This study addresses a crucial gap in the knowledge of normative spinal growth in children. The objective of this study is to provide detailed and accurate 3D reference values for global and segmental spinal dimensions in healthy children under the age of 11.

Radiographic spine examinations of healthy children conducted to rule out scoliosis were reviewed in four scoliosis referral centers in North America. All consecutive children aged three to eleven years old with EOS biplanar good quality x-rays, but without diagnosed growth-affecting pathologies, were included. Postero-Anterior and Lateral calibrated x-rays were used for spine 3D reconstruction and computation of vertebral body height and spine length. Median and interquartile range were calculated from cross-sectional data. Smooth centiles growth curves for 3D True Spinal Length (3DTSL) between T1 and S1, as well as for mid-vertebral heights of T5, T12 and L3, where fit and calibrated from data using the Lambda-Mu-Sigma method (GAMLSS package for R). This method automatically selects the best performing distribution from a familly of choices. Tables of centiles were then predicted from the computed models for selected ages.

A total of 638 full spine examinations from asymptomatic patients were reconstructed in 3D, 397 in girls and 241 in boys. Medians and interquartile ranges were calculated for 3DTSL (T1-S1): 285 (24) mm, 314 (26) mm and 349 (31) mm, and for selected vertebral heights T5: 10 (1) mm, 11 (1) mm and 12 (1) mm, T12: 13 (2) mm, 14 (1) mm and 16 (2) mm, and L3: 14 (1) mm, 16 (2) mm and 18 (2) mm, respectively for the 3–6, 6–8 and 8–11 age groups. Centile curves ready for clinical use of the 3DTSL (T1-S1) and of the vertebral heights of T5, T12 and L3 as a function of age were derived for the 5, 10, 25, 50, 75, 90 and 95th centiles. In general, boys presented linear relationships between spinal dimensions and age, and girls presented more diverging trends with increased variance for older ages. Accordingly curves for boys follow the Normal distribution whereas those for girls follow the original Box-Cox-Cole-Green distribution. Model diagnostic tests (normally distributed residuals, adequate wormplots and |Z statistics| < 2) confirmed adequacy of the models and the absence of significant misfit.

Accurate reference values were derived for spinal dimensions in healthy children. Spinal dimension charts showed that the spinal lengths and vertebral heights changed relatively constantly across the age groups closely resembling WHO total body height charts. The reference values will help physicians better assess their patients' growth potential. It could also be used to predict expected spinal dimensions at maturity or changes in pathologic conditions as well as to assess the impact of growth friendly interventions in the correction of spinal deformities.