Introduction

Vertebroplasty (VP) is a minimally invasive intervention used to treat osteoporotic vertebral compression fractures (OVCFs) through the injection of an acrylic cement into the collapsed vertebral body [1]. Balloon kyphoplasty (BKP) is another intervention derived from VP, relying on the inflation of a balloon into the collapsed vertebral body before injecting the acrylic cement with the intent of restoring the vertebral height [1]. In addition to VP and BKP, many other vertebral augmentation (VA) techniques have been reported, although VP and BKP are still the most common VA procedures used in clinical practice [1, 2].

VP proved to be cost-effective and clinically beneficial in relieving OVCF-related pain [3,4,5]. However, the publication of two double-blind randomized trials demonstrating lack of significant pain improvement after VP (versus a “sham” procedure) in patients with OVCFs [6, 7] raised questioning of the clinical appropriateness of VP (and therefore of all VA procedures) proposed on a large-scale for the treatment of OVCFs [8]. As a result, recent increased adoption of non-surgical management (NSM; i.e., bed rest, analgesics, and bracing) has been noted for OVCFs [9].

Multiple explanations have been raised to justify the underestimation of the analgesic effect of VP in these two studies [6, 7] (e.g., enrollment, inclusion/exclusion criteria, use of the “sham” procedure, follow-up), which have, moreover, substantially contributed to gather attention on the sole outcome dealing with post-operative pain, thus neglecting other potential clinical benefits yielded by VA procedures, including lower patients’ mortality and morbidity [10,11,12,13]. In fact, some recent retrospective studies and one recent meta-analysis [11, 14, 15] have reported a substantial reduction of the mortality risk following VP/BKP in OVCF patients as compared to NSM. Nevertheless, mortality data are still sparse and scarce, and it is not still completely clear how patient’s morbidity is impacted by VP/BKP. Therefore, we have conducted a systematic literature review and meta-analysis to gather evidence on OVCF patients’ risk decrease of mortality and morbidity after VP/BKP or NSM.

Materials and methods

No specific funding was received to carry out the work described in this article. Two authors (R.L.C. and J.G.) are advisors to and two (T.B. and M.S.) are employed of Medtronic. Authors with no financial disclosures relating to the medical devices had control of the data and information submitted for publication.

Selection criteria

A systematic literature search was conducted on Medline and EMBASE databases on 12 November 2019 for original publications on VP/BKP patients, scoring 1–3 on the Oxford quality scale [16]. The following keywords were entered: “vertebral compression fracture,” “vertebroplasty,” “kyphoplasty,” “conservative management,” “non-surgical management,” “mortality,” and “morbidity.” According to the “Population-Interventions-Comparators-Outcomes-Study” (PICOS) model (Table 1), 62 studies were included. After excluding studies with fractures from non-osteoporotic origin (n = 9) and those that did not compare PV and/or BKP to NSM (n = 31), 22 studies were obtained. Two studies did not report post-treatment patients’ mortality and/or morbidity and were, therefore, excluded. Another study was finally excluded since data were published twice. Therefore, 19 studies [4, 17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34] were finally considered for the quantitative analysis (Fig. 1).

Table 1 Inclusion and exclusion criteria adopted to select studies according to the PICOS (P, population; I, interventions; C, comparator group; O, outcomes; S, study design) model
Full size table
Fig. 1
figure 1

Flowchart of the study inclusion and exclusion process conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines

Full size image

Eight studies (8/19; 42%) reported data on mortality and 16 (16/19; 84%) on morbidity. In particular, mortality was assessed at 1- (n = 2), 6- (n = 5), 12- (n = 5), and 24-month (n = 2) follow-up, and morbidity at 3- (n = 1), 6- (n = 1), 12- (n = 11), and 24-month (n = 2) follow-up.

Outcomes

The primary analysis focused on the 12-month all-cause mortality and all-cause morbidity in OVCF patients undergoing VP/BKP or NSM. In both cases, analyses were set at 12-month due to the larger number of data reported at such interval by the included studies.

Recorded events to assess morbidity were classified into cardiovascular (e.g., deep vein thrombosis, angina pectoris, arrhythmia, myocardial infarction), pulmonary (e.g., cement and non-cement embolism, dyspnea), infection (e.g., urinary tract infections, pneumonia, spondylitis, sepsis), musculoskeletal (e.g., back pain), neurologic/psychiatric (e.g., sleep disorder, depression), gastrointestinal (e.g., gastrointestinal bleeding), and new vertebral compression fractures (VCFs) occurring after VP/BKP or NSM and requiring subsequent medical or surgical interventions.

Secondary analyses were conducted to:

  • identify covariates favoring VP/BKP or NSM in terms of 12-month all-cause morbidity;

  • compare VP/BKP and NSM in terms of 12-month infective morbidity from any origin;

  • compare VP/BKP and NSM in terms of cardio-pulmonary, infective, and new VCFs events at all the available pooled follow-ups. For this purpose, single dedicated analyses were conducted for each studied item.

Data analysis

Descriptive statistics were used to summarize data by reporting means and standard deviation for continuous variables, and counts and percentages for categorical variables.

For analyses of endpoints evaluated at 12 months, estimates for each study were reported as odds ratios (OR) along with their 95% confidence intervals (CI) and p values.

For analyses of endpoints with pooled follow-ups, estimates for each study were reported as incidence rates ratios (IRR) along with their 95% CI and p values. When either the number of events or non-events in the VP/BKP or NSM cohorts was equal to 0, continuity correction was applied. Studies with no events in both treatments’ groups were excluded from the analyses.

Pooled estimates were computed using a fixed effects linear model, which is equivalent to compute a weighted mean using the within-study variances as weights.

Studies’ heterogeneity was calculated with Cochran’s Q and Higgin’s I2 tests. When I2 was > 25%, random effect meta-analysis was performed.

The meta-regression analysis for covariates correlating with 12-month all-cause morbidity was performed using the same model as the one for pooled estimate computation. Publication year, patients’ mean age, percentage of females, and mean length of symptoms’ duration were the studied covariates. Furthermore, a meta-analysis was conducted separately for observational and randomized studies.

Publication bias was assessed using the Egger’s test. Furthermore, a sensitivity analysis was performed by computing pooled estimates iteratively with the exclusion of each study, in order to detect possible influential studies that unduly affected the pooled estimate.

All statistical analyses were performed with SAS software 9.4 (SAS Institute Inc.) and validated using Stata/SE 15.1 (Stata Corp LCC). All tests were based on a two-sided significance level of 0.05.

Results

Baseline characteristics

Baseline characteristics of included studies are summarized in Supplementary Table 1. Among the 19 included studies, 13 reported PV (13/19; 68%) as the main intervention, 4 BKP (4/19; 21%), and 2 both PV and BKP (2/19; 11%). In all these studies, NSM represented the control group. Seven (7/19; 37%) studies were randomized controlled trials and 12/19 (63%) were observational studies (6 prospective, 6 retrospective).

Primary outcomes

12-month all-cause mortality

Pooled OR across 5 studies [17, 19, 21, 30, 32] (Table 2) comparing VP/BKP with NSM showed a reduction of the 12-month all-cause mortality (OR: 0.81 [95% CI: 0.46–1.42]; p = .46) in favor of VP/BKP (Fig. 2). No single study was preponderant. Studies were quite homogeneous (Q = 3.25; p = .52; I2 = 0%), and the sensitivity analysis confirmed that no single study was a key contributor to the pooled estimate (Supplementary Table 2). The Egger’s test for publication bias was not significant (p = .64).

Table 2 12-month all-cause mortality and all-cause morbidity
Full size table
Fig. 2
figure 2

Meta-analysis of studies comparing 12-month all-cause mortality in patients with osteoporotic vertebral compression fractures receiving vertebroplasty or balloon kyphoplasty versus non-surgical management. OR, odds ratio; CI, confidence interval; PV/BKP, percutaneous vertebroplasty/balloon kyphoplasty; NSM, non-surgical management

Full size image

12-month all-cause morbidity

Pooled OR across 11 studies [4, 17, 20, 21, 24,25,26,27,28,29, 32] (Table 2) comparing VP/BKP with NSM (Fig. 3) showed a reduction of 12-month all-cause morbidity (OR: 0.64 [95% CI: 0.31–1.30]; p = .25) in favor of VP/BKP. No single study was preponderant. Studies were quite heterogeneous (Q = 28.55; p = .001; I2 = 65%) mainly due to older studies. The sensitivity analysis confirmed that no single study was a key contributor to the pooled estimate (Supplementary Table 2). The Egger’s test for publication bias was not significant (p = .94).

Fig. 3
figure 3

Meta-analysis of studies comparing 12-month all-cause morbidity in patients with osteoporotic vertebral compression fractures receiving vertebroplasty or balloon kyphoplasty versus non-surgical management. OR, odds ratio; CI, confidence interval; PV/BKP, percutaneous vertebroplasty/balloon kyphoplasty; NSM, non-surgical management

Full size image

Pooled OR across 7/11 (64%) observational studies showed a 47% reduction ([95% CI: − 86–108%]; p = .36) in favor of VP/BKP. Similarly, pooled OR from 4/11 (36%) randomized studies confirmed the trend in favor of VP/BKP over NSM (− 29% [95% CI: − 71–70%]; p = .44). Morbidity events per study are summarized in Table 3.

Table 3 Recorded events for the analysis of the morbidity outcomes
Full size table

Secondary outcomes

Covariates associated with 12-month all-cause morbidity

This analysis was conducted across 11 studies [4, 17, 20, 21, 24,25,26,27,28,29, 32], and pooled results showed that most of the studied covariates (publication year [OR: 0.89; [95% CI: 0.67–1.17]; p = .36], patients’ mean age [OR: 1.01 [95% CI: 0.81–1.27]; p = .90], percentage of female patients [OR: 1.08 [95% CI: 1.00-1.16]; p = .04]; mean length of symptoms’ duration [OR: 1.05 [95% CI: 0.35–3.18]; p = .89]) negatively correlated with the relative risk, which means that the higher the OR, the lower the relative benefit of VP/BKP over NSM.

Twelve-month infection morbidity

The three studies [4, 21, 29] included for this analysis were quite heterogeneous (Q = 5.03; p = .08; I2 = 60%) mainly due to differences in publication years. The most recent study [4] was the most precise and closest (OR: 0.23 [95% CI: 0.05–1.18]; p = .08) to the pooled estimate of risk reduction (OR: 0.23 [95% CI: 0.02–2.54]; p = .23). No publication bias was detected (p = .51).

Morbidity at pooled follow-ups

Data on cardio-pulmonary morbidity [4, 18, 29, 30, 33] were moderately heterogeneous (Q = 5.54; p = .23; I2 = 28%) but concordant in showing the non-inferiority of VP/BKP over NSM, with a risk reduction in favor of PVP/BKP (OR: 0.42 [95% CI: 0.18–1.02]; p = .05).

Data on infection-related morbidity [4, 21, 29, 30, 33] were quite heterogeneous (Q = 12.42; p = .03; I2 = 60%); nevertheless, they confirmed the benefit of VP/BKP over NSM in reducing the risk for infection (OR: 0.44 [95% CI: 0.13–1.53]; p = .20), in line with results from the 12-month infection analysis.

Studies reporting estimates for new VCFs were numerous (n = 14) [4, 17,18,19,20,21, 24,25,26,27,28,29,30,31,32,33], although heterogeneous (Q = 25.52; p = .02; I2 = 49%), and showed a higher incidence of new VCFs after VP/BKP compared to NSM (OR: 1.06 [95% CI: 0.67–1.69]; p = .80).

Discussion

Pain relief is a very relevant clinical outcome to be evaluated after VP/BKP in patients with OVCFs, and has been proven to be significantly and rapidly improved following VP/BKP in OVCF patients provided adequate selection and evaluation criteria [3, 4, 21, 35]. Nevertheless, it is largely believed among clinicians that evaluation of “pain relief” exclusively does not adequately represent the overall clinical benefits provided by VP/BKP. This has also been highlighted by physicians from one lead site of one of the 2009 trial, who continued to perform VP after the publication of their study discouraging VP, since they firmly believed that the benefits provided by VP largely outweighed the risks [36]. To further evaluate this, we conducted a meta-analysis with the intent of evaluating clinical benefits other than “pain relief” provided by VP/BKP in OVCF patients. In particular, we have studied the impact of VP/BKP in terms of 12-month all-cause mortality and morbidity as compared to NSM. Our primary analyses were fixed at 12-month given the larger amount of data available at this time-point in the majority of the studies included in the quantitative analysis. Furthermore, morbidity was also studied at pooled follow-ups.

Our findings showed that compared to NSM, VP/BKP reduced the 12-months risk of all-cause mortality and morbidity by 19% and 36%, respectively. Moreover, VP/BKP reduced the risk of infections from any origin by 77% at 12-month and by 56% at pooled follow-ups ranging between 3 and 24 months. Similarly, VP/BKP showed protective tendency in terms of cardio-pulmonary events, whose risk was reduced by 58% at pooled follow-ups ranging between 3 and 24 months. Lastly, new VCFs were the most common adverse event following VP/BKP and NSM, and the risk of occurrence was 6% higher with the former treatment.

Despite our results favor VP/BKP over NSM for most of the studied endpoints, they failed to reach statistical significance. This was felt to be secondary to the relatively low number of recorded events for the assessment of primary outcomes, which is confirmed by the large CI. Such a paucity of events was likely attributed to the strict selection criteria applied, allowing only the inclusion of comparative studies (VP/BKP versus NSM) with 1–3 level on the Oxford scale. As a result, we could only include high-quality studies, although this had resulted in inclusion of studies with relatively limited sample sizes and short follow-ups. This was not the case for a similar meta-analysis that studied mortality after VP/BKP, and which included a larger number of studies with sample sizes larger and follow-ups longer than ours, due to the different and less strict inclusion criteria not requiring comparison to NSM [11]. Accordingly, Hinde et al [11] proved a significant reduction of the mortality risk after VP/BKP at 2- and 5-year follow-up (hazard ratio [HR] 0.70, and HR 0.79; respectively). Their results were mainly due to the inclusion of the study by Ong et al [15] that represents one of the largest population study performed, with more than 2 million of patients derived from the US Medicare data set. In their analysis using “real-world” data, Ong et al [15] noted a dramatic increase of the mortality risk for OVCF patients not receiving VA (i.e., BKP and VP cohorts had respectively a 19% and 7% lower propensity-adjusted 10-year mortality risk compared the NSM cohort). Furthermore, the same team has recently showed that treating 15 OVCF patients with VA is enough to save one life at 1 year [37].

Although the analysis by Ong et al [15] could not highlight the cause of mortality, when pneumonia was entered as primary or secondary diagnosis 90 days within death, the 10-year mortality risk was 21% and 3% higher in the NSM cohort compared to the BKP and VP ones, respectively [15]. Moreover, in another similar study pooling patients from the same data set, Edidin et al [14] confirmed that compared to BKP, NSM carries a significantly higher adjusted risk of pneumonia, along with higher risk of myocardial infarction and cardiac complications, deep vein thrombosis, and urinary tract infection. These findings are in line with ours demonstrating that the most commonly recorded events in terms of morbidity (other than secondary VCFs) were lung and urinary tract infections, and deep vein thrombosis with or without pulmonary embolism.

The higher morbidity associated with NSM may be directly related to the higher mortality in this group, and from a physio-pathologic stand point, all these morbidity events are substantially favored by prolonged patients’ immobilization, which is a well-known factor supporting systemic degradation of patients’ health status [38, 39], and it has been also associated with the occurrence of pneumonia in patients receiving late delayed mobilization (i.e., 48 h after surgery) after elective spinal surgery (8.47% versus 1.51% in patients being mobilized 24 h within surgery) [40]. Similarly, in patients with acute spinal trauma, MacCallum et al [41] demonstrated that compared to long (> 72 h) periods of immobilization before treatment, short (< 72 h) periods of immobilization significantly reduce the occurrence of urinary tract infections (14.5% versus 6.0%). Moreover, in this study, pneumonia and deep vein thrombosis were more common in the delayed immobilization group compared to the early one (18.4% versus 11.9% and 9.5% versus 4.0%, respectively).

Consequently, it is not surprising that patients receiving NSM, which is essentially based on analgesics, bed rest, and bracing, are more likely to be exposed to infectious or venous thrombotic events compared to those receiving VP/BKP, which in turns allows rapid patients’ mobilization (< 24 h) and hospital discharge (on the same day of the procedure or 1–2 days within it), mainly due to fast pain relief [3, 4, 21, 35]. Therefore, it is important to note that evaluation of the outcome “pain relief” after VA in OVCF patients should be reasonably assessed in the short term (i.e., 2–4 weeks) rather than in the long term (e.g., 12 months), when spontaneous fracture healing logically hides the advantages of treatment [42]. Conversely, VA should be provided in the acute setting [3, 4, 43], and especially to patients with pain lasting no more than 6 weeks. Furthermore, it has been reported that VA rapidly and significantly improves gait in patients with OCVFs [44], and reduces the long-term (4-year) cumulative costs of patients’ management, mainly due to reduced analgesic consumption [45].

Finally, a slightly (6%) increased risk of new VCFs after VA compared to NSM was noted. However, given the relative small risk shown in comparison to the high number of new VCFs requiring subsequent treatment in both VP/BKP and NSM cohorts, it may be reasonably concluded that these new VCFs are more likely related to the systemic involvement of the osteoporotic disease rather than the altered plasticity induced by cemented vertebrae to adjacent non-cement vertebrae [43, 46, 47]. This view also highlights that VA solely treats the clinically evident consequence (i.e., fracture) of osteoporosis and not the osteoporosis itself, and in fact, a large study including 650 patients has recently proven that older age (OR: 2.48) and lower bone mineral density (OR: 0.31) are risk factors favoring the development of new VCFs after PV/PKP, whereas outdoor activity (OR: 0.38) played a protective role [48]. In line with these findings, Cao et al [49] reported that primary factors associated with new VCFs after VP were low bone mineral density (standardized mean difference − 0.37), steroid usage (OR: 2.63), and multiple treated vertebrae (OR: 2.03). Therefore, with poor bone mineral density being identified as the leading factor favoring secondary VCFs after VA, clinical management should not be solely restricted to VA, but should also include adequate medical treatment of the underlying osteoporotic disease with the intent of reducing the rate of new post-VA VCFs [50]. In the end, it has been showed [51] that poor adherence to the anti-osteoporotic medical treatment in OVCF patients increases the risk of mortality (HR: 1.75) mostly due to increased rates of infection (HR: 4.56). Therefore, further studies are needed to prove whether systematic combination of VA and anti-osteoporotic medications may further reduce patients’ morbidity and mortality.

This study has some limitations. First, we had a low number of censored events accounting for the analysis of the primary outcomes. This occurred due to our research of studies solely performed on Medline and EMBASE and, most of all, due to our strict inclusion criteria. As a result, statistical significance was not reached in the primary outcome analysis.

Secondly, statistical results may depend on the classification we have adopted to group censored events for the morbidity analysis, and we cannot exclude that results may change with different grouping criteria, althought it seems highly implausible.

In conclusion, compared to NSM, VP/BKP confer a risk reduction of 12-month all-cause mortality and morbidity by 19% and 36%, respectively. Moreover, following VP/BKP, the 12-month risk of infection from any origin is reduced by 77%. Although these data favor VP/BKP over NSM, probably due to fast pain relief and rapid patients’ mobilization, cautious interpretation is needed due to absence of statistical significance.