Balloon kyphoplasty or percutaneous vertebroplasty for osteoporotic vertebral compression fracture? An updated systematic review and meta-analysis

Abstract

BACKGROUND

Both kyphoplasty (KP) and vertebroplasty (VP) are effective for patients with osteoporotic vertebral compression fracture (OVCF), but which approach might be more effective remains unclear, so we decided to update earlier systematic reviews.

OBJECTIVE

Review and analyze studies published as of August 2015 that compared clinical outcomes and complications of KP versus VP.

DESIGN

Systematic review and meta-analysis.

SEARCH METHOD

Published reports up to August 2015 were found in PubMed, EMBASE and the Cochrane Central Register of Controlled Trials (CENTRAL).

SELECTION CRITERIA

Randomized controlled trials (RCTs) and prospective and retrospective cohort studies comparing KP and VP in patients with OVCF.

DATA COLLECTION AND ANALYSIS

Two authors independently assessed the studies and extracted data.

RESULTS

Thirty-two studies involving 3274 patients fulfilled the inclusion criteria. There were significant differences between the two groups in short- and long-term postoperative changes in measures of pain intensity and dysfunction (P<.01), in anterior and middle height (P<.01), kyphotic angle (P<.01), and time to injury, but not in posterior height (P=.178). There were no significant differences in the rate of postoperative fractures including adjacent and total fractures, but cement leakage to the intraspinal space was greater in the VP group (P=.035). KP surgery took longer and required a greater volume of injected cement.

CONCLUSIONS

KR resulted in better pain relief, improvements in Oswestry dysfunction and radiographic outcomes with less cement leakage, but further RCTs are needed to verify this conclusion.

LIMITATIONS

Only four RCTs with a certain of risk of bias. Most studies were observational.

The increasing elderly population throughout the globe has brought increasing attention to osteoporosis, the most important cause osteoporotic vertebral compression fractures (OVCF).^1,2 OVCF has a prevalence of more than 30% in the population older than 65 years.³ OVCF is associated with acute and chronic pain, progressive spinal deformity, a decreased quality of life, impaired physical function and increasing mortality.^4–8

One method to treat OVCF is conservative non-surgical management (NSM) which consists of bed rest, use of painkillers, and bracing.⁹ However, NSM does not improve vertebral height¹⁰ or reverse kyphotic deformities, and has undesirable effects such as bedsores, bone demineralization and deep vein thrombosis. ¹¹ Since 1987, vertebroplasty (VP) and kyphoplasty (KP) with polymethylmethacrylate (PMMA) augmentation has been increasingly advocated as treatment for OVCF.^12,13 Both of these minimally invasive techniques increase bone strength and reduce pain. Recently, two randomized controlled trials (RCT) showed that both methods were effective in reducing immediate pain, unlike conservative treatment.^14,15 Several studies have shown that KP achieves better restoration of the kyphotic angle and vertebral height compared with VP^16–18 Furthermore, KP reduced the cement leakage rate compared with VP.^19,20

The comparative effectiveness and complications of KP and VP have been assessed in a few systematic reviews and meta-analysis, all which pooled randomized contolled trials with observational studies. This systematic review updates previous analyses.^21–25

PATIENTS AND METHODS

Literature search

We performed a comprehensive systematic computer-based literature search of published reports before August 2015 in PubMed, EMBASE and the Cochrane Central Register of Controlled Trials (CENTRAL). The reference lists of the selected studies were also searched. The search terms were: “kyphoplasty” or “KP” AND “vertebroplasty” or “VP” AND “vertebral fracture” AND “osteoporotic” or “osteoporosis”. We selected randomized controlled trials (RCTs) and prospective and retrospective cohort studies that compared KP with VP with no language restrictions. The protocol was not registered.

Inclusion/Exclusion criteria

The inclusion criteria were that studies be comparative studies (RCTs, prospective and retrospective cohort studies) comparing KP and VP in patients with OVCF. Outcomes had to include the postoperative time to injury, the duration of the operation, pain relief and quality of life, postoperative radiographic data and complications. Studies were excluded from our meta-analysis if they were of vertebral fractures caused by any etiology other than osteoporosis, including neoplastic or invasive, infective and traumatic fracture. Studies involving any type of cement other than PMMA cement were excluded.

Quality assessment and data extraction

RCTs were carefully assessed by two authors (LL and XLC) and any disagreement resolved through discussion. Determination of the risk of bias in the RCTs included the following key domains: adequate sequence generation, allocation concealment, blinding, incomplete outcome data, free from selective reporting, and free from other bias. The prospective and retrospective cohort studies were assessed by the methodological index for non-randomized studies (MINORS), a validated instrument designed to assess the quality of comparative or non-comparative non-RCT studies. LL and XLC independently extracted the data from each article with a standard data extraction form. The data included authors, year of publication, study design, age of population, gender, numbers of vertebral bodies, surgical procedures, duration of follow-up and outcomes parameters. The extracted data were analyzed by YYZ.

Clinical outcomes

Pain intensity and functional disability was measured using the visual analog scale (VAS) and the Oswestry Disability Index (ODI). Radiographic outcomes included the height of the vertebral body (anterior, middle and posterior) and the kyphotic angle. Complication outcomes were cement leakage and new vertebral fracture. Injury time, operation time and the volume of injected cement were also extracted from the reports.

VAS and ODI were extracted and summarized by short-term (less than one week) and long-term (more than six months) follow-up. We defined the short-term period as less than one week and the long-term period as no less than 6 months.²⁵ If there were several time points in the long-term follow-up, we selected the longest follow-up. We defined the postoperative period as the first day after surgery and improvement as any change between the preoperative and postoperative periods.

Complications

We classified cement leakage as any intraspinal and extraspinal leakage. Intraspinal leakage means that cement leaked into the intraspinal space, including the disc and vertebral body; if cement leaked into an extraspinal space such as the external venous plexus, epidural tissue or spinal canal, we considered that extraspinal leakage. Fractures included re-fracture of the same postoperative vertebral body and fractures of an adjacent vertebral body.

Statistical analysis

We performed all meta-analysis with Stata version 12.0 (StataCorp, College Station, TX). For dichotomous outcomes, the odds ratio (OR) and the 95% confidence interval (95% CI) were assessed. For continuous outcomes, means and standard deviations were pooled to a weighted or standardized mean difference (WMD or SMD), a weighting by the individual variances for each study, and the 95% CI. A probability of P<.05 was regarded as statistically significant. Statistical heterogeneity was assessed using Q statistics. Analysis of the outcomes was divided to subgroups according to the time or the region, if possible. For the variables - extraspinal and total leakage, adjacent and total new fracture, posterior height-postoperation, we used a fixed-effects model; for the rest, we used a random-effects model.

RESULTS

Study characteristics

Of 1300 titles and abstracts reviewed preliminarily, 32 met the inclusion criteria for the meta-analysis.^17,26–56 (Figure 1). They included 4 RCTs,^27,31,41,53 14 prospective cohort studiess,^{17,32,33,36,39,40,41–49,54} and 14 retrospective cohort studies^{26,28–30,34,35,37,38,42,50–52,55,56} (Figures 2 and 3). There were a total of 3274 patients; 1653 patients underwent the KP surgery and 1621 underwent VP surgery. Individual study sample sizes ranged from 41 to 381 patients. The demographic characteristics of patients are summarized in Table 1.

Figure 1.

Flow diagram for selection of articles in the meta-analysis..

Figure 2.

Methodological quality of the randomized controlled trials (n=4) showing risk-of-bias assessment.

Figure 3.

Summarization of risk of bias as percentages for low, unclear and high for the randomized controlled trials (n=4).

Table 1.

Patient demographic and study characteristics of the 32 studies in the meta-analysis.

Study	Country	Year	Study design	Patient numbers		Age (years)		Follow-up period (KP/VP) (months)	MINORS scores
				KP	VP	KP	VP
Bozkurt et al26	Turkey	2014	Retrospective	200	96	57.5	57	40	14
Dohm et al27	United States	2014	RCT	191	190	75.6		24	-
Dong et al28	China	2013	Retrospective	51	35	69.8	70.5	21.3	14
Dong et al29	China	2009	Retrospective	20	18	69.5	70.2	3	11
Ee et al30	England	2012	Retrospective	97	148	75	77	24	15
Endres et al31	Germany	2011	RCT	20	21	63.3	71.3	5.8	-
Figueiredo et al32	Brazil	2011	Prospective	22	30	73	77	6	16
Folman et al33	Israel	2011	Prospective	31	14	70.7	75.6	12	16
Frankel et al34	United States	2007	Retrospective	17	29	70	72	3.5 years	14
Gan et al35	China	2014	Retrospective	41	38	69.1	67.1	43.5/41.4	15
Grohs et al36	Austria	2005	Prospective	28	23	70	70	24	17
Hiwatashi et al37	Japan	2008	Retrospective	40	66	75	77	NR	13
Kong et al38	China	2014	Retrospective	29	24	71.9	70.5	12	13
Kumar et al39	Canada	2009	Prospective	24	28	73	78	42.3/42.2	17
Li et al40	China	2012	Prospective	45	40	68.5	67.1	12	17
Liu JT et al41	Taiwan	2009	RCT	50	50	72.3	74.3	>6	-
Liu T et al42	China	2013	Retrospective	40	60	68.5	62.5	1 week	13
Lovi et al43	Italy	2009	Prospective	36	118	67.6	33m	17
Movrin et al44	Slovenia	2010	Prospective	46	27	67.8	72.9	1 year	16
Omidi-Kashani45	Iran	2013	Prospective	29	28	72.1	72.4	6m	13
Pflugmacher et al17	Germany	2005	Prospective	22	20	67	65	12	15
Qian et al46	China	2012	Prospective	53	9	66.2	3.9y	16
Rollinghoff et al47	Germany	2009	Prospective	53	51	68.9	1y	17
Santiago et al48	Span	2009	Prospective	30	30	65.9	73	1 year	16
Schofer et al49	Germany	2009	Prospective	30	30	72.5	73.8	13.5/13.7	17
Sun et al50	China	2010	Retrospective	31	28	74.2	72.3	18	14
Wu et al51	China	2014	Retrospective	20	20	65.1	66.3	1 year	15
Yan et al52	China	2011	Retrospective	98	94	76.9	77.2	14.3/15.2	14
Yang et al53	Korea	2014	RCT	112	109	73.4	73.3	NR	-
Yi et al54	China	2014	Prospective	79	90	61.3	49.4m	16
Yokoyama et al55	Japan	2013	Retrospective	38	28	75.5	74	NR	12
Zhang et al56	China	2013	Retrospective	30	29	68.7	66.2	25	13

NR = not reported. RCT = randomized controlled trial. Follow-up period is months unless reported otherwise.

The MINORS criteria include the following items: (1) a clearly stated aim; (2) inclusion of consecutive patients; (3) Prospective data collection; (4) endpoints appropriate to the aim of the study; (5) unbiased assessment of the study endpoint; (6) a follow-up period appropriate to the aims of the study; (7) less than 5% loss to follow-up; (8) Prospective calculation of the sample size; (9) an adequate control group; (10) contemporary groups; (11) baseline equivalence of groups; and (12) adequate statistical analysis. The items are scored as follows: 0 (not reported); 1 (reported but inadequate); 2 (reported and adequate). The ideal global score for comparative studies is 24.

Clinical outcomes

Eighteen studies reported short-term follow-up VAS scores.^{17,28,30,31,33,35,36,38–42,47,49–52,55} There was a significant difference between KP and VP (WMD=−0.2, 95% CI=−0.27 to −0.63; P<.01). Long-term VAS scores were available from 14 studies.^{17,28,30,32,35,36,38,41,45,47,49,51} The pooled result also showed a significant difference between the two groups (WMD=−0.46, 95% CI=−0.57 to −0.36; P<.01) (Figure 4 and Table 2). Adequate data on short-term ODI scores was present in 7 studies^{17,31,35,38,39,47,50} and the difference in overall estimate was statistically significant (WMD−17.56, 95% CI=−18.07 to −17.05; P<.01). Eight studies provided long-term ODI data.^{17,30,35,36,38–40,47} There was a significant difference between KP and VP (WMD=−2.41, 95% CI= −3.44 to −1.38; P<.01) (Figure 5 and Table 2).

Figure 4.

Forest plots for the meta-analysis of the visual analog scale scores.

Table 2.

Meta-analysis of clinical outcomes comparing the KP and VP groups.

Outcomes	No. of studies	No. of patients	Effect estimate (95% CI)	P
Visual analog scale
Short-term	18	1500	−0.2 (−0.27, −0.13)	<.01
Long-term	14	1071	−0.46 (−0.57, −0.36)	<.01
Oswestry Disability Index
Short-term	7	430	−17.56 (−18.07,−17.05)	<.01
Long-time	8	676	−2.41 (−3.44, −1.38)	<.01
Injury time	4	311	−1.31 (−3.37, 0.75)	<.01
Operation time	5	716	6.58 (5.47, 7.68)	<.01
Volume of injected cement	12	1764	0.51 (0.44, 0.56)	<.01

The effect estimate is weighted mean difference, CI=confidence interval.

Figure 5.

Forest plots for the meta-analysis of the Oswestry Disability Index scores.

The dates of injury were available for four trials. ^40,41,48,55 The pooled results demonstrated no significant difference between the KP and VP group (WMD=−1.31, 95% CI=−3.37 to 0.75; P<.01). Five reports reported the mean and standard deviation for operation time.^{27,31,41,43,51} VP required less time for the surgical procedure (WMD=6.58, 95% CI=5.47 to 7.68; P<.01) than the KP group (Table 2). The reported volume of injected cement analyzed in 12 studies^{26,27,38,39,41,44–46,52,53,55,56} was greater in the KP group (WMD=0.51, 95% CI=0.44 to 0.56; P<.01) (Figure 6 and Table 2).

Figure 6.

Forest plots for the meta-analysis of the volume of injected cement.

Radiographic outcome

In the 14 studies that reported the postoperative anterior height of the vertebral body,^{17,26,28,30,35,36,41,43,47,48,50–53} there was a significant difference in the immediate postoperative follow-up period (WMD=2.55, 95% CI=2.33 to 2.78, P<.01), the final follow-up (WMD=2.79, 95% CI=2.39 to 3.19; P<.01) and improvement (WMD=5.91, 95% CI=5.19 to 6.64; P=<.01) between the KP and VP groups, respectively. Patients who underwent the KP procedure had a better post-operative anterior height of the vertebral body than those who had the VP procedure (Table 3).

Table 3.

Results of meta-analysis of radiological outcome measures.

Outcomes	No. of studies	No. of patients	Effect estimate (95% CI)	P
Anterior height
Postoperative follow-up	10	1020	2.55 (2.33, 2.78)	<.01
Final follow-up	6	505	2.79 (2.39, 3.19)	<.01
Improvement	4	797	5.91 (5.19, 6.64)	<.01
Middle height
Postoperative follow-up	4	386	2.44 (2.14, 2.73)	<.01
Final follow-up	3	275	6.92 (6.31, 7.52)	<.01
Posterior height
Postoperative follow-up	3	344	0.5 (−0.03, 1.02)	.178
Final follow-up	3	344	1.78 (1.44, 2.11)	.033
Kyphotic angle
Postoperative follow-up	15	1365	−2.5 (−2.84, −2.16)	<.01
Final follow-up	9	641	−1.7 (−2.06, −1.33)	<.01
Improvement	7	916	4.79 (4.19, 5.32)	<.01

Effect estimates are weighted mean difference, CI = confidence interval, postoperative means immediate postoperative follow-up period.

The pooled measures of middle height included the immediate postoperative follow-up period (WMD=2.44, 95% CI=2.14 to 2.73; P<.01) and the final follow-up (WMD=6.92, 95% CI=6.31 to 7.52; P<.01) in four^17,35,43,45 and three studies,^17,35,43 respectively. Both showed a significant difference and demonstrated that the KP group had a better result than the VP group for changes in anterior and middle vertebral height, but in three reports there was no significant difference in pooled posterior height between KP and VP^28,43,47 (WMD=0.5, 95% CI=−0.03 to 1.02; P=.178/WMD=1.78, 95% CI=1.44 to 2.11; P=.033) (Table 3).

The kyphotic angle in the immediate postoperative was analyzed in 15 studies.^{17,28,33,35,38,40,41,44,47,49–53,56} The kyphotic angle improved more in the KP group than in the VP group (WMD=−2.5, 95% CI=−2.84 to −2.16; P<.01). Nine studies^{17,28,35,38,40,47,49,51,56} reported the kyphotic angle at the final follow-up (WMD=−1.7, 95%CI=−2.06 to −1.33; P<.01) and seven studies^{29,30,44,49,52,53,55} compared the improvement (WMD=4.76, 95%CI=4.19 to 5.32; P<.01). With the KP procedure there was more improvement in the kyphotic angle than with the VP procedure (Figure 7 and Table 3).

Figure 7.

Forest plots for the meta-analysis of the kyphotic angle.

Complications

Cement leakage in the VP group was significantly more frequent than in the KP group in the intraspinal space (OR=0.5, 95% CI=0.3 to 0.85; P=.035)^{31,32,34,36,37,40,43–45,47,48,52,55} in the extraspinal space (OR=0.36, 95% CI=0.21 to 0.62; P=.15)^{31,32,34,36,37,40,43,45,47–49,52} and in total leakage (OR=0.53, 95% CI=0.4 to 0.7; P=.051) (Figure 8 and Table 4).^{26,27,30–32,34–36,39,40,40–45,47–53,55,57} Thirteen studies reported complications related to fractures.^{26,34–36,39–41,43–45,47,52,54} The pooled analysis showed no significant difference between the KP and VP group (OR=0.94, 95% CI=0.59 to 1.49; P=.248). Of these, there were nine reports of adjacent fractures.^{26,34,36,39,41,43–45,47} There was no significant difference between the groups (OR=1.41, 95% CI=0.7 to 2.83; P=.283) (Table 4).

Figure 8.

Forest plots for the meta-analysis of leakage.

Table 4.

Differences in complications between the VP and K groups.

Outcomes	No. of studies	No. of patients	Effect estimate (95% CI)	P
Leakage
Intraspinal	13	1503	0.5 (0.3, 0.85)	.035
Extraspinal	12	1223	0.36 (0.21,0.62)	.15
Total	22	2773	0.53 (0.4, 0.7)	.051
New fractures
Adjacent	9	1070	1.41 (0.7, 2.83)	.283
Total	13	1628	0.94 (0.59, 1.49)	.248

Effect estimates are weighted mean difference, CI = confidence interval.

DISCUSSION

Our systematic review and meta-analysis included 4 randomized studies and 28 non-randomized studies that included 1653 patients treated with KP and 1621 patients treated with VP. The main outcome variables were pain intensity and dysfunction measured by VAS and ODI, kyphotic angle, and vertebral height at short-term and long-term follow-ups. Postoperative complications included new vertebral and adjacent fractures, as well as time of injury and duration of surgery.

Treatment of OVCF should lead to a lasting improvement in the pain. More than 90% of pain and dysfunction caused by OVCF can be relieved successfully by KP or VP. Both surgical procedures significantly relieve the pain and improved dysfunction in patients with OVCF. In our analysis, KP was more effective on the VAS and ODI assessments than the VP group. The mechanism of pain reduction reflected in Oswestry score improvements might result from the inhibition and immobility of micro-movements of the fractured vertebral body, as well as the cytotoxic effect of the PMMA cement.^57–59

We pooled the improvement in kyphotic angle and height, which included the anterior, middle and posterior vertebral body. Improvements in postoperative anterior and middle height were better in the KP group in the immediate postoperative period and at the final follow-up. Improvements in posterior height were similar. One study reported that a reduction in the kyphotic angle depends more on natural healing than surgical treatment.⁶⁰ Schofer et al⁴⁹ reported a reduction in the kyphotic angle by a mean of 3–6° after the KP procedure compared with a reduction of 1°, suggesting that the balloon-induced restoration had a positive effect.

Total new vertebral fracture did not differ between the KP and VP groups. There was also no difference in the rate of adjacent fractures. Whether bone cement injection causes an increased incidence of new vertebral fractures is an interesting topic of ongoing discussion. Hulme et al²⁰ showed that the incidence of new vertebral fractures did not increase in osteoporotic patients who had suffered vertebral fractures. New vertebral fractures may relate to the sustained loss of bone mass seen in the osteoporotic population, rather than the surgical procedure itself.

Cement leakage does not usually result in clinical symptoms. In our experience, the high injection pressure and low viscidity of the cement leads to a higher incidence of cement leakage during VP than during KP. The KP procedure creates a hole in which to package the cement with the help of a balloon. The KP group had a lower frequency of leakage than the VP group in our analysis. The intraspinal and extraspinal leakage were greater in the VP group.

An ideal meta-analysis would include only RCTs with little heterogeneity. However, RCTs are rare for surgical procedures. Patients will not usually agree to partake in a randomized surgical option. Every surgeon has his personal specialty and chooses the preferable procedure according to the specific condition. Because of the lack of RCTs, we included prospective and retrospective cohort studies of high quality and designed a baseline form to collect demographic characteristics in a manner that would limit the risk of bias.

In conclusion, we found that the KP procedure was more effective in pain relief, physical functional improvement, improving restoration of vertebral height and kyphotic angle with reduced cement leakage, but the KP surgery took longer and required a greater volume of injected cement. The KP procedure has a higher cost of hospitalization. Additional RCTs are needed to confirm these conclusions and to select the best surgical procedure for patients with OVCF.