Introduction

Intracranial dural arterio-venous fistulas (iDAVFs) are rare malformations characterized by pathological anastomoses connecting arterial dural branches and dural sinuses, mostly the cavernous sinus and/or cortical veins. Arterial branches may arise from the external and internal carotid arteries and/or from the vertebrobasilar system. IDAVFs account for 10–15% of intracranial vascular malformations [1], representing about 6% of all supratentorial and 35% of all infratentorial vascular malformations [2]. They are generally acquired and associated with several predisposing factors such as history of craniotomy, head trauma, previous dural sinus infection or thrombosis, and genetic thrombosis predisposing mutations (heterozygous factor V Leiden and protein C/S deficiency) [1]. The mean age of diagnosis is between the fifth and the sixth decades, with a male-to-female ratio of 1.

Clinically, iDAVFs can present with both hemorrhagic and non-hemorrhagic symptoms, depending on the grade and anatomical localization. Hemorrhagic symptoms are typically characterized by lobar (or sublobar) hemorrhages, venous infarctions, and subdural hematomas; non-hemorrhagic symptoms are extremely variable, ranging from non-localizing signs such as intracranial hypertension with papilledema, headache, and nausea/vomiting to focal signs like seizures, cranial neuropathies, and pulsatile tinnitus. Chronic complications such as glaucoma, hydrocephalus, dementia, parkinsonism, and slowly progressive myelopathy are also possible [1].

IDAVFs are usually classified according to Borden-Shucart’s [3] or Cognard’s classifications [4], both strictly related to prognosis: the higher the grade, the worse the prognosis (see Table 1).

Table 1 Cognard classification
Full size table

Cognard type V fistulas (CVFs) display perimedullary venous drainage and are associated with myelopathy in 50% of cases [4, 5]. These entities are extremely rare; until 2016, only 54 cases of CVFs had been described [2]. In 2020 Hou et al. reported 73 patients with CVFs, 57 of which presented with either paraparesis or tetraparesis [6]. CVFs’ clinical presentation is highly variable, but the classical picture is that of a middle-aged man with ascending tetraparesis (62%), sphincter dysfunction (34%), bulbar symptoms (31%), and a sensory level typically developing over several months; nevertheless, up to 50% of cases can present with acute onset [7]. Small vessel thrombosis, infarct or hemorrhage, are believed to be responsible for rapid worsening or acute onsets [2].

The pathophysiology of myelopathy and brainstem engorgement is similar to that described for type I–IV fistulas, involving congestion and dilation of the venous system, but with the involvement of perimedullary veins instead of cortical veins [8]. However, as noted by Brunereau and colleagues, not all CVFs cause myelopathy [9]. Some authors speculate that in a subset of patients, a medullary-radicular vein connecting the spinal perimedullary venous network to the epidural venous system at the cervical level may prevent the establishment of spinal cord venous hypertension, while the absence of the communicating vein may predispose to engorgement of cervico-thoracic perimedullary veins, leading to medullary edema and rarely spinal infarct due to poor arterial supply [9]. Two other possible theories to explain spinal cord involvement in CVFs are arterial steal and direct compression of the spinal cord by enlarged veins, clot, or varicose vessels [2].

Due to their rarity, these entities are seldom suspected, resulting in a diagnostic delay up to many years (average time 220–343 days) [10]. Whether this delay affects patients’ life expectancy and the grade of residual disability is still a matter of debate. It is noteworthy that El Asri et al. postulated the absence of correlation between diagnostic delay and the clinical outcome in patients with paraparesis, quadriparesis, or bulbar dysfunction. They also found that 38% patients with CVFs died or did not improve significantly despite the treatment, whereas 26% of patients showed an improvement after the treatment but still had a moderate disability, highlighting that the outcome of CVFs can often be life changing. Nevertheless, treatment resulted in complete recovery or noticeable improvement (defined as persistence of only mild symptoms) in 36% of cases [10].

The objective of this article is to review the literature describing the clinical and radiological spectrum of CVF presentation, starting with a representative case, and to investigate whether there are any reliable clinical or radiologic parameters that could help clinicians in suspecting the diagnosis. The possibility of a misdiagnosis due to many “iDAVFs mimics” is another key point of our study; CVFs may cause clinical and radiological findings very similar to a variety of inflammatory, infectious, or vascular diseases (i.e., spinal dural fistulas) affecting the spinal cord. This virtual absence of pathognomonic signs makes the diagnosis extremely challenging and currently possible only in a few specialized centers.

Furthermore, we aim to assess whether specific clinical and radiological findings impact diagnostic delay and prognosis, and if there is a correlation between clinical and/or radiological signs and clinical outcomes.

A representative case

In December 2019, a 52-year-old man presented with subacute onset of severe neck pain, vertigo, nausea, and vomiting which completely resolved within a month. In April 2020, he reported progressive difficulty in walking with tripping, climbing stairs, and manipulating small objects. He sought medical attention on May 1st when he experienced acute urinary retention requiring hospitalization and catheterization. He denied any history of fever, insect bites, trauma, or recent vaccinations. His past medical history included hypertension, gastroesophageal reflux disease, previous exposure to asbestos, and pulmonary fibrosis. There was no consanguinity between his parents, and family history was negative for neurological diseases. An urgent brain CT scan revealed a hypodense lesion in the right cerebellar hemisphere and subsequent MRI of the spine showed a gadolinium-enhancing spinal lesion suggestive of myelitis extending from the medulla to C7/D1. Cerebrospinal fluid (CSF) analysis was inconclusive, and search for common pathogens in the CSF was negative. The patient was started on high doses of steroids with mild clinical improvement. At discharge, his neurological examination showed left gaze evoked nystagmus, mild central left facial paresis, hyperreflexic quadriparesis with ankle clonus, abolished pain, and temperature and proprioception below D10 along with urinary and bowel incontinence. Despite initial improvement, the patient relapsed in August 2020 with worsening in his upper limb strength, increased lower limb spasticity, and altered mentation. MRI revealed extension of the previously described lesion to the pons (see Fig. 1). CFS analysis was once again uninformative and the extensive search even for tropical microorganisms was inconclusive. Vasculitides, neuromyelitis optica spectrum disorders, and other autoimmune systemic diseases were ruled out. A total-body PET study with 18-FDG did not show any findings consistent with neoplasm, and no malignant cells were found in the CSF. After a multidisciplinary discussion, neuroradiologists carefully reviewed the spinal MRI and identified the presence of tortuous vessels behind the cervical spinal cord. Subsequent cerebral angiography (digital subtraction angiography, DSA) confirmed the presence of an arterio-venous fistula between the posterior meningeal artery and the straight sinus with drainage into the perimedullary venous plexus at cervical level (Cognard type V fistula, see Fig. 2). The patient underwent endovascular treatment, which resulted in almost complete obliteration of the fistula; he was discharged to a rehabilitation center 2 weeks later. One year after the embolization, the patient’s neurological examination remained unchanged and he continued to use a wheelchair. In October 2021, he underwent successful retreatment with the endovascular approach, resulting in complete obliteration of the fistula, with only slight improvement in upper limb strength noted after the procedure.

Fig. 1
figure 1

MRI at the diagnosis. A Axial FLAIR sequence shows hyperintensities in the pons and in the right cerebellar hemisphere (white arrows). B Axial T1-weight image obtained after gadolinium administration reveals enhancement of the same areas (white arrows). C Sagittal T2-weighted image shows intramedullary hyperintensity from the medulla oblongata to D2 vertebra level, with swollen cervical spinal cord (white arrows). D Gd-enhancement of the lesion shown in C (white arrows)

Full size image
Fig. 2
figure 2

Cerebral angiography, vertebral artery injections. A The A-V shunt at the fistula site is indicated by the white arrow (LL view). The arterial feeder is the posterior meningeal artery (PMA), arising from the vertebral artery. B Same as in (A) but with digitally subtracted images showing the feeding artery (PMA, a), the fistula site (white arrow), and the precociously enhanced straight sinus (v). Retrograde venous drainage route is indicated by the black arrow. C Parenchimal-phase acquired image showing backward venous drainage route (black arrows) toward the perimedullary venous system. D Late acquisition image showing venous blood direction (black arrows) reaching the perimedullary venous system at cervical level (*)

Full size image

Materials and methods

Literature search

We started identifying the published case reports and case series of patients having CVFs using a search strategy developed by three authors (ADG, ES, and CM) through an iterative process. We searched for published articles that mentioned iDAVF in title, abstract, and keywords using the following strategy: “Intracranial fistula AND spinal drainage,” “Intracranial fistula AND spinal cord,” “Intracranial fistula AND myelopathy,” and “Intracranial fistula AND myelitis,” since the inception of the database to March 2021. The language was restricted to English and Italian. The search was conducted independently by three experienced neurologists (ADG, ES, and CM) and was performed both (a) in PubMed electronic database and (b) through manual searches (i.e., reference lists of previously reported case reports/series and systematic reviews on this topic identified during the search).

Study selection and data extraction

Following the procedure detailed by El Asri and colleagues and by Kamio et al. [10, 11], we collected all case reports published from inception to March 2021, thus providing a greater sample size of patients with CVF. We included all articles which (a) described a case or a series of cases of CVFs and that (b) were written in English or Italian. We excluded (1) articles in which full text could not be obtained; (2) papers concerning the pediatric populations; (3) unrelated papers (i.e., spinal fistulas).

The screening process was conducted as follows: first, the three authors (ADG, ES, and CM) independently reviewed all abstracts and titles for eligibility: after a manual screening, full-text reports were obtained if a study was deemed eligible or where eligibility was unclear. Then, reports were finally examined for inclusion, with disagreement resolved through consensus by the three authors.

Regarding the data extraction, we adopted a standardized coding scheme to collect data referring to (1) age and sex of patients, (2) type of CVF onset (acute, progressive, or multiphasic; see below), (3) presence of predisposing factors, (4) symptoms at onset (motor, sensory, sphincteric disturbances, ataxia, brainstem symptoms, dizziness/nausea/vomiting, and others), (5) symptoms at diagnosis, (6) time interval to definite diagnosis, (7) presence of an initial misdiagnosis, (8) MRI findings at diagnosis, (9) CVF localization, (10) feeding artery, (11) type of treatment (surgery or endovascular), (12) outcome (outlined as good recovery/complete regression, moderate disability, severe disability/death), (13) presence of a relapse, (14) length of post-treatment follow-up, and (15) angiography outcome. The authors coded all available information reported in any part of the articles, including tables and figures. Under “brainstem signs” we included the following: dysphagia, dysphonia, dysarthria, respiratory failure, diplopia, gaze-evoked nystagmus, decreased gag reflex, cranial nerves palsies, and hiccups. “Acute” onset was defined as an abrupt onset over 1 day or two, “multiphasic” onset was defined as bouts of symptoms with complete or almost complete recovery between the episodes, and “progressive” onset was preferred when disturbs developed over time with no definite poussées. For clinical outcome assessment, we defined “good recovery” (GR) as the complete regression of symptoms, “moderate disability” (MD) either as the ability to walk with assistance or neurologic sequalae interfering with daily activities but not determining loss of independence, and “severe disability (SD)/death” either as the inability to walk or as neurologic sequelae severely interfering with daily activities or as death. Missing data were not imputed.

Statistical analysis

Data extracted from case series or case reports were then analyzed through descriptive statistics, including means, standard deviations, frequencies, and percentages.

Further, we tested several hypotheses on the association between socio-demographic, clinical, and neurological variables, and CVF (i) type of onset (acute, progressive, or multiphasic), (ii) outcomes (dichotomized as good outcome vs disability/death), (iii) time interval to diagnosis, and (iv) presence of specific symptoms at onset, through univariate and multivariate statistics. Differences between frequencies of specific categories were tested through chi square tests (with the significantly different categories identified through the adjusted standardized residuals >|2| [12]. Further, non-parametric correlations, Mann–Whitney U tests and Kruskal–Wallis ANOVAs were adopted to test for the association between dichotomous and continuous variables and for the presence of significant differences between groups on continuous independent variables, respectively.

All analyses were performed with SPSS 26 (IBM, 2019). All statistical tests were two tailed, and a p ≤ 0.05 was deemed statistically significant.

Results

Literature search and study inclusion

The literature search initially yielded 382 articles. After screening titles and abstracts, 245 articles were excluded due to unrelatedness or because they were written in languages other than Italian or English. Additionally, 80 duplicates were removed, and 16 records were identified through other sources such as manual searches among reference lists of previously published reviews. The PRISMA flow diagram, depicted in Fig. 3, provides a graphical representation of the screening process. In total, 72 studies, including 60 case reports and 12 case series, were included in the analysis, providing data on a total of 100 patients with CVF. The median year of publication for the 72 studies was 2006, with a range from 1988 to 2020.

Fig. 3
figure 3

Flow chart of the searching strategy

Full size image

Socio-demographic, clinical, and radiological features

Table 2 summarizes the clinical characteristics of the patients with CVF. Most of the patients were middle aged, with a mean age of 56.20 ± 14.07 years and a range of 18–79 years and the majority were males (72%, N = 72). Predisposing factors, such as past head trauma, were reported in only 20.4% of the articles. The CVF’s onset was mostly progressive (N = 63, 64.9%), while multiphasic (N = 21, 21.6%) and acute (N = 13, 13.4%) onsets were less commonly reported. A total of 47 patients (58%) received an initial misdiagnosis. More details are provided in Fig. 4.

Table 2 Sociodemographic and clinical characteristics of patients with DAVF (N = 100)
Full size table
Fig. 4
figure 4

Misdiagnosis rate. Note. NA, not available

Full size image

The median time from symptoms onset to diagnosis was 5 months (range: 0–48 months). Figure 5 provides a graphical depiction of the prevalence of symptoms at onset.

Fig. 5
figure 5

Symptoms at onset

Full size image

MRI findings at diagnosis included flow voids (81.6%, N = 71), T2 hyperintensities (80.5%, N = 70), and swelling (56.3%, N = 49). DWI abnormalities, thrombosis, and T2* effects were rare (2 cases each, 2.3%), and contrast enhancement assessment was performed in only 55.8% of cases (N = 29).

As for clinical outcomes after treatment, 57 patients experienced a moderate-to-severe disability or died (N = 57; 67.1%; moderate disability, N = 33, 41.3%; severe disability/death = 19, 23.8%), while 28 experienced a partial-to-full recovery (32.9%).

Patients underwent endovascular treatment (N = 45, 48.9%), open surgery (N = 28, 30.4%), or both (N = 20, 21.3%); a total of 10 patients experienced a relapse after the first treatment attempt (11.9%).

Tables 2, 3, 4, 5, and 6 report all the clinical and radiological variables hitherto described.

Table 3 Additional clinical and neurological characteristics of patients with DAVF (N = 100)
Full size table
Table 4 Symptoms at onset among patients with DAVF (N = 100)
Full size table
Table 5 Symptoms at diagnosis among patients with DAVF (N = 100)
Full size table
Table 6 Brain MRI findings at diagnosis among patients with DAVF (N = 100)
Full size table

Associations between sociodemographic, clinical and MRI variables, and CVF’s onset and outcomes

We initially investigated the association between age, gender, and outcome among the 100 patients with CVFs through non-parametric correlations and crosstabs, respectively. In both cases, results were not significant, suggesting that outcome was not related to the age (r = 0.065, p = 0.56) or gender (χ2 = 3.163, p = 0.075). We then tested for an association between an initial misdiagnosis and the disease’s outcome, but the chi square test was not significant (χ2 = 0.194, p = 0.66), suggesting that those who had an initial misdiagnosis had similar outcomes compared to those whose CVFs were diagnosed correctly at symptoms’ onset.

As for the association between diagnostic delay (in months) and type of onset, a non-parametric ANOVA (Kruskal–Wallis test (2) = 15.540, p < 0.001) evidenced that those with an acute onset had a significantly lower interval to diagnosis compared to those with a progressive (p < 0.001) or multiphasic one (p = 0.049). All other comparisons were not significant. Interestingly, the association between diagnostic delay and outcome was also significant (U = 432.000, z =  − 1.960, p = 0.050), with patients who experienced a disability or exited receiving their diagnosis months later compared to patients who experienced a good recovery.

As for the association between the presence of specific symptoms at onset (e.g., ataxia, sphincteric disturbances, motor or sensory ones) and diagnostic delay, those who experienced sensory symptoms at onset received their diagnosis of CVFs later than those who did not experience them (U = 749.000, z = 2.247, p = 0.025), while all other comparisons were not significant. As a follow-up analysis, and to better understand the unique contribution of sensory symptoms in explaining the diagnostic delay, we tested for the presence of significant differences in diagnostic delay between those who experienced only sensory symptoms at onset (N = 8) and those who experienced sensory symptoms together with other ones (N = 16). Though the Mann–Whitney U test is non-significant (U = 75.500, z = 1.308, p = 0.20), the between-groups effect size was medium (Hedge’s g = 0.65), suggesting that—with a larger sample size—this comparison would have reached the significance. As for the association between the presence of specific symptoms at onset and outcome, none of the chi square tests reached the significance.

Finally, we examined the association between spinal MRI findings at diagnosis, diagnostic delay, and outcome. In these analyses we focused exclusively on MRI swelling, T2 hyperintensities, flow voids, and contrast enhancement due to extremely low incidence of other MRI findings (DWI, T2* abnormalities, and thrombosis). Results showed that MRI findings were unrelated with both diagnostic delay and CVF outcome.

Detailed results, including frequencies, percentages, means, and standard deviations, separately for each group, are reported in the Supplementary Materials.

Discussion

The diagnosis of CVFs is challenging and often requires the expertise of highly specialized centers, leading to potential delays in diagnosis that can impact clinical outcomes and patients’ quality of life. Many patients, including the case discussed, receive a correct diagnosis only months or years after the onset of symptoms, when irreversible damage may have already occurred. One of the major challenges in diagnosing CVFs is that they are rarely considered in the initial diagnostic workup of myelopathies. Our CVF case was a starting point to conduct an analysis on the possible way to improve the outcome and to look for reliable clinical or radiological signs that could aid an earlier diagnosis.

Demographic features

Our analysis outlined that the mean age of onset was 56 years old, with most patients being males. These findings are in line with what has been described in a similar review conducted in 2013 [10], confirming CVFs as being a disease mainly affecting middle-aged patients, even though a few pediatric cases have been reported [7].

Clinical characteristic

The most prevalent type of onset was “progressive,” while the “acute” one was much less represented (13%) compared to the 25% reported in the literature [2], possibly because “multiphasic” onsets were considered “acute” in those other studies. The most common complaints at beginning were motor deficits (either paraparesis or quadriparesis) followed by sensory symptoms, sphincteric disturbances, dizziness, ataxia, and brainstem symptoms (Fig. 5). Interestingly, we found that only 9% of patients had brainstem signs, whereas El Asri and colleagues reported their presence in one-third of the patients [10]; this discrepancy may be due to the different definition of “brainstem signs” between the studies. It was also found that patients presenting with brainstem signs tended to have a shorter time to reach a correct diagnosis (see the “Diagnostic delay” section), which is consistent with current literature [11]. This may be because patients with brainstem signs are often mistaken for having a stroke and are promptly admitted to the emergency room. In cases where the onset is progressive and the pattern is that of an ascending myelopathy (which is the most common pattern), brainstem signs are less likely to appear early, and by the time they do, other symptoms may already be irreversible [2]. Another interesting finding is that patients presenting with only sensory symptoms tended to receive a correct diagnosis much later than those presenting with other symptoms. The most likely explanation is that sensory symptoms are common, easily missed during neurological examination, and their importance is often underestimated by clinicians and by patients themselves. Sensory symptoms are considered less disabling than motor symptoms, so patients may not consult a neurologist until motor symptoms occur, while neurologists may underestimate the subtle onset of sensory findings, often attributing them to radiculopathies or peripheral neuropathies.

Diagnostic delay

In this study the mean diagnostic delay was 5 months, a result slightly shorter than what had been previously reported (6–12 months) [7, 10]; this minimal difference with studies conducted years ago imply that very few progresses have been made in diagnosing CVFs during the last few decades. Interestingly, our patients with acute symptoms were more likely to be diagnosed correctly and sooner compared to those with progressive or multiphasic onsets. This could be because patients with acute symptoms are more likely to seek medical attention promptly, while those with progressive symptoms may delay seeking medical help for months, as stated in the “Clinical characteristics” section. This concept is of utmost importance since our analysis outlined that diagnostic delay has a significant impact on the outcome (see the “Outcome” section). As a matter of fact, patients experiencing the poorest prognoses (severe disability or death) had the longest time-to-correct diagnosis interval implying a relationship between these two variables. In other words, a longer diagnostic delay was often associated with a worse clinical outcome, suggesting that early diagnosis could not only lead to a reduction in mortality rate but also to a noticeable reduction of the residual disability. Although several studies have drawn the same conclusion in the past, our study managed to statistically support this hypothesis. In contrast, another study by Kamio et al. did not find a correlation between disease duration and prognosis, but did emphasize the importance of prompt and accurate diagnosis for improving symptoms and avoiding poorer outcomes (see the “Outcome” section) [11]. Of note, in the past some authors reported that even paraplegia can be reversible if the fistula is treated before the occurrence of ischemic and gliotic changes, pointing out the importance of early diagnosis and treatment [13, 14].

Misdiagnosis

In this setting, reaching the correct diagnosis in the shortest possible time and minimizing the misdiagnosis rate is pivotal. According to our numbers, more than half of the patients were initially misdiagnosed as having other diseases, including our own patient. This is a much more discouraging result than the previous 40.2% misdiagnosis rate reported by Kun Hou et al. in their review [6]. The most common reported misdiagnoses were spinal dural A-V fistulas [9], myelitis [15], tumors (mainly lymphoma [16] and glioma [17]), and strokes [13] (see Fig. 4). In one case stroke was suspected twice before the fistula was discovered [18], suggesting that CVF diagnosis is still challenging. Even if in terms of outcome, we did not find any statistically relevant difference between patients who received misdiagnosis and the ones who did not; misdiagnosis could potentially contribute to diagnostic delay, which in turn is associated with poorer outcomes.

It is important to notice that (1) mildly elevated CSF protein and absence of CSF pleocytosis (“albumino-cytological dissociation”) may occur in arteriovenous fistula and therefore should not necessarily be attributed to idiopathic transverse myelitis or Guillain-Barrè syndrome; (2) post steroid worsening should raise the suspicion about a non-inflammatory disease of the spinal cord, particularly a spinal or an intracranial fistula [19].

Imaging

While conventional angiography is still to be considered the gold standard for definite diagnosis of CVFs, MRI can strongly aid the diagnosis and dramatically shorten the time to diagnosis, especially when MRA sequences or contrast studies are carried out. Abnormal vascular flow voids, which are tortuous and dilated veins generally found on the dorsal or ventral surface of the spinal cord, were eventually found in 81.6% of patients, even when they were not reported initially [20, 21]. A high index of clinical suspicion is then required to carefully evaluate MRI images looking for flow voids so to reduce the interval to diagnosis and, accordingly, achieve a better outcome. Moreover, in the appropriate clinical context, flow voids help distinguishing CVFs from all other mimics (except spinal fistulas). Unfortunately, all other imaging features (i.e., T2/FLAIR hyperintensities and spinal cord swelling) are nonspecific. An interesting description was made by Copelan et al. who reported spinal edema as having a “a tigroid pattern” with geographic central medullary edema and sparing of the periphery as well as internal linear segments [20]. However, they did not include all cases of CVFs, making this differentiation based on tigroid appearance less suitable for generalization. Several other studies tried to find peculiar image findings (i.e., the “black butterfly sign” by Enokizono and colleagues [22]) but these remain isolated observations.

Outcome

Our analysis did not disclose any relationship between age, sex, and outcome, supporting the current knowledge about CVFs [10] and implying that the prognosis can be severe even in otherwise healthy young subjects. In our study sample, the percentages of moderate and poor recovery/death were 41.3% and 23.8%, respectively, while good recovery was only 32.9% which is consistent with the literature [10] and highlights that CVFs can still result in moderate/severe disability in two-thirds of cases. Moreover, there was no statistical relationship found between the presence of a specific subset of symptoms at onset and the outcome, suggesting that more compromised patients at onset do not necessarily have a worse prognosis. Similar findings have been reported in the literature, particularly regarding the lack of correlation between the severity of symptoms at onset and prognosis, except when signs of brainstem dysfunction are present, possibly due to the involvement of respiratory and cardiovascular centers in the brainstem [10, 11]. Unfortunately, no highly suggestive pattern of CVF symptoms that could shorten the time to diagnosis and lead to a better prognosis was identified in the analysis (see the “Diagnostic delay” section above).

Limitations

Our review has some intrinsic limitations: (1) it only includes Italian- and English-written articles, excluding some potentially interesting reports written in other languages; (2) it encompasses studies ranging from 1988 to 2021 during which time myelography has been substituted by MRI and MRI itself has become progressively more sophisticated so it was sometimes difficult to compare radiological data among the studies; (3) publications are mostly limited to single case reports and small case series; (4) many patients were lost on follow-up or received a very close range follow-up so that their actual long-term outcome is unknown; (5) in some cases, clinical data were scarce.

Conclusions and future directions

CVFs are rare and treatable conditions but, since their first clinical description, few progresses have been made in their early diagnosis. Despite the several innovations in surgery and neuroimaging introduced during the last four decades, CVFs still carry a moderate/severe grade of disability in two-third of cases; among the reasons we recognize late diagnosis and treatment. Our analyses show that diagnostic delay is more often associated with worse clinical outcomes, suggesting that early diagnosis could not only lead to a reduction of mortality rate but also to a noticeable reduction in residual disability. Interestingly, the latter is not associated with the severity of clinical picture, so more compromised patients do not necessarily show a worse outcome. Misdiagnosis itself is not associated with a poorer outcome but it can increase diagnostic delay which is, in turn, associated with poorer recoveries.

Unfortunately, we were not able to recognize any highly suggestive (“red flags”) CVF’s pattern of symptoms to shorten the time to diagnosis, but we can empirically suggest considering CVFs and conduct an angiography including cerebral vessels in a patient with slowly progressing/relapsing myelopathy when myelitis routine work-up is inconclusive.

The findings also emphasize the importance of careful investigation of spinal flow voids in appropriate clinical contexts, as they can provide valuable clues for CVFs and help distinguish them from other mimics, except for spinal fistulas. Prompt extension of angiographic studies to intracranial vessels is suggested when spinal angiography is unremarkable in suspected cases of CVFs [9-11]. Other imaging features were found to be non-specific and could potentially lead to misdiagnosis. Suggestions on when to perform a cerebral angiography are reported in Fig. 6.

Fig. 6
figure 6

Red flags for performing a cerebral angiography

Full size image

In conclusion, a multidisciplinary approach is needed to better understand whether early treatment improves patients’ prognosis and quality of life, and whether a combined clinical-radiological predictive score could help to decide when to perform a cerebral angiography in a patient with otherwise unexplained myelopathy.