BACKGROUND

Lumbar spinal fusion is a common surgical treatment for many degenerative, traumatic, deformity, and destructive conditions of the lumbar spine. It has been shown to improve outcomes in disabling conditions such as spinal stenosis, degenerative and isthmic spondylolisthesis, and degenerative disc disease (DDD).¹ With advances in diagnostic tools, fusion techniques, subspecialty training, raised patient and physician awareness, and better understanding of the causes and consequences of these disorders, as well as their management, there has been a dramatic increase in overall rates of spinal fusion.² Spinal fusion surgery can be technically demanding, surgical costs are high, and complications can be significant.³ Population-based reoperation rates after spinal fusion in the early 1990s were 10% to 20%.^4,5 Nonunion of the spine is a common reason for pain and reoperation after spinal fusion.⁶ Harvesting iliac crest autograft for spinal fusion is another common causeof persistent pain and reduced functional outcome after lumbar spinal fusion surgery.^7–9

Bone morphogenetic proteins (BMPs), first identified by Urist in 1965,¹⁰ consist of protein components of bone matrix that can be extracted to induce the differentiation of osteoprogenitor cells into bone-producing osteogenic cells, thus stimulating the production of new bone.¹¹ Further research has been able to identify many different BMPs, of which BMP-2, BMP-6, and BMP-7 appear to have the most important roles.¹²

Earlier preclinical and clinical studies demonstrated that BMPs increase fusion rates. In a comprehensive review, Sandhu suggests that recombinant BMPs can be used as substitutes for autograft, and that in some circumstances, their efficacy for inducing fusion is superior.¹³

The rationale for BMPs in spinal fusion surgery is to avoid the complications of autogenous iliac crest bone graft harvesting, to reduce the risks for nonunion and reoperations, and to improve functional outcomes for patients. The main objective of this systematic review was to determine whether BMPs improve spinal fusion. In particular, do BMPs reduce the risk for nonunion and improve clinical outcomes for patients undergoing lumbar spinal fusion?

SYSTEMATIC REVIEW

From our systematic review (Table 9-1), we identified 75 articles, of which 14 were considered potentially relevant from abstract review. One additional published randomized, controlled trial (RCT) was identified from a manual bibliography search; however, abstract review excluded it based on inadequate length of follow-up.¹⁴ Hand searches of major meeting proceedings identified an additional RCT; however, this study has not yet been submitted for publication. The 14 potentially relevant trials were retrieved in full for data abstraction.^15–28 The baseline characteristics of these trials are presented in Table 9-2. Once all data were abstracted, the trials were then reviewed in an unblinded fashion. It became apparent that many of them included the same randomized patients presented in different publications.

TABLE 9-1 MEDLINE and EMBASE Search Strategies

We attempted to consolidate the trials that included the same patients to avoid over-reporting and to obtain mutually exclusive trials. Two trials included patients with DDD who underwent anterior lumbar interbody fusion (ALIF) with tapered cylindrical cages randomized to BMP-2 or autograft.^17,19 One trial included 45 patients at 1 center, whereas the larger of the two reported on 279 patients at 16 investigational sites. Because follow-up was similar in both trials and the smaller trial reported only radiographic outcomes, we decided to include only the larger of the two trials.¹⁷

Three trials included patients with DDD who underwent ALIF using threaded cortical allografts randomized to BMP-2 or autograft.^{18,20, 21} The smallest trial reported a pilot series that was reanalyzed in the two other larger studies. Of the larger studies, only one of them reported radiographic outcomes and focused on the healing patterns, whereas the other reported both clinical and radiographic outcomes for the entire cohort. Thus, we included only the larger trial with both clinical and radiographic outcomes.²⁰

Two trials of patients with degenerative spondylolisthesis who underwent posterolateral noninstr-umented fusion randomized to BMP-7 or autograft were identified.^27,28 One was a longer follow-up (to 24 months) of the same cohort and, therefore, the one only included for analysis.²⁸ Finally, two trialsof patients with DDD who underwent posterolateral fusion with pedicle screw fixation randomized to BMP-2 or autograft were identified.^22,23 The smaller of the two reported only radiographic results for a single institution of a multicenter trial with 12-month follow-up. The larger trial that included clinical and radiographic outcomes at 24 months for two of the investigating centers was retained for analysis.²² In all cases of trials that were eliminated, information abstracted was used to complete missing information for those retained. Therefore, our results are reported in detail for the nine “mutually exclusive” trials.^{15–17,20,22,24–26,28}

Overall, descriptions of randomization methods were poor. Only one of the nine trials (Burkus and colleagues, 2005²⁰) described an adequate method of randomization of subjects. One other trial (Johnsson and coworkers, 2002²⁵) indicated that randomization was blind to patient and surgeon; however, only until the time of surgery. Also, the method of randomization was not discussed to ensure allocation concealment. The other eight trials had unclear or no discussion of randomization methods.

Six of the nine trials had over 24-month follow-up data, and three trials had at least 12-month follow-up data. Four trials had near-complete follow-up with minimal dropouts. One trial (Vaccaro and coauthors, 2005²⁸) had incomplete 24-month follow-up data but imputed the missing values with 36-month data. One trial (Haid and colleagues, 2004²⁴) had unclear accounting of the final numbers of patients analyzed, although the discrepancy was relatively small. Two trials (Burkus and colleagues, 2002¹⁷; Burkus and colleagues, 2005²⁰) failed to perform intention-to-treat analyses and had removed patients who underwent reoperations from subsequent analyses. One other trial (Dimar and investigators, 2006²²) had a follow-up rate of less than 67% without a proper account of the dropouts.

Patients and investigators were not blinded to treatment in any of the trials at the time of outcome assessment. Radiologists and orthopedic surgeons who assessed radiographic fusion were blinded in seven of the trials. Two trials (Johnsson and coworkers, 2002²⁵; Kanayama and colleagues, 2006²⁶) did not mention whether the radiologists were blinded to treatment assignment at the time of their assessment.

Eight of the trials provided validated patient-oriented clinical outcome measures (ODI, 36-Item Short Form Health Survey [SF-36]). Six trials reported both outcome measures, and two of them (Burkus and colleagues, 2002¹⁷; Kanayama and colleagues, 2006²⁶) reported only one (Oswestry Disability Index [ODI]). One trial (Johnsson and coworkers, 2002²⁵) had no validated patient-oriented clinical outcome measure. Seven of the nine trials reported on adverse surgical events, and eight of the nine trials reported on the need for reoperation.

Four of the nine trials evaluated interbody fusion, all using BMP-2 in patients with DDD, and five evaluated posterolateral fusions with both BMP-2 and BMP-7 in patients with varying diagnoses. Of the interbody fusions, three utilized an anterior approach with two trials of metal ALIF cages and one using an allograft dowel. The fourth interbody fusion trial investigated posterior lumbar interbody fusion (PLIF) as a stand-alone construct.

Of the five posterolateral fusions, two trials used BMP-2 for patients with DDD in instrumented fusions. The other three trials all used BMP-7. One was in noninstrumented fusions in patients with isthmic spondylolisthesis. Two trials evaluated BMP-7 in posterolateral fusions in patients with degenerative spondylolisthesis, one noninstrumented and one instrumented. Commercial funding sources and/or conflicts of interest were reported for all of the trials. Commercial funding solely designated to research and education was reported in only one of the trials (Kanayama and colleagues, 2006²⁶).

EVIDENCE

A summary of our findings is shown in Table 9-3. The evidence is presented for three main clinical applications. The first is BMP-2 in interbody fusion for patients with DDD (four trials). The second application is the use of BMP-2 in posterolateral fusion for patients with DDD (two trials), and the final is the use of BMP-7 in posterolateral fusion for patients with spondylolisthesis (three trials).