Overcorrection of fractured vertebrae increases the incidence of adjacent fractures after balloon kyphoplasty: A retrospective study

Abstract

Purpose

To investigate whether the correction amount by balloon kyphoplasty (BKP) is associated with the incidence of adjacent vertebral fractures (AVF).

Methods

Data from 61 patients without and 25 patients with AVF were analyzed. A linear regression model was used between vertebral instability and corrected angle or height to divide patients into the overcorrection and undercorrection groups.

Results

Vertebral fractures overcorrected for instability led to a higher incidence rate of AVF [42.4% and 46.9% with overcorrection vs. 10.7% and 13.3% with undercorrection for angle and height, respectively].

Conclusions

Overcorrection of fractured vertebrae may associate with the increased incidence of AVF.

1. Introduction

Most osteoporotic vertebral fractures are successfully healed by conservative treatment, but there are some cases of nonunion, neurological deficits, remaining high-grade kyphosis.¹ Kyphoplasty has been reported as an effective and safe procedure for patients with osteoporotic vertebral fractures.² Kyphoplasty is less invasive and involves few complications, but adjacent vertebral fracture is a frequently observed problem.³

The prevention of adjacent vertebral fractures is key to improving outcomes. A greater degree of restoration height of the vertebral body and cement leakage into the disc are both reported risk factors for adjacent vertebral fractures.4, 5, 6, 7, 8 These factors can be controlled technically, but these facts do not suggest that the adjustment of the correction contributes to the prevention of adjacent fractures. Balloon kyphoplasty (BKP) is a procedure that restores vertebral height by inserting cement into the vertebral cavity created by inflating a balloon; however, the restoration is more dependent on the postural correction than balloon inflation⁹ and is mainly influenced by the instability of the fractured vertebra. If the instability of the fractured vertebra is severe, then the restoration height and the amount of cement used will be increased. If the vertebral endplates have substantial damage, then more leakage of the cement may occur. It is difficult to judge whether the incidence of adjacent vertebral fracture associates with the surgical procedure of kyphoplasty, whether this is dependent on the condition of the vertebra before treatment, or both.

The purpose of this study is to evaluate whether adjusting the correction of the fractured vertebrae may associate with the incidence of adjacent vertebral fracture. We conducted a retrospective study of patients who underwent kyphoplasty and evaluated the influence of vertebral instability and correction measures on the incidence of adjacent vertebral fractures.

2. Materials and methods

This retrospective analysis included 86 patients treated with BKP at our institution between April 2013 and March 2019. This study was approved by the Kyushu University institutional review board (IRB approval no., 28–324). Written informed consent was not needed due to the retrospective nature of the study. The number of patients was 61 without and 25 with adjacent vertebral fractures. Patients followed for over 6 months were included in this study, and the adjacent fracture was diagnosed at 6 months after BKP. New adjacent fractures were confirmed by evaluating vertebral instability between the sitting and supine positions from X-ray images, a 15% loss in height, and/or changes in bone signal intensity observed with magnetic resonance imaging (MRI, Siemens Medical Solutions, Erlangen, Germany). The fracture sites were located within the area between the tenth thoracic (T10) and the fifth lumbar (L5) vertebrae. Patients with pathological fractures from cancer or fractures accompanied by neurological symptoms were excluded. We also excluded patients who presented with collapsed vertebrae on both sides of a fracture, regardless of whether the fracture had already healed, because new fractures in the collapsed vertebrae are difficult to detect. The duration of time of corset use and bed rest depended on the patient's condition. BKP was performed under general anesthesia, and patients treated with BKP tended to get out of bed the next day with a soft corset.

The kyphosis angle of the fractured vertebra was determined by measuring the slopes of the endplates of the fractured vertebra in the sitting position.⁹ The vertebral height was calculated by adding together the heights of the anterior, middle, and posterior columns of the vertebra in the sitting position.⁹ Vertebral instability was determined by evaluating differences in the kyphosis angle and vertebral height between the sitting and supine positions (Fig. 1A). The corrected kyphosis angle was calculated by measuring differences between the angles of the fractured vertebral endplate before and after BKP in the sitting position (Fig. 1B). The corrected height difference was calculated by measuring differences in the vertebral heights before and after BKP in the sitting position (Fig. 1C). The measurement was performed by two orthopedic surgeons. The leakage of cement was indicated by the presence of cement in the disc space, as determined by X-ray.

Fig. 1.

X-ray films showing methods for measuring the fractured vertebral instability and the corrected amount. A) Vertebral instability (angle) = a-b. Vertebral instability (height) = (f + g + h)-(c + d + e). B) Corrected kyphosis angle = A-B. C) Corrected height = (F + G + I) - (C + D + E).

Correction of the vertebral deformity was defined by the method used linear regression analysis of the fractured vertebral instability and amount of correction to divide patients into two groups. The patients whose data fell above the regression line with 95% confidence interval (CI) were placed in the overcorrection group, and those whose data fell under the line were placed in the undercorrection group. The incidence rate of adjacent fractures was compared between the two correction groups. To clarify the prognostic risk factors for adjacent vertebral fractures, logistic regression was performed if there was a significant difference by univariate analysis.

Comparisons were made by univariate analysis between groups with and without adjacent fracture for age, sex, body mass index, site of fracture, existence of past vertebral fracture, kyphosis angle of the fractured vertebra, fractured vertebral instability, corrected angle and height, cement amount, and cement leakage. The relationship between fractured vertebral instability and the amount of correction was evaluated using Spearman's rank correlation. The chi-square test and Wilcoxon's signed-rank test were used for group comparisons. Continuous values were binarized around the median value when necessary. Spearman's rank correlation and linear regression models were used to assess the relationship between vertebral instability and the amount of vertebral correction. Odds ratios, as determined by logistic regression, were considered significant when the lower limit of a 95% CI exceeded 1.0. The numbers are presented as means ± standard deviations (SDs) in the tables, and P < 0.05 was considered to indicate statistical significance for all tests. All statistical analyses were carried out using JMP® 14 (SAS Institute Inc., Cary, NC, USA).

3. Results

The rate of adjacent vertebral fracture was significantly higher in patients with greater kyphosis angles of the fractured vertebrae, fractured vertebral instability, and corrected angle and height (Table 1). There was no significant difference in age, sex, body mass index, site of fracture, existence of past vertebral fracture, amount of cement used, or cement leakage between patients with and without adjacent fractures (Table 1).

Table 1.

Demographic and clinical data of patients with and without adjacent fractures.

	Adjacent fracture		P value
	Without (n = 61)	With (n = 25)
Age, n (%)			0.982
<80 y	27 (44.3)	11 (44.0)
≧80 y	34 (55.7)	14 (56.0)
Sex, n (%)			0.774
Male	20 (31.7)	9 (36.0)
Female	41 (68.3)	16 (64.0)
Body mass index, n (%)			0.278
18.5–25	39 (64.0)	19 (76.0)
<18.5, ≧25	22 (36.0)	6 (24.0)
History of vertebral fracture, n (%)			0.859
–	28 (45.9)	12 (48.0)
+	33 (54.1)	13 (52.0)
Site of fracture, n (%)			0.32
T11-L1	37 (61.0)	18 (72.0)
T10, L2-5	24 (39.0)	7 (28.0)
Kyphosis angle (°), n (%)			0.044
<20	29 (47.5)	6 (24.0)
≧20	32 (52.5)	19 (76.0)
Cement amount (ml), n (%)			0.4
<8	28 (45.9)	9 (36.0)
≧8	33 (54.1)	16 (64.0)
Cement leak, n (%)			0.108
–	36 (59.0)	10 (40.0)
+	25 (41.0)	15 (60.0)
Vertebral instability [angle (°)], mean ± SD	9.8 ± 5.1	13.5 ± 5.8	0.006
Vertebral instability [height (mm)], mean ± SD	13.2 ± 7.4	18.2 ± 9.0	0.038
Corrected angle (°), mean ± SD	10.8 ± 6.3	17.6 ± 5.5	<0.001
Corrected height (mm), mean ± SD	17.0 ± 8.9	26.3 ± 5.8	<0.001

Spearman's rank correlation showed a positive proportional relationship between vertebral instability and the amount of correction for both angle (ρ = 0.6447, P < 0.001) and height (ρ = 0.4909, P < 0.001). In this analysis, the data from three patients were removed because the X-ray images were unclear, and we were unable to measure the correction amount. To determine the over and undercorrection, patients between the CI of the regression line were removed from the analysis. The remaining 61 patients for angle and 62 patients for height were investigated for incidence rate of adjacent fracture. Patients above the regression line with CI were determined as having an overcorrection while those under were determined to be undercorrected (Fig. 2). The rates of adjacent fractures were 42.4% (14/33) and 10.7% (3/28) for overcorrected and undercorrected angles, respectively (P = 0.009, Table 2), while the rates of adjacent fractures were 46.9% (15/32) and 13.3% (4/30) for overcorrected and undercorrected height, respectively (P = 0.006, Table 2) (Fig. 3).

Fig. 2.

The definition of correction based on linear regression. A) model for the corrected angle versus vertebral instability, B) model for corrected height versus vertebral instability. ●, patient without adjacent fractures; × , patient with adjacent fractures. Straight line is a regression line, and dotted lines are 95% confidence interval. Overcorrection – patients with data above the upper dotted line; undercorrection – patients with data under the lower dotted line.

Table 2.

Univariate analysis of the correction amount based on the linear regression line.

	Adjacent fracture		P value
	Without (n = 44)	With (n = 17)
Correction [angle], n (%)			0.009
Under	25 (56.8)	3 (17.6)
Over	19 (43.2)	14 (82.4)
	Without (n = 43)	With (n = 19)	P value
Correction [height], n (%)			0.006
Under	26 (60.5)	4 (21.1)
Over	17 (39.5)	15 (78.9)

Fig. 3.

X-ray image showing over- and undercorrected vertebra. An example of a corrected vertebra that fell above the regression line (i.e., overcorrected) and one that fell under the regression line (i.e., undercorrected). The overcorrected vertebra shows the adjacent fracture under the fractured vertebra at 6 months after balloon kyphoplasty (BKP). The undercorrected vertebra shows the remaining instability of the fractured vertebra after BKP. This instability disappeared at 6 months after the BKP.

Overcorrection of the kyphosis angle and height was detected as a risk factor for adjacent vertebral fractures in the linear regression model. Logistic regression analysis was performed with risk factors detected by univariate analysis for corrected angle (Table 3) and corrected height (Table 4). For these analyses, the continuous value of vertebral instability was binarized around the median value. Patients with overcorrected angle and height had a 5.59 (95% CI 1.40–30.28, P = 0.014) and 5.07 (95% CI 1.50–20.50, P = 0.013) times higher likelihood of adjacent vertebral fractures, respectively, than patients with undercorrected angle and height (Table 3, Table 4).

Table 3.

Statistical analysis using a logistic regression model for corrected angle.

	Adjacent fracture		Univariate	Multivariate
	Without	With	P value	OR (95% CI)	P value
Kyphosis angle (°)			0.034	1.852 (0.350–10.758)	0.463
<20	27	5
≧20	32	19
Vertebral instability [angle] (°)			0.008	2.868 (0.670–14.973)	0.158
<10	31	5
≧10	28	19
Overcorrection [angle]			0.009	5.593 (1.396–30.284)	0.014
Without	25	3
With	19	14

OR odds ratio, CI confidence interval.

Table 4.

Statistical analysis using a logistic regression model for corrected height.

	Adjacent fracture		Univariate	Multivariate
	Without	With	P value	OR (95% CI)	P value
Kyphosis angle (°)			0.034	2.129 (0.606–8.237)	0.249
<20	27	5
≧20	32	19
Vertebral instability [height] (mm)			0.119	1.187 (0.356–4.021)	0.780
<15	38	11
≧15	21	13
Overcorrection [height]			0.006	5.069 (1.502–20.504)	0.013
Without	26	4
With	17	15

OR odds ratio, CI confidence interval.

4. Discussion

In this study, we report that an overcorrected fractured vertebra is a risk factor for adjacent vertebral fractures. This overcorrection contributed to the incidence of adjacent fractures even when the corrected amount was evaluated in association with the vertebral instability. These data suggest that overcorrection of fractured vertebrae associates with the incidence of adjacent fracture after BKP.

Several studies reported that the amount of restoration of the fractured vertebrae was one of the risk factors for adjacent fractures.4, 5, 6, 7, 8 However, the method of measurement of the restoration amount varied between different studies. Some studies measured restoration amount by the angle change of the fractured vertebral endplate before and after BKP, while others compared the height of the fractured vertebra after BKP to the height of the adjacent vertebrae. Additionally, posture at the time of measurement was not mentioned in most cases. The association between vertebral instability and the degree of correction depends on the method of measurement, and hence, we clearly mention that we measured the amount of restoration by measuring differences between the angles and height of the fractured vertebra before and after BKP in the sitting position.

We determined the correction by the relationship between fractured vertebral instability and the corrected amount using regression analysis. Separating patients according to the regression line was considered an objective reference point for determining overcorrection and undercorrection. Taking into consideration the error of measurement, we removed the patients within the CI of the regression line and compared them by two standards (angle and height). In the methods used for determining overcorrection and undercorrection, we observed a significant difference in the incidence of adjacent fractures in the group with the overcorrected amount compared to that with the undercorrected.

The incidence rate of adjacent vertebral fractures after BKP is controversial. We believe this controversy stems from the diversity of indication for the use of kyphoplasty for osteoporotic vertebral fracture. Randomized controlled trials reported no significant differences in incidence rates for adjacent fractures between kyphoplasty and conservative treatment^10,11; however, these trials enrolled patients with relatively acute phase fractures where vertebral collapse had not progressed, and the endplates were less damaged. We performed BKP only when conservative nonsurgical treatment failed; hence, the differences in indication between our cases and those reported in the clinical trials may have produced a different rate of adjacent fractures.

One of the limitations of our study is that we could not directly show the results of injecting differing quantities of restoration cement because our study was retrospective, and the surgeons performed optimum surgery in all cases. To verify the influence of the correction, we would need to compare the incidence rate of adjacent fracture by changing the correction amount between patients who had the same condition. However, we could not perform a study of this type due to the diverse nature of osteoporotic fractured vertebrae, and ethical considerations would likely not allow a study of this kind. We cannot deny the possibility that the corrected amount data were influenced by fractured vertebral conditions not evaluated by vertebral instability, such as disc conditions or endplate damage, in addition to the BKP procedure. Furthermore, lack of other important factors, such as osteoporosis, normalization in the after-treatment period, and the period of follow-up, may have influenced the results. We could not include the severity of osteoporosis for this analysis due to the lack of data. The adjacent fractures occurred within approximately 2 months post-surgery; therefore, we do not believe that the after-treatment (e.g., anti-osteoporotic drugs) influenced the incidence rate of adjacent fractures. The incidence rates of adjacent fractures after 2 months were similar to those of the conservative treatments in our studies. Therefore, we believe that the follow-up period of 6 months was sufficient to investigate complications associated with BKP.¹²

Our results on the amount of correction were generated by comparing X-ray images in the loading position both before and 1 week after surgery. As patients were in the prone position during surgery, the fractured vertebra spread without cementing¹¹; thus, we could not predict the corrected amount in the loading position when performing BKP. Therefore, we could not present an appropriate reference for moderate restoration during surgery to prevent adjacent fractures. However, from our results, performing excessive correction to restore vertebral height appears to be unnecessary because there were no problems caused by the correction loss or insufficient cement stability in the undercorrected group and the remaining instability did not influence the clinical outcome.^13,14

Our results showed that preventing overcorrection may contribute to a reduction in the incidence rate of adjacent vertebral fractures. The fractured vertebral conditions, fractured vertebral instability and kyphosis angle, were also identified as risk factors in the univariate analysis. In cases with high kyphotic angle and instability, there is a significant risk of adjacent fracture associated with BKP. Therefore, surgical procedures such as instrument surgery could be desirable to prevent adjacent fractures.

In summary, a higher correction angle and height were detected as a risk factor even when we evaluated the corrected amount based on vertebral instability. Preventing the overcorrection of fractured vertebrae may reduce the incidence of adjacent fracture when BKP is performed.

Authors’ contributions

KI contributed to the conception and design of the study, acquisition of the data, analysis and interpretation of the data, and drafting the manuscript. HK and TS aided in the acquisition of the data. KH participated in the design and coordination and helped in drafting the manuscript. All authors read and approved the final manuscript.

Declaration of competing interest

The Authors declare that there is no conflict of interest.