Criteria for successful correction of thoracolumbar/lumbar curves in AIS patients: results of risk model calculations using target outcomes and failure analysis.

INTRODUCTION: Failure to select the appropriate lowest instrumented vertebra (LIV) in selective lumbar fusion (SLF) for thoracolumbar/lumbar curves (LC) can result in adding-on in the lumbar curve (LC) or the need for fusion extension due to a decompensating thoracic curve (TC). The selection criteria that predict optimal outcomes still need to be refined. The objectives of the current study were to identify risk factors for failure of anterior scoliosis correction and fusion (ASF) as well as predictors of optimal outcomes and ASF efficacy for SLF.
MATERIALS AND METHODS: A retrospective review of all patients (n = 245) with AIS who had anterior SLF at one institution was conducted. Optimal outcomes were defined as a target LC ≤ 20° and a target TC ≤ 30°. The distance from the LIV to the SV was recorded. An increase in the LIV adjacent level disc angulation (LIVDA) ≥ 5° was defined as adding-on. An increase in the TC at follow-up was defined as TC-progression. Stepwise univariate and multivariate linear and logistic regression analyses were performed to identify criteria predicting the target LC and TC. A total of 68 % of the patients had the LIV at SV-2 (=2 levels above stable vertebra).
RESULTS: The patients' average age was 17 years, the average fusion length was 4.6 levels, and the average follow-up time was 32 months. The preoperative LC was 49 ± 14°, the LC-bending was 22 ± 13° (57 ± 18 %), and the follow-up LC was 25 ± 10°. LC correction was 59 ± 17% (p < 0.01). The preoperative TC was 39 ± 13°, the TC-bending was 21 ± 12°, and the follow-up TC was 29 ± 13°. The TC-correction was 32 ± 19% (p < 0.01). At follow-up, 85 patients (35%) had an LC ≤ 20°, 110 patients (45 %) had a TC ≤ 30°. The follow-up LC and an LC ≤ 20° were predicted by LC-bending (p < 0.01, r = 0.6), preoperative LC (p < 0.01, r = 0.6). The logistic regression models could define patients at risk for failing the target LC ≤ 20° or TC ≤ 30°. At follow-up, TC ≤ 30° was best predicted by the preoperative TC (p < 0.01, r = 0.8; OR 1.2) and TC-bending (p < 0.01, r = 0.8; OR 1.06), with the logistic regression model revealing a correct prediction in 84 % of all cases. Among the patients, 8 % required late posterior surgery. Patients achieving the target LC ≤ 20° had a significantly reduced risk for failure (p = 0.01). Selecting an LIV at SV-1 vs. SV-2 significantly increased the chance of achieving a target LC ≤ 20° (p = 0.01) and reduced the risk of adding-on (p < 0.01). Predictors for failure also included a high preoperative LC (p = 0.02; OR 0.97), TC-bending (p < 0.01), and preoperative TC (p = 0.01). A cut-off in the failure risk analysis was established at a TC of 38°. Additionally, a significant cut-off for risk of adding-on was established at LIVDA <3.5°.
CONCLUSION: A high chance of achieving a target LC ≤ 20° and a low risk of revision was dependent on LC-bending, preoperative LC and TC, and a LIV at SV-1 with non-parallel LIVDA. Our risk model analysis may support the selection of a safe LIV to achieve the target LC.