In our evaluation of predictive factors for de novo adjacent compression fractures following PVP, we uncovered statistically significant relationships between an increased risk of complications and the variables of age and BMD scores. Interestingly, decreased lumbar and hip BMD scores both indicated a high risk of adjacent VFs. Based on our LR results, patients 85 years of age or older with low BMD scores do not represent good surgical candidates for PVP. In contrast, those less than 75 years old with higher BMD scores have a relatively low risk of fracture, suggesting that these individuals are suitable for PVP intervention.
Several authors have reported that the mechanism for new fractures is a combination of the usually osteoporotic underlying condition and the hard cement injected into the vertebral body [11–14]. Trout et al. described that vertebral bodies adjacent to those treated with vertebroplasty had greater than 4 times the risk of fracture than non-adjacent vertebrae. The authors also showed that new-onset adjacent-level fractures after PVP occurred significantly sooner than non-adjacent ones [8]. Moreover, Uppin et al. revealed that many adjacent VFs occurred within a month after PVP [15]. In our preliminary subject screening, 18 of 19 new fractures (94.7 %) had affected adjacent vertebrae, 17 of 19 fractures (89.5 %) had occurred within a month after surgery, and 16 of 19 fractures (84 %) were both adjacent and manifested within a month. Indeed, it appears critical to prevent adjacent VFs in the month following PVP treatment. We therefore focused on these parameters when analyzing clinical factors, bone marker values, and BMD scores in the present study.
Adjacent fractures occurred more significantly in the upper vertebrae. In general, kyphosis occurs in thoraco-lumbar fractures due to the collapse of the anterio-superior vertebrae. When the regions are surgically repositioned, defects in the upper vertebrae may become present and cement might enter the fronto-tepomoral regions. Nagaraja et al. recently showed that vertebroplasty alters the biomechanics of the spine, which increases compression on adjacent vertebral bodies and intervertebral discs, especially in severely osteoporotic women [16]. Furthermore, in vitro studies have revealed that cement-filled bone is 36 times stronger than spinal cancellous bone [17]. Therefore, it is plausible that more fractures occur in the upper vertebrae that make contact with the solid cement after PVP.
Several risk factors have been described with respect to adjacent VFs following vertebroplasty [7, 18–20]. Zhang et al. identified the significance of low lumbar BMD, low BMI, and intradiscal cement leakage in de novo fractures [7], while Ryo et al. implicated osteoporosis and intervertebral discal cement leakage [20]. Based on these reports, the strongest prognostic clinical factors appear to be lumbar BMD-related conditions and cement leakage.
Smaller volumes of cement have been shown to decrease the risk of newly forming VFs while maintaining sufficient stability [16]. In the present series, PMMA was injected into the vertebral body until the cement reached the posterior one-fourth of the body, as indicated by another group [17], or until significant leakage occurred. We had anticipated that the amount of leakage would be relatively small due to our careful adjustment of the amount and distribution of cement administration. However, 4 and 3 patients experienced soft tissue and intradiscal leakage of PMMA, respectfully. The 7 patients in whom leakage was found had significantly lower BMD scores than the remainder of the cohort. This suggests that diminished BMD is also a serious risk factor with respect to cement leakage.
Syed et al. revealed that subsequent adjacent and non-adjacent fractures after PVP occurred at roughly equal frequencies in disc extravasation and non-disc extravasation groups. [21]. However, considerable evidence supports that cement leakage is a primary risk factor for new vertebral compression fractures [7, 10, 19]. As the 3 patients who experienced leakage into the disc space were all in the collapse group, we suspected that leakage of cement into this area influenced adjacent VF onset. The BMD of the 4 cases of cement leakage into soft tissue was also significantly lower than that of the remaining cases. However, such leakage did not lead to adjacent VF. Thus, it appears that cement leakage itself does not lead to VFs; rather, increased disc stiffness after leakage may have increased the risk of adjacent compression fracture.
The cumulative incidence of VF over 10 years of follow-up was shown to be between 5.1 and 22.2 % by Yoshimura et al. [22]. Zhang et al. also reported that age was a prominent risk factor in the occurrence of fractures. We witnessed that the collapse group was significantly older than the non-collapse group (Table 3). Furthermore, the OR of adjacent fractures when age was increased by 1SD was 4.5 (P = 0.001). Although age is a well known risk factor for fractures, there have been few reports on its relationship with PVP. Trout et al. stated that 186 new VFs occurred in 86 of 432 patients (19.9 %). In this study, 19 of 73 patients (26.0 %) experienced de novo VFs, which was comparably higher. One possible reason for this discrepancy is age; the median age was 75.2 years in Trout’s report [8] versus 77.2 years in ours. Nevertheless, such findings confirm that age plays an important role in the frequency of subsequent VFs. In the present study, the risk of fracture in women was high at an age of more than 85 years and low for an age of less than 75. Therefore, an increased fracture risk should be expected in elderly women aged 85 years or older.
It is widely known that fracture risk rises in patients with high bone turnover [23]. Our findings revealed that the level of urinary NTX in the collapse group was significantly higher than that in the non-collapse group (Table 3). However, there was no significant difference by logistic regression analysis (P = 0.073). Serum BAP levels were not apparently related to the incidence of new fractures. These results resembled those of Komemushi et al., who reported that a combination of high levels of bone resorption markers and normal levels of bone formation markers may be associated with an increased risk of de novo fractures after PVP [24]. Urinary NTX values in our previously reported cohort of osteoporotic outpatients was 62.2 on average, which had risen to 76.7 on average in patients over 80 years of age [25]. Taken together, it is conceivable that in accordance with aging, the elevated values of urinary NTX may be related to fracture risk.
According to the International Society for Clinical Densitometry and International Osteoporosis Foundation, one of the key clinical risk factors for osteoporosis is diminished hip BMD [26]. In the present study, the strongest risk factors for new VFs were lumbar as well as hip BMD. Trochanteric BMD showed the highest statistical correlation among hip BMD scores (Odds ratio: 9.6). It is currently unknown why trochanteric BMD might prognosticate VFs, although this may be due to the fact that the trochanteric region involves far more cancellous bone than cortical bone. The high risk values of BMD based on LR+ were less than −2.6SD for lumbar, less than −1.8SD for total hip, and less than −2.1SD for femoral neck. The low risk values of BMD based on LR- were −1.8SD for lumbar, less than −0.8SD for total hip, and less than +0.1SD for femoral neck. These results indicate that an important factor in the prediction of adjacent VFs may be hip BMD as much as lumbar BMD. Although correlations exist between spine and hip BMD, they are insufficient to assess the counterpart’s BMD value [27]. Therefore, not only vertebral BMD, but also hip BMD and especially trochanteric BMD, should be assessed before vertebroplasty.
The main limitations of this study are its retrospective design and relatively small sample size. However, as we performed strict statistical analyses to compensate for the latter shortcoming, we consider our findings to be applicable in the treatment of spinal fractures using PVP.
The merit of vertebroplasty is still controversial. Lindsay et al. reported that while some clinicians felt vertebroplasty was an effective procedure in select patients, others believed that other treatment modalities for osteoporotic compression fractures were preferable to surgery [28]. We earlier treated such cases conservatively with rest. Patients required 3 weeks on average to regain mobility following a reduction in pain [29]. Meanwhile, PVP has few complications and can improve pain very quickly, and thus represents an effective, low invasive option. However, the occurrence of adjacent fractures remains a pressing issue. We propose that not only technical surgical improvements, such as the amount of injected cement, but also preoperative risk evaluation for new adjacent fractures along with postoperative treatment with teriparatide and other biologics, may enhance the merits of PVP.