How Much Variance Exists Among Published Definitions of Proximal Junctional Kyphosis: A Retrospective Cohort Study of Adult Spinal Deformity

1. Introduction

Adult spinal deformity (ASD) is a common spinal disorder that is projected to become increasingly prevalent in the future [1]. ASD may be brought about by multiple etiologies, including degenerative disease, iatrogenic causes (e.g. prior spinal surgery), or conditions such as ankylosing spondylitis [2]. Regardless of the cause, ASD often must be addressed via long-segment posterior pedicle screw and rod constructs. These surgeries can be quite effective in restoring proper alignment and promoting adequate fusion. However, extensive rigid constructs also carry the risk of complications, notably proximal junctional kyphosis (PJK), broadly defined as kyphosis that develops at the interface of the cranial end of the construct and the adjacent mobile segments [1,3,4]. The prevalence of PJK after surgery for ASD may lie between 20-40%, though some sources report rates as high as 61% [5,6,7,8].

While a global definition of PJK is consistent among publications, various authors have put forth different nuanced criteria [3,5,9,10]. In general, PJK is usually defined as an excess PJK angle, a Cobb angle formed between the upper-instrumented vertebra (UIV) and a supra-adjacent vertebra (SAV) either one (UIV+1) or two (UIV+2) vertebral levels above the UIV. Depending on the study, a Cobb angle of at least 10-20 degrees is required for the diagnosis of PJK [3,5,9,10]. Different cutoff values for PJK angle have been analyzed with respect to their prognostic value as predictors of postoperative pain, functional recovery, or need for revision surgery [11,12,13]. Nonetheless, the results have not supported one strict definition of clinically significant PJK and no expert consensus currently exists as to the ideal magnitude to use for the PJK angle or which SAV should be used.

To that end, we used a large cohort of patients >65 years of age who underwent thoracolumbar fusions for ASD to better characterize the variance and overlap among multiple existing definitions of PJK used in the literature and to evaluate which definitions may be the most useful to compare across future studies.

2. Materials and Methods

Radiographic and demographic data were collected for thoracolumbar fusion from an institutional database for patients >65 years old with a diagnosis of ASD. Patients who underwent fusion of at least three segments within the thoracolumbar region with pelvic fixation at the authors’ institution between 2014-2024 were included, while patients with missing radiographs for required analysis were excluded. Preoperative and most recent follow-up (>6 months postoperative) full-body scoliosis radiographs were exported from the patient’s electronic medical record (EMR). The upper instrumented vertebrae (UIV) for each patient were identified from intraoperative/postoperative imaging, and the Cobb angle was measured between the inferior endplate of the UIV and the superior endplate of the vertebrae immediately cranial to the UIV (UIV+1) (Figure 1, blue). A second Cobb angle was measured between the inferior endplate of the UIV and the superior endplate of the vertebrae two levels cranial to the UIV (UIV+2) (Figure 1, red) [14]. Angles were measured from both the preoperative and most recent follow-up radiographs.

A PubMed search with key terms (“proximal junction* kyphosis” AND “definition”) and subsequent exploration of citations yielded 6 criteria for varying definitions of PJK based on different thresholds of the UIV to UIV+1 or UIV+2 cobb angle [3,5,9,10,15,16]. A series of statistical tests were performed with R: A Language and Environment for Statistical Computing version 4.4.1 (CRAN, 2024). A Pearson’s Chi Squared test was performed among the different rates of PJK based on the definitions identified from the PubMed search. A subsequent pairwise comparison of proportions was performed to evaluate the distinctness of each individual definition’s rate in comparison to one another. Rate of agreement among PJK definitions were also analyzed in a pairwise fashion, followed by an evaluation of the proportion of PJK diagnoses that would also be diagnosed by additional criteria. Statistical significance was established by an alpha <0.05, and 95% confidence intervals (95% CI) were calculated.

3. Results

A total of 116 patients met inclusion criteria, 79 of whom were female (68.1%). The average age of the cohort was 70.9 + 4.3 years, and the average BMI was 28.9 + 5.4 (Table 1). Analysis of PJK definitions requiring identification of the UIV+2 decreased the total cohort to 111 patients due to visual obstruction of the UIV+2 vertebrae from either image cutoff or low resolution for 5 patients. For consistency, analysis of all PJK definitions were performed among the 111 eligible patients.

The following 6 criteria for PJK were identified from the PubMed search: [1] PJK angle >20° with UIV+2 as the SAV; [2] PJK angle >10° with a >10° change from preoperative values with UIV+2 as the SAV; [3] PJK angle >2 standard deviations from average with UIV+1 as the SAV; [4] PJK angle >10° with a >10° change from preoperative values with UIV+1 as the SAV; [5] PJK angle >15° with UIV+1 as the SAV; and [6] PJK angle >30° with UIV+2 as the SAV, displaced rod fracture, or reoperation within 2 years for junctional failure, pseudoarthrosis, or rod fracture (Table 2) [3,5,9,10,16]. These PJK definitions will henceforth be referred to as [1] PJK20, [2] PJK10, [3] PJK2SD, [4] PJK10+10, [5] PJK15, and [6] PJK30. PJK rates, by each definition, were 1) 20.7% (95% CI: 13.8-29.7%), 2) 36.9% (95% CI: 28.1-46.7%), 3) 3.6% (95% CI: 1.2-9.5%), 4) 23.4% (95% CI: 16.1-32.6%), 5) 15.3% (95% CI: 9.4-23.7%), and 6) 10.8% (95% CI: 6.0-18.5%).

Pearson Chi-Squared testing revealed significant variance among rates of PJK by criteria (p=2.6*10^-9, χ2=48.7, 95% CI: 0.83-12.8) with a Cramer’s V of 0.27 (95% CI: 0.18-0.34). Post-hoc pairwise proportion testing with Holm p-value adjustment revealed 5 significantly distinct definition pairs and 10 non-significantly distinct pairs. The distinct pairs were PJK20&PJK2SD (p=2.63*10^-3, 95% CI: 0.088-0.25), PJK10&PJK2SD (p=4.86*10^-3, 95% CI: 0.24-0.43), PJK10&PJK15 (p=2.80*10^-8, 95% CI: 0.10-0.33), PJK10&PJK30 (p=1.50*10^-4, 95% CI: 0.15-0.37), and PJK2SD&PJK10+10 (p=4.90*10^-4, 95% CI: 0.11-0.28) (Table 3). In other words, the overall PJK rates for our cohort given by these distinct pairs were significantly different from each other. Meanwhile the overall PJK rates given by every other pair were not significantly different.

The pairwise agreements of the cohort classification as PJK positive or negative among these six definitions are shown in Table 4. These percentages represented the proportion of patients that were similarly diagnosed as having or not having PJK between the pair of definitions. The pairs with greatest agreement were PJK10+10&PJK15 at 90.1% (95% CI: 82.6-94.7%), PJK2SD&PJK30 at 89.2% (95% CI: 81.5-94.0%), and PJK2SD&PJK15 at 88.3% (95% CI: 80.5-93.4%) (Table 4). Furthermore, PJK30, PJK2SD, and PJK15 had the greatest agreement with reoperation at 82.9% (95% CI: 74.3-89.1%), 79.3% (95% CI: 70.3-86.2%), and 71.2% (95% CI: 61.7-79.2%) respectively.

Among patients with PJK according to a particular definition, the following percentages were identified as having PJK by at least one other definition: PJK20) 91.3%, PJK10) 78.0%, PJK2SD) 100%, PJK10+10) 88.5%, PJK15) 100%, PJK30) 29.4%. The percentages of each definition’s PJK cohort that met criteria for additional PJK definitions are shown in Table 5.

4. Discussion

Discussions of PJK are of much interest in spinal deformity literature. However, this condition is heterogeneously defined depending on the article and classification criteria used. In this study we used a large cohort of patients who underwent surgical correction of ASD to evaluate 6 different definitions of PJK identified in the current literature. We investigated which definition may encapsulate the most patients in our cohort to determine the definition that may be of most utility when comparing PJK among studies or for potential surgical decision-making. While we found that certain definitions of PJK captured similar patient subsets of our overall patient cohort, there were nevertheless significant differences between multiple criteria as to which patients would qualify as having PJK. In aggregate, our work demonstrates that comparison of results between studies that utilize distinct definitions of PJK may be difficult to interpret and of limited utility.

Only PJK definition PJK30 directly considered patient outcomes beyond imaging findings by integrating the need for reoperation or instrumentation failure [16]. Hence, we considered it the most clinically relevant definition for a symptomatic PJK. PJK30’s close relationship with patient function and greatest agreement with reoperation rates supported its use as a pseudo- “gold standard” for comparison with other definitions (Table 4).

Considering the raw rate of PJK-positive patients in the same cohort, definition PJK2SD was the strictest for assigning PJK, while definition PJK10 was the most lenient among the six PJK definitions. The low rate of PJK for definition PJK2SD suggests that the dispersion of UIV/UIV+1 angles was functionally small to yield such a stringent diagnosis [9]. Notably, the tightness of distribution of this UIV/UIV+1 angle was relatively more extreme than the already conservative gold-standard of definition PJK30. This means that in contrast to clinically relevant definition PJK30 (diagnosing 10.8% of the entire cohort), only 4.5% of the entire cohort experienced the kyphotic angle at the UIV varying outside expected ranges per definition PJK2SD. A conservative calculation, assuming all 4.5% PJK patients per definition PJK2SD were also accounted for by definition PJK30, found that 58.3% of patients who experienced symptomatic PJK by definition PJK30 did not meet the threshold for definition PJK2SD. Considering that only 89.2% of positive or negative PJK diagnoses from definitions PJK2SD and PJK30 overlapped (Table 4), 65.4% of patients with symptomatic PJK (qualified by definition PJK30) were not accounted for by definition PJK2SD (Figure 2). It should be noted that definition PJK2SD was historically applied to adolescent scoliosis, which features a distinctly different clinical, biomechanical, and pathophysiologic picture than the degenerative ASD featured in our cohort [17].

A comparison of raw PJK rates from definitions PJK10 and PJK10+10, which differ only by usage of the UIV+1 or UIV+2 vertebrae as the cranial vertebrae in Cobb angle measurements, demonstrates that use of definition PJK10+10, and thus the UIV+1 vertebrae, results in stricter criteria for PJK (Table 2) [3,10]. Again, if we were to use definition PJK30 as the standard for symptomatic PJK, use of definition PJK10, or the UIV+2 vertebrae, could be interpreted as too lenient of a definition in comparison to use of definition PJK10+10, or UIV+1.

Closer evaluation of the pairwise overlap between definition PJK30 and either definitions PJK10 or PJK10+10 confirms that definition PJK10+10, or use of the UIV+2, also diagnoses the PJK status of individual patients more similarly to that of definition PJK30 (Table 4). In fact, definition PJK10 had the least overlap with definition PJK30 compared to all other definitions, whereas definition PJK2SD, which used extreme deviations of PJK from average values, had the best overlap. This might suggest that while definition PJK2SD is the strictest criteria for PJK, those who met the high threshold to be considered PJK positive by PJK2SD were more likely to require reoperation. However, definition PJK30 also includes a criterion for extreme PJK angle magnitude, which may confound this conclusion.

Chi-square testing with post-hoc pairwise proportions analysis and p-value adjustment demonstrated that 10/15 (67%) definition pairs were not significantly different when evaluating overall cohort PJK rates (Table 3). Among the 5 distinct pairs, definitions PJK10 and PJK2SD were most common. Definition PJK10 was found to be too lenient and definition PJK2SD too strict based on raw PJK rates, which could explain why they was often statistically different from the other definitions and limited in practicality. Definitions PJK20, PJK10+10, PJK15, and PJK30 were not distinct from each other. These results suggest that definitions PJK10 and PJK2SD should be used with caution when consistency across studies is desired, while definitions PJK20, PJK10+10, PJK15, and PJK30 may be used more interchangeably.

From the evaluation of additional PJK criteria that were encompassed by each of the 6 definitions, as shown in Table 5, patients diagnosed as having PJK by definition PJK2SD would also be sufficiently diagnosed by most other definitions. In the context of definition PJK2SD’s stringent nature, this suggests that most other PJK definitions have sufficient coverage and that use of definition PJK2SD may be redundant. On the other end of the spectrum, the patients labeled PJK-positive by PJK30 were often not diagnosed by other PJK criteria. This finding further supports the utility of definition PJK30, including capturing unique information that is missed by the remaining PJK definitions.

Overall, our findings suggest that definition PJK15 most optimally balances the clinical relevance of definition PJK30 while not yielding statistically different rates of PJK in our cohort. Additionally, definition PJK15 utilizes the UIV+1 vertebrae, which found greater agreement with definition PJK30 than when measuring the UIV+2 vertebrae. Definition PJK15 may also identify earlier PJK than definition PJK30 because PJK15, which relies purely on imaging characteristics, is not contingent on the patient undergoing reoperation, allowing clinicians to identify early PJK that may not have yet progressed to such a degree to necessitate surgical intervention.

5. Limitations and Future Steps

This study has several limitations. First, its retrospective design inherently limits its ability to establish causal relationships between the definitions of PJK and outcomes such as reoperation rates. Secondly, data collection was performed at a single institution, which may reduce generalizability of our findings to broader populations. Third, the exclusive focus on patients aged 65 and older may not reflect PJK dynamics in younger cohorts undergoing spinal surgeries. Another limitation is the study’s reliance on reoperation as a proxy for clinical outcomes. Other important measures such as patient reported pain, functional recovery and quality of life were not analyzed, potentially leading to incongruence between radiographic outcomes and patient-centered outcomes. Furthermore, some definitions, such as PJK2SD, were historically applied to adolescent scoliosis, which presents distinct biomechanical and clinical characteristics compared to adult spinal deformity. These differences may limit the applicability of such definitions to the adult population examined in this study. Further studies should aim to validate these definitions prospectively in multi-institutional cohorts and consider broader clinical outcomes beyond reoperation to provide a more comprehensive understanding of PJK. Finally, it should be noted that no formal non-radiographic analysis was performed for definition PJK15 to definitively demonstrate its superiority in predicting clinical outcomes compared to other definitions (e.g. Delphi analysis). PJK15’s applicability was inferred from its statistical overlap with PJK30 – a definition with intrinsic clinical outcome considerations – and this concept should be evaluated more thoroughly in future studies.