Volume 19, Issue 2 , Pages 225-231, February 2008
New Vertebral Osteoporotic Compression Fractures after Percutaneous Vertebroplasty: Retrospective Analysis of Risk Factors
Article Outline
Purpose
To investigate risk factors for new vertebral compression fractures (VCFs) after vertebroplasty.
Materials and Methods
The authors analyzed the occurrence of new VCFs in 70 patients who had previously undergone vertebroplasty for the treatment of one VCF. The following covariates were analyzed: age, sex, body weight, height, body mass index (BMI), treated vertebral level, relative distance between treated vertebrae and new VCFs, pre-existing untreated VCFs, gas-containing vertebrae before treatment, and surgical approach. Surgical variables, including cement leakage into the disk, anterior vertebral height restoration, and kyphosis correction of treated vertebrae were also analyzed. A Cox proportional hazards regression analysis was used to determine the relative risk of new adjacent VCFs. The Kaplan-Meier method was used to calculate mean fracture-free rate over time.
Results
Seventy patients were reviewed, with a mean follow-up of 20.0 months ± 10.2 (range, 6–48 months). We identified 22 new fractures in 19 of the 70 patients (27%), with 16 adjacent and six nonadjacent VCFs. The mean time to new fracture was 10.6 months ± 9.5, and there was no significant difference in time to adjacent or nonadjacent VCF. Increased risk of VCF was associated with proximity to the treated vertebra, greater kyphosis correction, and low patient BMI. The 1-year fracture-free rate was 79.5%.
Conclusions
New VCFs are common in patients with a low BMI, which suggests osteoporosis as a mechanism of fracture.
Abbreviations: BMI, body mass index, ROC, receiver operating characteristic, VCF, vertebral compression fracture
PERCUTANEOUS vertebroplasty is used in the treatment of painful vertebral compression fractures (VCFs), as has been described (1, 2). Complications of vertebroplasty are uncommon, with reported rates ranging from 1%–3% in osteoporotic fractures (1) to 10% in the treatment of metastatic lesions (3, 4). Recently, new VCFs after vertebroplasty have been noted as potential late sequelae (5, 6, 7, 8, 9, 10, 11). Because of the occurrence of new VCFs soon after surgery and because new VCFs are prone to occur in adjacent vertebrae, a reasonable hypothesis is that new VCFs are a result of the vertebroplasty procedure itself.
Multiple covariate analysis has been used to identify risks factors for new VCFs (6, 7, 8, 9, 10, 11). Cement leakage into the vertebral disk after the procedure (7, 10) or the presence of preexisting VCFs (11) have been reported as predictors. The effect of other potentially important covariates on risk of new fracture, however, has not been evaluated. Covariates that remain to be investigated include low body mass index (BMI), which is associated with osteoporosis, gas-containing vertebrae, surgical approach to the pedicle (uni- or bilateral), anterior vertebral height restoration, and the degree of kyphosis correction in adjacent treated vertebrae. The purpose of this study was to evaluate the relationship between new VCFs and these potential covariates and to assess whether new VCFs are a result of vertebroplasty itself or of the disease process that led to the original vertebroplasty.
Materials and Methods
Patients
All patients signed an informed consent form prior to the original evaluation for surgery. We retrospectively reviewed 86 patients treated with one percutaneous vertebroplasty for VCF in our institution between November 2001 and February 2005. Pre-vertebroplasty radiologic evaluation of the patients included plain radiography and magnetic resonance (MR) imaging for all patients.
Patients were evaluated for vertebroplasty if they had acute and severe vertebral fracture pain associated with osteoporosis that failed to respond to medical treatment. The following MR imaging features indicative of acute fracture activity were part of the inclusion criteria: (a) acute vertebral marrow edema with low signal intensity on T1-weighted MR images, high signal intensity on T2-weighted images, and obvious enhancement after contrast medium injection; or (b) osteonecrosis with air or fluid collection in the vertebral body. MR imaging was also used to rule out other spinal diseases including infection or malignancy. If a pathologic fracture was suspected, vertebral body biopsy was performed before vertebroplasty was undertaken.
Exclusion criteria for vertebroplasty included the following: (a) compromise of the spinal canal by retropulsed fragments visualized at medical imaging and accompanied by neurologic signs or (b) collapse of the vertebral body with a residual height less than 10% of the expected value, so that it would be difficult to place a needle in the vertebral body.
Vertebroplasty Technique
The vertebroplasty procedure was performed according to an established technique (1). Patients were placed in the prone position on the examination table. The procedure was performed with the patient under intravenous conscious sedation with 25-mg diazepam (Dupin; China Chemical and Pharmaceutical, Taiwan) and 15–30 mg of codeine (National Bureau of Controlled Drugs, Department of Health, Taiwan) for pain control, with 25-mg meperidine (National Bureau of Controlled Drugs) administered intravenously if codeine was insufficient for pain control. To determine the optimal approach for injection of cement, an 11-gauge bone marrow biopsy needle (Hakko Electric Machine Works, Nagano, Japan) was used to puncture the collapsed vertebral body through either of the pedicles, and the needle was advanced to the anterior third of the vertebral body with biplane fluoroscopic guidance.
Intraosseous venography with 0.5–2 mL of iohexol contrast medium (Omnipaque; Amersham Health, Cork, Ireland) was performed by using a bone biopsy needle under direct biplane fluoroscopic visualization before the injection of the cement. This was done to determine whether there was a risk of cement leakage into the disk. We typically observed opacification of the bony trabeculae while visualizing surrounding veins or leakage of contrast medium into the spinal canal and disk space. If venography showed only opacification of the ipsilateral bony trabeculae with direct and fast venous communication, the contralateral pedicle approach was used.
Bone cement was prepared by mixing the copolymer powder with sterile barium sulfate (9:1 by weight), followed by the addition of a monomer polymerization liquid (OsteoBond; Zimmer, Warsaw, Ind). The cement was injected into the vertebral body under direct fluoroscopic control, and the procedure was immediately terminated if cement reached the posterior fourth of the vertebral body or migrated to draining veins. We used approximately 5–6 mL of cement per vertebra, although we do not have a high degree of confidence in this measure, as there is a potential for cement to be lost during the injection procedure. After the procedure, plain radiographs of each treated vertebral level were assessed to characterize the deposition of cement.
Data Collection
Eighty-six patients underwent vertebroplasty at our institution during the study period. Patients without available radiographs, with radiographs of poor quality, or with radiographs showing malpositioning of the patient were excluded from the study before analysis by a trained neuroradiologist.
Sixteen patients were excluded from the study for the following reasons: 10 patients were treated for more than one VCF, four patients lacked radiographs that were adequate for analysis, and two patients were lost to follow-up. Because we excluded patients who underwent more than one vertebroplasty, it is possible that our results are biased to patients with less severe disease.
Patient demographics, including age, sex, body weight, body height, and BMI, were recorded at the time of surgery. The duration of follow-up was calculated at review. Parameters related to vertebral body treatment were considered, including vertebral level(s) of VCF, presence of older untreated compression fractures, gas content in the vertebra before treatment, and whether the surgical approach was through the pedicle (uni- or bilateral). Vertebrae were categorized into two groups: those at the thoracolumbar junction (T10 through L2) and those outside the thoracolumbar junction (T4 through T9 or L3 through L5). The presence of an intraosseous vacuum cleft containing gas in the vertebrae was recorded before treatment. Gas-containing vertebrae were defined as “bony cleft” if the radiographs or MR images revealed air or fluid within the vertebrae. To clarify the role of distance between new VCFs and previously treated vertebrae, the distance between them was measured as the number of vertebrae (eg, the distance between L1 and L3 is two and that between T7 and T12 is five), which enables us to estimate if nearby vertebra are more prone to fracture after vertebroplasty.
All radiographs, including pre- and postvertebroplasty images, were obtained with the patient in the supine position. Plain radiographs of the patient before and after the vertebroplasty were reviewed, and the kyphosis angle and height of the anterior border of the collapsed vertebral body were measured by using conventional methods (12). All measurements were performed by one radiologist, taking care to avoid bias from cement protrusion, and all measurements were stored and verified by a second radiologist.
To correct for possible differences in magnification between the pre- and postvertebroplasty radiographs, a ratio was formed of the height of the collapsed vertebral body at the anterior border on the lateral view compared to the posterior border of the closest adjacent normal vertebral body. Increases in kyphosis angle and anterior-border height after vertebroplasty were calculated. Any leakage of cement into the disk space was evaluated on the radiographs. The development of any new VCFs was recorded at follow-up.
Statistical Analysis
Descriptive statistics are expressed as means ± standard deviations for continuous variables and as percentages for categorical variables. Differences between patients with and patients without new VCFs were assessed by using t test statistics for continuous variables and the Fisher exact test for categorical variables. The following covariates were evaluated for each patient: age, sex, body weight, body height, BMI, treated vertebrae level, preexisting VCFs, gas-containing vertebrae before treatment, height restoration, kyphosis correction after surgery, use of the pedicle approach during surgery, and cement leakage into disk. The Cox proportional hazards model was used to assess the effect of these characteristics on the risk of new VCF.
To determine the relationship between kyphosis correction and continuous covariates that were not significant from the Cox proportional hazards model, Pearson correlation coefficients were used for continuous covariates and t tests were used for categorical covariates. We also divided the 22 new VCFs identified into two subgroups: those adjacent to and not adjacent to previously cemented vertebrae. Differences between adjacent and nonadjacent VCFs were assessed by using the t test or Fisher exact test.
The mean fracture-free rate was estimated with the Kaplan-Meier method. Receiver operating characteristic (ROC) curves and ROC analysis were used to determine the cutoff value of BMI. A P value of less than .05 was considered statistically significant, and corresponding 95% confidence limits were calculated with confidence interval estimation.
Results
A total of 70 patients (61 women, nine men; mean age, 73.8 years ± 8.1) met study inclusion criteria. Patients were followed up after vertebroplasty for a mean of 20.0 months ± 10.2 (range, 6–48 months). Patients with and without new VCFs were followed up for a mean of 21.1 months ± 11.3 (range, 6–48 months) and 19.4 months ± 9.9 (range, 6–48 months), respectively (P = .51, not statistically significant). Eight of 70 patients (11%) were followed up for 6–12 months, 34 (48%) were followed up for 13–24 months, and 28 (40%) were followed up for more than 24 months. During the follow-up, we identified 22 new VCFs in 19 of the 70 patients (27%) (Fig 1). The Kaplan-Meier estimate of the 1-year fracture-free rate after vertebroplasty was 79.5% (Fig 2). The mean interval between vertebroplasty and new VCF was 10.6 months ± 9.5 (range, 1–33 months).
Figure 1.
Images obtained in a 78-year-old woman with a history of T12 fracture. (a) Sagittal fast spin-echo T1-weighted MR image shows low signal intensity in the bone marrow and a loss of vertebral body height at the T11 vertebra. A fluid-filled vacuum cleft was noted (arrow). (b) Sagittal MR image obtained with fat saturation shows high signal intensity within the T11 vertebra. A fluid-filled vacuum cleft was noted (arrow). (c) Lateral radiograph obtained after vertebroplasty shows restoration of the vertebral height and correction of spinal kyphosis. (d) Radiograph obtained 3 months after vertebroplasty, after the patient returned with a new onset of back pain, shows a new deformity of the T10 vertebra.
Figure 2.
Graph shows the fracture-free rate for neighboring vertebrae at different time intervals after vertebroplasty.
Demographics of patients stratified according to presence or absence of new VCFs are summarized in Table 1. Patient body weight, BMI, vertebral height restoration, and degree of kyphosis correction were all significantly different between patients with and patients without new VCFs (P < .01 for all). Patients with new VCFs had 10.6% of vertebral height restored, whereas patients free of new VCFs had 4.4% of height restored (P < .01). Among the 70 patients, seven (10%) had more than 10° kyphosis correction, 10 (14%) had 6°–10° correction, and 46 (75.7%) had less than 5° correction. The mean kyphosis correction was 2.4° ± 4.8 (range, −2° to 18°). The Pearson correlation coefficient showed a significant correlation between body weight and BMI and between vertebral height restoration and kyphosis correction. Therefore, the body weight of the patient and vertebral height restoration were not analyzed with the Cox proportional hazards regression analysis.
Table 1. Characteristics of Patients with and Patients without New VCF
| Variable | All Patients | Patients with New VCF (n = 19) | Patients without New VCF (n = 51) | P Value |
|---|---|---|---|---|
| Patient characteristics | N = 70 | n = 19 (27%) | n = 51 (73%) | ... |
| Age (y) | 73.8±8.1 | 74.3±7.7 | 73.6±8.4 | .77 |
| No. of women (%) | 61(87%) | 17(90%) | 44(86%) | .73 |
| Mean body weight (kg) | 58.0±9.6 | 53.1±6.3 | 59.8±10.0 | <.01⁎ |
| Mean body height (cm) | 154.8±7.5 | 155.0±8.3 | 154.8±7.3 | .92 |
| Mean BMI (kg/m2) | 24.2±3.5 | 22.1±2.4 | 24.9±3.6 | <.01⁎ |
| Mean follow-up (mo) | 20.0±10.2 | 21.1±11.3 | 19.4±9.9 | .51 |
| Imaging and technical characteristics | ||||
| Treated vertebral level of T10 through L2 | N = 47 | n = 10 (53%) | n = 37 (73%) | .12 |
| Other preexisting VCFs | 45(64%) | 13(68%) | 32(63%) | .78 |
| Gas-containing vertebra before treatment | 17(24%) | 5(26%) | 12(24%) | .81 |
| Pedicle approach (unilateral) | 54(77%) | 17(89%) | 37(73%) | .14 |
| Cement leakage into disk | 15(21%) | 4(21%) | 11(22%) | .96 |
| Vertebral height restoration (%) | 6.1±8.7 | 10.6±11.5 | 4.4±6.8 | <.01⁎ |
| Kyphosis correction (°) | 2.4±4.8 | 5.2±5.6 | 1.4±4.0 | <.01⁎ |
⁎Statistically significant. |
With use of unit-variable Cox proportional hazards regression analysis, VCF risk after vertebroplasty was associated with patient body weight, BMI, vertebral height restoration, degree of kyphosis correction, and distance to treated vertebra (P < .01 for all, Table 2). Because the distance to the treated vertebra cannot have a normal distribution, it was not included in the multivariate analysis. With use of multiple-variable Cox proportional hazards regression analysis, greater VCF risk after vertebroplasty was associated with low BMI and greater degree of kyphosis correction (P < .01, Table 3). None of the following covariates were associated with an increased risk of new VCF: age, sex, body height, duration of follow-up, level of vertebrae treated, surgical approach, gas-containing vertebrae before surgery, other preexisting VCFs before treatment, or cement leakage into disk.
Table 2. Results of Unit-Variable Cox Proportional Hazards Regression Analysis
| Variable | Hazard Ratio | 95% Confidence Interval | P Value |
|---|---|---|---|
| Patient characteristics | |||
| Age | 1.02 | 0.97,1.08 | .41 |
| Women | 1.39 | 0.32,6.07 | .66 |
| Body weight | 0.94 | 0.88,0.99 | .02⁎ |
| Body height | 1.00 | 0.94,1.06 | .99 |
| BMI | 0.78 | 0.65,0.95 | .01⁎ |
| Time of follow-up | 0.97 | 0.91,1.02 | .24 |
| Imaging and technical characteristics | |||
| Treated vertebral level of T10 through L2 | 1.86 | 0.75,4.59 | .18 |
| Other preexisting VCFs | 0.91 | 0.34,2.43 | .85 |
| Gas-containing vertebra before treatment | 0.84 | 0.30,2.33 | .73 |
| Unilateral pedicle approach | 2.64 | 0.61,11.45 | .19 |
| Cement leakage into disk | 0.93 | 0.31,2.8 | .89 |
| Vertebral height restoration | 1.07 | 1.02,1.12 | <.01⁎ |
| Kyphosis correction | 1.18 | 1.08,1.28 | <.01⁎ |
| Distance of new VCF to treated vertebra | 1.69 | 1.35,2.11 | <.01⁎ |
⁎Statistically significant. |
Table 3. Results of Multiple-Variable Cox Proportional Hazards Regression Analysis
| Variable | Hazard Ratio | 95% Confidence Interval | P Value |
|---|---|---|---|
| BMI | 0.73 | 0.58,0.91 | <.01⁎ |
| Kyphosis correction | 1.09 | 1.09,1.33 | <.01⁎ |
⁎Statistically significant. |
To further probe the contribution of other parameters, the relationships between spinal kyphosis correction and covariates not in the regression model were evaluated. Continuous and categorical variables were analyzed by using the Pearson correlation coefficient and the independent t test, respectively. We found that greater height restoration (P < .01) and gas-containing vertebrae (P = .01) were associated with greater spinal kyphosis correction.
Low BMI was a statistically significant predictor of new VCF after vertebroplasty (P < .01). We used ROC curves and ROC analysis to determine the cutoff value of BMI of the patients (Fig 3). The BMI was less than 22 kg/m2 in 19 patients and more than 22 kg/m2 in 51. In the group with new VCFs, 14 of the 19 patients (74%) had a BMI of less than 22 kg/m2; the remaining five patients (26%) had a BMI of more than 22 kg/m2. In the group without new VCF, only five of the 51 patients (10%) were found to have a BMI of less than 22 kg/m2; the remaining 46 (90%) had a BMI of more than 22 kg/m2. A BMI of less than 22 kg/m2 was thus strongly associated with new VCFs after vertebroplasty (P < .01). The sensitivity and specificity of a BMI of less than 22 kg/m2 for new VCFs were 91% and 74%, respectively.
Figure 3.
Graph shows the ROC curve for BMI. The area under the curve was 0.8.
The characteristics of patients with new VCFs adjacent versus not adjacent to cemented vertebrae are summarized in Table 4. Among 22 new VCFs, 15 (68%) were adjacent to treated vertebrae and seven (32%) were not. Among 14 patients with adjacent fractures, seven patients had new VCFs superiorly, six patients had new VCFs inferiorly, and one patient experienced both superior and inferior new VCFs. Among five patients with nonadjacent fractures, three had one new VCF and two had two new VCFs. The time to fracture was not significantly different between the two subgroups (P = .34). Neither greater kyphosis correction nor lower BMI showed statistically significant differences between the two subgroups.
Table 4. Characteristics of Patients with New VCFs Adjacent to and Not Adjacent to Cemented Vertebrae
| Variable | Adjacent VCF | Nonadjacent VCF | P Value |
|---|---|---|---|
| Patient characteristics | |||
| No. of patients | 14(74%) | 5(26%) | ... |
| Mean age (y) | 73.6±7.7 | 76.8±8.3 | .48 |
| No. of women | 12(86%) | 4(80%) | .47 |
| Mean body weight (kg) | 54.4±6.0 | 49.8±5.5 | .22 |
| Mean body height (cm) | 155.4±8.9 | 153.5±5.8 | .70 |
| Mean BMI (kg/m2) | 22.6±2.4 | 20.4±1.2 | .11 |
| Mean follow-up (mo) | 21.4±12.2 | 20.8±8.4 | .92 |
| Imaging and technical characteristics | |||
| No. of vertebrae | 15(68%) | 7(32%) | ... |
| Treated vertebra level of T10 through L2 | 8(57%) | 2(40%) | .24 |
| Other preexisting VCFs | 9(64%) | 3(60%) | .77 |
| Gas-containing vertebra before treatment | 5(36%) | 1(20%) | .95 |
| Pedicle approach (unilateral) | 13(93%) | 4(80%) | .47 |
| Cement leakage into disk | 4(29%) | 1(20%) | .84 |
| Height restoration (%) | 11.4±12.2 | 7.3±8.8 | .53 |
| Kyphosis correction (°) | 5.7±5.7 | 3.0±5.7 | .40 |
| Time to new fracture (mo) | 9.7±9.2 | 12.8±11.0 | .61 |
Discussion
The occurrence of VCFs after vertebroplasty has been investigated recently (5, 6, 7, 8, 9, 10, 11), and the reported incidence of new VCFs after vertebroplasty ranges from 8% to 52% (11). In our study, the Kaplan-Meier estimate of the 1-year fracture-free rate was 80% in vertebroplasty patients, whereas a 1-year fracture-free rate of 93% has been reported (6). This average patient age and sex balance was approximately the same in our study and the previous study (6), and the reported discrepancy in the fracture-free rate at 1 year may be too small to be meaningful. In our study, 27% of patients had a new VCF after vertebroplasty, which is comparable to the 24% rate of new VCFs in patients reported by Voormolen et al (11).
In the literature, most new VCFs occurred within 3 months after vertebroplasty (5, 6, 9, 10, 11). In our report, the mean time to fracture (10.6 months ± 9.5) was longer than that in most other studies. Survival analysis has shown a rapid decrease in the fracture-free rate within the first 2 months after surgery (6). Adjacent fractures occurred significantly sooner than nonadjacent fractures after vertebroplasty (9). Previous studies have suggested a relationship between vertebroplasty and subsequent fracture (9), but some recent studies provide conflicting results (3).
In our study, 68% of new fractures occurred adjacent to treated vertebrae after vertebroplasty. A review of published reports suggests that 41%–69% of new VCFs were immediately adjacent to the treated vertebra (5, 9, 10, 11). Other studies have also noted that there is an increased risk of new VCFs adjacent to a previously treated vertebra (6, 11). These results suggest either that vertebroplasty damages adjacent vertebrae or that vertebroplasty is usually done in a part of the spine that is already weakened, such that adjacent vertebra were more likely to fail even if the vertebroplasty procedure had not been done. We favor the latter explanation because there was some imaging evidence consistent with damage to adjacent vertebrae before vertebroplasty.
It is also possible that the occurrence of new VCFs after vertebroplasty is due to altered biomechanics in the treated area of the spine (13, 14, 15, 16). Belkkoff et al (16) used female cadavers to demonstrate that cement augmentation increases the strength and stiffness of individual fractured vertebral bodies, which may place greater stress on adjacent vertebrae. Unusual increases in strength and stiffness influence intervertebral load transfer, especially in the vertebrae immediately above or below the augmented structures (14). Therefore, a stronger study hypothesis is that pressure change in the adjacent nucleus pulposus leads to increased deflection of the endplate into the adjacent, untreated vertebra. A finite-element approach has been used (15) to demonstrate that rigid cement augmentation reduces the inward bulge of the endplates of augmented vertebra, resulting in increasing disk pressure and an inward bulge of the endplates adjacent to the augmented vertebrae. This mechanical change may do no harm to people with normal bone metabolism. In patients with osteoporosis, however, it is possible that the normal bone-remodeling cycle is compromised, so the unusual postvertebroplasty load distribution in vertebrae may result in structural failure.
In our study, a variable significantly associated with new VCF is the degree of kyphosis correction (Table 3). The risk of new VCF increased 9% per degree of kyphosis correction after vertebroplasty. Although anterior vertebral height restoration did not enable the prediction of new VCFs in our study, it is generally accepted that one of the effects of vertebroplasty is to alleviate kyphosis, which is corrected by height restoration in the collapsed vertebral body. The range of vertebral height restorations reported is 2.5–8.4 mm, or 17%–29% relative to the normal vertebral height (12, 17, 18, 19). The risk of new adjacent fracture is increased when vertebral height restoration exceeds 1 cm (6), and our results are consistent with that conclusion. In the earlier study (6), however, kyphosis correction and gas-containing vertebrae were not taken into consideration. Our study suggests that kyphosis correction results from height restoration and gas-containment before treatment, and that height restoration has a greater effect. If the vertebral fracture cleft can be sealed during the procedure, it is possible that clinicians should not seek greater kyphosis angle reduction because a greater degree of height restoration of the cemented vertebrae may increase the fracture risk of adjacent vertebrae (6).
People who are unusually thin are more likely to develop osteoporosis and to have VCFs (20). The prevalence of VCF in women aged 60–80 years is 79% for a BMI of 19, 48% for a BMI of 22, and 27% for a BMI of 28 (20). Other studies reported that VCFs are more common in subjects with a BMI of less than 24 than in those with a BMI of more than 26 (21, 22). In our study, low BMI was a significant predictor of new VCFs after vertebroplasty, especially if the BMI was less than 22 kg/m2. Our results suggest that new VCFs after vertebroplasty may be associated with low BMI rather than with a complication of the vertebroplasty procedure itself.
In patients with spinal compression fractures due to osteoporosis, the risk of new VCF is high even without percutaneous vertebroplasty (22). The annual incidence of new VCFs in patients conservatively treated after an osteoporotic VCF is approximately 20% (23). With the presence of two or more preexisting VCFs, the incidence of new VCF increases to 24% (23). In an analysis of the risk of vertebral fracture in 6,082 women in a double-blind clinical trial of alendronate for the prevention of osteoporotic fracture, the best predictor of VCF was bone mineral density (24). Each 1 standard deviation decrease in bone mineral density was associated with a 2.1-fold increase in the risk of fracture in the upper spine. In one study (11), having more than two preexisting VCFs was the only independent risk factor for the development of a new VCF after vertebroplasty.
Another risk factor for new VCFs is the migration of disk cement (7). Results of multivariate analysis suggest that the risk of new VCF is 4.6 times higher in patients with cement leakage than in patients free of leakage (10). Cement leakage into the disk raises the risk of failure in the adjacent vertebrae, with no significant effect on nonadjacent vertebrae. In another study, 58% of vertebral bodies adjacent to a disk with cement leakage fractured during follow-up, whereas just 12% of vertebral bodies adjacent to a disk without cement leakage failed (7). These findings suggest that leakage of cement into the disk during vertebroplasty is a marker for an already weakened disk or that leakage increases the risk of a new fracture. In our study, there was no statistically significant increase in the risk of new fracture of the adjacent vertebral body when there was cement leakage into the disk space (Table 1). This difference in outcome could be a result of differences in the procedure. We used a minimum volume of cement to seal the fracture region, whereas others (7) filled the vertebra with as much cement as possible. No empirical conclusions can be drawn, however, because data characterizing the total quantity of cement we injected are unreliable. In one retrospective study (25), a semiautomatic volumetric quantification of CT scans was performed to calculate the volume of cement injected. In the osteoporotic spine, the disk and its periphery were the primary sites of cement leakage after vertebroplasty, but no correlation was found between total cement injection and leakage volume (25).
A weakness of our study is that we do not have a direct measure of osteoporosis in our patients. Osteoporosis is a substantial risk factor for VCFs, and VCFs occur in at least 25% of postmenopausal women (26). The diagnosis of osteoporosis and the assessment of bone fracture risk require an accurate measurement of bone mineral density (27). Old age, which is associated with bone resorption, is also a substantial risk factor for VCFs (28). These findings suggest that measurements of bone mineral density would be a potential way to predict the risk of new or recurrent VCF.
In conclusion, new VCFs occur after vertebroplasty in about 30% of patients. Most new fractures occur in the vertebrae immediately adjacent to a vertebroplasty. Greater kyphosis correction increases the risk of new VCFs and should perhaps not be done if the fracture cleft can be sealed. Our findings show that patients with a BMI of less than 22 kg/m2 are significantly more likely to develop new VCFs after vertebroplasty. In the future, we hope to refine the methods described so that we can predict which patients could benefit from intervention to prevent VCFs.
References
- . Percutaneous polymethylmethacrylate vertebroplasty in the treatment of osteoporotic vertebral body compression fractures: technical aspects. AJNR Am J Neuroradiol. 1997;18:1897–1904
- The role of imaging studies of percutaneous vertebroplasty in 63 patients with osteoporotic compression fracture: preliminary report. Chin J Radiol. 2004;29:177–184
- . Spinal metastases: indications for and results of percutaneous injection of acrylic surgical cement. Radiology. 1996;199:241–247
- Percutaneous vertebroplasty for osteolytic metastases and myeloma: effects of the percentage of lesion filling and the leakage of methyl methacrylate at clinical follow-up. Radiology. 1996;200:525–530
- . Occurrence of new vertebral body fracture after percutaneous vertebroplasty in patients with osteoporosis. Radiology. 2003;226:119–124
- . Risk factors of new compression fractures in adjacent vertebrae after percutaneous vertebroplasty. Acta Radiol. 2004;45:440–445
- . Vertebroplasty: cement leakage into the disc increases the risk of new fracture of adjacent vertebral body. AJNR Am J Neuroradiol. 2004;25:175–180
- . Adjacent vertebral failure after vertebroplasty: a biomechanical investigation. J Bone Joint Surg Br. 2002;84:748–752
- . New fractures after vertebroplasty: adjacent fractures occur significantly sooner. AJNR Am J Neuroradiol. 2006;27:217–223
- Percutaneous vertebroplasty for osteoporotic compression fracture: multivariate study of predictors of new vertebral body fracture. Cardiovasc Intervent Radiol. 2006;29:580–585
- . The risk of new osteoporotic vertebral compression fractures in the year after percutaneous vertebroplasty. J Vasc Interv Radiol. 2006;17:71–76
- Kyphosis correction and height restoration effects of percutaneous vertebroplasty. AJNR Am J Neuroradiol. 2003;24:1893–1900
- . Percutaneous vertebroplasty. Radiology. 2003;229:27–36
- . The effect of cement augmentation on the load transfer in an osteoporotic functional spinal unit: finite-element analysis. Spine. 2003;28:991–996
- . Load shift of the intervertebral disc after a vertebroplasty: a finite-element study. Eur Spine J. 2003;12:421–426
- . The biomechanics of vertebroplasty: the effect of cement volume on mechanical behavior. Spine. 2001;26:1537–1541
- . Increase in vertebral body height after vertebroplasty. AJNR Am J Neuroradiol. 2003;24:185–189
- . Reporting height restoration in vertebral compression fractures. Spine. 2003;28:2517–2521
- . Does vertebral height restoration achieved at vertebroplasty matter?. J Vasc Interv Radiol. 2005;16:973–979
- . Determinants of bone mineral density and risk factors for osteoporosis in healthy elderly women. J Bone Miner Res. 1993;8:669–675
- . Measurements of bone mineral density by DXA total body absorptiometry in different skeletal sites in postmenopausal osteoporosis. Bone. 1992;13:173–178
- . Crush fracture syndrome in senile osteoporosis: a nutritional consequence?. J Bone Miner Res. 1992;7:317–319
- Risk of new vertebral fracture in the year following a fracture. JAMA. 2001;285:320–323
- . Association of prevalent vertebral fractures, bone density, and alendronate treatment with incident vertebral fractures: effect of number and spinal location of fractures. Bone. 1999;25:613–619
- . Volumetric quantification of cement leakage following percutaneous vertebroplasty in metastatic and osteoporotic vertebrae. J Neurosurg. 2003;99:56–59
- . Osteoporotic compression fractures of the spine: current options and considerations for treatment. Spine J. 2006;6:479–487
- Assessment of fracture risk. Osteoporos Int. 2005;16:581–589
- Risk factors of developing new symptomatic vertebral compression fractures after percutaneous vertebroplasty in osteoporotic patients. Eur Spine J. 2006;15:1777–1783
None of the authors have identified a conflict of interest.
PII: S1051-0443(07)01266-3
doi:10.1016/j.jvir.2007.09.008
© 2008 SIR. Published by Elsevier Inc. All rights reserved.
Volume 19, Issue 2 , Pages 225-231, February 2008