Indirect meta-analysis comparing clinical outcomes of total cervical disc replacements with fusions for cervical degenerative disc disease

Anterior cervical discectomy and fusion (ACDF) and total cervical disc replacement (TDR) are considered effective treatments for patients with cervical degenerative disc disease (CDDD). An indirect meta-analysis including 19 randomized controlled trials (5343 patients) was conducted to compare the clinical outcomes of ACDF with TDR. Primary outcomes including functional indicators (NDI [neck disability index] score, neurological success and patient satisfaction), secondary outcomes including surgical outcomes (operation time, blood loss and length of stay) and secondary surgical procedures (secondary surgery at an adjacent level, secondary surgery at the index level, secondary surgery at both levels, removal, reoperation, revision and supplemental fixation) were included in the study. TDR using the Bryan disc was associated with a greater improvement in NDI score than ACDF (MD = −5.574, 95% CrIs [credible intervals] −11.73–−0.219). For neurological success, the Bryan (odds ratio [OR] = 0.559, 95% CrIs 0.323–0.955) and Prestige (OR = 0.474, 95% CrIs 0.319–0.700) discs were superior to ACDF. However, no differences in the patient satisfaction rate were shown between TDR and ACDF. For patients with CDDD, ACDF using allograft and a plate is most effective for determining the surgical parameters. Moreover, TDR using the ProDisc-C, Mobi-C, Prestige and Bryan discs are good choices for improving functional outcomes and reducing secondary surgeries.


Risk of bias. The risk of bias of the 18 studies is shown in
. A total of six included studies had a low risk of bias for random sequence generation. None of the studies described allocation concealment. Four studies described the blinding participants and personnel. However, the participants and personnel were not blinded in one study. None of the included studies described blinding to outcome assessment. Twelve studies were considered at low risk for incomplete outcome data. Five studies were at high risk of incomplete outcome data. Only two studies were considered low risk for selective reporting. Table S1. For length of stay, high heterogeneity (I 2 = 89.0%) was shown in the comparisons of the Bryan disc vs. ACDF using allograft and a plate. For NDI scores, moderate heterogeneity (I 2 = 55.8%) was shown in the comparisons of Bryan disc vs. ACDF using allograft and a plate, indicating random variation between the investigations by chance. The remaining comparisons of TDR vs. ACDF showed minimal to low heterogeneity for all outcomes.

Assessments of heterogeneity. Heterogeneity for each outcome is shown in
Functional indicators. NDI score. The network plot of comparisons of NDI score is shown in Fig. 3. In total, 520 patients were assigned to ACDF using allograft and a plate, 166 patients to TDR using the ProDisc-C disc, 139 patients to TDR using the Bryan disc, and 280 patients to TDR using the Prestige ST disc.
Regarding NDI scores, the results showed that TDR with the Bryan disc was significantly more effective than ACDF using allograft bone and a plate (MD = −5.574, 95% CrI −11.73-−0.219) ( Table 2). Neurological success. In total, 1406 patients were assigned to ACDF using allograft and a plate, 754 patients to TDR using the Prestige disc, 342 patients to TDR using the ProDisc-C disc, and 507 patients to TDR using the Bryan disc. The current results indicated that TDR using the Prestige disc significantly improved neurological success than ACDF using allograft and a plate (OR = 0.474, 95% CrI 0.319-0.700). Moreover, TDR using the Bryan disc was significantly more efficacious in terms of neurological success than ACDF using allograft and a plate (OR = 0.559, 95% CrI 0.323-0.955) (Table S2).
Patient satisfaction. In all, 285 patients were assigned to ACDF using allograft and a plate, 58 patients to TDR using the Prestige disc, 109 patients to TDR using the ProDisc-C disc, and 338 patients to TDR using the Mobi-C disc.
The results showed no significant differences in the satisfaction rates between the pairwise comparisons (Table S3).

Surgical Parameters. Operation time.
In total, 547 patients were assigned to ACDF using allograft and a plate, 220 patients to TDR using the ProDisc-C disc, 338 patients to TDR using the Mobi-C disc, 68 patients to TDR using the Prestige ST disc, and 139 patients to TDR using the Bryan disc.
The patients who underwent ACDF using allograft and a plate experienced less operation time than those who underwent TDR using the ProDisc-C disc ( (Table S4).
Blood loss. In total, 435 patients were assigned to ACDF using allograft and a plate, 220 patients to TDR using the ProDisc-C disc, 338 patients to TDR using the Mobi-C disc, and 83 patients to TDR using the Bryan disc.
Length of stay. In total, 392 patients were assigned to ACDF using allograft and a plate, 176 patients to TDR using the ProDisc-C disc, 338 patients to TDR using the Mobi-C disc, and 83 patients to TDR using the Bryan disc.
For length of stay, the results showed no significant differences between the pairwise comparisons (Table S6).

Secondary surgical procedures. Secondary surgery at an adjacent level.
In total, 1257 patients were assigned to ACDF using allograft and a plate, 433 patients to TDR using the Prestige disc, 603 patients to TDR using the Bryan disc, 365 patients to TDR using the Mobi-C disc, and 135 patients to TDR using the Kineflex|C disc. ACDF using allograft and a plate showed a significantly higher rate of secondary surgery at an adjacent level than TDR with the Prestige disc (OR = 3.527, 95% CrI 1.396-9.439) and TDR with the Mobi-C disc (OR = 3.197, 95% CrI 1.185-8.908), respectively (Table S7).

Secondary surgery at the index level.
In all, 886 patients were assigned to ACDF using allograft and a plate, 547 patients to TDR using the Bryan disc, 503 patients to TDR using the Mobi-C disc, and 135 patients to TDR using the Kineflex|C disc.
The results showed no significant difference between the pairwise comparisons in the rate for rate of secondary surgery at the index level (Table S8).
Secondary surgery at both levels. In total, 822 patients were assigned to ACDF using allograft and a plate, 547 patients to TDR using the Bryan disc, 365 patients to TDR using the Mobi-C disc, and 135 patients to TDR using the Kineflex|C disc.   ACDF using allograft and a plate showed a significantly higher rate of secondary surgery at both levels than TDR with the Mobi-C disc (OR = 3.155, 95% CrI 1-10.78) (Table S9).
Removal. In all, 1016 patients were assigned to ACDF using allograft and a plate, 492 patients to TDR using the Prestige disc, 166 patients to TDR using the ProDisc-C disc, 248 patients to TDR using the Bryan disc, and 503 patients to TDR using the Mobi-C disc.
For the removal rate, TDR using the ProDisc-C disc showed a significantly higher rate of removal surgery than when the Prestige disc was used (OR = 16.9, 95% CrI 1.027-803.6) (Table S10).
Reoperation. In sum, 1511 patients were assigned to ACDF using allograft and a plate, 433 patients to TDR using the Prestige disc, 166 patients to TDR the ProDisc-C disc, 650 patients to TDR using the Bryan disc, 503 patients to TDR using the Mobi-C disc, and 135 patients to TDR using the Kineflex|C disc.
Regarding the reoperation rate, TDR using the Mobi-C disc showed a significantly lower rate of reoperation surgery than ACDF using allograft and a plate (OR = 0.275, 95% CrI 0.103-0.740) (Table S11).
Revision. In all, 812 patients were assigned to ACDF using allograft and a plate, 492 patients to TDR using the Prestige disc, 166 patients to TDR using the ProDisc-C disc, and 266 patients to TDR using the Bryan disc.
The results indicated that TDR with the Prestige disc was significantly associated with a reduced rate of revision surgery than ACDF using allograft and a plate (OR = 0.077, 95% CrI 0.003-0.767). Additionally, TDR using the ProDisc-C disc was also significantly associated with a reduced rate of revision than ACDF using allograft and a plate (OR = 0.037, 95% CrI 0.000-0.629) (Table S12).
Supplemental fixation. In all, 1040 patients were assigned to ACDF using allograft and a plate, 433 patients to TDR using the Prestige disc, 166 patients to TDR using the ProDisc-C disc, 331 patients to TDR using the Bryan disc, and 503 patients to TDR using the Mobi-C disc.
TDR with the Prestige disc was significantly associated with a reduced rate of supplemental fixation surgery than ACDF using allograft and a plate (OR = 0.033, 95% CrI 0.001-0.305). Additionally, TDR with the Mobi-C disc was also significantly correlated with a reduced rate of supplemental fixation surgery than ACDF using allograft and a plate (OR = 0.116, 95% CrI 0.013-0.714) ( Table S13).
Ranking of treatments. The ranking results of the treatments are shown in Tables 2 and S2 to S13. For reducing blood loss and operation time, ACDF using allograft and a plate may be the best choice. For reducing length of stay, secondary surgery rates at both levels and at the index level and the reoperation rate, TDR using the Mobi-C disc may be the best option. TDR using the Bryan disc for improving the NDI score and the Prestige disc for increasing neurological success, reducing secondary surgery at an adjacent level, removal, and supplemental fixation, and the ProDisc-C disc for improving satisfaction rate and reducing the revision rate were the best choices, respectively. ACDF ranked second in reducing length of stay but was second to last in reducing the removal rate. Whereas it ranked lowest in improving NDI, neurological success, patient satisfaction and reducing secondary surgery at an adjacent level, both levels and the index level, reoperation, revision and supplemental fixation.
Sensitivity analyses. The results of each outcome were not altered by the sensitivity analyses.

Discussion
All relevant studies available concerning treatment of CDDD patients with TDR using the Prestige/ProDisc-C/ Bryan/Mobi-C/Kineflex|C discs and ACDF using allograft and a plate were included in the present meta-analysis. In total, thirteen available outcomes were estimated in the indirect comparison using random-effects models. Six to fourteen studies were included for each outcome.
The ranking results showed that TDR with the Mobi-C disc may be the best choice to reduce the length of stay, secondary surgery rate at both levels, secondary surgery rate at the index level and reoperation rate. TDR using the Bryan disc may be the first choice to improve patient NDI scores. TDR using the Prestige disc may be the best choice to increase the neurological success rate and to reduce secondary surgery rate at an adjacent level, removal surgery rate and supplemental fixation rate. TDR with the ProDisc-C disc may optimally increase the satisfaction rate and reduce the revision rate. Interestingly, the ranking results indicated that ACDF using allograft and a plate may be the best choice only for reducing blood loss and the operation time. However, ACDF using allograft and a plate may be the worst choice for improving the NDI score, neurological success and the satisfaction rate, and for reducing the secondary surgery rate at an adjacent level, secondary surgery rate at both levels, secondary surgery at the index level, reoperation, revision and the supplemental fixation rate. For reducing the removal surgery rate, ACDF using allograft and a plate ranked second to last. That is, for patients with CDDD, TDR using replacements including the Mobi-C disc, Bryan disc, Prestige disc and ProDisc-C disc are better choices than ACDF using allograft and a plate.
In all, five meta-analyses 32,34,35,39,41 and the short-term results of one meta-analysis 11 indicated that no significant difference existed between TDR and ACDF in improving the NDI score, although another five studies 7,31,38,40,42 and the long-term results of one study 11 showed that TDR was superior to ACDF, similar to the meta-analyses. The reasons for these findings may be as follows: (1) number of included studies differed significantly; (2) unfitted data was included in published meta-analysis; e.g., the data of Heller JG 2009 19 included the NDI improvement from baseline but not NDI score at the last follow-up; (3) some outcome data lacked SD values such as those of Mummaneni P 2007in Xing D 2013 , which were excluded from the current study; and (4) comparison of Nabhan A 2011 in Xing D 2013 39 was between the ProDisc-C and ACDF using the Solis cage and a titanium plate, which was the only ACDF study included in the comparison.
For neurological success rate, twelve meta-analyses 6, 9, 31-34, 36-40, 42 compared TDR and ACDF. Almost all results of the included studies were in accord with the current study except for Ren C 2014 37 , the short-term results of Boselie TF 2013 31 and the midterm results of Zhang Y 2015 42 , perhaps because the numbers of patients included in those three studies were significantly less than the current study.
In total, two meta-analyses 7, 31 compared TDR with ACDF in terms of patient satisfaction. Muheremu A 2015 7 indicated that TDR was better than ACDF, which differed from the current study. The reasons for this findings may be that Heller JG 2009, Mummaneni P 2007 and Sun 2008 in Muheremu A 2015 7 were not included in the current study. Data of the three studies were not identified in the articles.
In total, four meta-analyses 6,9,11,32 and five studies 6,9,11,32,33 comparing TDR with ACDF investigated operation time reduction and length of stay separately. The results of these studies were in accord with the current study.
For blood loss, four meta-analyses 6,9,32,33 comparing TDR with ACDF and one meta-analysis 11 comparing TDR using the Bryan disc with ACDF were published. The results of two studies 6,9 indicated no significant difference existed between TDR and ACDF, which differed from the current study. Two reasons may explain this difference: (1) several studies included in the published meta-analyses were excluded from the current study because only one study was included in the individual treatment comparison. For example, the comparisons of Philips FM 2015 and Coric D 2011 in Luo J 2015 6 were TDR using the PCM disc vs. ACDF using a CSLP or SLIM LOC anterior plate and the Kineflex|C vs. ACDF using allograft and a plate, and (2) some outcome data lacked SD values such as those of Zhang X 2012 included in Rao M 2015 9 , which were excluded from the current study.
A total of thirteen meta-analyses 6, 9-11, 32-34, 37-42 investigated the secondary surgery rate of TDR and ACDF. Almost all the meta-analyses results were in accord with the current study except for four studies 10,38,40,42 . The results of secondary surgery at an adjacent level reported in two studies 10,40 and those of secondary surgery at the index level reported in three studies 38,40,42 differed from the current study. The reason for these differences may be the significantly different number of patients included.
In all, two meta-analyses 33, 35 described reoperation rates of TDR and ACDF. The results of Luo J 35 showed that TDRs was associated with a greater reduction in the reoperation rate than ACDF, which was in accord with the current study. However, the results of Gao Y 2013 33 showed that no significant difference existed between the two treatments, which differed from the current study. The only results on removal and supplemental fixation rates were also described in Gao Y 2013 33 , which also differed from the current meta-analysis. However, the article did not provide detailed outcome data about reoperation, removal and supplemental fixation. Therefore, the reason for the discrepancy between Gao Y 2013 33 and the current study could not be discerned. Two studies 33, 41 that compared TDR with ACDF investigated revision. The results of those studies were in accord with the current study.
The current meta-analysis has several strengths: (1) pairwise comparisons between ACDF using different implants and TDR using different replacements were conducted for the first time to provide comprehensive treatment information for patients with CDDD, and (2) available evidence was applied to the ranking of treatments to provide suggestions about the best choice for patients for each clinical outcomes. However, this article has some limitations: (1) this study lacks direct comparisons between TDR using different replacements and comparisons between different ACDF methods because only one study comparing outcomes between different TDRs was retrieved and less than 3 pairwise comparisons between different TDRs were included in the network plot for the outcome, (2) the numbers of patients included for each treatment were small, e.g., the sample sizes of the Bryan disc for blood loss and length of stay and the Prestige disc for operation time and satisfaction rate were less than 100; thus, the results presented in the current study should be interpreted cautiously, (3) different follow-up durations and races of patients in the included studies might be potential study confounders, (4) the results of outcomes including the SF-36 PCS, SF-36 MCS, VAS neck pain score, VAS arm pain score and return-to-work status were excluded from the current study and (5) the data of several studies were excluded because only one trial was included in each comparison. Therefore, the results of the current meta-analysis could not be assessed comprehensively.

Conclusion
Available randomized controlled study data show that ACDF using allograft and a plate is superior to TDR in terms of surgical parameters, whereas TDR using the Bryan, Prestige, Mobi-C, and ProDisc-C discs are more effective than ACDF using allograft and a plate in regard to functional parameter improvement and secondary surgical procedure reduction. More head-to-head RCTs that directly compare ACDF using different implants or TDR using different replacements in a pairwise fashion are needed to comprehensively confirm the current results.

Materials and Methods
Inclusion criteria. Patients. Adult patients with symptomatic cervical degenerative disc disease including radiculopathy and/or myelopathy.
Interventions and comparisons. We included each pairwise comparison between total disc replacements (TDRs) using different replacements and anterior cervical discectomy and fusions (ACDFs) using different implants.
Scientific RepoRts | 7: 1740 | DOI:10.1038/s41598-017-01865-3 Outcomes. The primary outcomes included functional outcomes (NDI, SF-36 PCS, SF-36 MCS, VAS neck pain score, VAS arm pain score, neurological success, patient satisfaction and return-to-work status) and secondary outcomes included surgical parameters (operative time, blood loss and length of stay) and secondary surgical procedures (secondary surgery at an adjacent level, secondary surgery at the index level, secondary surgery at both levels, removal, reoperation, revision and supplemental fixation).
Study design. Randomized controlled trials were included in the study.
Exclusion criteria. 1. Patients with tumor, metabolic bone disease or trauma. 2. Lack of a detailed description of the surgical implants, e.g., TDR using the Bryan disc, ACDF using allograft and a plate. 3. Results of outcomes were not synthesized in the network meta-analysis in the following situations: (1) less than 3 different pairwise comparisons were included in the network plot for any outcome and (2)  Data extraction. Extraction of the available data was conducted by two reviewers (Xu B and Ma JX) separately. Any disagreement regarding the eligibility of data generated between the two reviewers was resolved via discussion among the investigators. Information concerning the author names, publication year, country, search duration, comparison of interventions, sample size, age and gender, number of treated levels, length of follow-up period, and clinical outcomes were extracted.
Risk of bias assessment. The quality of the included studies was estimated using the Cochrane Collaboration tool 43 for estimating risk of bias. Items for assessment included sequence generation, allocation concealment, blinding of outcome assessor, incomplete outcome data and selective outcome reporting. Each item was assessed using three degrees including high risk of bias, low risk of bias and unclear.
Data synthesis and analysis. Each treatment effect was estimated using the mean difference (MD) with 95% credible intervals (CrIs) or odds ratio (OR) with 95% CrIs for continuous or dichotomous variables, respectively. Statistical heterogeneity was estimated for each outcome of each pairwise comparison using I 2 values. I 2 values larger than 25%, 50% and 75% indicated low, moderate and high heterogeneity, respectively 44 .
A Bayesian indirect meta-analysis with random-effects models was conducted for each outcome using WinBUGS version 1.4 (UK). The model convergence was estimated using trace plots and the Brooks-Gelman-Rubin statistic; a burn-in of 10,000 followed by another 50,000 iterations was considered appropriate for convergence for each outcome 45 . The surface under the cumulative ranking area (SUCRA) was calculated to rank each treatment. SUCRA values of 100% and 0% indicate the best treatment and the worst treatment 46 , respectively.
Next, sensitivity analyses were conducted for each outcome by excluding each study included.