Restoring the sagittal alignment in the correction surgery has gotten increasing recognition of importance among spine surgeons1–7. However, it is a hard task to fully grasp every sagittal parameter, because some morphological parameters such as PI remains constant before and after surgery, and the other parameters change greatly and are difficult to predict30. Lumbar lordosis is one of sagittal parameter that is feasible to restore directly in the correction surgery8, 28, 29. Furthermore, in order to achieve the optimal spinopelvic sagittal balance, it is necessary to reconstruct ideal lumbar lordosis17, 18, and accurately predict the magnitude of postoperative lumbar lordosis with high sensitivity before surgery26.
However, although several predictive formulas17, 22–25, 28, 29 have been established to determine the target value of lumbar lordosis, variables recruited in these formulas were various and the actual prediction effectiveness remained controversial. Only PI was recruited in predictive formula and LL could be determined using ‘LL = 0.45*PI + 31.8’ in Yamato et al.’s study28. In Rose et al.’s formula29, the LL might be predicted by the formula LL ≤ 45o-TK-PI. These formulas either used one-sided parameters or parameters change with spinopelvic position, so they cannot accurately predict the ideal lumbar lordosis. So we performed this retrospective study and introduce a novel predictive formula using more comprehensive spinopelvic parameters, and then evaluated its reliability of prediction.
In our study, maxLL was significantly associated with maxTK (r = 0.564, P < 0.001), SS (r = 0.783, P < 0.001), PI (r = 0.483, P < 0.001) and PT (r=-0.155, P = 0.021), which suggested that the larger maxTK, PI and SS of the population is, the larger their maxLL will be, and vice versa, the smaller PT, the larger maxLL. The compensatory mechanisms of keeping sagittal balance are so complicated that it is a result of a reaction between the ground and an ideal dynamic chain between the spine, pelvis and lower limbs31, 32. The spine of patients suffering from lumbar degenerative diseases and low back pain is characterized by the loss of LL and SS, anterior sagittal imbalance and increase of PT32, and several compensatory mechanisms occurred to keep the whole sagittal balance, including reduction of maxTK and pelvic retroversion (increase of PT and decrease of SS), which could be verified in our findings. However, although no association was found between maxLL and age (r=-0.031, P = 0.643), we also found negative relationship between age and maxLL, which was consistent with Xu et al.’s study (r=-0.37)17.
In the unadjusted regression analysis, variables significantly associated with maxLL were included and analyzed and the results showed that maxTK and SS were the primary contributors to the maxLL. The sagittal spinal balance has been described as reciprocal curves of TK and LL, and the relationship between TK and LL play a key role in the sagittal balance13, 26, 27. In addition, the extension of adjacent segments (changes of TK and LL) is also an important compensatory mechanism of keeping sagittal balance27, 32. Therefore, it could be easily understood why maxTK has significant impact on the morphology of lumbar lordosis. SS was a positional parameter changing with age and pelvic position and also a compensatory mechanism of keeping sagittal balance32, hence it was an importance contributor in the prediction of LL as well. However, the parameters of SS changes with the position of spine and pelvic, which is not stable enough to predict lumbar lordosis, in addition, the compensatory reaction often did not occur because of the lack of motion in the spino-pelvic junction in the postoperative unphysiological spine, making us difficult to predict the compensatory parameters (SS and PT) before the surgery28. Furthermore, how to restore the positional parameters like SS and PT accurately in correction surgery remains a challenge for surgeons. Therefore, it is not reasonable to use SS to predict ideal maxLL.
PI is a fixed morphological parameter, which remains constant with the increasing age and changes of position18. Compared with other pelvic parameters, PI plays a more dominant role in the shape of sagittal curves33. Therefore, many researchers used PI to estimate ideal LL based on this theory17, 22–25, 28, 29. Furthermore, significant correlation was observed between PI and SS in our study (r = 0.680, P < 0.001), which was consistent with Vialle et al.’s study (r = 0.81)24 and Mac-Thiong et al.’s study (r = 0.76)34 Considering the factor that it is difficult to predict SS before the surgery, that PI plays a dominant role in the shape of lumbar lordosis and that PI is significantly correlated to SS, we replaced PI for SS in our adjusted regression analysis (all P < 0.001), and established a novel ideal lumbar lordosis prediction formula as follows: maxLL = 0.6*maxTK + 0.5*PI + 3.
In the analysis of validation, the mean value of predicted maxLL using our formula was 49.32o, and there was no significant difference between actual maxLL and our predicted maxLL (P = 0.408), suggesting a potent power of prediction. When comparing actual maxLL with predicted maxLL yielded by Yamato et al.28, Lee et al.25 Legaye et al.22, Schwab et al.23 and Rose et al.29, we found significant difference between actual maxLL and predicted maxLL yielded by these authors. In Yamato et al.’s28, Lee et al.’s25, Legaye et al.’s22 and Schwab et al.’s study23, only PI was used to predict the ideal maxLL in their formulas. In our opinion, it is not appropriate if we only use PI to predict the ideal LL as lumbar lordosis has been verified to be affected by many variables17, which should not be neglected. Although TK was used as a parameter in Rose et al.’s formula29 along with PI, significant difference was also observed between actual maxLL and predicted maxLL yielded by their formula. In their study29, the predicted formula was proposed by their hypotheses rather than scientific methods, although they found that the inclusion of both TK and PI displayed the greatest sensitivity in predicting LL, which might be primary contributor to their formula’s low effectiveness of prediction. In Xu et al.’s study17 age was included in their predictive formula, and there was no significant difference observed between actual maxLL and predicted maxLL yielded by their formula. However, we believed that the prediction effectiveness of Xu et al.’s formula could be contributed to the correlation between age and maxTK(r = 0.283,P < 0.001), PT(r = 0.266, P < 0.001) and SS(r=-0.213, P = 0.001),rather than age itself, because age was not found to be significantly associated with maxLL in our correlation test and unadjusted regression analysis. This could be contributed to the different measure methods to assess lumbar lordosis (LL and maxLL) between our and Xu et al.’s study and measure errors between different surgeons as well. In addition, no significant difference was also found between actual maxLL and predicted maxLL yielded by Vialle et al.’s study24, nevertheless, the prediction stability of formula was limited due to influence of spine and pelvic position and surgery choice on PT measurement, and too many variables affect the convenience of the formula application.
Furthermore, other two important things should be noticed when using our predictive formula. Firstly, all the values of parameters used in our formula were pre-operative values. PI remains constant before and after surgery, however, we should notice that the mean value of maxTK used in our formula is also pre-operative value because maxTK in majority of patients with mild or moderate sagittal balance remains constant if we do not perform osteoectomy at thoracic vertebras. With regard to patients with severe sagittal imbalance, osteoectomy is usually needed to perform at lower thoracic segments even at upper thoracic segments, therefore, the impact of osteoectomy on maxTK should be considered when we use this formula. Secondly, the lumbar lordosis restored during the spinal surgery might be smaller than our predicted maxLL as spontaneous compensation of pelvic parameters should be considered.
Although a robust and reliable formula was proposed in our study, there are some limitations of this study that should be addressed. First, all the patients we recruited in our study came from a single-center study and might result in the selection bias and compromise the statistical power. Second, sagittal parameters have been verified to be different between males and females; therefore, the predicted formula might be different between males and females, which was not evaluated in our study. Therefore, large-scaled and multicenter studies should be performed to make a more comprehensive research.