The Spine Journal 15 (2015) 1277–1283
Clinical Study
Reoperation rates after anterior cervical discectomy and fusion versus posterior cervical foraminotomy: a propensity-matched analysis Daniel Lubelski, BAa,b,c, Andrew T. Healy, MDa,c, Michael P. Silverstein, MDa,d, Kalil G. Abdullah, MDe, Nicolas R. Thompson, MSf,g, K. Daniel Riew, MDh, Michael P. Steinmetz, MDi, Edward C. Benzel, MDa,b,c, Thomas E. Mroz, MDa,b,d,* a
Cleveland Clinic Center for Spine Health, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195, USA b Cleveland Clinic Lerner College of Medicine, 9500 Euclid Avenue, Cleveland, OH 44195, USA c Department of Neurosurgery, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195, USA d Department of Orthopaedic Surgery, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195, USA e Department of Neurosurgery, Hospital of the University of Pennsylvania, 3400 Spruce Street, Philadelphia, PA 19104, USA f Department of Quantitative Health Sciences, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195, USA g Neurological Institute Center for Outcomes Research and Evaluation, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195, USA h Washington University Orthopedics, Washington University School of Medicine, 4921 Parkview Pl, St. Louis, MO 63110, USA i Department of Neurological Surgery, Case Western Reserve University School of Medicine, MetroHealth Medical Center, 11100 Euclid Avenue, HAN 5042 Cleveland, OH 44106, USA Received 3 November 2014; revised 3 February 2015; accepted 18 February 2015
Abstract
BACKGROUND CONTEXT: Anterior cervical discectomy and fusion (ACDF) and posterior cervical foraminotomy (PCF) are both used to surgically treat patients with cervical radiculopathy and have been shown to have similar outcomes. Nonetheless, ACDF has become increasingly more commonplace compared with PCF, in part because of a pervasive belief that PCF has a higher incidence of required reoperations. PURPOSE: To determine the reoperation rate at the index level of ACDF versus PCF 2 years postoperatively. STUDY DESIGN: A retrospective case-control. PATIENT SAMPLE: All patients that underwent ACDF and PCF for radiculopathy (excluding myelopathy indications) between January 2005 and December 2011. OUTCOME MEASURES: Revision surgery within 2 years, at the index level, was recorded. METHODS: Propensity score analysis between the ACDF and PCF groups was done, matching for age, gender, race, body mass index, tobacco use, median income and insurance status, primary surgeon, level of surgery, surgery duration, and length of hospital stay. RESULTS: Seven hundred ninety patients met the inclusion/exclusion criteria, including 627 ACDF and 163 PCF. Before propensity matching, the PCF group was found to be significantly older and more likely to be male. After matching, there were no significant differences between groups
FDA device/drug status: Not applicable. Author disclosures: DL: Nothing to disclose. ATH: Nothing to disclose. MPSi: Nothing to disclose. KGA: Nothing to disclose. NRT: Nothing to disclose. KDR: Royalties: Osprey (D), Biomet (F), Medtronic Sofamor Danek (G, Paid directly to employer); Stock Ownership: Spineology (C), Spinal Kinetics (C), Nexgen Spine (B), Vertiflex (C), Benvenue (C), Paradigm Spine (C), Amedica (C); Consulting: AOSpine (B), New England Spine Group (B); Trips/Travel: Dubai Spine (B, reimbursement), SpineMasters (B, reimbursement), Broadwater (B, reimbursement); Board of Directors: CSRS (None), KASS (None), AO Spine International (C), JBJS Spine Highlights (None), Spine (None), Global Spine Journal (None). MPSt: Royalties: Biomet spine; Consulting: Biomet Spine, Intellirod, Stryker Spine, Medtronic. ECB: Royalties: Elsevier Pub (B), Thieme Pub (B); Stock Ownership: Axiomed (NA), Depuy (NA), Orthomems (NA), Turning Point (NA); Consulting: Axiomed; Speaking and/or http://dx.doi.org/10.1016/j.spinee.2015.02.026 1529-9430/Ó 2015 Elsevier Inc. All rights reserved.
Teaching Arrangements: Multiple (Varied); Trips/Travel: Multiple (Varied); Grants: OREF (F, Paid directly to institution), Rawlings (F, Paid directly to institution). TEM: Stock Ownership: PearlDiver Inc (No monies received); Consulting: Globus Medical (B); Speaking and/or Teaching Arrangements: AO Spine (B). The disclosure key can be found on the Table of Contents and at www. TheSpineJournalOnline.com. No grants and technical or corporate support were received in conducting this study or writing this manuscript. There are no relevant conflicts of interest. IRB approval (Study #13-1399) was obtained before start of study. * Corresponding author. Cleveland Clinic Center for Spine Health, Neurological Institute, The Cleveland Clinic, 9500 Euclid Ave., S-80, Cleveland, OH 44195, USA. Tel.: (216) 445-9232; fax: (216) 363-2040. E-mail address: [email protected] (T.E. Mroz)
1278
D. Lubelski et al. / The Spine Journal 15 (2015) 1277–1283
for any baseline characteristics. Reoperation rate at the index level was 4.8% for the ACDF group and 6.4% for the PCF group (p5.7) within 2 years of the initial surgery. Using equivalence testing, based on an a priori null hypothesis that a clinically meaningful difference between the two groups would be $5%, we found that the absolute difference of 1.6% was significantly (p5.01) less than our hypothesized difference. CONCLUSIONS: This study demonstrates that even after accounting for patient demographics, operative characteristics, and primary surgeon, there are no significant differences in 2-year reoperation rates at the index level between ACDF and PCF. The reoperation rates are statistically equivalent. Thus, spine surgeons can operate via the posterior approach without putting patients at increased risk for revision surgery at the index level. Ó 2015 Elsevier Inc. All rights reserved. Keywords:
Anterior cervical discectomy and fusion; Posterior cervical foraminotomy; Reoperation rates; Propensity matching; Foraminal stenosis; Cervical radiculopathy
Introduction Anterior cervical discectomy and fusion (ACDF) and posterior cervical foraminotomy (PCF) are two mainstay approaches to decompress foraminal stenosis in patients with radiculopathy that are refractory to conservative management. These surgical procedures were both initially described in the 1940s to 1950s and modified over time [1–4]. Although both have been shown to be safe and equally effective [4–7], ACDF has become more prevalent in recent years [8,9]. The relative advantages and disadvantages of each approach are well described. With the ACDF, there is a risk of pseudoarthrosis and adjacent segment degeneration, as well as ventral approach-related complications [4,7,10,11]. Posterior cervical foraminotomy does not require stabilization, which allows the surgeon and patient to circumvent the fusion related complications of pseudoarthrosis, and instrumentation failure, and also can preserve the motion-segment mobility. In addition, it avoids complications associated with an anterior approach, such as dysphagia, dysphonia, and injury to the vertebral artery, sympathetic chain, esophagus, and other structures. Exposure through the dorsal cervical musculature, however, has a potential for greater postoperative neck pain and blood loss [4,12–14]. Furthermore, PCF does not allow reconstruction and stabilization of the intervertebral space. It, therefore, may be associated with a greater incidence of revision surgery than ACDF [7]. In a retrospective cohort study, Wang et al. [7] demonstrated that relative to historical controls, PCF has a similar reoperation rate (5%). These findings, however, were limited in that they compared rates from two disparate populations. No randomized controlled trials (RCTs) or matched-cohort retrospective studies have compared the reoperation rates of ACDF versus PCF. Accordingly, we sought to use a propensitymatched analysis (the gold standard for retrospective studies, second only to prospective randomization) to compare the same-level reoperation rate of these two surgical approaches, while controlling for possible confounding variables such as demographic variables, operative characteristics, level of
surgery, and operating surgeon. We hypothesized that there would be no statistically significant difference in reoperation rates between ACDF and PCF and that the revision rate would be within 5% of each other.
Methods Study sample A retrospective study of all patients who underwent ACDF or PCF at C2–C7 for cervical radiculopathy at a single tertiary-care institution between 2005 and 2011 was performed. The electronic medical records were queried to retrieve patient data that met our criteria. Patients were excluded if they were younger than 18 years, if their surgical indication was for myelopathy, or if they had undergone a previous cervical spine surgery. Outcome was defined as whether revision surgery was performed within 2 years at the index level of surgery. The patient information collected included age, gender, race, body mass index, tobacco use, median income, insurance status, operating surgeon, year of surgery, surgery duration, level of surgery, length of hospital stay (LOS), and primary and revision surgery approach. Propensity matching was based on the aforementioned variables and is described below. Statistical analysis Summary statistics (eg, means, standard deviations, counts, percentages) were computed for patients undergoing ACDF and PCF. We knew, a priori, that the two treatment groups would likely differ on certain variables (eg, primary Surgeon and gender). To account for these group differences, we used propensity score matching. To measure the covariate balance in the two groups, we computed the standardized difference [15] for each variable, both before and after propensity score matching. We considered standardized differences less than 10.0 in absolute value to be balanced. In addition to examining standardized differences for each continuous variable, we compared density
D. Lubelski et al. / The Spine Journal 15 (2015) 1277–1283
1279
plots after propensity matching to ensure similar overall distributions in each treatment group. The propensity score model To obtain propensity scores, we created a mixed-effects logistic regression model where the response variable was defined by whether or not the patient had PCF. All variables listed in Table 1 were included in the propensity model with continuous variables treated as linear, and a random intercept was included for each patient’s primary surgeon. Propensity score matches were made using the predicted values on the logit scale from the final model, using the optmatch package [16] in R 3.0.1 (R core team, 2013, Vienna, Austria). Matches of similar propensity scores were made where the ratio of ACDF to PCF patients was allowed to vary from 3:1 to 1:2. Propensity scores were considered a match as long as they were within one unit of each other on the logit scale. Comparing 2-year revision rates Within the propensity score-matched patients, we computed the proportion of patients having revision surgery at the index surgical level within 2 years of their initial surgery date for both groups. To establish that revision rates for the two groups were equivalent, we assumed for the null hypothesis that the absolute difference in 2-year revision rates for the two groups was greater than or equal to 5%. The alternative hypothesis was that the absolute difference in 2-year revision rates was less than 5%. Because of small sample size, to calculate the two-tailed p value, we used the method by Gart [17] as described by Dunnett and Gent [18]. We set the significance level at a50.05. Missing data Because of the nature of our electronic health record, we anticipated varying amounts of missing data on the variables of interest. For the purpose of creating the propensity score model, we used multiple imputation [19] to create and analyze 10 imputed data sets. Incomplete variables were imputed under fully conditional specification [20]. Calculations were done in R 3.0.1 using the default settings of the mice 2.13 package [21]. Model parameters were estimated with mixed-effects logistic regression applied to each imputed data set separately. Predicted values on the linear scale were averaged over the 10 analyzed data sets to obtain the propensity scores.
Results Seven hundred ninety patients met our inclusion criteria, consisting of 627 ACDF and 163 PCF. Summary statistics for the two groups are provided for before (Table 1) and after (Table 2) propensity score matching along with standardized differences. Before propensity score matching, variables that were imbalanced between the groups based on
Context Posterior cervical foraminotomy (PCF) is considered to have a higher associated risk of reoperation, in part due to disease progression and limitation associated with the extent to which a posterior decompression can be performed without need for fusion. The authors sought to retrospectively compare cohorts receiving PCF or ACDF to evaluate the need for reoperation using a propensity matched analysis. Contribution This series included 790 patients, the overwhelming majority of whom received ACDF surgery. The propensity matched analysis included 188 ACDF patients and 140 PCF patients. Reoperation rates between the two groups showed no significant difference. The authors maintain that surgeons should feel comfortable using PCF on indicated patients without a concern of increased need for reoperation. Implications This paper is a retrospective review of patients treated at a single center and thus suffers from a component of selection, indication and possibly even information bias. While a propensity matched analysis was performed, this likely cannot address issues related to these possibilities for confounding. In particular, patients who may have had complications related to their PCF and could have benefited from a revision procedure, yet who were not offered surgery or elected not to undergo surgery, would not be seen as having the outcome of interest in this work. Given the limitations associated with single center design and the potential for confounding, the results of this work should rightly be seen as preliminary and further testing in prospective analysis is likely required. —The Editors
standardized differences greater than or equal to 10.0 in absolute value were age, gender, median income, insurance status, LOS, total levels, and levels C2, C3, C4, C5, and C7. For ACDF and PCF, respectively, average age was 51.2 and 53.6 years, gender was 53% and 69.3% male, and LOS was 2.3 and 2.9 days. Of this initial cohort, 26/ 627 (4.1%) ACDF patients and 10/163 (6.1%) PCF patients required reoperation at the index level(s) (p5.4). After propensity score matching, there were 188 ACDF patients matched to 140 PCF patients and all variables in Table 1 (except total levels and primary surgeon) had standardized differences less than 10.0 in absolute value. After propensity matching, none of the differences were statistically significant. The mean number of total levels operated
1280
D. Lubelski et al. / The Spine Journal 15 (2015) 1277–1283
Table 1 Summary statistics for ACDF and PCF patients before propensity score matching Characteristic
ACDF
PCF
Standard difference p Value
N 627 163 Age (y) 51.2 (10.7) 53.6 (12.8) 20.8 Gender: male, N (%) 332 (53.0) 113 (69.3) 34.1 BMI 29.1 (6.2) 28.7 (5.2) 7.3 Race White: N (%) 551 (87.9) 144 (88.3) 1.4 Black: N (%) 45 (7.2) 12 (7.4) 0.7 Other: N (%) 31 (4.9) 7 (4.3) 3.1 Tobacco use Never: N (%) 240 (38.3) 69 (42.3) 2.9 Former: N (%) 168 (30.2) 51 (32.9) 5.9 Current: N (%) 149 (26.8) 35 (22.6) 9.7 Median income 53.7 (17.8) 55.9 (18.0) 12.7 ($10,000) Insurance Private/other: N (%) 365 (69.1) 94 (63.5) 11.9 Medicaid: N (%) 27 (5.1) 1 (0.7) 26.7 Medicare: N (%) 136 (25.8) 53 (35.8) 21.9 Surgery duration (min) 211.3 (78.7) 216.8 (71.4) 7.3 Length of stay (d) 2.3 (4.2) 2.9 (5.9) 11.6 Level of surgery C2: N (%) 1 (0.2) 4 (2.5) 20.3 C3: N (%) 58 (9.3) 33 (20.2) 31.4 C4: N (%) 206 (32.9) 63 (38.7) 12.1 C5: N (%) 471 (75.1) 115 (70.6) 10.3 C6: N (%) 546 (87.1) 142 (87.1) 0.1 C7: N (%) 311 (49.6) 91 (55.8) 12.5 Total levels 2.5 (0.7) 2.7 (0.9) 25.9
.0230 .0002 .6527 .9404
.5573
.0865
.0061*
.2168 .3644 .0070* .0002 .1942 .2772 1.0000 .1839 .0065
ACDF, anterior cervical discectomy and fusion; PCF, posterior cervical foraminotomy; BMI, body mass index. Note: All values are presented as means (standard deviation) or number (%) for continuous and categorical variables, respectively. p Values were calculated using Mann-Whitney U tests for continuous variables and chisquare tests or Fisher exact tests for categorical variables. * p Value was calculated from Fisher exact test.
was 2.57 and 2.67 for ACDF and PCF, respectively (p5.3). Although balance among primary surgeons improved after propensity matching, there still was some imbalance; however, a Fisher exact test indicated that the differences in surgery rates among primary surgeons were not significantly different after propensity score matching (p5.08). Of the 188 ACDF patients, 9 (4.8%) had revision within 2 years of their initial surgery, whereas 9/140 (6.4%) PCF patients had revision within 2 years of their initial surgery. The absolute difference in revision rates for the two groups (6.4%4.8%51.6%) was significantly less than our hypothesized difference of 5% (p5.0131). Furthermore, a chi-square test indicated these percentages were not significantly different (p5.7).
Discussion Anterior cervical discectomy and fusion and PCF have been described as having similar postoperative outcomes
Table 2 Summary statistics for ACDF and PCF patients after propensity score matching Characteristic
ACDF
PCF
Standard difference p Value
N 188 140 Age (y) 53.3 (11.6) 52.9 (12.2) 2.3 Gender: male, N (%) 128 (68.1) 97 (69.3) 3.7 BMI 28.9 (5.4) 28.5 (5.2) 6.5 Race White: N (%) 166 (88.3) 123 (87.9) 1.4 Black: N (%) 13 (6.9) 10 (7.1) 0.9 Other: N (%) 9 (4.8) 7 (5.0) 1.0 Tobacco use Never: N (%) 70 (37.2) 59 (42.1) 7.4 Former: N (%) 62 (36.0) 45 (33.8) 3.4 Current: N (%) 40 (23.3) 29 (21.8) 4.8 Median income 56.7 (18.3) 55.9 (18.2) 4.5 ($10,000) Insurance Private/other: N (%) 111 (66.9) 82 (65.6) 2.7 Medicaid: N (%) 1 (0.6) 1 (0.8) 2.4 Medicare: N (%) 54 (32.5) 42 (33.6) 2.3 Surgery duration 218.1 (73.7) 216.9 (74.2) 1.3 (min) Length of stay (d) 2.5 (4.6) 2.6 (5.6) 3.1 Level of surgery C2: N (%) 1 (0.5) 1 (0.7) 2.3 C3: N (%) 25 (13.3) 23 (16.4) 8.8 C4: N (%) 60 (31.9) 51 (36.4) 9.5 C5: N (%) 132 (70.2) 98 (70.0) 0.5 C6: N (%) 165 (87.8) 122 (87.1) 1.9 C7: N (%) 101 (53.7) 79 (56.4) 6.5 Total levels 2.6 (0.7) 2.7 (0.8) 12.9
.7443 .9112 .4210 .9924
.8137
.5889
.9590*
.7739 .5550 1.0000* .5251 .4614 1.0000 1.0000 .7078 .2533
ACDF, anterior cervical discectomy and fusion; PCF, posterior cervical foraminotomy; BMI, body mass index. Note: All values are presented as means (standard deviation) or number (%) for continuous and categorical variables, respectively. p Values were calculated using Mann-Whitney U tests for continuous variables and chisquare tests or Fisher exact tests for categorical variables. * p Value calculated from Fisher exact test.
for pain, complications, and quality of life. Specifically, in various retrospective studies, the reported success rates have ranged from 87% to 94% for ACDF and 90% to 97% for PCF [7,22–26]. Reoperation rates have been described as similar for ACDF and PCF, with reports describing 6% to 9% for ACDF [25,27–36] and 5% for PCF [7]; however, none of these studies performed a direct comparison between the two. The RCTs [4,6] comparing these two approaches have similarly reported no differences in pain or complication outcomes. Several small retrospective cost analyses have been performed that have found that PCF costs less than ACDF [37–39] and that PCF is significantly more costeffective [39]. Nonetheless, in recent decades, the prevalence of ACDF has increased substantially faster than PCF [8,9]. Herkowitz et al. [6] randomized 33 patients with lateral disc herniations to either ACDF or laminotomy/foraminotomy and found no significant differences between the groups, showing subjectively ‘‘excellent or good’’ results
D. Lubelski et al. / The Spine Journal 15 (2015) 1277–1283
in 94% and 75% of ACDF and PCF patients, respectively. Wirth et al. [40] randomized 72 patients with cervical disc herniation and radiculopathy to PCF, ACDF, or anterior cervical discectomy without fusion between 1984 and 1991. They found that all three procedures had similar (no significant differences) LOS, perioperative symptoms, and shortterm (~2 months) and long-term (~5 years) postoperative improvement in pain and satisfaction from surgery. The authors also reviewed recurrent symptoms necessitating additional surgery and found no significant differences (reoperation rate was 27% for PCF and 28% for ACDF). Most recently, Ruetten et al. [4], in Germany, performed an RCT with 175 patients with lateral disc herniations and radiculitis or neurologic deficit to endoscopic PCF versus conventional microsurgical ACDF. They found no differences in complications, clinical outcomes for neck and arm visual analog scale, Hilibrand criteria outcomes, or radicular or neurologic outcomes. The reoperation rates were not significantly different between the groups, with 4.7% (4 of 86) in the ACDF group and 6.7% (6 of 89) in the PCF group. Although the previous studies represent Class I evidence, they are limited by a relatively small sample size. For this reason, the studies largely presented trends in the data and were only able to draw limited conclusions. Nonetheless, the RCTs found that ACDF and PCF have similar outcomes in terms of pain, complications, and reoperation rate. The largest of the RCTs, by Ruetten et al. [4], found very similar reoperation rates at index levels to the ones we found in the present study (ie, 4.7% in their cohort for ACDF vs. 4.7% in our study; 6.7% in their cohort for PCF vs. 6.4% in our study). Although retrospective, the present study had a substantially larger sample size that was propensity matched. Furthermore, through the use of an equivalence analysis, we were able to demonstrate that not only was there no significant difference between reoperation rates for ACDF and PCF, but also support our hypothesis that these rates were statistically equivalent. When interpreting these results it is important to understand that the data support the equivalency between the reoperation rates of ACDF and PCF, it does not indicate that one is superior to the other. As previously described, the outcomes of these two approaches, when the indication is equivocal, are often successful irrespective of the approach. There are advantages and disadvantages to each. As there has been a dramatic increase in the number of ACDF surgeries performed annually in the United States relative to PCF [9], our goal was to dispel the notion that one is superior to the other. It is important to understand that the benefits of PCF in avoiding fusion and fusion-related complications do not come at the expense of greater destabilization or predisposition to revision surgery. Rather, our results demonstrate that indexlevel reoperation rate at 2-year follow-up is equivalent between the two surgical approaches. It should be noted, however, that reoperation for adjacent segment disease is substantially more common after a fusion procedure,
1281
occurring in approximately 12% of patients undergoing ACDF [41] and only 2% to 3% after PCF [42]. The present study has several limitations that should be considered in the interpretation of our findings. It was a retrospective analysis, in which patient data were collected over a period of 7 years. The data in this study were objective, requiring minimal interpretation, but nonetheless, the nature of a retrospective analysis is that it is prone to data collection or interpretation bias. Two-year reoperation rates were determined, which we felt was sufficient time to determine index level failures, but studies with longer-term follow-up would reveal additional information regarding true costs and long-term complications. The retrospective design also meant that if some patients underwent reoperation at an outside institution, it would not be captured in our review. There is no reason to suspect, however, that there would be a difference in missed (outside hospital) reoperations between patients that underwent an index ACDF versus a PCF. Additionally, while propensity matching, the gold standard in retrospective studies, was used, we were only able to match based on the collected variables and there may have been differences in uncollected variables. Nonetheless, this is the largest study to date comparing reoperation rates for ACDF versus PCF. Furthermore, the propensity matching, ensuring similar demographic and operative characteristics between the groups, and the equivalence analysis strengthen the conclusions herein. Another limitation is that there may have been a selection bias for the types of patients that were treated with an ACDF versus a PCF. It is possible that those with an arthritic facet joint, severe neck pain, or a degenerative listhesis were preferentially treated with the fusion procedure. Additionally, anatomical considerations can influence selection; soft disc herniations that are exclusively localized laterally may be more easily addressed with a PCF, whereas a central/paracentral calcified disc can be more accessible with an ACDF. Therefore, our data should be interpreted as being valid for those patients who meet the indications for both procedures only and not those who would otherwise be contraindicated for the PCF. Conclusion After accounting for patient demographics and operative characteristics, ACDF and PCF have statistically equivalent 2-year reoperation rates at the index level, both of which are relatively low. Accordingly, spine surgeons can operate for cervical radiculopathy via a posterior approach in properly indicated patients without putting them at a comparatively increased risk for revision surgery. References [1] Abumi K, Panjabi MM, Kramer KM, Duranceau J, Oxland T, Crisco JJ. Biomechanical evaluation of lumbar spinal stability after graded facetectomies. Spine 1990;15:1142–7.
1282
D. Lubelski et al. / The Spine Journal 15 (2015) 1277–1283
[2] Brayda-Bruno M, Cinnella P. Posterior endoscopic discectomy (and other procedures). Eur Spine J 2000;9(1 Suppl):S24–9. [3] Mixter WJ, BARR JS. Rupture of the intervertebral disc with involvement of the spinal canal. N Engl J Med 1934;211:210–5. [4] Ruetten S, Komp M, Merk H, Godolias G. Full-endoscopic cervical posterior foraminotomy for the operation of lateral disc herniations using 5.9-mm endoscopes: a prospective, randomized, controlled study. Spine 2008;33:940–8. [5] Henderson CM, Hennessy RG, Shuey HM, Shackelford EG. Posterior-lateral foraminotomy as an exclusive operative technique for cervical radiculopathy: a review of 846 consecutively operated cases. Neurosurgery 1983;13:504–12. [6] Herkowitz HN, Kurz LT, Overholt DP. Surgical management of cervical soft disc herniation. A comparison between the anterior and posterior approach. Spine 1990;15:1026–30. [7] Wang TY, Lubelski D, Abdullah KG, Steinmetz MP, Benzel EC, Mroz TE, et al. Rates of anterior cervical discectomy and fusion after initial posterior cervical foraminotomy. Spine J 2013 Jul 16. [Epub ahead of print]. [8] Patil PG, Turner DA, Pietrobon R. National trends in surgical procedures for degenerative cervical spine disease: 1990-2000. Neurosurgery 2005;57:753–8; discussion 753–758. [9] Wang MC, Kreuter W, Wolfla CE, Maiman DJ, Deyo RA. Trends and variations in cervical spine surgery in the United States: Medicare beneficiaries, 1992 to 2005. Spine 2009;34:955–61; discussion 962–963. [10] Gao Y, Liu M, Li T, Huang F, Tang T, Xiang Z. A meta-analysis comparing the results of cervical disc arthroplasty with anterior cervical discectomy and fusion (ACDF) for the treatment of symptomatic cervical disc disease. J Bone Joint Surg Am 2013;95:555–61. [11] Fountas KN, Kapsalaki EZ, Nikolakakos LG, Smisson HF, Johnston KW, Grigorian AA, et al. Anterior cervical discectomy and fusion associated complications. Spine 2007;32:2310–7. [12] Skovrlj B, Gologorsky Y, Haque R, Fessler RG, Qureshi SA. Complications, outcomes, and need for fusion after minimally invasive posterior cervical foraminotomy and microdiscectomy. Spine J 2014;4: 2405–11. [13] Albert TJ, Vacarro A. Postlaminectomy kyphosis. Spine 1998;23: 2738–45. [14] Mikawa Y, Shikata J, Yamamuro T. Spinal deformity and instability after multilevel cervical laminectomy. Spine 1987;12:6–11. [15] Rosenbaum PR, Rubin DB. Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat 1985;39:33. [16] Hansen BB, Klopfer SO. Optimal full matching and related designs via network flow. J Comput Graph Stat 2006;609–27. [17] Gart JJ. The comparison of proportions: a review of significance tests, confidence intervals and adjustments for stratification. Rev Inst Int Stat 1971;39:148. [18] Dunnett CW, Gent M. Significance testing to establish equivalence between treatments, with special reference to data in the form of 2X2 tables. Biometrics 1977;33:593–602. [19] Rubin DB. Multiple imputation for nonresponse in surveys. Wiley, 2004. [20] Van Buuren S, Brand JP, Groothuis-Oudshoorn CG, Rubin DB. Fully conditional specification in multivariate imputation. J Stat Comput Sim 2006;76:1049–64. [21] Van Burren S, Groothuis-Oudshoorn C. Multivariate imputation by chained equations in R. J Stat Software 2011;45:1–6. [22] Jacobs W, Willems PC, Kruyt M, van Limbeek J, Anderson PG, Pavlov P, et al. Systematic review of anterior interbody fusion techniques for single- and double-level cervical degenerative disc disease. Spine 2011;36:E950–60. [23] Jagannathan J, Sherman JH, Szabo T, Shaffrey CI, Jane JA. The posterior cervical foraminotomy in the treatment of cervical disc/osteophyte disease: a single-surgeon experience with a minimum of 5
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
years’ clinical and radiographic follow-up. J Neurosurg Spine 2009;10:347–56. Adamson TE. Microendoscopic posterior cervical laminoforaminotomy for unilateral radiculopathy: results of a new technique in 100 cases. J Neurosurg 2001;95:51–7. Marotta N, Landi A, Tarantino R, Mancarella C, Ruggeri A, Delfini R. Five-year outcome of stand-alone fusion using carbon cages in cervical disc arthrosis. Eur Spine J 2011;20(1 Suppl): S8–12. Korinth MC, Kr€uger A, Oertel MF, Gilsbach JM. Posterior foraminotomy or anterior discectomy with polymethyl methacrylate interbody stabilization for cervical soft disc disease: results in 292 patients with monoradiculopathy. Spine 2006;31:1207–14; discussion 1215–1216. Bhadra AK, Raman AS, Casey AT, Crawford RJ. Single-level cervical radiculopathy: clinical outcome and cost-effectiveness of four techniques of anterior cervical discectomy and fusion and disc arthroplasty. Eur Spine J 2009;18:232–7. Song KJ, Taghavi CE, Lee KB, Song JH, Eun JP. The efficacy of plate construct augmentation versus cage alone in anterior cervical fusion. Spine 2009;34:2886–92. Burkus JK, Haid RW, Traynelis VC, Mummaneni PV. Long-term clinical and radiographic outcomes of cervical disc replacement with the Prestige disc: results from a prospective randomized controlled clinical trial. J Neurosurg Spine 2010;13:308–18. Woodworth RS, Molinari WJ, Brandenstein D, Gruhn W, Molinari RW. Anterior cervical discectomy and fusion with structural allograft and plates for the treatment of unstable posterior cervical spine injuries. J Neurosurg Spine 2009;10:93–101. Jensen WK, Moore TA, Tribus CB, Anderson PA, Zdeblick TA. Use of patella allograft for anterior cervical diskectomy and fusion. J Spinal Disord Tech 2009;22:392–8. Ruetten S, Komp M, Merk H, Godolias G. Recurrent lumbar disc herniation after conventional discectomy: a prospective, randomized study comparing full-endoscopic interlaminar and transforaminal versus microsurgical revision. J Spinal Disord Tech 2009;22:122–9. Mummaneni PV, Robinson JC, Haid RW. Cervical arthroplasty with the PRESTIGE LP cervical disc. Neurosurgery 2007;60:310–4; discussion 314–315. Anderson PA, Sasso RC, Riew KD. Comparison of adverse events between the Bryan artificial cervical disc and anterior cervical arthrodesis. Spine 2008;33:1305–12. Coric D, Cassis J, Carew JD, Boltes MO. Prospective study of cervical arthroplasty in 98 patients involved in 1 of 3 separate investigational device exemption studies from a single investigational site with a minimum 2-year follow-up. Clinical article. J Neurosurg Spine 2010;13:715–21. Coric D, Numley PD, Guyer RD, Musante D, Carmody CN, Gordon CR, et al. Prospective, randomized, multicenter study of cervical arthroplasty: 269 patients from the KineflexjC artificial disc investigational device exemption study with a minimum 2-year follow-up: clinical article. J Neurosurg Spine 2011;15:348–58. Mansfield HE, Canar WJ, Gerard CS, O’Toole JE. Single-level anterior cervical discectomy and fusion versus minimally invasive posterior cervical foraminotomy for patients with cervical radiculopathy: a cost analysis. Neurosurg Focus 2014;37:E9. Tumialan LM, Ponton RP, Gluf WM. Management of unilateral cervical radiculopathy in the military: the cost effectiveness of posterior cervical foraminotomy compared with anterior cervical discectomy and fusion. Neurosurg Focus 2010;28:E17. Alvin MD, Lubelski D, Abdullah KG, Whitmore RG, Benzel EC, Mroz TE. Cost-utility analysis of anterior cervical discectomy and fusion with plating (ACDFP) versus posterior cervical foraminotomy (PCF) for patients with single-level cervical radiculopathy at 1-year follow-up. J Spinal Disord Tech 2014 Mar 27. [Epub ahead of print].
D. Lubelski et al. / The Spine Journal 15 (2015) 1277–1283 [40] Wirth FP, Dowd GC, Sanders HF, Wirth C. Cervical discectomy. A prospective analysis of three operative techniques. Surg Neurol 2000;53:340–6; discussion 346–348. [41] Xu R, Bydon M, Macki M, De la Garza-Ramos R, Sciubba DM, Wolinsky JP, et al. Adjacent segment disease after anterior cervical
1283
discectomy and fusion: clinical outcomes after first repeat surgery versus second repeat surgery. Spine 2014;39:120–6. [42] Clarke MJ, Ecker RD, Krauss WE, McClelland RL, Dekutoski MB. Same-segment and adjacent-segment disease following posterior cervical foraminotomy. J Neurosurg Spine 2007;6:5–9.
Clinical Study
Reoperation rates after anterior cervical discectomy and fusion versus posterior cervical foraminotomy: a propensity-matched analysis Daniel Lubelski, BAa,b,c, Andrew T. Healy, MDa,c, Michael P. Silverstein, MDa,d, Kalil G. Abdullah, MDe, Nicolas R. Thompson, MSf,g, K. Daniel Riew, MDh, Michael P. Steinmetz, MDi, Edward C. Benzel, MDa,b,c, Thomas E. Mroz, MDa,b,d,* a
Cleveland Clinic Center for Spine Health, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195, USA b Cleveland Clinic Lerner College of Medicine, 9500 Euclid Avenue, Cleveland, OH 44195, USA c Department of Neurosurgery, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195, USA d Department of Orthopaedic Surgery, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195, USA e Department of Neurosurgery, Hospital of the University of Pennsylvania, 3400 Spruce Street, Philadelphia, PA 19104, USA f Department of Quantitative Health Sciences, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195, USA g Neurological Institute Center for Outcomes Research and Evaluation, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, OH 44195, USA h Washington University Orthopedics, Washington University School of Medicine, 4921 Parkview Pl, St. Louis, MO 63110, USA i Department of Neurological Surgery, Case Western Reserve University School of Medicine, MetroHealth Medical Center, 11100 Euclid Avenue, HAN 5042 Cleveland, OH 44106, USA Received 3 November 2014; revised 3 February 2015; accepted 18 February 2015
Abstract
BACKGROUND CONTEXT: Anterior cervical discectomy and fusion (ACDF) and posterior cervical foraminotomy (PCF) are both used to surgically treat patients with cervical radiculopathy and have been shown to have similar outcomes. Nonetheless, ACDF has become increasingly more commonplace compared with PCF, in part because of a pervasive belief that PCF has a higher incidence of required reoperations. PURPOSE: To determine the reoperation rate at the index level of ACDF versus PCF 2 years postoperatively. STUDY DESIGN: A retrospective case-control. PATIENT SAMPLE: All patients that underwent ACDF and PCF for radiculopathy (excluding myelopathy indications) between January 2005 and December 2011. OUTCOME MEASURES: Revision surgery within 2 years, at the index level, was recorded. METHODS: Propensity score analysis between the ACDF and PCF groups was done, matching for age, gender, race, body mass index, tobacco use, median income and insurance status, primary surgeon, level of surgery, surgery duration, and length of hospital stay. RESULTS: Seven hundred ninety patients met the inclusion/exclusion criteria, including 627 ACDF and 163 PCF. Before propensity matching, the PCF group was found to be significantly older and more likely to be male. After matching, there were no significant differences between groups
FDA device/drug status: Not applicable. Author disclosures: DL: Nothing to disclose. ATH: Nothing to disclose. MPSi: Nothing to disclose. KGA: Nothing to disclose. NRT: Nothing to disclose. KDR: Royalties: Osprey (D), Biomet (F), Medtronic Sofamor Danek (G, Paid directly to employer); Stock Ownership: Spineology (C), Spinal Kinetics (C), Nexgen Spine (B), Vertiflex (C), Benvenue (C), Paradigm Spine (C), Amedica (C); Consulting: AOSpine (B), New England Spine Group (B); Trips/Travel: Dubai Spine (B, reimbursement), SpineMasters (B, reimbursement), Broadwater (B, reimbursement); Board of Directors: CSRS (None), KASS (None), AO Spine International (C), JBJS Spine Highlights (None), Spine (None), Global Spine Journal (None). MPSt: Royalties: Biomet spine; Consulting: Biomet Spine, Intellirod, Stryker Spine, Medtronic. ECB: Royalties: Elsevier Pub (B), Thieme Pub (B); Stock Ownership: Axiomed (NA), Depuy (NA), Orthomems (NA), Turning Point (NA); Consulting: Axiomed; Speaking and/or http://dx.doi.org/10.1016/j.spinee.2015.02.026 1529-9430/Ó 2015 Elsevier Inc. All rights reserved.
Teaching Arrangements: Multiple (Varied); Trips/Travel: Multiple (Varied); Grants: OREF (F, Paid directly to institution), Rawlings (F, Paid directly to institution). TEM: Stock Ownership: PearlDiver Inc (No monies received); Consulting: Globus Medical (B); Speaking and/or Teaching Arrangements: AO Spine (B). The disclosure key can be found on the Table of Contents and at www. TheSpineJournalOnline.com. No grants and technical or corporate support were received in conducting this study or writing this manuscript. There are no relevant conflicts of interest. IRB approval (Study #13-1399) was obtained before start of study. * Corresponding author. Cleveland Clinic Center for Spine Health, Neurological Institute, The Cleveland Clinic, 9500 Euclid Ave., S-80, Cleveland, OH 44195, USA. Tel.: (216) 445-9232; fax: (216) 363-2040. E-mail address: [email protected] (T.E. Mroz)
1278
D. Lubelski et al. / The Spine Journal 15 (2015) 1277–1283
for any baseline characteristics. Reoperation rate at the index level was 4.8% for the ACDF group and 6.4% for the PCF group (p5.7) within 2 years of the initial surgery. Using equivalence testing, based on an a priori null hypothesis that a clinically meaningful difference between the two groups would be $5%, we found that the absolute difference of 1.6% was significantly (p5.01) less than our hypothesized difference. CONCLUSIONS: This study demonstrates that even after accounting for patient demographics, operative characteristics, and primary surgeon, there are no significant differences in 2-year reoperation rates at the index level between ACDF and PCF. The reoperation rates are statistically equivalent. Thus, spine surgeons can operate via the posterior approach without putting patients at increased risk for revision surgery at the index level. Ó 2015 Elsevier Inc. All rights reserved. Keywords:
Anterior cervical discectomy and fusion; Posterior cervical foraminotomy; Reoperation rates; Propensity matching; Foraminal stenosis; Cervical radiculopathy
Introduction Anterior cervical discectomy and fusion (ACDF) and posterior cervical foraminotomy (PCF) are two mainstay approaches to decompress foraminal stenosis in patients with radiculopathy that are refractory to conservative management. These surgical procedures were both initially described in the 1940s to 1950s and modified over time [1–4]. Although both have been shown to be safe and equally effective [4–7], ACDF has become more prevalent in recent years [8,9]. The relative advantages and disadvantages of each approach are well described. With the ACDF, there is a risk of pseudoarthrosis and adjacent segment degeneration, as well as ventral approach-related complications [4,7,10,11]. Posterior cervical foraminotomy does not require stabilization, which allows the surgeon and patient to circumvent the fusion related complications of pseudoarthrosis, and instrumentation failure, and also can preserve the motion-segment mobility. In addition, it avoids complications associated with an anterior approach, such as dysphagia, dysphonia, and injury to the vertebral artery, sympathetic chain, esophagus, and other structures. Exposure through the dorsal cervical musculature, however, has a potential for greater postoperative neck pain and blood loss [4,12–14]. Furthermore, PCF does not allow reconstruction and stabilization of the intervertebral space. It, therefore, may be associated with a greater incidence of revision surgery than ACDF [7]. In a retrospective cohort study, Wang et al. [7] demonstrated that relative to historical controls, PCF has a similar reoperation rate (5%). These findings, however, were limited in that they compared rates from two disparate populations. No randomized controlled trials (RCTs) or matched-cohort retrospective studies have compared the reoperation rates of ACDF versus PCF. Accordingly, we sought to use a propensitymatched analysis (the gold standard for retrospective studies, second only to prospective randomization) to compare the same-level reoperation rate of these two surgical approaches, while controlling for possible confounding variables such as demographic variables, operative characteristics, level of
surgery, and operating surgeon. We hypothesized that there would be no statistically significant difference in reoperation rates between ACDF and PCF and that the revision rate would be within 5% of each other.
Methods Study sample A retrospective study of all patients who underwent ACDF or PCF at C2–C7 for cervical radiculopathy at a single tertiary-care institution between 2005 and 2011 was performed. The electronic medical records were queried to retrieve patient data that met our criteria. Patients were excluded if they were younger than 18 years, if their surgical indication was for myelopathy, or if they had undergone a previous cervical spine surgery. Outcome was defined as whether revision surgery was performed within 2 years at the index level of surgery. The patient information collected included age, gender, race, body mass index, tobacco use, median income, insurance status, operating surgeon, year of surgery, surgery duration, level of surgery, length of hospital stay (LOS), and primary and revision surgery approach. Propensity matching was based on the aforementioned variables and is described below. Statistical analysis Summary statistics (eg, means, standard deviations, counts, percentages) were computed for patients undergoing ACDF and PCF. We knew, a priori, that the two treatment groups would likely differ on certain variables (eg, primary Surgeon and gender). To account for these group differences, we used propensity score matching. To measure the covariate balance in the two groups, we computed the standardized difference [15] for each variable, both before and after propensity score matching. We considered standardized differences less than 10.0 in absolute value to be balanced. In addition to examining standardized differences for each continuous variable, we compared density
D. Lubelski et al. / The Spine Journal 15 (2015) 1277–1283
1279
plots after propensity matching to ensure similar overall distributions in each treatment group. The propensity score model To obtain propensity scores, we created a mixed-effects logistic regression model where the response variable was defined by whether or not the patient had PCF. All variables listed in Table 1 were included in the propensity model with continuous variables treated as linear, and a random intercept was included for each patient’s primary surgeon. Propensity score matches were made using the predicted values on the logit scale from the final model, using the optmatch package [16] in R 3.0.1 (R core team, 2013, Vienna, Austria). Matches of similar propensity scores were made where the ratio of ACDF to PCF patients was allowed to vary from 3:1 to 1:2. Propensity scores were considered a match as long as they were within one unit of each other on the logit scale. Comparing 2-year revision rates Within the propensity score-matched patients, we computed the proportion of patients having revision surgery at the index surgical level within 2 years of their initial surgery date for both groups. To establish that revision rates for the two groups were equivalent, we assumed for the null hypothesis that the absolute difference in 2-year revision rates for the two groups was greater than or equal to 5%. The alternative hypothesis was that the absolute difference in 2-year revision rates was less than 5%. Because of small sample size, to calculate the two-tailed p value, we used the method by Gart [17] as described by Dunnett and Gent [18]. We set the significance level at a50.05. Missing data Because of the nature of our electronic health record, we anticipated varying amounts of missing data on the variables of interest. For the purpose of creating the propensity score model, we used multiple imputation [19] to create and analyze 10 imputed data sets. Incomplete variables were imputed under fully conditional specification [20]. Calculations were done in R 3.0.1 using the default settings of the mice 2.13 package [21]. Model parameters were estimated with mixed-effects logistic regression applied to each imputed data set separately. Predicted values on the linear scale were averaged over the 10 analyzed data sets to obtain the propensity scores.
Results Seven hundred ninety patients met our inclusion criteria, consisting of 627 ACDF and 163 PCF. Summary statistics for the two groups are provided for before (Table 1) and after (Table 2) propensity score matching along with standardized differences. Before propensity score matching, variables that were imbalanced between the groups based on
Context Posterior cervical foraminotomy (PCF) is considered to have a higher associated risk of reoperation, in part due to disease progression and limitation associated with the extent to which a posterior decompression can be performed without need for fusion. The authors sought to retrospectively compare cohorts receiving PCF or ACDF to evaluate the need for reoperation using a propensity matched analysis. Contribution This series included 790 patients, the overwhelming majority of whom received ACDF surgery. The propensity matched analysis included 188 ACDF patients and 140 PCF patients. Reoperation rates between the two groups showed no significant difference. The authors maintain that surgeons should feel comfortable using PCF on indicated patients without a concern of increased need for reoperation. Implications This paper is a retrospective review of patients treated at a single center and thus suffers from a component of selection, indication and possibly even information bias. While a propensity matched analysis was performed, this likely cannot address issues related to these possibilities for confounding. In particular, patients who may have had complications related to their PCF and could have benefited from a revision procedure, yet who were not offered surgery or elected not to undergo surgery, would not be seen as having the outcome of interest in this work. Given the limitations associated with single center design and the potential for confounding, the results of this work should rightly be seen as preliminary and further testing in prospective analysis is likely required. —The Editors
standardized differences greater than or equal to 10.0 in absolute value were age, gender, median income, insurance status, LOS, total levels, and levels C2, C3, C4, C5, and C7. For ACDF and PCF, respectively, average age was 51.2 and 53.6 years, gender was 53% and 69.3% male, and LOS was 2.3 and 2.9 days. Of this initial cohort, 26/ 627 (4.1%) ACDF patients and 10/163 (6.1%) PCF patients required reoperation at the index level(s) (p5.4). After propensity score matching, there were 188 ACDF patients matched to 140 PCF patients and all variables in Table 1 (except total levels and primary surgeon) had standardized differences less than 10.0 in absolute value. After propensity matching, none of the differences were statistically significant. The mean number of total levels operated
1280
D. Lubelski et al. / The Spine Journal 15 (2015) 1277–1283
Table 1 Summary statistics for ACDF and PCF patients before propensity score matching Characteristic
ACDF
PCF
Standard difference p Value
N 627 163 Age (y) 51.2 (10.7) 53.6 (12.8) 20.8 Gender: male, N (%) 332 (53.0) 113 (69.3) 34.1 BMI 29.1 (6.2) 28.7 (5.2) 7.3 Race White: N (%) 551 (87.9) 144 (88.3) 1.4 Black: N (%) 45 (7.2) 12 (7.4) 0.7 Other: N (%) 31 (4.9) 7 (4.3) 3.1 Tobacco use Never: N (%) 240 (38.3) 69 (42.3) 2.9 Former: N (%) 168 (30.2) 51 (32.9) 5.9 Current: N (%) 149 (26.8) 35 (22.6) 9.7 Median income 53.7 (17.8) 55.9 (18.0) 12.7 ($10,000) Insurance Private/other: N (%) 365 (69.1) 94 (63.5) 11.9 Medicaid: N (%) 27 (5.1) 1 (0.7) 26.7 Medicare: N (%) 136 (25.8) 53 (35.8) 21.9 Surgery duration (min) 211.3 (78.7) 216.8 (71.4) 7.3 Length of stay (d) 2.3 (4.2) 2.9 (5.9) 11.6 Level of surgery C2: N (%) 1 (0.2) 4 (2.5) 20.3 C3: N (%) 58 (9.3) 33 (20.2) 31.4 C4: N (%) 206 (32.9) 63 (38.7) 12.1 C5: N (%) 471 (75.1) 115 (70.6) 10.3 C6: N (%) 546 (87.1) 142 (87.1) 0.1 C7: N (%) 311 (49.6) 91 (55.8) 12.5 Total levels 2.5 (0.7) 2.7 (0.9) 25.9
.0230 .0002 .6527 .9404
.5573
.0865
.0061*
.2168 .3644 .0070* .0002 .1942 .2772 1.0000 .1839 .0065
ACDF, anterior cervical discectomy and fusion; PCF, posterior cervical foraminotomy; BMI, body mass index. Note: All values are presented as means (standard deviation) or number (%) for continuous and categorical variables, respectively. p Values were calculated using Mann-Whitney U tests for continuous variables and chisquare tests or Fisher exact tests for categorical variables. * p Value was calculated from Fisher exact test.
was 2.57 and 2.67 for ACDF and PCF, respectively (p5.3). Although balance among primary surgeons improved after propensity matching, there still was some imbalance; however, a Fisher exact test indicated that the differences in surgery rates among primary surgeons were not significantly different after propensity score matching (p5.08). Of the 188 ACDF patients, 9 (4.8%) had revision within 2 years of their initial surgery, whereas 9/140 (6.4%) PCF patients had revision within 2 years of their initial surgery. The absolute difference in revision rates for the two groups (6.4%4.8%51.6%) was significantly less than our hypothesized difference of 5% (p5.0131). Furthermore, a chi-square test indicated these percentages were not significantly different (p5.7).
Discussion Anterior cervical discectomy and fusion and PCF have been described as having similar postoperative outcomes
Table 2 Summary statistics for ACDF and PCF patients after propensity score matching Characteristic
ACDF
PCF
Standard difference p Value
N 188 140 Age (y) 53.3 (11.6) 52.9 (12.2) 2.3 Gender: male, N (%) 128 (68.1) 97 (69.3) 3.7 BMI 28.9 (5.4) 28.5 (5.2) 6.5 Race White: N (%) 166 (88.3) 123 (87.9) 1.4 Black: N (%) 13 (6.9) 10 (7.1) 0.9 Other: N (%) 9 (4.8) 7 (5.0) 1.0 Tobacco use Never: N (%) 70 (37.2) 59 (42.1) 7.4 Former: N (%) 62 (36.0) 45 (33.8) 3.4 Current: N (%) 40 (23.3) 29 (21.8) 4.8 Median income 56.7 (18.3) 55.9 (18.2) 4.5 ($10,000) Insurance Private/other: N (%) 111 (66.9) 82 (65.6) 2.7 Medicaid: N (%) 1 (0.6) 1 (0.8) 2.4 Medicare: N (%) 54 (32.5) 42 (33.6) 2.3 Surgery duration 218.1 (73.7) 216.9 (74.2) 1.3 (min) Length of stay (d) 2.5 (4.6) 2.6 (5.6) 3.1 Level of surgery C2: N (%) 1 (0.5) 1 (0.7) 2.3 C3: N (%) 25 (13.3) 23 (16.4) 8.8 C4: N (%) 60 (31.9) 51 (36.4) 9.5 C5: N (%) 132 (70.2) 98 (70.0) 0.5 C6: N (%) 165 (87.8) 122 (87.1) 1.9 C7: N (%) 101 (53.7) 79 (56.4) 6.5 Total levels 2.6 (0.7) 2.7 (0.8) 12.9
.7443 .9112 .4210 .9924
.8137
.5889
.9590*
.7739 .5550 1.0000* .5251 .4614 1.0000 1.0000 .7078 .2533
ACDF, anterior cervical discectomy and fusion; PCF, posterior cervical foraminotomy; BMI, body mass index. Note: All values are presented as means (standard deviation) or number (%) for continuous and categorical variables, respectively. p Values were calculated using Mann-Whitney U tests for continuous variables and chisquare tests or Fisher exact tests for categorical variables. * p Value calculated from Fisher exact test.
for pain, complications, and quality of life. Specifically, in various retrospective studies, the reported success rates have ranged from 87% to 94% for ACDF and 90% to 97% for PCF [7,22–26]. Reoperation rates have been described as similar for ACDF and PCF, with reports describing 6% to 9% for ACDF [25,27–36] and 5% for PCF [7]; however, none of these studies performed a direct comparison between the two. The RCTs [4,6] comparing these two approaches have similarly reported no differences in pain or complication outcomes. Several small retrospective cost analyses have been performed that have found that PCF costs less than ACDF [37–39] and that PCF is significantly more costeffective [39]. Nonetheless, in recent decades, the prevalence of ACDF has increased substantially faster than PCF [8,9]. Herkowitz et al. [6] randomized 33 patients with lateral disc herniations to either ACDF or laminotomy/foraminotomy and found no significant differences between the groups, showing subjectively ‘‘excellent or good’’ results
D. Lubelski et al. / The Spine Journal 15 (2015) 1277–1283
in 94% and 75% of ACDF and PCF patients, respectively. Wirth et al. [40] randomized 72 patients with cervical disc herniation and radiculopathy to PCF, ACDF, or anterior cervical discectomy without fusion between 1984 and 1991. They found that all three procedures had similar (no significant differences) LOS, perioperative symptoms, and shortterm (~2 months) and long-term (~5 years) postoperative improvement in pain and satisfaction from surgery. The authors also reviewed recurrent symptoms necessitating additional surgery and found no significant differences (reoperation rate was 27% for PCF and 28% for ACDF). Most recently, Ruetten et al. [4], in Germany, performed an RCT with 175 patients with lateral disc herniations and radiculitis or neurologic deficit to endoscopic PCF versus conventional microsurgical ACDF. They found no differences in complications, clinical outcomes for neck and arm visual analog scale, Hilibrand criteria outcomes, or radicular or neurologic outcomes. The reoperation rates were not significantly different between the groups, with 4.7% (4 of 86) in the ACDF group and 6.7% (6 of 89) in the PCF group. Although the previous studies represent Class I evidence, they are limited by a relatively small sample size. For this reason, the studies largely presented trends in the data and were only able to draw limited conclusions. Nonetheless, the RCTs found that ACDF and PCF have similar outcomes in terms of pain, complications, and reoperation rate. The largest of the RCTs, by Ruetten et al. [4], found very similar reoperation rates at index levels to the ones we found in the present study (ie, 4.7% in their cohort for ACDF vs. 4.7% in our study; 6.7% in their cohort for PCF vs. 6.4% in our study). Although retrospective, the present study had a substantially larger sample size that was propensity matched. Furthermore, through the use of an equivalence analysis, we were able to demonstrate that not only was there no significant difference between reoperation rates for ACDF and PCF, but also support our hypothesis that these rates were statistically equivalent. When interpreting these results it is important to understand that the data support the equivalency between the reoperation rates of ACDF and PCF, it does not indicate that one is superior to the other. As previously described, the outcomes of these two approaches, when the indication is equivocal, are often successful irrespective of the approach. There are advantages and disadvantages to each. As there has been a dramatic increase in the number of ACDF surgeries performed annually in the United States relative to PCF [9], our goal was to dispel the notion that one is superior to the other. It is important to understand that the benefits of PCF in avoiding fusion and fusion-related complications do not come at the expense of greater destabilization or predisposition to revision surgery. Rather, our results demonstrate that indexlevel reoperation rate at 2-year follow-up is equivalent between the two surgical approaches. It should be noted, however, that reoperation for adjacent segment disease is substantially more common after a fusion procedure,
1281
occurring in approximately 12% of patients undergoing ACDF [41] and only 2% to 3% after PCF [42]. The present study has several limitations that should be considered in the interpretation of our findings. It was a retrospective analysis, in which patient data were collected over a period of 7 years. The data in this study were objective, requiring minimal interpretation, but nonetheless, the nature of a retrospective analysis is that it is prone to data collection or interpretation bias. Two-year reoperation rates were determined, which we felt was sufficient time to determine index level failures, but studies with longer-term follow-up would reveal additional information regarding true costs and long-term complications. The retrospective design also meant that if some patients underwent reoperation at an outside institution, it would not be captured in our review. There is no reason to suspect, however, that there would be a difference in missed (outside hospital) reoperations between patients that underwent an index ACDF versus a PCF. Additionally, while propensity matching, the gold standard in retrospective studies, was used, we were only able to match based on the collected variables and there may have been differences in uncollected variables. Nonetheless, this is the largest study to date comparing reoperation rates for ACDF versus PCF. Furthermore, the propensity matching, ensuring similar demographic and operative characteristics between the groups, and the equivalence analysis strengthen the conclusions herein. Another limitation is that there may have been a selection bias for the types of patients that were treated with an ACDF versus a PCF. It is possible that those with an arthritic facet joint, severe neck pain, or a degenerative listhesis were preferentially treated with the fusion procedure. Additionally, anatomical considerations can influence selection; soft disc herniations that are exclusively localized laterally may be more easily addressed with a PCF, whereas a central/paracentral calcified disc can be more accessible with an ACDF. Therefore, our data should be interpreted as being valid for those patients who meet the indications for both procedures only and not those who would otherwise be contraindicated for the PCF. Conclusion After accounting for patient demographics and operative characteristics, ACDF and PCF have statistically equivalent 2-year reoperation rates at the index level, both of which are relatively low. Accordingly, spine surgeons can operate for cervical radiculopathy via a posterior approach in properly indicated patients without putting them at a comparatively increased risk for revision surgery. References [1] Abumi K, Panjabi MM, Kramer KM, Duranceau J, Oxland T, Crisco JJ. Biomechanical evaluation of lumbar spinal stability after graded facetectomies. Spine 1990;15:1142–7.
1282
D. Lubelski et al. / The Spine Journal 15 (2015) 1277–1283
[2] Brayda-Bruno M, Cinnella P. Posterior endoscopic discectomy (and other procedures). Eur Spine J 2000;9(1 Suppl):S24–9. [3] Mixter WJ, BARR JS. Rupture of the intervertebral disc with involvement of the spinal canal. N Engl J Med 1934;211:210–5. [4] Ruetten S, Komp M, Merk H, Godolias G. Full-endoscopic cervical posterior foraminotomy for the operation of lateral disc herniations using 5.9-mm endoscopes: a prospective, randomized, controlled study. Spine 2008;33:940–8. [5] Henderson CM, Hennessy RG, Shuey HM, Shackelford EG. Posterior-lateral foraminotomy as an exclusive operative technique for cervical radiculopathy: a review of 846 consecutively operated cases. Neurosurgery 1983;13:504–12. [6] Herkowitz HN, Kurz LT, Overholt DP. Surgical management of cervical soft disc herniation. A comparison between the anterior and posterior approach. Spine 1990;15:1026–30. [7] Wang TY, Lubelski D, Abdullah KG, Steinmetz MP, Benzel EC, Mroz TE, et al. Rates of anterior cervical discectomy and fusion after initial posterior cervical foraminotomy. Spine J 2013 Jul 16. [Epub ahead of print]. [8] Patil PG, Turner DA, Pietrobon R. National trends in surgical procedures for degenerative cervical spine disease: 1990-2000. Neurosurgery 2005;57:753–8; discussion 753–758. [9] Wang MC, Kreuter W, Wolfla CE, Maiman DJ, Deyo RA. Trends and variations in cervical spine surgery in the United States: Medicare beneficiaries, 1992 to 2005. Spine 2009;34:955–61; discussion 962–963. [10] Gao Y, Liu M, Li T, Huang F, Tang T, Xiang Z. A meta-analysis comparing the results of cervical disc arthroplasty with anterior cervical discectomy and fusion (ACDF) for the treatment of symptomatic cervical disc disease. J Bone Joint Surg Am 2013;95:555–61. [11] Fountas KN, Kapsalaki EZ, Nikolakakos LG, Smisson HF, Johnston KW, Grigorian AA, et al. Anterior cervical discectomy and fusion associated complications. Spine 2007;32:2310–7. [12] Skovrlj B, Gologorsky Y, Haque R, Fessler RG, Qureshi SA. Complications, outcomes, and need for fusion after minimally invasive posterior cervical foraminotomy and microdiscectomy. Spine J 2014;4: 2405–11. [13] Albert TJ, Vacarro A. Postlaminectomy kyphosis. Spine 1998;23: 2738–45. [14] Mikawa Y, Shikata J, Yamamuro T. Spinal deformity and instability after multilevel cervical laminectomy. Spine 1987;12:6–11. [15] Rosenbaum PR, Rubin DB. Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat 1985;39:33. [16] Hansen BB, Klopfer SO. Optimal full matching and related designs via network flow. J Comput Graph Stat 2006;609–27. [17] Gart JJ. The comparison of proportions: a review of significance tests, confidence intervals and adjustments for stratification. Rev Inst Int Stat 1971;39:148. [18] Dunnett CW, Gent M. Significance testing to establish equivalence between treatments, with special reference to data in the form of 2X2 tables. Biometrics 1977;33:593–602. [19] Rubin DB. Multiple imputation for nonresponse in surveys. Wiley, 2004. [20] Van Buuren S, Brand JP, Groothuis-Oudshoorn CG, Rubin DB. Fully conditional specification in multivariate imputation. J Stat Comput Sim 2006;76:1049–64. [21] Van Burren S, Groothuis-Oudshoorn C. Multivariate imputation by chained equations in R. J Stat Software 2011;45:1–6. [22] Jacobs W, Willems PC, Kruyt M, van Limbeek J, Anderson PG, Pavlov P, et al. Systematic review of anterior interbody fusion techniques for single- and double-level cervical degenerative disc disease. Spine 2011;36:E950–60. [23] Jagannathan J, Sherman JH, Szabo T, Shaffrey CI, Jane JA. The posterior cervical foraminotomy in the treatment of cervical disc/osteophyte disease: a single-surgeon experience with a minimum of 5
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
years’ clinical and radiographic follow-up. J Neurosurg Spine 2009;10:347–56. Adamson TE. Microendoscopic posterior cervical laminoforaminotomy for unilateral radiculopathy: results of a new technique in 100 cases. J Neurosurg 2001;95:51–7. Marotta N, Landi A, Tarantino R, Mancarella C, Ruggeri A, Delfini R. Five-year outcome of stand-alone fusion using carbon cages in cervical disc arthrosis. Eur Spine J 2011;20(1 Suppl): S8–12. Korinth MC, Kr€uger A, Oertel MF, Gilsbach JM. Posterior foraminotomy or anterior discectomy with polymethyl methacrylate interbody stabilization for cervical soft disc disease: results in 292 patients with monoradiculopathy. Spine 2006;31:1207–14; discussion 1215–1216. Bhadra AK, Raman AS, Casey AT, Crawford RJ. Single-level cervical radiculopathy: clinical outcome and cost-effectiveness of four techniques of anterior cervical discectomy and fusion and disc arthroplasty. Eur Spine J 2009;18:232–7. Song KJ, Taghavi CE, Lee KB, Song JH, Eun JP. The efficacy of plate construct augmentation versus cage alone in anterior cervical fusion. Spine 2009;34:2886–92. Burkus JK, Haid RW, Traynelis VC, Mummaneni PV. Long-term clinical and radiographic outcomes of cervical disc replacement with the Prestige disc: results from a prospective randomized controlled clinical trial. J Neurosurg Spine 2010;13:308–18. Woodworth RS, Molinari WJ, Brandenstein D, Gruhn W, Molinari RW. Anterior cervical discectomy and fusion with structural allograft and plates for the treatment of unstable posterior cervical spine injuries. J Neurosurg Spine 2009;10:93–101. Jensen WK, Moore TA, Tribus CB, Anderson PA, Zdeblick TA. Use of patella allograft for anterior cervical diskectomy and fusion. J Spinal Disord Tech 2009;22:392–8. Ruetten S, Komp M, Merk H, Godolias G. Recurrent lumbar disc herniation after conventional discectomy: a prospective, randomized study comparing full-endoscopic interlaminar and transforaminal versus microsurgical revision. J Spinal Disord Tech 2009;22:122–9. Mummaneni PV, Robinson JC, Haid RW. Cervical arthroplasty with the PRESTIGE LP cervical disc. Neurosurgery 2007;60:310–4; discussion 314–315. Anderson PA, Sasso RC, Riew KD. Comparison of adverse events between the Bryan artificial cervical disc and anterior cervical arthrodesis. Spine 2008;33:1305–12. Coric D, Cassis J, Carew JD, Boltes MO. Prospective study of cervical arthroplasty in 98 patients involved in 1 of 3 separate investigational device exemption studies from a single investigational site with a minimum 2-year follow-up. Clinical article. J Neurosurg Spine 2010;13:715–21. Coric D, Numley PD, Guyer RD, Musante D, Carmody CN, Gordon CR, et al. Prospective, randomized, multicenter study of cervical arthroplasty: 269 patients from the KineflexjC artificial disc investigational device exemption study with a minimum 2-year follow-up: clinical article. J Neurosurg Spine 2011;15:348–58. Mansfield HE, Canar WJ, Gerard CS, O’Toole JE. Single-level anterior cervical discectomy and fusion versus minimally invasive posterior cervical foraminotomy for patients with cervical radiculopathy: a cost analysis. Neurosurg Focus 2014;37:E9. Tumialan LM, Ponton RP, Gluf WM. Management of unilateral cervical radiculopathy in the military: the cost effectiveness of posterior cervical foraminotomy compared with anterior cervical discectomy and fusion. Neurosurg Focus 2010;28:E17. Alvin MD, Lubelski D, Abdullah KG, Whitmore RG, Benzel EC, Mroz TE. Cost-utility analysis of anterior cervical discectomy and fusion with plating (ACDFP) versus posterior cervical foraminotomy (PCF) for patients with single-level cervical radiculopathy at 1-year follow-up. J Spinal Disord Tech 2014 Mar 27. [Epub ahead of print].
D. Lubelski et al. / The Spine Journal 15 (2015) 1277–1283 [40] Wirth FP, Dowd GC, Sanders HF, Wirth C. Cervical discectomy. A prospective analysis of three operative techniques. Surg Neurol 2000;53:340–6; discussion 346–348. [41] Xu R, Bydon M, Macki M, De la Garza-Ramos R, Sciubba DM, Wolinsky JP, et al. Adjacent segment disease after anterior cervical
1283
discectomy and fusion: clinical outcomes after first repeat surgery versus second repeat surgery. Spine 2014;39:120–6. [42] Clarke MJ, Ecker RD, Krauss WE, McClelland RL, Dekutoski MB. Same-segment and adjacent-segment disease following posterior cervical foraminotomy. J Neurosurg Spine 2007;6:5–9.
