Objective. To elucidate the risk factors for a subsequent vertebral compression fracture following percutaneous vertebroplasty, we analyzed the potential predictors of vertebral compression fractures adjacent to or remote from fractures previously treated with percutaneous vertebroplasty.
Design. This is a retrospective cohort study.
Background. A major concern after percutaneous vertebroplasty in patients with osteoporosis is the occurrence of subsequent vertebral compression fractures in the untreated vertebral bodies. The risk factors for the development of subsequent vertebral compression fractures after percutaneous vertebroplasty are unclear.
Methods. Two hundred four consecutive patients underwent percutaneous vertebroplasty for acute vertebral compression fractures between January 2007 and December 2008. Forty-nine patients were excluded. Subsequent vertebral compression fractures were diagnosed by bone edema changes on magnetic resonance imaging. Patient's demographic data were used for univariate and multivariable binary logistic regression analyses.
Results. Forty-three (27.7%) of the 155 patients had subsequent vertebral compression fractures within 2 years of percutaneous vertebroplasty, with 21 (48.8%) of these patients having fractures detected within 3 months. Adjacent vertebral compression fractures tended to occur sooner, although not significantly (log-rank test, P = 0.112). On multivariate analyses, only the T-score of bone mineral density was significantly associated with subsequent vertebral compression fractures (P < 0.0001; odds ratio = 0.27; 95% confidence interval, 0.15–0.49).
Conclusions. The only risk factor significantly associated with subsequent vertebral compression fractures following percutaneous vertebroplasty was a low bone mineral density T-score. Patients with lower bone mineral density have a higher incidence of vertebral compression fractures and thus need more intensive clinical and radiological follow-up.
Bone Mineral Density
The majority of osteoporotic vertebral compression fractures (VCFs) are not very painful. However, occasionally acute osteoporotic VCF with refractory pain is a crippling disorder, frequently resulting in severe and prolonged back pain, lengthy hospitalization, physical decline, and a potential risk of increased mortality [1,2]. Bed rest, analgesic medications, external bracing, and physical therapy were once the only available therapies and had limited success . Percutaneous vertebroplasty (PV), the injection of bone cement into the fractured vertebral body, was introduced as an alternative treatment option for acute osteoporotic VCFs and has gained widespread acceptance as an effective method of pain relief [4–6]. Although two randomized trials [7,8] to assess the efficacy of PV with a sham control intervention showed no improvement over controls, interpretation of these studies is hampered by inclusion of some chronic VCFs instead of only acute osteoporotic VCFs and inconsistent use of bone edema on magnetic resonance imaging (MRI) as an inclusion criterion . Another randomized study  showed that PV is effective and safe in a subgroup of patients with acute osteoporotic VCFs and persistent pain.
A major concern after PV in patients with osteoporosis is the occurrence of new VCFs in the untreated vertebral bodies at other levels. Some authors believe that new VCFs after PV are caused by augmented stiffness of the treated vertebrae related to the amount of injected cement or by cement leakage into the adjacent vertebral disc space [9–15]. It has also been proposed that ongoing osteoporosis induces new VCFs [16–19]. The purpose of this study was to analyze the potential predictors of new VCFs adjacent to or remote from VCFs previously treated with PV.
Patients and Methods
This retrospective study was approved by the Institutional Review Board of our hospital. Between January 2007 and December 2008, we performed 244 PV procedures on 244 vertebral bodies in 204 consecutive patients. Patients with osteoporosis or osteopenia were selected to receive vertebroplasty if they had acute vertebral fracture pain that was refractory to medical treatment. Prevertebroplasty radiographic evaluation included plain radiography, computed tomography, and MRI, which were performed to exclude other causes of back pain. The inclusion criteria were acute pain (lasting < 6 weeks), low signal intensity on T1-weighted and high signal intensity on T2-weighted MRI images of the fractured vertebrae, VCFs with >20% loss of height, focal tenderness at the fractured level, and decreased bone mineral density (BMD) with a T-score of −1 . Exclusion criteria were pathological fractures (due to malignancy/myeloma), osteomyelitis, major retropulsion of bony segments into the spinal canal, collapse of the vertebral body with residual height < 10% that made PV difficult, coagulopathy, and a follow-up period of <24 months. One hundred fifty-five patients met the study inclusion criteria (Figure 1).
PV was performed under strict sterile conditions in an operating room. All patients received conscious intravenous sedation prior to the procedure to make sure they were comfortable. Local anesthesia with 2% lidocaine was administered through the skin to the periosteum of the targeted pedicle. Patients were placed in the prone position under fluoroscopic guidance, and the skin overlying the target area was prepared and draped. The procedure was performed using a unipedicle method . Targeted pedicles were localized under fluoroscopic guidance. After a small incision was made, an 11-gauge bone biopsy needle (Stryker Instruments, Kalamazoo, MI, USA) was pushed through the cortex, made to traverse the pedicle, and then directed into the anterior third of the vertebral body. Bone cement was prepared from methyl methacrylate polymer (15 g of powder in 10 mL of solvent) and sterile barium (5 g of powder to increase radio-opacity). Cement was injected forcibly into the vertebral body using a 1-mL syringe under continuous fluoroscopic imaging guidance. After the procedure, radiographs of the treated vertebral bodies were obtained to identify cement leakage or other local complications. All patients were observed for 12–24 hours after treatment and were instructed to wear an orthosis for at least 3 months after the operation. Patients were followed up in the outpatient department monthly for 3 months after the operation and subsequently at 3-month interval. Patients were encouraged to immediately undergo radiography or MRI for recurrent back pain. Complications related to PV and new VCFs were evaluated during follow-up. Subsequent VCFs were defined by a decrease of >20% body height and bone edema changes as determined by MRI.
Patient data including age, gender, body mass index (BMI), history of excessive smoking and alcohol use, use of steroids, and nonsteroidal anti-inflammatory drugs (NSAIDs), bisphosphonate and cholecalciferol therapy, vertebroplasty level, cement volume, cement leakage into the disc, vacuum clefts, and lumbar spine BMD were recorded and analyzed as potential risk factors. BMD data were available for 147 patients (94.8%). T10 to L2 was defined as the thoracolumbar (TL) junction. BMD values were obtained using dual-energy X-ray absorptiometry (GE Lunar PRODIGY Advance Fan-Beam, GE Healthcare, WI, USA) before the PV procedure. A database of BMD values of the Chinese population was used as a reference . Patients were prescribed bisphosphonate and cholecalciferol therapy after BMD data became available.
Categorical variables were presented as numbers (percentages) and continuous variables as mean ± standard deviation (SD). Comparisons of patient characteristics between the two groups (no new VCFs vs new VCFs) were made by the chi-square test for categorical variables and the unpaired t-test for continuous variables. Potential predictors of new VCFs were examined using multivariable binary logistic regression analyses with the all-enter method. Results were presented as adjusted odds ratio (OR) with corresponding 95% confidence interval (95% CI). The Kaplan–Meier survival curve and log-rank test were used to compare time to diagnosis of new VCFs between fractures adjacent to the treated levels and fractures not adjacent to the treated levels. All statistical analyses were carried out using SPSS for Windows (SPSS 15.0; SPSS Inc., Chicago, IL, USA).
One hundred thirty-one patients were women (84.5%). The patients' mean age was 73.3 ± 9.8 years (range: 43–94 years). Among the patients, 135 received vertebroplasty at one level, 18 at two levels, and 2 at three levels. Table 1 lists the demographic characteristics of the patients. Forty-nine patients were excluded from the analysis, including 31 who were lost to follow-up, 10 without qualified radiographs for analysis, 5 who died within 24 months of the follow-up period, and 3 who had VCFs due to malignancy.
Forty-three patients (27.7%) had 47 new VCFs within 2 years of PV; in 21 cases (48.8%), the VCFs occurred within 3 months (Figure 2). Thirty-nine patients had new VCFs at one level and four patients had new VCFs at two levels. Among the 47 new VCFs, 25 vertebrae (53.2%) were rostral to those previously treated with PV. The mean time to diagnosis of new VCFs was 24.6 ± 26.1 weeks (range: 2–100 weeks). Twenty-four (55.8%) of the 43 patients had new VCFs in vertebrae adjacent to the vertebral body treated by vertebroplasty, whereas 19 (44.2%) had nonadjacent VCFs. Mean time to new adjacent VCFs was 18.1 weeks with a standard error of the mean (SEM) of 4.5 (median, 8; SEM, 3.7). However, mean time to nonadjacent VCFs was 32.8 weeks with an SEM of 6.7 (median, 14; SEM, 21.8). Adjacent VCFs showed a trend toward occurring sooner, although it did not reach statistical significance (log-rank test, P = 0.112; Figure 3).
Kaplan–Meier survival curve depicting time to new fractures adjacent and nonadjacent to treated levels.
Univariate analyses showed that only the T-score of BMD was significantly associated with new VCFs within 2 years of PV (P < 0.0001). Vertebral compression fractures occurred more frequently in women (P = 0.07), possibly because they had lower T-scores (−2.57 ± 0.89) than men (−1.94 ± 0.99). Patients with a history of smoking had less frequent VCFs (P = 0.058), although the difference did not reach statistical significance. Age, BMI, alcohol excess, use of steroids, use of NSAIDs, bisphosphonate and cholecalciferol therapy, TL junction location, cement volume, and cement leakage into the disc, and vacuum clefts were not associated with new VCFs (Table 2). Furthermore, multivariable binary logistic regression analyses revealed that the BMD T-score (P < 0.000,1, OR = 0.27, 95% CI, 0.15–0.49) was the only independent prognostic factor for new VCFs within 2 years of a PV procedure.
Comparison of the 24 patients with adjacent VCFs with 131 patients without adjacent VCFs (19 patients with nonadjacent VCFs) in the present study showed that the T-score of BMD is also lower in patients with adjacent VCFs than in those without adjacent VCFs (3.0 ± 0.9 vs 2.3 ± 0.9, P = 0.003).
The most frequent complication was cement leakage, which was found in 39 of the 155 patients (25.1%). However, most patients were asymptomatic. There were no major complications except for one procedure during which cement migrated toward the lungs. The procedure was halted when this was observed. The patient had no clinical symptoms of pulmonary emboli.
In the present study, we analyzed the potential predictive factors affecting new VCFs in vertebral bodies adjacent to or remote from the vertebrae previously treated with PV. Mean time to new adjacent VCFs was 18.1 weeks and to nonadjacent VCFs was 32.8 weeks. Although this difference did not reach statistical significance, it indicated that adjacent VCFs tend to occur sooner. Univariate and multivariate logistic regression analyses showed that only the T-score of the lumbar spine was significantly associated with new VCFs within 2 years. As expected, patients with a lower BMD had a higher incidence of new VCFs.
There are many risk factors that have been reported for the occurrence of new VCFs after vertebroplasty, including old age , use of steroids , location at the TL junction , osteoporosis [19,25], prior vertebral fractures [22,26,27], proximity to the initial fracture site [14,19,24,27–30], cement leakage into the discs [14,22,31], and vacuum clefts within the compression fracture . We were unable to document the relationship between the risk of new VCFs and age, gender, use of steroids, location at the TL junction, proximity to the initial fracture site, cement leakage into the discs, and vacuum clefts within the compression fracture. In our study, women had a higher incidence of new fractures (30.5%) than men (12.5%), possibly because women had severer osteoporosis and lower T-scores.
Cement leakage into the discs [14,22,31] and vacuum clefts within the compression fracture  could be important risk factors for subsequent adjacent fractures. Polikeit et al.  found the overall load transfer in the adjacent levels to the vertebrae treated with PV to be markedly changed with increased stress and strain. Pressure changes in the intervertebral discs due to cement leakage into the discs have been shown to lead to an enlarged deflection of the endplate into the adjacent untreated vertebrae, which could induce subsequent adjacent fractures. The stiffer vertebrae after vertebroplasty reduced the inward bulge of the endplates of the augmented vertebra. Baroud et al.  found that the bulge of the augmented endplate resulted in a stiffening of the intervertebral joint and of the whole motion segment. The intervertebral pressure increase and the inward bulge of the endplate adjacent to the augmented vertebrae may be the cause of the adjacent fractures.
The hard cement results in increased mechanical pressure and may lead to new adjacent fractures. The increased mechanical pressure is especially pertinent in patients who increase their daily activities as their back pain improves after PV, which places additional stress on the vertebral bodies [14,22,31]. One study observed 53 patients treated with PV and found that 8 (20.5%) of the 39 patients with vacuum clefts had subsequent VCFs . However, in our study, vacuum clefts were not associated with new VCFs. We could not demonstrate that these factors were associated with subsequent VCFs. Perhaps these effects only have some impact on adjacent VCFs.
Some authors  have suggested that a larger volume of cement injected during PV may result in an increased tendency for cement leakage and a higher subsequent VCF rate. In our study, similar to some reports [8,14], the cement volume used was smaller in the new VCF group, and thus, we did not confirm the above-mentioned observation.
Our earlier study  demonstrated that osteoporosis is a predisposing factor for the occurrence of new VCFs as 62.5% of new VCFs occurred in patients who had a lumbar T-score < −2.5. The present study, which included more patients than our prior study, demonstrated that low BMD was the only independent prognostic factor for new VCFs within 2 years following PV. Few studies have provided BMD data for patients with new VCFs following PV. Further investigations are needed to clarify the role of osteoporosis in VCFs following PV.
There is controversy as to whether PV predisposes patients to new VCFs. In this study, the incidence of new VCFs after PV was 27.7%. It has been reported to be as high as 48% within 1 year . Some authors believe the incidence of new VCFs following PV to be higher than those that occur naturally [6,34,35]. For patients treated with conservative therapy, significant back pain further hindered mobility and ambulation. This may have a protective effect against new VCFs. Thus, it is unclear whether PV affects the occurrence of new VCFs. The present study suggests that the most important risk factor for new VCFs could be the underlying osteoporosis. Fragile bones due to severe osteoporosis are highly susceptible to subsequent VCFs. Patients with a lower BMD have a higher incidence of VCFs and need more intensive clinical and radiological follow-up.
Osteoporosis and pre-existing VCFs increase the incidence of additional VCFs after PV [34,36]. In a large cohort study for serial identification of new VCFs, Ross et al. found that the risk for subsequent fracture increased fivefold after the first VCF and 12-fold after there were 2 or more VCFs .
In the present study, bisphosphonate and cholecalciferol therapy seemed to be ineffective for decreasing new VCFs. However, only 74% of patients received bisphosphonate and cholecalciferol therapy because the Bureau of National Health Insurance of Taiwan does not pay for bisphosphonate and cholecalciferol therapy for patients with a T-score of BMD > −2.5. Patients with a lower BMD T-score were prescribed bisphosphonate and cholecalciferol therapy. The efficacy of bisphosphonate and cholecalciferol therapy needs further investigation.
This study has some limitations. Although patients were encouraged to undergo imaging for recurrent back pain, some asymptomatic new VCFs might not have been detected because appropriate imaging was unavailable. This may have had an influence on the incidence of new VCFs and thus might be a source of bias. Data on the patients' physical activity levels were not available. This is an important limitation because decreased mobility or inactivity secondary to pain may have a protective effect in preventing new VCFs. It is too early to discuss the efficacy of bisphosphonates and vitamin D3 because some fractures occur so early after treatment.
In this study, 27.7% of patients with one or more VCFs treated with PV developed new VCFs within 2 years of follow-up. The only risk factor significantly related to new VCFs was the BMD T-score. Fragile bones due to severe osteoporosis are highly susceptible to subsequent VCFs. Patients with a lower BMD have a higher incidence of VCFs and need more intensive clinical and radiological follow-up. It is also important to consider patients' activity level when studying the rates of new VCFs after PV. The efficacy of bisphosphonate and cholecalciferol therapy also needs more investigation. Future research correlating activity level with new VCFs following PV would be helpful in this analysis.