews: most acute episodes resolve on their own, so reassurance, staying active and patience are reasonable first steps. The chronic result is more encouraging, with sham tracking natural history almost exactly (+0.4 per 100), so gains from manipulation and exercise look real rather than a placebo artifact. Even so, only about one in three chronic patients was modeled to improve. Exercise is cheap, carries little risk and offers wider health benefits, which argues for making it the default. Caveats: this is a preprint awaiting peer review, natural-history estimates come from different studies than the treatment trials, and trial heterogeneity and sham quality in manual therapy remain contested. The framework is a useful refinement; the clinical conclusions are largely confirmatory.