Introduction

Progressive adult spinal deformity (ASD) is a debilitating condition leading to severe disability [1]. With improved medical care and higher life expectancy in the United States [2], some reports suggest a 68% prevalence of ASD in individuals older than 65 years [3]. Although ASD is symptomatic in a smaller subset of patients, progressive spinal deformity can result in axial back pain, gait instability, and neurological impairment [4].

Despite improvements in surgical treatment of ASD, the risk of complications is high, even in a perfectly executed ASD surgery [5, 6]. Prior studies endorse a reoperation rate of 15–20% following ASD surgery [7], with complications as high as 70% [8], and a 20–50% incidence of mechanical complications [9], which include proximal/distal junctional kyphosis (PJK/DJK), rod fracture, pseudarthrosis, and implant failure [10]. Recent data have shown that outcomes between patients with and without mechanical complications are similar, as long as the mechanical complication is adequately treated [10].

Researchers and surgeons have often grouped mechanical complications together as a single entity; however, it remains unknown what differences exist between mechanical complication sub-types. Moreover, strategies to treat each respective mechanical complication differ. While PJK occurs at the proximal end of a construct and requires extension of fusion, pseudarthrosis and implant failure often occur at more distal segments involving the lumbosacral junction. The complexity and invasiveness of how to treat each mechanical complication varies considerably.

Considering previously unstudied differences between mechanical complications in ASD surgery, we sought to further evaluate mechanical complication sub-types with respect to one another. In a group of ASD patients undergoing surgery, we aimed to: (a) describe the time course of each mechanical complication, and (b) compare different types of mechanical complications amongst each other in the areas of demographic factors, operative variables, radiographic measurements, and preoperative patient-reported outcomes measures (PROMs).

Methods

Study design

A single-institution, retrospective, case–control study was designed using prospectively collected data from our institution’s spine outcomes registry from 2009–2017. The registry team includes three full-time employees whose role includes contacting patients to collect PROMs data at several pre- and postoperative timepoints. A total of 12 fellowship-trained neurosurgery and orthopedic spine surgeons have contributed patients in the decade of the registry’s existence. Institutional review board (IRB) approval was obtained for this study (IRB#211290). Informed consent was obtained for all patients.

Patient population

Registry data were obtained for patients who underwent elective ASD surgery from 2009–2017. Inclusion criteria were: ≥ 5-level fusions with sagittal/coronal deformity and/or regional scoliosis requiring surgical reconstruction. All patients had a minimum of 2-year follow-up to assess of the occurrence of a mechanical complication. Therefore, all patients were included in the same analysis regardless of their specific follow-up period in an effort to increase study power. Patients who were lost to follow-up were also noted.

Exposure variables

Several exposure variables were evaluated and included the following: (1) demographic factors, (2) operative variables, (3) radiographic measurements, and (4) preoperative PROMs. Demographic variables included age, sex, body mass index (BMI), comorbidities, and Hounsfield Units. The latter were measured on three axial slices of one vertebra, either at the upper instrumented vertebra (UIV) itself or at a vertebra within UIV ± 4 from computed tomography (CT) scans preoperatively [11]. Operative variables included: UIV region, UIV implant, total instrumented levels (TIL). Radiographic variables included pelvic incidence (PI), pelvic tilt (PT), thoracic kyphosis (TK), sagittal vertical axis (SVA), lumbar lordosis (LL) L1-S and L4-S1, PI:LL mismatch, and lordosis distribution index (LDI). Though LDI is an imperfect measurement, as it does not consider whether the total LL is enough for the PI, it nonetheless captures the distribution of lordosis. All radiographic variables were taken preoperatively and at 6 weeks postoperatively and were measured by an orthopedic/neurosurgery resident. Any complex radiographs were confirmed by a fellowship-trained spine surgeon. Preoperative PROMs included Oswestry Disability Index (ODI), Numeric Rating Scale (NRS)-Back, NRS-Leg, EuroQol-5 Dimension (EQ-5D).

Outcome variables

The primary outcomes were occurrence of a mechanical complication and time to mechanical complication. To fit the case–control design, patients with mechanical complications were obtained, and several exposure variables were retrospectively evaluated after dividing our sample into those with and without mechanical complications. In keeping with prior literature [12], mechanical complications were defined as follows. PJK occurred if there was ≥ 10° increase in kyphosis between the upper instrumented vertebrae (UIV) and UIV + 2 on postoperative imaging [13]. Proximal junctional failure (PJF), a common complication of PJK, was defined as ≥ 10° postoperative increase in kyphosis between the UIV and UIV + 2, along with one or more of the following features: fracture of the vertebral body of the UIV or UIV + 1, posterior osseo-ligamentous disruption, or pullout of UIV instrumentation [14]. DJK occurred if there was ≥ 10° increase in kyphosis between lowest instrumented vertebra (LIV) and LIV-1 on postoperative radiographs. A rod fracture was defined as a single or double rod breakage. Lastly, implant failure included any case of screw pullout, breakage, loosening, or dislodgement.

Due to the common co-occurrence of rod fracture with pseudarthrosis, these two diagnoses were grouped together to form a composite outcome. Though both rod fracture and pseudarthrosis can occur independently of one another, we a-priori decided to group these complications together due to their similarity. Pseudarthrosis increases the risk of rod fracture by almost 29-fold [15], and anecdotally, many times pseudarthrosis can be missed in the presence of a rod fracture if not assessed for diligently. Hence, combining rod fracture and pseudarthrosis in the same category was deemed in accordance with previous literature [16,17,18,19].

In addition to the occurrence of a mechanical complication, the time that each complication occurred was recorded in relation to the index surgery. Of note, for the initial baseline analysis, mechanical complication was treated as an outcome. However, for subsequent analysis comparing types of mechanical complication to each other, type of mechanical complication was treated as an independent, exposure variable.

Statistical analysis

Descriptive statistics were compiled for all demographic, preoperative, and postoperative characteristics. Mean ± standard deviation (SD) and median (IQR) for continuous variables and frequency for categorical variables were computed. Continuous data was compared using student’s t test and mean difference (MD) was reported. Nominal data were compared by Chi-square or Fisher’s Exact test, and a one-way analysis of variance tests (ANOVA) was used to differentiate among four groups of continuous variables. The log-rank test was used to differentiate the time to reoperation between PJK and pseudarthrosis/rod fracture groups. Both pseudarthrosis and rod fracture occurred in 23 (15.8%) of patients. As rod fracture commonly occurs following pseudarthrosis [15], and due to the overlapping nature of patients with pseudarthrosis and rod fracture in our sample, patients with either complication were analyzed as a single group. Given the likely event that there was some sort of non-union present with most rod fractures, it was decided a priori to analyze this group as a single group. An alpha level of 0.05 was regarded as statistically significant. To the best of our ability, we included 95% CI with all p values when appropriate to provide more granular details of the data and variance of the data. The analysis was performed using Stata version 14 (StataCrop LP).

Results

Patient sample

A total of 145 patients underwent ASD surgery during the study period with completed 2-year follow-up. An additional 153 patients underwent ASD surgery but did not have complete follow-up, including both x-rays and complete PROMs to the 2-year mark, and were thus excluded from analysis. A baseline comparison of patients with and without 2-year follow-up is presented in Supplementary Information. Of note, those lost to follow-up appeared to have some notable differences, specifically having more males, more comorbidities (COPD, CHF, HTN), more prior fusions, less often fused to the sacrum, and lower TIL.

In patients with 2-year follow-up, mean age was 63.8 ± 11.3 years, 118 (81.4%) were female, and 56 (38.6%) had undergone a previous fusion operation. Median (range) follow-up was 26.5 (24.0–48.7) months. A total of 131 (90.3%) were fused to sacrum, and 113 (77.9%) had iliac/S2 Alar-iliac (S2AI) screws inserted. Mechanical complications occurred in 85 (58.6%) patients. Patients who developed mechanical complications had more comorbidities (p = 0.005) including higher rates of COPD (p = 0.033), and higher NRS-leg pain preoperatively (p = 0.011) compared to patients without mechanical complications. The presence or absence of mechanical complications did not differ in terms of having undergone a prior fusion (p = 0.272), Hounsfield unit average (p = 0.065), total instrumented levels (p = 0.668), or any preoperative radiographic variables. Full demographic and perioperative characteristics are shown in Table 1 and Table 2.

Table 1 Demographics, operative variables, and PROMs between those with and without mechanical complications
Full size table
Table 2 Preoperative radiographic measurements
Full size table

Time to mechanical complications

Of the 85 patients (58.6%) suffering from mechanical complications, 59 (40.7%) required reoperation. Of the 47 patients with PJK, 27 (47.4%) had PJF, and 30 (63.8%) required reoperation at a median of 16.3 months. Of note, 6/30 (20%) of the patients required reoperation within 6 months of the index surgery. Of the 31 patients with pseudarthrosis/rod fracture, 27 (87.1%) underwent reoperation at a median of 18.4 months. Patients with PJK had significantly higher reoperation rate than pseudarthrosis/rod fracture (63.8% vs. 87.1%, Χ2 =  −0.23, 95% CI −0.41, −0.05, p = 0.023). Cox regression analysis assessing the difference between the time to reoperation in pseudarthrosis/rod fracture vs. PJK showed no significant difference (HR = 0.97, 95% CI 0.85–1.11, p = 0.686) (Fig. 1). DJK (N = 3; 2 reoperation) and implant failures (N = 4; 0 reoperations) had low numbers, which limited further analysis. Reoperation rate for each mechanical complication along with time to reoperation are summarized in Table 3.

Fig. 1
figure 1

Graphical display of time to reoperation (months) of PJK and pseudarthrosis/rod fracture

Full size image
Table 3 Mechanical complications and median time to mechanical complication
Full size table

Demographic and preoperative factors

Demographic and preoperative variables for each mechanical complication subtype are compared in Table 4. Among preoperative factors, patients with PJK had significantly lower Hounsfield Units preoperatively (138.2 ± 43.8 vs. 160.3 ± 41.0, MD =  −22.1, 95% CI −41.8, −2.4, p = 0.028) and underwent more prior fusion operations (51.1% vs. 25.8%, Χ2 = 0.253, 95% CI 0.41, 0.46, p = 0.026), compared with patients with pseudarthrosis/rod fracture. Furthermore, though patients with PJK were more frequently fused to the sacrum compared to three other groups (p < 0.001), no significant difference in fusion to sacrum was found between PJK and pseudarthrosis/rod fracture (p = 0.817). Patients with PJK also had significantly fewer instrumented levels compared to patients with pseudarthrosis/rod fracture (9.2 ± 2.6 vs. 10.7 ± 2.5, MD =  −1.5, 95% CI −2.7, −0.31, p = 0.013).

Table 4 Demographics and operative variables by mechanical complication
Full size table

Radiographic factors

No differences were seen in any measured preoperative radiographic values (Table 5). Postoperatively, patients with PJK had higher postoperative thoracic kyphosis (TK) compared to patients with pseudarthrosis/rod fracture (46.3 ± 12.7 vs. 34.9 ± 10.6, MD = 11.4, 95% CI 5.9, 16.9, p < 0.001). A similar but non-significant difference was seen for C7 SVA, where PJK patients had a higher C7 SVA than pseudarthrosis/rod fracture patients (80.7 ± 72.1 vs. 51.9 ± 57.3, MD = 28.8, 95% CI − 1.9, 59.5, p = 0.066), but again, this did not achieve statistical significance. Preoperative and postoperative radiographic measurements are presented in Table 5.

Table 5 Radiographic measurements preoperatively, postoperatively, and change from preoperative to postoperative
Full size table

Preoperative patient-reported outcomes

No differences were seen with regards to preoperative ODI, NRS-Back, NRS-Leg, EQ-5D. Table 6 summarizes preoperative PROMs in different mechanical complication groups.

Table 6 Preoperative PROMs of different mechanical complication groups
Full size table

Discussion

While the ASD literature discusses the importance of mechanical complications at length, to our knowledge, few studies have examined the clinical impact of different types of mechanical complications. First, our study found that patients with mechanical complications had more comorbidities, higher rates of COPD, and higher preoperative NRS-leg pain. In addition, reoperation rates were higher in patients with pseudarthrosis/rod fracture compared to those with PJK, but time to reoperation did not significantly differ between pseudarthrosis/rod fracture and PJK. Furthermore, PJK patients had higher postoperative TK compared to patients with pseudarthrosis/rod fractures. PJK patients also had lower preoperative Hounsfield Units, more often had a prior fusion procedure, and had fewer instrumented levels. Lastly, no difference in preoperative PROMs were found when comparing each mechanical complication with one another. The results of this preliminary study may provide useful information in avoiding and treating mechanical complications after ASD surgery.

Specific distinguishing features existed between each type of mechanical complication, specifically that pseudarthrosis/rod fracture had a higher reoperation rate than PJK, but similar time to reoperation. A meta-analysis of 26 studies reported an average time of 33.0 months to diagnose pseudarthrosis after ASD surgery [20], much longer than our current cohort (mean 21.8 months). Jung et al. [21] found a 9.2% revision rate following rod fractures after ASD surgery, and mean time to rod fracture of 27.3 ± 6.9 months (range 20–42 months), with no differences in demographic variables and radiological parameters when compared to patients without rod fractures. Similar revision rates for rod fractures were found in Yamato et al. [22] where 11.8% of the patients with rod fracture underwent reoperation. Both these values were significantly lower than our study, which may have been due to us grouping pseudarthrosis/rod fracture as one group. With respect to PJK, one study [23] compared the average onset of PJK in distal thoracic versus proximal thoracic surgery and found a significantly earlier onset of PJK in the distal thoracic group (65 ± 14 vs. 560 ± 322 days, p < 0.01). A systematic review performed by Lau et al. [24] indicated that 66% of PJK occurs within 3 months and 80% within 18 months after surgery, with reported revision rates ranging from 13 to 55%. Our reoperation rates for PJK were lower than for pseudarthrosis/rod fracture but higher than previously reported in the literature. Though our time to reoperation for PJK was 16.3 months, 20% of patients required reoperation within 6 months.

PJK patients had lower Hounsfield units, more prior fusions, and fewer instrumented levels than those with pseudarthrosis/rod fracture. While lower bone density is a well-known risk factor for PJK, the number of prior surgeries is not a well-described risk factor for PJK [25]. One potential explanation is that patients had several smaller operations leading to adjacent segment disease or PJK, without addressing the underlying malalignment and lack of distal lumbar lordosis. Moreover, if a patient had PJK once, they may be more likely to have it again during subsequent spine operations. Prior studies have shown that older age, number of fused vertebrae, and thoracic kyphosis > 20° are all considered risk factors for pseudarthrosis [26], and these factors are also closely associated with developing PJK. With respect to rod fractures, some studies suggest that failure to restore optimal sagittal alignment increases the risk of rod fracture [27]. Thus, appropriate correction of sagittal alignment is paramount to preventing pseudarthrosis by reducing stress in the screw/rod construct, and subsequently decreasing the odds of rod fractures [28,29,30].

PJK is a vexing complication even to the most expert deformity surgeons, and controversies exist surrounding the etiology of PJK. Risk factors for PJK include end vertebrae selection, facet violation proximal to UIV, significant soft tissue disruption, and loss of posterior tension band integrity [13, 31]. While some studies found no relationship between PJK and appropriateness of sagittal restoration [26, 32], others have found a higher risk of PJK after a large amount of correction due to potential overcorrection [23, 33]. In our study, although preoperative malalignment did not increase the odds of mechanical complications, higher postoperative TK and C7 SVA was found to be more associated with PJK compared to pseudarthrosis/rod fracture. Of note, postoperative TK and C7 SVA were recorded in the immediate postoperative period before the development of PJK. These results seem to agree with the literature, that persistent deformity, particularly a high postoperative TK, is a risk factor for PJK.

Newer areas to assess spinal alignment may hold promise for further study of mechanical complications. Yilgor et al. [12] found that the Global Alignment and Proportion (GAP) score accurately correlated with the development mechanical complications (AUC = 0.92, CI = 0.85–0.98, p < 0.001). Jacobs et al. [34] showed that the GAP score significantly predicted mechanical complications (p = 0.003) and proved to be superior compared with Schwab classification. Other studies found no association between the GAP score and the incidence of mechanical complications [35, 36]. Our study showed that the LDI—a component of the GAP score—had no difference in type of mechanical complications. While the GAP score may have utility in predicting the occurrence of a mechanical complication, it appeared to provide little assistance in distinguishing types of mechanical complications. The Roussouly classification has also been extensively studied in the context of mechanical complications after ASD surgery. Pizones et al. [37] performed a study on 96 patients with ASD and found that postoperative Roussouly-type mismatch (OR = 41.9; CI = 5.5–315.7; p < 0.001), iliac instrumentation (OR = 19.4; CI = 2.6–142.5; p = 0.004), and age (OR = 1.1; CI = 1.02–1.16; p = 0.004) are the most important predictors of mechanical complications. Whether the Roussouly classification can help predict different types of mechanical complication remains to be seen.

The current study has several limitations that are worth discussing. First, data was collected from a single institution, and sample size was relatively small for each type of mechanical complication. A larger, multi-center dataset may yield differing results as ASD practice patterns may vary between institutions. Second, only preoperative PROMs were retrieved and included in the analysis, as it was not clear in our registry if postoperative PROMs were recorded before or after the occurrence of mechanical complications. It would have been imprecise to compare PROMs between patients who encountered a reoperation with those who did not. Postoperative PROMs are an important component of assessing outcomes of ASD surgery and further studies are warranted to assess differences in PROMs following mechanical complications in ASD surgery. Third, we did not group the patients according to the follow-up period in aim of increasing the number of patients who encountered a mechanical complication or required reoperations with such a tight follow-up period (range 24.0–48.7). We included patients who only had 2-year follow-up. Fourth, we imposed strict inclusion criteria of having completed 2-year X-rays and 2-year PROMs, which led to a high exclusion and lost to follow-up rate of 153 patients. Though some variables were missing from the Supplementary Information, the baseline comparison provided between patients with and without 2-year follow-up showed that patients with 2-year follow-up had more comorbidities, which means that our analysis may have focused on more sick individuals with more comorbidities and undergoing more complex surgeries. Patients that were indeed lost to follow-up appeared to be a healthier, less complex population, which might have created a selection bias with a subsequent decreased generalizability of our findings. Other reasons contributing to the selection bias originate from having a greater proportion of males and revision surgeries in the non-attendant group, Fifth, we did not perform a separate analysis for patients with and without fusion to the sacrum or ilium, as the vast majority of patients (90.3%) being fused to the sacrum, and 77.9% having additional iliac screws below S1 screws. Sixth, we grouped PJK as a single entity regardless of severity, in keeping with prior literature [13, 24, 33, 38, 39, 39, 40], though this may bring forth some inaccuracy in reporting reoperations, as some PJKs are very mild and non-significant, and others can lead to PJF. Seventh, though the use of composite variables is rarely reproducible and difficult to interpret [41], combining pseudarthrosis with rod fracture seemed to be in accordance with previous literature [16, 18, 19], serving the purpose of increasing the sample size while preserving external validity. Eighth, we did not have a way of following patients who chose not to follow-up with their original surgeon. Several patients may have had complications and gone to another surgeon, and we had no way of obtaining this information. Future investigation should include prospective cohort studies with larger sample sizes, to discern further differences and provide additional insights of each type of mechanical complications following ASD surgery.

Conclusion

In a single-center, preliminary study of patients undergoing ASD surgery, those with mechanical complications had more comorbidities, higher rates of COPD, and higher preoperative NRS-leg pain compared to those without mechanical complications. Patients with pseudarthrosis/rod fracture had a higher rate of reoperation compared to patients with PJK, though time to reoperation was similar between both groups. Patients with PJK also had higher postoperative TK, lower preoperative Hounsfield Units, more prior fusions, and fewer instrumented levels compared to patients with pseudarthrosis/rod fracture. Preoperative PROMs were similar between all types of mechanical complications. Taken together, even though mechanical complications are often regarded as a single outcome, several important differences may exist between them.