Key risk factors and a prognostic model for vascular myelopathy

Full Text

BACKGROUND

Diagnosing vascular myelopathies (spinal cord ischemia, spinal cord infarction) can be challenging, given the absence of pathognomonic clinical and neuroimaging markers to reliably distinguish them from other spinal cord (SC) lesions, especially in the early stages. According to international research, magnetic resonance imaging (MRI) findings in SC infarction are consistent with demyelinating, inflammatory, or compressive myelopathies in 60%–70% of cases, delaying diagnosis by an average of 72–96 h [1–3]. Thus, the assessment and interpretation of vascular myelopathy risk factors is critical for the differential diagnosis of this condition. Identifying reliable predictors of SC vascular injuries will allow clinicians to suspect the disease’s ischemic origin before receiving imaging findings, which can have a considerable impact on prognosis [4, 5].

Furthermore, determining and analyzing risk factors will allow identifying groups of patients who need special attention and more thorough angiography and neuroimaging examinations. This will accelerate the diagnosis of SC infarction, reduce diagnostic errors, and ultimately improve clinical outcomes. Thus, an in-depth examination of vascular myelopathy factors and predictors is relevant in modern neurology.

Aim

The work aimed to systematize known and newly identified clinically relevant risk factors for spinal cord ischemia and to develop a prognostic model of vascular myelopathy.

METHODS

A single-center observational study with prospective and retrospective analysis was conducted. The study included patients with acute and subacute myelopathies. The main group included patients with vascular myelopathy confirmed by Zalewski criteria [6, 7]; the comparison group included patients with other myelopathies.

Inclusion criteria in the main group were as follows: male and female patients over 18 years; clinical symptoms of acute myelopathy; ischemic myelopathy confirmed by SC MRI findings.

The comparison group had comparable clinical criteria; however, it included patients with non-ischemic myelopathy confirmed by laboratory and imaging findings.

Exclusion criteria in both groups were as follows: traumatic SC injuries, neoplasms, history of stroke or transient ischemic attack, and incomplete medical records.

The study was conducted at the neurology clinic of Pavlov University.

A standard examination was performed in all patients, including clinical and neurological examination, comprehensive laboratory examination (coagulation, lipid, and carbohydrate profiles; inflammatory and endothelial dysfunction markers) with a liquor test, and spinal MRI with mandatory T1, T2, and STIR sequences.

A comparative analysis was performed between the main group (patients with confirmed vascular myelopathy) and the comparison group (patients with other acute or subacute myelopathies).

The clinical assessment followed a standard protocol, using the 2019 American Spinal Injury Association (ASIA) score [8], and included a detailed neurological assessment. The motor function was assessed by ten key muscle groups on a 0–5 scale. Light touch and pain sensitivity was assessed in 28 dermatomes on a 0–2 scale. Special attention was given to injury level, severity (from A = complete injury to E = normal), and pelvic organ function.

The following laboratory parameters of the hemostatic system were assessed: activated partial thromboplastin time, prothrombin time with international normalized ratio, and fibrinogen, D-dimer, antithrombin III, and tissue-type plasminogen activator levels. The lipid profile included total cholesterol, low-density lipoproteins, very low-density lipoproteins, high-density lipoproteins, lipoprotein (а), and atherogenic and pro-atherogenic indices. Carbohydrate metabolism was assessed by fasting blood glucose and glycated hemoglobin levels. Furthermore, serum total protein, creatinine, urea, vitamin B12, and folic acid levels were assessed. The liquor test included protein levels, cell count, and immunoglobulin G (IgG) synthesis type.

MRI was performed on GE Signa HD1,5Т following a standard protocol with sagittal and axial T1- and T2-weighted images and sagittal STIR sequence. Diffusion-weighted imaging with diffusion coefficient mapping was used in some patients. Arteriovenous malformations (AVMs) were detected using selective spinal angiography and/or contrast-enhanced spinal MR angiography. The location and extent of myelopathy foci, changes in MR signal, diffusion restriction, the condition of spinal arteries and veins, signs of SC compression, and concomitant degenerative changes in the spine were assessed. All MRI findings were interpreted by two independent neuroradiologists.

All findings were entered into a standardized electronic database.

Statistical analysis was performed using IBM SPSS Statistics 21. Categorical values were presented as absolute and relative frequencies. The chi-squared test (χ²) or the Fisher’s exact test for small samples was used for intergroup comparisons. The Shapiro–Wilk and Kolmogorov–Smirnov tests were used for normality testing of numerical values. Data were presented as median ± standard deviation (normal distribution) or median [25th percentile; 75th percentile] (non-normal distribution). The Student’s t-test or the Mann–Whitney U test was used for intergroup comparisons of quantitative variables. The associations between risk factors and vascular myelopathy were assessed using odds ratios (ORs) with 95% confidence intervals (CIs). Binary logistic regression adjusted for potential confounders was used for the multivariate analysis. ROC analysis was used to assess the prognostic value of the logistic regression model. The significance level was set as p < 0.05.

RESULTS

The study included 177 patients, who were divided into two groups. The main group (n = 77, 40 [51.9%] females) included patients with vascular myelopathy (SC infarction) confirmed by a comprehensive examination. The comparison group (n = 100, 55 [55.0%] females) included patients with other acute and subacute myelopathies. These included demyelinating diseases of the central nervous system (acute transverse myelitis [n = 29], neuromyelitis optica spectrum disorders [n = 18], multiple sclerosis [n = 1], acute disseminated encephalomyelitis [n = 1]), compressive myelopathy (n = 18), subacute combined SC degeneration (n = 17), and unspecified myelopathies (n = 16). The mean age was 53.1 ± 15.6 years.

In the main group, arterial and venous SC infarction was reported in 64 (83.1%) patients and 13 (16.9%) patients, respectively. Patients in the main group had no clinical or neuroimaging signs of intramedullary hemorrhage (hematomyelia). There were characteristic differences in the topography of lesions. Lumbar enlargement lesions were significantly more common in the main group (41.6% vs 10.0% in the comparison group; p < 0.0001), whereas cervical spine lesions were less common (29.9% vs 49.0%; p = 0.010). There were no significant intergroup differences in the incidence of thoracic spine lesions (28.5% vs 41.0%; p = 0.087).

A comprehensive laboratory and imaging examination with comorbidity assessment (diseases, conditions, and/or laboratory syndromes) identified the following groups of factors potentially associated with myelopathies: cardiovascular, metabolic, vertebrogenic, infectious and immunological, and iatrogenic (Fig. 1).

 

Fig. 1. Comorbidities detected in the study groups.

 

A significant association was observed between vascular myelopathy and cardiovascular disorders. In patients with confirmed SC infarction, dyslipidemia and advanced atherosclerosis were 3.4 times more common than in the comparison group: n = 58 (75.3%) vs n = 22 (22.0%); χ²_{df = 1} = 49.94, р < 0.0001. A comparable pattern was observed for hypertension (n = 57 (74.0%) vs n = 34 (34.0%); χ²_{df = 1} = 27.90, р < 0.0001) and aortic diseases (n = 39 (50.6%) vs n = 7 (7.0%); р < 0.0001). Aortic atherosclerosis was reported in 32 patients, abdominal aortic aneurysm in four patients, aortic dissection in two patients, and severe aortic stenosis in one patient.

Cardiac arrhythmias, despite the low prevalence (n = 12), were also significantly more common in the main group than in the comparison group (11.7% vs 3.0%; p = 0.033). Moreover, significant intergroup differences were observed for heart failure: n = 27 (35.1%) in the main group vs n = 16 (16.0%) in the comparison group; χ²_{df = 1} = 8.60, p = 0.005.

Of special interest are differences in the incidence of metabolic disorders. Diabetes mellitus, a well-known risk factor for microangiopathy, was reported in 22 (28.6%) patients in the main group and 5 (5.0%) patients in the comparison group (р < 0.0001). Furthermore, significant differences were observed for conditions with laboratory signs of hypercoagulation and thrombosis: n = 22 (26.0%) in the main group vs n = 2 (2.0%) in the comparison group (р < 0.0001).

SC vascular disorders are particularly interesting as a cause of venous myeloischemia. Spinal AVMs, including dural arteriovenous fistulas (dAVFs), were reported as a diagnostic clue in 16 (20.8%) patients in the main group and 3 (3.0%) patients in the comparison group (p = 0.0002).

An analysis of vertebrogenic factors showed that the incidence of disc herniation and other degenerative changes in the spine was comparable in the two groups: n = 41 (53.2%) in the main group vs n = 44 (44.0%) in the comparison group; χ²_{df = 1} = 1.49, p = 0.22. A history of uncomplicated spinal injury was more common in the main group (n = 10 [13.0%] vs n = 5 [5.0%] in the comparison group); however, the differences were not significant (p = 0.10).

There were no significant intergroup differences in immunological and infectious factors. The incidence of autoimmune diseases (autoimmune thyroiditis, systemic lupus erythematosus, ankylosing spondylitis) was 14.3% (n = 11, with 10 autoimmune thyroiditis cases) in the main group and 21.0% (n = 21) in the comparison group (p = 0.25). In 5 (6.5%) patients in the main group, myelopathy was preceded by novel coronavirus infection (COVID19); in the comparison group, there were 9 (9.0%) patients with COVID19 (p = 0.54). Comparable findings were obtained for chronic infections (chronic hepatitis B and С virus, HIV infection, syphilis): n = 4 (5.2%) in the main group vs n = 7 (7.0%) in the comparison group (p = 0.62).

The study included 17 (9.6%) patients with myelopathy that occurred during or within a day after a therapeutic intervention. There were 14 (18.2%) such patients in the main group (endovascular aortic repair: n = 5; aortic stenting: n = 3; coronary stenting: n = 3; diskectomy: n = 2; paravertebral block: n = 1). In the comparison group, there were 3 (3.0%) patients with post-vaccination myelopathy, with significant intergroup differences (р = 0.001).

Binary logistic regression was used to assess the associations between these factors and vascular myelopathy, as well as to predict the presence or absence of SC infarction based on specific factors. The regression analysis used data from all 177 patients.

The regression analysis included all assessed parameters (cardiovascular, metabolic, vertebrogenic, infectious and immunological, and iatrogenic) as vascular myelopathy predictors. The stepwise regression analysis (conditional inclusion) was completed in the fifth step. In the first step, Dyslipidemia/Atherosclerosis parameter was included in model 1; SC Venous Diseases was included in the second step, Aortic Diseases in the third step, Hypercoagulation/Thrombosis in the fourth step, and Uncomplicated Spinal Injury in the fifth step (Table 1).

 

Table 1. Primary analysis of the association between spinal cord infarction and identified predictors

Factors	Parameter B assessment	Standard error	χ²	р-value	OR EXP(B)	95% CI for EXP(B)
						lower	upper
Dyslipidemia/atherosclerosis	2.102	0.511	16.934	<0.0001	8.182	3.007	22.266
Aortic diseases	2.590	0.573	20.443	<0.0001	13.328	4.337	40.956
Hypercoagulation/thrombosis	3.656	0.913	16.051	<0.0001	38.721	6.473	231.623
SC venous diseases	3.586	0.814	19.415	<0.0001	36.091	7.322	177.886
Spinal injury	2.522	0.751	11.292	0.001	12.454	2.861	54.219
Constant	–2.928	0.458	40.918	<0.0001	0.053	–	–

 

Model 1 had a high prognostic value (χ² = 119.8, p < 0.0001, R² = 65.9%) with classification accuracy of 86.4% (sensitivity 80.5%, specificity 91.0%). The ROC curve for the model is presented in Fig. 2. The area under the curve (AUC) for the model was 0.924 (95% CI: 0.88–0.96), indicating excellent quality of regression model 1.

 

Fig. 2. ROC curve for binary logistic regression model 1 for spinal cord infarction.

 

The disadvantages of this model include the integration of significant associations, making it difficult to interpret the contribution of any specific parameter. For example, there were significant associations between Dyslipidemia/Atherosclerosis and Aortic Diseases parameters (р < 0.0001), as well as between Dyslipidemia/Atherosclerosis and Hypercoagulation/Thrombosis (р = 0.002). There were no significant associations between other parameters (p > 0.05). The multicollinearity of Dyslipidemia/Atherosclerosis parameter necessitated adjusting the model and removing it from the list of predictors.

Repeated stepwise regression analysis was completed in the fourth step. In the first step, Aortic Diseases parameter was included in model 2; Hypercoagulation/Thrombosis was included in the second step, SC Venous Diseases in the third step, and Uncomplicated Spinal Injury in the fourth step (Table 2).

 

Table 2. Repeated analysis of the association between spinal cord infarction and identified predictors

Factors	Parameter B assessment	Standard error	χ²	р-value	OR EXP(B)	95% CI for EXP(B)
						lower	upper
Aortic diseases	3.337	0.529	39.86	<0.0001	28.132	9.984	79.266
Hypercoagulation/thrombosis	3.595	0.833	18.621	<0.0001	36.419	7.115	186.421
SC venous diseases	3.064	0.737	17.282	<0.0001	21.414	5.05	90.802
Spinal injury	2.403	0.659	13.301	<0.0001	11.051	3.039	40.19
Constant	–2.04	0.32	40.564	<0.0001	0.13	–	–

 

Model 2 retained a high prognostic value (χ² = 100.9, p < 0.0001, R² = 58.3%) with classification accuracy of 85.3% (sensitivity 87.0%, specificity 84.0%). ROC analysis confirmed the model’s adequacy (AUC = 0.88, 95% CI: 0.83–0.94) (Fig. 3).

 

Fig. 3. ROC curve for binary logistic regression model 2 for spinal cord infarction.

 

Therefore, the proposed prognostic model 2 can be used for risk stratification at admission in patients with myelopathy, especially with atypical clinical symptoms. The model’s high specificity (84%–91%) makes it a valuable tool for differential diagnosis of non-vascular myelopathies.

DISCUSSION

This work identified a set of factors associated with vascular myelopathy, with macrovascular diseases (atherosclerosis, aortic diseases), hypercoagulation, and SC venous diseases being the major contributors. The multivariate analysis confirmed the independent contribution of these parameters in the prognostic model, with a high diagnostic accuracy (AUC = 0.88). Cardiac disorders (arrhythmia, heart failure) and diabetes mellitus also showed a significant association with vascular myelopathy, whereas there were no significant intergroup differences for vertebrogenic, autoimmune, and infectious factors.

Summary of primary results

The identified association between aortic diseases and vascular myelopathy (OR = 28.1) is consistent with traditional research on angioarchitecture and impaired SC vascularization. According to Skoromets et al. [9], the artery of Adamkiewicz provides up to 85% of the blood supply to the lower thoracic and lumbar regions. In 75% of cases, it originates from the thoracic aorta between segments Th8 and L1. Our findings confirm that atherosclerosis and aortic aneurysms increase the risk of SC ischemia by impairing perfusion in this sensitive area. This explains why lumbar enlargement lesions were significantly more common in our patients with vascular myelopathy (p < 0.0001).

The identified association between hypercoagulation and vascular myelopathy (OR = 36.4) is of particular interest in terms of modern research into thrombophilia and microcirculation diseases. Our findings are consistent with those of Khoueiry et al. and Kubota et al. [10, 11], who found that even subclinical hemostasis disorders can promote thrombosis of small spinal arteries. These findings highlight the need for an expanded hemostasiological examination (including D-dimer, antiphospholipid antibodies, antithrombin III, protein C, and protein S levels) in all patients with ischemic myelopathy, even in the absence of systemic symptoms of thrombophilia. Furthermore, the findings emphasize the prospects of preventive anticoagulant therapy in patients with multiple risk factors; further research is warranted.

SC venous diseases, including dAVFs, were another major predictor of vascular myelopathy, with rates seven times higher than in other myelopathies (p = 0.0002). This prevalence of venous SC ischemia is consistent with the findings of Vuong et al. [12] and Kiyosue et al. [13]. It requires special attention in clinical practice, given that venous disorders of spinal blood flow are frequently underestimated.

Cardiac risk factors were not included in the final model; however, they also deserve attention. The identified incidence of heart failure (35.1%, p = 0.005) and arrhythmia (11.7%, p = 0.033) in the main group confirms the pathogenetic relevance of these factors. Hypoperfusion in the watershed areas of the SC makes it particularly vulnerable to ischemia when cardiac output is reduced [14]. The effect of arrhythmias may be associated with both spinal artery embolization and hemodynamic abnormalities in paroxysmal conditions [15]. These findings highlight the need for an expanded cardiovascular examination in patients with unspecified myelopathy.

In contrast, there was no significant association between vascular myelopathy and vertebrogenic factors. Despite the high incidence of disc herniation in both groups (53.2% and 44.0%; p = 0.22), they had a minor role in vascular myelopathy. This is inconsistent with classical concepts of compressive myeloischemia. However, it aligns with modern neuroimaging findings, which show that even severe degenerative changes in the spine rarely cause a critical decrease in spinal blood flow [5, 16].

Notably, while uncomplicated spinal injuries had no significant impact, they were included in the final logistic regression model as an independent predictor. This apparent paradox can be explained by the fact that the multivariate analysis considered interactions between variables, and trauma as a trigger had the most pronounced effect in combination with other risk factors such as aortic diseases or hypercoagulation.

The absence of an association between vascular myelopathy and autoimmune (р = 0.25) or infectious (р = 0.54, р = 0.62) diseases was not unexpected. Nonetheless, these findings are relevant in clinical practice, because they allow distinguishing between vascular myelopathy and inflammatory SC lesions, in which autoimmune and infectious triggers play a critical role [3, 17].

Of special interest is the identified association between vascular myelopathy and therapeutic interventions in the aorta and its branches (18.2% in the main group and 3.0% in the comparison group; p = 0.001). These findings are consistent with the increasing reports of iatrogenic spinal ischemia [5, 9, 18]. In our work, SC infarction was the most prevalent in aortic surgeries (8 of 14 cases), causing both hypoperfusion and reperfusion injuries of the SC. The following can be recommended in patients with multiple risk factors during elective thoracic or abdominal aortic surgeries: preoperative MR angiography mapping of the artery of Adamkiewicz; intraoperative neurophysiological monitoring; improved perfusion pressure during long surgeries; and prolonged neurological monitoring in the postoperative period.

The findings highlight the need for a comprehensive assessment of vascular myelopathy risk factors and allow supplementing the existing diagnostic algorithm (MRI in standard sequences) with the following: a) angiography of the aorta and spinal arteries, including spinal MR angiography, if lumbosacral or other myelopathy is detected; b) screening for hypercoagulation, even in patients with no history of thrombosis; and c) expanded cardiovascular examination (Holter monitor, echocardiography) in patients with risk factors.

The limitations of this study include a single-center, partially retrospective design, which may affect the sample’s representativeness, and lack of long-term follow-up. Furthermore, the study did not take into account the drug therapy received by patients (e. g., antithrombotic or lipid-lowering therapy), which could have altered the assessed risks for vascular myelopathy.

CONCLUSION

This study confirms and expands the existing multifactorial concept of vascular myelopathy, highlighting the key pathogenetic role of macrovascular (especially aortic) diseases and hypercoagulation. The findings are consistent with recent international research, emphasizing the critical role of systemic vascular risk factors and the need for a multifaceted approach to diagnosis. However, uncertainties remain regarding the precise ischemia mechanisms in subclinical hypercoagulation, the best approaches for early detection of venous malformations, and strategies for preventing SC infarction during aortic surgery.

Our study addresses these issues by confirming the independent influence of each identified risk factor and proposing a clinically relevant predictive model with high diagnostic accuracy (AUC = 0.88). The identified association between iatrogenic factors and vascular myelopathy highlights the need to reconsider preoperative examinations in patients at risk. The findings can be used to develop new diagnostic and differential diagnostic algorithms, which is especially relevant for myelopathies with overlapping clinical symptoms. Further research is needed to validate the proposed model in prospective multicenter studies and assess the efficacy of personalized prevention and treatment strategies based on the identified predictors. Applying the findings in clinical practice will improve the early diagnosis and prognosis in patients with SC infarction.

ADDITIONAL INFO

Author contributions: G.V. Ponomarev: conceptualization, methodology, investigation, formal analysis, writing—original draft; A.V. Amelin, A.A. Skoromets: conceptualization, methodology, formal analysis, writing—review & editing. All authors have read and approved the final version of the manuscript prior to publication.

Funding sources: The study was not supported by any external sources of funding.

Competing interests: The authors declare that they have no competing interests.

Consent for publication: Written consent was obtained from the patients for publication of relevant medical information within the manuscript. All information provided is anonymized, photographs are not published.

Ethics approval: The study was approved by the local ethics committee of the Pavlov First Saint Petersburg State Medical University No. 268 of 26.12.2022.

Disclosure of potential conflicts of interest: The research materials are part of the application for invention No. 2025118338 dated 01.07.2025, filed with the Federal Institute of Industrial Property (FIPS, Russia).

About the authors

Grigory V. Ponomarev

Pavlov First Saint Petersburg State Medical University

Author for correspondence.
Email: grigoryponomarev@yandex.ru
ORCID iD: 0000-0002-6219-8855
SPIN-code: 1143-4227

MD, Cand. Sci. (Medicine)

Russian Federation, Saint Petersburg

Aleksandr V. Amelin

Pavlov First Saint Petersburg State Medical University

Email: grigoryponomarev@yandex.ru
ORCID iD: 0000-0001-6437-232X
SPIN-code: 2402-7452

MD, Dr. Sci. (Medicine), Professor

Russian Federation, Saint Petersburg

Aleksandr A. Skoromets

Pavlov First Saint Petersburg State Medical University

Email: grigoryponomarev@yandex.ru
ORCID iD: 0000-0002-5884-3110
SPIN-code: 6273-8033

MD, Dr. Sci. (Medicine), Professor, Academician of the RAS

Russian Federation, Saint Petersburg

References

Supplementary files