Percutaneous Vertebral Augmentation for Acute Traumatic Vertebral Fractures: A TQIP Database Study

Abstract

Patients with vertebral fractures may be treated with percutaneous vertebroplasty (VP) and kyphoplasty (KP) for pain relief. Few studies examine the use of VP and KP in the setting of an acute trauma. In this study, we describe the current use of VP/KP in patients with acute traumatic vertebral fractures. All patients in the ACS Trauma Quality Improvement Program (TQIP) 2016 National Trauma Databank with severe spine injury (spine AIS ≥3) met inclusion criteria, including patients who underwent PVA. Logistic regression was used to assess patient and hospital factors associated with PVA; odds ratios and 95% confidence intervals are reported. 20,769 patients met inclusion criteria and 406 patients received PVA. Patients aged 50 or older were up to 6.73 (2.45 – 27.88) times more likely to receive PVA compared to younger age groups and women compared to men (1.55 [1.23–1.95]). Hospitals with a Level II trauma center and with 401–600 beds were more likely to perform PVA (2.07 [1.51–2.83]) and (1.82 [1.04–3.34]) respectively. African American patients (0.41 [0.19–0.77]), isolated trauma (0.64 [0.42–0.96]), neurosurgeon group size > 6 (0.47 [0.30–0.74]), orthopedic group size > 10, and hospitals in the Northeastern and Western regions of the U.S. (0.33 [0.21–0.51] and 0.46 [0.32–0.64]) were less likely to be associated with PVA. Vertebroplasty and kyphoplasty use for acute traumatic vertebral fractures significantly varied across major trauma centers in the United States by multiple patient, hospital, and surgeon demographics. Regional and institutional practice patterns play an important role in the use of these procedures.

INTRODUCTION

Percutaneous vertebral augmentation (PVA) is a minimally invasive procedure used for pain relief in the treatment of osteoporotic vertebral fracture and is subdivided into percutaneous vertebroplasty (VP) and kyphoplasty (KP). VP is performed with the injection of medical cement, polymethylmethacrylate (PMMA) or calcium phosphate into the fractured vertebral body.^6,14 KP consists of inserting a cannula into the vertebral body, and an inflatable bone tamp is used to move the vertebral body back to its original height, allowing a cavity to be created to be filled with bone cement. Advantages of KP include restoration of vertebral body height and correction of kyphosis as compared to VP; however clinical outcome profiles between the two procedures are very similar.¹

Several randomized controlled trials have investigated the clinical utility of VP and KP in osteoporotic and neoplastic pathologic vertebral compression fractures, with some studies indicating a clinical benefit to PVA compared to sham treatment in patients that fail conservative therapy. The VERTOS I, VERTOS II, and VAPOUR trials identified VP as a safe and effective treatment for patients with acute osteoporotic vertebral compression fractures and persistent pain.^4,15,22 However, on the contrary, the INVEST, VERTOS IV, and the Buchinder et al sham trials were unable to prove significant differences in back pain intensity, overall pain, perceived improvement, functional disability, or quality of life between VP and conservative therapies.^2,7,14 With respect to KP, the multicenter, international Fracture Reduction Evaluation (FREE) trial determined KP to be an effective and safe procedure for patients with acute vertebral compression fractures.²³ Smaller case series have suggested that PVA can also be used for the treatment of painful burst fractures, correction of kyphosis, and in conjunction with fusion and instrumented fixation to improve stiffness and pullout strength.^3,8,21,24

PVA, through the use of VP and KP, was deemed by the American Association of Neurological Surgeons, Congress of Neurological Surgeons, American Society of Spine Radiology, Society of NeuroInterventional Surgery, among others as a safe, efficacious, and durable procedure in appropriate patients with osteoporotic and neoplastic fractures in 2014.¹ Between 1993 and 2004 there was a rapid rise of PVA use, outpacing the rate of increase in other major orthopedic procedures.^17,19 Similarly, between 2004 and 2008 there was a noted increase in PVA usage with KP significantly outpacing VP.^10,19 However, after the negative trials involving PVA sparked controversy, there was a noted decline in annual volume of PVA in 2009, but the overall frequency of PVA in the Medicare population did not change significantly between 2005 and 2015.²⁰ Most previous studies^10,17–20 have explored vertebroplasty and kyphoplasty use in the setting of osteoporotic or neoplastic pathologic fractures, but little has been published on the use of PVA in acute traumatic compression and burst fractures.^12,18 In this study, we look to investigate the current state of practice of PVA in traumatic vertebral compression and burst fractures as an isolated procedure.

MATERIAL AND METHODS

Patient selection

The American College of Surgeons (ACS) Trauma Quality Improvement Program (TQIP) database was retrospectively reviewed for all patients from January 1, 2016 through December 31, 2016 across 463 participating facilities. All ages were included.

Inclusion criteria included all patients in the TQIP database with severe spine injury who had vertebral fractures and underwent either vertebroplasty or kyphoplasty. Exclusion criteria was defined as patients who did not undergo VP or KP, patients who had vertebral fracture secondary to neoplasm or non-traumatic osteoporosis, and patients who had VP or KP in conjunction with spinal fusion, spinal stabilization, or spinal instrumentation.

Severe spine injury was defined as a spine Abbreviated Injury Severity (AIS) ≥ 3. Patients with vertebral fractures were identified according to their associated ICD-9 or ICD-10 diagnostic codes. Patients who underwent percutaneous vertebral augmentation were then identified by the associated vertebroplasty or kyphoplasty ICD-9 or ICD-10 procedural codes. Spinal fusion, stabilization, and instrumentation were also identified by their corresponding ICD-9 or ICD-10 procedural codes.

Data collection

Of the patients with severe spine injury who underwent PVA, we identified baseline patient demographics, noted as gender, age, race, and insurance payer (Table 1). Clinical patient demographics such as Glasgow coma scale (GCS), polytrauma, injury severity score (ISS), and discharge disposition were described as well (Table 1). Hospital demographic information included United States region (Midwest, Northeast, South, and West), neurosurgeon group size, orthopedic surgeon group size, teaching status (University, Community, or Non-Teaching), trauma level, and hospital size (Table 2). Injury morphology, defined as compression fracture or burst fracture, was identified by the associated ICD-10 diagnostic codes (Table 3).

Table 1:

Patient Demographics

		Patients who received PVA n = 406 (%)	Odds Ratio (95% CI) C-Stat = 0.866
Age:	16–19	3 (0.7%)	Reference
	20–29	7 (1.7%)	0.89 (0.24 – 4.18)
	30–39	8 (2.0%)	1.18 (0.34 – 5.45)
	40–49	13 (3.2%)	1.70 (0.53 – 7.52)
	50–59	50 (12.3%)	4.00* (1.44 – 16.65)
	60–69	77 (19.0%)	5.75* (2.09 – 23.75)
	70–70	91 (22.4%)	6.36* (2.30 – 26.36)
	80–89	122 (30.0%)	6.73* (2.45 – 27.88)
	≥ 90	35 (8.6%)	4.44* (1.51 – 18.97)
Gender:	Male	148 (36.5%)	Reference
	Female	258 (63.5%)	1.55* (1.23 – 1.95)
Race:	Caucasian	365 (89.9%)	Reference
	African American	10 (2.5%)	0.41* (0.19 – 0.77)
	Other	31 (7.6%)	0.91 (0.60 – 1.34)
Insurance:	Gov	282 (69.5%)	Reference
	Private	109 (26.8%)	0.88 (0.67 – 1.16)
	Other	15 (3.7%)	0.88 (0.48 – 1.50)
GCS:	≤ 8	3 (0.7%)	Reference
	≥ 9	357 (87.9%)	2.26 (0.83 – 9.26)
Polytrauma:	Isolated Trauma	372 (91.6%)	Reference
	Polytrauma	34 (8.4%)	0.64* (0.42 – 0.96)
ISS:	Median	9	0.90* (0.88 – 0.93)
	IQR	4

Results shown as Odds Ratio (95% CI).

^*indicates p < 0.05. Regression Strength shown as concordance statistic (C-Stat). Abbreviations: GCS – Glasgow Coma Scale; ISS – Injury Severity Score

Table 2:

Hospital Demographics

		Patients who received PVA n = 406 (%)	OR (95% CI) C-Stat = 0.866
Region:	Midwest	145 (35.7%)	Reference
	Northeast	27 (6.7%)	0.33* (0.21 – 0.51)
	South	180 (44.3%)	1.12 (0.86 – 1.46)
	West	54 (13.3%)	0.44* (0.30 – 0.62)
Neurosurgeon Group Size:	1–2	57 (14.0%)	Reference
	3–5	244 (60.0%)	0.86 (0.59 – 1.27)
	≥ 6	105 (25.9%)	0.47* (0.30 – 0.74)
Orthopedic Surgeons:	1–4	73 (18.0%)	Reference
	5–6	90 (22.2%)	0.71 (0.50 – 1.02)
	7–9	106 (26.1%)	0.78 (0.55 – 1.10)
	10–15	100 (24.6%)	0.55* (0.39 – 0.78)
	> 15	37 (9.1%)	0.61* (0.38 – 0.95)
Teaching Status:	University	120 (29.6%)	Reference
	Community	184 (45.3%)	0.98 (0.71 – 1.35)
	Non-teaching	102 (25.1%)	1.12 (0.75 – 1.68)
Trauma Level:	Level I	144 (35.5%)	Reference
	Level II	239 (58.9%)	2.07* (1.51 – 2.83)
	Level III	9 (2.2%)	1.79 (0.56 – 4.71)
	Level 0†	14 (3.4%)	2.74* (1.31 – 5.36)
Hospital Size:	≤ 200 Beds	19 (4.7%)	Reference
	201–400 Beds	124 (30.5%)	1.15 (0.68 – 2.04)
	401–600 Beds	153 (37.7%)	1.82* (1.04 – 3.34)
	> 600 Beds	110 (27.1%)	1.01 (0.57 – 1.90)

Results shown as Odds Ratio (95% CI).

^*indicates p < 0.05.

^†Level 0 trauma indicates no ACS verification or state designation. Regression Strength shown as concordance statistic (C-Stat).

Table 3:

Injury Morphology

	Patients who received PVA n = 406 (%)	OR (95% CI) C-Stat = 0.866
Neither Burst nor Compression Fracture	172 (42.4%)	Reference
Burst Fracture:	37 (9.1%)	1.89* (1.24 – 2.80)
Compression Fracture:	190 (46.8%)	5.06* (3.99 – 6.44)
Both Burst and Compression Fractures:	7 (1.7%)	3.14* (1.20 – 6.77)

Results shown as Odds Ratio (95% CI).

^*indicates p < 0.05. Regression Strength shown as concordance statistic (C-Stat).

Several allowances had to be made when considering the variables available in the 2016 TQIP NTDB. The 2016 TQIP NTDB considers all patients aged 90 years or greater to be of equivalent age for patient privacy protection purposes. Patients were thereby stratified by decile in order to allow for meaningful comparisons between this age group and others. Additionally, the 2016 TQIP NTDB contains seven primary racial identification values, some of which contained too few members to support analysis. Patients were summarily recategorized into Caucasian, African American, and Other categories. The Other category itself consists as patients who were primarily identified as Asian, Native Hawaiian or Other Pacific Islander, Other Race, Not Known/Not Recorded, or Not Applicable. Similarly, the nine insurance status values possible in the 2016 TQIP NTDB were consolidated to enable meaningful analysis. Government insurance refers to patients labeled as “Medicare”, “Medicaid”, or “Other Government Insurance”. Private insurance refers to patients labeled as “Private/Commercial Insurance” or “Self-Pay”. Other insurance refers to patients labeled as “Other”, “Not Applicable”, “Not Known/Not Recorded”, or “Not Billed”. Finally, Trauma Level was categorized to prioritize ACS-verified levels over state-designated levels. A center labeled Level I by ACS and Level II by a state is thereby considered to be Level I while a center labeled Level II by ACS and Level I by a state is considered to be Level II for our analysis.

Statistical analysis

Descriptive statistics were used to define the proportions of each variable. Logistic regressions were used to evaluate the association between baseline patient demographics, patient clinical demographics, and hospital demographics and PVA. Regressions were adjusted for gender, age by decile, race, insurance payer, GCS, polytrauma, ISS, and fracture type as well as region, teaching status, hospital size as determined by total bed number, trauma level, hospital neurosurgeon group size, and hospital orthopedic surgeon group size. Regression strength is reported as concordance statistic (C-stat) and regression results are reported as odds ratios (OR) and 95% confidence intervals.

Analysis of this de-identified dataset was deemed exempt by our Institutional Review Board. Statistical analysis was performed using R, version 4.0.1 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Patient Demographics

We identified 406 patients with acute traumatic spine fractures undergoing PVA. Most of the patients were female (63.5%). Isolated lumbar fractures (127, 31.3%), and isolated thoracic vertebral fractures (86, 21.2%) were the most common vertebral level. Spinal fractures involving both the thoracic and lumbar regions were the least common (21, 5.2%). Government insurance (282, 69.5%) was the most common insurance type and approximately 92% of the patient cohort experienced isolated trauma and the majority had a GCS greater than 9 (Table 1).

Hospital Demographics

PVA was most frequently performed in hospitals in the Southern United States (44.3%), hospitals containing between 400–600 beds (37.7%), as well as community hospitals (45.3%) and Level II trauma centers (58.9%). Neurosurgery practices consisting of 3–5 surgeons performed these procedures more often than orthopedic surgeons of any practice size (Table 2). Approximately half of the patients in this cohort were discharged home (47.8%), and the other half were discharged to skilled nursing facilities (125, 30.8%), rehab (75, 18.5%), or other/unknown (9, 2.2%). Three patients died after the procedure (Table 4).

Table 4:

Discharge Disposition

	Patients who received PVA n = 406 (%)
Home	194 (47.8%)
Rehab	75 (18.5%)
Skilled Nursing Facility	125 (30.8%)
Deceased	3 (0.7%)
Other/Unknown	9 (2.2%)

Predictors of PVA

Grouped in deciles, patients aged 50 or older (4.00 [1.44–16.64], 5.75 [2.09–23.36], 6.36 [2.30–26.36], 6.73 [2.45 – 27.88], 4.44 [1.51 – 18.97] more likely to receive KP or VP compared to patients in younger age groups as well as women compared to men (1.55 [1.23–1.95]). Hospitals with a Level II trauma center were more than twice as likely to treat patients with KP or VP who experienced vertebral compression fractures (2.07 [1.51–2.83]). The size of the hospital was also a positive predictor for receiving KP or VP with patients at hospitals with 401–600 beds almost two times as likely compared to hospitals of other sizes (1.82 [1.04–3.34]). Compared to other vertebral fractures, patients with burst and compression fractures (1.89 [1.24–2.80], 3.14 [1.20–6.77]) were more likely to receive VP or KP, with compression fractures having the highest odds at 5.06 [3.99–6.44] (Table 3).

African American patients were less likely to receive PVA in the trauma setting compared to other races (0.41 [0.19–0.77]). Those who experienced isolated trauma were also less likely to undergo VP or KP than those with polytrauma (0.64 [0.42–0.96]). Patients who reside in the Northeastern and Western portion of the United States were less likely to receive PVA compared to patients living in other regions (0.33 [0.21–0.51] and 0.46 [0.32–0.64]) respectively. Surgeon group size predicted lower odds of receiving KP or VP with practices consisting of greater than six neurosurgeons being 0.47 times less likely, and orthopedic surgeon practices between 10–15 surgeons and greater than or equal to 15 surgeons were 0.55 times and 0.61 times less likely to perform PVA, respectively (0.47 [0.30–0.74], 0.55 [0.39–0.78], and 0.61 [0.38–0.95]) (Table 3). Teaching status, GCS, and insurance type did not statistically impact the likelihood of patients receiving KP or VP (Table 3).

DISCUSSION

Our study utilizes a large national trauma database to evaluate the clinical factors associated with PVA in the setting of a hospitalization for acute trauma. We demonstrate that there is great variability in the use of PVA in traumatic vertebral fractures. Variations exist based on patient specific demographics, hospital demographics, and injury characteristics. To our knowledge, this is the only multi-center trauma study to investigate the use of PVA in the setting of an acute trauma. Prior randomized control trials were performed in patients with osteoporotic and neoplastic compression fractures outside of any acute trauma hospitalization and thus little is known about use of PVA for acute traumatic compression and burst fractures.^5,9,11,13,16.

This study further suggests that PVA is more frequently performed at smaller hospitals with fewer neurosurgeons and orthopedic surgeons on staff and tended to be performed on patients with isolated spine trauma as opposed to polytrauma. Patients with more complex polytrauma are likely transferred to larger level I trauma centers for management of their acute medical concern, whereas isolated spine trauma may be adequately managed at smaller hospitals. Following discharge from the hospital, biomechanically stable compression and burst fractures are likely managed on an outpatient basis at community hospitals and clinics geographically close to the patient’s home.

Compression and burst fractures were both found to have higher rates of PVA use than other fractures (AO spine B and AO spine C), as these higher-grade fracture patterns typically require instrumented fusion as opposed to PVA. There are some reports of VP and KP as a form of anterior column support in conjunction with open surgical fixation.^3,8,21,24 Our study excluded patients who received instrumented fusion/fixation surgeries, however, follow-up studies should evaluate the frequency at which PVA is utilized for additional anterior column support.

Using the National Inpatient Sample (NIS), Lad et al describe a 12,900% increase in the use of PVA between 1993 and 2004. They also describe that approximately 60% of patients were female and aged 65 to 84 years, while PVA was performed mostly at large-sized hospitals and hospitals located in the southern United States. They also go on to describe that most patients were discharged home.¹⁷ These findings are similar to our study as we found that 80% of our patients were over the age of 60, and approximately 60% of our patients were female. Additionally, the southern region of the United States was where PVA was primarily performed for acute traumatic vertebral fractures. However, in our study, 60% of our patients received PVA at middle-sized hospitals, ranging from 201–600 beds (68.2%).

Leake et al used NIS data to track trends in PVA and found a 741% increase in PVA between 2001 through 2008.¹⁹ Here they found significant differences between VP and KP, including length of hospital stay and discharges to long-term facilities. Similar to Lad et al and our study, Leake et al found that most patients were female and aged 60 and above. Rabei et al investigated PVA use through The Centers for Medicare and Medicaid Services annual Medicare Physician Supplier Procedure Summary database, finding increased annual claims for PVA in the Medicare population between 2005 and 2015.²⁰ Although our study does not differentiate between different forms of government insurance, we found that government insurance was the most used insurance form for our patient population.

Although the aforementioned studies identified increasing trends in PVA usage from 1993–2015, these databases include both traumatic and non-traumatic etiologies.^17,19,20 The use of PVA in the setting of trauma has been described by Joseph et al, in a single-institution study who found that PVA for patients who had traumatic fractures experienced significant pain relief and was a safe and efficacious treatment to manage painful VCF.¹³

Overall, our study suggests that VP and KP for traumatic vertebral compression patients occur outside of the setting of polytrauma, in smaller community hospitals for compression and burst fractures. There still exists a need for the use of PVA in traumatic vertebral compression fractures as current usage relies on studies that were primarily for osteoporotic and neoplastic indications. Such information may be useful to spine surgeons who manage patients with compression or burst fracture and can model their practice of vertebral augmentation based on current community practices. Further studies should include prospective clinical studies to evaluate outcomes in a primarily trauma population following use of vertebral augmentation.

Limitations

Our study has limitations, however. The use of a retrospective data set allows us to only identify associations between factors rather than causation. The TQIP dataset is also limited to certain data secondary to the Healthcare Insurance Portability and Accountability Act, for example patients older than 89 were all coded as having an age of 90 years, which can influence regression. National databases are also limited by the list of variables, which does not allow us to investigate granular data to better understand the hospital course of each patient. Lastly, we relied heavily on ICD diagnostic and procedural coding to make associations. This is largely dependent on how accurately the clinician and hospital systems coded for these variables.

CONCLUSION

In 2016, vertebroplasty and kyphoplasty use for traumatic vertebral fractures significantly varied by gender, race, injury severity score, injury type, region, hospital size, trauma level, neurosurgeon and orthopedic group size, amongst other factors. Regional and institutional practices may play an important role in the use of these procedures. Future prospective studies should explore the efficacy of PVA in the setting of acute trauma.