Predictors of failure of non-operative management in neurologically intact thoracolumbar burst fractures: a systematic review and meta-analysis.

purposeNon-operative management of stable thoracolumbar burst fractures fails in a subset of patients, who require surgical conversion. A 2022 systematic review identified candidate predictors of failure only qualitatively and did not examine AOSpine A3 versus A4 morphology as a predictor. This review pools these predictors quantitatively and tests whether radiographic and clinical trajectories diverge by morphology.

methodsFour databases were searched for studies of neurologically intact patients with AOSpine A3/A4 (or posterior-ligamentous-complex-intact, or Thoracolumbar Injury Classification and Severity Score < 4) thoracolumbar burst fractures managed non-operatively. Failure rate (conversion to surgery) was pooled with a Freeman-Tukey double-arcsine random-effects model. Continuous predictors were pooled as Hedges' g standardised mean differences (SMD) between failure and success groups, with per-unit odds ratios (OR) pooled separately where at least three studies reported a comparable scale. Between-study variance was estimated by restricted maximum likelihood with Hartung-Knapp confidence intervals, and 95% prediction intervals are reported.

resultsThe pooled failure rate was 3.8% (95% CI 0.1-10.8%; 95% prediction interval 0.0-24.7%; k = 6, N = 629, I²=78%), and 2.5% (95% CI 0.2-6.4%, k = 5, I²=5%) after excluding the one cohort defined by a TLICS threshold rather than by morphological classification. Definitions of failure differed across studies, from pain preventing mobilisation to radiographic progression. Canal compromise was the only predictor robust to the Hartung-Knapp adjustment (SMD + 0.97, 95% CI 0.52-1.41, k = 6, I²=67%), its heterogeneity traceable to one study. Kyphotic angle was associated with failure (SMD + 1.26, 95% CI 0.15-2.37, k = 5, I²=93%) with a prediction interval crossing zero. Interpedicular distance had the largest point estimate (SMD + 2.36, k = 3, I²=97%) but a Hartung-Knapp interval including zero (95% CI - 0.76 to 5.48; OR 1.52 per percentage-point, 95% CI 0.64-3.65). Age was not associated with failure as an SMD (k = 6, I²=95%), one study finding an inverse association; as a per-year odds ratio it was homogeneous but imprecise (OR 1.08, 95% CI 1.00-1.15, k = 3, I²=0%). In the one cohort reporting both outcomes by morphology, AOSpine A4 fractures progressed radiographically more than A3 fractures (Cobb change + 6.1° versus + 4.1°, p = 0.021) with nearly identical clinical failure (3.9% versus 3.3%); a second cohort reproduced the radiographic pattern, and the largest predictor study found no association between morphology and clinical failure. Too few studies compared bracing with no bracing to test effect modification by severity.

conclusionsCanal compromise at admission is the most consistent quantitative predictor of failure of non-operative management in this population. Interpedicular distance and kyphotic angle show large but imprecise associations that need confirmation; age predicts failure within a cohort but does not transfer as a fixed threshold across populations. AOSpine A4 morphology predicts worse radiographic progression than A3 morphology without a corresponding increase in clinical failure. Whether bracing modifies outcomes differentially by fracture severity remains unresolved for lack of comparative data.