AO Spine Clinical Practice Recommendations for Mild Degenerative Cervical Myelopathy: Contemporary Concepts

Abstract

Study Design

Literature review with critical appraisal and clinical recommendations.

Objective

To highlight contemporary concepts involving the management of mild degenerative cervical myelopathy (DCM) and facilitate integration into clinical practice.

Methods

Four articles relating to mild DCM were selected and critically appraised. Clinical practice recommendations were developed and evaluated using the GRADE framework and expert opinion.

Results

Based on low evidence in the Tetreault et al paper, it is conditionally recommended that the severity of DCM should be based on the mJOA score with the following thresholds: mild (15–17), moderate (12–14), and severe (≤11). No clinical recommendation, based on very low level of evidence for the Al-Shawwa et al article, on the routine use of MRI parameters to predict neurological deterioration or guide decisions on timing of surgical intervention in mild DCM. With low quality evidence presented by Badhiwala et al, there is a conditional recommendation for surgery in mild DCM patients with high levels of baseline pain and Neck Disability Index scores. A conditional recommendation, based on very low quality of evidence in the Ozaki et al paper, is that patients should be counseled that upper extremity sensory deficits frequently persist after decompression and that angular-edged spinal cord deformity on axial MRI should be considered a risk factor for residual postoperative upper extremity sensory impairment.

Conclusions

The management of mild DCM is an evolving field. Using established nomenclature and understanding the nuances of clinical and radiographic presentation may allow for more appropriate counselling and surgical candidate selection.

Introduction

Cervical myelopathy is one of the more common issues encountered by spine surgeons. ¹ In 2015, the term Degenerative Cervical Myelopathy (DCM) was introduced to encompass the various degenerative disorders that result in cervical myelopathy.^1,2 Davies et al published a unifying term and definition of the disease in 2024, as part of the AO Spine RECODE-DCM project, further strengthening the term DCM for use in patient care and research, defining it “as a progressive spinal cord injury caused by narrowing of the cervical spinal canal.” ³ The RECODE-DCM project further defined the clinical signs of DCM and the importance of awareness of what DCM is and translating it to timely treatment.^4-6

Disease states have various levels of severity that affect the treatment strategy. DCM is no exception. The clinical practice guidelines for the management of patients with DCM were published in 2017. ⁷ In these guidelines, there were strong recommendations regarding surgical intervention, based on moderate quality of evidence, for severe and moderate DCM. The recommendation regarding mild DCM, however, was considered weak and based on very low to low quality of evidence. Given the inherent challenges associated with surgical decision making around mild DCM, it was felt that further exploration of this topic was warranted.

The purpose of this article is to provide a concise update of the literature regarding the management of mild DCM. The topic of mild DCM, article selection, and comprehensive review was made by the AO Spine Knowledge Forum Spinal Cord Injury, representing an international group of clinicians and researchers. The Grading of Recommendation, Assessment, Development, and Evaluation (GRADE) framework, along with expert opinion, was utilized to consider the quality of the evidence and the strength of the recommendation. ⁸

Article 1: Tetreault L, Kopjar B, Nouri A, et al. The modified Japanese Orthopaedic Association scale: establishing criteria for mild, moderate and severe impairment in patients with degenerative cervical myelopathy. Eur Spine J. Jan 2017; 26(1):78-84. doi:10.1007/s00586-016-4660-8 ⁹

Clinical Rationale

Patients with DCM present with a broad spectrum of neurological symptoms and signs. Historically, definitions of “mild”, “moderate”, and “severe” DCM have been inconsistent, limiting comparability between studies and introduced ambiguity into clinical decision-making when using these terms.

The modified Japanese Orthopaedic Association (mJOA) score is an 18-point scale used to assess neurological impairment in patients with DCM. The version used in this study was the 18-point mJOA, based on the Benzel modification. ¹⁰ A score of 18 indicates normal neurological function, whereas lower scores reflect increasing disability and functional impairment. Previous studies have proposed mJOA cut-offs to classify DCM severity. ¹¹ However, these cut-offs were largely consensus-based rather than empirically derived and varied across studies, underscoring the need for validated, data-driven mJOA thresholds to standardize DCM severity classification.

Study Summary

This study analysed 740 patients with clinically and radiologically confirmed DCM that were enrolled in AO Spine North America and International prospective observational studies between 2005 and 2011. The investigators aimed at determining evidence-based cut-offs for mild, moderate and severe DCM. The 18-point mJOA was used, with a mean score of 12.50±2.74 (range 3-17 points). Receiver operating curve (ROC) analysis demonstrated good to near excellent predictive performance (AUC 0.81 for mild vs. moderate; 0.87 for moderate vs. severe DCM). The authors identified mJOA scores of 15–17 as mild, 12–14 as moderate, and ≤11 as severe, using the Nurick score as an external anchor to optimize sensitivity and specificity. Furthermore, functional impairment, disability, quality of life, and number of neurological signs and symptoms aligned with DCM severity, supporting the validity of the identified cut-offs. The study concluded that these mJOA-based thresholds provide a standardized and clinically relevant framework for classifying DCM severity, with direct implications for patient counselling, treatment planning, outcome interpretation, and guideline development.

Methodological Review

A key methodological strength is the anchor-based approach to severity classification. The Nurick score served as an external clinical anchor, and ROC curve analysis was applied to identify mJOA cut-offs optimizing sensitivity and specificity between severity categories. Construct validity was assessed through comparisons of functional performance, disability, quality of life, and neurological burden across severity groups.

Several limitations warrant consideration. The Nurick score is predominantly weighted toward gait and lower-extremity function. Consequently, using Nurick as the external anchor may have increased the influence of ambulatory impairment on the derived mJOA severity thresholds and may not fully capture upper-extremity-predominant manifestations of DCM. Patients with very mild neurological impairment were underrepresented, potentially limiting generalizability at the mild end of the disease spectrum. Clinicians should also recognize the limitations of the mJOA, including reduced sensitivity to subtle neurological impairment and the absence of patient-reported outcome components.^12,13 In addition, the mJOA has limited ability to capture pain and its impact on function and quality of life. Disease severity should therefore be interpreted in conjunction with clinical examination, patient-reported symptoms, functional status, and imaging findings.

Nonetheless, the large sample size, prospective data collection, and consistent associations across multiple outcome measures support the internal validity of the proposed severity thresholds.

Quality of Evidence: Low

Recommendation for Integrating Into your Clinical Practice

We suggest the use of the mJOA score to define disease severity in patients with DCM, using the following thresholds: mild (15–17), moderate (12–14), and severe (≤11). (Conditional recommendation, low quality evidence).

These data-driven severity categories provide a standardized and clinically meaningful framework for patient assessment, longitudinal monitoring, outcome interpretation, and comparison across studies. Adoption of these thresholds supports shared decision-making and aligns with prior AO Spine Clinical Practice Guidelines for the management of DCM. ⁷

Article 2: Al-Shawwa A, Craig M, Ost K, et al. Focal compression of the cervical spinal cord alone does not indicate high risk of neurological deterioration in patients with a diagnosis of mild degenerative cervical myelopathy. J Neurol Sci. Jun 15, 2024; 461:123042. https://doi.org/10.1016/j.jns.2024.123042 ¹⁴

Clinical Rationale

This observational study by Al-Shawwa et al evaluated whether patterns of cervical spinal cord compression and anterior–posterior spinal canal diameter (SCD) predict neurological deterioration in patients managed non-operatively for mild DCM. The study addresses an important clinical problem: identifying patients with mild DCM who may be at higher risk of neurological deterioration and therefore benefit from earlier surgical intervention. This question is particularly relevant given the clinical equipoise surrounding the management of mild DCM, where both operative and non-operative approaches are commonly considered. Furthermore, there are currently no imaging or clinical features that can reliably predict neurological progression in DCM and help clinicians determine which patients should undergo early surgical intervention and which can be safely monitored with conservative management.

Study Summary

In this longitudinal study non-operative patients with mild DCM, the authors investigated whether focal compression of the cervical spinal cord and SCD can reliably predict neurological deterioration at six months. Using manual MRI measurements of SCD and patterns of cord compression, the authors applied machine learning clustering techniques to identify patient subgroups and their associated risk profiles.

Overall, the findings indicate that focal cord compression alone is not a reliable predictor of neurological deterioration in mild DCM. The predictive value of pattern of compression is less relevant when the spinal canal is either very narrow (≤14.5 mm) or very wide (>15.75 mm). Importantly, the study found that severe spinal cord compression at two-disc levels, even when occurring in the context of an otherwise normal SCD, was associated with a higher likelihood of deterioration. In contrast, patients with normal SCD and no severe multi-level compression tended to have a lower risk of neurologic worsening, even if some cord compression was present.

Using dimensionality reduction and k-means clustering, the authors identified five distinct patient groups defined by combinations of SCD ranges and cord compression patterns, each with differing risks of neurological decline. These results suggest that the interaction between spinal canal morphology and compression pattern, rather than focal compression alone, is more accurate for risk stratification in mild DCM.

Methodological Review

From a GRADE perspective, the baseline level of evidence for this study is low given that it is an observational cohort study. This was further downgraded to very low due to risk of bias and imprecision. All patients diagnosed with DCM were enrolled in the ongoing Canadian Spine Outcomes Research Network prospective observational study, supporting systematic patient identification and reducing selective enrollment. A strength of this study is that patients who underwent surgery were excluded, leaving a cohort composed predominantly of individuals with mild myelopathy. While this may limit generalizability, the resulting cohort represents the population in whom prognostic stratification is most clinically relevant, as treatment decisions in mild DCM are often uncertain. Confounding bias is also an important consideration, as neurological deterioration may be influenced by multiple clinical and imaging factors beyond focal spinal cord compression or cervical SCD. At present, there is limited evidence regarding which specific variables should be controlled for in prognostic analyses of DCM, making it difficult to fully account for potential confounders. Furthermore, the mJOA score may not be sufficiently sensitive to detect subtle but clinically meaningful changes in neurological function, potentially leading to underestimation of neurological deterioration.

It is unclear whether the assessors of the imaging parameters were blinded to clinical status or outcomes, potentially introducing measurement bias. Additionally, the study included more MRIs than patients, indicating that some individuals underwent multiple scans. It is uncertain which scans were used for analysis or whether temporal progression of imaging findings may have been more relevant to neurological deterioration. Nevertheless, the study used well-defined, clinically practical imaging protocols with good to very good inter-rater reliability. The authors also acknowledged that their analysis focused on absolute SCD rather than relative measures (e.g. cord–canal ratio) that may better reflect predisposition to spinal injury.

Imprecision may represent a substantial limitation. Although there was a relatively balanced distribution of outcomes (35 patients deteriorated and 45 did not), the overall sample size was modest, which is particularly relevant given the use of machine learning approaches. Small datasets in machine learning analyses can produce patterns driven by random noise rather than true underlying structure, resulting in unstable cluster assignments and limited reproducibility. This concern is particularly relevant given the substantial overlap between the likelihood clusters, raising questions about the clinical separability of the identified groups. Moreover, the use of unsupervised machine learning techniques in a relatively small dataset raises concern about overfitting and may limit the reproducibility and generalizability of the patient subgroups.

Quality of Evidence: Very Low

Recommendation for Integrating Into your Clinical Practice

A recommendation cannot be made at this time for the routine use of these MRI parameters to predict neurological deterioration or guide decisions on timing of surgical intervention in DCM. Prospective validation in larger datasets is required before implementation in practice.

Article 3: Badhiwala JH, Hachem LD, Merali Z, et al. Predicting Outcomes After Surgical Decompression for Mild Degenerative Cervical Myelopathy: Moving Beyond the mJOA to Identify Surgical Candidates. Neurosurgery. 2020 Apr 1; 86(4):565-573. https://doi.org/10.1093/neuros/nyz160. PMID: 31225604. ²⁴

Clinical Rationale

Uncertainty remains around the role of surgical management for mild DCM. ¹⁵ While some argue for surgery in this group as a preventative intervention, others argue that a wait-and-see approach is best-practice. ⁷ The lack of data on the surgical outcomes for mild DCM has led to uncertainty around best practice management.^16,17 This study aimed to help inform clinical decision making by identifying predictors of recovery after surgery among people with mild DCM undergoing surgical decompression.

Study Summary

In this observational study, 185 patients with mild DCM, identified by a score of 15 to 17 on the mJOA, who had decompressive surgery with or without fusion were included. Patients included were part of the broader prospective, multicenter, AO Spine CSM North America or AO Spine CSM International cohort study. One-hundred and fifty-six (84%) patients had one-year follow-up data. For patients with mild DCM, surgery may lead to improved symptoms at 1 year. For those mild DCM patients with higher baseline symptoms, they may report greater improvements in their bodily pain and NDI scores, but not mJOA or most of the SF-36 components.

Methodological Review

This was a post hoc subgroup analysis from the AO Spine International and AO Spine North America CSM studies, which recruited patients from multiple sites in North America, Latin America, Europe, and Asia. This provides generalizability to the results, which is a strength of the study.

For the primary analysis, a k-means clustering approach was used to generate two separate groups from the included patients. This aimed to identify if specific clustering of outcomes could predict changes in quality of life at one-year post-surgery (compared to baseline). The result was a cluster 1 group with 79 included patients, and a cluster 2 group with 106 included patients. At baseline, no significant difference existed in features between clusters for age, symptom duration, or comorbidities; but significant differences existed on neck pain (Cluster 1: 32% vs Cluster 2: 11%), gait difficulty (Cluster 1: 57% vs Cluster 2: 40%), and motor weakness (Cluster 1: 75% vs Cluster 2: 60%). At one-year follow-up, the more disabled cluster 1 group had significant improvements compared with the less disabled cluster 2 group for the Neck Disability Index (NDI), and three components of the SF-36 (physical functioning, role physical and bodily pain), but not the mJOA or the other components of the SF-36.

A limitation of the study was that the patients were not randomized (they all underwent surgery), and the surgical approach and procedure details were at the surgeon’s discretion. The study was also a secondary analysis of an existing dataset, which can introduce bias. The authors minimized this with clear and transparent methodology being provided in the study. Additional limitations include that the follow-up was only to one year, which is insufficient to determine how stable the changes were.

Quality of Evidence: Low.

Recommendation for Integrating Into your Clinical Practice

We suggest that surgery be offered to carefully selected patients with mild DCM when they report high levels of baseline pain and NDI scores. (Conditional recommendation, low quality evidence) An understanding of what symptoms tend to improve in post-operative mild DCM patients allows for better selection of surgical candidates and pre-operative counselling.

Article 4: Ozaki M, Nagoshi N, Yamane J, et al. Symptom-specific Analysis of Surgical Outcomes and Predictors of Residual Symptoms in Patients with Mild Degenerative Cervical Myelopathy: An Analysis of Cases with Severe Cord Compression or Progressive Symptoms. Spine (Phila Pa 1976). Jun 9, 2025; https://doi.org/10.1097/brs.0000000000005423 ¹⁸

Clinical Rationale

Previous studies investigating surgical outcomes in mild DCM have relied primarily on total JOA or mJOA scores to evaluate treatment effects, without detailed assessment of symptom-specific changes across individual scoring domains. This is a significant limitation because mild DCM patients may exhibit a ceiling effect in composite scores, obscuring meaningful differences in recovery across specific functional domains. Understanding which symptoms are most likely to persist after surgery would enable more informed preoperative counseling and patient-centered decision-making. This study sought to fill that gap by performing a domain-level analysis of the JOA score and identifying predictors of residual symptoms following surgical decompression in mild DCM.

Study Summary

This was a retrospective multicenter study conducted across three institutions in Japan. From 679 consecutive patients who underwent surgical decompression with or without fusion for DCM between 2011 and 2021, 104 patients (15.3%) were classified as having mild DCM, defined as a JOA score of 14.5–16.5 (equivalent to mJOA 15–17 using an established conversion formula). ¹⁹ The mean age was 61.9 ± 11.8 years, and 80.8% were male. The posterior approach was used in 88.5% of patients, with laminoplasty being the most common procedure (67.3%). Notably, the mean maximum canal compromise was 49.7% and the mean maximum spinal cord compression was 37.2%, indicating substantial structural compression despite mild clinical presentation. The predominant trigger for surgery was symptom progression (94.2%).

Analysis of preoperative and 2-year postoperative JOA scores showed statistically significant improvements in finger motor function, upper extremity sensory function, and total JOA score. The average recovery rate was 49.3%. VAS scores for neck pain, upper extremity pain, and lower extremity pain also improved significantly.

The key finding was in the symptom-specific analysis. Upper-extremity sensory impairment was the most common preoperative symptom, present in 88.4% of patients, followed by finger motor dysfunction (52.8%). Among patients with preoperative upper extremity sensory deficits, 60.9% had persistent impairment at 2-year follow-up, a rate significantly higher than all other JOA domains. Multivariable logistic regression identified angular-edged spinal cord deformity on axial MRI as the sole independent predictor of persistent upper-extremity sensory impairment (OR: 4.264, 95% CI: 1.312–13.854, P = 0.016). These findings suggest that structural severity on imaging may not directly correlate with symptom reversibility, particularly for sensory deficits.

Methodological Review

This study addresses important and underexplored questions. Which specific symptoms in mild DCM are most resistant to surgical intervention, and what imaging features predict residual deficits?

The inclusion of both JOA sub scores and VAS pain measures provides a multidimensional assessment of outcomes. The evaluation of angular-edged deformity on axial MRI, with demonstrated good-to-excellent intra- and interobserver reliability, introduces a novel, clinically actionable imaging biomarker. The statistical approach was appropriate, with univariable screening followed by multivariable logistic regression. The surgical approach in this cohort adhered to institutional algorithms that shaped treatment characteristics and, by extension, the generalizability of outcomes.

Several important limitations must be considered. This is a retrospective study without a non-surgical control group. The sample size is relatively small (n = 104), which limits statistical power for the multivariable regression and may explain the borderline significance of intramedullary signal change (P = 0.055). The JOA score is an investigator-rated measure, which introduces potential interviewer bias. The cohort is predominantly Japanese, with a high proportion of posterior decompressions (88.5%) and laminoplasties (67.3%), which may limit the generalizability of findings. The study used the JOA scoring system rather than the more internationally recognized mJOA, potentially limiting comparability with the broader DCM literature. Finally, the definition of “symptom persistence” was binary (any residual score deterioration vs. full recovery), which does not capture partial improvement. This limitation may lead to an oversimplification of postoperative outcomes, particularly in mild DCM where partial improvements are clinically meaningful.

While the study provides novel and hypothesis-generating findings regarding symptom-specific outcomes and the role of spinal cord morphology, the retrospective design, small sample size, absence of a control group, and limited generalizability outside of the Japanese surgical context all contribute to considerable uncertainty in the effect estimate.

Quality of Evidence: Very Low

Recommendation for Integrating Into your Clinical Practice

We suggest that when counseling patients with mild DCM who are being considered for surgery because of progressive symptoms, clinicians communicate that upper extremity sensory deficits frequently persist after decompression. (Conditional recommendation, very low-quality evidence) Furthermore, we suggest angular-edged spinal cord deformity on axial MRI be considered a risk factor for residual postoperative upper extremity sensory impairment. (Conditional recommendation, very low-quality evidence).

This study provides clinically relevant insights to inform preoperative counseling for patients with mild DCM. Additionally, this study reinforces that mild DCM patients who come to surgery typically do so because of symptom progression (94.2%) and often have substantial spinal cord compression on imaging. This supports the current consensus that observation with structured follow-up is appropriate for stable mild DCM, whereas surgery should be considered when symptoms progress, even if composite scores remain in the “mild” range.

Discussion

This paper highlights 4 articles focusing on mild DCM and evaluating them using the GRADE methodology. ⁸ These studies illustrate the key challenges in the management of mild DCM: defining disease severity, understanding the risk of progression, identifying appropriate surgical candidates and counselling patients regarding postoperative expectations.

Article 1 was selected to provide the background and rationale to clearly define mild DCM. To accurately discuss a disease, there needs to be commonality in the terms used and how they are defined. Based on the results, there is a conditional recommendation that the severity of DCM should be classified using the mJOA score with the following thresholds: mild (15–17), moderate (12–14), and severe (≤11). While the mJOA has known limitations, there are ongoing efforts to better incorporate clinical signs and patient symptoms in defining myelopathy and its severity.^20,21

Article 2 was selected to address the natural history of mild DCM and the use of MRI to help determine those patients who would benefit from surgical intervention from the undefined risk of neurological decline. The study was probably underpowered and very low-quality evidence. No MRI findings were identified as prognostic factors and thus no recommendation could be made for the routine use of the MRI parameters to predict neurological deterioration or direct the timing of surgery.

Article 3 was selected to concentrate on making surgical treatment decisions in patients with mild DCM and highlights some of the limitations associated with the existing grading systems for early or less severe symptoms; methodological review was low quality evidence. A conditional recommendation for surgery in patients with mild DCM when they report high levels of baseline pain and NDI scores is made. For patients with mild DCM, surgery is likely to improve their NDI and pain symptoms at 1 year. Of note, it is increasingly recognized that neck and arm pain are important determinants of quality of life in patients with DCM and pain is an important determinant in surgical decision making for patients.^22,23 Furthermore, the mJOA has limitations around its ceiling effect and ability to detect symptom changes among people with mild DCM.

Article 4 was selected to focus on surgical outcomes for patients with mild DCM, looking at specific symptoms. The study was very low level of evidence. Surgery for DCM is done not only to prevent worsening, but also to improve the patient’s presenting complaints. Understanding the limitations of surgery is an important aspect of the pre-operative informed consent process. A conditional recommendation is made that patients should be counseled that upper extremity sensory deficits frequently persist after decompression, and that angular-edged spinal cord deformity on axial MRI should be considered a risk factor for residual postoperative upper extremity sensory impairment.

The AO Spine Knowledge Forum SCI is a group of international experts in the field of DCM treatment and research. While not a comprehensive clinical practice guideline, this paper provides clinical recommendations using the GRADE system, that may be useful for spine surgeons in their clinical practice. ⁸

(Summary of articles selected for critical review and key recommendations using the GRADE system: Table 1).

Table 1.

The quality of evidence and key recommendation for each article using the GRADE methodology

Study	Quality of Evidence	Recommendation
Tetreault L, Kopjar B, Nouri A, et al. The modified Japanese Orthopaedic Association scale: establishing criteria for mild, moderate and severe impairment in patients with degenerative cervical myelopathy	Low	Conditional recommendation to use the mJOA score to define disease severity in patients with DCM, using the following thresholds: mild (15–17), moderate (12–14), and severe (≤11). Implementation : Into daily clinical practice Caveats: Further efforts are ongoing to continue to evolve the classification of disease severity in DCM.
Al-Shawwa A, Craig M, Ost K, et al. Focal compression of the cervical spinal cord alone does not indicate high risk of neurological deterioration in patients with a diagnosis of mild degenerative cervical myelopathy	Very Low	No recommendation for the routine use of the defined MRI parameters to predict neurological deterioration or guide decisions on timing of surgical intervention in DCM. Implementation: Further study required before recommendations can be made regarding its use. Caveats: Severe spinal cord compression at two-disc levels, even when occurring in the context of an otherwise normal spinal canal diameter, was associated with a higher likelihood of deterioration.
Badhiwala JH, Hachem LD, Merali Z, et al. Predicting Outcomes After Surgical Decompression for Mild Degenerative Cervical Myelopathy: Moving Beyond the mJOA to Identify Surgical Candidates. Neurosurgery	Low	Conditional recommendation that surgery be recommended for people with mild DCM when they report high levels of baseline pain and NDI scores Implementation : Consider incorporating into clinical practice. Caveats : The mJOA has limitations around its ceiling effect and ability to detect symptom changes among people with mild DCM.
Ozaki M, Nagoshi N, Yamane J, et al. Symptom-specific Analysis of Surgical Outcomes and Predictors of Residual Symptoms in Patients with Mild Degenerative Cervical Myelopathy: An Analysis of Cases with Severe Cord Compression or Progressive Symptoms	Very Low	Conditional recommendation that when counseling patients with mild DCM communicate that upper extremity sensory deficits frequently persist after decompression and that angular-edged spinal cord deformity on axial MRI should be considered a risk factor for residual postoperative upper extremity sensory impairment. Implementation : Consider incorporating into clinical practice. Caveats : The mJOA scoring system was not used in the study. The paper was not designed to record partial improvement.

ORCID iDs

Chris J. Neal https://orcid.org/0000-0002-5072-6454

Paul Arnold https://orcid.org/0000-0002-4622-7695

Mahmoud Fouad Ibrahim https://orcid.org/0000-0001-6353-3689

Konstantinos Margetis https://orcid.org/0000-0002-3715-8093

Charles André Carazzo https://orcid.org/0000-0001-6339-2214

Michael G. Fehlings https://orcid.org/0000-0002-5722-6364