When Cervical Deformity Correction Fails — and Why: A Longitudinal Map of Alignment Durability After Surgery
An ISSG multicenter survival analysis finds that fewer than one-third of cervical deformity patients maintain acceptable alignment at one year — and identifies a predictable, time-dependent sequence of failure mechanisms that has direct implications for surgical planning and postoperative surveillance.
OPENING SUMMARY
Cervical deformity surgery can substantially improve neck pain, disability, and neurological function. But surgical success at the time of the operation and durable success at one year are not the same thing — and the gap between them, in cervical deformity correction, is larger than the field has fully reckoned with. This study, drawing on the ISSG's prospective multicenter cervical deformity database, applies rigorous longitudinal statistical methods to characterize how alignment durability unfolds across the first postoperative year. What it finds is both clinically important and methodologically significant: fewer than one-third of patients with complete follow-up maintained acceptable alignment at one year, failure mechanisms shifted in a predictable temporal sequence, and radiographic deterioration — while real and meaningful — did not translate into significant differences in patient-reported outcomes at the one-year horizon. Together, these findings reframe radiographic alignment loss as a surveillance signal rather than a clinical verdict, and argue for individualized risk stratification before surgery and structured longitudinal monitoring afterward.
STUDY SNAPSHOT
Study type – Retrospective cohort study using a prospectively maintained multicenter ISSG database
Total operative patients – 155
Patients with complete 4-point radiographic follow-up – 89 (57.4%)
Alignment success definition – Cervical sagittal vertical axis (cSVA) <4 cm
Follow-up timepoints – Immediate postoperative, 3 months, 6 months, 1 year
Trajectory groups – A: Maintained correction (34%), B: Corrected then deteriorated (19%), C: Never aligned (47%)
Statistical methods – Cox proportional hazards regression, Kaplan–Meier with RMST, mixed-effects logistic regression
Cox model sample – 47 initially aligned patients; 17 alignment loss events
Mixed-effects model – 356 observations across 89 patients; ICC = 0.36
Deformity subtypes – Focal cervical kyphosis, cervical flatback, cervicothoracic deformity
Patient-reported outcomes – NDI, EQ-5D, mJOA
Published in – Global Spine Journal, 2026, Vol. 0(0), 1–13 (online ahead of print)
DOI – 10.1177/21925682261490470
Corresponding author – Alexander Kucherina, MS — NYU Langone Health, New York
Open access – Yes — Creative Commons CC BY-NC-ND 4.0
WHY THIS MATTERS
Cervical deformity — abnormal alignment of the neck — is a complex condition causing pain, functional impairment, and neurological compromise. Surgical correction restores cervical sagittal balance and produces meaningful improvements in validated patient-reported outcomes, with established gains in neck disability and quality of life at one year. But the durability of that correction has been incompletely characterized, and prior studies have often relied on repeated cross-sectional comparisons that treat each postoperative timepoint as statistically independent — an approach that violates the repeated-measures structure of longitudinal data and can inflate false-positive findings.
This study addresses both the durability gap and the methodological limitation simultaneously. By applying Cox regression, Kaplan–Meier survival analysis, and mixed-effects logistic regression to the same dataset — three complementary frameworks that each capture different aspects of longitudinal alignment behavior — it provides the most rigorous characterization to date of when and how cervical deformity correction fails.
The clinical stakes are high. Junctional failures — both proximal (PJK) and distal (DJK) — occur in 15–35% of cervical deformity patients. Postoperative cervical malalignment correlates with inferior health-related quality of life. And revision surgery, when it becomes necessary, carries compounding risks. Understanding the mechanisms, timing, and predictors of alignment failure is essential for improving outcomes — both through better surgical planning before the operation and through better surveillance after it.
THE THREE ALIGNMENT TRAJECTORIES
The study defined three patient groups based on alignment trajectory across the four radiographic follow-up timepoints:
Group A — Maintained Correction (34%, n=30)
These patients achieved cSVA <4 cm and maintained it through one year. Their radiographic parameters improved substantially from baseline: cSVA decreased by a mean of 17.8 mm, C2 slope decreased by 15.7°, and T1 slope decreased by 9.8° over the follow-up year. This group had the lowest DJK rates (3.3%) and the best maintained alignment at every timepoint.
Group B — Corrected Then Deteriorated (19%, n=17)
These patients achieved acceptable alignment after surgery but subsequently lost it. They had the smallest baseline cSVA of any group (40.5 mm) — a finding the authors interpret as potentially reflecting relative undercorrection, where patients corrected close to the 40 mm threshold had little margin before crossing back into failure. They worsened across all cervical radiographic parameters over the follow-up year, with a mean cSVA increase of 12.1 mm. DJK rates were strikingly elevated (47.1%). Within Group B, 4 patients lost correction by 3 months, 6 more by 6 months, and 7 more by one year — alignment loss was not sudden but a progressive attrition.
Group C — Never Aligned (47%, n=42)
Nearly half of patients never achieved cSVA <4 cm at any postoperative timepoint. This group was significantly older than Groups A and B (65.9 vs. 61.8 vs. 58.8 years; p=0.038) and had the largest baseline cSVA (60.1 mm). Their aggregate cSVA changed minimally over the year (+1.7 mm on average), but the authors note this stability likely conceals a subset with progressive worsening — a heterogeneous group that may contain both stably undercorrected patients and those on a declining trajectory that longer follow-up would reveal.
KEY FINDINGS
1. Fewer than one-third of patients maintained acceptable alignment at one year.
Among the 89 patients with complete four-point radiographic follow-up, the proportion well-aligned fell from 48.3% at 3 months to 33.7% at one year. The authors carefully contextualize this figure: because complete follow-up patients are disproportionately those with symptoms or suspected complications — who are most likely to be re-imaged — this figure likely represents a conservative lower bound. The 64.5% well-aligned rate in the full 155-patient cohort (which includes patients seen only at some timepoints) likely represents an optimistic upper bound. The true population rate lies between them, and both figures should be read as bounds rather than point estimates.
2. Distal junctional kyphosis was the most consistent predictor of alignment failure.
DJK was the single most powerful predictor of alignment loss across all three statistical frameworks:
Cox regression: HR 5.21 (95% CI 1.94–13.98; p=0.001)
Kaplan–Meier RMST: 4.1 months maintained alignment with DJK versus 9.8 months without (difference 5.7 months; p=0.001)
Mixed-effects logistic regression: OR 0.09 (95% CI 0.04–0.22; p<0.001)
Patients who developed DJK maintained correction for less than half the observation period before crossing back into malalignment. The breadth of the Cox confidence interval (spanning approximately twofold to fourteenfold hazard) reflects the limited number of events and should temper inference about the precise effect magnitude — but the direction and statistical significance were consistent across all models.
3. Lower postoperative cSVA was paradoxically associated with higher failure risk.
One of the study's more counterintuitive findings is that lower postoperative cSVA — meaning better immediate correction — was associated with higher subsequent failure hazard (HR 0.58 per 10 mm increase; p=0.002). The authors explicitly caution against overinterpreting this finding: the model contained only 17 events, and neither correction magnitude nor deformity-specific alignment targets were directly assessed. The most plausible interpretation is that patients corrected only modestly end surgery close to the 40 mm failure threshold, leaving little margin for the settling and junctional changes that occur as fusion consolidates — so that even small anatomic changes are sufficient to cross back into malalignment. This remains a hypothesis rather than a confirmed finding, and the authors present it as such.
4. Older patients deteriorated faster.
The mixed-effects model identified a significant Time × Age interaction (OR 0.88; p=0.022), indicating that the negative association between age and alignment probability strengthened at each successive postoperative visit. Older patients did not simply start worse — they declined faster. Group C, the never-aligned group, was significantly older than both Groups A and B (p=0.038). This finding argues for age-stratified surveillance protocols with more frequent monitoring of older patients during the first postoperative year.
5. Failure mechanisms shifted in a predictable temporal sequence.
Among the 17 Group B patients who lost correction, the mechanism responsible for that loss shifted systematically over time:
At 3 months: DJK was the dominant mechanism (50% of failures at that visit)
At 6 months: PJK became dominant (75% of failures at that visit)
At 1 year: Loss of global sagittal compensation was most common (71% of failures at that visit)
Across all Group B failures combined, global compensation loss was most common overall (47%), followed by PJK (33%) and DJK (20%). The authors interpret this temporal pattern as potentially reflecting the biomechanical progression of the fusion construct: distal junctional stress is greatest before fusion consolidation; as distal fixation matures, the proximal junction becomes the vulnerable zone; and late failures reflect exhaustion of spinopelvic and thoracolumbar compensatory reserve. This remains a hypothesis given the small number of events, but the pattern is mechanistically coherent. Deformity subtype also influenced failure pattern: focal kyphosis patients contributed disproportionately to late global compensation failures, while cervicothoracic patients contributed disproportionately to DJK events.
6. Radiographic failure was not accompanied by significant patient-reported outcome differences at one year.
Change in NDI and EQ-5D from baseline to one year did not differ significantly between trajectory groups. Only mJOA at 6 months separated the groups, with Group B showing greater worsening. The absence of patient-reported outcome differences does not mean alignment deterioration is clinically inconsequential — the study was not powered to detect HRQL differences across three groups, the one-year horizon may be too short for the functional signal to emerge, and the near-doubling of revision rate in Group B (29.4% vs. 13.3% in Group A; p=0.377) points toward clinical consequences without establishing them statistically. The authors' framing is precise and important: radiographic alignment loss should be treated as a surveillance signal warranting closer monitoring, not as established clinical failure.
7. No surgical parameters differentiated trajectory groups.
Levels fused, upper and lower instrumented vertebrae, surgical approach, rates of three-column osteotomy, estimated blood loss, and operative time did not differ significantly between trajectory groups. This finding is important: it suggests that the difference between patients who maintain correction and those who do not is not primarily explained by what was done surgically — it is driven by patient characteristics, particularly DJK status and age, that precede and exist independently of surgical technique choices.
THE FAILURE MECHANISM TIMELINE: A CLINICAL FRAMEWORK
The temporal shift in failure mechanisms has direct implications for how surveillance should be structured after cervical deformity surgery:
Early (0–3 months): Watch for distal junctional kyphosis. This is the period when DJK is most likely to manifest, likely reflecting the mechanical demands on the distal construct before fusion consolidation. Radiographic assessment of the distal junction at the 3-month visit is particularly important.
Mid-term (3–6 months): Watch for proximal junctional kyphosis. As distal fixation consolidates, the proximal junction becomes the zone of greatest vulnerability. Full-spine sagittal imaging at 6 months allows identification of emerging PJK before it becomes clinically symptomatic.
Late (6–12 months): Watch for global sagittal decompensation. The dominant failure mode in the second half of the first postoperative year is not junctional collapse but progressive loss of global sagittal compensation — a more diffuse and potentially harder to reverse deterioration. Full-length sagittal radiographs at one year are essential for capturing this pattern.
The authors explicitly support interval-specific imaging: junctional assessment early, full-spine sagittal radiographs at 6 months and one year.
UNDERSTANDING THE STATISTICAL APPROACH
This study's use of three complementary longitudinal methods — rather than the repeated cross-sectional comparisons common in prior literature — is a methodological advance worth noting.
Cox proportional hazards regression modeled the time to first alignment loss among initially aligned patients, accounting for the fact that patients are followed over time and that the event (alignment loss) occurs at different points for different patients. It identifies factors associated with the hazard (instantaneous risk) of losing correction.
Kaplan–Meier survival analysis with restricted mean survival time (RMST) estimated how long patients in different subgroups maintained alignment on average, providing an intuitive and clinically meaningful summary of durability differences between groups stratified by DJK status and baseline cSVA.
Mixed-effects logistic regression modeled alignment status across all four timepoints simultaneously for all 89 patients with complete data, accounting for the repeated-measures structure of the data and the correlation of observations within the same patient over time. The intraclass correlation coefficient (ICC) of 0.36 confirmed substantial within-patient correlation, validating the mixed-model approach.
Each method captures different aspects of the longitudinal data, and the convergence of DJK as a significant predictor across all three frameworks substantially strengthens confidence in that finding.
PRACTICAL IMPLICATIONS
DJK prevention should be a central goal of cervical deformity surgical planning.
The convergent significance of DJK across all three statistical frameworks makes it the most actionable finding in the study. Risk stratification for DJK — using validated scoring tools and careful LIV selection — should be a standard component of preoperative planning. Patients at high DJK risk should be counseled accordingly and monitored with particular attention at the 3-month postoperative visit.
Surveillance protocols should be structured around the failure mechanism timeline.
The temporal shift from DJK to PJK to global decompensation argues for a specific surveillance approach rather than a generic "see them at fixed intervals" protocol. Junctional imaging early, full-spine sagittal imaging at 6 months and one year, and more frequent monitoring of older patients are all supported by the data.
Age-stratified surveillance is warranted.
The Time × Age interaction is identifiable only through longitudinal modeling and would not have emerged from a simpler analysis. Older patients deteriorate at each successive postoperative visit at an accelerating rate. Protocols that treat all patients identically regardless of age are likely suboptimal, and the data support differentiating monitoring intensity by age.
The alignment durability findings should inform preoperative patient counseling.
Patients considering cervical deformity surgery should understand that achieving durable correction is challenging — the majority of patients in this study with complete follow-up did not maintain acceptable alignment at one year. This does not mean surgery fails to provide benefit: patient-reported outcomes improved across all groups, and radiographic deterioration at one year did not translate into significant NDI or EQ-5D differences. But it does mean that expectations should be appropriately calibrated, and that radiographic follow-up — not just symptom-based follow-up — is an important component of the postoperative plan.
The absence of surgical parameter differences between groups has important implications.
If levels fused, instrumented vertebrae, surgical approach, and technique factors did not differentiate trajectory groups, then patient selection and preoperative risk stratification — not primarily intraoperative decisions — may be the most important determinant of durability. This finding argues for investing heavily in preoperative characterization of DJK risk, age-related vulnerability, and alignment targets before the patient enters the operating room.
Radiographic alignment loss should trigger reassessment, not alarm.
The authors' framing of alignment loss as a surveillance signal rather than clinical failure is important for clinical communication. A patient who shows deteriorating radiographic alignment at 6 months should be assessed more closely — not told their surgery failed. The revision rate in Group B was higher than in Group A, but the difference was not statistically significant, and patient-reported outcomes did not separate the groups at one year. The surveillance signal should prompt reassessment and individualized decision-making rather than automatic intervention.
LIMITATIONS
The study's authors are commendably transparent about its limitations, which are worth summarizing for readers who go on to read the full paper:
Complete four-timepoint follow-up was available in only 89 of 155 patients (57.4%), introducing potential selection bias: patients who return for all imaging visits may be systematically different from those who do not, and the direction of this bias is uncertain — though the authors judge it more likely to overestimate than underestimate failure rates. The Cox model contained only 17 alignment loss events, limiting it to two covariates in the primary model and rendering all expanded-model estimates exploratory. Correction magnitude and deformity-specific alignment targets were not assessed, so the undercorrection interpretation of the lower cSVA finding remains hypothetical. Failure mechanisms were adjudicated radiographically and, for non-revised patients, inferred from serial imaging rather than confirmed clinically — the apparent temporal progression rests on small numbers at each visit and is descriptive. Patient-reported outcomes did not differ between groups, but the study was not powered to detect such differences. Follow-up was limited to one year, and the maintenance rates reported here are unlikely to improve with longer observation, as late failures from global compensation loss were still accruing at the final timepoint.
CONCLUSION
This study provides the most rigorous and comprehensive longitudinal characterization to date of alignment durability after cervical deformity correction surgery. Its central findings are sobering but actionable: fewer than one-third of patients with complete follow-up maintained acceptable alignment at one year; DJK was the most consistently powerful predictor of failure across multiple statistical frameworks; older patients deteriorated faster; and failure mechanisms followed a predictable temporal sequence from distal junctional failure early to global decompensation late. At the same time, the study offers an important clinical caveat: radiographic deterioration at one year was not accompanied by significant differences in patient-reported neck disability or quality of life, and should be framed as an early warning signal warranting surveillance rather than as established clinical failure. The path forward requires individualized preoperative risk stratification — centered on DJK risk and patient age — and structured postoperative surveillance calibrated to the temporal failure pattern this study has now described.
READ THE FULL STUDY
This article is published open access under a Creative Commons CC BY-NC-ND 4.0 license.
