Abstract
Study Design
Retrospective cohort study.
Objectives
To determine whether thoracic compensation (TC) assessed against the deformity burden is associated with proximal junctional kyphosis (PJK) after lower thoracic fusion for adult spinal deformity (ASD).
Methods
We reviewed 343 patients fused from T9-L1 to the sacrum or pelvis with a pelvic incidence (PI) minus lumbar lordosis (LL) mismatch above 20°. Expected thoracic kyphosis (TK) was derived from the PI-LL mismatch using published equations, and a measured TK below it defined TC. PJK was defined as a proximal junctional angle of 20° or more, or revision, within two years. Multivariable logistic regression adjusted for patient, operative, and radiographic covariates including two measures of correction adequacy.
Results
PJK occurred in 60 of 124 patients (48.4%) without TC and 68 of 219 (31.1%) with TC (P = 0.002). Correction was equivalent for PI-LL, the L1 pelvic angle (L1PA), and pelvic tilt, yet the TC present group restored more TK (19.1° vs. 11.0°) while reaching a lower postoperative value (21.5° vs. 33.7°, both P < 0.001). TC remained associated with a lower risk of PJK after adjustment (odds ratio 0.431, 95% CI 0.261-0.713, P = 0.001), consistently across eight prespecified subgroups. Relative to TC without overcorrection, the odds ratio reached 3.41 (95% CI 1.37-8.46) when TC was absent and the L1PA offset indicated overcorrection.
Conclusions
TC assessed against the deformity burden was associated with a lower risk of PJK despite a larger reciprocal increase in TK. Risk was highest when TC was absent and lordosis overcorrected.
Keywords: Adult spinal deformity, Thoracic compensation, Thoracic kyphosis, Proximal junctional kyphosis, Sagittal alignment, L1 pelvic angle
Introduction
Adult spinal deformity (ASD) is a major source of disability in the aging population, and sagittal malalignment correlates closely with pain and functional impairment. 1 Loss of lumbar lordosis (LL) is the principal driver of global sagittal malalignment in many ASD cases. Once the lumbar spine progresses to kyphotic deformity, patients recruit thoracic compensation (TC) by reducing the thoracic kyphosis (TK) (i.e., flattening of the thoracic spine) in an effort to shift the trunk posteriorly and preserve horizontal gaze. 2 The capacity to recruit this response reportedly varies between patients according to age, severity of the lumbar deformity, paraspinal musculature, and the flexibility of the thoracic segments.3,4
Proximal junctional kyphosis (PJK) is among the most frequent mechanical complications after ASD surgery. Although multifactorial causes are involved in PJK development,5-11 its final pathway ends in unfavorable biomechanical loads at the proximal junction. When the deformity is confined to the lumbar spine, as frequently seen in many degenerative ASD cases, fusion terminates at the lower thoracic spine, leaving most of the thoracic spine unfused and free to change after surgery. The restoration of LL is the primary goal of corrective surgery, and individualized alignment targets have been developed to guide this correction.12-16 Understanding how the unfused thoracic spine behaves before and after surgery is as clinically meaningful as LL correction, since the proximal junction lies at the base of the unfused thoracic spine and is therefore affected by how that segment is aligned. We hypothesized that preoperative TC status would influence the development of PJK by shaping the postoperative thoracic alignment.
Few studies have examined the relationship between preoperative TC status and PJK occurrence. We therefore aimed to investigate whether and how the preoperative TC status influences the development of PJK in patients undergoing lower thoracic fusion for ASD. To isolate the impact of preoperative TC, potential confounders were adjusted, including the magnitude of lordosis correction and patient and surgical factors.
Methods
Study design and population
This retrospective study of a prospectively maintained adult spinal deformity (ASD) database was approved by the institutional review board of our institution, which waived the requirement for informed consent (IRB no. 2026-06-095). Consecutive patients who underwent long instrumented fusion for ASD at a single tertiary referral center between 2015 and 2024 were reviewed. Inclusion criteria were: (1) ASD with significant sagittal malalignment (pelvic incidence [PI]-LL mismatch greater than 20°, corresponding to the SRS-Schwab ++ modifier); (2) long instrumented fusion with the upper instrumented vertebra (UIV) between T9 and L1 and the lower instrumented vertebra at the sacrum or pelvis; and (3) completion of the two-year follow-up. Patients were excluded for non-degenerative deformity; for primarily coronal deformity without sagittal malalignment; or when revision surgery was performed for reasons other than proximal junctional complications. A total of 343 patients were analyzed (Figure 1).
Figure 1.
Patient selection. Patients who underwent corrective surgery for adult spinal deformity between 2015 and 2024 were screened. Those not meeting the inclusion criteria and those without a complete two-year follow-up were excluded, leaving 343 patients for analysis. PJK was defined as a proximal junctional angle of 20° or more, or revision surgery performed for a proximal junctional complication, within two years. ASD, adult spinal deformity; UIV, upper instrumented vertebra; PI, pelvic incidence; LL, lumbar lordosis; PJK, proximal junctional kyphosis
Data collection
Demographic variables included age, sex, body mass index (BMI), bone mineral density (BMD, T-score), American Society of Anesthesiologists (ASA) physical status, and frailty score based on 5-item modified frailty index (mFI-5). Surgical factors included surgical approach (combined anterior vs. posterior-alone), pedicle subtraction osteotomy, pelvic fixation, cement augmentation at UIV, transverse process hook at UIV, and number of fused segments. For radiographic analysis, standing whole-spine radiographs were obtained before surgery, at the first postoperative standing examination, and at each subsequent visit. The following parameters were measured: PI, PI minus LL (PI-LL), L1 pelvic angle (L1PA), pelvic tilt (PT), TK (between the superior endplate of T4 and the inferior endplate of T12), and sagittal vertical axis (SVA). Two measures of correction adequacy were derived from the first postoperative standing radiograph. The L1PA offset was the measured L1PA minus the ideal value of 0.5 × PI - 21, 15 and the age-adjusted PI-LL offset was the measured PI-LL minus the age-adjusted target of (age - 55)/2 + 3. 17 A negative value of either offset indicated relative overcorrection.
Definition of thoracic compensation
Thoracic compensation (TC) was defined by comparing the measured preoperative TK with the value expected for the degree of lumbar deformity, using the reference equations of Li et al. 18 Expected TK was obtained as -0.51 × (PI-LL) + 31.24 for patients with a PI below 50° and as -0.54 × (PI-LL) + 37.19 for patients with a PI of 50° or above. Because a lower TK reflects greater compensatory flattening, a measured TK below the expected value indicated that compensation had been recruited, and patients were assigned to the TC present group or the TC absent group accordingly.
Outcome
The outcome was PJK, defined as a PJA of 20° or more, or revision surgery performed for proximal junctional complications, within two years of the index operation. 19 The time to PJK was taken as the interval from surgery to the first radiograph meeting this definition or to the revision procedure, and patients without the event were censored at two years.
Statistical analysis
Continuous variables are presented as the mean with standard deviation and categorical variables as the number with percentage. The independent t test was applied to continuous variables that were normally distributed on the Shapiro-Wilk test, and the Mann-Whitney U test to the remainder. The chi-square test or Fisher exact test was used for categorical variables. The cumulative incidence of PJK was analyzed by the Kaplan-Meier method and compared with the log-rank test. Univariable logistic regression was performed for every demographic, operative, and radiographic variable. Variables with a P value below 0.10 were considered for entry into a multivariable model alongside the TC status. Among the radiographic variables, only those available before surgery or set during surgery were eligible for the multivariable model, since the model was intended to describe risk that can be judged at the time of planning. Postoperative PI-LL and postoperative L1PA were withheld because they overlap with the age-adjusted PI-LL offset and the L1PA offset, and postoperative TK was withheld because it is not modifiable or measurable intraoperatively. The final model comprised TC together with age, sex, BMD, ASA grade, number of fused segments, postoperative L1PA offset, and postoperative age-adjusted PI-LL offset. The performance of the multivariable model was examined in three respects. Discrimination was quantified by the area under the receiver operating characteristic curve (AUC) and corrected for optimism by bootstrap resampling with 1,000 replications. Calibration was assessed by the Hosmer-Lemeshow test and collinearity by variance inflation factors. The model was then refitted with preoperative TK and with postoperative TK entered in turn, and the resulting estimates are reported descriptively rather than as direct or indirect effects. The offset between the measured and the expected TK was also entered as a continuous variable in place of the TC classification. Models carrying the offset, preoperative TK, and the TC classification were compared by the Akaike information criterion and by likelihood ratio tests against the model with covariates alone. Patients were also cross-classified by L1PA offset (>0° vs. ≤0°) and TC status, and adjusted odds ratios (ORs) were obtained for the four groups with the same covariate set excluding the L1PA offset. Subgroup analyses were performed within strata of eight prespecified variables, using the same covariate set with the stratifying variable omitted from the multivariable model. Heterogeneity of the association across strata was assessed by a likelihood ratio test of the product term between TC and the subgroup indicator. Statistical analyses were performed with Python version 3.12 (Python Software Foundation, Wilmington, DE, USA). Statistical significance was set at P < 0.05.
Results
Cohort characteristics and incidence of PJK
Of 343 patients, 124 were TC absent and 219 TC present. The mean age was 69.9 ± 6.8 years, 303 patients (88.3%) were women, and a mean of 7.1 ± 1.4 segments were fused (Table 1). The groups differed only in age (71.7 ± 6.2 vs 68.9 ± 7.0 years, P < 0.001), while sex, BMI, BMD, ASA grade, frailty score, and all operative characteristics were comparable. PJK occurred in a total of 128 patients (37.3%) with significantly higher incidence in the TC absent group (60/124, 48.4%) than in the TC present group (68/219, 31.1%) (P = 0.002). The Kaplan-Meier curve showed a lower cumulative incidence in the TC present group than in the TC absent group (log-rank P = 0.001) (Figure 2). Among the 128 patients who met the PJK definition, 106 (82.8%) reached a proximal junctional angle of 20° or more without revision and 22 (17.2%) underwent revision surgery for a proximal junctional complication. Events occurred early, with 68 of 128 (53.1%) within three months and 95 (74.2%) within six months.
Table 1.
Baseline demographic and operative characteristics according to the presence of preoperative thoracic compensation
| Variable | Total (n = 343) | TC absent (n = 124) | TC present (n = 219) | P |
|---|---|---|---|---|
| Demographic characteristics | ||||
| Age, years | 69.9 ± 6.8 | 71.7 ± 6.2 | 68.9 ± 7.0 | <0.001 |
| Female sex, n (%) | 303 (88.3) | 104 (83.9) | 199 (90.9) | 0.078 |
| BMI, kg/m2 | 26.0 ± 3.8 | 26.2 ± 3.3 | 25.8 ± 4.1 | 0.132 |
| BMD, T-score | -1.49 ± 1.33 | -1.37 ± 1.45 | -1.56 ± 1.25 | 0.493 |
| ASA physical status, n (%) | 0.325 | |||
| Grade 1 | 16 (4.7) | 3 (2.4) | 13 (5.9) | |
| Grade 2 | 271 (79.0) | 101 (81.5) | 170 (77.6) | |
| Grade 3 | 56 (16.3) | 20 (16.1) | 36 (16.4) | |
| Frailty score (mFI-5) | 1.34 ± 0.89 | 1.36 ± 0.90 | 1.33 ± 0.90 | 0.809 |
| Operative characteristics | ||||
| Combined anterior surgery, n (%) | 249 (72.6) | 92 (74.2) | 157 (71.7) | 0.709 |
| Pedicle subtraction osteotomy, n (%) | 34 (9.9) | 11 (8.9) | 23 (10.5) | 0.766 |
| Pelvic fixation, n (%) | 263 (76.7) | 97 (78.2) | 166 (75.8) | 0.706 |
| Cement augmentation at UIV, n (%) | 102 (29.7) | 30 (24.2) | 72 (32.9) | 0.117 |
| Transverse process hook at UIV, n (%) | 88 (25.7) | 31 (25.0) | 57 (26.0) | 0.936 |
| Number of fused segments, levels | 7.1 ± 1.4 | 7.2 ± 1.4 | 7.0 ± 1.4 | 0.273 |
| Two-year outcome | ||||
| Proximal junctional kyphosis, n (%) | 128 (37.3) | 60 (48.4) | 68 (31.1) | 0.002 |
Values are presented as mean ± standard deviation or number (%).
Bold P values indicate statistical significance.
TC, thoracic compensation; BMI, body mass index; BMD, bone mineral density; ASA, American Society of Anesthesiologists; mFI-5, 5-item modified frailty index; UIV, upper instrumented vertebra.
Figure 2.
Cumulative incidence of PJK according to the presence of thoracic compensation, estimated by the Kaplan-Meier method. Shaded area represent 95% confidence intervals. TC, thoracic compensation; PJK, proximal junctional kyphosis
Sagittal parameters
PI, PI-LL, and L1PA did not differ between the groups before or after surgery, and neither did the postoperative L1PA offset (3.9 ± 5.0 vs 4.4 ± 4.8°, P = 0.414). PT did not differ before surgery, but was significantly higher in the TC absent group (19.3 ± 8.6 vs. 17.0 ± 8.8°, P = 0.016). TK was higher in the TC absent group both before (22.6 ± 13.5 vs 2.3 ± 13.8°) and after surgery (33.7 ± 9.0 vs 21.5 ± 9.7°, both P < 0.001), as was SVA. The age-adjusted PI-LL offset was more negative in the TC absent group, indicating greater overcorrection (-6.2 ± 11.3 vs -3.2 ± 11.5°, P = 0.018) (Table 2). The perioperative change in PI, PI-LL, L1PA, and PT was equivalent (P = 0.228 to 0.636). The TC present group showed a greater increase in TK (19.1 ± 12.7 vs 11.0 ± 12.5°, P < 0.001) and a smaller reduction in SVA (-51.0 ± 55.0 vs -68.0 ± 61.9 mm, P = 0.020) (Figure 3).
Table 2.
Sagittal radiographic parameters before and after surgery according to the presence of preoperative thoracic compensation
| Parameters | Preoperative | Postoperativent | ||||
|---|---|---|---|---|---|---|
| TC absent | TC present | P | TC absent | TC present | P | |
| PI (°) | 53.7 ± 10.9 | 53.9 ± 11.1 | 0.866 | 53.5 ± 10.1 | 53.6 ± 10.7 | 0.970 |
| PI-LL (°) | 38.0 ± 21.5 | 41.7 ± 18.6 | 0.105 | 5.1 ± 10.8 | 6.7 ± 11.1 | 0.177 |
| L1PA (°) | 18.6 ± 9.3 | 20.5 ± 9.6 | 0.070 | 9.7 ± 6.7 | 10.1 ± 7.1 | 0.383 |
| PT (°) | 33.7 ± 11.3 | 31.9 ± 10.7 | 0.140 | 19.3 ± 8.6 | 17.0 ± 8.8 | 0.016 |
| TK (°) | 22.6 ± 13.5 | 2.3 ± 13.8 | <0.001 | 33.7 ± 9.0 | 21.5 ± 9.7 | <0.001 |
| SVA (mm) | 93.3 ± 60.5 | 69.2 ± 50.9 | <0.001 | 25.3 ± 30.3 | 18.2 ± 30.3 | 0.038 |
| Postoperative alignment offsets | ||||||
| Age-adjusted PI-LL offset (°) | - | - | - | -6.2 ± 11.3 | -3.2 ± 11.5 | 0.018 |
| L1PA offset (°) | - | - | - | 3.9 ± 5.0 | 4.4 ± 4.8 | 0.414 |
Values are presented as mean ± standard deviation.
Bold P values indicate statistical significance.
TC, thoracic compensation; PI, pelvic incidence; LL, lumbar lordosis; L1PA, L1 pelvic angle; PT, pelvic tilt; TK, thoracic kyphosis; SVA, sagittal vertical axis; Age-adjusted PI-LL offset, difference between the postoperative PI-LL and the age-adjusted PI-LL target; L1PA offset, difference between the measured and the ideal L1 pelvic angle.
Figure 3.
Comparison of perioperative changes in sagittal parameters according to the presence of preoperative thoracic compensation. (A) PI, (B) PI-LL, (C) L1PA, (D) PT, (E) TK, and (F) SVA. TC, thoracic compensation; PI, pelvic incidence; LL, lumbar lordosis; L1PA, L1 pelvic angle; PT, pelvic tilt; TK, thoracic kyphosis; SVA, sagittal vertical axis
Factors associated with PJK
TC was associated with a lower risk of PJK on univariable analysis (OR 0.480, 95% CI 0.305-0.756) and after multivariable adjustment (OR 0.431, 95% CI 0.261-0.713, P = 0.001) (Table 3). Female sex (OR 3.656, 95% CI 1.501-8.905), ASA grade (1.751, 95% CI 1.003-3.059), the number of fused segments (0.798, 95% CI 0.668-0.955), and the postoperative L1PA offset (0.932, 95% CI 0.880-0.988) were also independently associated with PJK, whereas no preoperative radiographic parameter was (P = 0.163 to 0.872). The model showed an AUC of 0.723, which fell to 0.696 after bootstrap correction for optimism, indicating little overfitting. Calibration was adequate on Hosmer-Lemeshow test (chi-square 2.72, df 8, P = 0.951), and the maximum variance inflation factor was 1.54, indicating no collinearity among the variables. Because TC is defined from TK, the model was refitted with each TK term entered in turn. Adding preoperative TK strengthened the association (OR 0.355, 95% CI 0.191-0.662, P = 0.001), whereas adding postoperative TK attenuated it (OR 0.683, 95% CI 0.380-1.227, P = 0.202). The offset between the measured and the expected TK was also analysed as a continuous variable. A higher offset, indicating less compensation, was associated with PJK after adjustment for the same covariates (OR 1.023, 95% CI 1.002-1.044, P = 0.035), whereas preoperative TK entered in its place was not (OR 1.008, 95% CI 0.993-1.023, P = 0.309). The model carrying the offset showed better fit and discrimination than the model carrying preoperative TK (AIC 429.2 vs 432.7 and AUC 0.711 vs 0.699). The model carrying TC as a dichotomy fitted better still (AIC 422.8 and AUC 0.723), and when the dichotomy and the offset were entered together only the dichotomy remained associated with PJK (OR 0.378, 95% CI 0.179-0.798, P = 0.011 and OR 0.993, 95% CI 0.963-1.023, P = 0.641).
Table 3.
Univariable and multivariable logistic regression analysis for proximal junctional kyphosis
| Variable | Univariable analysis | Multivariable analysis | ||
|---|---|---|---|---|
| OR (95% CI) | P | OR (95% CI) | P | |
| Thoracic compensation | ||||
| Preoperative thoracic compensation, present vs absent | 0.480 (0.305-0.756) | 0.002 | 0.431 (0.261-0.713) | 0.001 |
| Demographic characteristics | ||||
| Age, per year | 1.033 (0.999-1.069) | 0.059 | 1.005 (0.966-1.046) | 0.811 |
| Female sex, vs male | 2.623 (1.169-5.886) | 0.019 | 3.656 (1.501-8.905) | 0.004 |
| BMI (kg/m2), per unit | 1.019 (0.963-1.079) | 0.514 | | |
| BMD (T-score), per unit | 0.841 (0.707-1.000) | 0.050 | 0.863 (0.704-1.057) | 0.154 |
| ASA physical status, per grade | 1.671 (1.015-2.750) | 0.044 | 1.751 (1.003-3.059) | 0.049 |
| Frailty score (mFI-5), per point | 1.047 (0.820-1.338) | 0.711 | | |
| Operative characteristics | ||||
| Combined anterior surgery, yes vs no | 1.385 (0.838-2.289) | 0.205 | | |
| Pedicle subtraction osteotomy, yes vs no | 0.575 (0.259-1.274) | 0.173 | | |
| Pelvic fixation, yes vs no | 1.316 (0.775-2.234) | 0.310 | | |
| Cement augmentation at UIV, yes vs no | 1.057 (0.656-1.704) | 0.819 | | |
| Transverse process hook at UIV, yes vs no | 0.723 (0.433-1.210) | 0.217 | | |
| Number of fused segments (levels), per level | 0.865 (0.738-1.014) | 0.074 | 0.798 (0.668-0.955) | 0.014 |
| Preoperative radiographic parameters | ||||
| PI (°), per degree | 0.994 (0.975-1.014) | 0.574 | | |
| PI-LL (°), per degree | 1.001 (0.990-1.012) | 0.872 | | |
| L1PA (°), per degree | 0.984 (0.961-1.007) | 0.163 | | |
| PT (°), per degree | 1.013 (0.993-1.033) | 0.214 | | |
| TK (°), per degree | 1.008 (0.994-1.021) | 0.264 | | |
| SVA (mm), per mm | 1.000 (0.996-1.003) | 0.820 | | |
| Postoperative radiographic parameters | ||||
| PI (°), per degree | 0.992 (0.972-1.013) | 0.472 | | |
| PI-LL (°), per degree | 0.971 (0.951-0.991) | 0.005 | Not entered * | |
| L1PA (°), per degree | 0.946 (0.914-0.978) | 0.001 | Not entered * | |
| LDI (%), per percentage | 1.006 (0.995-1.017) | 0.269 | * | |
| PT (°), per degree | 0.996 (0.971-1.021) | 0.751 | | |
| TK (°), per degree | 1.052 (1.029-1.075) | <0.001 | Not entered † | |
| SVA (mm), per mm | 1.000 (0.993-1.008) | 0.895 | | |
| Postoperative alignment offsets | ||||
| Age-adjusted PI-LL offset (°), per degree | 0.967 (0.948-0.987) | 0.001 | 0.982 (0.957-1.007) | 0.158 |
| L1PA offset (°), per degree | 0.911 (0.869-0.956) | <0.001 | 0.932 (0.880-0.988) | 0.018 |
Bold P values indicate statistical significance.
Variables with a P value below 0.10 on univariable analysis were considered for the multivariable model.
TC, thoracic compensation; BMI, body mass index; BMD, bone mineral density; ASA, American Society of Anesthesiologists; mFI-5, 5-item modified frailty index; UIV, upper instrumented vertebra; PI, pelvic incidence; LL, lumbar lordosis; L1PA, L1 pelvic angle; LDI, lordosis distribution index; PT, pelvic tilt; TK, thoracic kyphosis; SVA, sagittal vertical axis; Age-adjusted PI-LL offset, difference between the postoperative PI-LL and the age-adjusted PI-LL target; L1PA offset, difference between the measured and the ideal L1 pelvic angle; OR, odds ratio; CI, confidence interval.
*Postoperative PI-LL and postoperative L1PA were also withheld because they overlap with the age-adjusted PI-LL offset and the L1PA offset, respectively.
†Postoperative TK was withheld because it is not modifiable/measurable intraoperatively.
Risk stratification by TC and L1PA offset
Patients were divided into four groups by TC and the L1PA offset; these groups comprised 177, 42, 94, and 30 patients (Figure 4). The incidence of PJK rose stepwise, from 28.2% with TC present and no overcorrection to 42.9% with TC present and overcorrection, 44.7% with TC absent and no overcorrection, and 60.0% with TC absent and overcorrection (P = 0.001). With the first group of these as the reference, the adjusted OR was 1.55 (95% CI 0.73-3.33, P = 0.256) for the second group, 2.29 (95% CI 1.31-4.01, P = 0.004) for the third, and 3.41 (95% CI 1.37-8.46, P = 0.008) for the fourth. The two factors acted independently (interaction P = 0.941).
Figure 4.
Incidence of PJK according to the presence of thoracic compensation and the postoperative L1PA offset. (A) Observed incidence in each of the four groups, with error bars showing 95% confidence intervals. (B) Adjusted odds ratios from a multivariable logistic regression model including age, sex, bone mineral density, American Society of Anesthesiologists grade, number of fused segments, and postoperative age-adjusted PI-LL. The postoperative L1PA offset was not entered as a covariate because it defines the groups. Overcorrection was defined as a postoperative L1PA offset ≤ 0°. The horizontal bars represents 95% confidence intervals. TC, thoracic compensation; PJK, proximal junctional kyphosis; L1PA, L1 pelvic angle; PI-LL, pelvic incidence minus lumbar lordosis; OR, odds ratio; CI, confidence interval
Consistency across subgroups
Eight prespecified subgroups were examined, defined by age, osteoporosis, ASA grade, number of fused segments, preoperative TK and PI-LL, postoperative PI-LL, and the age-adjusted PI-LL offset (Figure 5). Adjusted ORs ranged from 0.22 to 0.64, no stratum favored the TC absent group, and no interaction term approached significance (P = 0.105 to 0.736).
Figure 5.
Forest plot of the adjusted association between thoracic compensation and PJK across prespecified subgroups. Odds ratios were estimated within each stratum by multivariable logistic regression adjusted for age, sex, bone mineral density, American Society of Anesthesiologists grade, number of fused segments, postoperative L1PA offset, and postoperative age-adjusted PI-LL, with the stratifying variable omitted from its own model. The P value for interaction was obtained from a likelihood ratio test of the product term between thoracic compensation and the subgroup indicator. The horizontal bars represents 95% confidence intervals. PJK, proximal junctional kyphosis; L1PA, L1 pelvic angle; PI-LL, PI-LL, pelvic incidence minus lumbar lordosis; ASA, American Society of Anesthesiologists; TK, thoracic kyphosis; OR, odds ratio; CI, confidence interval
Representative cases
Figures 6 and 7 show two patients with a comparable preoperative deformity and an identical postoperative L1PA offset, in whom PJK developed in the TC absent patient but not in the TC present patient.
Figure 6.
Representative case without preoperative thoracic compensation. A 74-year-old woman underwent fusion from T10 to the pelvis. (A) Preoperatively, PI was 66° and LL was 13°, giving a PI-LL mismatch of 53°. The TK (17°) exceeded the expected TK (8.6°) so that thoracic compensation was absent. (B) Postoperatively, LL was corrected to 63°, giving a PI-LL mismatch of 3° and an L1PA offset of 4°. (C) PJK developed six months after surgery. TC, thoracic compensation; TK, thoracic kyphosis; eTK, expected thoracic kyphosis; PI-LL, pelvic incidence minus lumbar lordosis; LL, lumbar lordosis; L1PA, L1 pelvic angle
Figure 7.
Representative case with preoperative thoracic compensation. A 78-year-old woman underwent fusion from T10 to the pelvis. (A) Preoperatively, PI was 58° and LL was 4°, giving a PI-LL mismatch of 54°, and the expected TK was comparable to that of the case in Figure 5. The TK (5°) was lower than the expected TK (8.0°) so that thoracic compensation was present. (B) Postoperatively, LL was corrected to 53°, giving a PI-LL mismatch of 5° and an L1PA offset of 4°, so the magnitude of correction was comparable to the case in Figure 6. (C) PJK did not develop until postoperative two years
Discussion
This study showed that TC assessed against the deformity burden was associated with a lower risk of PJK after lower thoracic fusion for ASD. PJK occurred in 48.4% of patients without TC and 31.1% of those with TC, and the association persisted in multivariable analysis adjusting for patient, surgical, and radiographic covariates with OR of 0.431. It was also consistent across all eight prespecified subgroups.
Research in ASD has concentrated on the lumbar correction, and individualized targets have been proposed to guide it.12-16 Overcorrection relative to either the age-adjusted PI-LL target or the ideal L1PA has repeatedly been linked to PJK.15,16,20,21 Our results are consistent with these reports. A smaller age-adjusted PI-LL offset raised the risk of PJK on univariable analysis, and a smaller L1PA offset did so on both univariable and multivariable analysis (Table 3). Compared with the lumbar correction, the unfused thoracic spine has received far less attention, and preoperative TC status in particular has rarely been examined. The few studies addressing preoperative TK have been inconsistent. Ohba et al. found a higher baseline TK in patients with PJK (32.0° vs. 20.9°), 22 whereas Yagi et al. found no difference (16.8° vs. 21.4°). 23 The absolute TK value itself does not convey how much thoracic flattening the lumbar deformity demanded. Two patients with the same TK may be classified differently once the PI-LL mismatch is considered. Two observations support this. First, although preoperative TK was significantly smaller in the TC present group than in the TC absent group (2.3° vs. 22.6°, Table 2), it did not discriminate PJK development in logistic regression analysis (P = 0.264, Table 3). Second, entering preoperative TK into the model strengthened rather than weakened the association with TC (OR 0.431 to 0.355), indicating that the magnitude of TK had been masking the effect of compensation.
The behavior of the unfused thoracic spine explains the apparent paradox in Figure 3. After LL was restored, the thoracic spine returned toward kyphosis in both groups. However, the increase was significantly larger when TC was present (19.1° vs. 11.0°, P < 0.001), consistent with previous studies reporting that a more flattened thoracic spine has more room to rebound.4,24,25 Protopsaltis et al. observed the same larger reciprocal change in compensated patients and concluded that it conferred a higher risk of PJK, 4 which contradicts our findings. Two explanations seem plausible. The first concerns the study cohort. Our patients were required to have a PI-LL mismatch greater than 20°, and their mean mismatch was 40.4° with a mean preoperative TK of 9.7°, indicating an established lumbar deformity with thoracic compensation already operating. The cohort of Protopsaltis et al. was not restricted in this way and carried a larger baseline TK (mean TK, 26.1°) even with a smaller deformity burden (mean PI-LL mismatch, 27.4°), suggesting a lesser degree of TC compared with our cohort. A given reciprocal change may therefore carry different consequences in these two settings. The second concerns the definition of TC. In that study, the expected TK was derived as PI minus 20°, which follows from the relationship LL = (PI+TK)/2 + 10 with LL treated as equal to PI, and therefore presupposes preserved lordosis and excludes the deformity burden. For a PI of 50°, the expected TK is 30° whether the mismatch is 0° or 40°, so compensation is counted regardless of the lumbar deformity.
PJK is multifactorial.5-11 In the present study, female sex, higher ASA grade, shorter fusion, and a smaller L1PA offset were each independently associated with PJK, and risk is therefore better judged from the combination of factors than from any one of them. In a cross-classification by TC and L1PA overcorrection, when TC was absent the risk was more than doubled even without overcorrection (OR 2.29) and rose more than threefold when overcorrection was also present (OR 3.41). Preoperative TC status may be useful for risk stratification. A patient without TC carried more than twice the adjusted odds of PJK even without overcorrection, and more than three times the odds when overcorrection was also present. Where TC is absent, closer attention to the correction target and to postoperative surveillance appears reasonable. Within the T9 to L1 range, a more cranial UIV may be preferred. Every construct in this study ended at the sacrum or pelvis, so a longer fusion means a higher UIV, and each additional level was associated with lower odds of PJK in our model (OR 0.798 per level). Whether preoperative TC status should guide the choice between an upper and a lower thoracic UIV was not addressed in this study, as all patients had the UIV between T9 and L1, and we offer this as a hypothesis for prospective investigation. 26 More than half of the events occurred within three months, so this decision carries its consequence early.
Several limitations should be acknowledged. First, the retrospective single-center design carries a risk of selection bias, and the UIV was chosen at the surgeon’s discretion rather than by fixed criteria. Second, the reference equations for expected TK were derived in an asymptomatic cohort whose PI-LL mismatch averaged near zero, 18 so their application to a cohort with a mean mismatch of 40.4° requires extrapolation, and TC defined in this way should be regarded as an approximation. Third, paraspinal muscle quality was not assessed although it probably underlies the compensatory capacity. 27 Fourth, the model was built to estimate the association with TC rather than to predict PJK, so its discrimination should not be read as predictive utility. Finally, the cohort was Asian and predominantly female, and external validation is required.
Conclusions
TC assessed against the deformity burden was associated with a lower risk of PJK after lower thoracic fusion for ASD, and the association persisted after adjustment for the correction achieved. Risk was highest when TC was absent and lordosis was overcorrected, so preoperative compensation status warrants consideration alongside the lordosis correction planned.
Footnotes
Funding: The authors received no financial support for the research, authorship, and/or publication of this article.
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
ORCID iDs
Tae Soo Shin https://orcid.org/0000-0003-4020-7690
Jin-Sung Park https://orcid.org/0000-0001-6517-8609
Dong-Ho Kang https://orcid.org/0000-0002-9114-0150
Joon Young Jung https://orcid.org/0000-0002-8561-7522
Se-Jun Park https://orcid.org/0000-0002-2412-9437
Ethical Considerations
The institutional review board approved this study (IRB number: SMC 2026-06-095) and waived the requirement for informed consent due to its retrospective design.
References
- 1.Glassman SD, Bridwell K, Dimar JR, Horton W, Berven S, Schwab F. The impact of positive sagittal balance in adult spinal deformity. Spine (Phila Pa 1976). 2005;30:2024-2029. [DOI] [PubMed] [Google Scholar]
- 2.Barrey C, Roussouly P, Le Huec JC, D'Acunzi G, Perrin G. Compensatory mechanisms contributing to keep the sagittal balance of the spine. Eur Spine J. 2013;22(Suppl 6):S834-S841. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Diebo BG, Ferrero E, Lafage R, et al. Recruitment of compensatory mechanisms in sagittal spinal malalignment is age and regional deformity dependent: a full-standing axis analysis of key radiographical parameters. Spine (Phila Pa 1976). 2015;40:642-649. [DOI] [PubMed] [Google Scholar]
- 4.Protopsaltis TS, Diebo BG, Lafage R, et al. Identifying Thoracic Compensation and Predicting Reciprocal Thoracic Kyphosis and Proximal Junctional Kyphosis in Adult Spinal Deformity Surgery. Spine (Phila Pa 1976). 2018;43:1479-1486. [DOI] [PubMed] [Google Scholar]
- 5.Tanaka T, Katayanagi J, Konuma H, et al. Impact of osteoporosis on perioperative complications in patients undergoing surgical treatment for lumbar spinal stenosis: a nationwide retrospective study. Asian Spine J. 2025;19:794-802. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Singh V, Oppermann M, Evaniew N, et al. Estimation of proximal junctional failure and associated risk factors in adult spine deformity surgery: an observational study from a single institution. Asian Spine J. 2025;19:452-462. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Park JS, Kim HJ, Park SJ, Kang DH, Lee CS. A comprehensive review of risk factors and prevention strategies: how to minimize mechanical complications in corrective surgery for adult spinal deformity. Asian Spine J. 2025;19:463-475. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Echt M, Ranson W, Steinberger J, Yassari R, Cho SK. A Systematic Review of Treatment Strategies for the Prevention of Junctional Complications After Long-Segment Fusions in the Osteoporotic Spine. Global Spine J. 2021;11:792-801. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Zou L, Liu J, Lu H. Characteristics and risk factors for proximal junctional kyphosis in adult spinal deformity after correction surgery: a systematic review and meta-analysis. Neurosurg Rev. 2019;42:671-682. [DOI] [PubMed] [Google Scholar]
- 10.Park SJ, Park JS, Kang DH, Kang M, Jung K, Lee CS. Different Risk Factors Between Bony and Ligamentous Proximal Junctional Failure in Patients Undergoing Thoracolumbar Fusion to Pelvis for Adult Spinal Deformity. Spine (Phila Pa 1976). 2025;50:1065-1073. [DOI] [PubMed] [Google Scholar]
- 11.Park SJ, Park JS, Kang DH, et al. Different characteristics between acute and delayed proximal junctional failure in elderly patients undergoing corrective surgery for adult spinal deformity: comparative analysis of risk factor, failure mode, and clinical consequences. Spine J. 2024;24:2377-2388. [DOI] [PubMed] [Google Scholar]
- 12.Park SJ, Kim HJ, Park JS, Kang DH, Lee CS. Sagittal alignment goals in adult spinal deformity surgery: a narrative review focusing on proximal junctional complications and clinical outcomes. Asian Spine J. 2026. doi: 10.31616/asj.2025.0661. E-pub ahead of print. [DOI] [PubMed] [Google Scholar]
- 13.Lafage R, Smith JS, Elysee J, et al. Sagittal age-adjusted score (SAAS) for adult spinal deformity (ASD) more effectively predicts surgical outcomes and proximal junctional kyphosis than previous classifications. Spine Deform. 2022;10:121-131. [DOI] [PubMed] [Google Scholar]
- 14.Yilgor C, Sogunmez N, Boissiere L, et al. Global Alignment and Proportion (GAP) Score: Development and Validation of a New Method of Analyzing Spinopelvic Alignment to Predict Mechanical Complications After Adult Spinal Deformity Surgery. J Bone Joint Surg Am. 2017;99:1661-1672. [DOI] [PubMed] [Google Scholar]
- 15.Hills J, Mundis GM, Klineberg EO, et al. The T4-L1-Hip Axis: Sagittal Spinal Realignment Targets in Long-Construct Adult Spinal Deformity Surgery: Early Impact. J Bone Joint Surg Am. 2024;106:e48. [DOI] [PubMed] [Google Scholar]
- 16.Park SJ, Park JS, Kang DH, Lee CS, Kim HJ. Establishing Optimal L1 Pelvic Angle Targets to Minimize Both Proximal Junctional Kyphosis and Pelvic Nonresponse in Adult Spinal Deformity Surgery. Neurospine. 2025;22:987-997. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Lafage R, Schwab F, Glassman S, et al. Age-Adjusted Alignment Goals Have the Potential to Reduce PJK. Spine (Phila Pa 1976). 2017;42:1275-1282. [DOI] [PubMed] [Google Scholar]
- 18.Li X, Wang D, Chen X, et al. Thoracic and pelvic compensation during lumbar degeneration: a retrospective study of 625 asymptomatic adults. Eur Spine J. 2025;34:4475-4485. [DOI] [PubMed] [Google Scholar]
- 19.Bridwell KH, Lenke LG, Cho SK, et al. Proximal junctional kyphosis in primary adult deformity surgery: evaluation of 20 degrees as a critical angle. Neurosurgery. 2013;72:899-906. [DOI] [PubMed] [Google Scholar]
- 20.Byun CW, Cho JH, Lee CS, Lee DH, Hwang CJ. Effect of overcorrection on proximal junctional kyphosis in adult spinal deformity: analysis by age-adjusted ideal sagittal alignment. Spine J. 2022;22:635-645. [DOI] [PubMed] [Google Scholar]
- 21.Park SJ, Park JS, Kang DH, Jung K, Kang M, Lee CS. Association of age-adjusted pelvic incidence minus lumbar lordosis correction with long-term radiographic and clinical outcomes in adult spinal deformity surgery. J Neurosurg Spine. 2025;43:396-404. [DOI] [PubMed] [Google Scholar]
- 22.Ohba T, Koji F, Koyama K, et al. Preoperative Radiographic Evaluation of Thoracic Flexibility and Compensation for Adult Spinal Deformity Surgery. How to Select Optimal Upper Instrumented Vertebra to Prevent Proximal Junctional Kyphosis. Spine (Phila Pa 1976). 2022;47:144-152. [DOI] [PubMed] [Google Scholar]
- 23.Yagi M, Hosogane N, Okada E, et al. Factors affecting the postoperative progression of thoracic kyphosis in surgically treated adult patients with lumbar degenerative scoliosis. Spine (Phila Pa 1976). 2014;39:E521-E528. [DOI] [PubMed] [Google Scholar]
- 24.Shimizu T, Lehman RA, Jr., Sielatycki JA, et al. Reciprocal change of sagittal profile in unfused spinal segments and lower extremities after complex adult spinal deformity surgery including spinopelvic fixation: a full-body X-ray analysis. Spine J. 2020;20:380-390. [DOI] [PubMed] [Google Scholar]
- 25.Jang JS, Lee SH, Min JH, Maeng DH. Influence of lumbar lordosis restoration on thoracic curve and sagittal position in lumbar degenerative kyphosis patients. Spine (Phila Pa 1976). 2009;34:280-284. [DOI] [PubMed] [Google Scholar]
- 26.Kumar RP, Adida S, Lavadi RS, et al. A guide to selecting upper thoracic versus lower thoracic uppermost instrumented vertebra in adult spinal deformity correction. Eur Spine J. 2024;33:2742-2750. [DOI] [PubMed] [Google Scholar]
- 27.Wang G, Li Y, Zhang C, Liu P, Sun J. Erector Spinae Atrophy Correlates with Global Sagittal Imbalance and Postoperative Proximal Junctional Kyphosis Incidence in Lumbar Degenerative Kyphosis. Asian Spine J. 2024;18:50-57. [DOI] [PMC free article] [PubMed] [Google Scholar]