Musculoskeletal Imaging • Original Research Stensby et al. Imaging of Transforaminal Lumbar Interbody Fusion
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Musculoskeletal Imaging Original Research
Radiographic Appearance of Transforaminal Lumbar Interbody Fusion Performed With and Without Recombinant Human Morphogenetic Protein–2 J. Derek Stensby 1,2 Ryan W. Kaliney 1,3 Bennett Alford1 Francis H. Shen 4 James T. Patrie5 Michael G. Fox1 Stensby JD, Kaliney RW, Alford B, Shen FH, Patrie JT, Fox MG Keywords: bone resorption, lumbosacral region, osteogenesis, recombinant human bone morphogenetic protein–2, spinal fusion DOI:10.2214/AJR.15.14503 Received February 6, 2015; accepted after revision May 13, 2015. F. H. Shen has consulting agreements with DePuy Synthes Spine, Globus Medical, and Medtronic Spine. He is on the Medical Board of Trustees for Musculoskeletal Transplant Foundation, and has royalties from Globus Medical and Elsevier Publishing. Based on a presentation at the Radiological Society of North America 2011 annual meeting, Chicago, IL. 1
Department of Radiology and Medical Imaging, University of Virginia, 1218 Lee St, Box 800170, Charlottesville, VA 22908. Address correspondence to M. G. Fox ([email protected]). 2 Present address: Mallinckrodt Institute of Radiology, St. Louis, MO. 3
Present address: Jefferson Radiology, Hartford, CT.
4
Department of Orthopedic Surgery, University of Virginia, Charlottesville, VA. 5
Department of Public Health Sciences, University of Virginia, Charlottesville, VA.
This article is available for credit. AJR 2016; 206:588–594 0361–803X/16/2063–588 © American Roentgen Ray Society
588
OBJECTIVE. The purpose of this study is to determine whether recombinant human morphogenetic protein–2 (rhBMP-2) alters the findings on routine radiographs performed after transforaminal lumbar interbody fusion (TLIF). MATERIALS AND METHODS. A retrospective review of 256 TLIF procedures in 200 patients was performed over a 4-year period. The rhBMP-2 group included 204 TLIFs in 160 patients, and the control group included 52 TLIFs in 40 patients. Two musculoskeletal radiologists reviewed the postoperative radiographs for endplate resorption, resorption resolution, new bone formation, bridging bone, and allograft migration. Statistical analysis was performed using logistic regression. RESULTS. The median age was 53 years in the rhBMP-2 group and 54 years in the control group (p = 0.182). The groups were similar with regard to sex (p = 0.517), single or multilevel TLIF (p = 0.921), specific TLIF levels (p = 0.53), and median radiographic follow-up (373 vs 366 days; p = 0.34). Findings that were more common in the rhBMP-2 group than in the control group included endplate resorption (38% [78/204] vs 12% [6/52]; odds ratio [OR], 4.67; 95% CI, 1.99–12.54; p < 0.001), resorption resolution (59% [46/78] vs 0% [0/6]; OR, 8.09; 95% CI, 1.41 to ∞; p = 0.022), new bone formation (84% [171/204] vs 67% [35/52]; OR, 2.51; 95% CI, 1.24–4.99; p = 0.011), bridging bone (55% [112/204] vs 31% [16/52]; OR, 2.73; 95% CI, 1.43–5.34; p = 0.002), and allograft migration (17% [35/204] vs 2% [1/52]; OR, 6.30; 95% CI, 0.91–151.41; p = 0.065). CONCLUSION. A statistically significant higher frequency of endplate resorption, new bone formation, and bone bridging is present in TLIF augmented by rhBMP-2 compared with TLIF performed without rhBMP-2. Endplate resorption resolves without treatment in most cases after rhBMP-2 use. utologous iliac crest bone graft has long been considered the method of choice for augmentation of vertebral body fusion. Harvesting the autologous iliac bone graft is an additional procedure associated with major complications in 0.7–25% and minor complications in 9.4–39% of cases [1]. Increased complications, increased operative time, increased blood loss, and insufficient autograft for complex or revision procedures have led to the use of alternative graft types [2–4]. One alternative for graft augmentation introduced in 1991 is demineralized bone matrix; however, differences in the harvesting and sterilization processes resulted in variable osteoconductive properties in the commercially available varieties [5, 6]. The potential for infectious disease transmission [6, 7], the use of toxic binders in certain prep-
A
arations of demineralized bone matrix, and the desire to increase fusion rates fueled the search for acceptable alternatives to augment spinal fusion [8, 9]. Bone morphogenic proteins (BMPs) were discovered in 1965, when new bone formation was observed after implantation of devitalized bone into animal models in a dosedependent manner [10]. Not until 2002 did recombinant human BMP–2 (rhBMP-2) receive U.S. Food and Drug Administration (FDA) approval for use in anterior lumbar interbody fusion [11]; the commercially available product is InFUSE (Medtronic Sofamor Danek) [12]. A subsequent meta-analysis evaluating the use of rhBMP-2 as a supplement to or replacement for autologous bone graft for lumbar spinal fusion reported increased fusion rates [13]. Given its potential benefits, the use of rhBMP-2 in spinal sur-
AJR:206, March 2016
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Imaging of Transforaminal Lumbar Interbody Fusion gery increased from less than 1% of posterior lumbar interbody fusions in 2002 to more than 40% in 2011 [14]. Previous studies have compared the postoperative complication rate after lumbar fusion both with and without rhBMP-2 augmentation [15, 16]. A search of the radiology literature for descriptions of the radiographic appearance of transforaminal lumbar interbody fusion (TLIF) after the use of rhBMP-2 yielded one study of 36 cases without a control group for comparison [17]. To our knowledge, the radiographic appearance of TLIF performed with and without the use of rhBMP-2 in the disk space has not been directly compared. We sought to determine whether the radiographic appearance after TLIF varied between patients who had received allograft augmented by rhBMP-2 and patients who had received allograft alone. Materials and Methods
A retrospective radiographic analysis of patients who underwent TLIF at a single academic institution (University of Virginia) between August 2005 and September 2009 was performed after institutional review board approval. TLIF is the preferred method for performing posterior lumbar spinal fusion at our institution because it requires less nerve root retraction than does posterior lumbar interbody fusion as a result of the more lateral transforaminal exposure, although it requires complete resection of the facet joint complex unilaterally, unlike a posterior lumbar interbody fusion, which preserves greater than 50% of the facet joint complex but resects varying degrees of the lamina and spinous process. Two surgeons performed the TLIFs using two varieties of allografts; one group of allografts had a curved cylindric shape and the other group had a more blocklike shape. The shape of the allograft and the use of rhBMP-2 (InFUSE, Medtronic Sofamor Danek) was decided by the surgeon, with the rhBMP-2 placed on an absorbable collagen sponge and placed in the disk space posterior to the allograft. The surgical technique used to perform the TLIF was otherwise similar, with the cartilaginous endplates removed, but the vertebral endplate along with the corresponding subchondral bone left intact to provide endplate support to reduce the risk of graft migration. Because rhBMP-2 is FDA approved for only anterior lumbar interbody fusion when used with the Lumbar Tapered Fusion Device (Medtronic Sofamor Danek), the use of rhBMP-2 for TLIF is considered off label [11]. We did not study anterior lumbar interbody fusion because this technique requires an anterior approach and involves the assistance of a
TABLE 1: Summary of Disk Space Levels Fused Disk Space Level
Group Treated With Human Recombinant Morphogenetic Protein–2
Control Group
T12–L1
1
0
L1–2
1
0
L2–3
11
2
L3–4
26
6
L4–5
83
25
L5–S1
82
19
Note—Data are number of patients. p = 0.53 for all comparisons.
vascular surgeon to gain exposure to the anterior disk space. Even so, over 80% of the spinal fusions augmented with rhBMP-2 in the United States have been performed using an off-label technique [14, 18].
Patients
The study included 200 patients who underwent fusion of 256 intervertebral disk spaces; 160 patients underwent fusion of 204 intervertebral disk spaces augmented by rhBMP-2, and 40 patients underwent fusion of 52 intervertebral disk spaces without rhBMP-2 treatment. Patients younger than 18 years (n = 1), patients clinically suspected of having deep soft-tissue infection or discitis or osteomyelitis (n = 1), and those undergoing revision surgeries were excluded (n = 6). There were 63 men and 97 women in the rhBMP-2 group and 18 men and 22 women in the control group (p = 0.517). The median age was 53 years in the rhBMP-2 group and 54 years in the control group (p = 0.182). Single-level fusion was performed in 76 (47.5%) patients treated with rhBMP-2 and in 22 (55%) of the patients in the control group (p = 0.921). Individual surgical levels are displayed in Table 1. The study was not performed in a prospective manner, and some variability in the interval and duration of radiographic follow-up was unavoidable. Occasionally, patients were lost to follow-up or had intermittent follow-up visits (e.g., missed follow-up at 6 months but presented for follow-up at 12 months). Nevertheless, the median duration of follow-up and
the follow-up intervals between the groups were not statistically significantly different (Table 2). Anteroposterior and lateral radiographs performed at routine postoperative visits were reviewed in consensus by two fellowship-trained musculoskeletal radiologists, with 8 years and 1 year of experience, using a PACS workstation (Carestream, version 10.2, Carestream Health). Radiographs were compared with the baseline immediate postoperative radiograph for the development of endplate resorption, resolution of resorption, new bone formation, bridging bone, and graft migration. Endplate resorption was categorized as none, mild, moderate, or severe, as described by Vaidya et al. [19] (Fig. 1). Resorption was considered to have resolved when bone density replaced the endplate defect (Fig. 2). New bone formation was present when ossification was identified within the intervertebral disk space, with bone bridging occurring when the ossification joined the superior and inferior endplates (Fig. 3). Graft migration was diagnosed when a change in allograft position, from the initial postoperative radiograph, occurred (Fig. 4).
Statistical Analysis
The odds of endplate resorption, resorption resolution, new bone formation, osseous bridging, and allograft migration were compared between the patients who received and did not receive rhBMP-2 via exact logistic regression. Note that exact logistic regression is comparable to logistic regression in that the method is appropriate for analyzing of outcome variables that can be sum-
TABLE 2: Radiographic Follow-Up Intervals, by Patient Group Follow-Up Interval (Postoperative Days)
Group Treated With Human Recombinant Morphogenetic Protein–2 (n = 204)
1–91 92–183
Control Group
p
99.0 (202)
100 (52)
0.47
77.5 (158)
67.3 (35)
0.13
184–365
69.1 (141)
73.1 (38)
0.58
366–730
56.4 (115)
50 (26)
0.41
Median
373
366
0.34
Note—Data are percentage (no.) of patients.
AJR:206, March 2016 589
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Stensby et al. Fig. 1—23-year-old woman who underwent transforaminal lumbar interbody fusion at L3–4 and L4–5 augmented with human recombinant morphogenetic protein–2. A, Initial radiograph obtained 11 days after surgery shows grafts (asterisks) at L3–4 and L4–5 with intact endplates (solid arrows). Difference in appearance of rectangular-shaped grafts is due to long axis of graft at L3–4 level being rotated in more anteroposterior orientation (dotted arrow). B, Follow-up radiograph obtained 102 days after surgery shows marked resorption of inferior L3 endplate (solid arrows) and minimal resorption of superior and inferior L4 endplates with minimal graft resorption (asterisks). New bone formation is present in posterior L3–4 disk space, and between grafts and adjacent endplates (dotted arrows). It is important to be aware that resorption and bone formation can occur simultaneously.
A
B
marized as either being absent (Y = 0) or present (Y = 1). However, contrary to logistic regression, in which statistical inference is based on asymptotic large sample theory, exact logistic regression is based on the exact conditional distribution of the outcome variable and, hence, allows valid statistical inferences to be made even when the sample size or the frequency of one of the outcome categories (i.e., Y = 0 or Y = 1) is small. Univariate and multivariate exact conditional logistic analyses—In conducting the exact conditional logistic regression analyses, we used a two-step analytical process. For each outcome
variable, we initially conducted a set of univariate exact logistic regression analyses to determine whether one or more potential concomitant variables (i.e., confounders) was associated with the outcome variable. The set of potential concomitant variables included patient sex and age, graft shape, single or multiple fusion level, interbody fusion level, and surgeon. For patient age, three age groups were determined ( 61 years) to best obtain an equal distribution. For each univariate analysis, we used a p ≤ 0.05, decision rule as the criterion for rejecting the null hypothesis that the concomitant and outcome
variable were not associated. In step two of the analytical process, we added the rhBMP-2 status variable to a regression model that included all of the concomitant variables, if any, that were found to be associated with the outcome variable. Endplate resorption analysis—For the exact condition logistic regression analysis, Yi was assigned the value 1 if endplate resorption was detected for subject i, and Yi was assigned the value 0 otherwise. The model predictor variables included an indicator variable for rhBMP-2 use (yes or no) and indicator variable for surgeon. The estimate for the conditional adjusted odds ratio (OR) was
A
B
C
Fig. 2—49-year-old woman who underwent transforaminal lumbar interbody fusion at L4–5 and L5–S1 levels augmented with human recombinant morphogenetic protein–2. A, Initial postsurgical radiograph shows appropriate positioning of graft with intact endplate (arrow). B, Radiograph obtained as part of routine follow-up on day 92 after surgery shows moderate endplate resorption (arrows). C, Subsequent lateral radiograph obtained for routine follow-up at 393 days after surgery shows resolution of endplate resorption with new bone formation adjacent to allograft (arrows). Note that slight change in graft position is likely related to slight rotation of radiograph and not migration.
590
AJR:206, March 2016
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Imaging of Transforaminal Lumbar Interbody Fusion Fig. 3—53-year-old man who underwent transforaminal lumbar interbody fusion at L4–5 augmented with human recombinant morphogenetic protein–2. A, Initial radiograph shows graft (asterisk). B, Follow-up radiograph obtained at 195 days after surgery shows robust osseous bridging of disk space (arrows). Graft (asterisk) is also shown.
A
B
determined on the basis of the exact method, and the p value for the exact test of no association between rhBMP-2 status and resolved status was determined using the mid-p method of SAS version 9.2 (SAS Institute). Resorption resolution analysis—For the exact conditional logistic regression analysis, Yi was assigned the value 1 if resorption resolution was detected for subject i, and Yi was assigned the value 0 otherwise. The model predictor variables included an indicator variable for rhBMP-2 use (yes or no), and a categoric variable for subject age ( 61 years). The estimate for the conditional adjusted OR was determined on the basis of the median unbiased estimate, and the p value for the exact test of no association between rhBMP-2 status and resolved status was determined via the mid-p method (SAS Institute). New bone formation analysis—For the exact conditional logistic regression analysis, Yi was assigned the value 1 if new bone was detected for subject i, and Yi was assigned the value 0 otherwise. The sole model predictor variable was an indicator for rhBMP-2 use (yes or no). Osseous bridging analysis—For the exact conditional logistic regression analysis, Yi was assigned the value 1 if osseous bridged bone was detected for subject i, and Yi was assigned the value 0
otherwise. The sole model predictor variables was an indicator variable for rhBMP-2 use (yes or no). Allograft migration analysis—For the exact conditional logistic regression analysis, Yi was assigned the value 1 if allograft migration was detected for subject i, and Yi was assigned the value 0 otherwise. The model predictor variables included an indicator variable for rhBMP-2 use (yes or no), a categoric variable for subject age ( 61 years), graft shape, and an indicator for surgeon. With regard to hypothesis testing, for each of the aforementioned exact logistic regression analysis, the estimate for the conditional OR for rhBMP-2 equals yes versus rhBMP-2 equals no was determined on the basis of the exact methods of the Logistic procedure of SAS version 9.2 (SAS Institute), and the p value for the exact test of no association was determined via the mid-p method of the same SAS procedure. All of the statistical analyses were conducted by a professional biostatistician.
A
Results Endplate Resorption Endplate resorption was seen in 38% (78/204) of the disk spaces in patients treated with rhBMP-2 versus 12% (6/52) of the disk spaces in the control patients (p < 0.001).
B
There was no statistically significant association between resorption and patient sex (p = 0.588), patient age (p = 0.246), graft shape (p = 0.691), or number of levels fused (p = 0.586). There was a statistically significant difference in rate of resorption and the surgeons (p = 0.032). Multivariate analysis (indicator variable of rhBMP-2 use and a categoric variable of surgeon) resulted in an OR of 4.67 (95% CI, 1.99–12.54; p < 0.001) for resorption occurring in the rhBMP-2 group compared with patients who did not receive rhBMP-2 (Table 3). Endplate resorption was identified at a median of 55 postoperative days in the rhBMP-2 group and 92 days in the group without rhBMP-2 (p = 0.414). Resorption Resolution Endplate resorption resolved in 59% (46/78) of the rhBMP-2-treated disk spaces, compared with 0% (0/6) of the control group (p = 0.007). There was no statistically significant association between resorption resolution and patient sex (p = 0.18), graft shape (p = 0.827), surgeon (p = 0.115), or number of levels fused (p = 0.502). Resorption resolution did occur more often in younger patients (p ≤ 0.001). Multivariate analysis
Fig. 4—49-year-old woman who underwent transforaminal lumbar interbody fusion at L4–5 level augmented with human recombinant morphogenetic protein–2. A, Initial radiograph shows graft (asterisk) within disk space. B, Subsequent routine follow-up radiograph obtained 40 days after surgery shows marked posterior migration of allograft (asterisk). Note that apparent increase in anteroposterior length of graft is because graft has curvilinear cylindric shape and, on migrating posteriorly, it also rotated (arrows), resulting in apparent increase in size.
AJR:206, March 2016 591
Stensby et al. TABLE 3: Summary of Radiographic Findings by Group Radiographic Finding
Group Treated With Human Recombinant Morphogenetic Protein–2
Control Group
p
Odds Ratio (95% CI)
38 (78/204)
12 (6/52)
< 0.001
4.67 (1.99–12.54)
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Endplate resorptiona Resolution of resorptionb
59 (46/78)
0 (0/6)
0.022
8.09 (1.41 to ∞)
New bone formation
84 (171/204)
67 (35/52)
0.011
2.51 (1.24–4.99)
Osseous bridging
55 (112/204)
31 (16/52)
0.002
2.73 (1.43–5.34)
Allograft migrationa,b,c
17 (35/204)
2 (1/52)
0.065
6.30 (0.91–151.41)
Note—Data are percentage (no./total) of patients. Multivariate analysis was adjusted when necessary. a Adjusted for surgeon. bAdjusted for age. cAdjusted for graft shape.
(indicator variable of rhBMP-2 use and a categoric variable of patient age) resulted in an OR of 8.09 (95% CI, 1.41 to ∞; p = 0.022) for resorption resolution occurring in the rhBMP-2 group compared with those who did not receive rhBMP-2 (Table 3). New Bone Formation New bone formation was identified in 84% (171/204) of the disk spaces in the rhBMP-2 group and in 67% (35/52) of disk spaces in the control group (p = 0.011). There was no statistically significant association between new bone formation and patient sex (p = 0.334), patient age (p = 0.345), graft shape (p = 1.00), surgeon (p = 0.316), or number of levels fused (p = 0.521). Multivariate analysis (indicator variable of rhBMP-2 use) was performed, which resulted in an OR of 2.51 (95% CI 1.24–4.99; p = 0.011) for new bone formation occurring in the rhBMP-2 group compared with the control group (Table 3). The median time until new bone formation was identified was 99 postoperative days in the rhBMP-2 group and 164 days in the control group (p = 0.063). Osseous Bridging Osseous bridging of the disk space occurred in 55% (112/204) of the rhBMP-2 group and 31% (16/52) of the control group (p = 0.003). There was no statistically significant association between osseous bridging and patient sex (p = 0.798), patient age (p = 0.39), graft shape (p = 0.381), surgeon (p = 0.228), or number of levels fused (p = 0.898). Multivariate analysis (indicator variable of rhBMP-2 use) resulted in an OR of 2.73 (95% CI 1.43–5.34; p = 0.002) for osseous bridging in the rhBMP-2 group compared with the control group (Table 3). Osseous bridging occurred earlier in the rhBMP-2 group, at a median of 190 versus 353 days after TLIF (p = 0.008).
592
Allograft Migration Allograft migration was more frequent in the rhBMP-2 group, 17% (35/204), compared with 2% (1/52) in the control group (p = 0.003). There was no statistically significant association between allograft migration and patient sex (p = 0.466) or number of levels fused (p = 0.197). Allograft migration was more frequent in older patients (p = 0.022), and it occurred more than twice as often with grafts having the curved cylindric shape (p = 0.007). There was also a statistically significant association between allograft migration and surgeon (p = 0.003). Multivariate analysis (indicator variable of rhBMP-2 use and categoric variables of patient age, graft shape, and surgeon) resulted in an OR of 6.30 (95% CI, 0.91–151.41; p = 0.065) for allograft migration in the rhBMP-2 group compared with the control group. Allograft migration was identified at a median of 35 postoperative days in the rhBMP-2 group. Discussion Recombinant rhBMP-2 was used in more than 40% of posterior lumbar interbody fusions in 2011, the most recent year for which data are available [14]. Given the frequent use of rhBMP-2 in lumbar spinal fusions at many institutions and the routine evaluation of postoperative spinal fusion with radiographs, it is essential that radiologists are cognizant of the expected radiographic imaging findings after TLIF with rhBMP-2 to avoid misconstruing these findings as postoperative complications. We found endplate resorption to be a common finding after intervertebral disk fusions augmented with rhBMP-2 on radiographs performed at routine postoperative followup clinic visits; it was present in 38% of the treated disks in our study. Endplate resorption has been theorized to result from dosedependent stimulation of osteoclasts and
local inflammatory response [20, 21]. Regardless of the physiologic cascade induced by BMPs that leads to endplate resorption, it is important for the radiologist to know that endplate resorption may be an expected postoperative finding when rhBMP-2 is used during a TLIF. It is also important to note that endplate resorption usually occurs within the first 90 days after TLIF; the median was 55 days in our study. If there is clinical concern for infection, differentiating early osteomyelitis from endplate resorption associated with rhBMP-2 use may not be possible radiographically, and the use of pertinent clinical and laboratory data are indicated. The use of MRI is also important to exclude a paraspinal mass or epidural collection, findings that are not present with rhBMP-2 use but that are usually associated with osteomyelitis and discitis [22]. Because infection was not a clinical concern for any of the patients included in our study, we were able to observe that the endplate resorption was transient, resolving in nearly 60% of the involved endplates on follow-up radiographs as a result of new bone formation. We evaluated the TLIF levels, both with and without the use of rhBMP-2, in a similar manner and found that rhBMP-2 results in a statistically higher incidence of new bone formation and osseous bridging of the disk space compared with TLIFs performed without rhBMP-2. Radiographically evident new bone formation was reported in 83% of patients 1 year after TLIF by Sethi et al. [17], findings similar to our rate of bone formation in the rhBMP-2 group. New bone formation and bone bridging occurred earlier in the TLIFs for which rhBMP-2 was used compared with the control group, with bone bridging usually occurring between 6 and 9 months after surgery. All but one of the 36 lumbar fusions showing allograft migration was augmented by rhBMP-2. However, after adjusting
AJR:206, March 2016
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Imaging of Transforaminal Lumbar Interbody Fusion for the three variables also associated with a statistically higher incidence of graft migration (graft shape, surgeon, and patient age), the use of rhBMP-2 was only a marginally significant factor in graft migration. Our findings suggest that the graft shape may be a more important factor in graft migration, with grafts shaped like curved cylinders more prone to migration than square or block-shaped grafts. Graft type was reported to be statistically significantly (p ≤ 0.001) associated with higher rates of graft migration by Sethi et al. [17]. These findings raise the importance of scrutinizing the position of the interbody graft on postoperative radiographs, especially after TLIF and when more cylindric grafts are used, because posterior migration of the allograft may lead to impingement on the neural foramina or central canal. Our study design was limited by the fact that it was a retrospective review. As such, the number and frequency of imaging studies each patient received was variable, and some patients were lost to follow-up. However, patients are lost to follow-up even in prospective studies, and the percentage of patients who remained in our study at the designated follow-up time periods, as well as the average duration of follow-up, were not statistically significantly different between the two study groups. Another limitation is that radiographs and not CT were used to access the studied variables. Although CT is more sensitive for the findings recorded [23], radiographic follow-up is the standard of care in most institutions for monitoring patients after TLIF surgery, and it is likely that our results will be more widely applicable than studies that used only CT for follow-up. Isolated use of radiographs precluded a quantitative evaluation of subsidence because of the variation in the centering of the x-ray beam, which did not allow an accurate assessment of 10% disk height loss, which was the definition of subsidence in a prior study [17]. Although the dose of rhBMP-2 used in the surgery was recorded in all cases, the percentage of rhBMP-2 used in the disk space to that used in the fusion of the posterior elements was not always clearly delineated, precluding a detailed analysis. However, a recent study evaluating dose-related complications in over 500 patients found that the rate of complications was so infrequent that a correlation with a particular dose was not possible [24]. All of the patients in our study received doses similar to that used
by C randall et al. [24]. Finally, although we would anticipate that our findings associated with rhBMP-2 use in TLIF with allograft would likely be similar in patients with anterior lumbar interbody fusion and posterior lumbar interbody fusion surgeries or in patients in which autograft was used, we did not directly address this issue in our study. In conclusion, rhBMP-2 use during TLIF has increased dramatically since it received FDA approval. A statistically significantly higher frequency of endplate resorption, new bone formation, and bone bridging is present in TLIF augmented by rhBMP-2 compared with TLIF performed without rhBMP-2; endplate resorption resolves without treatment in most cases. Although the endplate resorption can mimic early osteomyelitis radiographically, our patients exhibited no signs or symptoms of infection, and none was treated for discitis or osteomyelitis. Because radiographs are the primary modality used to image postoperative patients, awareness of the time sequence of resorptive and healing changes both with and without rhBMP-2, as well as the relative frequency of these events, is integral to interpreting these examinations. References
1. Banwart JC, Asher MA, Hassanein RS. Iliac crest bone graft harvest donor site morbidity: a statistical evaluation. Spine 1995; 20:1055–1060 2. Boden SD, Zdeblick TA, Sandhu HS, Heim SE. The use of rhBMP-2 in interbody fusion cages: definitive evidence of osteoinduction in humans—a preliminary report. Spine 2000; 25:376–381 3. Burkus JK, Gornet MF, Dickman CA, Zdeblick TA. Anterior lumbar interbody fusion using rhBMP-2 with tapered interbody cages. J Spinal Disord Tech 2002; 15:337–349 4. Dimar JR, Glassman SD, Burkus KJ, Carreon LY. Clinical outcomes and fusion success at 2 years of single-level instrumented posterolateral fusions with recombinant human bone morphogenetic protein-2/compression resistant matrix versus iliac crest bone graft. Spine 2006; 31:2534–2539; discussion, 2540 5. Bae HW, Zhao L, Kanim LE, Wong P, Delamarter RB, Dawson EG. Intervariability and intravariability of bone morphogenetic proteins in commercially available demineralized bone matrix products. Spine 2006; 31:1299–1306; discussion, 1307–1308 6. Finkemeier CG. Bone-grafting and bone-graft substitutes. J Bone Joint Surg Am 2002; 84-A:454–464 7. Malarkey M, Solomon R, Witten C, et al.; Centers for Disease Control and Prevention (CDC). Brief report: investigation into recalled human tissue
for transplantation—United States, 2005–2006. CDC website. www.cdc.gov/mmwr/preview/ mmwrhtml/mm5520a6.htm. Published May 25, 2006. Accessed April 4, 2014 8. Bostrom MP, Yang X, Kennan M, Sandhu H, Dicarlo E, Lane JM. An unexpected outcome during testing of commercially available demineralized bone graft materials: how safe are the nonallograft components? Spine 2001; 26:1425–1428 9. Wang JC, Kanim LE, Nagakawa IS, Yamane BH, Vinters HV, Dawson EG. Dose-dependent toxicity of a commercially available demineralized bone matrix material. Spine 2001; 26:1429–1435; discussion, 1435–1436 10. Urist MR. Bone: formation by auto induction. Science 1965; 150:893–899 11. U.S. Food and Drug Administration (FDA). InFUSE™ Bone Graft/LT-CAGE™ Lumbar Tapered Fusion Device—P000058. FDA website. www.fda.gov/MedicalDevices/ProductsandMedicalProcedures/DeviceApprovalsandClearances/ Recently-ApprovedDevices/ucm083423.htm. Published 2013. Accessed April 4, 2014 12. Deyo RA, Ching A, Matsen L, et al. Use of bone morphogenetic proteins in spinal fusion surgery for older adults with lumbar stenosis: trends, complications, repeat surgery, and charges. Spine 2012; 37:222–230 13. Papakostidis C, Kontakis G, Bhandari M, Giannoudis PV. Efficacy of autologous iliac crest bone graft and bone morphogenetic proteins for posterolateral fusion of lumbar spine: a meta-analysis of the results. Spine 2008; 33:E680–E692 14. Singh K, Nandyala SV, Marquez-Lara A, Fineberg SJ. Epidemiological trends in the utilization of bone morphogenetic protein in spinal fusions from 2002 to 2011. Spine 2014; 39:491–496 15. Williams BJ, Smith JS, Fu KM, et al.; Scoliosis Research Society Morbidity and Mortality Committee. Does bone morphogenetic protein increase the incidence of perioperative complications in spinal fusion? A comparison of 55,862 cases of spinal fusion with and without bone morphogenetic protein. Spine 2011; 36:1685–1691 16. Singh K, Ahmadinia K, Park DK, et al. Complications of spinal fusion with utilization of bone morphogenetic protein: a systematic review of the literature. Spine 2014; 39:91–101 17. Sethi A, Craig J, Bartol S, et al. Radiographic and CT evaluation of recombinant human bone morphogenetic protein-2–assisted spinal interbody fusion. AJR 2011; 197:[web]W128–W133 18. Ong KL, Villarraga ML, Lau E, Carreon LY, Kurtz SM, Glassman SD. Off-label use of bone morphogenetic proteins in the United States using administrative data. Spine 2010; 35:1794–1800 19. Vaidya R, Sethi A, Bartol S, Jacobson M, Coe C, Craig JG. Complications in the use of rhBMP-2 in
AJR:206, March 2016 593
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Stensby et al. PEEK cages for interbody spinal fusions. J Spinal Disord Tech 2008; 21:557–562 20. Muchow RD, Hsu WK, Anderson PA. Histopathologic inflammatory response induced by recombinant bone morphogenetic protein-2 causing radiculopathy after transforaminal lumbar interbody fusion. Spine J 2010; 10:e1–e6 21. Otsuka E, Notoya M, Hagiwara H. Treatment of myoblastic C2C12 cells with BMP-2 stimulates
vitamin D-induced formation of osteoclasts. Calcif Tissue Int 2003; 73:72–77 22. Fox MG, Goldberg JM, Gaskin CM, et al. MRI of transforaminal lumbar interbody fusion: imaging appearance with and without the use of human recombinant bone morphogenetic protein-2 ( rhBMP-2). Skeletal Radiol 2014; 43:1247–1255 23. Krestan CR, Noske H, Vasilevska V, et al. MDCT versus digital radiography in the evaluation of
bone healing in orthopedic patients. AJR 2006; 186:1754–1760 24. Crandall DG, Revella J, Patterson J, Huish E, Chang M, McLemore R. Transforaminal lumbar interbody fusion with rhBMP-2 in spinal deformity, spondylolisthesis, and degenerative disease. Part 2. BMP dosage-related complications and long-term outcomes in 509 patients. Spine 2013; 38:1137–1145
F O R YO U R I N F O R M AT I O N
This article is available for CME and Self-Assessment (SA-CME) credit that satisfies Part II requirements for maintenance of certification (MOC). To access the examination for this article, follow the prompts associated with the online version of the article.
594
AJR:206, March 2016
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Musculoskeletal Imaging Original Research
Radiographic Appearance of Transforaminal Lumbar Interbody Fusion Performed With and Without Recombinant Human Morphogenetic Protein–2 J. Derek Stensby 1,2 Ryan W. Kaliney 1,3 Bennett Alford1 Francis H. Shen 4 James T. Patrie5 Michael G. Fox1 Stensby JD, Kaliney RW, Alford B, Shen FH, Patrie JT, Fox MG Keywords: bone resorption, lumbosacral region, osteogenesis, recombinant human bone morphogenetic protein–2, spinal fusion DOI:10.2214/AJR.15.14503 Received February 6, 2015; accepted after revision May 13, 2015. F. H. Shen has consulting agreements with DePuy Synthes Spine, Globus Medical, and Medtronic Spine. He is on the Medical Board of Trustees for Musculoskeletal Transplant Foundation, and has royalties from Globus Medical and Elsevier Publishing. Based on a presentation at the Radiological Society of North America 2011 annual meeting, Chicago, IL. 1
Department of Radiology and Medical Imaging, University of Virginia, 1218 Lee St, Box 800170, Charlottesville, VA 22908. Address correspondence to M. G. Fox ([email protected]). 2 Present address: Mallinckrodt Institute of Radiology, St. Louis, MO. 3
Present address: Jefferson Radiology, Hartford, CT.
4
Department of Orthopedic Surgery, University of Virginia, Charlottesville, VA. 5
Department of Public Health Sciences, University of Virginia, Charlottesville, VA.
This article is available for credit. AJR 2016; 206:588–594 0361–803X/16/2063–588 © American Roentgen Ray Society
588
OBJECTIVE. The purpose of this study is to determine whether recombinant human morphogenetic protein–2 (rhBMP-2) alters the findings on routine radiographs performed after transforaminal lumbar interbody fusion (TLIF). MATERIALS AND METHODS. A retrospective review of 256 TLIF procedures in 200 patients was performed over a 4-year period. The rhBMP-2 group included 204 TLIFs in 160 patients, and the control group included 52 TLIFs in 40 patients. Two musculoskeletal radiologists reviewed the postoperative radiographs for endplate resorption, resorption resolution, new bone formation, bridging bone, and allograft migration. Statistical analysis was performed using logistic regression. RESULTS. The median age was 53 years in the rhBMP-2 group and 54 years in the control group (p = 0.182). The groups were similar with regard to sex (p = 0.517), single or multilevel TLIF (p = 0.921), specific TLIF levels (p = 0.53), and median radiographic follow-up (373 vs 366 days; p = 0.34). Findings that were more common in the rhBMP-2 group than in the control group included endplate resorption (38% [78/204] vs 12% [6/52]; odds ratio [OR], 4.67; 95% CI, 1.99–12.54; p < 0.001), resorption resolution (59% [46/78] vs 0% [0/6]; OR, 8.09; 95% CI, 1.41 to ∞; p = 0.022), new bone formation (84% [171/204] vs 67% [35/52]; OR, 2.51; 95% CI, 1.24–4.99; p = 0.011), bridging bone (55% [112/204] vs 31% [16/52]; OR, 2.73; 95% CI, 1.43–5.34; p = 0.002), and allograft migration (17% [35/204] vs 2% [1/52]; OR, 6.30; 95% CI, 0.91–151.41; p = 0.065). CONCLUSION. A statistically significant higher frequency of endplate resorption, new bone formation, and bone bridging is present in TLIF augmented by rhBMP-2 compared with TLIF performed without rhBMP-2. Endplate resorption resolves without treatment in most cases after rhBMP-2 use. utologous iliac crest bone graft has long been considered the method of choice for augmentation of vertebral body fusion. Harvesting the autologous iliac bone graft is an additional procedure associated with major complications in 0.7–25% and minor complications in 9.4–39% of cases [1]. Increased complications, increased operative time, increased blood loss, and insufficient autograft for complex or revision procedures have led to the use of alternative graft types [2–4]. One alternative for graft augmentation introduced in 1991 is demineralized bone matrix; however, differences in the harvesting and sterilization processes resulted in variable osteoconductive properties in the commercially available varieties [5, 6]. The potential for infectious disease transmission [6, 7], the use of toxic binders in certain prep-
A
arations of demineralized bone matrix, and the desire to increase fusion rates fueled the search for acceptable alternatives to augment spinal fusion [8, 9]. Bone morphogenic proteins (BMPs) were discovered in 1965, when new bone formation was observed after implantation of devitalized bone into animal models in a dosedependent manner [10]. Not until 2002 did recombinant human BMP–2 (rhBMP-2) receive U.S. Food and Drug Administration (FDA) approval for use in anterior lumbar interbody fusion [11]; the commercially available product is InFUSE (Medtronic Sofamor Danek) [12]. A subsequent meta-analysis evaluating the use of rhBMP-2 as a supplement to or replacement for autologous bone graft for lumbar spinal fusion reported increased fusion rates [13]. Given its potential benefits, the use of rhBMP-2 in spinal sur-
AJR:206, March 2016
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Imaging of Transforaminal Lumbar Interbody Fusion gery increased from less than 1% of posterior lumbar interbody fusions in 2002 to more than 40% in 2011 [14]. Previous studies have compared the postoperative complication rate after lumbar fusion both with and without rhBMP-2 augmentation [15, 16]. A search of the radiology literature for descriptions of the radiographic appearance of transforaminal lumbar interbody fusion (TLIF) after the use of rhBMP-2 yielded one study of 36 cases without a control group for comparison [17]. To our knowledge, the radiographic appearance of TLIF performed with and without the use of rhBMP-2 in the disk space has not been directly compared. We sought to determine whether the radiographic appearance after TLIF varied between patients who had received allograft augmented by rhBMP-2 and patients who had received allograft alone. Materials and Methods
A retrospective radiographic analysis of patients who underwent TLIF at a single academic institution (University of Virginia) between August 2005 and September 2009 was performed after institutional review board approval. TLIF is the preferred method for performing posterior lumbar spinal fusion at our institution because it requires less nerve root retraction than does posterior lumbar interbody fusion as a result of the more lateral transforaminal exposure, although it requires complete resection of the facet joint complex unilaterally, unlike a posterior lumbar interbody fusion, which preserves greater than 50% of the facet joint complex but resects varying degrees of the lamina and spinous process. Two surgeons performed the TLIFs using two varieties of allografts; one group of allografts had a curved cylindric shape and the other group had a more blocklike shape. The shape of the allograft and the use of rhBMP-2 (InFUSE, Medtronic Sofamor Danek) was decided by the surgeon, with the rhBMP-2 placed on an absorbable collagen sponge and placed in the disk space posterior to the allograft. The surgical technique used to perform the TLIF was otherwise similar, with the cartilaginous endplates removed, but the vertebral endplate along with the corresponding subchondral bone left intact to provide endplate support to reduce the risk of graft migration. Because rhBMP-2 is FDA approved for only anterior lumbar interbody fusion when used with the Lumbar Tapered Fusion Device (Medtronic Sofamor Danek), the use of rhBMP-2 for TLIF is considered off label [11]. We did not study anterior lumbar interbody fusion because this technique requires an anterior approach and involves the assistance of a
TABLE 1: Summary of Disk Space Levels Fused Disk Space Level
Group Treated With Human Recombinant Morphogenetic Protein–2
Control Group
T12–L1
1
0
L1–2
1
0
L2–3
11
2
L3–4
26
6
L4–5
83
25
L5–S1
82
19
Note—Data are number of patients. p = 0.53 for all comparisons.
vascular surgeon to gain exposure to the anterior disk space. Even so, over 80% of the spinal fusions augmented with rhBMP-2 in the United States have been performed using an off-label technique [14, 18].
Patients
The study included 200 patients who underwent fusion of 256 intervertebral disk spaces; 160 patients underwent fusion of 204 intervertebral disk spaces augmented by rhBMP-2, and 40 patients underwent fusion of 52 intervertebral disk spaces without rhBMP-2 treatment. Patients younger than 18 years (n = 1), patients clinically suspected of having deep soft-tissue infection or discitis or osteomyelitis (n = 1), and those undergoing revision surgeries were excluded (n = 6). There were 63 men and 97 women in the rhBMP-2 group and 18 men and 22 women in the control group (p = 0.517). The median age was 53 years in the rhBMP-2 group and 54 years in the control group (p = 0.182). Single-level fusion was performed in 76 (47.5%) patients treated with rhBMP-2 and in 22 (55%) of the patients in the control group (p = 0.921). Individual surgical levels are displayed in Table 1. The study was not performed in a prospective manner, and some variability in the interval and duration of radiographic follow-up was unavoidable. Occasionally, patients were lost to follow-up or had intermittent follow-up visits (e.g., missed follow-up at 6 months but presented for follow-up at 12 months). Nevertheless, the median duration of follow-up and
the follow-up intervals between the groups were not statistically significantly different (Table 2). Anteroposterior and lateral radiographs performed at routine postoperative visits were reviewed in consensus by two fellowship-trained musculoskeletal radiologists, with 8 years and 1 year of experience, using a PACS workstation (Carestream, version 10.2, Carestream Health). Radiographs were compared with the baseline immediate postoperative radiograph for the development of endplate resorption, resolution of resorption, new bone formation, bridging bone, and graft migration. Endplate resorption was categorized as none, mild, moderate, or severe, as described by Vaidya et al. [19] (Fig. 1). Resorption was considered to have resolved when bone density replaced the endplate defect (Fig. 2). New bone formation was present when ossification was identified within the intervertebral disk space, with bone bridging occurring when the ossification joined the superior and inferior endplates (Fig. 3). Graft migration was diagnosed when a change in allograft position, from the initial postoperative radiograph, occurred (Fig. 4).
Statistical Analysis
The odds of endplate resorption, resorption resolution, new bone formation, osseous bridging, and allograft migration were compared between the patients who received and did not receive rhBMP-2 via exact logistic regression. Note that exact logistic regression is comparable to logistic regression in that the method is appropriate for analyzing of outcome variables that can be sum-
TABLE 2: Radiographic Follow-Up Intervals, by Patient Group Follow-Up Interval (Postoperative Days)
Group Treated With Human Recombinant Morphogenetic Protein–2 (n = 204)
1–91 92–183
Control Group
p
99.0 (202)
100 (52)
0.47
77.5 (158)
67.3 (35)
0.13
184–365
69.1 (141)
73.1 (38)
0.58
366–730
56.4 (115)
50 (26)
0.41
Median
373
366
0.34
Note—Data are percentage (no.) of patients.
AJR:206, March 2016 589
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Stensby et al. Fig. 1—23-year-old woman who underwent transforaminal lumbar interbody fusion at L3–4 and L4–5 augmented with human recombinant morphogenetic protein–2. A, Initial radiograph obtained 11 days after surgery shows grafts (asterisks) at L3–4 and L4–5 with intact endplates (solid arrows). Difference in appearance of rectangular-shaped grafts is due to long axis of graft at L3–4 level being rotated in more anteroposterior orientation (dotted arrow). B, Follow-up radiograph obtained 102 days after surgery shows marked resorption of inferior L3 endplate (solid arrows) and minimal resorption of superior and inferior L4 endplates with minimal graft resorption (asterisks). New bone formation is present in posterior L3–4 disk space, and between grafts and adjacent endplates (dotted arrows). It is important to be aware that resorption and bone formation can occur simultaneously.
A
B
marized as either being absent (Y = 0) or present (Y = 1). However, contrary to logistic regression, in which statistical inference is based on asymptotic large sample theory, exact logistic regression is based on the exact conditional distribution of the outcome variable and, hence, allows valid statistical inferences to be made even when the sample size or the frequency of one of the outcome categories (i.e., Y = 0 or Y = 1) is small. Univariate and multivariate exact conditional logistic analyses—In conducting the exact conditional logistic regression analyses, we used a two-step analytical process. For each outcome
variable, we initially conducted a set of univariate exact logistic regression analyses to determine whether one or more potential concomitant variables (i.e., confounders) was associated with the outcome variable. The set of potential concomitant variables included patient sex and age, graft shape, single or multiple fusion level, interbody fusion level, and surgeon. For patient age, three age groups were determined ( 61 years) to best obtain an equal distribution. For each univariate analysis, we used a p ≤ 0.05, decision rule as the criterion for rejecting the null hypothesis that the concomitant and outcome
variable were not associated. In step two of the analytical process, we added the rhBMP-2 status variable to a regression model that included all of the concomitant variables, if any, that were found to be associated with the outcome variable. Endplate resorption analysis—For the exact condition logistic regression analysis, Yi was assigned the value 1 if endplate resorption was detected for subject i, and Yi was assigned the value 0 otherwise. The model predictor variables included an indicator variable for rhBMP-2 use (yes or no) and indicator variable for surgeon. The estimate for the conditional adjusted odds ratio (OR) was
A
B
C
Fig. 2—49-year-old woman who underwent transforaminal lumbar interbody fusion at L4–5 and L5–S1 levels augmented with human recombinant morphogenetic protein–2. A, Initial postsurgical radiograph shows appropriate positioning of graft with intact endplate (arrow). B, Radiograph obtained as part of routine follow-up on day 92 after surgery shows moderate endplate resorption (arrows). C, Subsequent lateral radiograph obtained for routine follow-up at 393 days after surgery shows resolution of endplate resorption with new bone formation adjacent to allograft (arrows). Note that slight change in graft position is likely related to slight rotation of radiograph and not migration.
590
AJR:206, March 2016
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Imaging of Transforaminal Lumbar Interbody Fusion Fig. 3—53-year-old man who underwent transforaminal lumbar interbody fusion at L4–5 augmented with human recombinant morphogenetic protein–2. A, Initial radiograph shows graft (asterisk). B, Follow-up radiograph obtained at 195 days after surgery shows robust osseous bridging of disk space (arrows). Graft (asterisk) is also shown.
A
B
determined on the basis of the exact method, and the p value for the exact test of no association between rhBMP-2 status and resolved status was determined using the mid-p method of SAS version 9.2 (SAS Institute). Resorption resolution analysis—For the exact conditional logistic regression analysis, Yi was assigned the value 1 if resorption resolution was detected for subject i, and Yi was assigned the value 0 otherwise. The model predictor variables included an indicator variable for rhBMP-2 use (yes or no), and a categoric variable for subject age ( 61 years). The estimate for the conditional adjusted OR was determined on the basis of the median unbiased estimate, and the p value for the exact test of no association between rhBMP-2 status and resolved status was determined via the mid-p method (SAS Institute). New bone formation analysis—For the exact conditional logistic regression analysis, Yi was assigned the value 1 if new bone was detected for subject i, and Yi was assigned the value 0 otherwise. The sole model predictor variable was an indicator for rhBMP-2 use (yes or no). Osseous bridging analysis—For the exact conditional logistic regression analysis, Yi was assigned the value 1 if osseous bridged bone was detected for subject i, and Yi was assigned the value 0
otherwise. The sole model predictor variables was an indicator variable for rhBMP-2 use (yes or no). Allograft migration analysis—For the exact conditional logistic regression analysis, Yi was assigned the value 1 if allograft migration was detected for subject i, and Yi was assigned the value 0 otherwise. The model predictor variables included an indicator variable for rhBMP-2 use (yes or no), a categoric variable for subject age ( 61 years), graft shape, and an indicator for surgeon. With regard to hypothesis testing, for each of the aforementioned exact logistic regression analysis, the estimate for the conditional OR for rhBMP-2 equals yes versus rhBMP-2 equals no was determined on the basis of the exact methods of the Logistic procedure of SAS version 9.2 (SAS Institute), and the p value for the exact test of no association was determined via the mid-p method of the same SAS procedure. All of the statistical analyses were conducted by a professional biostatistician.
A
Results Endplate Resorption Endplate resorption was seen in 38% (78/204) of the disk spaces in patients treated with rhBMP-2 versus 12% (6/52) of the disk spaces in the control patients (p < 0.001).
B
There was no statistically significant association between resorption and patient sex (p = 0.588), patient age (p = 0.246), graft shape (p = 0.691), or number of levels fused (p = 0.586). There was a statistically significant difference in rate of resorption and the surgeons (p = 0.032). Multivariate analysis (indicator variable of rhBMP-2 use and a categoric variable of surgeon) resulted in an OR of 4.67 (95% CI, 1.99–12.54; p < 0.001) for resorption occurring in the rhBMP-2 group compared with patients who did not receive rhBMP-2 (Table 3). Endplate resorption was identified at a median of 55 postoperative days in the rhBMP-2 group and 92 days in the group without rhBMP-2 (p = 0.414). Resorption Resolution Endplate resorption resolved in 59% (46/78) of the rhBMP-2-treated disk spaces, compared with 0% (0/6) of the control group (p = 0.007). There was no statistically significant association between resorption resolution and patient sex (p = 0.18), graft shape (p = 0.827), surgeon (p = 0.115), or number of levels fused (p = 0.502). Resorption resolution did occur more often in younger patients (p ≤ 0.001). Multivariate analysis
Fig. 4—49-year-old woman who underwent transforaminal lumbar interbody fusion at L4–5 level augmented with human recombinant morphogenetic protein–2. A, Initial radiograph shows graft (asterisk) within disk space. B, Subsequent routine follow-up radiograph obtained 40 days after surgery shows marked posterior migration of allograft (asterisk). Note that apparent increase in anteroposterior length of graft is because graft has curvilinear cylindric shape and, on migrating posteriorly, it also rotated (arrows), resulting in apparent increase in size.
AJR:206, March 2016 591
Stensby et al. TABLE 3: Summary of Radiographic Findings by Group Radiographic Finding
Group Treated With Human Recombinant Morphogenetic Protein–2
Control Group
p
Odds Ratio (95% CI)
38 (78/204)
12 (6/52)
< 0.001
4.67 (1.99–12.54)
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Endplate resorptiona Resolution of resorptionb
59 (46/78)
0 (0/6)
0.022
8.09 (1.41 to ∞)
New bone formation
84 (171/204)
67 (35/52)
0.011
2.51 (1.24–4.99)
Osseous bridging
55 (112/204)
31 (16/52)
0.002
2.73 (1.43–5.34)
Allograft migrationa,b,c
17 (35/204)
2 (1/52)
0.065
6.30 (0.91–151.41)
Note—Data are percentage (no./total) of patients. Multivariate analysis was adjusted when necessary. a Adjusted for surgeon. bAdjusted for age. cAdjusted for graft shape.
(indicator variable of rhBMP-2 use and a categoric variable of patient age) resulted in an OR of 8.09 (95% CI, 1.41 to ∞; p = 0.022) for resorption resolution occurring in the rhBMP-2 group compared with those who did not receive rhBMP-2 (Table 3). New Bone Formation New bone formation was identified in 84% (171/204) of the disk spaces in the rhBMP-2 group and in 67% (35/52) of disk spaces in the control group (p = 0.011). There was no statistically significant association between new bone formation and patient sex (p = 0.334), patient age (p = 0.345), graft shape (p = 1.00), surgeon (p = 0.316), or number of levels fused (p = 0.521). Multivariate analysis (indicator variable of rhBMP-2 use) was performed, which resulted in an OR of 2.51 (95% CI 1.24–4.99; p = 0.011) for new bone formation occurring in the rhBMP-2 group compared with the control group (Table 3). The median time until new bone formation was identified was 99 postoperative days in the rhBMP-2 group and 164 days in the control group (p = 0.063). Osseous Bridging Osseous bridging of the disk space occurred in 55% (112/204) of the rhBMP-2 group and 31% (16/52) of the control group (p = 0.003). There was no statistically significant association between osseous bridging and patient sex (p = 0.798), patient age (p = 0.39), graft shape (p = 0.381), surgeon (p = 0.228), or number of levels fused (p = 0.898). Multivariate analysis (indicator variable of rhBMP-2 use) resulted in an OR of 2.73 (95% CI 1.43–5.34; p = 0.002) for osseous bridging in the rhBMP-2 group compared with the control group (Table 3). Osseous bridging occurred earlier in the rhBMP-2 group, at a median of 190 versus 353 days after TLIF (p = 0.008).
592
Allograft Migration Allograft migration was more frequent in the rhBMP-2 group, 17% (35/204), compared with 2% (1/52) in the control group (p = 0.003). There was no statistically significant association between allograft migration and patient sex (p = 0.466) or number of levels fused (p = 0.197). Allograft migration was more frequent in older patients (p = 0.022), and it occurred more than twice as often with grafts having the curved cylindric shape (p = 0.007). There was also a statistically significant association between allograft migration and surgeon (p = 0.003). Multivariate analysis (indicator variable of rhBMP-2 use and categoric variables of patient age, graft shape, and surgeon) resulted in an OR of 6.30 (95% CI, 0.91–151.41; p = 0.065) for allograft migration in the rhBMP-2 group compared with the control group. Allograft migration was identified at a median of 35 postoperative days in the rhBMP-2 group. Discussion Recombinant rhBMP-2 was used in more than 40% of posterior lumbar interbody fusions in 2011, the most recent year for which data are available [14]. Given the frequent use of rhBMP-2 in lumbar spinal fusions at many institutions and the routine evaluation of postoperative spinal fusion with radiographs, it is essential that radiologists are cognizant of the expected radiographic imaging findings after TLIF with rhBMP-2 to avoid misconstruing these findings as postoperative complications. We found endplate resorption to be a common finding after intervertebral disk fusions augmented with rhBMP-2 on radiographs performed at routine postoperative followup clinic visits; it was present in 38% of the treated disks in our study. Endplate resorption has been theorized to result from dosedependent stimulation of osteoclasts and
local inflammatory response [20, 21]. Regardless of the physiologic cascade induced by BMPs that leads to endplate resorption, it is important for the radiologist to know that endplate resorption may be an expected postoperative finding when rhBMP-2 is used during a TLIF. It is also important to note that endplate resorption usually occurs within the first 90 days after TLIF; the median was 55 days in our study. If there is clinical concern for infection, differentiating early osteomyelitis from endplate resorption associated with rhBMP-2 use may not be possible radiographically, and the use of pertinent clinical and laboratory data are indicated. The use of MRI is also important to exclude a paraspinal mass or epidural collection, findings that are not present with rhBMP-2 use but that are usually associated with osteomyelitis and discitis [22]. Because infection was not a clinical concern for any of the patients included in our study, we were able to observe that the endplate resorption was transient, resolving in nearly 60% of the involved endplates on follow-up radiographs as a result of new bone formation. We evaluated the TLIF levels, both with and without the use of rhBMP-2, in a similar manner and found that rhBMP-2 results in a statistically higher incidence of new bone formation and osseous bridging of the disk space compared with TLIFs performed without rhBMP-2. Radiographically evident new bone formation was reported in 83% of patients 1 year after TLIF by Sethi et al. [17], findings similar to our rate of bone formation in the rhBMP-2 group. New bone formation and bone bridging occurred earlier in the TLIFs for which rhBMP-2 was used compared with the control group, with bone bridging usually occurring between 6 and 9 months after surgery. All but one of the 36 lumbar fusions showing allograft migration was augmented by rhBMP-2. However, after adjusting
AJR:206, March 2016
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Imaging of Transforaminal Lumbar Interbody Fusion for the three variables also associated with a statistically higher incidence of graft migration (graft shape, surgeon, and patient age), the use of rhBMP-2 was only a marginally significant factor in graft migration. Our findings suggest that the graft shape may be a more important factor in graft migration, with grafts shaped like curved cylinders more prone to migration than square or block-shaped grafts. Graft type was reported to be statistically significantly (p ≤ 0.001) associated with higher rates of graft migration by Sethi et al. [17]. These findings raise the importance of scrutinizing the position of the interbody graft on postoperative radiographs, especially after TLIF and when more cylindric grafts are used, because posterior migration of the allograft may lead to impingement on the neural foramina or central canal. Our study design was limited by the fact that it was a retrospective review. As such, the number and frequency of imaging studies each patient received was variable, and some patients were lost to follow-up. However, patients are lost to follow-up even in prospective studies, and the percentage of patients who remained in our study at the designated follow-up time periods, as well as the average duration of follow-up, were not statistically significantly different between the two study groups. Another limitation is that radiographs and not CT were used to access the studied variables. Although CT is more sensitive for the findings recorded [23], radiographic follow-up is the standard of care in most institutions for monitoring patients after TLIF surgery, and it is likely that our results will be more widely applicable than studies that used only CT for follow-up. Isolated use of radiographs precluded a quantitative evaluation of subsidence because of the variation in the centering of the x-ray beam, which did not allow an accurate assessment of 10% disk height loss, which was the definition of subsidence in a prior study [17]. Although the dose of rhBMP-2 used in the surgery was recorded in all cases, the percentage of rhBMP-2 used in the disk space to that used in the fusion of the posterior elements was not always clearly delineated, precluding a detailed analysis. However, a recent study evaluating dose-related complications in over 500 patients found that the rate of complications was so infrequent that a correlation with a particular dose was not possible [24]. All of the patients in our study received doses similar to that used
by C randall et al. [24]. Finally, although we would anticipate that our findings associated with rhBMP-2 use in TLIF with allograft would likely be similar in patients with anterior lumbar interbody fusion and posterior lumbar interbody fusion surgeries or in patients in which autograft was used, we did not directly address this issue in our study. In conclusion, rhBMP-2 use during TLIF has increased dramatically since it received FDA approval. A statistically significantly higher frequency of endplate resorption, new bone formation, and bone bridging is present in TLIF augmented by rhBMP-2 compared with TLIF performed without rhBMP-2; endplate resorption resolves without treatment in most cases. Although the endplate resorption can mimic early osteomyelitis radiographically, our patients exhibited no signs or symptoms of infection, and none was treated for discitis or osteomyelitis. Because radiographs are the primary modality used to image postoperative patients, awareness of the time sequence of resorptive and healing changes both with and without rhBMP-2, as well as the relative frequency of these events, is integral to interpreting these examinations. References
1. Banwart JC, Asher MA, Hassanein RS. Iliac crest bone graft harvest donor site morbidity: a statistical evaluation. Spine 1995; 20:1055–1060 2. Boden SD, Zdeblick TA, Sandhu HS, Heim SE. The use of rhBMP-2 in interbody fusion cages: definitive evidence of osteoinduction in humans—a preliminary report. Spine 2000; 25:376–381 3. Burkus JK, Gornet MF, Dickman CA, Zdeblick TA. Anterior lumbar interbody fusion using rhBMP-2 with tapered interbody cages. J Spinal Disord Tech 2002; 15:337–349 4. Dimar JR, Glassman SD, Burkus KJ, Carreon LY. Clinical outcomes and fusion success at 2 years of single-level instrumented posterolateral fusions with recombinant human bone morphogenetic protein-2/compression resistant matrix versus iliac crest bone graft. Spine 2006; 31:2534–2539; discussion, 2540 5. Bae HW, Zhao L, Kanim LE, Wong P, Delamarter RB, Dawson EG. Intervariability and intravariability of bone morphogenetic proteins in commercially available demineralized bone matrix products. Spine 2006; 31:1299–1306; discussion, 1307–1308 6. Finkemeier CG. Bone-grafting and bone-graft substitutes. J Bone Joint Surg Am 2002; 84-A:454–464 7. Malarkey M, Solomon R, Witten C, et al.; Centers for Disease Control and Prevention (CDC). Brief report: investigation into recalled human tissue
for transplantation—United States, 2005–2006. CDC website. www.cdc.gov/mmwr/preview/ mmwrhtml/mm5520a6.htm. Published May 25, 2006. Accessed April 4, 2014 8. Bostrom MP, Yang X, Kennan M, Sandhu H, Dicarlo E, Lane JM. An unexpected outcome during testing of commercially available demineralized bone graft materials: how safe are the nonallograft components? Spine 2001; 26:1425–1428 9. Wang JC, Kanim LE, Nagakawa IS, Yamane BH, Vinters HV, Dawson EG. Dose-dependent toxicity of a commercially available demineralized bone matrix material. Spine 2001; 26:1429–1435; discussion, 1435–1436 10. Urist MR. Bone: formation by auto induction. Science 1965; 150:893–899 11. U.S. Food and Drug Administration (FDA). InFUSE™ Bone Graft/LT-CAGE™ Lumbar Tapered Fusion Device—P000058. FDA website. www.fda.gov/MedicalDevices/ProductsandMedicalProcedures/DeviceApprovalsandClearances/ Recently-ApprovedDevices/ucm083423.htm. Published 2013. Accessed April 4, 2014 12. Deyo RA, Ching A, Matsen L, et al. Use of bone morphogenetic proteins in spinal fusion surgery for older adults with lumbar stenosis: trends, complications, repeat surgery, and charges. Spine 2012; 37:222–230 13. Papakostidis C, Kontakis G, Bhandari M, Giannoudis PV. Efficacy of autologous iliac crest bone graft and bone morphogenetic proteins for posterolateral fusion of lumbar spine: a meta-analysis of the results. Spine 2008; 33:E680–E692 14. Singh K, Nandyala SV, Marquez-Lara A, Fineberg SJ. Epidemiological trends in the utilization of bone morphogenetic protein in spinal fusions from 2002 to 2011. Spine 2014; 39:491–496 15. Williams BJ, Smith JS, Fu KM, et al.; Scoliosis Research Society Morbidity and Mortality Committee. Does bone morphogenetic protein increase the incidence of perioperative complications in spinal fusion? A comparison of 55,862 cases of spinal fusion with and without bone morphogenetic protein. Spine 2011; 36:1685–1691 16. Singh K, Ahmadinia K, Park DK, et al. Complications of spinal fusion with utilization of bone morphogenetic protein: a systematic review of the literature. Spine 2014; 39:91–101 17. Sethi A, Craig J, Bartol S, et al. Radiographic and CT evaluation of recombinant human bone morphogenetic protein-2–assisted spinal interbody fusion. AJR 2011; 197:[web]W128–W133 18. Ong KL, Villarraga ML, Lau E, Carreon LY, Kurtz SM, Glassman SD. Off-label use of bone morphogenetic proteins in the United States using administrative data. Spine 2010; 35:1794–1800 19. Vaidya R, Sethi A, Bartol S, Jacobson M, Coe C, Craig JG. Complications in the use of rhBMP-2 in
AJR:206, March 2016 593
Downloaded from www.ajronline.org by Gazi Universitesi on 02/28/16 from IP address 194.27.18.18. Copyright ARRS. For personal use only; all rights reserved
Stensby et al. PEEK cages for interbody spinal fusions. J Spinal Disord Tech 2008; 21:557–562 20. Muchow RD, Hsu WK, Anderson PA. Histopathologic inflammatory response induced by recombinant bone morphogenetic protein-2 causing radiculopathy after transforaminal lumbar interbody fusion. Spine J 2010; 10:e1–e6 21. Otsuka E, Notoya M, Hagiwara H. Treatment of myoblastic C2C12 cells with BMP-2 stimulates
vitamin D-induced formation of osteoclasts. Calcif Tissue Int 2003; 73:72–77 22. Fox MG, Goldberg JM, Gaskin CM, et al. MRI of transforaminal lumbar interbody fusion: imaging appearance with and without the use of human recombinant bone morphogenetic protein-2 ( rhBMP-2). Skeletal Radiol 2014; 43:1247–1255 23. Krestan CR, Noske H, Vasilevska V, et al. MDCT versus digital radiography in the evaluation of
bone healing in orthopedic patients. AJR 2006; 186:1754–1760 24. Crandall DG, Revella J, Patterson J, Huish E, Chang M, McLemore R. Transforaminal lumbar interbody fusion with rhBMP-2 in spinal deformity, spondylolisthesis, and degenerative disease. Part 2. BMP dosage-related complications and long-term outcomes in 509 patients. Spine 2013; 38:1137–1145
F O R YO U R I N F O R M AT I O N
This article is available for CME and Self-Assessment (SA-CME) credit that satisfies Part II requirements for maintenance of certification (MOC). To access the examination for this article, follow the prompts associated with the online version of the article.
594
AJR:206, March 2016
