The Neural Generation Hypothesis of Adolescent Idiopathic Scoliosis: Converging Evidence from Genetics, Proprioception, and Predictive Coding

Adolescent idiopathic scoliosis (AIS), a three-dimensional deformity of the spine (a sideways curve together with rotation of the vertebrae) without an identified cause, affects about 1-3% of children aged 10 to 16. This integrative review proposes the neural generation hypothesis (NGH): that, where the sensorimotor loop is intact, the actively reducible component of the scoliotic curve (what active self-correction removes beyond passive flexibility) is maintained by the body schema (the brain's non-conscious and precision-weighted predictive model of the body) and that along one entry route degraded sensing can precede and help initiate the curve. Curve magnitude in the growing spine is governed by growth-dependent asymmetric loading, which NGH does not claim to explain. In its revised, multi-pathway form, NGH separates how a curve begins from how its reducible, schema-mediated component is maintained. Several distinct entry routes (a sensory-prediction channel in any of several senses whose precision is degraded, by genetic predisposition or, as proposed, by other causes; an unregulated expression of the lateral visceral and diaphragmatic asymmetry present in nearly all humans; and, in secondary scoliosis, a primary structural load) are proposed to converge on a single neural lock, in which the schema's slow learning loop adapts the postural reference (the goal prior the body is held to) toward the deviated geometry, so that the deviation ceases to be signalled as error (predictive entrenchment).

Predisposition is polygenic and spans two compartments. One of the strongest and most replicated genetic signals in AIS lies at the LBX1 locus. LBX1 specifies dorsal-horn somatosensory interneurons, which makes a sensory route plausible, but the causal gene and tissue at that locus are not yet confirmed: the same locus also acts in muscle and axis patterning. A second compartment shapes structural substrate (PAX1 and GPR126/ADGRG6: growth plate, cartilage and disc), and the chromatin remodeller CHD7 is a candidate regulator with weaker human evidence. These are small-effect, low-penetrance variants: a liability substrate, not a deterministic trunk.

Onset clusters at the growth spurt, and NGH proposes that this is partly because growth-driven drift can stay below what a degraded sensory channel can detect and correct, while the schema's slow learning loop gradually absorbs the deviated geometry as its new normal (a proposal, not yet measured). That loss of position sense can start a curve is supported in animals: mice lacking proprioceptors on both sides develop scoliosis at puberty without a prior vertebral defect, though the mice are ataxic, and one-sided lesions produce curves (dorsal-root section in monkeys and rabbits, giving curves of a neuromuscular type, with similar curves after motor-root section in rabbits; somatosensory cortex in rats). Human conditions that remove or reduce position sense (familial dysautonomia, partial dorsal rhizotomy, PIEZO2 deficiency) come with scoliosis, with confounds. We found no study that measured position sense in humans before an idiopathic curve appeared, so the human ordering remains open; where a structural load initiates the curve, the schema adapts to it late.

Initiation and progression are separate problems. Once a curve exists, its magnitude is driven by growth-dependent asymmetric loading (Hueter-Volkmann), which the neural lock does not govern. Three coupled loops (proprioceptive, autonomic, and psychosocial, the latter two lowering sensory precision without specifying geometry) are proposed to convert ordinary slow-loop convergence into entrenchment, making the reducible component of the curve a loop-maintained basin of attraction; the structural component is not. The evidence is uneven and is stated link by link: the genetic predisposition and growth-driven progression are well supported; sensory sufficiency is supported in animals; human sensory findings are associations measured after onset, which a curve-driven explanation also fits; and the links specific to NGH (rate mismatch, symmetry breaking, the lock, the precision cascade, the vestibular reference-frame proposal, the autonomic and psychosocial loops, reversibility) are untested.

The hypothesis states the results that would refute it, with proposed thresholds. It predicts that a pre-onset perception and sensory-weighting measure should stratify who curves, that perception at brace weaning should predict later correction loss in skeletally mature patients, and that measured asymmetry before onset should predict curve convexity in the same person. One test can use data that already exist (balance records in a birth cohort), and one can be added to an existing brace cohort. It makes no claim that any intervention reverses established deformity or displaces monitoring, bracing, or surgery. "Idiopathic" may name a measurement gap: current tools assess the structural output, while NGH proposes that a mechanism holding the curve's reducible component operates across the nervous system, the epigenetic landscape, autonomic state and the psychosocial environment.