Osteotomy is an effective method to correct kyphosis and restore the sagittal alignment in patients with AS, which greatly improves their daily living activities and quality of life [5, 9, 12, 15]. However, clinically, not all patients with kyphosis who undergo osteotomy can achieve a satisfactory sagittal balance at the follow-up, leading to an increased risk of implant failure, delayed union, pseudoarthrosis, and correction loss [8, 16]. Until now, few studies have attempted to specifically determine the preoperative predictions with clear threshold values to predict sagittal imbalance in patients with AS, and the optimal postoperative goal for preventing sagittal imbalance in these patients is still less known [9, 12, 13].
In this study, the patients in group A had larger LL, PI − LL, TPA, and SVA values than those in group B preoperatively, which indicated that the patients with follow-up sagittal imbalance often had more severe preoperative sagittal imbalance and spinopelvic deformity. These patients were supposed to undergo a matching larger correction to construct their sagittal alignment with the apparently preoperative sagittal deformity. However, in fact, they only received the same amount of correction as those in the group B, leaving much residual postoperative deformity. As a result, patients in group A were more likely to experience failed sagittal realignment than those in group B at the follow-up. This was consistent with that observed by Schwab et al. , who indicated that one-level three-column osteotomy may not always achieve a satisfactory outcome, particularly in those with severe preoperative sagittal imbalance and lumber kyphosis; additional osteotomy was recommended for these patients [13, 17].
The Pearson coefficient analysis demonstrated that the follow-up SAV value significantly corelated with preoperative parameters, which made it possible to predict sagittal imbalance at the final follow-up with preoperative parameters. Based on the significant relationship between the follow-up SVA value and preoperative parameters, ROC analysis showed that the top three AUC values were those for TPA, PI − LL, and SVA, which indicated that these three preoperative parameters were the optimal predictions with the best accuracy for predicting sagittal imbalance. Moreover, the optimal threshold value of these preoperative predictions for predicting sagittal imbalance was TPA >40.9°, PI − LL >32.5°, and SVA >13.7 cm. Similarly, Qian et al.  found that the SVA and PI were the radiographic predictors for the postoperative sagittal imbalance; however, the optimal threshold value of these two predictors could not be evaluated. In this study, the preoperative predictions were the three most relevant parameters with the largest AUC; and their threshold values were evaluated clearly for predicting sagittal imbalance, making them useful in surgical decision-making. Although the preoperative LL and PT values showed a statistical relevance with sagittal imbalance, the accuracy and coefficient were relatively lower than the other three parameters; thus, they were not included in the further analysis. Additionally, the efficacy of these three predictions was validated with the present cohort. The results showed that patients in whom two or three predictions met the threshold values significantly increased the rate of sagittal imbalance at the final follow-up. This revealed that the predictive ability of these predictions was effective when two or three predictions met their threshold values concurrently. Only one prediction meeting the threshold value might not be enough to predict sagittal imbalance accurately. Therefore, according to the results, two or three preoperative predictions were recommended to be taken into consideration concurrently when doing preoperative planning and judging the postoperative sagittal realignment.
Predicting sagittal realignment in advance was the first step for preventing sagittal imbalance, while correcting kyphosis and constructing sagittal alignment should be the most important step. The main cause for failed sagittal realignment in most patients has always been inadequate intraoperative correction . Schwab et al.  suggested that the goal for postoperative SVA should be <4.7 cm for ASD after osteotomy. However, for kyphosis in patients with AS, the optimal postoperative SVA value for correction has not been reported yet. In this study, based on the relationship between the follow-up SVA and postoperative parameter, we investigated the key parameter for sagittal imbalance with logistic regression analysis. The top three postoperative parameters (PI − LL, TPA, and SVA) with the highest coefficient were entered into logistic regression analysis to determine the independent key parameter of sagittal imbalance. The results demonstrated that the postoperative SVA was the independent key factor of sagittal imbalance. Subsequently, the optimal postoperative SVA was analyzed with the ROC, and the result showed that the optimal threshold of postoperative SVA for preventing sagittal imbalance was <7.4 cm, with an AUC of 0.941, a sensitivity of 88.9%, and a false-positive rate of 5.6%.
To further verify the effectiveness of the postoperative SVA with a threshold of 7.4 cm, the cohort was divided into two groups to compare the differences in SVA and the incidence of sagittal imbalance at the follow-up. Reliably, the results indicated that patients with postoperative SVA <7.4 cm had a smaller SVA and lower incidence of sagittal imbalance than those with postoperative SVA of >7.4 cm. This result was in line with that of Kim et al.  who reported that restoration and maintenance of postoperative SVA <8 cm was important to the ultimate sagittal reconstruction in fixed sagittal imbalance. In this study, these patients with an immediate postoperative SVA of <7.4 cm maintained an acceptable sagittal realignment at the follow-up. Similarly, Wang et al.  also reported that patients with AS with the postoperative SVA of 8.6 cm usually did not show an obvious correction loss or severe sagittal imbalance at the follow-up. Kim and his colleague observed 248 AS patients and found that the patients with SVA of 7.0 cm or less obtained the best clinical outcome after surgery. These results commonly reminded that for AS patients with severe kyphosis, the SVA would not be required to correct to a normal range, and the postoperative SVA of ≤7.4 cm might be enough for most patients to obtain satisfied clinical outcome.
Interestingly, a phenomenon that patients with AS with a larger immediate postoperative SVA might partially decrease to a smaller range at the follow-up was found in this study. In general, the spine of AS patients was stiff and rigid in the advanced stage, which could not be reflexed and extended. However, the pelvis functioned as a compensatory mechanism with the femoral heads as the fulcrum to rotate anteriorly after surgery. As the pelvis rotated backward to compensate sagittal imbalance preoperatively, the backward rotated pelvis could anteriorly rotate to some extent postoperatively provided the hip joints were not ankylosed. Of note, the pelvis rotation anteriorly could not be accomplished immediately after surgery because the AS with severe pelvic retroversion usually complicated with the contracture of hip joints for the tense ligaments, which could be gradually corrected to a certain extent with rehabilitation exercise [4, 19, 20]. Second, the patients with AS were still accustomed to the preoperative anterior center of gravity; thus, they kept on flexing the trunk when standing until they got used to the new center of mass after a period. And these interesting findings also were observed in previous studies [18, 21].