Determinants of Thoracic Reciprocal Change and Its Relationship to Proximal Junctional Kyphosis Following Lower Thoracic Fusion for Lumbar Flatback Deformity Authors: Tae Soo Shin, Jin-Sung Park, Dong-Ho Kang, Chong-Suh Lee, Se-Jun Park Published: 2026 ABSTRACT BackgroundLumbar flatback deformity (LFD) is characterized by the loss of lumbar lordosis (LL) and global sagittal malalignment, often necessitating thoracolumbar corrective surgery. Lumbar correction induces adaptive changes in unfused thoracic segments, known as thoracic reciprocal change (RC); however, its independent clinical significance remains debatable. This study aims to identify factors associated with thoracic RC after lower thoracic fusion for LFD and to evaluate its independent effect on the development of proximal junctional kyphosis (PJK).MethodsThis retrospective study included 190 patients who underwent lower thoracic fusion (upper instrumented vertebra at T9 or T10) to the sacrum or pelvis for LFD, with a 2-year follow-up. Sagittal parameters were evaluated preoperatively at 6 weeks postoperatively, and at 2 years. Thoracic RC was defined as the change in thoracic kyphosis (TK) from the preoperative baseline to 6 weeks postoperatively. Thoracic flexibility was assessed using the standing-supine TK difference. Multivariate linear regression was used to identify factors influencing thoracic RC, and multivariate logistic regression was performed to determine independent risk factors for PJK.ResultsMean TK increased from 5.7° (supine) preoperatively to 26.0° at 6 weeks, resulting in a mean thoracic RC of 20.3°. Thoracic RC was significantly associated with lower preoperative TK, greater thoracic flexibility, greater LL correction, and lower postoperative pelvic incidence (PI) minus LL (PI–LL) mismatch (adjusted R2 = 0.647). PJK occurred in 47 patients (24.7%). However, the magnitude of thoracic RC did not differ between patients with and without PJK (20.9° vs. 20.0°, p = 0.678). Higher preoperative TK and lower postoperative PI–LL mismatch were independently associated with PJK, whereas thoracic RC was not.ConclusionsThoracic RC after lower thoracic fusion is primarily determined by preoperative thoracic characteristics and the magnitude of lordosis correction. Although associated with an early increase in the postoperative junctional angle, thoracic RC itself does not independently predict the development of PJK. These findings suggest that thoracic RC appears to behave as a compensatory adaptation to sagittal realignment rather than an independent driver of PJK. Lumbar flatback deformity (LFD) represents a significant clinical challenge characterized by loss of lumbar lordosis (LL) and resultant global sagittal malalignment, often necessitating surgical correction to restore LL and sagittal alignment.1) Such correction inevitably induces adaptive changes in unfused thoracic segments, commonly referred to as thoracic reciprocal change (RC).23) Thoracic RC, typically manifesting as increased thoracic kyphosis (TK), serves as a compensatory response to restored LL and is an important part of the global sagittal realignment process. The clinical significance of thoracic RC extends beyond radiographic observations; excessive reciprocal increases in TK can undermine long-term surgical goals by increasing proximal junctional stress or reducing overall sagittal balance.45) Therefore, assessing thoracic RC and its potential clinical impact is crucial for understanding postoperative alignment behavior and mechanical complications following corrective surgery for LFD. Proximal junctional kyphosis (PJK) remains one of the most common mechanical complications following multilevel fusion surgery.67) Although previous research has suggested that greater thoracic RC predisposes patients to junctional failure,248) this association may be confounded by multiple interrelated factors that independently influence both RC magnitude and PJK risk. Specifically, factors known to affect thoracic RC—such as preoperative TK profiles and the magnitude of lordosis correction—have also been independently associated with PJK development. This overlap complicates the interpretation of whether thoracic RC is a pathological driver for PJK or merely reflects the natural realignment process following lumbar correction. To clarify the clinical significance of thoracic RC, it is essential to evaluate RC in conjunction with its contributing factors rather than as a single isolated parameter. Therefore, the purposes of this study were to identify the factors associated with thoracic RC following lower thoracic fusion for LFD and to evaluate whether thoracic RC independently contributes to the development of PJK after adjusting for key confounding sagittal alignment factors. METHODS A retrospective cohort study was conducted using a prospectively collected adult spinal deformity database from a tertiary referral center. The Institutional Review Board of Samsung Medical Center approved this study (IRB No. SMC 2024-07-144) and waived the requirement for informed consent due to its retrospective nature. Study Population The study included consecutive patients who underwent corrective surgery for symptomatic LFD between 2015 and 2022. Inclusion criteria were as follows: (1) LFD defined as a pelvic incidence (PI) minus LL (PI–LL) greater than 20°; (2) uppermost instrumented vertebra (UIV) at T9 or T10, extending to the sacrum or pelvis; and (3) completion of a 2-year follow-up. To avoid UIV-related bias, only those with a UIV in the lower thoracic spine at T9 or T10 were included. Pelvic fixation using conventional iliac screws was routinely performed for patients undergoing long fusion involving the sacrum. However, pelvic fixation was waived for patients with milder deformities who had preexisting lumbosacral solid fusion, either due to prior surgery or the presence of a lumbosacral transitional vertebra (e.g., L5 sacral assimilation). All surgeries were performed by 1 of 3 board-certified spine surgeons, with preferred lateral access surgery throughout the study period. Exclusion criteria included (1) non-degenerative conditions (e.g., trauma, infection, or tumor); (2) primarily coronal deformities without significant sagittal malalignment (e.g., adult idiopathic scoliosis); (3) patients with diffuse idiopathic skeletal hyperostosis involving the thoracic spine; and (4) revision surgeries performed for reasons other than junctional complications (e.g., pseudarthrosis or surgical site infection). Data Collection Demographic variables included age, sex, American Society of Anesthesiologists (ASA) performance status grade, body mass index (BMI), bone mineral density (BMD), and history of prior lumbar fusion. Surgical parameters included pelvic fixation and UIV levels (T9 and T10). Radiographic parameters were assessed preoperatively, at 6 weeks postoperatively, and at the final follow-up (i.e., 2 years postoperatively). These parameters included PI–LL mismatch, sacral slope (SS), pelvic tilt (PT), TK, T1 pelvic angle (T1PA), sagittal vertical axis (SVA), and proximal junctional angle (PJA). When early PJK developed prior to the 6-week postoperative evaluation, radiographic parameters were measured on images acquired before discharge (approximately 2 weeks postoperatively). None of these patients demonstrated PJK on before-discharge radiographs, confirming that RC was measured prior to PJK development in all cases. For patients who underwent revision surgery due to PJK, radiographic data obtained at the time of revision were considered as the final follow-up. TK was measured as Cobb angle between T4 and T12. Additionally, supine TK was evaluated preoperatively using sagittal scout images from magnetic resonance imaging. To assess the preoperative thoracic flexibility, the difference in TK between the standing and supine positions (supine ΔTK) was evaluated, with higher values indicating greater thoracic flexibility.5) Thoracic RC was defined as the difference in TK between the preoperative and 6-week postoperative measurements as described in a previous study.2) The PJA was measured as the angle between the caudal endplate of the UIV and the cranial endplate of the 2 supra-adjacent vertebrae.9) PJK was defined as a PJA ≥ 20° according to previously established criteria.10) In addition to the occurrence of PJK, revision surgery for PJK was also evaluated. Statistical Analysis Variables are presented as means and standard deviations for continuous variables, and as frequencies and percentages for categorical variables. Paired t-tests were used to compare preoperative and 6-week postoperative sagittal parameters. Multiple linear regression analysis was conducted to identify factors affecting thoracic RC. Based on previous literature, clinically relevant factors known to influence thoracic RC, such as preoperative TK, preoperative supine ΔTK, change in LL, and postoperative PI–LL, were incorporated into the analysis.234511) Subsequently, the influence of thoracic RC on PJK development was evaluated. Given that the variables included in the linear regression model are also known to affect PJK development, these parameters, along with thoracic RC, were included in both the univariate analysis and multivariate logistic regression analysis to identify independent risk factors for PJK. Multicollinearity among covariates in the multivariate logistic model was assessed using the variance inflation factor. To investigate the characteristics of patients with low and high thoracic RC, patients were stratified into the 2 groups based on the median change in TK (median, 20.5°). Demographic and surgical data, sagittal parameters (preoperatively, at 6 weeks, and at 2 years), and rates of PJK and revision surgery were compared between the 2 RC groups using independent t-tests or chi-square tests. Repeated measures analysis of variance was conducted to analyze time-dependent changes in TK and PJA across the different time points (preoperatively, at 6 weeks, and at 2 years), comparing results by RC group and PJK occurrence. Finally, changes in PJA between consecutive time points (preoperative–6 weeks, and 6 weeks–2 years) were compared using independent t-tests. To assess the potential impact of radiograph substitution in patients with early PJK, a sensitivity analysis was performed excluding the 14 patients whose 6-week radiographic parameters were measured before discharge. All statistical analyses were performed by professional statisticians using Python version 3.10 (Python Software Foundation). A p-value of < 0.05 was considered statistically significant. RESULTS Study Population A total of 219 patients who underwent lower thoracic fusion (T9 or T10) to the sacrum or pelvis for LFD were initially screened. After excluding patients with insufficient radiographic data (n = 8), diffuse idiopathic skeletal hyperostosis involving the thoracic spine (n = 3), revision surgery for reasons unrelated to proximal junctional pathology (n = 10), and those with less than 2 years of follow-up (n = 8), 190 patients were included in the final analysis (Fig. 1). The mean age at surgery was 70.3 ± 6.1 years, and 172 patients (90.5%) were female (Table 1). Pelvic fixation was performed in 172 patients (90.5%). The UIV was located at T9 in 56 patients (29.5%) and T10 in 134 patients (70.5%). Postoperative Changes in Sagittal Parameters Significant improvements in sagittal alignment were observed following surgery (Table 1). The mean PI–LL mismatch improved from 46.6° ± 19.4° preoperatively to 6.1° ± 11.7° at 6 weeks postoperatively. TK increased significantly from 5.7° ± 17.2° preoperatively to 26.0° ± 11.7° at 6 weeks. Consequently, the mean magnitude of thoracic RC was 20.3°. A postoperative increase in TK was observed in 181 patients (95.2%), while 9 patients (4.7%) exhibited no change or a slight decrease. Factors Associated with Thoracic RC Multiple linear regression analysis identified 4 factors significantly associated with thoracic RC (adjusted R2 = 0.647) (Table 2). Lower preoperative TK (β = -0.24, p = 0.001) and greater thoracic flexibility, as measured by preoperative supine ΔTK (β = 0.62, p < 0.001), were associated with greater thoracic RC. Additionally, larger LL correction (β = 0.16, p < 0.001) and lower postoperative PI–LL mismatch (β = –0.17, p = 0.004) significantly influenced the magnitude of thoracic RC, suggesting that patients with greater lordosis correction exhibited greater thoracic compensatory changes. Risk Factors for PJK During the follow-up period, PJK developed in 47 patients (24.7%). Patients who developed PJK had significantly higher preoperative TK values (14.5° vs. 9.4°, p = 0.001), lower preoperative supine ΔTK (1.9° vs. 5.9°, p = 0.040), and lower postoperative PI–LL mismatch (2.3° vs. 7.4°, p = 0.009) compared to those without PJK (Table 3, Fig. 2C). In contrast, the magnitude of thoracic RC, defined as the change in TK, was not significantly different between the 2 groups (20.9° vs. 20.0°, p = 0.678). In multivariate logistic regression analysis, higher preoperative TK (odds ratio [OR], 1.05; 95% confidence interval [CI], 1.01–1.09; p = 0.020) and lower postoperative PI–LL (OR, 0.97; 95% CI, 0.94–0.99; p = 0.030) were independently associated with the development of PJK. Comparison between Low and High RC Groups Based on the median thoracic RC value of 20.5°, patients were divided into a low RC group (n = 94) and a high RC group (n = 96). Demographic and surgical variables were comparable between the groups, except for sex, which showed a higher female predominance in the high RC group (Table 4). The high RC group had a significantly lower preoperative TK (7.0° vs. 14.4°, p < 0.001) and exhibited a more severe preoperative sagittal deformity, as indicated by greater PI–LL mismatch, higher PT, higher T1PA, and greater SVA. The magnitude of sagittal correction was generally greater in the high RC group compared to the low RC group. Notably, the magnitude of LL correction (i.e., change in LL) was significantly greater in the high RC group (50.9° vs. 29.9°, p < 0.001). Postoperatively, the high RC group demonstrated better alignment compared to the low RC group in terms of PI–LL (4.4° vs. 8.0°, p = 0.033), T1PA (13.5° vs. 16.5°, p = 0.022), and SVA (13.0 mm vs. 25.3 mm, p = 0.010) at 6 weeks, as well as SVA (32.3 mm vs. 44.2 mm, p = 0.034) at 2 years. However, the differences in alignment status between the groups were generally smaller compared to those in preoperative sagittal parameters. Specifically, TK at both 6 weeks (26.1° vs. 25.9°, p = 0.929) and 2 years (33.9° vs. 34.7°, p = 0.704) was not significantly different between the groups (Fig. 2A, Table 4). The incidence of PJK was comparable between the 2 groups (23.4% vs. 26.0%, p = 0.674), as were revision surgery rates (9.6% vs. 11.5%, p = 0.645). Changes in PJA over Time The temporal behavior of the PJA varied according to thoracic RC and PJK status (Fig. 2B and D). From preoperatively to 6 weeks postoperatively, the increase in PJA was significantly greater in the high RC group than in the low RC group (9.4° vs. 5.4°, p < 0.001) (Table 5). However, this early increase was not significantly different between patients who developed PJK and those who did not (7.9° vs. 7.3°, p = 0.490). From 6 weeks to 2 years postoperatively, PJA progression was similar between the low and high RC groups (p = 0.346) but was markedly greater in the PJK group compared to the non-PJK group (21.4° vs. 4.9°, p < 0.001). A representative case illustrating substantial thoracic RC without subsequent PJK development is shown in Fig. 3. Sensitivity Analysis We performed a sensitivity analysis excluding the 14 patients with before-discharge radiograph substitution (n = 176). In the linear regression model for RC determinants, all 4 predictors retained statistical significance with virtually identical effect sizes: preoperative TK (β = -0.19, p = 0.015), preoperative supine ΔTK (β = 0.60, p < 0.001), change in LL (β = 0.18, p < 0.001), and postoperative PI–LL mismatch (β = -0.17, p = 0.003). The adjusted R2 was 0.655, comparable to that of the primary analysis (0.647). In the logistic regression model for PJK predictors, preoperative TK (OR, 1.05; 95% CI, 1.00–1.09; p = 0.030) and postoperative PI–LL mismatch (OR, 0.97; 95% CI, 0.93–1.00; p = 0.034) remained independent predictors of PJK, whereas thoracic RC remained non-significant, consistent with the primary analysis. DISCUSSION Surgical correction of LFD has traditionally focused on adequately restoring LL, with extensive research establishing alignment targets and predictive models for surgical outcomes.12131415) Compared to the issue regarding lordosis correction, postoperative changes occurring in unfused thoracic segments have received comparatively less attention in the literature.811) This relative scarcity of research on thoracic segment behavior is noteworthy, given that thoracic spine alignment plays a crucial role in maintaining global sagittal balance. The thoracic spine is a key component of the compensatory chain that supports upright posture, and alterations in TK directly influence global alignment status.1617) Therefore, understanding behavioral changes occurring within unfused thoracic segments has important clinical implications for predicting postoperative alignment evolution and junctional complications. The present study investigated factors influencing thoracic RC following lower thoracic fusion for LFD and evaluated whether thoracic RC independently contributes to the development of PJK. Our principal findings demonstrate that thoracic RC is primarily determined by preoperative thoracic characteristics, such as preoperative TK and thoracic flexibility, as well as lordosis correction profiles, including the magnitude of LL correction and postoperative PI–LL (Table 2). Importantly, after controlling for these confounding variables, the magnitude of thoracic RC itself did not independently predict PJK development. Instead, greater preoperative TK and lower postoperative PI–LL emerged as independent risk factors for PJK (Table 3). Additionally, a comparison between the low and high RC groups revealed that despite substantial differences in preoperative sagittal deformity severity and sagittal correction magnitude, the postoperative TK at both 6 weeks and 2 years postoperatively converged to similar values between groups. These findings collectively suggest that thoracic RC appears to behave as a compensatory adaptation towards normalized global sagittal alignment following lordosis correction, rather than an independent driver of junctional failure. Our finding that lower preoperative TK predicts greater thoracic RC aligns with previous observations, which reported that patients with lower baseline TK demonstrated greater postoperative increases in TK.211) This relationship likely reflects the greater reserve for adaptive kyphotic change in relatively straight thoracic spines compared to those that are already more kyphotic preoperatively. Additionally, we identified preoperative thoracic flexibility, measured as the standing-supine TK difference, as the strongest predictor of RC (β = 0.62). This finding also corroborates recent evidence demonstrating that patients with more flexible thoracic segments possess a greater biomechanical capacity to undergo compensatory kyphotic change in response to LL restoration.518) The magnitude of LL correction in terms of the change in LL and postoperative PI–LL also significantly influenced thoracic RC in our cohort, consistent with prior studies showing that greater lordosis restoration induces proportionally larger thoracic compensatory changes.2311) Jang et al.3) studied 53 patients with LFD and reported that the magnitude of postoperative LL is significantly correlated with TK increase (r = 0.620). This relationship reflects the biomechanical principle that substantial changes in 1 spinal region necessitate compensatory adjustments in adjacent unfused segments to maintain global sagittal balance. While a greater magnitude of thoracic RC may be expected to increase the risk of PJK,25) our findings do not support this assumption. In the present study, the magnitude of TK increase was comparable between patients with and without PJK, and thoracic RC was not identified as an independent risk factor in multivariate analysis. This apparent discrepancy can be explained by the confounding influence of preoperative thoracic alignment and lordosis correction profiles. In the high RC group, the postoperative PI–LL mismatch was significantly smaller than in the low RC group, a condition that could potentially increase the risk of PJK (Table 4). However, preoperative TK was also lower in the high RC group, which may have mitigated the risk. Consequently, the opposing effects of these 2 factors may have counterbalanced each other, leading to similar PJK rates between the groups. This relationship is further supported by the observed changes in the PJA, which may be more directly associated with the development of PJK (Table 5, Fig. 2B and D). Although patients with high RC demonstrated a significantly greater early postoperative increase in PJA, this early junctional angle change did not differentiate those who subsequently developed PJK from those who did not. In contrast, progressive junctional kyphosis occurring between 6 weeks and 2 years postoperatively was strongly associated with PJK development, regardless of the magnitude of initial RC. This temporal dissociation suggests that the early junctional angle increase related to thoracic RC represents an acute biomechanical adaptation to sagittal realignment, whereas late progressive junctional kyphosis reflects a distinct pathologic process underlying junctional failure. In our cohort, the PJK group demonstrated a lower mean postoperative PI–LL (2.3° vs. 7.4°), which may reflect relative overcorrection in the PJK group. Prior studies have shown that overcorrection of LL relative to PI–LL targets significantly increases PJK risk and that PJK may itself develop as a compensatory response to sagittal overcorrection.141519) Biomechanically, excessive lordosis restoration shifts the UIV posteriorly relative to the center of gravity, concentrating bending and shear forces at the proximal junction. RC represents a multi-segmental adaptive response to global alignment changes, whereas the mechanical failure at the UIV is a focal segment-level phenomenon driven by the magnitude of correction. This explains why, in our cohort, high thoracic RC did not independently predict PJK despite successfully normalizing TK, whereas low postoperative PI–LL did. This study provides important clinical insights into the interpretation of thoracic RC following long-segment fusion for LFD. The absence of an independent association between RC magnitude and PJK incidence challenges the assumption that large RC directly predisposes patients to junctional failure. Surgeons should recognize RC as a physiological adaptive response rather than a risk factor. Greater emphasis should be placed on understanding the underlying sagittal alignment profile—particularly preoperative TK and proximal junctional biomechanics—rather than on the absolute magnitude of postoperative thoracic change. Accordingly, risk assessment and preventive strategies for PJK should focus on appropriate lordosis correction rather than attempts to modulate thoracic RC itself. This study has several limitations that should be acknowledged. First, its retrospective design may have introduced selection bias, despite the use of a prospectively collected database. Second, although a relatively large cohort was analyzed, our cohort consisted predominantly of elderly female patients with LFD undergoing lower thoracic fusion (UIV at T9 or T10), and the results may not apply to younger patients, male patients, or those with different deformity patterns or fusion levels. Despite this limitation, the use of strict inclusion criteria based on diagnosis and UIV level likely reduced bias associated with heterogeneity in deformity type and fusion length. Third, thoracic RC was evaluated based on radiographic parameters obtained at 6 weeks postoperatively; thus, alignment changes occurring beyond this early postoperative period may not have been fully captured. Longitudinal assessment of thoracic alignment evolution over extended follow-up periods may provide additional insights.19) Fourth, although multiple alignment-related variables were incorporated into the multivariate analyses, several measured variables—including age, BMD, BMI, prior fusion status, and UIV level—were not entered into the multivariate PJK model, as the model was designed to focus on radiographic sagittal parameters relevant to both RC and PJK. Additionally, genuinely unassessed factors—such as soft-tissue integrity, paraspinal muscle quality, and surgical technique-related variables—may also influence the development of PJK.72021) However, the primary objective of this study was not to comprehensively analyze risk factors for PJK, but rather to evaluate the effect of thoracic RC on PJK development; therefore, our analyses focused on radiographic parameters known to influence both phenomena. Finally, clinical outcomes were not directly correlated with thoracic RC or PJK occurrence in this study, and future investigations incorporating patient-reported outcomes may provide additional insight into the clinical relevance of these radiographic findings. Thoracic RC after lower thoracic fusion for LFD was primarily influenced by preoperative thoracic characteristics, thoracic flexibility, and the magnitude of LL correction. Although greater thoracic RC was associated with early postoperative increases in PJA, it did not independently predict the development of PJK. These findings suggest that thoracic RC appears to behave as a compensatory adaptation to sagittal realignment rather than an independent driver of PJK.