0
Reviews |

Safety and Effectiveness of Recombinant Human Bone Morphogenetic Protein-2 for Spinal Fusion: A Meta-analysis of Individual-Participant Data FREE

Mark C. Simmonds, PhD, MA; Jennifer V.E. Brown, MSc, BA; Morag K. Heirs, MSc, MA; Julian P.T. Higgins, PhD, BA; Richard J. Mannion, PhD; Mark A. Rodgers, MSc, BSc; and Lesley A. Stewart, PhD, MSc, BSc
[+] Article and Author Information

From the Centre for Reviews and Dissemination, University of York, York, and MRC Biostatistics Unit and Addenbrookes Hospital, Cambridge, United Kingdom.

Note: Annals peer review materials (original and revised manuscripts and communications, including peer reviewer, editorial, statistical, and author comments) are available at www.annals.org (see Supplement 2).

Disclaimer: This study used data obtained from the YODA Project that included 17 clinical trials of rhBMP-2 funded by Medtronic. The analysis, interpretation, and reporting of these data were completed independently on an unrestricted basis, are solely the responsibility of the authors, and do not necessarily represent the official views of the YODA Project.

Acknowledgment: The authors thank Leah Carreon for providing the IPD from the trial by Glassman and colleagues (18), Kath Wright for performing the literature search, and Charlotte Seneschall for assistance in extracting data for the safety evaluation.

Grant Support: By a competitive grant from the YODA Project.

Potential Conflicts of Interest: Disclosures can be viewed at www.acponline.org/authors/icmje/ConflictOfInterestForms.do?msNum=M12-2603.

Reproducible Research Statement: Study protocol: Available in Supplement 1. Statistical code: Not available because of large voulme. Data set: Available from the YODA Project (http://medicine.yale.edu/core/projects/yodap/index.aspx).

Requests for Single Reprints: Lesley A. Stewart, PhD, MSc, BSc, Centre for Reviews and Dissemination, University of York, York YO10 5DD, United Kingdom.

Current Author Addresses: Drs. Simmonds, Higgins, and Stewart; Ms. Brown; Ms. Heirs; and Mr. Rodgers: Centre for Reviews and Dissemination, University of York, York YO10 5DD, United Kingdom.

Dr. Mannion: Cambridge University Hospitals NHS Foundation Trust, Hills Road, Cambridge CB2 0QQ, United Kingdom.

Author Contributions: Conception and design: M.C. Simmonds, J.P.T. Higgins, R.J. Mannion, M.A. Rodgers, L.A. Stewart.

Analysis and interpretation of the data: M.C. Simmonds, M.K. Heirs, J.P.T. Higgins, R.J. Mannion, M.A. Rodgers, L.A. Stewart.

Drafting of the article: M.C. Simmonds, J.V.E. Brown, M.K. Heirs, R.J. Mannion, M.A. Rodgers, L.A. Stewart.

Critical revision of the article for important intellectual content: M.C. Simmonds, M.K. Heirs, J.P.T. Higgins, R.J. Mannion, M.A. Rodgers, L.A. Stewart.

Final approval of the article: M.C. Simmonds, J.V.E. Brown, J.P.T. Higgins, R.J. Mannion, M.A. Rodgers, L.A. Stewart.

Statistical expertise: M.C. Simmonds, J.P.T. Higgins, L.A. Stewart.

Obtaining of funding: J.P.T. Higgins, R.J. Mannion, L.A. Stewart.

Administrative, technical, or logistic support: R.J. Mannion, M.A. Rodgers.

Collection and assembly of data: M.C. Simmonds, J.V.E. Brown, M.K. Heirs, M.A. Rodgers.


Ann Intern Med. 2013;158(12):877-889. doi:10.7326/0003-4819-158-12-201306180-00005
Text Size: A A A

Background: Recombinant human bone morphogenetic protein-2 (rhBMP-2) is widely used to promote fusion in spinal surgery, but its safety has been questioned.

Purpose: To evaluate the effectiveness and safety of rhBMP-2.

Data Sources: Individual-participant data obtained from the sponsor or investigators and data extracted from study publications identified by systematic bibliographic searches through June 2012.

Study Selection: Randomized, controlled trials of rhBMP-2 versus iliac crest bone graft (ICBG) in spinal fusion surgery for degenerative disc disease and related conditions and observational studies in similar populations for investigation of adverse events.

Data Extraction: Individual-participant data from 11 eligible of 17 provided trials sponsored by Medtronic (Minneapolis, Minnesota) (n = 1302) and 1 of 2 other eligible trials (n = 106) were included. Additional aggregate adverse event data were extracted from 35 published observational studies.

Data Synthesis: Primary outcomes were pain (assessed with the Oswestry Disability Index [ODI] or Short Form-36), fusion, and adverse events. At 24 months, ODI scores were 3.5% lower (better) with rhBMP-2 than with ICBG (95% CI, 0.5% to 6.5%) and radiographic fusion was 12% higher (CI, 2% to 23%). At or shortly after surgery, pain was more common with rhBMP-2 (odds ratio, 1.78 [CI, 1.06 to 2.95]). Cancer was more common after rhBMP-2 (relative risk, 1.98 [CI, 0.86 to 4.54]), but the small number of events precluded definite conclusions.

Limitation: The observational studies were diverse and at risk of bias.

Conclusion: At 24 months, rhBMP-2 increases fusion rates, reduces pain by a clinically insignificant amount, and increases early postsurgical pain compared with ICBG. Evidence of increased cancer incidence is inconclusive.

Primary Funding Source: Yale University Open Data Access Project.


Recombinant human bone morphogenetic protein-2 (rhBMP-2) is widely used as an alternative to iliac crest bone graft (ICBG) to promote fusion in spinal surgery (12). Since the U.S. Food and Drug Administration (FDA) approved rhBMP-2 for anterior lumbar interbody fusion (ALIF) surgery (3), its use has grown rapidly, including off-label indications (2, 4). A review of publicly available data suggesting that the risk for adverse events is 10 to 50 times higher than reported in trial publications (5) raised concerns about the safety of rhBMP-2.

The Yale University Open Data Access (YODA) Project team invited Medtronic (Minneapolis, Minnesota) to provide full data from all of its trials of rhBMP-2 to allow independent reanalysis. The project team subsequently invited proposals to undertake independent evaluation and funded the Centre for Reviews and Dissemination and 1 other group to do so, thus enabling meta-analysis of individual-participant data (IPD), which is regarded as a “gold standard” approach to evidence synthesis.

We embedded our IPD meta-analysis within a systematic review and sought to examine all relevant evidence. Investigation of comparative effectiveness was restricted to randomized, controlled trials (RCTs). In addition, to investigate the safety of rhBMP-2, we sought all observational studies of its use in spinal surgery that reported adverse events. Findings of our investigation of reporting bias are presented in our full report and have been submitted for publication elsewhere (112).

Eligibility, Search, Data Collection, and Critical Assessment

Methods were prespecified (in advance of detailed knowledge of the IPD to be provided) in a protocol (Supplement 1) that was registered in PROSPERO in February 2012 (CRD42012001907) (6).

All RCTs that compared rhBMP-2 with ICBG in spinal fusion surgery regardless of spinal level or surgical approach were eligible for inclusion in our principal analysis. We included trials of the licensed INFUSE formulation (rhBMP-2 concentration, 1.5 g/L) and unlicensed AMPLIFY (rhBMP-2 concentration, 2 g/L) and biphasic calcium phosphate (BCP) formulations. For supporting analyses of safety, we included all studies of more than 10 adult participants that compared rhBMP-2 with any other spinal fusion technique and reported adverse events.

Our reanalysis of the Medtronic data was done in the context of a full systematic review. We performed a systematic literature search of the Cochrane Central Register of Controlled Trials, MEDLINE, EMBASE, and Science Citation Index in January 2012 and automated “current awareness” searches up to June 2012 to identify eligible studies not provided by Medtronic. We also searched ClinicalTrials.gov to identify ongoing or unpublished randomized trials and published a call for evidence. Detailed search strategies are provided in Appendix 1.

One researcher collated IPD from across the SAS data files (SAS Institute, Cary, North Carolina) provided by Medtronic, and a second researcher checked the process. The data were checked for completeness, internal consistency, improbable values, and balance of patient characteristics across treatment groups (see Appendix 2 for details). Three researchers independently working in pairs performed study selection and data extraction from published reports. Disagreements were resolved by discussion or by referring to the third researcher. We assessed risk of bias by using the Cochrane Collaboration's “risk-of bias” tool in the RCTs and a modified form of the Newcastle-Ottawa Scale for nonrandomized studies (89) (see Appendix 2 for details). Risk of bias was assessed by at least 2 researchers independently, with disagreements resolved by discussion.

Effectiveness Analysis

Our prespecified primary outcomes were those likely to be important to patients. The Oswestry Disability Index (ODI) and Neck Disability Index for cervical spinal surgery measure lower back and neck pain, respectively, on a scale from 0% (no pain) to 100% (extreme pain). The Short Form-36 (SF-36) Physical Component Summary (PCS) assesses pain and physical function on a scale from 0% (worst) to 100% (best). Medtronic also provided data on back and leg pain, which were measured on a scale from 0 (no pain) to 20 (extreme pain). Spinal fusion was assessed as success or failure according to Medtronic's radiographic definition, which required evidence of bridging trabeculae, no evidence of motion (<3-mm difference in translation and <5-degree difference in angular motion), and no evidence of radiolucency. We analyzed these outcomes at 6 weeks and 3, 6, 12, and 24 months after surgery. We did not analyze the limited data that were available for longer follow-up times for few participants.

We also considered 4 secondary outcomes: duration of hospital stay, operating time, successful return to work or usual activity, and use of pain medication.

Statistical Methods

Appendix 2 provides details of statistical methods and analyses. In the analyses of effectiveness for continuously distributed outcomes (such as ODI score), we calculated mean differences between treatment groups in the change in score from preoperative values. For dichotomous outcomes (such as successful fusion), we calculated relative risks (RRs). Both were calculated separately for every trial at every time point. We then used standard random-effects meta-analytic techniques (1011) to combine effect estimates across trials. Separate meta-analyses were done for each of the specified time points. Linear and logistic random-effects regression models were used to combine all data from all trials in “1-stage” meta-analyses as sensitivity analyses (1213). We used multiple imputation to explore the influence of missing observations.

Heterogeneity was assessed in all meta-analyses by using the Higgins I2 statistic (14) and the Cochran Q test. We performed a subgroup analysis (stratified by trial) to examine whether effects varied according to the type of spinal surgery or by rhBMP-2 formulation (INFUSE or AMPLIFY). We investigated whether patient-level factors (age, sex, smoking, alcohol consumption, body mass index, diabetic status, and history of spinal surgery for back pain) were associated with the effectiveness of rhBMP-2 surgery by using a 1-stage random-effects regression model (12) that included interaction terms between patient-level factors and treatment.

Safety Analysis

We examined numbers of adverse events, including cancer, provided in the IPD according to the Medtronic classifications of adverse events. Medtronic did not provide case report forms, but summaries of cases were given for most patients in the clinical study reports. A full clinical assessment of these narratives was beyond the scope of this project, but checks in 1 trial (Inter Fix PLIF [15]) showed that the Medtronic classifications of events seemed appropriate. To our knowledge, these data represent all adverse events that occurred during the follow-up periods of these trials. Because the number of specific adverse events in most trials was generally small, we used 1-stage random-effects logistic regression meta-analysis models (13) to analyze these data (see Appendix 2 for model details). Results of these analyses are presented as odds ratios.

We extracted data on number of adverse events from other published studies that compared spinal fusion surgery using rhBMP-2 with ICBG and other comparators in at least 10 adult participants, as specified in our protocol.

Role of the Funding Source

This review was funded by the YODA Project, which provided the IPD for and other materials relating to the Medtronic trials but was not involved in the analyses of these data or in the production of this paper. There was no direct contact with Medtronic or the other evaluation team. Our full report to the YODA Project is publicly available and examines a wide range of issues around the effectiveness and safety of rhBMP-2 in spinal surgery (1617).

The YODA project team provided IPD from 17 Medtronic trials. Eleven of these were RCTs comparing rhBMP-2 with ICBG surgery and were eligible for inclusion in our principal evaluation of effectiveness. Of the others, 4 were single-group trials of rhBMP-2 and 1 (7) used a different comparator. We included these only in our supporting consideration of adverse events. We did not consider 1 trial that was stopped early after recruiting only 3 patients.

Figure 1 shows search results. In addition to the trials supplied by Medtronic, we identified 2 eligible randomized trials not conducted by Medtronic that compared rhBMP-2 with ICBG surgery. We requested IPD from the authors and obtained them from 1 study (18). Data from the other trial, which involved 40 patients having single-level bilateral posterior lateral interbody fusion, were unavailable (19). This trial reported 100% fusion in both groups and no difference in back pain or ODI score at 12 months and therefore these published aggregate data could not contribute to our analyses. We identified 1 ongoing trial for which recruitment had been suspended, but we could not include it because it had not been closed (20).

Grahic Jump Location
Figure 1.

Summary of evidence search and selection.

ICBG = iliac crest bone graft; IPD = individual-patient data; RCT = randomized, controlled trial.

* One additional RCT (19) was eligible, but IPD were not provided.

Grahic Jump Location

Our IPD meta-analysis was based on data from 1302 patients in 11 Medtronic RCTs that compared rhBMP-2 with ICBG surgery, as well as data from 106 participants in 1 additional RCT (18) (Table 1).

Table Jump PlaceholderTable 1. Summary of Trials That Provided Individual-Participant Data Used in the Analyses 

Appendix Table 1 provides details of publications associated with each Medtronic trial. However, our analyses used the supplied IPD rather than data reported in these publications.

Table Jump PlaceholderAppendix Table 1. Details of Publications of Medtronic Trials 

For all of the included trials, data on all pain outcomes were available at all time points from 6 weeks to 24 months. Data on spinal fusion were available for all Medtronic trials except the LT-CAGE pilot trial (31) at all times from 6 months onward. The trial by Glassman and colleagues (18) provided fusion data in a different format and is not included in these analyses. Appendix Table 2 summarizes the levels of missing data for the main pain and fusion outcomes. At 24 months after surgery, outcome data were not available for approximately 15% of participants.

Table Jump PlaceholderAppendix Table 2. Missing Outcome Data in Medtronic Trials 

Our assessment of risk of bias was the same for all Medtronic trials. Randomization and allocation concealment procedures were adequate for all trials. Neither patients nor physicians were blinded to the treatment received, and all pain and function outcomes were patient-assessed, so there was a potential for bias in these outcomes. Successful fusion was assessed by researchers blinded to the treatment received.

Effectiveness
Pain

Figure 2 shows results of meta-analyses across the 12 RCTs for 4 pain outcomes (the SF-36 PCS also incorporates an assessment of physical function). From 6 months after surgery onward, the use of rhBMP-2 generally achieved greater pain reduction (from preoperative values) than did ICBG. Among rhBMP-2 recipients, the ODI score was approximately 3.5 percentage points better (mean difference, −3.48 percentage points [95% CI, −6.47 to −0.49 percentage points]; I2 = 38%) and back pain was better by more than 1 point on the 20-point scale used (mean difference, −1.58 [CI, −2.65 to −0.51]; I2 = 44%) at 24 months after surgery. The SF-36 PCS score was 1.93 percentage points higher for rhBMP-2 recipients (CI, 0.63 to 3.22 percentage points; I2 = 0%) at 24 months. We found no evidence of a difference in leg pain reduction between treatment groups (mean difference, −0.59 [CI, −1.27 to 0.09]; I2 = 0%). In general, patients in both groups improved considerably over time such that the extra benefit of rhBMP-2 over ICBG surgery was small in comparison (Appendix Figure 1). The ODI score improved by approximately 26 percentage points at 24 months for ICBG recipients and by 30 percentage points for rhBMP-2 recipients.

Grahic Jump Location
Figure 2.

Meta-analyses of pain outcomes at 6 wk and 3, 6, 12, and 24 mo after surgery.

Points on the plot represent mean differences in changes in scores (from preoperative values) at each time point, and vertical lines show the 95% CIs. For the first 3 outcomes, points below the dotted line indicate a benefit of rhBMP-2; for the SF-36 PCS score, points above the line indicate a benefit of rhBMP-2. ICBG = iliac crest bone graft; PCS = Physical Component Summary; rhBMP-2 = recombinant human bone morphogenetic protein-2; SF-36 = Short Form-36.

Grahic Jump Location
Grahic Jump Location
Appendix Figure 1.

Reduction in pain scores from preoperative results, by treatment received.

ICBG = iliac crest bone graft; PCS = Physical Component Summary; rhBMP-2 = recombinant human bone morphogenetic protein-2; SF-36 = Short Form-36.

Grahic Jump Location

Results of the 1-stage linear regression meta-analysis models of pain outcomes were almost identical to the results presented earlier, as were those from models that used multiple imputation of missing pain data (see Appendix Figure 2).

Grahic Jump Location
Appendix Figure 2.

Results of 1-stage meta-analyses of pain and function outcomes at 6 wk and 3, 6, 12, and 24 mo after surgery.

ICBG = iliac crest bone graft; PCS = Physical Component Summary; rhBMP-2 = recombinant human bone morphogenetic protein-2; SF-36 = Short Form-36.

Grahic Jump Location
Fusion

Figure 3 shows a forest plot for successful fusion 24 months after surgery, at which time rhBMP-2 increased fusion rates by 12% (RR, 1.12 [CI, 1.02 to 1.23]). Increased fusion rates were also identified at 6 months (RR, 1.20 [CI, 1.00 to 1.44]) and 12 months (RR, 1.11 [CI, 1.00 to 1.22]) after surgery. However, we found substantial heterogeneity of the RR for successful fusion across trials (I2 = 97%, 80%, and 76% at 6, 12, and 24 months, respectively).

Grahic Jump Location
Figure 3.

Forest plot of RR for successful fusion 24 mo after surgery.

ALIF = anterior lumbar interbody fusion; BCP = biphasic calcium phosphate; ICBG = iliac crest bone graft; PLIF = posterior lumbar interbody fusion; rhBMP-2 = recombinant human bone morphogenetic protein-2; RR = relative risk.

Grahic Jump Location

Results from the 1-stage logistic regression meta-analysis model for successful fusion were almost identical to the results presented earlier, as were those from models that used multiple imputation of missing fusion data.

Investigation by Surgical Approach

We performed subgroup analyses to investigate whether the effectiveness of rhBMP-2 varied among patients who had anterior lumbar fusion, posterior lumbar fusion, or anterior cervical fusion (Table 1). Appendix Figure 3 shows the results of these analyses for ODI score and successful fusion 24 months after surgery. A test for heterogeneity showed moderate evidence of a difference between surgery types for ODI score (P = 0.065), but this was primarily due to the large benefit of rhBMP-2 on the Neck Disability Index score observed in the single small cervical surgery trial (n = 23). Excluding this trial resulted in no clear difference in the effectiveness of rhBMP-2 between anterior or posterior approaches (P = 0.171). We found no evidence of a difference in the RRs for successful fusion (P = 0.88) or any other outcome at 24 months across surgery types.

Grahic Jump Location
Appendix Figure 3.

Subgroup analyses for Oswestry Disability Index (top) and successful fusion (bottom), by surgical approach.

ALIF = anterior lumbar interbody fusion; BCP = biphasic calcium phosphate; ICBG = iliac crest bone graft; MD = mean difference; PLIF = posterior lumbar interbody fusion; rhBMP-2 = recombinant human bone morphogenetic protein-2; RR = relative risk.

Grahic Jump Location

Evidence of interactions between rhBMP-2 and the patient-level factors (age, sex, smoking, alcohol consumption, body mass index, diabetic status, and history of spinal surgery) was generally lacking—that is, each factor benefited from rhBMP-2 to the same extent. One possible exception was that, for persons with a previous spinal surgery, there was no difference in the effectiveness of rhBMP-2 and ICBG at reducing ODI score or improving fusion rates. Given the number of analyses done, this may be a chance finding. These results are available in our full report (17).

Secondary Outcomes

We found no evidence of difference in duration of hospital stay (mean difference, −0.15 days [CI, −0.33 to 0.03 days]) or that rhBMP-2 surgery increased the probability of returning to work or usual activity earlier compared with ICBG (RR at 24 months, 1.01 [CI, 0.88 to 1.17]). Using rhBMP-2 shortened operating times by 21 minutes (CI, 15 to 27 minutes) (Appendix Figure 4) from an average of 135 minutes. We found no evidence that analgesic use differed between the rhBMP-2 and ICBG groups at any time (Appendix Figure 5).

Grahic Jump Location
Appendix Figure 4.

MD in operating time (minutes) between rhBMP-2 and ICBG in the Medtronic trials.

Details on the Medtronic trials are available in Appendix Table 1. ALIF = anterior lumbar interbody fusion; BCP = biphasic calcium phosphate; ICBG = iliac crest bone graft; MD = mean difference; PLIF = posterior lumbar interbody fusion; rhBMP-2 = recombinant human bone morphogenetic protein-2.

Grahic Jump Location
Grahic Jump Location
Appendix Figure 5.

Meta-analysis of use of 4 types of pain medication in Medtronic trials.

ICBG = iliac crest bone graft; rhBMP-2 = recombinant human bone morphogenetic protein-2.

Grahic Jump Location

We investigated the association between successful fusion and change in pain score by comparing the mean difference in pain at 24 months after surgery with the RR for successful fusion at the same time point for each of the 4 pain outcomes (ODI score, SF-36 PCS score, back pain, and leg pain) (Appendix Figure 6). We found no evidence of a consistent relationship between improvements in fusion due to rhBMP-2 and improvements in pain or function. If successful fusion resulted in reduced pain, we would expect trials that showed higher fusion rates with rhBMP-2 to show greater improvement in pain scores, but this did not seem to be the case. In particular, trials where fusion was less common in the rhBMP-2 recipients (Inter Fix ALIF pilot [27] and BCP U.S. [32]) still showed a benefit of rhBMP-2 in terms of improved SF-36 PCS score. The BCP U.S. trial also showed a benefit of rhBMP-2 on ODI score and back pain. Therefore, the apparent small benefits of rhBMP-2 in pain reduction do not seem to be due to increased fusion rates.

Grahic Jump Location
Appendix Figure 6.

Relationship between relative risk for fusion and mean difference in pain outcomes across trials 24 mo after surgery.

PCS = Physical Component Summary; SF-36 = Short Form-36.

Grahic Jump Location
Safety

All Medtronic trials provided data on adverse events at all specified time points and also at or shortly after (that is, up to 4 weeks after) surgery. Because reporting of adverse events in the trial by Glassman and colleagues (18) was not consistent with that in the Medtronic trials, it was not included in these analyses; however, the data are included in our full report (17).

We note that pain was reported as an adverse event in the Medtronic IPD as well as being assessed as an effectiveness outcome using the pain scales discussed earlier. The reasons for this were not clear, but for completeness, we analyze pain reported as an adverse event, particularly because pain immediately after surgery was not recorded on the pain scales—its presence or absence was recorded only as an adverse event.

Figure 4 shows the results of the 1-stage meta-analyses for adverse events across the 11 Medtronic RCTs at or shortly after surgery. Risks for arthritis and bursitis, implant-related events, neurologic events, other pain, retrograde ejaculation, wound complications, and vascular events increased by at least 50% among rhBMP-2 recipients. Because there were few events, CIs were wide and findings were inconclusive. For back and leg pain, we found clear evidence of a higher incidence among rhBMP-2 recipients (odds ratio, 1.92 [CI, 1.14 to 3.25]; P = 0.004). Appendix Figure 7 shows the results of the 1-stage meta-analyses for adverse events over all times up to 24 months after surgery. As in the analysis in Figure 4, results were generally inconclusive.

Grahic Jump Location
Figure 4.

Meta-analysis of adverse events at or shortly after surgery in 11 Medtronic trials.

Details on the Medtronic trials are available in Appendix Table 1. ICBG = iliac crest bone graft; OR = odds ratio; rhBMP-2 = recombinant human bone morphogenetic protein-2.

Grahic Jump Location
Grahic Jump Location
Appendix Figure 7.

Meta-analysis of all adverse events in 11 Medtronic trials.

Details on the Medtronic trials are available in Appendix Table 1. ICBG = iliac crest bone graft; OR = odds ratio; rhBMP-2 = recombinant human bone morphogenetic protein-2.

Grahic Jump Location

Appendix Figure 8 shows the results of meta-analyses for 4 key adverse event categories (implant-related, infections, neurologic, and any pain) across all periods. Events were uncommon, so findings are mostly inconclusive. The only clear evidence of a difference was for pain at or shortly after surgery, which was more common in rhBMP-2 recipients (odds ratio, 1.78 [CI, 1.06 to 2.95]; P = 0.007). This contrasts with the results seen in the analyses of ODI score, SF-36 PCS score, and back pain in the effectiveness analyses, where pain reduction was greater in the rhBMP-2 recipients from 3 months after surgery onward.

Grahic Jump Location
Appendix Figure 8.

Meta-analyses of adverse events.

ICBG = iliac crest bone graft; rhBMP-2 = recombinant human bone morphogenetic protein-2.

Grahic Jump Location
Cancer

Table 2 summarizes the cancer cases observed in the 11 Medtronic RCTs, 5 of which observed at least 1 case. It excludes preexisting cancer but includes 3 cases in the LT-CAGE open pivotal trial (33) that were identified during extended follow-up of only the rhBMP-2 recipients. However, these 3 cases were not included in the quantitative analyses because this would have biased against rhBMP-2 because any equivalent cases occurring in the ICBG group had not been sought. A 1-stage random-effects meta-analysis model found that cancer was nearly twice as common among rhBMP-2 recipients (RR, 1.98 [CI, 0.86 to 4.54]), but the 95% CIs were consistent with risk in rhBMP-2 recipients being anywhere from 14% lower to 454% higher. The absolute risk for cancer was low (3% in rhBMP-2 recipients). A forest plot for the equivalent 2-stage analysis is shown in Figure 5 (RR, 1.84 [CI, 0.81 to 4.16]). The RR for cancer was similar across trials. In particular, the RR for cancer in the AMPLIFY trial (3435), which used a different preparation of rhBMP-2 at a higher dose, was no greater than in trials that used INFUSE (P = 0.82). We note that in addition to the 3 cancer cases identified during additional follow-up, 3 cases were seen among rhBMP-2 recipients in the Maverick trial (7) and 2 were seen in a single-group Medtronic trial (3638). These were not included in our analyses because neither trial had an ICBG comparator.

Table Jump PlaceholderTable 2. Incidence of Cancer in the Medtronic Trials 
Grahic Jump Location
Figure 5.

Forest plot of cancer incidence in the Medtronic trials.

The number of cancer cases is the total number occurring during the 2-y follow-up. Details on the Medtronic trials are available in Appendix Table 1. BCP = biphasic calcium phosphate; ICBG = iliac crest bone graft; rhBMP-2 = recombinant human bone morphogenetic protein-2; RR = relative risk.

Grahic Jump Location
Adverse Events Reported in the Literature

We identified 35 observational studies of at least 10 adult patients (in 43 publications) that reported adverse effects of rhBMP-2. There were 14 studies of posterior lumbar fusion, 5 of anterior lumbar fusion, 10 of cervical fusion, and 8 that used multiple spinal fusion procedures. These studies used various spinal fusion techniques as controls, including ICBG, local bone graft, allograft, and bone marrow aspirates. Other than the Medtronic Maverick trial (7), we did not have IPD from any of them. The studies are summarized in Appendix Table 3. Quality assessment found that all of the studies included patients who were representative of those likely to receive treatment in practice and clearly established exposure to treatment. However, most made no attempt to match or control for potential confounding factors, and data on the comparability of the treatment groups were generally unreported or limited. Given the heterogeneity of these studies and their potential for bias, we did not meta-analyze these data.

Table Jump PlaceholderAppendix Table 3. Details of Published Studies Reporting on the Safety of rhBMP-2 Surgery 

Despite the methodological issues, we found evidence suggestive of higher rates of particular adverse events among rhBMP-2 recipients (Figure 6). Among studies reporting at least 1 event, heterotopic bone formation (reported in 5 studies) was more common among rhBMP-2 recipients, although whether this led to any clinical consequences for these patients was unclear. Leg pain and radiculitis (4 studies) seemed to be more common, as had been observed in the Medtronic trials. Osteolysis was more common, but only 2 studies reported on this event. Dysphagia (6 studies) seemed to be more common among rhBMP-2 recipients having cervical spinal surgery, although results of these studies were inconsistent. Comer and colleagues (39) compared 4 consecutive-patient cohorts undergoing ALIF with or without rhBMP-2 and reported a higher rate of retrograde ejaculation among rhBMP-2 recipients (6.3% vs. 0.9%; P = 0.001). Further adverse event outcomes were examined in our full report.

Grahic Jump Location
Figure 6.

Heterotopic bone formation, radiculitis, and dysphagia in nonrandomized studies of rhBMP-2.

ACDF = anterior cervical discectomy and fusion; ICBG = iliac crest bone graft; CSA = cervical spinal arthrodesis; LIF = lumbar interbody fusion; PCF = posterior cervical fusion; PLF = posterolateral lumbar fusion; PLIF = posterior lumbar interbody fusion; rhBMP-2 = recombinant human bone morphogenetic protein-2; RR = relative risk; SCP = silicated calcium phosphate; TLIF = transforaminal lumbar interbody fusion.

Grahic Jump Location

Our principal analyses were based on data from 1408 individual participants in 11 eligible RCTs, including all trials sponsored by Medtronic (published and unpublished) and 1 additional trial. We found the randomization procedures to be adequate in all trials, but participants were not blinded to treatment. Although assessment of some outcomes, such as radiologic assessment of fusion, was blinded, patient-reported outcomes related to pain were not. Follow-up was reasonably complete up to our final analysis time point of 24 months. Although there is some potential for bias associated with patient-reported outcomes, in general, we consider the body of evidence for comparative effectiveness to be strong.

We found clear evidence that rhBMP-2 improves rates of fusion compared with ICBG; however, the Medtronic definitions of fusion that we used may have been stringent given that only 69% of ICBG recipients achieved fusion within 24 months, which is lower than would be expected generally. Inconsistency across trials was high, with large I2 values at all time points.

We also found that rhBMP-2 improves back pain and quality of life compared with ICBG at between 6 and 24 months after surgery. However, these improvements in pain fall below previously described, clinically meaningful thresholds (estimated as between 4 and 17 percentage points for ODI score and ≥5.4 points for SF-36 PCS [4041]).

In general, successful fusion and pain reduction do not seem to be strongly correlated. Trials with higher fusion rates for rhBMP-2 did not also achieve greater pain reduction. In the trials with lower fusion rates among rhBMP-2 recipients, pain reduction was still greater among these patients. It therefore seems that either rhBMP-2 surgery has an effect on pain beyond that from fusion—which seems medically unlikely—or the interpretation of pain was biased. Because participants were not blinded to the treatment received or their fusion status, they may have reported exaggerated benefits of the “new” treatment, thus biasing assessment in favor of rhBMP-2.

In contrast, the analysis of adverse events reported in the IPD showed an increased risk for pain associated with rhBMP-2 in the immediate postsurgical period. Although this may seem to contradict the finding that rhBMP-2 reduces pain from 6 months onward, rhBMP-2 surgery may lead to increased pain shortly after surgery but reduced pain in the longer term.

The IPD also indicate that rhBMP-2 may be associated with an increased risk for cancer, with nearly double the number of new cancer cases compared with ICBG recipients. The overall absolute risk for cancer is low in both groups, however, so whether this increased risk is genuine is uncertain, but it is consistent with the literature suggesting a possible link between BMP and cancer (42).

Adverse event data in the literature raise concerns that rhBMP-2 may increase the risk for heterotopic bone formation, osteolysis, radiculitis, and retrograde ejaculation. However, these findings should be interpreted cautiously because they are based on only published nonrandomized studies, most of which provided little information about the comparability of groups.

Our review differs from the existing review (5) in that we had access to more extensive and detailed data than did Carragee and colleagues, who used aggregate data extracted from publications of industry-sponsored trials and publicly available FDA summaries and public meeting documents. The FDA materials seem to provide incomplete outcome data from a subset of trials evaluating rhBMP-2. We analyzed IPD from all Medtronic-sponsored trials regardless of whether they had been published or submitted to the FDA. This review included the licensed INFUSE preparation of rhBMP-2 and the unlicensed AMPLIFY and BCP preparations, which use a higher dose of rhBMP-2. We found no evidence of a difference in effectiveness, safety, or cancer risk between licensed and unlicensed preparations.

The use of rhBMP-2 in spinal fusion surgery increases the likelihood of successful fusion at up to 24 months, but this does not seem to translate into a clinically significant reduction in pain. The small improvements in fusion and in the level of pain reduction, which manifest after 6 months, also seem to come at the expense of more frequent pain in the immediate postoperative period and, possibly, an increased number of cancer cases. We believe that it is important that clinicians explain these findings to patients so that they can make informed choices about the type of surgery they would prefer.

Appendix 1: Search Strategies and Call for Evidence
Search Strategies
OVID MEDLINE (1948 to Present)

1 bone morphogenetic proteins/ or bone morphogenetic protein 2/ (10872)

2 ((bone morphogen$ or osteogen$ or osteoinduct$) adj (protein$ or factor$ or polypeptide$ or poly-peptide$)).af. (16131)

3 (bmp or bmp2 or bmp-2).af. (11384)

4 (rhbmp or rhbmp2 or rhbmp-2).af. (1262)

5 (rh-bmp or rh-bmp2 or rh-bmp-2).af. (72)

6 (infuse or amplify).af. (15590)

7 or/1-6 (33622)

8 Spinal Fusion/ (14504)

9 (spine or spinal).af. (322484)

10 spondylosyndes$.af. (7)

11 spondylodes$.af. (631)

12 lumbar interbody arthrodesis.af. (22)

13 ((lumbar or cervical or posterior or anterior or lumbosacral or transforminal or posterolateral) adj3 fusion$).af. (7931)

14 fusion cage.af. (125)

15 or/8-14 (323015)

16 7 and 15 (1312)

17 exp animals/ not humans.sh. (3715340)

18 16 not 17 (827)

EMBASE (1974 to Present)

1 bone morphogenetic protein/ or bone morphogenetic protein 2/ (12676)

2 ((bone morphogen$ or osteogen$ or osteoinduct$) adj (protein$ or factor$ or polypeptide$ or poly-peptide$)).af. (19870)

3 (bmp or bmp2 or bmp-2).af. (11817)

4 (rhbmp or rhbmp2 or rhbmp-2).af. (1435)

5 (rh-bmp or rh-bmp2 or rh-bmp-2).af. (90)

6 (infuse or amplify).af. (17204)

7 or/1-6 (38898)

8 Spine Fusion/ (13270)

9 (spine or spinal).af. (369003)

10 spondylosyndes$.af. (10)

11 spondylodes$.af. (1709)

12 lumbar interbody arthrodesis.af. (23)

13 ((lumbar or cervical or posterior or anterior or lumbosacral or transforminal or posterolateral) adj3 fusion$).af. (9892)

14 fusion cage.af. (196)

15 or/8-14 (369756)

16 7 and 15 (1898)

17 animal experiment/ (1579120)

18 16 not 17 (1542)

Cochrane Central Register of Controlled Trials (CENTRAL)

#1 MeSH descriptor Bone Morphogenetic Proteins, this term only

#2 MeSH descriptor Bone Morphogenetic Protein 2, this term only

#3 (morphogen* NEXT (protein* or factor* or polypeptide* or poly-peptide*)) or (osteogen* NEXT (protein* or factor* or polypeptide* or poly-peptide*)) or (osteoinduct* NEXT (protein* or factor* or polypeptide* or poly-peptide*))

#4 (bmp or bmp2 or bmp-2)

#5 (rhbmp or rhbmp2 or rhbmp-2)

#6 (rh-bmp or rh-bmp2 or rh-bmp-2)

#7 (infuse or amplify)

#8 (#1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7)

#9 MeSH descriptor Spinal Fusion, this term only

#10 (spine or spinal)

#11 (spondylosyndes*)

#12 (spondylodes*)

#13 (“lumbar interbody arthrodesis”)

#14 ((lumbar or cervical or posterior or anterior or lumbosacral or transforminal or posterolateral) NEAR/3 fusion*)

#15 (“fusion cage”)

#16 (#9 OR #10 OR #11 OR #12 OR #13 OR #14 OR #15)

#17 (#8 AND #16)

Science Citation Index Expanded (SCI-EXPANDED) (1899 to Present)

#24 #14 not #23

#23 #22 OR #21 OR #20 OR #19 OR #18 OR #17 OR #16 OR #15

#22 Title=(genera or taxonomy or species or fauna or habitat or marine or ecology)

#21 Title=(cow or cattle or bovine or livestock or swine or poultry)

#20 Title=(rabbit or rabbits or moss or mosses or fungus or fungi)

#19 Title=(fossil or fossils or lichen or lichens or mushroom or mushrooms)

#18 Title=(bat or bats or bee or bees or grass or grasses or bird or birds or avian)

#17 Title=(bovine or sheep or fly or flies or fish or fishes or fisheries or horse or horses or equine)

#16 350,781 Title=(animal or animals or dog or dogs or canine or cat or cats or feline)

#15 Title=(rat or rats or mouse or mice or hamster or hamsters)

#14 #13 AND #6

#13 #12 OR #11 OR #10 OR #9 OR #8 OR #7

#12 Topic=(“fusion cage”)

#11 Topic=((lumbar or cervical or posterior or anterior or lumbosacral or transforminal or posterolateral) NEAR/3 fusion*)

#10 Topic=(“lumbar interbody arthrodesis”)

#9 Topic=(spondylodes*)

#8 Topic=(spondylosyndes*)

#7 Topic=(spine or spinal)

#6 #5 OR #4 OR #3 OR #2 OR #1

#5 Topic=(infuse or amplify)

#4 Topic=(rh-bmp or rh-bmp2 or rh-bmp-2)

#3 Topic=(rhbmp or rhbmp2 or rhbmp-2)

#2 Topic=(bmp or bmp2 or bmp-2)

#1 14,240 Topic=((“bone morphogen*” or osteogen* or osteoinduct*) NEAR/1 (protein* or factor* or polypeptide* or “poly-peptide*”))

Similar searches were performed in PubMED, DARE (Database of Abstracts of Reviews of Effects), HTA, Biosis Previews, and Toxfile.

Call for Evidence
Systematic Review of Bone Morphogenic Protein-2 (rhBMP-2) for Spinal Fusion

The Centre for Reviews and Dissemination (CRD) is undertaking a systematic review and individual participant data (IPD) meta-analysis of the comparative effectiveness of rhBMP-2 (marketed as INFUSE) for spinal fusion. The review has been commissioned by the Yale University Open Data Access (YODA) initiative as part of an overarching project to systematically review the safety and effectiveness of rhBMP-2, including reanalysis of IPD that have been made available to Yale on an unrestricted basis by the manufacturer (Medtronic Inc.). YODA aims to improve access to patient-level data from clinical trials and provide independent, scientifically rigorous, objective, and fair analyses of such data.

CRD will undertake a comprehensive and rigorous systematic review and meta-analysis of individual participant data (IPD) of all relevant randomised controlled trials that have compared rhBMP-2 with standard bone graft therapy.

We will include all relevant randomised controlled trials irrespective of whether conducted by the manufacturer or not, and irrespective of whether published or not. We are therefore interested in hearing from anyone who has conducted, or is aware of, unpublished or partially published research in this area. For example, trials which have been presented at conferences but not fully reported elsewhere.

We are currently aware of 17 trials funded by the manufacturer and have searched the published literature but welcome any information regarding further unpublished research. If you know of any such trials please contact [CRD details deleted].

Appendix 2: Methods
Data Sources Used

Analyses of efficacy were restricted to RCTs in spinal fusion surgery that compared rhBMP-2 with conventional ICBG surgery. Single-group trials of rhBMP-2 or trials with comparators other than IBCG were excluded.

One trial (rhBMP-2/BCP US pilot RCT) had 2 rhBMP-2 groups using different fixation procedures. Only the primary group was used in these analyses. The second rhBMP-2 group (consisting of 11 patients) was combined with the first in sensitivity analyses (not presented in this paper).

We performed all analyses using the patient-level data supplied by Medtronic and from Glassman and colleagues' (18) trial. Although intention-to-treat analyses were intended, this was not possible because many randomly assigned patients withdrew before surgery; therefore, no outcome data were available for them.

Outcomes of Interest
Primary Outcomes

Disease-specific pain and functionality:

ODI or Neck Disability Index scores for cervical spinal surgery; this measures lower back (or neck) pain on a scale from 0% (no pain) to 100% (extreme pain).

SF-36 PCS score, which assesses both pain and physical function on a scale from 0% (worst) to 100% (best).

Back and leg pain; both measured on a scale from 0 (no pain) to 20 (extreme pain).

Successful spinal fusion:

Defined radiographically by Medtronic as requiring all of the following: evidence of bridging trabeculae, no evidence of motion (<3-mm difference in translation, <5-degree difference in angular motion), and no evidence of radiolucency.

Secondary Outcomes

Duration of hospital stay

Operating time

Successful return to work or usual activity

Use of pain medication (not prespecified)

Safety Outcomes

We analyzed adverse events supplied by Medtronic that we considered to be potentially related to spinal surgery according to the categorizations provided in the Medtronic IPD. We also considered the following broad categories of adverse events:

Pain: back, leg, lower extremity, arm, neck, and upper extremity pain

Implant-related (hardware failure): displacement, breakage, loosening, malpositioning, and subsidence

Infection

Neurologic events: including numbness, tingling, “pins and needles”

Cancer

For adverse events not generally reported by Medtronic, we searched the wider literature and analyzed the following:

Leg pain (including radiculitis)

Heterotopic bone formation

Dysphagia (in cervical spinal surgery)

Retrograde ejaculation

Osteolysis (further outcomes were considered in our full report)

Patient-Level and Trial-Level Factors Affecting Effectiveness

We investigated how the effectiveness of rhBMP-2 might be influenced by trial- and patient-level characteristics. We investigated the following trial-level factors:

Spinal location of surgery (such as cervical or lumbosacral)

Surgical approach (such as anterior lumbar fusion and posterior lumbar interbody fusion)

We also investigated the following patient-level factors:

Previous spinal surgical interventions

Age

Sex

Smoking status

Diabetic status

Body mass index

Time Points

These outcomes were analyzed at a range of different time points after surgery: 6 weeks and 3, 6, 12, and 24 months. Data on successful fusion were available only from 6 months after surgery onward. Data on adverse events were provided at these times and were also available and analyzed at or immediately after surgery (up to 4 weeks). For all trials, data were provided at all the described time points.

Data Management and Checking of IPD

Data were provided by Medtronic for each trial in a range of separate SAS-format data files according to the types of outcomes reported. From these, we collated individual-level data on all the available effectiveness outcomes at all the time points listed above.

Data from the single non-Medtronic trial (18) were provided as a single Excel (Microsoft, Redmond, Washington) spreadsheet.

For each trial for which IPD were provided, we checked the consistency of the data, both to ensure that the data as provided were valid and to check for errors in our data collation process. Data-checking procedures included the following:

Checking uniqueness and consistency of patient identification numbers

Checking consistency of treatment allocation records

Ensuring all demographic data (e.g., ages) were within plausible ranges

Ensuring all pain and function measurements (e.g., ODI score, SF-36 PCS) were within range, with no outliers

Checking for balanced randomization in terms of age, sex, and other patient-level factors

Checking that summary scores (e.g., ODI score) agreed with the raw scores from each question from the questionnaire

It was not possible to check judgments of fusion status without access to the raw radiologic data and a radiologist. However, if fusion status was assessed by blinded experts, any possible reporting bias or inconsistencies in judgment should be shared across both rhBMP-2 and the control groups. We have assumed that the assessments of spinal fusion provided by Medtronic are valid.

Quality Assessment and Risk of Bias

In addition to the data-checking procedures, we also assessed each randomized trial using the Cochrane risk-of-bias tool (8). The Cochrane Collaboration developed this tool to assess aspects of trial design and conduct that have been demonstrated empirically to affect estimates of treatment effect. The tool does not address aspects of trial design that relate to applicability or generalizability.

The risk-of-bias tool covers 6 key areas of potential bias: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessors, incomplete outcome data, and selective reporting of outcomes. Each area was given a judgment of high risk of bias, unclear risk of bias, or low risk of bias, and a reason for each judgment was recorded in the main data extraction spreadsheet. We used the guidelines from the Cochrane Collaboration on what constituted high, low, and unclear risk of bias. Additional details were based on discussions with the clinical member of the team.

Two reviewers independently completed risk-of-bias assessments for all trials included in the efficacy analyses. Judgments were made for each type of outcome reported in the trials: fusion, patient-reported, and adverse events.

These decisions were based on the full trial protocols provided by Medtronic and the brief protocol provided by Glassman and colleagues.

To address incomplete outcome data, we used standard data-checking procedures described in the previous section to compare loss to follow-up in each group. We also checked that we had been given IPD for all outcomes that were listed in the trial protocols. We requested from Medtronic any available data for patients who were recruited but not reported in the trials. Some tabulated information was provided on why these patients did not undergo surgery and so were excluded from the IPD, but further data were unavailable. No postrandomization data had been collected for these individuals. We also checked loss to follow-up and missing outcome data (by treatment group) for main outcomes at each analysis time point.

Quality of the nonrandomized studies was assessed by using a domain-based approach, based on a modified version of the Newcastle-Ottawa scale (9). However, we avoided scoring or assigning star values to the studies because doing so risks producing an uninformative summary score. Instead, we tabulated the relevant information for each of the following domains:

Representativeness of the exposed cohort (did the patients require spinal fusion surgery? Did they form a particular subgroup [for example, all smokers]?)

Selection of the control group (were they drawn from the same source [for example, the same hospital or database]?)

Ascertainment of exposure (given the nature of the topic, this was usually via a secure record in which the exposure to treatment was clearly known)

Outcome of interest not present at start of the study (was this explicitly checked for [for example, preexisting cancer]?)

Confounders or other factors used to match or controlled for in analysis

Outcome assessment (independent/blind assessment, via secure medical record, self-report, or no details)

Follow-up (we considered adequacy of duration in relation to the specific adverse events reported and whether this was similar across the 2 groups)

Assessments were checked, with disagreements resolved by discussion and/or consultation with a third researcher.

Statistical Analysis

Our primary statistical method for estimating the efficacy of rhBMP-2 surgery for all the specified outcomes was to use standard 2-stage meta-analytic techniques (1011). Individual-patient data from each trial were analyzed separately, using the same methods across trials, for all the efficacy outcomes. Separate meta-analyses were performed at each of the specified time points.

We also used a “1-stage” meta-analysis approach in some analyses, primarily as a sensitivity analysis to confirm the results of 2-stage analyses. These approaches are described in more detail below.

All main analyses used a complete-case approach in which participants with missing data were excluded from the analysis.

Estimates of Effect
Continuously Distributed Outcomes.

For the continuously distributed outcomes (ODI, SF-36 PCS, back pain, leg pain), we assessed efficacy in terms of the mean difference in outcome between the rhBMP-2 and ICBG groups.

For each patient, the change in the score from baseline to the time point of interest was calculated. These were then averaged for each intervention in each trial and the difference in means within each trial calculated, and these mean differences were then combined across trials. This mean difference, along with its associated SE, was calculated for each trial.

Dichotomous Outcomes.

For dichotomous outcomes (successful fusion, successful return to work, use of pain medication, cancer), we assessed efficacy in terms of the RR for the outcome between the rhBMP-2 and ICBG groups.

In the 1-stage random-effects meta-analyses, RRs could not be calculated because algorithms did not converge successfully (that is, they crashed). For these models, results were therefore calculated in terms of the odds ratio, with its corresponding 95% CI.

Two-Stage Meta-Analyses

We combined the effect estimates from each trial (mean difference or RR) across trials using a standard DerSimonian–Laird random-effects meta-analysis to account for potential heterogeneity in effects across trials (55). Separate analyses were performed at each time point. We present the results as summary plots across all times.

This is called a “2-stage” approach because it is performed in 2 stages: First, we estimate effects within trials, and then we combine results across trials in a meta-analysis.

One-Stage Meta-Analyses

We performed 1-stage meta-analyses of the pain and function outcomes as a comparison to confirm the validity of the 2-stage analyses.