Key Questions

The review is defined by 13 Key Questions that address the effectiveness and harms of treatments for cervical degenerative disease (CDD), as well as how effectiveness and harms may differ by patient and disease characteristics (e.g., age, gender, severity of disease, vertebral level(s) of involvement). Two Contextual Questions were also included to help inform the report. Contextual Questions are not reviewed using systematic review methodology.

Key Question 1: In patients with radiographic spinal cord compression and no cervical spondylotic myelopathy, what are the comparative effectiveness and harms of surgery compared to non-operative treatment or no treatment?

Key Question 3: In patients with cervical degenerative disease, what are the comparative effectiveness and harms of surgical compared to non-operative treatment?

Key Question 4: In patients with cervical degenerative disease, what are the comparative effectiveness and harms of therapies added on to surgery (pre- or post-operative) compared with the same surgery alone?

Key Question 5: In patients with cervical radiculopathy due to cervical degenerative disease, what are the comparative effectiveness and harms of posterior versus anterior surgery?

Key Question 6: In patients with cervical degenerative disease, what are the comparative effectiveness and harms of posterior versus anterior surgery in patients with greater than or equal to three level disease?

Key Question 7: In patients with cervical spondylotic myelopathy due to cervical degenerative disease, what are the comparative effectiveness and harms of cervical laminectomy and fusion compared to cervical laminoplasty in patients?

Key Question 8: In patients with cervical spondylotic radiculopathy or myelopathy at one or two levels, what are the comparative effectiveness and harms of cervical arthroplasty compared to anterior cervical discectomy and fusion?

Key Question 9: In patients undergoing anterior cervical discectomy and fusion, what are the comparative effectiveness and harms of surgery based on interbody graft material or device type?

Key Question 10: In patients with pseudarthrosis after prior anterior cervical fusion surgery, what are the comparative effectiveness and harms of posterior approaches compared to revision anterior arthrodesis?

Key Question 11: In patients with cervical spondylotic myelopathy, what is the prognostic utility of preoperative magnetic resonance imaging (MRI) findings for neurologic recovery after surgery?

Key Question 12: What is the sensitivity and specificity of imaging assessment for identifying symptomatic pseudarthrosis after prior cervical fusion surgery?

Key Question 13: In patients with cervical spondylotic myelopathy, what are the comparative effectiveness and harms of intraoperative neuromonitoring (e.g., with somatosensory or motor evoked potential measurements) versus no neuromonitoring on clinical outcomes in patients undergoing surgery?

For purposes of these Key Questions, we focused on symptomatic CDD; with the exception of Key Question 1, evaluation and management of asymptomatic disease is beyond the scope of this review.

Contextual Questions

Contextual Question 1: What is the prevalence of cervical degenerative disease with spinal cord compression in asymptomatic patients?

Contextual Question 2: What is the natural history of untreated spinal cord compression in patients with cervical degenerative disease?

Findings in Relation to the Decisional Dilemmas

Cervical degenerative disease (CDD), which affects millions of older Americans, may lead to neck pain, radiculopathy, and myelopathy. Treatment of CDD, initially limited to conservative therapies (e.g., neck collar, traction, physiotherapy), has evolved to include instrumented and noninstrumented surgeries to decompress nerve roots and/or the spinal cord. Decisional dilemmas concerning best management of CDD include determination of whether one or more nonoperative treatments instead of surgery or in addition to surgery is preferred, and, if surgery is indicated, the determination of the most effective operative approaches and techniques for each individual patient.

Fifty-seven randomized controlled trials (RCTs) (in 82 publications), 56 nonrandomized studies (in 57 publications), and 1 systematic review provided evidence for this review. The highest-quality evidence was for cervical arthroplasty versus anterior cervical discectomy and fusion (ACDF) in patients with cervical radiculopathy and/or myelopathy. Evidence for nonsurgical interventions was particularly limited. Similarly, there was no evidence to guide treatment for asymptomatic patients with radiographic spinal cord compression.

Conservative (nonoperative) therapy or operative treatment. There was insufficient evidence to determine the effectiveness of nonoperative compared with operative treatment for CDD, and limited evidence to suggest no important difference in pain beyond two weeks when a postoperative cervical collar was added to laminoplasty (SOE: Low). Post-operative pulsed electro-magnetic field stimulation in addition to ACDF was associated with a greater likelihood of fusion than ACDF alone (SOE: Low). Evidence for exercise therapy was insufficient.

Anterior or posterior surgery. Anterior approaches were primarily ACDF and included anterior cervical foraminotomy (ACF) and anterior decompression without fusion; posterior approaches included posterior cervical discectomy and fusion, laminoplasty and posterior cervical foraminotomy. Single-level surgery was performed in patients with radiculopathy and two or more levels in patients with myelopathy. There was low strength of evidence of no difference between these approaches for improvement in pain, function, quality of life, or reoperation in patients with fewer than three operated levels (SOE: Low). There was limited evidence to suggest that a posterior approach is associated with increased likelihood of experiencing any serious adverse event in patients with greater than or equal to 3-level disease (SOE: Low). Selection bias, inadequate adjustment for potential confounding factors (e.g. age, comorbidities), confounding by indication, and other methodological limitations in NRSIs resulted in low or insufficient SOE for all outcomes, particularly in patients with ≥3-level diseases.

Laminoplasty or laminectomy and fusion. In patients with cervical spondylotic myelopathy, there was moderate strength evidence indicating similar benefits on postoperative function between laminectomy and fusion compared with laminoplasty and no important difference in reoperation rates, although limited evidence suggests laminoplasty may be associated with fewer complications than laminectomy and fusion (SOE: Low).

Disc replacement or fusion. In patients with radiculopathy and/or myelopathy at one level, there was moderate strength evidence of no important difference between cervical arthroplasty and ACDF in pain or function. Cervical arthroplasty was associated with substantially decreased likelihood of reoperation (SOE: High) and slightly lower likelihood of any serious adverse event in the short term (SOE: Low), but there was no important difference between cervical arthroplasty and ACDF in serious adverse events longer term (SOE: Low). However, index level reoperation rates for ACDF may be influenced by removal an existing plate to treat adjacent segment disease. This may artificially inflate the reported reoperation rate for ACDF versus C-ADR. Studies did not consistently specify reasons for revision. Additionally, MRI artifact created by the artificial disc may obscure pathology while concerns related to fusion may be more apparent, leading to more revisions with ACDF vs. ADR. The actual impact of these factors on reported reoperation rates is unclear.

Study findings were similar in patients with 2-level cervical arthroplasty or ACDF in pain and function and likelihood of reoperation at the index level, but the likelihood of an adverse event was slightly lower at 24 with months with cervical arthroplasty and no different at 120 months (SOE: Low). Evidence was sparse for this comparison beyond two levels. The majority of these cervical arthroplasty were industry funded and were frequently authored by individuals with industry-related conflicts of interest.

In patients with pseudarthrosis after ACDF, evidence on comparative effectiveness and harms of revision anterior arthrodesis versus a posterior approach was lacking.

ACDF graft choices. In patients undergoing ACDF, there was moderate strength evidence of no important difference between use of a standalone cage or a plate and cage in fusion rate, postoperative arm pain, function, quality of life, or subsidence. In a comparison of titanium/titanium-coated cages versus polyetheretherketone (PEEK) cages in ACDF, there was limited evidence to suggest that use of a PEEK cage results in a greater likelihood of fusion and function improvement than use of a titanium/titanium-coated cage (SOE: Low). In patients undergoing ACDF, there was also low strength evidence to suggest an increased risk of complications with the use of bone morphogenetic protein 2 (BMP-2) in the cervical spine compared with fusion without the use of BMP-2 (i.e., use of other osteogenic materials).

Other decisional dilemmas included the use of pre- and post-operative imaging findings and associations with better or worse outcomes, and the use or nonuse of intraoperative neuromonitoring on patients undergoing cervical spine surgery.

Role of imaging. Evidence for imaging to predict neurologic recovery was heterogeneous, as various study methods were used (e.g., different type and basis of classification of increased signal intensity, different outcomes, and different statistical analysis methods), thus making comparisons across studies challenging. In patients with cervical myelopathy, there was limited evidence to suggest that multisegmental T2-weighted increased signal intensity, sharp T2-weighted increased signal intensity, and increased SIR are associated with poorer neurologic recovery (SOE: Low).

In an asymptomatic and symptomatic populations, there was limited evidence suggesting that postoperative ACDF dynamic radiographs can predict pseudarthrosis with surgical exploration used as the gold standard (SOE: Low).

Intraoperative neuromonitoring or no monitoring. There was limited evidence to suggest that patients undergoing anterior cervical discectomy and fusion with IONM had similar likelihood of neurological complications as patients undergoing surgery without IONM (SOE: Low). Two databases (NIS and PearlDiver) were included, but only the NIS analysis used propensity score matching. The PearlDiver study did not match or control for confounders, but had similar results. In the total NIS sample, 42 percent of participants had radiculopathy alone and 31 percent had myelopathy (proportions not reported in the matched sample), 66 percent had a CCI of 0, and 84 percent had 1-2 level fusion. The PearlDiver study did not report baseline radiculopathy, myelopathy, comorbidities or levels fused. These findings apply only to ACDF procedures; neither study evaluated posterior cervical procedures. Both studies relied on claims data to distinguish patients that had IONM versus those who did not, which may significantly underreport the number who received IONM.

Implications for Clinical and Policy Decisions

This review was sponsored by the Congress of Neurological Surgeons (CNS) to update their 2009 guidelines on the management of CDD. Our review provides additional evidence that operative approaches to management of CDD generally result in improvement in pain, function, and quality of life postoperatively, as well as successful fusion (if a fusion surgery). In many cases patient-centered benefit outcomes between compared operative approaches and techniques were similar. The likelihood of general or specific adverse events, such as need for reoperation/revision surgery, were where most differences between therapies were observed and may help guide decision making regarding best operative approach for any given patient.

Our review provides additional support to the 2009 finding that preoperative magnetic resonance imaging (MRI) can help predict better or worse outcomes and to the 2009 recommendation discouraging use of BMP-2 in the cervical spine. Standalone cages for cervical fusion represent a newer design (Zero-P approved for use in the United States in 2008) and not covered in the 2009 guidelines. Although a more modern design, we did not find it superior to the use of anterior plating for most outcomes.

Gaps in the evidence make it difficult to create recommendations and inform policy. For example, challenges remain in determining the preferred course of action in patients with incidental findings of spinal cord compression on MRI. Although the natural history of non-myelopathic spinal cord compression is poorly understood, limited evidence suggests that some patients develop myelopathy over time, but it is not clear if any treatment provided prior to the development of symptoms results in better outcomes than treating symptomatic disease. Another challenge remaining is determining when conservative treatment may be preferred and what therapies are most effective compared with operative management or result in better outcomes when added to surgery. Good quality comparative evidence on conservative treatment was sparse in this review.

This review does not provide cost information.

Strength and Limitations of the Systematic Review Process

Strengths. This review appears to provide the most comprehensive synthesis of evidence related to the comparative effectiveness of surgical treatment of CDD and identifies important gaps in the comparative evidence for many of them. Important strengths of this review include the use of a “best evidence” approach, where we focused our efforts on studies with least risk of bias, particularly RCTs when available and supplemented with nonrandomized studies that adjusted for potential prognostic variables where appropriate. We avoided use of nonrandomized studies that did not adjust for potential confounding (e.g., propensity score matching, statistical control for confounding variables) as the conclusion from such studies may differ from RCT evidence and are more likely to suffer from various important biases (see below). Another strength is our focus on outcomes of primary importance to patients including pain, function, and quality of life as improved patient outcomes may lead to higher quality patient care, as well as patient satisfaction with care. Additionally, interpretation of clinically important differences in mean change for continuous variables is challenging. A strength of our review is our categorization of the magnitude of effects for function and pain outcomes using the system described in our previous reviews to facilitate interpretation of results across trials and interventions by providing a level of consistency and objective benchmarks for comparison. We also added two Contextual Questions (on the natural history of untreated spinal cord compression and on the prevalence of CDD with spinal cord compression in asymptomatic patients) to provide context for this review.

Limitations. For many Key Questions, quantitative synthesis of evidence was not possible due to the poor quality of evidence available and lack of comparative evidence for some Key Questions. For some Key Questions evidence was limited to one study per comparison, making it difficult to draw conclusions about any specific treatment. While we did include nonrandomized studies of interventions (NRSIs) that made comparisons of interest, results from such studies should be interpreted cautiously. Limitations of these studies generally led to determination of insufficient evidence for many outcomes. Confounding by indication, lack of adequate control for confounding on important prognostic factors, as well as failure to adequately account for selection of patients and loss to followup in NRSIs were common methodologic concerns. For subjective patient-reported outcomes such as pain, NRSI results may be misleading due to the subjective nature of pain and the impact of nonspecific effects related to patient expectations regarding treatment and attention received. Analysis of data from large administrative claims-based databases present additional methodological challenges. Coding related to conditions, procedures and outcomes in such databases is focused on optimizing billing and there is a potential for misclassification of exposures and outcomes. Such databases are unable to account for some potential confounders or for factors that may impact decision-making regarding the appropriateness of a given procedure (e.g., use of an anterior versus posterior procedure). The large sample sizes available for administrative data may facilitate evaluation of rare outcomes and may demonstrate statistical significance when results may be of unclear clinical importance.

Other limitations of our review include the following:

1. lack of RCT data for many comparisons and small sample sizes in most trials that precluded analyses on differential effectiveness and harms of interventions based on patient demographics, social determinants of health, severity of radiculopathy or myelopathy, number of vertebral levels involved, and other factors;
2. poor reporting of adverse events in many studies and heterogeneity in what harms and adverse events were described;
3. studies reporting vertebral levels affected (e.g., number of levels with pseudarthrosis, subsidence, needing reoperation) while not reporting the number of individuals experiencing a specific adverse event such as pseudarthrosis, thereby limiting the ability to use such studies in a pooled analysis in conjunction with studies reporting results in people rather than vertebral levels;
4. heterogeneity in research design, interventions, and reported outcomes for several Key Questions that limit ability to draw conclusions on effectiveness across studies;
5. in most cases we were not able to assess for publication bias using graphical or statistical methods to evaluate any potential impact of small sample sizes due to insufficient number of studies per comparison; and
6. limiting the evidence to English-language publications is a potential limitation, however we did not identify large numbers of non-English-language articles in our review of bibliographies.

Applicability

According to a NIS trend study of patients who underwent cervical fusion in 2013 for cervical spondylotic myelopathy (N=8181), the average patient was 60.6 years, slightly more likely to be male (54.3%), White (71.5%), with a CCI ≤ 2 (65.7%), have Medicare (44.6%) or private insurance (39.6%), and live in the South (43.8%). In the absence of more recent data, this represents a “best guess” at defining the typical patient seen in clinical practice today. There were similarities and differences between the typical study participant in our review and the typical patient as described above.

Reasons for greater applicability of this body of evidence to clinical practice include: (1) many studies required enrolled study participants to have failed several weeks or months of conservative therapies, which is considered a valid approach to the management of mild degenerative cervical myelopathy (as is an operative approach); (2) studies enrolled a balance of males and females; (3) most studies did not limit the upper age of enrollment and included individuals in their 60s or 70s (although the mean age of participants in most studies was in the 40s and 50s); and (4) studies often enrolled patients with a combination of radiculopathy and myelopathy, likely reflecting the condition of many US patients. Additionally, approximately 45 percent of studies included in this review were conducted in the United States.

Reasons for lower applicability to clinical practice include the exclusion of participants with a variety of common health conditions such as inflammatory arthritis, obesity, and diabetes. The risk of CDD increases with age and so do many other health conditions and comorbidities. For example, a large proportion of the US population is overweight or obese and an increasing proportion have diabetes. Excluding these populations from surgical intervention studies, because postoperative improvement may be reduced, decreases the applicability of study findings to many US patients needing operative management of their CDD. Additionally, few studies reported race or ethnicity. While those that did tended to enroll white participants, it is unclear how differences in access in populations of color may impact results.

Future Research

While it may not always be feasible to perform RCTs for surgical procedures, well-designed prospective comparative NRSIs with protocols using methods for patient selection and treatment allocation that mitigate possible selection bias and imbalances in prognostic factors and that follow protocols established a priori for comparable evaluation, measurement and treatment of groups would provide a valuable contribution to the evidence base. In order to evaluate the differential impact of patient characteristics and other factors, adequately powered RCTs are needed. Additionally, more explicit evaluation of procedure-specific (or device-specific) harms and adverse events is needed in future studies; ideally such studies would be powered to detect rare events. Future studies should also report the proportion of patients who experience a clinically important improvement in pain or function. This would provide valuable insight to complement data on average changes in continuous measures of pain, function, and quality of life for which there is difficulty describing clinically important effects. Studies should also estimate the minimally important between-group differences for included outcomes to facilitate interpretation of study findings.