Background
Adult spinal deformity (ASD) is a debilitating condition that often requires surgical correction. The aging population, combined with an increasing number of spine fusion procedures, is increasing the prevalence of ASD [
1,
2]. Osteotomy procedures, including Smith-Peterson (SPO) and pedicle subtraction (PSO) are effective procedures used to correct deformity and lead to improved clinical outcomes [
3‐
6]. One important difference between these osteotomies is that SPOs only involve resection of the posterior column of the spine while PSOs involve a greater resection including the pedicle and wedge of vertebral body [
7,
8]. As a result, one PSO can achieve greater correction (30–40°) than a single SPO (approximately 10°) [
9‐
13]. Due to the physical differences in osteotomies, it would seem plausible that after PSO vs. SPO the spine and fusion construct would be subjected to different strains and subsequent complications. However, while mechanical complications after PSO have been well described, these complications after SPO haven’t specifically been of focus. Multiple authors have described increased complications [
4,
6,
14‐
16], pseudarthrosis [
3,
4,
14,
15,
17], and instrumentation failure including rod fracture (RF) in PSO patients [
4,
6,
14,
18]. Of note, rod fracture often requires reoperation and studies have postulated a link between pseudarthrosis and RF [
4,
6,
14,
18]. However, prior studies have not statistically confirmed this linkage due to study limitations.
Rod fracture can be differentiated into two main subgroups: clinically significant rod fracture (CSRF) and non-CSRF (CSRF was termed “symptomatic rod fracture” in a previous study) [
18]. CSRF can be defined as symptoms prompting evaluation (e.g., pain, neurological symptoms) and diagnosis of fractured rods via radiographic imaging or re-operation. CSRFs are often accompanied with radiographic signs of pseudarthrosis, new symptoms such as pain or radiculopathy, or resulting in loss of correction [
18], while non-CSRFs by definition aren’t associated with patient symptoms or loss of correction. Previous studies have found that the incidence of RF varies widely after spinal correction, with a recent study citing a post-PSO incidence of 22 % for all RF including non-CSRF and CSRF [
4,
14,
18,
19]. The largest study to date studying CSRF only, reported a lower incidence rate of 6.8 % in a population of ASD patients treated with long, >5 level posterior instrumented fusion, with a higher incidence in a subset of patients who underwent osteotomy at 15.8 % [
18]. Of note, no large series to date has cited the incidence of RF and CSRF subset in a combined cohort of adult spinal deformity patients treated with either PSO and/or SPO and compared the incidence between the two procedures.
Prior authors have suggested subdividing the etiology of RF into two main categories: early and late RF [
18]. Early RF (<12 months) is thought to be attributed to primary instrumentation failure and may lead to delayed union while late RF (>12 months) may be partially attributed to pseudarthrosis [
18].
Biomechanical risk factors for rod fracture studied to date can be divided into two main categories: intrinsic and extrinsic properties. Intrinsic properties include material type (mainly stainless steel, titanium, and cobalt chromium alloys) and diameter (usually ranging from 5.5–6.35 mm). Pertaining to metal type and rod diameter, studies have either found no statistically significant differences or have published conflicting results [
18,
19]. Extrinsic factors can be divided into several previously described biomechanical notions. The first notion states that primary contouring or metal bending (including the use of French benders) has shown to weaken metal and decrease fatigue strength [
20,
21]. Further, repeated bending or rod contouring to an extreme angle has been hypothesized to decrease rod fatigue strength [
20‐
25]. While biomechanical studies have shown a link between “excessive contouring” and decreased fatigue strength, no clinical study to date has confirmed if this “excessive contouring” poses a significant risk in real-world surgical applications. Additionally, “notch sensitivity,” results from defects in the surface of metal, particularly titanium, which decrease the strength of the metal [
26,
27]. Third, cyclic loading causes repeated metal strain leading to inherent metal fatigue and failure, particularly in titanium constructs [
22‐
24,
28], while single incident instrumentation overloading is less common [
28]. Similarly, notch sensitivity and cyclic loading haven’t received statistical confirmation in clinical studies.
Several risk factors have been confirmed statistically in clinical applications. A recent study on RF in patients undergoing instrumented posterior fusion (including a subset of PSO patients) found statistically significant risk factors for RF in this population including age, body mass index (BMI), baseline sagittal imbalance, baseline pelvic incidence minus lumbar lordosis (PI-LL) mismatch, and greater sagittal imbalance correction [
19]. This study found no statistically significant risk for RF pertaining to smoking, levels fused, and rod diameter. Further, the literature includes a limited number of clinical studies that have proposed risk factors for mechanical complications (some encompassing RF) following posterior spinal fusion in various populations. Some of these variables are thought to have an association with instrumentation failure by increasing construct strain and include under-corrected sagittal vertical axis (SVA) (
OR = 17.5) [
29], number of instrumented vertebra (
OR = 2.23–11.46 depending on number of levels) [
14], and fusion across transitional spine junctions such as the lumbosacral junction (
OR = 2.8) [
14,
30]. Other variables linked to mechanical complications include absence of anterior column support (i.e., interbody fusion), insufficient distal foundation, and fusion to the sacrum [
16,
18,
30‐
32]. However, these studies often included a broad population of instrumented posterior fusion, did not analyze RF and CSRF specifically, and did not provide risk assessment analysis for osteotomy patients. Therefore, analyzing risk factors for RF, specifically CSRF after osteotomy is important. The purpose of this study was to determine incidence and risk factors for RF and CSRF after posterior instrumented correction of adult spinal deformity with osteotomy.
Discussion
This study is a case-series of 75 consecutive ASD patients assessed for incidence and risk factors for RF and CSRF after instrumented posterior fusion and osteotomy (SPO and PSO). Total incidence of RF was 9.3 % with 16.2 % after PSO and 2.6 % after SPO. Clinically significant total rod fracture rate was 5.3 % and post-PSO rate was 8.1 %. Following risk factor analysis, statistically significant risk factors for RF were obtained: fusion construct crossing 2 spine junctions, sagittal rod contour >60°, presence of dominos and/or parallel connectors at date of fracture, and pseudarthrosis at ≥1 year follow-up. Statistically significant risk factors for the clinically significant subset were fusion to pelvis and pseudarthrosis at ≥1 year follow-up. Interestingly, most variables with increased risk, including those meeting statistical significance or borderline significance, were similar between total RF group and CSRF subset. However, due to the low number of only 4 CSRFs, most variables only met borderline significance.
Prior studies have reported a variable incidence of rod fracture after PSO, with two the largest studies to date reporting an incidence of 22 % for all rod fractures and 15.8 % for symptomatic rod fracture [
18,
19]. Our overall post-PSO RF incidence of 16.2 % is comparable to what has been previously described, but our post-PSO CSRF rate of 8.1 % is approximately 1/2 the rate of what has been previous reported [
18]. However, prior rates cannot be perfectly compared to our rates due to differences in study design including inclusion/exclusion criteria, for example prior study inclusion of ≥5 fusion levels vs. current study inclusion of ≥3 fusion levels. Further, we hypothesize that our lower rate may be underestimated due to minimum follow-up of 1 year in some of these patients, or due to differences in surgical technique and instrumentation trends at our institution. Several institutional trends associated with decreased rod strain include use of polyaxial screws (100 %) and use of anterior support including interbody fusion in a majority of our patients. A recent finite element analysis of a PSO model comparing mono-axial vs. poly-axial screws showed rod contour affected the location of bending moments and stress [
34]. Anterior support below the PSO level reduced bending moments along the rod (−26 %) [
34]. The lower incidence of CSRF could also be explained by patient demographic (Colorado population, patient selection) and surgeon expertise.
Our study found no significant impact of patient demographics or baseline characteristics on the RF or CSRF risk. It has been shown recently that patients with RF had higher BMI and age than those without RF [
19]. Our results may differ based on the lower statistical power of our study or differences in design including: multicenter vs. single center and inclusion/exclusion criteria.
Surgical variable risk factors meeting statistical or borderline significance for RF and CSRF included: fusion constructs crossing two junctions and fusion to the pelvis. In prior studies, fusion across junctional regions of the spine, lumbosacral or thoraco-lumbar has been shown to be a risk factor for complications and increased strain on instrumentation [
9,
14,
30,
32]. However, pelvic fixation alone has not been previously described as having association with RF. We propose that the association between CSRF and pelvic fixation could be the result of a few possible underlying etiologies. First, pelvic fixation may be a confounder to fusion across the lumbosacral junction of the spine, as pelvic fixation can only be present in those with fusion across the lumbosacral junction. Supporting evidence showed a high risk, with borderline significance, toward a greater proportion of RF and CSRF subjects with fusion to the sacrum compared to non-CSRF subjects. Alternatively, pelvic fixation alone may increase rod strain as pelvic fixation has been shown to increase overall spinal construct stiffness, but may inflict pinpoint strain on the rods [
35]. Supporting evidence is seen with multiple rods fracturing near or at iliac bolt connector. Lastly, an extremely unbalanced spine requiring additional support can be a factor necessitating pelvic fixation and thus contributing to rod stress. Therefore, a spine with a high degree of imbalance at baseline, rather than the pelvic fixation alone, may explain the association between CSRF and pelvic fixation. Supporting literature found increased risk for RF with increased baseline sagittal imbalance and increased SVA correction [
19]. Another surgical variable showing increased risk for RF, but only meeting borderline significance, was osteotomy type - PSO vs. SPO. Prior studies haven’t quantified this risk. However, PSOs have a couple factors that may increase rod strain including absence of pedicle screws at PSO level and greater degree of correction obtained with PSO vs. SPO.
Our findings support previous experimental data that excessive rod contouring causes notching and internal strain that decreases rod fatigue strength and increases risk of RF [
20‐
24,
26,
27,
36]. In particular: sagittal rod bend >60° and presence of dominos and/or parallel-connectors were found to be statistically significant instrumentation risk factors for RF. These two variables also showed increased risk, but only held borderline significance in the CSRF subset. Over half of fractured rods had failed at the apex of rod bend with rod bends averaging over 60°. One CSRF occurred at an area previously noted in the operative note that the rod was “bent out of way” and subsequently bent back into a new iliac bolt. Two RF subjects had rod fractures located at or near a domino connector. Presence of domino or parallel connectors was evaluated at date of fracture, to account for a subsequent operation that may have used these connections to connect to original instrumentation. This finding is important because revision surgery and connection to prior instrumentation often calls for re-contouring of the end of the rod to match new instrumentation. Other variables associated with increased risk but only meeting borderline statistical significance included ≥2 crosslinks vs. <2 crosslinks (RF and CSRF) and screw density, complete vs. incomplete (CSRF only). Quantified risk pertaining to these factors has not been published in prior studies.
Pseudarthrosis was revealed in over half of RF subjects and ¾ of CSRF subjects, showing statistically significant association with rod fracture. This result provides statistic confirmation of previous findings, and can be explained by effect of cyclic loading at a non-fused segment allowing micro-movements to increase construct strain and risk of instrumentation failure [
4,
6,
14,
18].
Previous studies suggested that residual postoperative sagittal imbalance may be associated with risk of instrumentation failure; however, these studies are without statistical confirmation [
18,
29]. A recent study on RF found the amount of SVA correction, rather than postoperative SVA was a significant risk for RF [
19]. The risk analysis performed in our study did not reveal significant association between postoperative spinopelvic parameters and RF/CSRF. This finding could simply be contributed to our small cohort, or may signify that differences in surgical technique and instrumentation may protect against under correction. Of note, 40 % of non-RF subjects in our study had postoperative sagittal imbalance of >5 cm and didn’t experience a rod fracture. Only 3 of the 25 total subjects with postoperative sagittal imbalance developed RF during the follow-up period.
Our study had several limitations, the main of which are: retrospective design with the inherent risk of selection bias; and relatively small sample size likely cause underestimation of significance for a few studied risk factors. Further, missing data including the type of interbody device and bone graft and rod characteristics in multiple patients should be regarded as a limitation. Therefore, the presented results should be viewed as preliminary basis for further prospective studies with a larger cohort.
Based on these preliminary findings found in our retrospective review case series, it may be reasonable for clinicians to try and eliminate or reduce these risk factors for rod fracture in their ASD patients to possibly decrease the risk of rod fracture. With consideration of the limitations in our study, incidence of rod fracture may be decreased by: 1) pursuing all means to avoid post-operative pseudarthrosis 2) use the smallest sagittal rod contour possible while providing adequate correction, in particular <60° 3) decrease use of domino and/or parallel rod connectors when possible 4) if unnecessary, do not implement pelvic fixation. Finally, patients with radiographic evidence of pseudarthrosis after 1 year post-operatively should be regarded as having increased risk of rod fracture and may require more careful observation.
Competing interests
This study has no competing interests to disclose.
Authors’ contributions
CB performed primary data collection, spinopelvic parameter measurement, and created primary draft of manuscript. AN participated in the study design, provided statistical analysis, and helped draft the manuscript. CK helped draft the manuscript. CC helped draft the manuscript. VP helped draft the manuscript. EB conceived the study, participated in its design and coordination, made inclusion/exclusion decisions, and helped draft the manuscript. All authors read and approved the final manuscript.