Abstract
PURPOSE: Intervertebral disc degeneration progresses with normal aging; yet common disc grading schemes do not account for age. Degeneration progression also varies between spine levels and is similarly not accounted for by current grading schemes. These limitations inhibit differentiation between discs with normal and expected aging (non-pathological) and discs with accelerated degeneration (which may be pathological). We sought to develop a statistical model to quantify normal age and spine level dependent disc degeneration. METHODS: Eighty-four asymptomatic adult subjects ranging evenly from 18 to 83 years old underwent magnetic resonance imaging (MRI) of the lumbar spine. Subject traits, MRI-derived disc geometry, and MRI biomarkers of T2 relaxation time were evaluated and used to develop a statistical model to predict effective disc age, the age at which normal aging would produce a disc's observed phenotype. RESULTS: After evaluating several models, a 4-predictor model utilizing 1) subject height, 2) nucleus pulposus T2 relaxation time, 3) disc mid-sagittal area and 4) disc 3D volume, optimally estimated effective disc age. The effective age closely tracked true age for spine levels L1-L5 (R(2) ≈ 0.7, RMSE ≈ 10 years) and moderately tracked true age for L5-S1 (R(2) = 0.4, RMSE = 14 years). The uncertainty in the effective disc age prediction was ± 3 years as assessed by fivefold cross validation. CONCLUSION: We offer a data-driven, quantitative tool to quantify normal, expected intervertebral disc aging. This effective age model allows future research to target discs with accelerated degeneration.