Mid- to Long-Term Outcomes of Cervical Disc Arthroplasty versus Anterior Cervical Discectomy and Fusion for Treatment of Symptomatic Cervical Disc Disease: A Systematic Review and Meta-Analysis of Eight Prospective Randomized Controlled Trials

Purpose This study aimed to investigate the mid- to long-term outcomes of cervical disc arthroplasty (CDA) versus anterior cervical discectomy and fusion (ACDF) for the treatment of 1-level or 2-level symptomatic cervical disc disease. Methods Medline, Embase, and the Cochrane Central Register of Controlled Trials databases were searched to identify relevant randomized controlled trials that reported mid- to long-term outcomes (at least 48 months) of CDA versus ACDF. All data were analyzed by Review Manager 5.3 software. The relative risk (RR) and 95% confidence intervals (CIs) were calculated for dichotomous variables. The weighted mean difference (WMD) and 95%CIs were calculated for continuous variables. A random effect model was used for heterogeneous data; otherwise, a fixed effect model was used. Results Eight prospective randomized controlled trials (RCTs) were retrieved in this meta-analysis, including 1317 and 1051 patients in CDA and ACDF groups, respectively. Patients after an ACDF had a significantly lower rate of follow-up than that after CDA. Pooled analysis showed patients in CDA group achieved significantly higher rates of overall success, Neck Disability Index (NDI) success, neurological success and significantly lower rates of implant/surgery-related serious adverse events and secondary procedure compared with that in ACDF group. The long-term functional outcomes (NDI, Visual Analog Scale (VAS) neck and arm pain scores, the Short Form 36 Health Survey physical component score (SF-36 PCS)), patient satisfaction and recommendation, and the incidence of superior adjacent segment degeneration also favored patients in CDA group with statistical difference. Regarding inferior adjacent segment degeneration, patients in CDA group had a lower rate without statistical significance. Conclusions This meta-analysis showed that cervical disc arthroplasty was superior over anterior discectomy and fusion for the treatment of symptomatic cervical disc disease in terms of overall success, NDI success, neurological success, implant/surgery-related serious adverse events, secondary procedure, functional outcomes, patient satisfaction and recommendation, and superior adjacent segment degeneration.


Search Strategy
The study was conducted following the Preferred Reporting Items for systematic Reviews and Meta-analysis (PRISMA) guidelines [42]. Medline, Embase, the Cochrane Central Register of Controlled Trials databases were searched through January 2016 by using the following key terms: "cervical disc arthroplasty", "fusion", "arthrodesis", and "randomized controlled trial" (S1 Fig). The searches were limited to studies published in English. The reference lists of selected articles and relevant reviews were also reviewed to identify studies not identified in the original search. Two investigators independently reviewed all subjects, abstracts, and the full text of articles that were potentially eligible based on abstract review. Any disagreements between the investigators were discussed and resolved by consensus.

Eligibility Criteria
We included studies that met the following conditions: (1) prospective randomized controlled trials comparing CDA with ACDF with a minimum 48 months of follow-up; (2) subjects who were older than 18 years of age and had 1-level or 2-level symptomatic cervical disc disease unresponsive to non-operative treatment for at least 6 weeks; (3) at least one desirable outcome should be reported. Articles were excluded if they had any of the following characteristics: (1) non-RCTs, retrospective studies, or case series; (2) follow-up duration was less than 48 months; (3) duplicated publications from the same investigational site.

Methodological Quality Evaluation
Two reviewers independently performed the quality of the included studies using the 12 criteria recommended by the Cochrane Back Review Group. According to the recommendation by the Cochrane Back Review Group, studies were rated as having "low risks of bias" when at least 6 of the 12 criteria were met without serious flaws. Otherwise, the studies were rated as having "high risk of bias".

Data Extraction
Two reviewers independently extracted the following data: study country, publication year, study design, sample size, follow-up duration, patient demographics, prosthesis type, overall success, neurological success, Neck Disability Index (NDI) success, patient satisfaction and recommendation, implant/surgery-related serious adverse events (classified as WHO grade 3 or 4), secondary procedure, NDI, neck and arm pain scores, the Short Form 36 Health Survey physical component score (SF-36 PCS), and adjacent segment degeneration.

Data Analysis
The analysis was carried out using Reviewer Manager 5.3 software (Cochrane Collaboration, Oxford, UK). For dichotomous variables, the relative risk (RR) and 95% confidence intervals (CIs) were calculated. For continuous variables, the weighted mean difference (WMD) and 95% CIs were calculated. The level of significance was set as P< 0.05. Standard errors, confidence intervals, P values for difference in means, and interquartile ranges were transformed into standard deviation (SD), where necessary, according to the Cochrane Handbook for Systematic Reviews of Interventions. Statistical heterogeneity was evaluated using the chi-square test and Higgin's I 2 test. A P value of chi-square test < 0.10 or I 2 > 50% indicated statistical heterogeneity, prompting a random effects modeling estimate. Otherwise, a fixed effects model was used. Subgroup analysis was performed on patients with only 1-level cervical disc disease.

Methodological Quality Assessment
The methodological quality of all included studies is presented in Table 2. All eight studies were rated as "low risk of bias" according to the Cochrane Back Review Group criteria. One study [59] failed to clearly report adequate randomization and only one study provided the information of allocation concealment [59]. Blinding of patients, surgeons, and assessors were not considered achieved because of the nature of the studies and evident difference of implant design. Missing information such as the absence of intention-to-treatment analysis and followup loss were presented in seven studies [53][54][55][56][57][58]60].

Overall Success, NDI Success, and Neurological Success
Overall success was considered achieved if a patient met all of the following items: NDI success, neurological success, absences of implant/surgery-related serious adverse events and secondary procedure. Two studies [53,58] reported the overall success data. Pooled analysis showed that patients in CDA group had a higher rate of overall success compared with that in ACDF group (RR = 1.17; 95%CI: 1.07, 1.28; P = 0.0005; I 2 = 0%; Fig 2). NDI success was defined as postoperative NDI score improvement of at least a 15-point increase from preoperative score. Three studies [53,57,58] reported the NDI success data. Pooled analysis revealed a higher rate of NDI success in CDA group (RR = 1.10; 95%CI: 1.04, 1.18; P = 0.002; I 2 = 17%; Fig 2). Neurological success was determined as postoperative maintenance or improvement in each of the individual neurological evaluations (muscle strength, sensory deficit, and reflex functions) compared with the preoperative status. Six studies [53,[55][56][57][58]60] reported the neurological   Fig 3).

Subgroup Analysis
Subgroup analyses were performed on patients with only 1-level cervical disc disease. The results are shown in Table 3.

Discussion
Although artificial cervical discs have been utilized in spinal surgery for several years, ACDF remains the gold standard for the treatment of symptomatic cervical disc disease. It is partly attributable to the uncertainty of the long-term outcomes of cervical disc arthroplasty compared with the well-perceived long-term success of ACDF. To our knowledge, there are many meta-analysis studies available in the literature comparing the efficacy and safety of CDA with ACDF for the treatment of symptomatic cervical disc disease. However, most of them included the studies with short-term follow-up, which made it impossible to conclude the long-term comparativeness. Therefore, we performed a meta-analysis of eight RCTs with at least 48 months follow-up to determine whether CDA was superior over ACDF. This meta-analysis found that patients in CDA group had a significantly higher overall success rate compared with that in ACDF group. Pooled analysis of NDI success and neurological success data also revealed to be in favor of CDA. Moreover, we extracted NDI, VAS, and SF-36 scores at last follow-up to evaluate functional outcomes. Pooled estimates of these data showed superiority in CDA except for arm pain score improvement data, which showed no significant difference. These findings suggested that CDA seemed to be more effective than ACDF for the treatment of cervical spondylosis.
Secondary procedure is an important clinical event with substantial clinical and financial burdens for the patient as well as additional cost for the payor. In this meta-analysis, we found that CDA was superior to ACDF regarding the rate of total secondary procedures. Pooled results of the data of secondary procedure involving index level or adjacent level also revealed superiority in CDA group. These results were consistent with Wu et al.' findings [63]. However, they only included four randomized controlled trials with only 921 patients in total.
We observed that most of the adverse events reported in the included studies were medical problems unrelated to the index surgery or the cervical spine. Therefore, we chose implant/ surgery-related serious adverse events for the assessment of safety. Pooled results showed a lower rate in CDA patients, suggesting CDA seemed to be surgically safer than ACDF for the treatment of symptomatic cervical disc disease. In addition, three studies reported the data of patient satisfaction and patient recommendation. Regarding these self-assessed data, this metaanalysis revealed better results reported in CDA patients, supporting the superior efficacy in CDA over ACDF. Adjacent segment degeneration has been considered as one major concern for patients undergoing ACDF for degenerative disc disease [3][4][5][6][7]. Compared to cervical fusion, disc arthroplasty provides theoretical biomechanical advantage of motion preservation and stress reduction at adjacent levels [12][13][14][15][16]. However, it remains unclear whether CDA can decrease the incidence of adjacent segment degeneration compared to ACDF [33,36,40,49,64]. In this meta-analysis, no studies reported the rate of symptomatic adjacent segment disease while three studies reported the rate of radiological adjacent segment degeneration where pooled outcomes demonstrated a significantly lower rate of superior adjacent segment degeneration and an insignificantly lower rate of inferior adjacent segment degeneration in CDA patients. These findings suggested that CDA seemed to have positive effect on the process of adjacent segment degeneration. We noticed the statistical heterogeneity was high for these outcomes. This level of heterogeneity might be due to the difference of radiological criteria determined for adjacent segment degeneration and number of surgical levels. Of note, radiological adjacent segment degeneration is known to not directly correlate with symptomology [6,65]. Therefore, prospective RCTs with long-term follow-up reporting symptomatic adjacent segment disease as an outcome are warranted to clarify this question.
Several potential limitations should be acknowledged in our meta-analysis. First, only eight RCTs with follow-up between 4 to 7 years were included in this meta-analysis. Further studies with larger sample sizes and longer follow-up are warranted. Second, there were some methodological weaknesses in the included studies, such as unclear methods of allocation concealment and inadequate blinding procedures. Moreover, missing information such as the absence of ITT analysis and follow-up loss was presented in almost every study. All these methodological drawbacks would weaken the credibility of pooled outcomes. Third, patients undergoing an ACDF had a tendency to have poorer follow-up rate than those in CDA group (P = 0.04), which might lead to biased results. The reasons for this bias are not clear and probably multifactorial because of the lack of blinding of patients in all studies. Fourth, almost all the studies utilized a non-inferiority study design, which is typically less stringent in demonstrating efficacy than standard clinical trials. Despite these limitations, we still believe that this meta-analysis supports the superiority of CDA over ACDF on efficacy and safety for the treatment of symptomatic cervical disc disease in mid-to long-term follow-up.

Author Contributions
Conceived and designed the experiments: YH GL. Performed the experiments: YH SR. Analyzed the data: YH GL. Contributed reagents/materials/analysis tools: YH SR. Wrote the paper: YH GL SR DJ.