➢ The use of recombinant human bone morphogenetic protein-2 (rhBMP-2) during spine surgery is controversial, requiring surgeons to be intimately familiar with the available data.
➢ The use of rhBMP-2 during spine surgery has been associated with a number of complications, many of which can be mitigated with proper usage, including the application of an appropriate dose, use of associated carriers that help confine the extravasation protein, and timely recognition and treatment of any complications.
➢ The Yale Open Data Access Project provides an unbiased, comprehensive review of the available patient data from U.S. Food and Drug Administration investigational device exemption clinical trials and adds to our understanding of rhBMP-2 and its clinical effects.
The era of using bone morphogenetic protein (BMP) to augment spinal arthrodesis has been noteworthy in a number of ways. Since the U.S. Food and Drug Administration (FDA) approved the use of recombinant human bone morphogenetic protein-2 (rhBMP-2) in a titanium cage for anterior lumbar interbody arthrodesis in 2002, the widespread effects of its use have been well publicized, both in the evidence-based literature and in the lay press, with a variety of connotations. Initially perceived as a revolutionary, fail-safe bone-graft substitute that obviated the need for iliac crest bone graft, rhBMP-2 is now perceived as a potentially high-risk material and has been vilified in academic circles, the media, and conversations between surgeons and patients. This review will highlight the important historical milestones in the clinical and basic-science research on BMP and will assess the current state of its use in spine surgery.
The Use of Autograft
The use of iliac crest bone graft has long been the so-called gold standard for the augmentation of spinal arthrodesis. With the ability to obtain large amounts of graft through either an anterior or a posterior approach, the technique has evolved greatly to include a smaller exposure and less disruption to the cortex; some surgeons have even suggested backfilling of the harvest site with graft to help limit the destruction required to remove a portion of the patient’s anatomy1-3. Despite the improvement of bone-graft harvest technique, the procedure has been associated with a number of complications, including wound hematoma, sacral fracture, and infection, all of which have necessitated reoperation at times4-9. More importantly, the harvest of autograft has been associated with substantial pain for as long as two years after the operation10. Consequently, the stigma associated with use of autograft has affected patient opinions and has influenced the preoperative discussion regarding bone-graft options to achieve a successful fusion.
More recently, investigators have suggested that the pain and complications resulting from such a procedure have been overestimated11,12. Proponents of the use of autograft have claimed that graft harvest does not lead to extensive postoperative long-term pain if the procedure is performed with attention to preservation of the inner cortex, formation of a bone window, and proper wound closure13. Nevertheless, the number of autograft procedures performed in association with spinal arthrodesis has decreased steadily since 200014.
Recent studies have suggested that, aside from iliac crest bone graft, local bone graft from the site of an open decompression can provide a sufficient osteoinductive stimulus to achieve a successful spine fusion15-17. In fact, evidence on posterolateral and posterior interbody arthrodesis has shown that, as long as a threshold volume of local graft is available, results similar to those associated with the use of iliac crest autograft can be expected17. However, in many situations, the local bone graft that is harvested from the spinous process and lamina (<15 cc per level) is insufficient to achieve this fusion mass. Furthermore, with the increasing popularity of minimally invasive spine surgery, more procedures are being performed that do not provide access to local graft. In these situations, when an open iliac crest bone-graft harvest would defeat the purpose of minimally invasive surgery, spine surgeons are still searching for adequate bone-graft technology that would lead to a successful fusion.
Initially described by Dr. Marshall Urist18 in 1965 as the “bone autoinduction” principle and later identified as “bone morphogenetic proteins,”19 these factors have gained a substantial role in bone healing and spinal fusion enhancement. It was not until twenty-three years later, in 1988, that Wozney was able to molecularly clone rhBMP-2, widely increasing its availability20. After this development, another fourteen years passed before enough basic-science and clinical data were available to allow for FDA approval of the use of rhBMP-2 in a lumbar tapered titanium cage (LT-CAGE Lumbar Tapered Fusion Device; Medtronic Sofamor Danek, Memphis, Tennessee) for anterior lumbar interbody arthrodesis. The use of rhBMP-2, including for off-label indications, steadily grew after the initial FDA approval in 2002. In fact, it was estimated that, from 2002 to 2007, 85% of BMP-2 use, including that for posterolateral arthrodeses, was off-label21. While additional clinical studies were performed to investigate the use of rhBMP-2 in the posterolateral aspect of the spine, interbody regions of the lumbar spine, and anterior aspect of the cervical spine22-26, case series and case reports of adverse effects resulting from the use of rhBMP-2 in different anatomic environments also surfaced27-30. The culmination of these reports came in 2011, with the publication of a focus issue of The Spine Journal dedicated to the discussion of adverse effects of rhBMP-2 use, including its apparent associations with increased rates of cancer and retrograde ejaculation among exposed patients31. Shortly following the release of the focus issue, Medtronic Sofamor Danek commissioned the Yale Open Data Access Project for an independent review of individual patient data, industry-sponsored clinical trials, and published case series. These reports were released in June 201332,33 (Fig. 1).
More than twenty different types of BMPs have been identified, with BMP-2, 4, 6, 7, and 9 having the greatest osteogenic characteristics34. Members of the transforming growth factor (TGF)-β superfamily, BMPs promote bone regeneration through the initiation of signal transduction via the binding of a membrane-based heterodimeric complex of two transmembrane serine-threonine kinase receptors, BMP receptor types I and II. The activated receptor kinase then phosphorylates Smads 1, 5, and 8; these transcription factors translocate to the nucleus (with Smad 4), leading to the increased expression of target genes involved in osteogenesis35.
In vitro data suggest that BMPs facilitate osteogenic differentiation of mesenchymal stem cells through multiple pathways36. BMP-7-treated stem cells show enhanced secretion of osteogenic lineage-specific markers, including phosphoprotein-1, osteocalcin, and osterix37. Mesenchymal stem cells that are treated with BMP-2 in vitro demonstrate increased osteocalcin release, whereas neutralization of BMP-2 decreases alkaline phosphatase levels38.
The mechanisms of rhBMP-2 action that may have led to the complications detailed in case reports have been investigated. For example, a preclinical rodent model demonstrated that the in vivo host response to rhBMP-2 may be associated with circulating proinflammatory and osteoclastic cytokines such as tumor necrosis factor-α, macrophage inflammatory protein 1-alpha, and interleukin 1-β39. Furthermore, BMPs have been found to stimulate angiogenesis through the induction of vascular endothelial growth factor secretion40 and chemotaxis of circulating endothelial cell precursors41. This response may lead to the seroma formation that has been reported postoperatively42.
Finally, because certain cancer cell lines have been shown to have BMP receptors43, the local administration of this growth factor has led to the stimulation of cell growth of a number of cancer lines in vitro44. While preclinical animal models have demonstrated the increased growth rate of certain cancer lines when BMP is applied locally43,45,46, rhBMP-2 remains a member of the TGF-β family, which has demonstrated tumor suppression through the inhibition of the cell-cycle phase G1 and has been shown to decrease the tumor size of breast cancer cells after implantation in a preclinical rat spine metastasis model47.
The efficacy of rhBMP-2 to generate an osseous fusion mass has been well established in several preclinical spine models. Boden et al. reported successful fusion, as seen with computed tomography (CT), following laparoscopic anterior lumbar interbody arthrodesis with rhBMP-2 inside a titanium-threaded interbody fusion cage in five rhesus monkeys48. The authors of that study also described a dose-response phenomenon, with increased bone formation being associated with the use of a higher concentration of rhBMP-2. Unfortunately, the concentration required for a successful fusion differed depending on the species of animal used. Data from subsequent clinical trials led to a recommended concentration of 1.5 mg/mL for interbody fusion implants in humans49.
Preclinical animal models of posterolateral spinal arthrodesis also have demonstrated enhancement of fusion with use of rhBMP-2. Schimandle et al. performed posterolateral intertransverse process spinal arthrodesis at L5-L6 in fifty-six rabbits with use of different doses of rhBMP-2 delivered in a collagen carrier or with use of autogenous bone graft50. All of the rabbits that were treated with rhBMP-2 achieved solid spinal fusion as determined on the basis of manual palpation and radiographic evaluation, whereas only 42% of the rabbits in the autograft control group achieved fusion.
Martin et al. performed posterolateral intertransverse process arthrodesis at L4-L5 in twenty-nine rhesus monkeys with use of various doses and carriers of rhBMP-2 and found that tissue compression of the collagen sponge carrier with inadequate binding times could prevent bone induction at standard BMP doses51. The authors suggested that premature compression of the collagen sponge could cause extravasation of protein and dilution of the osteogenic effect. Consequently, the authors recommended that a sufficient rhBMP-2 loading time on the collagen carrier be used along with mechanical protection from the immediate soft-tissue compression. Suh et al. evaluated biphasic ceramic phosphate granules as an alternative carrier to collagen for rhBMP-2 in a posterolateral spinal arthrodesis at L5-L6 in six rhesus monkeys52. The authors demonstrated that the biologically compatible biphasic ceramic phosphate carrier matrices of rhBMP-2 induced fusion, resulting in stiffer fusion masses than autograft did without tissue compression of collagen sponge carriers.
Boden et al., in 2000, performed what we believe to have been the first prospective, randomized clinical study investigating the delivery of rhBMP-2 in an interbody fusion cage. In that study, fourteen patients with single-level degenerative disc disease that was refractory to nonoperative treatment were randomized to anterior lumbar interbody arthrodesis with use of a titanium cylindrical cage packed with rhBMP-2 on a collagen sponge (n = 11) or autogenous iliac crest bone graft (n = 3)49. All eleven patients who received rhBMP-2 had fusion, compared with two of the three patients who received autogenous bone graft. Burkus et al. subsequently performed a multicenter, prospective, randomized controlled study in which 279 patients underwent anterior lumbar interbody arthrodesis with use of either rhBMP-2 (study group) (n = 143) or autogenous iliac crest bone graft (control group) (n = 136); the overall fusion rate as seen with CT imaging was 94.5% in the rhBMP-2 treatment group, compared with 88.7% in the control group53. Clinical outcome measures (Oswestry Disability Index, back pain, and leg pain) were statistically similar between the two groups at the six, twelve, and twenty-four-month follow-up intervals; however, 32% of patients in the iliac crest bone-graft group reported graft-site discomfort at twenty-four months. FDA approval of the use of rhBMP-2 with an absorbable collagen sponge (INFUSE; Medtronic Sofamor Danek) in this clinical setting was granted in 2002.
Although preliminary studies demonstrated high fusion rates in association with the use of INFUSE in the posterolateral aspect of the spine54, many investigators postulated that INFUSE alone was not sufficient to form a large-enough osseous mass to withstand stresses in this region. Dawson et al. performed a prospective randomized study investigating the use of rhBMP-2 combined with a ceramic-granule bulking agent as a replacement for autogenous iliac crest bone graft for single-level posterolateral lumbar arthrodesis with instrumentation22. At twenty-four months, the rate of radiographic fusion was 95% (eighteen of nineteen) in the experimental group, compared with 70% (fourteen of twenty) in the iliac crest bone-graft group (p = 0.12). The data in that study suggest that rhBMP-2 with an accompanying “bulking agent” can serve as a replacement for autograft during single-level posterolateral lumbar arthrodesis with instrumentation.
Two years after the FDA approval of INFUSE for anterior lumbar interbody arthrodesis, the first case series and reports detailing unexpected adverse events related to the use of rhBMP-2 were published. Many of the reported complications involved anatomic areas that were not specified in FDA approval23-25, such as the cervical spine, posterior interbody region of the lumbar spine, and posterolateral aspect of the lumbar spine. Consequently, rhBMP-2 use is not recommended unless it is used in an on-label setting that has been cleared by the FDA (i.e., for anterior lumbar interbody arthrodesis with use of a titanium lumbar tapered cage).
Baskin et al. reported that the use of rhBMP-2 in the interbody space of the anterior aspect of the cervical spine was associated with 100% fusion rates at six, twelve, and twenty-four months of follow-up26; however, other investigators have described greater numbers of devastating complications, including postoperative prevertebral swelling, hematoma formation, and dysphagia, in association with the use of rhBMP-2 as compared with iliac crest bone graft27,28. Furthermore, Yaremchuk et al. reported that the rates of reintubation for the treatment of acute airway obstruction were significantly greater for patients managed with rhBMP-2 than for those managed with autograft (p = 0.003)29. The results of these and other, similar studies led to an FDA-administered warning in 2008 against the use of rhBMP-2 in the anterior aspect of the cervical spine55. Because the fusion rates following a one or two-level anterior cervical arthrodesis with use of bone-graft substitutes such as allograft are historically high, many surgeons have questioned the routine use of rhBMP-2 in this anatomic region for any reason. These reports indicate that the potentially devastating risks associated with the use of INFUSE may outweigh any benefits. While we are not aware of any study that has directly linked the dose-response relationship with the rate of complications in the cervical region30, it is commonly thought among surgeons and scientists alike that the use of higher doses increases the likelihood of an adverse event. In any case, on the basis of the available evidence, involvement of the anterior aspect of the cervical spine is a relative contraindication for the use of rhBMP-2.
Although Level-I evidence is not available, the results of several clinical studies have suggested that the use of rhBMP-2 for posterior cervical arthrodesis is associated with higher rates of infection, wound seromas, and reoperation56,57. Some surgeons have chosen to use INFUSE for upper cervical spine arthrodeses because of historically higher rates of pseudarthrosis in this region than in the subaxial region; however, complications have been documented in this area as well58. The potential clinical utility of rhBMP-2 in the posterior aspect of the cervical spine is unclear.
A number of complications, including radiculitis, vertebral body resorption, seroma and/or hematoma formation, and heterotopic ossification, have been reported following posterior lumbar interbody arthrodesis and transforaminal interbody arthrodesis23-25,42,59,60. Although radiculitis can be caused by nerve-root retraction and/or the host inflammatory response resulting from the use of rhBMP-2, patients managed with rhBMP-2 for transforaminal lumbar interbody arthrodesis or posterior lumbar interbody arthrodesis have demonstrated a higher rate of radiculitis61; some of them had chronic radiculitis. Mitigating techniques such as the use of a hydrogel sealant in the disc space have been suggested to limit the rate of postoperative pain61, and some investigators have postulated that complications can be minimized with the use of smaller doses61-63.
Anterior lumbar interbody arthrodesis with use of INFUSE has been associated with osteolysis and retrograde ejaculation. Pradhan et al. demonstrated a high rate of osteolysis and pseudarthrosis (five [56%] of nine patients) when rhBMP-2 was used in a femoral ring allograft without posterior instrumentation, necessitating reoperation in many cases64. Carragee et al. reported a higher rate of retrograde ejaculation among sixty-nine patients who underwent anterior lumbar interbody arthrodesis with the use of rhBMP-2 than among 174 patients who underwent the procedure without the use of rhBMP-265. The association between retrograde ejaculation and rhBMP-2 use is controversial, as this complication is known to be affected by the technique and exposure used to access this area. For example, Sasso et al. reported higher rates of retrograde ejaculation following transperitoneal exposure as compared with retroperitoneal exposure66. Furthermore, because this adverse event has been only recently scrutinized, detection methods for retrograde ejaculation have only just been analyzed. For instance, Tepper et al. demonstrated in 2013 that the use of sperm counts in semen both preoperatively and postoperatively to objectively identify retrograde ejaculation did not correlate with information obtained from patient questionnaires67, which historically have been used as the gold standard to identify this complication. Nonetheless, despite other clinical studies that have demonstrated no association between rhBMP-2 use and retrograde ejaculation in patients managed with anterior lumbar interbody arthrodesis67,68, because of the strong study design and large sample size in the study by Carragee et al.65, this potential risk should be discussed thoroughly, especially with young male patients, prior to clinical use of rhBMP-2.
Finally, perhaps the most concerning and devastating potential complication associated with the use of rhBMP-2 for spine surgery is the formation of new cancers. Because BMP is a factor that leads to the proliferation of cell growth when placed in certain biologic environments, the theoretical risk for cancer exists in vivo. It appears that the potential association between rhBMP-2 use and cancer formation is multifactorial, including the type of cancer cell, the phase of growth, and the anatomic area involved. Carragee et al. reviewed the data on the use of rhBMP-2 from industry-sponsored publications and FDA summaries and concluded that there was a significantly higher rate of new cancers among patients who had been exposed to this growth factor (3.8% compared with 0.89%; p = 0.5 to 0.1)31. The authors also pointed out that because cancer is a catastrophic complication, the typical accepted level of significance (p < 0.05) may not be applicable and instead should be altered to p < 0.1 to accommodate for such occurrences in the data. For this reason, Carragee et al. concluded that exposing patients to rhBMP-2 increases the risk of malignant neoplasm31.
The controversy regarding these findings stems from the fact that cancer formation as an adverse event is a very complex issue that involves many variables, most of which are not familiar to orthopaedic surgeons. For example, grouping all cancers together and reporting their collective incidence may not represent the true nature of these cells because of inherent differences in cell behavior, molecular interactions, and the mechanism of metastasis. For this reason, a subsequent study on this topic has focused on the twenty-four invasive cancer types as defined in the National Cancer Institute’s Surveillance Epidemiology and End Results (SEER) program, which are commonly utilized to report cancer incidence69. Furthermore, other factors, such as the stage of development of cell types, greatly affect the development of clinically important tumors. For example, a stage-IV metastatic pancreatic cancer that is diagnosed four weeks after growth-factor exposure would have a much weaker association than would a local soft-tissue sarcoma that is diagnosed two years after exposure to rhBMP-2 in that particular area. Grouping both of these occurrences together may cloud any statistical association found in a clinical study. Finally, the mechanism of action by which a growth factor leads to mutagenesis (the creation of a new cancer) is vastly different from that associated with metastasis (the propagation of an existing cancer). Any such mechanism would be important to identify; however, because of the limited nature of the existing detection methods for early cancer, this exact mechanism is largely unknown.
The debate regarding the relative risk between rhBMP-2 use and increased cancer incidence is currently unresolved. While data from the FDA investigational device exemption trials as reported by Carragee et al.31 are important to consider, authors of other studies, including meta-analyses and systematic reviews, have concluded that no association with cancer exists44,70,71. Notably, none of these studies (including that by Carragee et al.) were designed to investigate cancer incidence. These adverse events were reported in studies designed to detect differences in the fusion rate. This risk should be carefully evaluated and considered with future studies involving rhBMP-2.
The Yale Open Data Access Project
In response to both the frequency of rhBMP-2-related complications and the discussion regarding these complications, Medtronic Sofamor Danek, in an unprecedented decision in the medical-device and pharmaceutical industry, commissioned an independent review of the available patient data in industry-sponsored clinical trials involving INFUSE through the Yale Open Data Access (YODA) Project in 2011. In June 2013, two studies in the Annals of Internal Medicine that had been performed by research groups at different institutions (University of York and Oregon Health & Science University [OHSU]) presented an unbiased, comprehensive review of individual patient data, patient data from clinical trials sponsored by Medtronic Sofamor Danek, and selected case series and case reports from the peer-reviewed literature32,33. Both research groups had access to the same data sets, and each was allowed to determine its own inclusion and exclusion criteria, statistical methods, and conclusions.
In both studies, there was substantial agreement on the majority of conclusions derived from the data. For example, both sets of authors concluded that while rhBMP-2 use led to high fusion rates in the lumbar spine, there were no significant differences in clinical outcomes when compared with those for patients managed with iliac crest bone graft (the control group). Both studies demonstrated a significantly higher incidence of leg and/or back pain in the immediate postoperative period (at six weeks) among patients who received rhBMP-2; however, no other significant differences were noted between the groups in terms of any of the postoperative adverse events, including retrograde ejaculation. There was agreement that the use of rhMBP-2 in the anterior aspect of the cervical spine posed a clear safety risk. Nonetheless, although the York study demonstrated a nonsignificantly higher incidence of cancer in the rhBMP-2 group, the OHSU study demonstrated a significantly higher incidence of cancer at the twenty-four-month time point32. The research groups at both institutions ultimately concluded that although there may be an increased risk of neoplasm after exposure to rhBMP-2, any absolute risk is small.
The two different reports highlight the importance of the rigors of statistical analysis and demonstrate how minor variations in inclusion criteria can markedly change the results. For example, the OHSU report included one additional study72 in the calculation of cancer incidence that the York group excluded. That study72 included a comparison group that was managed with a Maverick Total Disc Arthroplasty System (Medtronic Sofamor Danek) that the York group believed was not appropriate for inclusion because the rest of the control groups received iliac crest bone graft. Consequently, the inclusion of this one study changed the statistical analysis, leading the OHSU group to report a significantly higher incidence in new cancer associated with rhBMP-2 exposure that was not revealed in the data analysis performed by the York group32.
It is important to point out that the design of the studies that were included ultimately led to the conclusions made by the YODA investigators. For example, the industry-sponsored FDA trials were statistically powered to demonstrate clinical outcomes equivalent, and not superior, to autogenous bone. Consequently, conclusions of equivalent outcomes between iliac crest bone graft and rhBMP-2 demonstrate the success of the investigational group and not failure.
The contrasting reports also show how different conclusions can be drawn from the same available data. Although both groups agreed that rhBMP-2 does not significantly improve clinical outcomes at the twenty-four-month follow-up, Simmonds et al. (the York group) concluded that rhBMP-2 use increases fusion rates without significantly increasing the risk of postoperative complications, other than early postsurgical pain at the six-week time point33. On the other hand, Fu et al. (the OHSU group) editorialized that “on the basis of the currently available evidence, it is difficult to identify clear indications for rhBMP-2 in spinal fusion.”32
The accumulation of data on rhBMP-2 provides a number of key lessons that spine surgeons can apply to everyday practice. First, while rhBMP-2 use unquestionably leads to high fusion rates, it is not required routinely for patients with spinal disorders. Beyond the less-popular practice of using iliac crest bone graft, there are a number of other options—such as the use of allograft for anterior cervical discectomy and arthrodesis73,74, the use of local bone graft for single-level lumbar posterolateral arthrodesis16,75, and the use of bone-graft extenders (synthetic carriers and demineralized bone matrix) when the volume of bone graft is insufficient76—that have been shown, in the evidence-based literature, to lead to high fusion rates. Second, the irresponsible use of rhBMP-2 can lead to devastating complications, as indicated by the studies involving the anterior aspect of the cervical spine. Third, health-care practitioners must be aware of the potential complications that have been reported to be associated with rhBMP-2 use, including heterotopic ossification, wound seromas, and retrograde ejaculation. Even if the data do not demonstrate a significantly greater incidence of these complications, awareness of these complications can lead to timely and proper treatment if problems are encountered.
With these points in mind, it is important to keep these data in perspective as one evaluates the future of rhBMP-2 use in spine surgery. Well before the YODA project began, surgeons and researchers were calling for a judicious algorithm for the use of rhBMP-2 in patients. For example, many have supported the use of rhBMP-2 only in cases in which the risk of pseudarthrosis is high (such as in smokers) or when iliac crest bone-graft harvest is not available (such as after previous surgery). Furthermore, with the increasing popularity of minimally invasive spine surgery among patients, there are cases in which local bone graft is not available and the prospect of an open iliac crest bone-graft harvest is unacceptable to the patient. In these cases, rhBMP-2 may prove to be the best option to lead to a solid fusion and improved pain scores. Consequently, the statement by Fu et al. that “…it is difficult to identify clear indications for rhBMP-2 in spinal fusion,”32 which is based solely on the information from individual patient data in investigational device exemption trials, may not be correct. The studies that were included in the meta-analysis were not designed to include real-life, individual scenarios in which existing bone-graft options are not available or not sufficient and when rhBMP-2 may present the best option.
As the spine surgery community moves forward with this new information, it is important to keep all data in perspective. rhBMP-2 must be recognized for what it is: a potent osteoinductive growth factor that has allowed surgeons to manage patients who have life-altering spinal disorders and that also can lead to harm if used irresponsibly. Although there are strong conflicting opinions on both sides of the debate regarding the exact indications for rhBMP-2 use in spine surgery, few would disagree that there is at least some role for such a powerful technology. The spine community and the patients we manage must remain informed about the profile of this bone-graft substitute that can offer benefit if used cautiously and responsibly.
Source of Funding: There was no external funding source for this study.
Investigation performed at the Department of Orthopaedic Surgery, Feinberg School of Medicine, Northwestern University, Chicago, Illinois
Disclosure: The author did not receive payments or services, either directly or indirectly (i.e., via his institution), from a third party in support of any aspect of this work. The author, or his institution, has had a financial relationship, in the thirty-six months prior to submission of this work, with an entity in the biomedical arena that could be perceived to influence or have the potential to influence what is written in this work. The author has not had any other relationships, or has not engaged in any other activities, that could be perceived to influence or have the potential to influence what is written in this work. The complete Disclosures of Potential Conflicts of Interest submitted by authors are always provided with the online version of the article.
- Copyright © 2014 by The Journal of Bone and Joint Surgery, Incorporated