The examination of reoperation risk among patients undergoing thoracic to pelvis surgery for ASD in the U.S., based on billing data since 2010, revealed a reoperation rate of 11%, 17%, and 22% at 1-, 2-, and 5-years, respectively. Preoperative kyphosis and ≥ 13 levels of posterior instrumentation was associated with higher reoperation risk, while the use of interbody cages was protective. Age, medical comorbidities, and osteotomies did not predict reoperations. This real-world cohort, with a minimum 2-year follow-up, is over six times larger than the current largest prospective dataset in the literature [7].

Reoperations among spine deformity patients are unwanted but prevalent [5, 8]. Since the emergence of prospective studies indicating surgical intervention is more effective than nonoperative treatment in patients with ASD [9, 10], efforts have been made to estimate a patient’s risk for adverse surgical outcomes and improve safety and effectiveness [7, 11]. As a result, postoperative complications and reoperations have reportedly halved over the last decade despite increased patient enrollment, according to the largest available prospective registries [7]. While these prospective registries provide high quality data, one must consider if the results align with practices across the U.S. and surgeons outside the study group. Similar to the ISSG and ESSG registries, our study included patients with 2 years of follow-up but come from a mixed payor national database from all U.S. states and territories. Notably, the limitations of coding prevented us from creating a direct comparison of patients based on their selection criteria of scoliosis ≥ 20°, sagittal vertical axis (SVA) ≥ 5 cm, pelvic tilt ≥ 25°, or thoracic kyphosis (TK) ≥ 60° [7]. Nevertheless, we were able to create a pure cohort of surgeries billed to insurance as T12-pevlis or greater in patients with associated deformity diagnosis codes.

Historically, postoperative complications among patients have been linked to patient age, nicotine use, and medical comorbidities [12,13,14]. However, in this study, age, nicotine use, and medical comorbidities did not independently predict reoperation, at least at the level of granularity afforded by an administrative database. This dataset is likely subject to some selection bias, such that older, sicker patients may not be deemed surgical candidates to begin with. Or, if they do undergo surgery and present years later with mechanical complications, they may not be deemed surgical candidates at that time and therefore not undergo subsequent operations [15]. Age not being associated with reoperation aligns with a recent meta-analysis that explored heterogeneity in treatment outcomes among patients with and without surgery, focusing on age as a potential source of variability [5]. Sub-group analysis by age revealed that surgical treatment was superior to nonsurgical treatment in terms of functional and symptomatic improvement. These findings, and our own, suggest that age alone should not be used as an absolute factor to deny surgery. An individualized decision-making process, considering a patient’s goals of treatment and a more robust evaluation of risk including functional capacity, sarcopenia, frailty, and local bone quality, is likely a better approach [16]. In the context of this study, there was not a reliable way to identify patients who were active smokers and those who were told they had to quit smoking prior to surgery. Considering this and the rate of patients who return to smoking after quitting [17], we elected to encompass current and past nicotine users in our smoking cohort. Consequently, our finding that there was an increased, but not statistically significant, risk with reoperation may have been affected by capturing more former nicotine users rather than active nicotine users. Nonetheless, it is well known that outcomes in spinal surgery for patients who smoke are worse than those that do not [13, 14].

The risk for mechanical complications is affected by the severity of the deformity type and degree of malalignment. Our finding that kyphosis was associated with higher reoperation rates, while scoliosis, spondylolisthesis, and sagittal plane deformities were not, highlight the importance that not all subtypes of ASD carry the same risk for mechanical complications. Previous studies have shown that greater pre-existing thoracic kyphosis and magnitude of sagittal plane correction correlates with the risk for proximal junction kyphosis [PJK) [18, 19]. Maruo et al. identified that patients with pre-existing thoracic kyphosis greater than 30 degrees and a change in lumbar lordosis greater than 30 degrees were independent risk factors for PJK [19]. These findings are supported by several other studies and reinforce the importance of preoperative planning with consideration of surgical planning and individualized counseling regarding goals of care [16, 20, 21].

One theorized factor contributing to the higher mechanical complication rates associated with more severe spinal deformities is the greater degree of surgical invasiveness required to realign the spine. In this study we found that patients undergoing ≥ 13 levels of posterior instrumentation were at greater risk of reoperation compared to patients with 7–12 levels. However, neither the use of posterior column nor three-column osteotomies predicted reoperation in this series. In patients with spinal deformities, it is often challenging to identify the level of upper instrumented vertebra to achieve optimal patient outcomes while minimizing complication rates. Previous studies have found the incidence of proximal junction failure varies when selecting the level of UIV, however the evidence is not clear [22]. The decision of UIV involves numerous factors and can be challenging, even for experienced deformity surgeons. Consequently, future studies are needed to evaluate the length of fusion and long-term outcomes. Like length of fusion, osteotomies involve greater surgical invasiveness and are often required to increase flexibility and achieve optimal alignment in larger rigid deformities. Though effective, osteotomies are technically challenging with potential for perioperative complications, such as pseudoarthrosis and rod fracture [1]. Unfortunately, we could not assess the severity of presenting malalignment and the efficacy of deformity correction with osteotomies within a billing database. Notably, it is possible that our findings may be influenced by surgeon sampling errors, non-representativeness, heterogeneity of treatment effect, and possibility of confounding variables. Other study designs would be more suited to investigate the influence of osteotomies with degree of correction and patient outcomes.

Surgical management of ASD through a posterior approach is often supplemented with circumferential fusion with the use of interbodies [1, 23]. In this study we found that the addition of interbody cages to posterior instrumentation was protective against reoperation. These findings validate several smaller studies showing satisfactory clinical and radiological outcomes [1, 24]. Mechanical advantages to interbodies include increased anterior column support, lower incidence of posterior implant failure, ability to restore sagittal alignment, and achieve indirect decompression [25].

This study is not without limitations. Although database research is useful for obtaining data for treatments with low incidences, it is limited by the amount of detail provided and the nature of information available. Notably, there was an inability to evaluate radiographic data, operative reports, degree of disability, sarcopenia, surgeon experience, technical expertise, and indications for operation. All of which are valuable for informing treatment decisions, determining the necessity for surgery, and understanding the reasons for reoperation in individual patients. Additionally, database research relies on billing data, which may contain some degree of misclassification of symptomatology or misrepresent reoperations. Furthermore, the billing codes utilized in this study lack information about the specific spinal level where surgeries were performed, thus restricting the ability to explore index versus adjacent segment pathology. Additionally, billing codes lack data on implant types (e.g., titanium vs. cobalt-chromium, or the use of pedicular screws, sublaminar wires, or hooks). Consequently, we were unable to analyze whether different types of implants had any correlation with the rate of reoperation. We acknowledge that such information could provide valuable insights into potential associations between implant choice and clinical outcomes. Despite the limitations inherent in billing database studies, it is crucial for practitioners to understand and critically review this type of data, as payors may use it to evaluate the durability and efficacy of treatments. As the emphasis on both short- and long-term healthcare cost-effectiveness becomes more important, this data is likely to have an increasingly significant impact on clinical care.