INTRODUCTION
Adult spinal deformity (ASD) is frequently associated with sagittal spinal malalignment, leading to significant functional disability. Since the pathological loss of lumbar lordosis (LL) relative to pelvic incidence (PI) (i.e., PI–LL mismatch) is a primary driver of global sagittal malalignment, the PI–LL parameter has been widely used over the years to assess the adequacy of correction [1]. However, concerns have been raised about its clinical utility. First, the PI–LL parameter was established based on patient disability in nonoperative patients, limiting its effectiveness in predicting mechanical complications such as proximal junctional kyphosis (PJK) [2]. Second, because the PI–LL measurement does not incorporate the pelvis, it is challenging to predict the position of the L1 vertebral body and the version of the pelvis (i.e., pelvic tilt [PT]) in a standing position [3].
L1 pelvic angle (L1PA), as a lumbar component of T1 pelvic angle (T1PA), was introduced to provide more comprehensive insights into the position of the L1 vertebral body and the version of the pelvis in relation to the hip joint (Fig. 1) [4]. Because L1PA is independent of patient positioning, the intraoperatively measured values can be reliably reproduced postoperatively. Accordingly, establishing the target value of L1PA can help achieve proper alignment correction during surgery. Since the L1PA is geometrically equivalent to PT minus L1 tilt, excessively low L1PA may be associated with posterior shift of the L1 vertebral body, increasing the risk of PJK. Conversely, correcting L1PA to an overly high value is associated with residual pelvic retroversion (represented by pelvic nonresponse [PNR]), potentially leading to suboptimal clinical outcomes [5]. Therefore, the ideal L1PA value should lie within thresholds that prevent both PJK and PNR.
The purpose of this study was to assess how L1PA influences the development of PJK and PNR and to determine the optimal L1PA targets for minimizing both conditions in patients with ASD undergoing low thoracic spine-to-pelvis fusion. Since it is reported that the normal value of L1PA depends on PI [6], the analysis was conducted across different PI categories.
MATERIALS AND METHODS
1. Study Population
This study was approved by the Institutional Review Board (IRB) of Samsung Medical Center (IRB No. 2024-07-144). Informed consent was waived due to the retrospective nature of the study. Patient records from a prospectively collected ASD database were analyzed retrospectively. The study cohort consisted of consecutive patients who underwent corrective surgery for symptomatic ASD, either degenerative flatback syndrome (DFBS) or degenerative lumbar kyphoscoliosis (DLKS), between 2014 and 2022. Patients were included if they met at least one of the following radiographic criteria: PI–LL mismatch ≥10°, PT ≥25°, sagittal vertical axis (SVA) ≥5 cm, or coronal Cobb angle ≥20°. The cohort was further refined to include only patients with an uppermost instrumented vertebra (UIV) at low thoracic vertebrae (T9–12) and a lower instrumented vertebra at the sacrum (with or without pelvic fixation). Iliac fixation was routinely performed using conventional iliac screws, except for patients with relatively mild deformity having solid fusion at the lumbosacral junction due to prior fusion surgery. Patients were excluded if they had spinal deformities secondary to acute trauma, tumors, acute infections, or neuromuscular diseases; had severe hip and knee arthritis; underwent revision surgery for reasons other than PJK such as rod fracture or infection; or failed to complete a 2-year follow-up.
2. Data Collection
We collected demographic data on sex, age, preoperative diagnosis (DFBS or DLKS), body mass index (BMI), modified frailty index-5 (mFI-5) score [7], osteoporosis (defined as the lowest T-score of the spine or hip≤-2.5), and Hounsfield units at the UIV [8]. For surgical factors, we evaluated the use of anterior lumbar interbody fusion (ALIF) at L5–S1, lateral lumbar interbody fusion (LLIF) at ≥L4–5, pedicle subtraction osteotomy (PSO), pelvic fixation, cement augmentation at UIV, hook fixation at UIV+1, and total fusion length.
Radiographic parameters, PI–LL, sacral slope (SS), PT, thoracolumbar junctional angle (TLJA, T11-L1), thoracic kyphosis (TK), T1PA, and SVA were measured preoperatively and at 6 weeks, and 2 years after surgery. In cases where PJK was detected within 6 weeks, immediate postoperative radiographs were used. To evaluate the shape of the LL, the 6-week LL was divided into upper LL (ULL, L1–4) and lower LL (LLL, L4–S1), then the lordosis distribution index (LDI) was calculated as LLL divided by total LL, multiplied by 100. L1PA was measured on the 6-week radiograph as the angle formed by a line from the center of the L1 vertebral body to the femoral head axis and a line from the femoral head axis to the center of the S1 endplate (Fig. 1). Lastly, L1 tilt was evaluated using the following equation: L1 tilt=PT–L1PA.
3. Outcome Measures
The primary outcomes of interest were PJK and PNR at 2-year follow-up duration. To specifically evaluate the impact of sagittalplane correction on the development of PJK and PNR, we limited the follow-up period to 2-years postsurgery. We assumed that the late development of PJK or PNR might not be directly influenced by the initial surgical correction but could instead result from biologic factors [9]. PJK was defined radiographically as a proximal junctional angle of ≥20° [10]. PNR was defined as undercorrection relative to the patient’s age-adjusted PT target [PT=(age–55)/3+20] at 6 weeks, according to a previous study [5]. PT undercorrection indicates that the postoperative PT exceeds the age-adjusted PT value expected for a person 10 years older than the patient’s actual age. This PT must be maintained for 2 years to be considered undercorrected. Based on the presence of PJK and PNR, patients were categorized into 3 groups: no PJK/PNR, PJK, and PNR. At 2 years postoperatively, clinical outcomes were evaluated using the Oswestry Disability Index (ODI) and Scoliosis Research Society-22r questionnaire (SRS-22r). Additionally, achievement of minimal clinically important difference (MCID) was evaluated using thresholds of 18.9 for ODI and 0.98 for the SRS-22r subtotal score [11].
4. Statistical Analysis
Baseline sagittal parameters, including PI–LL, SS, PT, TLJA, TK, T1PA, and SVA, were compared according to preoperative diagnoses (DFBS vs. DLKS) using independent t-tests. Preoperative and 6-week postoperative sagittal parameters, including PI–LL, SS, PT, TLJA, TK, T1PA, and SVA, were compared using paired t-tests. Risk factor analysis was conducted separately for PJK and PNR by comparing variables between the no PJK/PNR and PJK groups as well as between the no PJK/PNR and PNR groups. Inter- and intraobserver reliability for radiographic measurements was assessed using the intraclass correlation coefficient (ICC). ICC values <0.5, 0.5–0.75, 0.75–0.9, and >0.9 are indicative of poor, moderate, good, and excellent reliability, respectively. Time-dependent changes for TLJA and TK were evaluated according to the occurrence of PJK [12]. Demographic, operative, and radiographic data were compared using independent t-tests and chi-square tests, followed by multivariate logistic regression analysis utilizing significant variables identified in the univariate analysis. Receiver operating characteristic (ROC) curve analysis was performed to calculate the cutoff values of L1PA for the development of PJK and PNR. The threshold of L1PA for PJK indicates the lowest permissible L1PA value that does not lead to PJK, while the threshold of L1PA for PNR represents the highest permissible L1PA that does not result in PNR. Since the normal L1PA varies based on an individual’s PI [6], the ROC curve analysis to determine the cutoff value of L1PA was conducted separately according to 3 PI categories (PI<45°, 45°≤PI<60°, and PI≥60°) [13]. Based on the upper and lower thresholds of L1PA, patients were classified into low, ideal, and high L1PA groups. The L1PA was deemed ideal when it fell between the calculated upper and lower thresholds. We then compared the 6-week key sagittal parameters and 2-year outcomes across the 3 L1PA groups using analysis of variance or chi-square tests. Linear regression model was established to identify the relationship between L1PA and other intraoperatively measurable sagittal parameters including PI, ULL, and LLL. All statistical analyses were performed by professional statisticians using IBM SPSS Statistics ver. 29.0 (IBM Co., USA). A p-value of <0.05 was considered statistically significant.
RESULTS
1. Baseline Characteristics
A total of 323 patients (88.5% females; mean age, 69.5 years) were included in the study (Table 1). Preoperative diagnoses included DFBS in 194 patients (60.1%) and DLKS in 129 (39.9%). The mean BMI was 25.9 kg/m2 and the mean mFI-5 was 1.3. Osteoporosis was present in 65 patients (20.1%). The mean Hounsfield unit was 113.9. Regarding surgical techniques, 44.9% of patients underwent ALIF at L5–S1, and 66.9% had LLIF at ≥L4–5. PSO was performed in 32 patients (9.9%). Pelvic fixation was carried out in 266 patients (82.4%). Cement augmentation and hook fixation were performed in 96 (29.7%) and 66 patients (20.4%), respectively. The mean number of fused segments was 8.0. There were no significant differences in the baseline sagittal parameters between DFBS and DLKS except for SS, which was slightly greater in DLKS than in DFBS (22.6° vs. 20.0°, p=0.039) (Fig. 2). All sagittal parameters were significantly changed postoperatively, including PI–LL (41.4° to 6.9°), SS (21.1° to 34.9°), PT (32.6° to 18.7°), TLJA (6.6° to 9.4°), TK (12.1° to 25.6°), T1PA (32.6° to 15.7°), and SVA (80.1 mm to 20.8 mm) (Table 2). The ICC values for intraobserver reliability were 0.91 for PI–LL, 0.95 for SS, 0.93 for PT, 0.94 for TLJA, 0.86 for TK, 0.96 for T1PA, 0.93 for C7–SVA, and for 0.94 for L1PA. The ICC values for interobserver reliability were similar to those of intraobserver reliability with 0.89 for PI–LL, 0.94 for SS, 0.91 for PT, 0.94 for TLJA, 0.81 for TK, 0.93 for T1PA, 0.90 for C7–SVA, and 0.93 for L1PA. Particularly, the intra- and interobserver reliabilities for L1PA measurement were consistent with previous study reporting ICC values between 0.91 and 0.97 [14].
2. Risk Factors of PJK and PNR
At 2 years, PJK and PNR were identified in 101 (31.3%) and 49 patients (15.2%), respectively. Eight patients (2.5%) presented with both PJK and PNR. Univariate analysis for PJK showed that patients in the PJK group were significantly older than those in the non-PJK/PNR group (70.7 years vs. 69.3 years, p=0.040) (Table 3). Postoperative PI–LL (2.8° vs. 5.7°, p=0.016) and L1PA (7.5° vs. 10.0°, p=0.001) were significantly lower in the PJK group than in the non-PJK/PNR group, whereas TLJA (13.1° vs. 8.1°, p<0.001) and TK (30.2° vs. 24.2°, p<0.001) were significantly higher in the PJK group. Time-dependent changes in TLJA demonstrated that PJK group had significantly greater values than the non-PJK group at the preoperative, 6-week, and 2-year time points (Fig. 3A). Similarly, TK was significantly higher in the PJK group at 6 weeks and 2 years postoperatively; however, no significant difference was observed in preoperative TK between the non-PJK and PJK groups (Fig. 3B). In univariate analysis for PNR, patients in the PNR group underwent significantly fewer LLIF procedures than those in the non-PJK/PNR group (42.9% vs. 68.5%, p<0.001). Postoperative PI–LL (21.1° vs. 5.7°, p<0.001), PT (32.6° vs. 16.2°, p<0.001), T1PA (26.0° vs. 14.1°, p<0.001), and L1PA (20.0° vs. 10.0°, p<0.001) were significantly higher in the PNR group, while postoperative SS was significantly lower in the PNR group than in the non-PJK/PNR group (31.2° vs. 35.4°, p=0.005). Multivariate logistic regression analysis revealed that postoperative high TLJA (odds ratio [OR], 1.078; 95% confidence interval [CI], 1.029–1.129; p=0.001), high TK (OR, 1.040; 95% CI, 1.009–1.071; p=0.012), and low L1PA (OR, 0.927; 95% CI, 0.869–0.988; p=0.019) were significant risk factors for PJK, while high PI–LL (OR, 1.101; 95% CI, 1.039–1.167; p<0.001) and high L1PA (OR, 1.249; 95% CI, 1.133–1.377; p<0.001) were significant risk factors for PNR (Table 4).
3. Calculation of L1PA Thresholds
In patients with PI less than 45°, the lower threshold of L1PA (to avoid PJK) was 2.5° (area under the curve [AUC], 0.694; p=0.003) and upper threshold (not to cause PNR) was 4.5° (AUC, 0.979; p<0.001) (Fig. 4A and B). In patients with PI between 45° and 60°, the lower and upper thresholds of L1PA were 8.7° (AUC, 0.709; p<0.001) and 12.6° (AUC, 0.881; p<0.001), respectively (Fig. 4C and D). In patients with PI greater than 60°, the lower and upper thresholds of L1PA were 15.1° (AUC, 0.728; p<0.001) and 17.3° (AUC, 0.861; p<0.001), respectively (Fig. 4E and F).
4. Six-Week Sagittal Parameters and 2-Year Outcomes According to L1PA Groups
Using predefined L1PA thresholds, patients were categorized into low, ideal, and high L1PA groups with 132, 65, and 126 patients, respectively (Table 5). In 6-week sagittal parameters, L1PA (5.4° vs. 10.8° vs. 16.0°), PI–LL (0.3° vs. 6.2° vs. 14.2°), and PT (14.0° vs. 17.8° vs. 24.0°) were significantly lower in the low L1PA group, followed by the ideal and high L1PA groups. While ULL did not differ significantly among the 3 groups, a significant difference in LLL was observed with the highest values in the low L1PA group, followed by the ideal and high L1PA groups (33.4° vs. 28.6° vs. 23.9°, p<0.001). The LDI and L1 tilt values were not significantly different across the L1PA groups. The incidence of PJK varied significantly among the L1PA groups (p<0.001). In subanalysis, significant differences in PJK rates were observed between the low and ideal L1PA groups (p<0.001), but not between the ideal and high L1PA groups (p=0.388). For the PNR, the incidence also varied significantly across the L1PA groups (p<0.001). In contrast to PJK, the difference in PNR rates was significant between the ideal and high L1PA groups (p<0.001), while no significant difference was found between the low and ideal L1PA groups (p=0.166). Revision surgery was performed more frequently in the low L1PA group; however, statistically significant differences were observed among the groups. Regarding clinical outcomes, the ODI and SRS-22r scores were best in the ideal L1PA group than in the low and high L1PA groups. The proportions of patients achieving MCID for ODI and SRS-22r were also significantly higher in the ideal L1PA group compared to the low and high L1PA groups. A linear regression model demonstrated that L1PA is associated with intraoperatively measurable parameters, including PI, ULL, and LLL, as follows: L1PA=-7.18+0.64×PI−0.22×ULL−0.43×LLL (adjusted R2=0.725).
DISCUSSION
PJK and PNR are representative unfavorable events that compromise the improvement of clinical outcomes following ASD surgery [15]. Therefore, the prevention of both PJK and PNR is critical for the success of surgery. From the perspective of lordosis correction, PJK and PNR exhibit opposing characteristics: overcorrection of lordosis is associated with a higher risk of PJK, while undercorrection may predispose patients to PNR [16]. Confirming this phenomenon, Passias et al. [5] recommended that surgeons should find a middle ground of PI–LL correction goals to mitigate both PJK and PNR. It is well established that PJK results from the overcorrection of spinopelvic alignment [17-19]; however, recent studies have demonstrated that its risk can also be increased in cases of PNR secondary to undercorrection [20,21]. However, considering that PJK developed most frequently in the low L1PA group followed by the ideal and high L1PA groups in the present study, the occurrence of PJK appears to be more strongly influenced by L1PA overcorrection than by undercorrection. We also noticed that 8 patients (2.5%) experienced both PJK and PNR, indicating that there may be partial overlap in their occurrence despite the reportedly opposing mechanisms. Moreover, PJK still developed in 15.1% of patients even in the high L1PA group (i.e., L1PA undercorrection) (Table 5). These results suggest that any single parameter, either L1PA or another one, could not completely prevent both adverse events. These findings are the reasons why we included 8 patients with both PJK and PNR in the analysis.
Although the L1PA can be measured intraoperatively in the prone position, its clinical utility needs to be evaluated in relation to PJK and PNR, as both conditions occur in the standing position. In this study, we found that the L1PA significantly influenced the development of both PJK and PNR. The L1PA was significantly lower in the PJK group and significantly higher in the PNR group than in the non-PJK/PNR group (Tables 3 and 4). The optimal L1PA to avoid both PJK and PNR was calculated to be between 2.5° and 4.5° for patients with PI<45°, between 8.7° and 12.6° for patients with PI between 45° and 60°, and between 15.1° and 17.3° for patients with PI≥60° (Fig. 3). These cutoff values effectively discriminated the development of PJK and PNR (Table 5). In the subanalysis, the incidence of PJK was significantly higher in the low L1PA group compared to the ideal L1PA group, but not significantly different between the ideal and high L1PA groups. PNR occurred more frequently in the high L1PA group than in the ideal L1PA group, while the incidence of PNR was comparable between the low and ideal L1PA groups. This finding supports the clinical utility of our L1PA cutoff values in predicting both PJK and PNR. Notably, the optimal range of L1PA was very narrow: approximately 2° for patients with PI ≤45°, 4° for those with PI between 45° and 60°, and 2° for those with PI≥60°. The ideal ranges of previous alignment schemes were reported as 18° for PI–LL in Schwab criteria [22], 25° for ideal LL in the global alignment and proportion score [23], and 10° for PI–LL in the age-adjusted metric [24]. Compared to other alignment schemes that suggest broader ideal ranges for LL or PI–LL parameters, the narrow range of L1PA enhances its clinical applicability by allowing surgeons to target specific values (e.g., 3° for PI<45°, 10° for PI between 45° and 60°, and 16° for PI≥60°) rather than a broad range.
Several optimal L1PA values have been suggested in previous studies. Protopsaltis et al. [4], who first introduced the L1PA, studied the relationship between L1PA and health-related quality of life (HRQoL) in nonoperative ASD patients. They found that a high L1PA was associated with poor HRQoL and reported a cutoff value of 7.2°, corresponding to an ODI of 20. This method of defining an ideal L1PA cutoff is similar to that of Schwab et al. [22], who proposed a PI–LL cutoff value of ±9° based on ODI. Accordingly, their L1PA cutoff may be insufficient to address alignment-related issues such as PJK and PNR. In agreement with our results, Duvvuri et al. found that the L1PA was significantly lower in the PJK group than in the non-PJK group (7.5° vs. 10.4°, p=0.001) [25]. Although they presented a relatively narrow range of ideal L1PA (probably between 7.5° and 10.4°), its clinical application may be limited as they did not account for individual PI values. Lastly, Hills et al. proposed normal values for L1PA in a study on sagittal alignment in asymptomatic healthy adults [6]. They found that PI was strongly associated with L1PA and established the following: L1PA=0.5×PI–21° (R2=0.58). They also observed that for a given PI, a higher L1PA is associated with a lower LL (i.e., a higher PI–LL mismatch), and vice versa.
Consistent with Hills’s study, we found that L1PA positively correlated with PI and PI–LL mismatch (Table 5). When dividing LL into ULL and LLL, L1PA appears to be more influenced by LLL than by ULL. This study revealed that LLL was highest in the low L1PA group (i.e., L1PA overcorrection) and lowest in the high L1PA group (i.e., L1PA undercorrection), whereas ULL showed no significant differences among the 3 L1PA groups (Table 5). This relationship between L1PA and ULL/LLL is also well demonstrated in the linear regression analysis; L1PA is more strongly influenced by LLL (β=-0.43) than by ULL (β=-0.22). Although L1PA is a parameter that can be measured during surgery, it cannot be directly manipulated by surgeons unlike L1–S1 lordosis because L1PA encompasses the non-fused joints (i.e., hip joints). Instead, L1PA values are passively determined by the amount of LL correction (ULL and LLL). Therefore, understanding the relationship between L1PA and ULL/LLL may help guide the achievement of an appropriate L1PA during surgery. The fact that LLL was highest in the low L1PA group corresponded with a trend toward higher LDI in the same group; however, this finding was not statistically significant. Finally, the occurrence of PJK has been reported to be associated with the posterior orientation of the L1 vertebra, such as an increased L1 tilt angle or L1 slope [26]. Since the L1 tilt angle is arithmetically PT minus L1PA, one might assume that a lower L1PA may lead to a higher L1 tilt angle. This assumption may be true if the pelvis is a fixed structure. In the present study, we found that L1 tilt angle was comparable among the 3 L1PA groups. This result may be explained by the fact that as L1PA decreases, PT also decreases significantly in a concurrent manner. Furthermore, although the L1 tilt angle in the standing position is certainly an important factor in PJK development, it cannot be directly manipulated intraoperatively, unlike L1PA. Therefore, L1PA may serve as a more practical parameter for predicting and preventing PJK in clinical settings.
This study has some limitations. First, the retrospective nature of the study, relying on patients from a single institution, may introduce selection and recall bias. Additionally, surgical outcomes may vary depending on the surgeon’s experience and skills, which could lead to performance bias. Second, the study cohort included patients with a UIV in the lower thoracic spine (T9–12), so our findings may not be fully generalizable to patients undergoing fusion extending to the upper thoracic spine. We selected these patients for the following reasons: (1) only a few patients at our institution underwent fusion to the upper thoracic spine, and (2) the significant distance between the L1 vertebra and UIV in cases with fusion to the upper thoracic spine may reduce the direct load generated by L1PA correction on the proximal junction. Third, we acknowledge that in the multivariate analysis, not only L1PA but also high TLJA and TK were significant risk factors for PJK occrrence. However, since this study included cases with UIV at T12 and T11, these 2 variables of TLJA and TK may be influenced by the development of PJK, unlike L1PA. Thus, while high TLJA and TK may serve as risk factors, they could also be consequences of PJK. Fourth, while L1PA was identified as a common risk factor for both PJK and PNR, other factors, such as thoracolumbar soft tissue integrity and muscle function, may also contribute to these occurrences [27,28]. Lastly, although we proposed optimal values for L1PA, it remains unclear whether this metric is superior to previously established alignment schemes. Further studies are needed to address this question.
CONCLUSION
This study demonstrated that optimal correction based on these L1PA targets reduced the risk of both PJK and PNR. The optimal L1PA ranges were 2.5° to 4.5° for patients with PI<45°, 8.7° to 12.6° for patients with PI between 45° and 60°, and 15.1° to 17.3° for patients with PI≥60°. These L1PA target values can serve as reliable alignment goals to optimize the surgical outcomes in ASD surgery.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Full text via PMC
Download Citation
Print
