Each year, 185,000 spinal fusions are performed in the United States to treat various spine disorders with use of bone grafts or bone-graft substitutes1. Demineralized bone matrix, an allograft derived from processed human bone, is used to augment grafting material in order to enhance bone formation and fusion success. Bone morphogenetic proteins (BMPs) are the major osteoinductive components within demineralized bone matrix and have a regulatory role in promoting osteogenesis in submuscular and other mesenchymal tissues2.
An increasing number of demineralized bone matrix-based products are commercially available as bone-graft substitutes, but few of these so-called off-the-shelf demineralized bone matrix-based products have been evaluated for reliability and fusion efficacy. Comparison studies of demineralized bone matrix-based products3-7 have revealed significant differences in osteogenic potential from one demineralized bone matrix-based product formulation to another, illustrating the impact that processing techniques, carrier, sterilization method, and storage conditions can have on product safety and performance6,8-10. Surprisingly, this osteoinductive variability may be even greater across the donor-specific production lots of the same demineralized bone matrix-based product formulation, where production method is no longer a variable11-14. In other words, although production lots from the same demineralized bone matrix-based product are processed similarly, lot-to-lot differences in efficacy still exist within one brand. Since this variability undermines clinical product reliability, the source of intraproduct variability has become a strong focus of recent investigations.
The inconsistency in fusion outcomes resulting from intraproduct variability7 necessitates the terminal screening of demineralized bone matrix-based product lots for potential predictors of osteoinductivity prior to distribution for clinical use. Recent studies with use of rat models have established a relationship between extracted levels of known osteoinductive BMPs (from a demineralized bone matrix-based product) and ectopic bone formation15,16. However, this relationship may be more challenging for extracted BMPs and spinal fusion, an event that requires bone formation and connectivity to existing structures (transverse processes) as well as de novo bone formation. The use of rat assays is also inefficient and requires numerous animals. Other assays for alkaline phosphatase, a marker for osteoblast differentiation, have also been used to predict the osteoinductivity of unprocessed demineralized bone matrix12, but results have been variable17 and carriers in commercial demineralized bone matrix-based products may be confounders (unpublished data).
The purpose of this study was to assess the lot-to-lot variability across a single commercial demineralized bone matrix-based product with use of both in vitro and in vivo measures. Specifically, the goal was to determine whether BMP-2 and/or BMP-7 levels assayed from demineralized bone matrix-based product lots could sufficiently and accurately predict fusion performance in athymic rats.
Study Design
In vitro, ELISAs (enzyme-linked immunosorbent assays) for BMP-2 and BMP-7 were performed on ten different production lots from the same demineralized bone matrix-based product.
In vivo, posterolateral lumbar fusion procedures were performed on forty athymic rats with use of aliquots from each of ten commercial demineralized bone matrix-based product lots (four rats per demineralized bone matrix lot).
Demineralized Bone Matrix-Based Product Protein Extraction
Ten production lots of a commercially available demineralized bone matrix-based product (EBI InterGro DBM Putty; Interpore Cross, Irvine, California) were obtained. InterGro DBM Putty contains AlloGro DBM (40% demineralized bone matrix by weight) in a lipid (lecithin) carrier with coral-hydroxyapatite composite granules. The company lot numbers were 268309, 221820, 268308, 279577, 254526, 279562, 279565, 213334, 221808, and 221811, which were relabeled from 1 to 10, respectively, for convenience.
According to a protocol previously used by Bae et al.11, soluble proteins were extracted from the organic matrix by incubating 200-mg aliquots of each demineralized bone matrix-based product lot in 2 mL of 4-M guanidine-HCl (GuHCl) in 0.5-M sodium acetate at pH 5.8. The mixture was stirred at 4°C for seven days, and the protein extract was collected. Another 2 mL of 4-M GuHCl in 0.5-M sodium acetate containing 0.5-M EDTA at pH 5.8 was used to extract the remaining demineralized bone matrix and was stirred for seven days at 4°C. Protein extracts were independently dialyzed with use of tubing with a 6000 to 8000-Da molecular weight cutoff (Fisherbrand Regenerated Cellulose Dialysis Tubing; Fisher Scientific, Pittsburgh, Pennsylvania) against four changes of distilled water at 4°C for forty-eight hours. They were then lyophilized to dryness and dissolved in 1% sodium dodecyl sulfate phosphate-buffered saline solution.
ELISA: BMP-2 and BMP-7
Detection of BMP-2 and BMP-7 was done with use of the Quantikine BMP-2 and BMP-7 ELISA kits (R&D Systems, Minneapolis, Minnesota). The sandwich method was employed with use of microplates precoated with mouse anti-BMP-2 or anti-BMP-7 immunoglobulin G and secondary antibodies conjugated to horseradish peroxidase. Substrate was added, and the absorbance was measured at 450 nm. BMP-4 was not assayed because previously the measurable levels were insignificant11.
Athymic Rat Models
Forty female athymic (rNu/rNu) rats (Harlan Sprague Dawley, Indianapolis, Indiana), weighing between 137 and 217 g, underwent posterolateral spinal fusion with bilateral implantation of an aliquot of a demineralized bone matrix-based product (commercial demineralized bone matrix with carrier). Each of ten lots of a demineralized bone matrix-based product was implanted bilaterally at the L4-L5 vertebral level in four rats (after approval of the animal protocol was obtained from the Animal Research Committee at the University of California, Los Angeles). Each rat was implanted with only one lot of demineralized bone matrix-based product (the same lot was used on both sides). The L4-L5 transverse processes were exposed through a midline incision and were decorticated with use of a high-speed burr. The facet joints were left intact. Lactated Ringer solution and Baytril (enrofloxacin) were used as irrigation. A 0.3-mL aliquot of demineralized bone matrix was implanted bilaterally into the gutter of the transverse processes on each side. A 4.0 absorbable suture and 4.0 Vicryl suture (polyglactin; Ethicon, Somerville, New Jersey) were used for fascial layer and subcutaneous closure, respectively.
Evaluation
Radiographs were made at intervals until the animals were killed at eight weeks postoperatively. Investigators were blinded to the demineralized bone matrix lot assignments and ELISA results. The two evaluators independently assigned high-resolution radiographs a score from 1 to 4 on the basis of the degree of bone density and remodeling (with 1 indicating no bone or density between transverse processes; 2, osseous nodules or density evident between transverse processes but not attached to transverse processes with <50% fill; 3, bridging from one transverse process to the other with >50% fill; and 4, continuous bridging bone with mass extending to >75% of the interspace filled).
Lumbar spines were excised and tested for fusion by means of manual palpation (the so-called gold standard) performed by investigators blinded to lot assignment. Any motion detected at L4-L5 was deemed a fusion failure. One-sided fusions were downgraded to nonfusions.
Statistical Analyses
The number of rats needed for each experimental condition was calculated with use of an alpha set at 0.05 and power of 0.80 (an 80% chance of detecting a difference). The main response variable, fusion on manual palpation, was given a code of 1 for spines that had fused (no motion) or 0 for spines that had not fused (motion). To test the effect of 0% (zero of four rats) compared with 100% (four of four rats) successful manual fusion, a minimum sample size of approximately four rats was required for each lot to yield a right-sided p value of 0.01. A minimum of four rats per experimental group allowed statistical testing when there would be little variability within a lot-based group.
Chi-square analysis with the Fisher exact test was applied to fusion status to measure the pattern of variability of fusion rates across different demineralized bone matrix lots and for each lot in the lot-to-lot comparison. The frequency of success was compared with the desired manual fusion rate of 100%.
Logistic regression with an exact analysis algorithm was implemented with use of rat as the unit of analysis. Manual fusion (dependent variable) was tested as a function of assayed demineralized bone matrix dose by entering either the term BMP2 or BMP7 into the equation as an independent term (p < 0.05 was significant). This algorithm is specifically used for small, skewed, or sparse data sets18. Correlation analysis (R2 value) was used for the assessment of effect size. For quantification, statistical probit analyses were used to construct BMP concentration-dosing curves to determine the effective doses of BMPs required for successful spinal fusion on manual palpation.
Source of Funding
This study was funded entirely by the Spine Research Foundation.
Variability of Fusion Rates
After killing the animals at eight weeks, 96% of the rats that underwent posterolateral fusion with demineralized bone matrix implantation showed de novo bone formation. However, only 23% (nine) of the forty rats had completely fused spines at eight weeks on the basis of manual palpation. Sample radiographs showing a nonfusion and a fusion are provided for visual comparison (Fig. 1). The highest fusion rate determined by palpation for a demineralized bone matrix-based product lot was 75% (three of four rats), while the lowest was 0% (zero of four). On the basis of chi-square analyses, production lots varied significantly in their ability to promote spinal fusions, suggesting systematic osteoinductive variability across lots of the same formulation (p < 0.04). Two of the ten lots almost always promoted fusion (lots 7 and 8), while five of the ten lots consistently failed (lots 1 through 4 and 10) (Fig. 2-A). There was a positive relationship between manual fusion status and radiographic status. Radiographic scores of 3 and 4 strongly predicted fusion success, and all rats with a fusion by palpation received radiographic scores of =3 (R2 = 0.76, p < 0.0001) (Fig. 2-B). Kappa values of 0.86 and 0.87 indicated good agreement between two coders for radiographic score and manual palpation status, respectively.
Variability of BMP Concentrations
ELISAs revealed variable BMP-2 and BMP-7 concentrations across the demineralized bone matrix-based product lots (Fig. 3-A). The ranges spanned were considerable, from 22 to 110 pg of BMP-2 per milligram of demineralized bone matrix-based product and from 44 to 125 pg of BMP-7 per milligram of demineralized bone matrix-based product. The average concentration (and standard deviation) across all lots was 46 ± 25 pg of BMP-2 per milligram of demineralized bone matrix-based product and was nearly double this for BMP-7 at 82 ± 24 pg/mg of demineralized bone matrix-based product. The BMP-2 concentration in lot 8 (110 pg/mg of demineralized bone matrix-based product) was considerably higher, at a level that was at least twofold greater than that in any other lots; lot 8 also contained the highest concentration of BMP-7. Lot 9 possessed the lowest concentration of BMP-7 (44.5 pg/mg of demineralized bone matrix-based product) and lot 10, the lowest concentration of BMP-2 (21.6 pg/mg of demineralized bone matrix-based product). There was a highly positive correlation between BMP-2 and BMP-7 levels across demineralized bone matrix lots (r = 0.77, p < 0.0001) (Fig. 3-B). As such, the four lots with the highest BMP-7 levels also contained the greatest levels of BMP-2.
Predictive Power of BMP-2 and BMP-7 Concentrations for Fusion
Exact logistic regression implicated increasing concentrations of BMP-2 and BMP-7 (picograms of BMP per milligram of demineralized bone matrix-based product) as strongly positive predictors of spinal fusion (R2 = 0.32, p < 0.01 for BMP-2, and R2 = 0.22, p < 0.009 for BMP-7) (Fig. 4). BMP-2 and BMP-7 concentrations also positively predicted radiographic scores of 4 (p < 0.013 for BMP-2 and p < 0.010 for BMP-7). The finding was stable when adjusting for the total amount in picograms of BMP implanted. No significant increase in predictive power was found when the potential interaction of BMP-2 and BMP-7 combined was analyzed.
The same two lots (7 and 8) with the highest fusion rate of three of four rats also contained the highest concentrations of both BMP-2 and BMP-7. In fact, the four lots with the highest levels of BMP-2 and BMP-7 (lots 5 through 8) all promoted fusion in at least one of four rats. Lots 2 and 10 (no fusions) contained concentrations BMP-7 similar to lot 5 (one of four fused), but they possessed 35% to 55% less BMP-2 than lot 5 and significantly less BMP-7 than lots 7 and 8 (three of four fused). Slight decreases in the levels of BMP-2 or BMP-7, or both, led to lower fusion rates and lower radiographic scores, indicating the sensitivity of BMP-2 and BMP-7 concentrations for fusion success. The thresholds required to achieve at least one fusion among four rats in our study was 51.4 pg of BMP-2 per milligram of demineralized bone matrix-based product and 90.5 pg of BMP-7 per milligram of demineralized bone matrix-based product (Fig. 4). Notably, lot 9 promoted one fusion in four rats even with BMP levels well below these two thresholds, and radiographs of spines implanted with lot 9 showed substantial bone formation with scores of =2 (Fig. 2-B). Truth tables constructed on the basis of the empirical thresholds classified lot 9 as the only false negative (data not shown). For all other lots, higher BMP concentrations resulted in higher fusion rates. Statistical probit analyses for the construction of a BMP concentration-dosing curve projected that approximately 100 pg of BMP-2 per milligram of demineralized bone matrix-based product and 141 pg of BMP-7 per milligram of demineralized bone matrix-based product are required for 80% fusion probability (Fig. 5). Additional studies with use of greater sample sizes are currently underway to validate the accuracy of minimally effective doses (ED80% and ED100%; 80% and 100% fusion probability).
Intraproduct Variability of a Demineralized Bone Matrix-Based Product
This is the first of a series of studies to test in vitro predictors of osteoinductivity by an analysis of the lot-to-lot variability within a single formulation of a demineralized bone matrix-based product. Although the same processing procedure, carrier, storage temperature, and light exposure were used in manufacturing each production lot, a clear lot-to-lot variability in both demineralized bone matrix BMP levels and fusion efficacy is evident. Despite the manufacturer's claim that every lot was bioassayed for at least a minimum osteoinductivity, a fusion rate of 0% was obtained for half the lots tested, illustrating the potential inconsistency of the use of a demineralized bone matrix-based product mixed with bone for fusion clinically. Indeed, autogenous bone graft from the iliac crest remains the gold standard for fusion procedures19-22. Better quality control in the demineralized bone matrix commercial industry is needed to improve clinical fusion outcomes in demineralized bone matrix-based product applications. It would be valuable to correlate experimental in vivo fusion rates with the osteoinductive scores predicted by the manufacturer in future studies.
The root cause of demineralized bone matrix variability across lots from the same demineralized bone matrix-based product remains unclear. Some studies have attributed differential BMP levels to the characteristics of the bone donors, such as age, sex, and other factors14,16,23. In one study, BMP-2 and BMP-7 concentrations were found to be higher in female donors14. In another, BMP-4 levels in demineralized bone matrix were shown to be age-dependent16. Unfortunately, such studies have often provided conflicting data, and there is still no scientifically established relationship between donor traits and osteoinductivity. Given the results of our study, simple in vitro assays such as ELISA, rather than donor characteristics, may be used to more accurately predict the osteoinductivity of final demineralized bone matrix-based product lots.
The significant intraproduct variability of both osteoinductive potential and growth factor levels across demineralized bone matrix-based product lots from the same commercial product is consistent with previous demineralized bone matrix studies11-14. For example, our group previously uncovered differences in BMP-2 and BMP-7 levels across different production lots from the same demineralized bone matrix-based product comparable with the variability found in this study11. This study, however, goes one step further by establishing a connection between the variable BMP levels in a demineralized bone matrix-based product (in vitro) and the differential posterolateral spinal fusion rates promoted by those demineralized bone matrix-based product lots in rats (in vivo). The positive relationship observed between assayed BMP-2 and BMP-7 and spinal fusion in rats provides strong evidence substantiating the claims of our previous work, in which we hypothetically attributed differential fusion outcomes to variable BMP concentrations across demineralized bone matrix-based product lots. Still, despite such variability, the osteoinductivities of final off-the-shelf production lots of many demineralized bone matrix-based products are not tested. Perhaps screening for BMPs with use of simple in vitro procedures will be another step toward better quality control.
BMP Quantities in Demineralized Bone Matrix-Based Products and Spinal Fusion Rates
The most notable finding in this study was the predictive relationship observed between extracted BMP concentrations and fusion success among production lots from a single demineralized bone matrix-based product. In general, the higher the concentration of BMP-2 and/or BMP-7 in the demineralized bone matrix-based product lot used, the higher the rate of fusion. This observed relationship is consistent with the role that BMPs play in osteoblast differentiation and bone formation1. Soluble rhBMP-2 and rhBMP-7 are especially effective at stimulating the differentiation of mesenchymal stem cells into bone-forming osteoblasts. One study has revealed the significantly greater fusion rate induced by an implant of soluble rhBMP-7-collagen (OP-1 putty; 100% fusion) compared with a demineralized bone matrix-based product (Grafton DBM Putty [Osteotech, Eatontown, NJ]; 39%) at six weeks after surgery3. However, as implants containing pure recombinant BMPs are expensive, demineralized bone matrix-based products may offer a less costly alternative when graft augmentation is needed for spinal fusion24.
Past studies have used ELISAs to demonstrate a significant relationship between BMP concentrations extracted from unprocessed demineralized bone matrix powder and de novo bone formation15,16. Using an in vivo assay for ectopic bone formation, one study showed that the amount of BMP-2 in demineralized bone matrix was the best predictor of osteogenic potency25. Consistent with that study, our research also found BMP-2 (as well as BMP-7) concentration to be about equally predictive of spinal fusion success.
Interestingly, on the basis of our data, low concentrations of BMP-2 and BMP-7 do not automatically or absolutely preclude successful fusion. Lot 9, for example, contained the lowest amount of BMP-7 but promoted the same rate of fusion (one of four rats) as did lots with substantially more BMPs. This might suggest the presence of additional osteogenic factors in lot 9, such as transforming growth factor beta or fibroblast growth factor, that could have compensated for fewer BMPs. Carrier effects and variable BMP binding do not appear to be the issue since the carriers and demineralized bone matrix-based product processing were the same for every lot in this study. Differences in demineralized bone matrix carriers are currently being evaluated for their potential contributions to fusion probability.
The positive correlation found between BMP-2 and BMP-7 agrees with previous demineralized bone matrix studies14. It is uncertain how well a relatively high BMP-2 concentration can compensate for a low BMP-7 concentration or vice versa. When lots containing high levels of BMP-7 (such as lots 7 and 8) are compared, it appears that the addition of a high concentration of BMP-2, as in lot 8, does little to improve the fusion rate. Statistically, the interaction effect or combination of BMP-2 and BMP-7 did little to increase the predictive power for spinal fusion in this analysis. It is probable that the variability (in fusion rates) individually explained by BMP-2 and BMP-7 largely overlaps because of their highly correlated nature. For future analyses, it may be valuable to construct BMP-2 and BMP-7 dosing-concentration curves with a larger sample (for demineralized bone matrix-based products) to characterize the effects of potential synergistic interactions and threshold concentrations on efficacy.
Carrier Effects and Limitations of ELISA
ELISA is a sensitive procedure that has commonly been used to quantify BMPs and may be applied to demineralized bone matrix quality control11,14,26. The monoclonal antibodies used, however, detect the common epitope found on both active and inactive forms of BMP and cannot differentiate between the two. As such, ELISAs alone may be unreliable for measuring BMP activity and availability in demineralized bone matrix, both of which are affected by steps in demineralized bone matrix sterilization and manufacturing6,8-10. However, because all of the lots analyzed in this study originated from the same demineralized bone matrix-based product formulation, all BMPs were likely activated or inhibited to the same relative extent. Variable BMP binding and delivery due to differences in storage temperatures, processing, and/or sterilization techniques may have been less of an issue as a result. The quantities of BMP-2 and BMP-7 predicted the probability of fusion sufficiently well regardless of activity. Alternatively, when lots from different manufacturers are compared, ELISAs for BMPs may not be as predictive of fusion since additional variables from carriers, storage temperatures, processing, and sterilization come into play, and many of them can alter osteoinductivity independent of BMP concentrations3,5-7. In these cases, other tests for osteoinductivity may be combined with ELISA for better predictions. Physicians should not generalize the results of this demineralized bone matrix-based commercial product to other demineralized bone matrix-based products because different carriers and processing methods have variable effects on BMP concentrations and fusion rates. However, our study shows that the relative values of BMP-2 and BMP-7 (as measured by means of the ELISA method) predict the relative in vivo potency (fusion rates) of a demineralized bone matrix-based commercial product. ELISAs might be useful as an effective and simpler alternative to alkaline phosphatase assays and ectopic rat assays for predicting the relative likelihood of fusion for individual lots from the same demineralized bone matrix-based product.
To improve the consistency of demineralized bone matrix-mediated fusions in clinical settings, simple in vitro terminal screenings for BMP-2, BMP-7, and other putative growth factors predictive of bone formation are warranted prior to commercial distribution of demineralized bone matrix-based products. Such data should also be provided by demineralized bone matrix-based product manufacturers to allow surgeons to make decisions in an informed manner.
Note: The authors thank UCLA student research program coordinator Pamela Wong and participants Gaurav Mehta, Siamak Yasmeh, Leo Thai, and Justin Houman.