**Genetics of Degenerative Cervical Myelopathy: A Systematic Review and Meta-Analysis of Candidate Gene Studies**

### **Daniel H. Pope 1, Benjamin M. Davies 2, Oliver D. Mowforth 1, A. Ramsay Bowden 3,4 and Mark R. N. Kotter 2,5,\***


Received: 17 November 2019; Accepted: 14 January 2020; Published: 20 January 2020

**Abstract:** Degenerative cervical myelopathy (DCM) is estimated to be the most common cause of adult spinal cord impairment. Evidence that is suggestive of a genetic basis to DCM has been increasing over the last decade. A systematic search was conducted in MEDLINE, EMBASE, Cochrane, and HuGENet databases from their origin up to 14th December 2019 to evaluate the role of single genes in DCM in its onset, clinical phenotype, and response to surgical intervention. The initial search yielded 914 articles, with 39 articles being identified as eligible after screening. We distinguish between those contributing to spinal column deterioration and those contributing to spinal cord deterioration in assessing the evidence of genetic contributions to DCM. Evidence regarding a total of 28 candidate genes was identified. Of these, 22 were found to have an effect on the radiological onset of spinal column disease, while 12 genes had an effect on clinical onset of spinal cord disease. Polymorphisms of eight genes were found to have an effect on the radiological severity of DCM, while three genes had an effect on clinical severity. Polymorphisms of six genes were found to have an effect on clinical response to surgery in spinal cord disease. There are clear genetic effects on the development of spinal pathology, the central nervous system (CNS) response to bony pathology, the severity of both bony and cord pathology, and the subsequent response to surgical intervention. Work to disentangle the mechanisms by which the genes that are reviewed here exert their effects, as well as improved quality of evidence across diverse populations is required for further investigating the genetic contribution to DCM.

**Keywords:** genetics; single nucleotide polymorphism; degenerative cervical myelopathy; ossification posterior longitudinal ligament; severity; surgery

### **1. Introduction**

Degenerative cervical myelopathy (DCM) is estimated to be the most common cause of spinal cord impairment in the adult population and its incidence is expected to rise as the population continues to age [1]. The term DCM is relatively new, and it was proposed to unify degenerative pathologies with a common injury mechanism (subacute, progressive spinal cord injury) and treatment (decompressive surgery) [1]. This includes both cervical spondylosis (such as degenerative disc disease or osteophyte formation) and the ossification of the posterior longitudinal ligament (OPLL) or ligamentum flavum (OLF) [1–4]. These aetiologies were often previously separately considered, as cervical spondylotic myelopathy (CSM) and OPLL.

The trajectory of DCM between patients is heterogenous and currently unpredictable and unexplained [3]. For example, mechanical compression is an imaging hallmark of the disease. However, the location and amount of compression does not correlate with the disease symptoms [5–7]. In fact, the clinical phenotype can range from asymptomatic to severe disability, nearly independent from the amount of compression. Furthermore, patients' response to surgical decompression, the mainstay of treatment, is variable: it achieves excellent improvements in some patients, whereas in others these do not occur [8]. Such variation between patients has led to increasing interest in the genetic basis of this condition. One study reported a relative risk of 5.21 for the development of DCM in first-degree relatives of patients [9].

So far, the effects of genes involved in inflammation, bone, and lipid metabolism have been linked to both the pathogenesis of DCM and the response to surgical intervention [10,11]. However, these studies have failed to disentangle their relationship to spinal degeneration and myelopathy. This is important, as the fact that symptom progression and severity of spinal cord compression correlate poorly suggests that the genetic polymorphisms that contribute to spinal column degeneration may be distinct from those that influence the development of myelopathy in response to the resulting spinal cord compression.

Moreover, reviews have focused on CSM or OPLL, as opposed to DCM. Genes that influence how the spinal cord copes with mechanical stress may be identifiable in studies that investigate the severity of myelopathy and, in particular, the response to surgery.

Therefore, the objectives of this review are to provide a synthesis of the published literature on a genetic contribution to the susceptibility to develop degenerative spinal column changes that lead to DCM, the heterogeneity in severity of the clinical manifestation of DCM, and the heterogeneity in response to surgery, in order to evaluate the genes that are specifically linked to the onset and recovery of myelopathy.

### **2. Methods**

A systematic review was conducted in accordance with the PRISMA guidelines; a PRISMA checklist is presented in the Supplementary Data [12]. A search was conducted in MEDLINE, EMBASE, Cochrane, and HuGENet databases for all relevant papers from database origin up to 14th December 2019. The full search strategy is presented in the Supplementary Data and it was developed in conjunction with the Medical Library at the University of Cambridge School of Clinical Medicine. Reference lists of key articles were systematically examined to identify further eligible articles.

Titles and abstracts were screened for relevance and, subsequently, full text papers were screened for eligibility, according to the following inclusion criteria:


Animal studies, case reports, letters, editorials, reviews, technical notes, commentaries, proposals, and corrections were excluded. In addition, articles meeting the following criteria were excluded:


### • Focus on thoracic or lumbar spine

Two authors independently assessed the full-texts of potentially relevant articles (DHP and BMD), with any disagreements being resolved through discussion until agreement was reached.

Data that were extracted from the eligible articles included: study design, number of cases, number of controls, participant demographics, patient disease profile, gene studied, polymorphism/haplotype studied, and effects of polymorphisms and haplotypes on DCM susceptibility/severity/response to surgery (principal summary measures: odds ratios). The risk of bias was assessed through an evaluation of study design, methods of study population selection, matching of controls to cases, and the consideration of publication source. The MINORS methodological items were used to give structure to this process [13]. The GRADE guidelines were used to rate the quality of evidence for each candidate gene, and across genes for each of the three main questions (susceptibility, severity, response) [14].

Meta-analysis using the Cochrane Review Manager 5.3 software was used for polymorphisms, where more than one study had investigated the same polymorphism and the requisite data were available.

### **3. Results**

After removing duplicates, a total of 914 articles were screened and 39 were eligible for inclusion (Figure 1). In total, 37 articles addressed the genetics of susceptibility to developing DCM, 13 articles addressed the genetics of heterogeneity in DCM severity (either radiological or clinical severity) and six addressed the genetics of response to surgery. A total of 28 genes were identified, with key information regarding each candidate gene presented in Tables 1–3.

**Figure 1.** PRISMA flow diagram of search and screening.

### *3.1. What are the Genetic E*ff*ects on Susceptibility to Development of DCM?*

Evidence regarding the onset of DCM/OPLL was identified for 28 genes: ACE, APOE, BID, BMP2, BMP4, BMP9, COL6A1, COL9A2, COL11A2, FGF2, FGFR1, FGFR2, HIF1A, IL1B, IL15RA, IL18RAP, leptin receptor, NPPS, OPG, OPN, RUNX2, TGFB1, TGFB3, TGFBR2, TLR5, VDBP, VDR, and VKORC1. Of these 28 genes, 22 were found to be associated with the radiological onset of spinal pathology, while 12 were associated with the clinical development of DCM (i.e., spinal cord pathology). For six genes, no significant effect of polymorphisms has been found by the studies reviewed to date: FGF2, FGFR2, IL18RAP, leptin receptor, TLR5, and VDBP. Most of the genes (19, 68%) have been investigated by only a single study. Bone morphogenetic protein genes (9, 32%) and collagen genes were the most studied gene groups (8, 29%). Table 1 presents full information for each gene.


**Table 1.** Susceptibility to radiological or clinical degenerative cervical myelopathy (DCM).



*J. Clin. Med.* **2020** , *9*, 282


[22]

members

intron 33 (+20) G allele

 0.625


### *J. Clin. Med.* **2020**, *9*, 282




*J. Clin. Med.* **2020** , *9*, 282

### 3.1.1. Spinal Pathology

The majority of studies investigating the genetics of susceptibility to DCM used the radiological definition of cases. Therefore, these studies assess the development of bony spinal pathology (an initial stage in overall DCM development).

Kim et al. (2014) investigated the *ACE* gene, finding the deletion/deletion genotype of the intron 16 polymorphism (rs4646994) to be associated with an increased risk of developing radiological OPLL (AOR 2.20, *p* = 0.002) [15]. Similarly, two SNPs of the *BID* gene (rs8190315, rs2072392) were associated with the development of OPLL (OR 2.66, *p* = 0.005 for both) [18].

Four studies have investigated the role of variants in *BMP2*. Wang et al. (2008) found no significant effect of the Ser87Ser SNP, but found the Ser37Ala SNP was associated with an increased risk of OPLL development (*p* < 0.001) [19]. Interestingly, however, patients with the GG genotype of Ser87Ser had significantly greater number of ossified vertebrae, which suggested the A allele restricts ectopic ossification in OPLL. Meanwhile, the Ser37Ala SNP had no significant effect on the number of ossified vertebrae.

Yan et al. (2013) also found the Ser37Ala SNP to be associated with increased risk (*p* < 0.001) [21], although a more recent study that compared OPLL patients to their family members found no effect of either the Ser87Ser or Ser37Ala SNPs on risk of OPLL (*p* = 0.411, *p* = 0.670, respectively) [22]. Additionally, the 570A>T SNP in the *BMP2* gene was not found to be significantly associated with risk of OPLL [21]. Liu et al. (2010) used a patient cohort that included OPLL, OLF, and OPLL + OLF patients, but found no effect of the rs1005464 intronic SNP on the susceptibility of radiological DCM development [20].

In the *BMP4* gene, the 6007C>T SNP was found to be associated with an increased risk of developing radiological OPLL in male patients (OR 1.57, *p* = 0.014), although the effect is lost when males and females are considered together (*p* = 0.493) [23]. In the same SNP, the CT and TT genotypes were associated with a greater number of ossified vertebrae (*p* = 0.043) [23], as was a haplotype (TGGGCTT) containing seven SNPs (*p* = 0.002). Ren et al. (2012a) identified three SNPs that significantly increase the risk of OPLL: rs54419150 (OR 3.48, *p* < 0.001), rs17563 (OR 2.22, *p* < 0.001), and rs76335800 (OR 1.99, *p* < 0.001). Linkage disequilibrium studies also identified the haplotype block TGGGCTT containing these three SNPs to be significantly associated with the occurrence of OPLL (OR 2.54, *p* < 0.001) [24].

In the *BMP9* gene, two SNPs and a haplotype containing four SNPs were found to be associated with an increased risk of OPLL development: rs75024165 (OR 1.82, *p* < 0.001), rs34379100 (OR 1.95, *p* = 0.003), and haplotype CTCA (OR 2.37, *p* < 0.001). The haplotype was also associated with development of a greater number of ossified vertebrae (*p* = 0.001). A further SNP (rs9421799) was found to be protective (OR 0.69, *p* = 0.004), while three SNPs had no significant effect [26].

Wang et al. (2018) investigated the *BMPR1A* gene, finding two SNPs (-349C>T, 4A>C) that were associated with an increased risk of OPLL development (*p* < 0.001 both), and two (1327C>T, 1395G>C) with no significant effect [27]. Furthermore, patients with the C allele of the 4A>C SNP were more likely to have a greater number of ossified vertebrae on lateral cervical radiograph (*p* < 0.001).

The *COL6A1* gene has been the subject of four studies. Tanaka et al. (2003) investigated 32 SNPs in the *COL6A1* gene, of which 21 were significantly associated with OPLL (see Table 1) [28]. Further work by Kong et al. (2007) was consistent with these findings, with intron 32 (-29) C allele conferring a greater risk of OPLL (OR 1.89, *p* = 0.004) [29]. However, Liu et al. (2010) reported no significant effect of the rs2276255 SNP on the risk of OPLL or OLF development [20], in contrast to Tanaka et al.'s finding of a weak significant effect (*p* = 0.048). Further contradiction in the *COL6A1* gene is seen in Kong et al.'s (2007) finding that the promoter (−572) SNP T allele was associated with a 2.94 times greater risk of OPLL (*p* = 0.0003), while Kim et al. (2014) found no significant effect (*p* = 0.282) [22]. Liu et al. (2010) found no effect of one additional SNP (rs9978314) on the risk of OPLL or OLF development [20].

In the *COL11A2* gene, the intron 6 (−4) polymorphism was associated with a greater risk of OPLL development in two studies (OR 1.99, *p* = 0.0003; *p* = 0.0004) [31,32]. Similarly, the exon 6 (+28) polymorphism was associated with an odds ratio of 1.84 of developing OPLL (*p* = 0.0012) [32].

Jun & Kim (2012) investigated the *FGF2*, *FGFR1*, and *FGFR2* genes in 157 OPLL patients and 222 age- and sex-matched controls [34]. Three SNPs of the *FGF2* gene showed no significant effect on the likelihood of OPLL development, as did three SNPs of the *FGFR2* gene. However, the rs13317 SNP in the *FGFR1* gene was associated with an increased risk (OR 2.0, *p* = 0.02).

Kim et al. (2011) investigated two SNPs of the *IL15RA* (*IL15R*α) gene [36]. The A allele of rs2228059 conferred a 1.52 times risk of radiological OPLL (*p* = 0.009), while the rs2296139 SNP had no significant effect.

The A861G polymorphism of the leptin receptor gene had no effect on the likelihood of OPLL development in a study of 156 OPLL patients and 93 age-matched controls [38].

In the *NPPS* gene, two studies both found no significant effect of the IVS20-11delT SNP on the likelihood of radiological OPLL (*p* = 0.512, *p* = 0.093) [38,41]. However, patients that were homozygous for the T deletion of the IVS20-11delT polymorphism had fewer ossified vertebrae and less thick ossification of their cervical vertebrae (*p* < 0.001 for both) [41].

The IVS15-14T>C and C973T SNPs were associated with an increased risk of radiological OPLL (*p* = 0.026, *p* < 0.001) [41]. Furthermore, patients with the T allele of the IVS15-14T>C SNP also had both a greater number of ossified vertebrae and greater thickness of ossification of their vertebrae (*p* < 0.001, *p* = 0.017, respectively). For the C973T SNP, the T allele was associated with increased thickness of ossified vertebrae (*p* = 0.007), but it had no effect on number of ossified vertebrae (*p* = 0.248). There was no effect of the A533C polymorphism on the likelihood of radiological OPLL development, or number of ossified vertebrae, or thickness of ossified vertebrae (*p* = 0.430, *p* = 0.363, *p* = 0.947) [41].

In a case-control study of OPLL, OLF, and OPLL+OLF patients, 11 SNPs of the *RUNX2* gene had no significant association with radiological development of OPLL/OLF [20]. However, patients with the C allele of the rs16873379 SNP had a greater number of ossified vertebrae (*p* = 0.001), as did patients with the A allele of the rs1406846 SNP (*p* = 0.020), and patients with the C allele of the rs2677108 SNP (*p* = 0.044).

In the *TGFB1 (TGF*β*1)* gene, the CC genotype of the 869T>C polymorphism was found to be associated with an increased risk of radiological OPLL development in one study (OR 4.5, *p*=0.0004) [45], but it had no such association in a recent study that involved almost double the number of cases (*p* = 0.656) [46]. On meta-analysis, there was no significant effect of the 869T>C polymorphism on the susceptibility to OPLL development (OR 1.50, 95% CI 0.97–2.32, *p* = 0.07; Figure 2). The 509C>T was found to have no association with radiological OPLL development [46].


**Figure 2.** Forest plot for *TGFB1* 869T>C polymorphism.

Jekarl et al. (2013) investigated three SNPs of the *TGFBR2* (*TGF*β*R2*) gene, finding that two were associated with increased likelihood of OPLL development. The 445T>A polymorphism conferred a 2.81 times increased risk (*p* = 0.007), while the 571G>A polymorphism was associated with 8.73 times risk (*p* = 0.024) [47].

The *TLR5* gene has been investigated by one study, which found no association of three SNPs with the likelihood of OPLL development [48].

In the *VDR* gene, Kobashi et al. (2008) found the *FokI* polymorphism to be associated with 2.33 times increased risk of OPLL development (*p* = 0.0073) [50]. Similarly, Liu et al. (2010) found an association between the rs11574079 polymorphism and OPLL/OLF risk (OR 2.68, *p* = 0.0714) [20].

The *VKORC1* gene was investigated in 98 OPLL patients and 200 control subjects, with the −1639G> A polymorphism having a significant effect in female patients (OR 5.22, *p* = 0.004), but not when both sexes were considered together (*p* > 0.05) [52].

In the *NPPS* gene, He et al. (2013) examined the effect of four SNPs on the progression of OPLL ossification on lateral radiograph. The AA genotype of the A533C SNP and the homozygous T deletion genotype of the IVS20-11delT SNP were both associated with better responses to surgical intervention (OR 3.11, *p* = 0.029; OR 3.35, *p* = 0.007). The other two polymorphisms were not associated with any difference in response to surgery (good response defined as <2 mm increase in ossified mass of the posterior longitudinal ligament) [41].

### 3.1.2. Spinal Cord Pathology

Multiple studies used clinical signs and symptoms of DCM alongside positive radiological findings. Such combination interrogates the development of cord pathology, rather than simply the development of spinal pathology.

In the *APOE* gene, the ε4 allele was found to be associated with an increased risk of myelopathy in a case-control study, where the controls had cervical spondylosis without myelopathy (OR 3.50, *p* = 0.008) [16]. However, a study in an Indian population found the ε2 allele to be associated with increased risk of myelopathy when compared to both the ε3 and ε4 alleles (OR 4.4, *p* = 0.002; OR 6.69, *p* = 0.009) [17].

In the *BMP4* gene, Wang et al. (2013) found the 6007C>T SNP to be protective for the development of clinical signs and symptoms of CSM (OR 0.51, *p* < 0.001) [25]. This is in contradiction to the evidence described above, in which this SNP was shown to be associated with an increased risk of radiological OPLL development [23,24].

The Trp2(+) allele of the *COL9A2* gene was associated with an increased risk of CSM development (OR 1.78, *p* = 0.048), a risk that was worsened by heavy smoking (OR 5.56, *p* < 0.001), while the Trp3 allele had no significant effect [30].

Koga et al. (1998) identified three polymorphisms of the *COL11A2* gene associated with DCM development: promoter (−182), exon 43 (+24) and exon 46 (+18) [31]. Horikoshi and colleagues investigated two additional SNPs of the *COL11A2* gene, but found no significant effect for either [33].

In the *HIF1A* (*HIF-1*α) gene, Wang et al. (2014) found no effect of the 1772C>T SNP, while the 1790G>A polymorphism was associated with an increased risk of CSM development (OR 1.62, *p* < 0.001) [35].

In the *IL15RA* gene, Guo et al. (2014) found a significant effect of the A allele of the rs2228059 SNP on DCM development (OR 1.63, *p* < 0.001) [37]. However, there was no effect of the rs2296139 SNP on the likelihood of developing symptomatic DCM. This is in commonality with the above findings of Kim et al. (2011) who showed rs2296139 had no effect on likelihood of developing radiological OPLL while the rs2228059 SNP did [36].

In the *IL18RAP* gene, Diptiranhan et al. (2019) found no significant effect of either the rs1420106 or rs917997 SNPs on the development of myelopathy (*p* > 0.05) [17].

Three studies have looked at the *NPPS* gene in relation to clinical onset of spinal cprd disease [33,39,40]. Nakamura et al. (1999) found the IVS20-11delT polymorphism to be associated with an increased risk of development of DCM (*p* = 0.0029) [39]. There is conflicting evidence of the effect of the IVS15-14T>C polymorphism: one study found it to be associated with a 3.01 times risk of myelopathy development (*p* = 0.022) [40], while another found no significant effect (*p* = 0.320) [33].

Yu et al. (2018) found no significant effect of the 1181G>C and 163A>G polymorphisms in the osteoprotegerin (*OPG*) gene, but found the C allele of the 950T>C SNP to be associated with a greater risk of myelopathy (*p* < 0.01) [42].

Wu et al. (2014) studied three SNPs of the osteopontin (*OPN*) gene [43]. Two showed no significant effect, but the G allele of the -66T>G SNP was associated with an odds ratio of 1.55 of clinical onset of DCM (*p* = 0.002).

In the *RUNX2* gene, Chang et al. (2017) found the SNPs rs967588 and rs16873379 to be protective for DCM development (OR 0.47, *p* = 0.033; OR 0.48, *p* = 0.033) [44]. The rs1406846 SNP was, on the other hand, strongly associated with DCM development (OR 5.67, *p* < 0.001). Four further SNPs had no significant effect.

Horikoshi et al. (2006) studied the *TGFB1* (*TGF*β*1*) and *TGFB3* (*TGF*β*3)* genes [33]. There was no significant effect of the IVS2+114G>A SNP of *TGFB1*, while the CC genotype IVS1-1284G>C SNP of *TGFB3* was associated with an increased risk of DCM development (OR 1.46, *p* = 0.044).

Song et al. (2018) found no significant effect of the Thr20Lys polymorphism of the *VDBP* gene (OR 0.973, *p* = 0.834) [49].

In the *VDR* gene, Wang et al. (2010) found no significant effect of *FokI* polymorphism on CSM risk [51]. The *BsmI* polymorphism also had no effect on CSM risk, but the *ApaI* and *TaqI* polymorphisms conferred a 2.88 times and 4.67 times increased CSM risk (both *p* < 0.001). In opposition to Wang et al.'s findings, Song et al. (2018) found the ff genotype of the *FokI* polymorphism to be associated with a 1.985 times greater risk of myelopathy (*p* = 0.003) [49].

### *3.2. What Are the Genetic E*ff*ects on Clinical Severity of DCM?*

Seven studies investigated the genetic effects on the clinical severity of DCM, while 11 investigated radiological severity (four studies investigated both). Polymorphisms of 8 genes affected radiological severity, while three genes affected clinical severity. Table 2 presents the full results.


**Table 2.** Radiological or clinical severity of DCM.


*Cont*.

> **Table 2.**

*J. Clin. Med.* **2020** , *9*, 282

CSM patients with the Val66Met polymorphism of the *BDNF* gene had more severe disease, as assessed by functional survey: worse SF-36 scores for physical functioning and physical health summary than their counterparts without the polymorphism (*p* < 0.05) [53].

Wang et al. (2014) studied the effect of two polymorphisms of the *HIF1A* gene on CSM: 1772C>T and 1790G>A [35]. While the former conferred no significant difference in mJOA score, in the latter patients with the A allele had significantly worse mJOA scores than their G allele counterparts (*p* < 0.001).

Yu et al. (2018) found the TT genotype of the 950T>C polymorphism in the *OPG* gene to be associated with higher mJOA scores and fewer ossified vertebrae (*p* < 0.05); the TT genotype appears to be protective [42].

Wu et al. (2014) investigated four polymorphisms of the *OPN* gene in 187 CSM patients, finding no significant difference of all four polymorphisms on the mJOA score [43].

There was no effect of the Thr420Lys polymorphism of the *VDBP* gene on mJOA score or the number of ossified segments in 318 CSM patients [49]. Similarly, four polymorphisms of the *VDR* gene (*FokI*, *BsmI*, *ApaI*, *TaqI*) were found to have no significant effect on mJOA score in two studies [49,51].

### *3.3. What Are the Genetic E*ff*ects on Response to Surgery in DCM?*

The polymorphisms of five genes were associated with clinical response to surgery in DCM: *APOE*, *BMP4*, *HIF1A*, *OPN*, and *RUNX2*. The *NPPS* gene was studied for radiological response to surgery. Table 3 presents the results.

In the *APOE* gene, the ε4 allele was associated with an increased risk of poor response to ACDF surgery. In a multivariate model, it was associated with an 8.6 times risk of worsening or no change in mJOA score (*p* = 0.004) [54].

The 6007C>T polymorphism of the *BMP4* gene was associated with greater likelihood of post-surgical improvement of mJOA score (OR 1.53, *p* = 0.002), but the -5826G>A polymorphism had no significant effect (*p* = 0.053) [25].

In the *HIF1A* gene, the 1790G>A polymorphism was also associated with a greater likelihood of post-surgical improvement of the mJOA score (OR 1.55, *p* = 0.024) [35].

In the *OPN* gene, the GG genotype of the −66T>G SNP was found to be associated with worse response to surgical intervention, as assessed by mJOA score (OR 3.62, *p* = 0.007) [43]. Good surgical response was defined as >50% improvement in mJOA score.

Seven polymorphisms of the *RUNX2* gene were investigated for their effect on pre- vs. post-surgical mJOA score. The patients with the CC genotype of the rs16873379 SNP improved less (52.4%) than patients with TT genotype (61.7%), an effect that is mirrored by patients with the AA genotype of the rs1406846 SNP and patients with the CC genotype of the rs2677108 SNP. Patients with the AA genotype of the rs6908650 SNP improved more (66.8%) than their counterparts with the GG genotype (57.4%). The three other polymorphisms had no significant effect on mJOA score improvement [44].

In the NPPS gene, the AA genotype of the A533C polymorphism was associated with a 3.11 times greater likelihood of radiological improvement after surgical intervention. Similarly the IVS20-11delT homozygous T deletion was associated with a 3.35 greater likelihood of improvement. For both polymorphisms, improvement was defined as an increase of <2 mm in the ossified mass of the posterior longitudinal ligament over a mean follow-up length of 3.1 years [41].



*J. Clin. Med.* **2020**, *9*, 282

### **4. Discussion**

The aim of this study was to critically appraise the current evidence on the genetic contribution to DCM, with specific focus on distinguishing spinal column disease from spinal cord disease. Studies were identified evaluating the susceptibility, severity, and responsiveness to surgery in DCM. Studies on spinal column disease focused on the radiological outcomes of OPLL. Evidence was identified for a number of genes, including many in the *TGF*β superfamily and many known to be associated with bone development.

By further focusing on studies evaluating relationships with clinical function, versus radiological measures, a shortlist of genes that were related to spinal column disease or 'myelopathy' and not 'spondylosis' was identified: specifically, 12 genes that were associated with susceptibility, three genes with clinical severity, and five genes with response to surgical intervention. Table 4 presents a summary of the evidence for genetic effects on 'myelopathy', including GRADE rating for each gene. Across the three focuses of this review (susceptibility, severity, response to surgery), the GRADE rating of quality of evidence is baseline low, as all studies are observational. For all three, the quality of evidence is upgraded due to the large effects across genes, but downgraded due to inconsistency between studies.


*clinical* onset/severity/response to surgery rather than radiological). Colour coded for evidence red:noevidenceornotyetinvestigated).GRADEratingofqualityofevidencegivenforeach

> **Table 4.**

Summary of candidate genes a

ffecting myelopathy

 (i.e.,

### *4.1. Spinal Column Disease: Focus on OPLL*

The greatest focus of research to date has been on the bone morphogenetic proteins, a group of multifunctional growth factors that fall within the *TGF*β superfamily and are involved in cartilage development and the induction of bone formation [55]. Four genes within this family of growth factors have been associated with both altered susceptibilities to bony spinal pathology and altered susceptibility to the development of myelopathy: *BMP2*, *BMP4*, *BMP9*, and *BMPR1A*. The 4A>C SNP in the *BMPR1A* gene is associated with a significantly greater likelihood of radiological OPLL and a significantly greater number of ossified vertebrae [27]. Similarly, the CTCA haplotype of the *BMP9* gene is associated with a significantly increased risk of developing OPLL (OR 2.37), as well as a greater number of ossified vertebrae [26]. In the *BMP4* gene, a haplotype of 7 SNPs is associated with both greater susceptibilities to OPLL and worse disease [24]. Moreover, the 6007C>T SNP in the *BMP4* gene is associated with not only greater likelihood of developing bony pathology and greater severity of radiological disease, but also a greater likelihood of post-operative improvement of the mJOA score [23,25].

The dual role of 6007C>T SNP in the *BMP4* gene merits further discussion. The T allele of the polymorphism was found to be protective for spinal cord disease [25] (AOR 0.51) and it was associated with better outcomes in mJOA score after surgery (AOR 1.53 of being in the 'improvement' group). Conversely, Meng et al. found the same T allele to be associated with a greater likelihood of radiological OPLL (OR 1.57) [23]. The contrasting effect of the same allele suggests the effect of the *BMP4* gene is not limited to spinal pathology and the development of bony compression, but it may also influence the spinal cord response to such compression. It is unclear whether this effect is due to an intrinsic effect of *BMP4* on CNS resilience or regeneration, or a treatment artifact that faster compression elicited by the 6007C>T polymorphism giving more severe bony pathology results in faster decompression and better post-operative outcomes. Nonetheless, it is clear that bone morphogenetic protein genes may have extensive influences in the pathogenesis and symptoms of DCM.

Alongside the *BMP* genes, several other genes should be highlighted. In the *NPPS* gene, the C973T polymorphism significantly affected both the susceptibility of OPLL development and the thickness of ossified vertebrae, but notably did not affect the number of ossified vertebrae.

*NPPS* gene polymorphisms were implicated in post-surgical improvements of spinal column disease affecting the thickness of ossified vertebrae (C973T), while others (IVS15-14T>C) affect the number of ossified vertebrae and others affect both (IVS20-11delT) [41].

Evaluation of the network of genes that were found to be associated with the development of spinal column pathology shows that, while each gene has an independent effect on susceptibility to pathology, there is clear connectedness within and across gene families (Figure 3).

**Figure 3.** STRING Evidence Network for genes associated with spinal column disease.

### *4.2. Spinal Cord Disease*

The ε4 allele of the *APOE* gene, an allele that is well known for its associations with both cardiovascular disease and Alzheimer's disease, was associated with both a significantly increased likelihood of DCM development (OR 3.50) [16] and a significantly greater likelihood of failing to gain post-operative improvement (AOR 8.60 no improvement) [54]. However, this effect might not be universal across ethnicities; a study in an Indian population found the ε2 allele to be associated with development of myelopathy (OR 6.69) [17].

The 1790G>A polymorphism of the *HIF1A* gene displayed the opposite effect: it was associated with significantly greater likelihood of DCM development (OR 1.62), and worse disease but a greater likelihood of post-surgical improvement (OR 1.55) [35].

Reductions in Hif1α expression have been shown to be associated with the neuroprotective benefits of hyperbaric oxygen in spinal cord injury mouse models [56]. It is possible that such a mechanism is also the mediator of the *HIF1A* polymorphism's effect on susceptibility, severity, and post-operative response in DCM.

The *APOE* gene and its product, the apolipoprotein E transporter, are well-known to be involved remyelination, with defective clearance of myelin debris by the transporter limiting the potential for remyelination [57]. In the case of both *HIF1A* and *APOE*, their effects appear to be directly exerted on the cord's response to bony pathology, rather than via the bony pathology itself.

There appears to be delineation between genetic factors contributing to the development of bony pathology in the cervical spine, and those contributing to the CNS response to such insult. That an SNP of brain-derived neurotrophic factor (*BDNF*) is associated with the severity of disability (i.e., CNS response to insult) gives further weight to such a distinction [53].

As with genes that are associated with spinal pathology, the genes studied with relation to spinal cord disease have independent, but connected, effects (Figure 4).

**Figure 4.** STRING Evidence Network diagram for genes associated with spinal cord disease.

### *4.3. Conflicting Evidence*

The frequency of conflicting evidence is one striking aspect of much of the work reviewed here. The best example of this is perhaps seen in the *RUNX2* gene; the rs1406846 SNP A allele is associated with 5.67 times greater likelihood of developing DCM in one study [44], but it has no significant effect in a further study using a similar number of participants from the same country [20]. Similarly the 869T>C SNP in the *TGFB1* gene was associated with an odds ratio of 4.50 in one study [45], but a larger, more recent study found no significant effect of the same allele [46], with the result of meta-analysis showing no significant effect. Further examples of conflicting evidence include the IVS20-11delT polymorphism of the *NPPS* gene, in which one study found a significant effect on DCM susceptibility [39], but two others found no significant effect [38,41], while in the IVS15-14T>C polymorphism, two studies found an effect on susceptibility [40,41], with a further study showing no significant effect [33]. Such inconsistency might reflect the relatively small sample sizes of much of the work described here, and it indicates the need for large, well powered genetic investigations.

### *4.4. Limitations of Current Work*

Limitations of the current work on the genetics of DCM are multiple. Firstly, many of the studies that were reviewed in this article scored poorly on the MINORS methodological items assessment [13]. None published information regarding prospective calculation of study size, few reported whether the cases and controls were demographically matched, and some did not report how participants were recruited (e.g., consecutively). As mentioned above, the sample sizes remain in the hundreds rather than thousands, which limits the degree to which their conclusions can be considered valid. Moreover, in reporting the results, many omit odds ratios, instead of reporting only *p*-values, which limits the degree to which such results can be interpreted.

Many of the studies reviewed here focused exclusively on Japanese, Chinese, or South Korean participants, and specifically OPLL. Interestingly, in the *APOE* gene ethnicity appears to result in conflicting genetic effects, with the ε2 allele associated with myelopathy in Indian populations and the ε4 allele associated with myelopathy in Chinese populations [16,17]. It is widely acknowledged that there is a greater prevalence of OPLL within Asian populations, and this might explain their disproportionate representation in the literature [1]. However, without further work across ethnicities, it remains speculation as to whether the conclusions from these studies are globally relevant and across the spectrum of DCM pathologies.

There is significant diversity in the assessment of disease severity between studies. One study used the SF-36 quality of life survey [53], three used the mJOA score [35,43,51] (a clinical score commonly used in DCM research [58–61]), while others used radiographic measures [19,23,24,26,27,41,44,51]. A similar situation is found within the literature while considering response to surgery, with one study using a cut-off for 'improvement' as +1 point on mJOA score [54], some using >50% increase in mJOA score [25,35,43], one using a t-test of % improvement on mJOA between homozygous groups [44], and

one paper while using a radiographic definition of disease progression [41]. Such heterogeneity of outcome measures limits the degree to which the effects of genes on severity of DCM and response to surgery can be compared. The removal of surrogate outcome measures and more consistent use of a single form of outcome measure would permit more readily comparable conclusions to be drawn across different studies. We are currently undertaking RECODE DCM, an international consensus process to standardize the reporting of data elements in DCM research, and this would clearly hold benefit here (www.recode-dcm.com) [62]. For the reasons that are outlined above, the GRADE ratings of quality of evidence for each candidate gene were 'low' across all genes.

### *4.5. Future Directions*

It is clear that interest in this field is building, with increasing numbers of studies focusing on genetic effects in DCM (Figure 5). However, more than half the that are genes reviewed here have been investigated by only a single study, often with small sample sizes, which suggests more intensive work in larger populations is required to further describe the genetic basis of DCM. Furthermore, all of the studies included in this review focused on individual candidate genes. While some considered the effects of haplotypes consisting of several SNPs within a single gene [24,26,29], no work has yet combined SNPs across different genes. Such combinations may exhibit effect sizes of greater magnitude than those in the current body of literature, with potential for such genetic profiles permitting greater personalization of treatment strategies. Future work should also seek to characterize the mechanism by which the genes that were reviewed here exert their effects in the pathobiology of DCM.

**Figure 5.** Bar graph of number of papers investigating candidate genes in DCM in each calendar year.

#### **5. Conclusions**

While a number of limitations of the current work do exist, there is clear evidence of genetic effects of single nucleotide polymorphisms and haplotypes in DCM. Some of the genes exert their influence on the development of bony pathology, while others have effects on the spinal cord itself. Further investigation of the genetic basis of DCM requires larger study sizes, using more consistent measures of disease severity and response to surgery. The current evidence base is insufficient for translation to clinical practice for use in prognostication and management, but the potential for genetic profiles to be used in this way may well be realized once greater characterization of the genetic basis of DCM is achieved.

*J. Clin. Med.* **2020**, *9*, 282

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/1/282/s1, Data S1: search strategy used for MEDLINE and EMBASE databases. Data S2: search strategy used for Cochrane and HuGENet databases. Data S3: PRSIMA checklist.

**Author Contributions:** Conceptualization, B.M.D. and M.R.N.K.; methodology, B.M.D. and D.H.P.; formal analysis B.M.D. and D.H.P.; writing—original draft preparation, D.H.P.; writing—review and editing, B.M.D., D.H.P., O.D.M., A.R.B. and M.R.N.K.; supervision, M.R.N.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** Research in the senior author's laboratory is supported by a core support grant from the Wellcome Trust and MRC to the Wellcome Trust-Medical Research Council Cambridge Stem Cell Institute. MRNK is supported by a NIHR Clinician Scientist Award.

**Acknowledgments:** The authors are grateful to the Medical Library at the University of Cambridge for their assistance in the design of the search strategy for this article.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **Snake-Eye Myelopathy and Surgical Prognosis: Case Series and Systematic Literature Review**

**Marco Maria Fontanella 1,\*, Luca Zanin 1, Riccardo Bergomi 1, Marco Fazio 2, Costanza Maria Zattra 1, Edoardo Agosti 3, Giorgio Saraceno 1, Silvia Schembari 3, Lucio De Maria 1, Luisa Quartini 4, Ugo Leggio 5, Massimiliano Filosto 6, Roberto Gasparotti <sup>7</sup> and Davide Locatelli <sup>3</sup>**


Received: 9 May 2020; Accepted: 9 July 2020; Published: 12 July 2020

**Abstract:** The prognostic value of "snake-eyes" sign in spinal cord magnetic resonance imaging (MRI) is unclear and the correlation with different pathological conditions has not been completely elucidated. In addition, its influence on surgical outcome has not been investigated in depth. A literature review according to PRISMA (Preferred reporting items for systematic review and meta-analysis protocols) guidelines on the prognostic significance of "snake-eyes" sign in operated patients was performed. Clinical, neuroradiological, and surgical data of three institutional patients, were also retrospectively collected. The three patients, with radiological evidence of "snake-eyes" myelopathy, underwent appropriate surgical treatment for their condition, with no new post-operative neurological deficits and good outcome at follow-up. The literature review, however, reported conflicting results: the presence of "snake-eyes" sign seems a poor prognostic factor in degenerative cervical myelopathy, even if some cases can improve after surgery. "Snake-eyes" myelopathy represents a rare form of myelopathy; pathophysiology is still unclear. The frequency of this myelopathy may be greater than previously thought and according to our literature review it is mostly a negative prognostic factor. However, from our experience, prognosis might not be so dire, especially when tailored surgical intervention is performed; therefore, surgery should always be considered and based on the complete clinical, neurophysiological, and radiological data.

**Keywords:** snake-eye; owl sign; Hirayama disease; degenerative cervical myelopathy (DCM)

### **1. Introduction**

The "snake-eyes" appearance (SEA) or sign, also referred to as "owl-eyes" or "fried-eggs" sign, is a unique radiological finding appearing as bilateral hyperintense symmetric, circular or ovoid foci on T2-weighted (T2W) axial magnetic resonance imaging (MRI) sequences in the anterior horn cells of the spinal cord. It was first reported by Jenkins and Al-Mefty in 1986 [1]. The prognostic significance of this radiological finding has been debated in several articles with conflicting results [2–5].

SEA appearance is described in association with several clinical conditions like anterior spinal artery ischemia [6], chronic compressive myelopathy [7], degenerative cervical myelopathy (DCM) [4], Hirayama disease [8,9] or monomelic amyotrophy of the upper limb, amyotrophic lateral sclerosis [10], and spinal muscular atrophy [11]. The relationship between these ailments and the pathophysiology of SEA is not totally clear at present. It has been speculated that SEA is a reversible condition [3]. This claim is in contrast with its histopathology: in fact, SEA is the result of cystic necrosis at the junction of the central grey matter near the ventrolateral posterior column [12].

There is, therefore, a need to better understand SEA prognostic significance, and especially its influence on surgical outcome [13]. In this study, we evaluated the prognostic role of SEA through a systematic literature review and an analysis of our most recent patients who underwent surgical treatment.

### **2. Experimental Section**

### *2.1. Literature Review*

The systematic review of the literature was performed in March 2020 according to PRISMA guidelines [14]. Synthesis Without Meta-analysis (SWiM) guidelines were applied [15].

PubMed, Ovid MEDLINE, and Ovid EMBASE databases were searched using the keywords: "snake-eye myelopathy", "owl-eye myelopathy", "fried-eggs sign", "snake-eye appearance", "owl-eye appearance", and their variations. English studies published between December 1989 and December 2019 were included.

Inclusion criteria were: (1) studies with description of MRI-evident SEA myelopathy and surgery; (2) studies concerning a specific pathology related to SEA myelopathy, with patients undergoing surgical treatment; (3) studies with a clinical follow-up of surgically treated SEA patients. Exclusion criteria were: (1) absence of prognostic results about "snake-eyes sign"; (2) other radiological findings than SEA myelopathy; (3) absence of long-term follow-up.

For each study, we extracted the following baseline information: type of study, number of cases, clinical background, and prognostic value. The primary endpoint of the review was clinical outcome following surgical treatment in patients with SEA myelopathy.

### *2.2. Case Series*

We reviewed data of all our patients with a diagnosis of DCM, who underwent surgical treatment over the past year. Patients were included in our study if (1) they had a record of "snake-eye" sign on their T2W MRI sequences and they (2) gave consent to use of their information for research purposes.

Recorded information included: baseline demographic and clinical data (age at presentation and gender; symptoms and signs at presentation), treatment strategy, outcome at discharge and follow-up. Modified Japanese Orthopedic Association Score (mJOA score) [16] and Medical Research Council (MRC) Muscle Scale [17] were adopted for pre- and post-operative neurological evaluation.

### **3. Results**

### *3.1. Literature Review*

A total of 77 papers were identified after duplicates removal. After title and abstract analysis, 40 articles were identified for full-text analysis. Eligibility was ascertained for three articles. PRISMA flow chart is shown in Scheme 1. SWiM scheme is reported in Table 1.

**Scheme 1.** Literature review process, according to PRISMA guidelines.

Choi (2005) [4] reported 47 retrospective cases of DCM, Mizuno (2003) [12] described a total of 144 retrospective cases of degenerative cervical myelopathy with a mean pre-operative mJOA score of 10.8, and Zhang (2010) [2] reported 106 retrospective cases with a diagnosis of DCM with a mean pre-operative mJOA of 8.70.

In detail, subgroup analysis reported a total of 81 patients with ossification of posterior longitudinal ligament (OPLL).

Regarding DCM, the "snake-eye" appearance was regarded as a negative prognostic factor in 144 cases (48.5%). In particular, in Mizuno's study, the improvement ratio determined by the JOA score was 32.2% in SEA (mean post-operative mJOA score of 12.9), 47.1% in NSEA, and 50% (*p* < 0.01) in control cases in which high signal intensity was absent.


**Table 1.** Synthesis Without Meta-analysis (SWiM) in systematic reviews.

*3.2. Case Series*

Case 1. A 21-year-old man presented with a one-year history of numbness in the upper limbs with severe loss of hand sensation, especially on the right side. The neurological examination documented upper extremities hypoesthesia and pain, especially in the right arm, with no clear root distribution (mJOA upper extremity sensory subscore 1 of 3) and mild weakness of wrist extensor muscles (MRC grade 4/5), again worse on the right side, with sporadic dropping of objects (mJOA upper extremity motor subscore 4 of 5). No other neurological deficits were detected with a total mJOA score of 15/18. His past medical history included a traumatic injury secondary to hyperflexion of the cervical spine, causing transient acute tetraplegia and distal sensory loss, when he was 4 years old. Cervical spine MRI documented SEA at C5–C6 without stenosis of the vertebral canal. Dynamic flexion MRI showed reduction in the spinal canal diameter with subsequent medullary compression, especially at the C5–C6 level (Figure 1).

(**c**) (**d**)

**Figure 1.** Dynamic sagittal and axial cervical T2-MRI scan: (**a**) sagittal T2-MRI scan with visible myelopathy at C5–C6 level; (**b**) axial T2-MRI scan with an already visible SEA appearance; (**c**) sagittal T2-MRI scan in flexion, showing reduced spinal cord canal diameter, with subsequent spinal cord compression; (**d**) axial T2-MRI scan in flexion showing SEA at C5–C6 level and spinal cord compression.

An electromyography (EMG) of the upper limbs showed signs of bilateral chronic motor axonal neuropathy in C6 myotome; low amplitude pre-operative motor evoked potentials (MEPs) were detected. The patient underwent an anterior discectomy and fusion at the C5–C6 level with an intersomatic cage. Intra-operative MEP monitoring was performed (Figure 2).

The post-operative clinical course was uneventful with no evidence of new neurological deficits. At the post-operative neurological examination, the patient had a mJOA score of 17 over 18, with an upper extremity motor improvement of 1 point and an upper extremity sensory improvement of 1 point. Wrist extensor muscles weakness resolved completely (MRC grade 5/5).

**Figure 2.** Intra-operative motor evoked potentials (MEPs) prior (**a**) and following (**b**) stabilization. electromyography (EMG) needles record from a hemisphere of about 1 mm. Within this volume there are some 100 muscle fibers. EMG amplitude is the sum of the electric potential differences within a muscle relating to all the active motor units in the vicinity of the electrodes on the skin [18]. Trans-cranial stimulation was performed to elicit MEPs recorded with sub-dermal needle electrodes placed bilaterally in the biceps brachii (BB), extensor radialis carpi (ERC), opponens pollicis (OPP), tibialis anterior (TA), and flexor digitorum brevis (FBD). Prior to stabilization, MEPs were elicited from all monitored muscles on the right side (red) and only from BB, ERC, and TA on the left side (blue) because of technical reasons. Following stabilization, MEPs were essentially unchanged.

Case 2. A 44-year-old man suffered a severe traumatic brain injury that required decompressive craniectomy and subsequent cranioplasty. Years later he developed arm cramps, and he was subjected to a cervical MRI scan showing a post-traumatic anterior pseudomeningocele extending from C2 to C5. He underwent multiple lumbar punctures for cerebrospinal fluid (CSF) drainage and even a spinal-peritoneal shunt, which temporarily improved his symptoms, as previously suggested [14]. However, after some time, the pain recurred, along with progressive diparesis (MRC grade 3/5 for both proximal and distal movements) and hypoesthesia, which severely affected his quality of life. Neurological evaluation detected a mJOA score of 13/18 (2/5 upper motor extremity subscore, 1/3 upper sensory extremity subscore). A new MRI scan showed an extension of the already known pseudomeningocele and a new-onset cervical snake-eyes myelopathy at the C5–C6 level (Figure 3).

Superficial upper extremities EMG confirmed denervation in the upper of both arms and low arm myotomes; low-amplitude pre-operative MEPs were detected. It was decided to perform a C3–C7 spine posterior decompression and stabilization. During surgery the patient's intraoperative MEP did not show any worsening compared to the preoperative ones (Figure 4).

Surgery was uneventful, and, at six months outpatient follow-up, the patient regained significant strength in his arms, especially distally with a MRC grade 4/5 and a mJOA score of 16/18, gaining two points on the upper limbs motor scale.

Case 3. A 56-year-old woman reported pain in both arms radiating to her hands, especially on the right side, for about five months. The patient's medical history was unremarkable. She denied recent or past trauma. More recently, she also reported the development of grip loss in her right hand, which affected her daily activities. Neurological examination detected weakness of distal right arm movements (MRC grade 4/5) with mJOA score of 15/18 (upper extremity motor subscore 4/5, upper extremity sensory subscore 1/3). Cervical MRI scan showed cervical spondylosis with associated snake-eyes myelopathy at C5–C6 level (Figure 5).

(**a**) (**b**)

**Figure 3.** Pre-operative cervical T2-MRI scan: (**a**) sagittal view, showing an anterior pseudomeningocele extending from C3 to T5 with subsequent central canal stenosis and T2-high cord signal consistent with myelomalacia; (**b**) axial view, showing the edematous and T2-hyperintense anterior grey columns, with the characteristic snake-eyes sign.

**Figure 4.** Intra-operative motor evoked potentials (MEPs) prior (**a**) and following (**b**) stabilization. Transcranial stimulation was performed to elicit MEPs recorded with sub-dermal needle electrodes placed bilaterally in the extensor radialis carpi (ERC), opponens pollicis (OPP), tibialis anterior (TA), and flexor digitorum brevis (FBD). Prior to stabilization lower amplitude MEPs were elicited on the right side (red) compared to the left side (blue). Following stabilization higher amplitude MEPs were elicited from the upper limbs especially on the right side.

**Figure 5.** Pre-operative cervical T2-MRI scan: (**a**) axial view, showing cervical spondylosis with associated snake-eyes myelopathy at C5–C6 level; (**b**) sagittal view, showing the cervical canal stenosis with compression of the spinal cord.

The patient underwent a C3–C6 laminectomy and cervical arthrodesis. Surgery was uneventful. At discharge pain was reduced, especially in the right arm. The strength in her right hand was completely recovered (MRC grade 5/5 for distal arm movements) at the six month outpatient follow-up, with a mJOA score of 17/18. Post-op MRI scan is shown in Figure 6.

**Figure 6.** Post-operative cervical T2-MRI scan: (**a**) axial view through the C5–C6 level, showing resolution of the spinal canal stenosis, with persistence of the SEA signal; (**b**) sagittal view, showing resolution of the spinal cord compression.

### **4. Discussion**

SEA was first presented in a computer tomography (CT) myelography study of seven DCM patients in 1986 [1]. Subsequent anatomopathological studies confirmed that the main modifications were cystic necrosis at the junction of the central grey matter and the posterior ventrolateral column, combined with cell loss in the anterior horn [12]. SEA was reported in other forms of myelopathy too. A clinical randomized trial [8] showed that SEA myelopathy appears during the late stage of Hirayama disease, considered as an anterior horn disorder resulting from local ischemia, triggered by arterial compression from an anterior shifting of the posterior cervical dura upon neck flexion [12].

Undoubtedly, chronic mechanical compression and vascular insufficiency can be among the main promoters of SEA [8,18]. However, the pathogenetic mechanism is not completely understood and research is still ongoing. Although there is no clear data about the exact prevalence and incidence of SEA, some studies suggest that it is much more common than it might be believed [5,8,12].

A thorough literature review shows an inconsistency of results about the prognostic significance of SEA in surgical and non-surgical patients.

According to some studies, SEA does not affect the prognosis of patients who underwent corpectomy and fusion for treatment of DCM [4,5]. Another literature review stated that intense T2W SEA is associated with poorer surgical outcome in patients with DCM, while T2W SEA post-operative regression correlates with better functional outcomes [13]. No report about T1W hypointensity is reported about SEA

Li and Remmel stated that SEA is an irreversible lesion and a predictor of poor prognosis [19]. Mizuno et al. [12] assumed that SEA is an unfavorable prognostic factor for the recovery of upper extremity motor strength and that this is related to neuronal loss in the anterior horn.

A 2015 literature review [20] suggested that SEA can help in the differential diagnosis of spinal cord ischemia, indicating anterior horns infarction caused by anterior spinal artery ischemia.

The pathogenesis of the "snake-eye" myelopathy might be a matter of some debate, but it is interesting that a cervical hypermobility could lead to an anterior compression in flexion, in the absence of a spinal canal stenosis and compression in a neutral position. The institutional Case 1 seems to show this pathogenic mechanism: a dynamic MRI study was useful for a correct therapeutic decision. The surgical strategy, in this type of patient, might be a valid choice.

Regarding the follow-up data, especially the radiological one, the entity of spinal cord damage and therefore the reversibility of the "snake eyes" sign is very difficult to verify with a post-operative MRI study, mainly for the presence of implanted materials, that could distort the signal and preclude the identification of such a modest lesion. Of our three cases, only one already underwent post-operative MRI (Case 3), showing the persistence of the snake-eyes myelopathy, despite clinical improvement. For the other two cases, only clinical follow-up is available so far. Most importantly, all three patients had a clinical benefit from surgery, despite the radiological evidence of SEA.

The MRI picture of "snake eyes" has always been described in relation to a clinical picture of myelopathy, but another relevant problem is related to those cases that have a "snake-eyes" MRI picture and are asymptomatic or pauci-symptomatic like Case 1. Low-intensity signal on T1WI is considered as a sign of advanced disease due to a significant neural tissue damage and correlates with poor post-operative neurological outcome [21]. Signal changes in T1WI usually appear with an increase in T2WI [22]. T2 hyperintensity in isolation cannot predict a worse post-operative outcome: the combination of both signal alteration on T1 and T2, and the presence of long segments with T2 hyperintensity are better correlated with negative neurological outcome after surgery [23].

For other myelopathies, it is clear that there may be signs of hyperintensity in MRI that are not related to a clinical myelopathy and in these cases it is not clear what needs to be done [24]. Unfortunately, given our small case series, the less severely affected pre-operative disability in respect of cases presented in literature and the fact that SEA is not a frequent entity presenting in a variety of heterogenous diseases, it was not possible to determine its independent prognostic value.

Nevertheless, it is widely accepted that the baseline neurological status is the strongest predictor of post-operative outcome.

In addition, older age is related to a worse outcome after surgery [16]. Other clinical factors such as body mass index and baseline severity score are not predictive of complications [25]. Duration of symptoms is not considered uniquely as a negative clinical factor associated with a negative outcome after surgery [16,25] We suggest that, from the pathophysiological point of view, a more severe cervical myelopathy and a longer duration of symptoms could associate with different histological damages in the spinal cord that could be reversible or not. Therefore, to achieve a good clinical outcome, it is mandatory to identify early clinical signs of cervical myelopathy [26]. In this regard, different outcome measures are reported in the literature regarding functional impairment (mJOA scale), disability (Nurick scale and Neck Disability Index), and generic short form health survey (SF-36 scale). Finally, new neurophysiological tests, like contact-heat evoked potentials, could be of help in early detection of cervical myelopathy; in fact, they seem to exhibit a superior sensitivity compared to somato-sensory evoked potentials in detecting spinal cord ischemia caused by compression of the anterior spinal artery. Concerning this, the application of classical neurophysiological techniques might be a limitation of this study [27].

### **5. Limitations of the Study**

Our study suffers from several limitations. As for the analysis of the literature, few studies about DCM take into consideration surgical prognosis, and most of them have a suboptimal or no follow-up period. In addition, most studies presented an insufficient number of patients to reach the possibility of statistical inference on the general population. The variety of different pathologies that have been related to SEA myelopathy, an element that causes further difficulties in prognostic analysis, must also be considered. Furthermore, the term SEA is used unevenly in the scientific community and this may have led to loss of information. To partially compensate for this, a large number of synonyms for SEA have been used in the searching process, trying to cover the full range of terms used in the literature to describe this radiological sign. As for our case series we were unable to achieve optimal follow-up on all patients.

### **6. Conclusions**

"Snake eyes" myelopathy represents a rare form of myelopathy with a prognosis that is generally defined as unfavorable. Its pathophysiology is still unclear, and its frequency might be greater than previously thought.

The literature review and personal experience of surgically treated cases shows that SEA represents a negative surgical prognosis sign in a minority (22–28%) of patients, but the baseline neurological status remains crucial to determine patients' outcome.

**Author Contributions:** L.Z., C.M.Z., E.A., G.S., S.S., L.D.M., L.Q., U.L., M.F., R.G., D.L., M.M.F., and D.L. collected, analyzed, and interpreted the patient data; R.B. and M.F. performed the surgical procedures; R.B., M.F., U.L., L.Q., M.M.F., D.L., L.Z., and C.M.Z. followed the patients during the hospital course and performed follow-up analysis. All authors have read and agree to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Degenerative Cervical Myelopathy: How to Identify the Best Responders to Surgery?**

**Rocco Severino 1, Aria Nouri 2,\* and Enrico Tessitore <sup>2</sup>**


Received: 1 February 2020; Accepted: 5 March 2020; Published: 11 March 2020

**Abstract:** Surgery is the only definitive treatment for degenerative cervical myelopathy (DCM), however, the degree of neurological recovery is often unpredictable. Here, we assess the utility of a multidimensional diagnostic approach, consisting of clinical, neurophysiological, and radiological parameters, to identify patients likely to benefit most from surgery. Thirty-six consecutive patients were prospectively analyzed using the modified Japanese Orthopedic Association (mJOA) score, MEPs/SSEPs and advance and conventional MRI parameters, at baseline, and 3- and 12-month postoperatively. Patients were subdivided into "normal" and "best" responders (<50%, ≥50% improvement in mJOA), and correlation between Diffusion Tensor Imaging (DTI) parameters, mJOA, and MEP/SSEP latencies were examined. Twenty patients were "best" responders and 16 were "normal responders", but there were no statistical differences in age, T2 hyperintensity, and midsagittal diameter between them. There was a significant inverse correlation between the MEPs central conduction time and mJOA in the preoperative period (*p* = 0.0004), and a positive correlation between fractional anisotropy (FA) and mJOA during all the phases of the study, and statistically significant at 1-year (r = 0.66, *p* = 0.0005). FA was significantly higher amongst "best responders" compared to "normal responders" preoperatively and at 1-year (*p* = 0.02 and *p* = 0.009). A preoperative FA > 0.55 was predictor of a better postoperative outcome. Overall, these results support the concept of a multidisciplinary approach in the assessment and management of DCM.

**Keywords:** degenerative cervical myelopathy (DCM); surgical outcome; MRI; DTI; FA; ADC; signal changes spinal canal; neurophysiology; SSEP; MEP

### **1. Introduction**

Degenerative cervical myelopathy (DCM) is typically a chronic condition, commonly involving patients older than 55 years [1], and represents the most common cause of spinal cord injury in the industrialized world [2]. The progressive reduction of spinal canal diameter due to degeneration of the cervical spine, including the vertebrae, posterior longitudinal ligament, ligamentum flavum, intervertebral disk [3], results in a compression of the spinal cord, arterial perfusion to the nervous tissue, and consequent spinal cord ischemia [4,5]. DCM can be a highly disabling condition causing motor and sensory dysfunction that ultimately result in a reduced quality of life.

The diagnosis of DCM is based on clinical examination, and subsequently confirmed using imaging, and sometimes neurophysiological techniques such as sensory (SSEPs) and motor evoked potentials (MEPs). Studies using conventional MRI have shown that specific characteristics can correlate with neurological status and surgical outcome. The most commonly studied parameters include T1-weighted hypointensity or T2-weighted hyperintensity signals of the spinal cord and the number of compressed levels. It is believed that signal changes represent a wide-ranging set of pathological sequelae. Edema and gliosis are thought to result in demyelination and Wallerian degeneration, and are typically associated with T2 hyperintensity signal changes in the absence of T1 hypointensity [6]. After prolonged compression or significant dynamic injury, myelomalacia and loss of grey matter occurs [7–10], typically reflected by T1 hypointensity signal changes. However, T2 hyperintensity presents in 58%–85% of DCM patients, but it is present in 2.3% of people in the general population as well, making it a sensitive measure for diagnosis, but limiting it in terms of predicting surgical outcome. Contrarily, T1 hypointensity has been found to be a good predictor of suboptimal surgical outcome but its low prevalence in DCM of about 20% of patients limits its clinical utility [11].

In recent years, a newer MRI technique, the diffusion tensor imaging (DTI), has demonstrated an ability to identify the degenerative changes of the compressed spinal cord even in the early phases of DCM. [12,13] In the neural tissue, DTI imaging estimates the directionality and the diffusivity of water molecules through tissues and nervous fibers through two values: The fractional anisotropy (FA) and the apparent diffusion coefficient (ADC). These scalar parameters are inversely proportional and, in a damaged and demyelinated spinal cord water molecules diffuse in all different directions resulting in lower FA and higher ADC studies values [10,14], while in the normal population FA has higher scores because of intact myelin sheaths. [14,15] Moreover, several studies have shown a correlation between preoperative FA, ADC values, and clinical condition of DCM patients [13,16].

Neurophysiological studies are also useful diagnostic tools in the detection of functional alterations in nerve conduction in DCM. Several studies have demonstrated significant alterations in both SSEPs and MEPs in DCM patients [17–19]; specifically, a predictive value for the postsurgical outcome has been shown for median nerve SSEPs [19–21].

In a previous study, we were able to demonstrate the importance of combining clinical, radiological, and neurophysiological data in the assessment of DCM patients, in order to better identify those with optimal surgical outcomes [22]. The aim of the present study is to evaluate the correlation between DTI, neurophysiological parameters and neurological status, in both the preoperative and postoperative periods, in order to define a new multidisciplinary diagnostic approach that could identify the best candidates for decompressive surgery.

### **2. Methods**

We performed a prospective analysis of clinical, radiological, and neurophysiological data of thirty-six consecutive patients (13 males, 23 females; mean age: 57.05 years) suffering from DCM and operated between June 2012 and June 2018. Inclusion criteria are detailed in Table 1. All subjects gave their written informed consent for inclusion before they participated in the study. Research ethics board approval was obtained by the Institutional Ethical Board (NAC n. 11-194). A complete clinical, radiological, and neurophysiological evaluation of DCM, consisting of cervical MRI with DTI sequences, the modified Japanese Orthopedic Association (mJOA) score by Keller et al. [23], and neurophysiological assessments (both MEPs and SSEPs), was performed for all patients in the preoperative period and repeated at 3- and 12-months after surgery. The choice of the surgical approach depended on the predominant side of compression and on the surgeon's preferences. The posterior approach was chosen in patients with a multilevel compression and with preservation of the cervical lordosis on plain X-rays. Fusion was added in cases with radiological instability according to White and Panjabi criteria [24].

The mJOA score was used to measure the severity of DCM and segregated into patients with normal function (mJOA = 16–17), grade 1 (mJOA = 12–15), and grade 2 (mJOA = 8–11) myelopathy. We calculated the improvement of DCM using the mJOA recovery rate proposed by Hirabayashi, as follows: [(Postoperative mJOA-preoperative mJOA)/(17-preoperative mJOA) × 100] [25]. "Best responders" were identified by improvement of 50% or more in the postoperative period [26]. Patient improving below 50% or remaining stable (at least no deterioration) were defined as "normal responders".


**Table 1.** Inclusion and exclusion criteria.

### *2.1. Radiological Assessment*

All patients underwent a cervical 3 Tesla MR with the following sequences: Sagittal fast spin echo T2 (FSET2) TE 102 ms, TR 3500 ms, slice thickness 3 mm, axial gradient echo T2 TE 14 ms, TR 676 ms, slice thickness 3 mm, sagittal FSE T1, TE 10 ms, TR 900 ms, slice thickness 3 mm, sagittal diffusion tensor imaging (DTI) TE 58 ms, TR 4000 ms, *b*-value 700, slice thickness 2 mm, 25 directions with calculation of the apparent diffusion coefficient (ADC) and fractional anisotropy (FA).

The images were obtained in the preoperative period and at 3- and 12-months after surgery. The FA and the ADC were measured with isometric ROIs at three levels: The surgical level, or the narrowest point of the cervical stenosis in case of multilevel compression (level 2), and the nearest noncompressed intervertebral levels above and below (level 1 and 3, respectively—Figure 1). The presence and the extension of T2 hyperintensity, and the presence of T1 hypointensity was also investigated.

**Figure 1. A**: Measurement levels for fractional anisotropy (FA) and the apparent diffusion coefficient (ADC): The surgical level, or the narrowest point of the cervical stenosis in patients with multilevel compression (level 2), and the intervertebral levels above and below it (level 1 and 3). **B, C**: Diffusion tensor imaging (DTI) sequences for FA and ADC measurement, respectively.

The midsagittal diameter of the spinal canal at the site of greatest compression, was calculated on sagittal T2 images in both preoperative and postoperative controls (at three months) and used this to examine the expansion rate of the spinal canal after the surgery, using the formula: [(Postoperative–Preoperative AP diameter)/Preoperative AP diameter] × 100.

### *2.2. Electrophysiological Assessment*

All patients had undergone electrophysiological evaluation with SSEPs and MEPs. The Quadriceps Combined Test [27] was used for the analysis of the MEPs. Furthermore, calculation of the preoperative and postoperative (3-months and 1-year) central conduction time (TCC), corrected for the age and size, and the amplitude ratio of the MEP was obtained in all patients.

For SSEPs, the peak latencies of responses were recorded at Erb's point (N9), the C2 spinous process (N13), and the scalp (N20) for the median nerve. For the tibial nerve, we calculated the latencies N8–N22 (popliteal fossa to L1) and N22 (L1 spinous process). Unfortunately, several patients did not perform a complete postoperative SSEP analysis according to our criteria as a result of a mismatch between different neurophysiological diagnostic units. Therefore, only the SSEP analysis on the preoperative data was conducted.

### *2.3. Statistical Analysis*

Data were analyzed using an unpaired t-test to compare the mJOA, midsagittal diameter of the spinal canal, DTI parameters, and MEP/SSEP values, during all the phases of the study, in the "best responders" and "normal responders" to surgery. A Pearson correlation analysis was then performed to assess the correlation between pre and postoperative FA and ADC values, mJOA score, and MEP/SSEP values. Fischer exact tests were performed to determine a specific preoperative FA threshold that could be predictive of a better postoperative outcome. Fisher exact tests were also used to investigate if general "risk factors" (T2 hyperintensity, Tobacco use, diabetes, and clinical history >6 months) were related to worse outcome. A *p*-value < 0.05 was considered to be statistically significant.

### **3. Results**

### *3.1. Clinical Demographics and Outcome*

Twenty-six patients (72.2%) were operated by an anterior approach, including an anterior discectomy and fusion (ACDF) in 20 patients and an anterior corpectomy and fusion in six patients. Ten patients (27.3%) were treated with posterior decompression by laminectomy with or without fusion. Based on the mJOA score, there were 5/36 patients with normal function, 25/36 patients with grade 1 myelopathy, and 6/36 patients with grade 2 myelopathy. There were no patients with a preoperative grade 3 myelopathy.

The preoperative mJOA average value was 13.5, while the postoperative mean values were 14.9 at 3-months and 15.1 at 1-year, respectively. According to the Hirabayashi recovery ratio, 20 patients (55.5%) were considered "best responders"; the difference between the mJOA improvement of the "best responders" and "normal responders" patients at 1-year was statistically significant (*p* = 0.001, Figure 2). The difference between the mean age of the "best responders" and "normal responders" was not statistically significant (58.9 ± 13.2 vs. 54.6 ± 13.1, *p* = 0.34).

**Figure 2.** Difference of trends of the average modified Japanese Orthopedic Association (mJOA) scores in the "best responders" and "normal responders" patients. The improvement from the preoperative score to the 1-year value in the "best responders" group was significant (*p* = 0.001), as the difference between the 1-year values of the "best responders" (mean = 16.3) and the "normal responders" (mean = 13.3) patients (*p* = 0.001).

#### *J. Clin. Med.* **2020**, *9*, 759

There was no statistical difference between the two groups concerning both age (58.9 ± 13.2 vs. 54.6 ± 13.1, p: 0.34) and the investigated "risk factors" (T2 hyperintensity, p: 0.13; smoke, p: 0.22; diabetes, p: 0.83; clinical history > 6 months, p: 0.14. Table 2).


**Table 2.** Differences in characteristics between best responders and normal responders.

\*: Statistically signifant result.

### *3.2. Radiological Results*

The mean midsagittal canal diameter was 5.12 ± 1.4 and 8.92 ± 2 mm in the preoperative and postoperative period, respectively, with a mean expansion rate of 97.4%. Considering our cases as "best responder" and "normal responder" patients, we found no significant differences between the average values of both postoperative midsagittal diameters (8.98 ± 2.3 vs. 8.84 ± 1.6 mm, *p* = 0.8) and expansion rates (100.1% vs. 93.9%, *p* = 0.8).

Concerning the preoperative DTI parameters, the preoperative FA values were significantly higher in the "best responders" than the "normal responders" (0.63 ± 0.06 vs. 0.57 ± 0.08, *p* = 0.03). Six patients were excluded from the postoperative analysis because they presented with artefacts on their MRIs related to implanted metallic devices. In the remaining 30 patients, the average FA value remained higher in the "best responders" group at both 3-months (0.62 ± 0.08 vs. 0.58 ± 0.09) and statistically significantly different at 1-year (0.68 ± 0.07 vs. 0.55 ± 0.11, *p* = 0.004, Table 2). Furthermore, FA at the most stenotic level was significantly lower in the "normal responder" group preoperatively and at 1-year (*p* = 0.02 and *p* = 0.009, respectively—Figure 3, Table 2).

The "best responders" group had a preoperative FA > 0.55 in 71.5% of patients compared with only 28.5% of the "normal responders". This difference was statistically significant (p = 0.014), and suggests that a preoperative FA > 0.55 can be considered as a predictor of a better postoperative outcome.

T2 hyperintensity in the preoperative MRI was found in 12/16 (75%) of the "normal responders" patients and in 10/20 (50%) in the "best responders" group; however, this difference was not significantly different (*p* = 0.13). The preoperative average FA was similar in patients with and without T2 hyperintensity (0.595 vs. 0.596, *p* = 0.9). No T1 hypointensity was evident in any of our patients.

**Figure 3.** Differences between fractional anisotropy average values of the most compressed level in best responders (blue) and normal responders (red) patients.

Concerning the ADC, the average value between the "normal responder" group compared to the "best responders" group was higher in the preoperative period (1.54 vs. 1.40) and at the 3-month follow-up (1.40 vs. 1.27), but lower at the 1-year control (1.37 vs. 1.40 in the "best responders" group). These results did not show a statistically significant difference, and we did not consider a lower ADC as a predictive factor for good recovery.

### *3.3. Neurophysiological Results*

Concerning the MEPs, a significant inverse correlation between the CCT and mJOA values in the preoperative period (*p* = 0.0004, R = −0.59, Figure 4) was found.

**Figure 4.** Preoperative abnormal values of motor evoked potentials (MEPs) were related to worse mJOA scores: This inverse correlation was statistically significant (r = −0.59, *p* = 0.0004).

We found no correlation between abnormal preoperative SSEPs and preoperative mJOA scores. Regarding the relationship between SSEP and FA values, we observed a significant inverse correlation between preoperative FA and N22, N8–N22 latencies (*p* = 0.001 and *p* = 0.007, respectively (Figure 5), there was no statistically significant correlation with the other SSEPs otherwise. The same correlation was found postoperatively but due to the mismatch of the test conducted in different diagnostic centers we were not able to conduct any statistical analysis.

**Figure 5. A:** Inverse correlation between preoperative FA values and L1 spinous process (N22) (*p* = 0.001). **B**: Inverse correlation between preoperative FA values and popliteal fossa to L1 (N8–N22) (*p* = 0.007).

### *3.4. Correlation between FA, mJOA Values, and Neurophysiological Parameters*

A positive correlation between FA values and corresponding mJOA scores during all the phases of the study was found. However, this correlation was significant only for the 1-year postoperative values (*p* = 0.0005, R = 0.66, Figure 6).

**Figure 6.** Positive correlation between preoperative FA and mJOA at 1 year (*p* = 0.004, r = 0.66).

Moreover, a direct correlation was found between higher preoperative FA values, and the postoperative variation of the mJOA at 1 year. This result was significant considering both the FA at the most compressed level and the average value of all the three considered levels (*p* = 0.002, r = 0.66 and *p* = 0.0002, r = 0.75, respectively—Figure 7).

**Figure 7. A:** Significant correlation between preoperative FA value at the most compressed level and the 1-year postoperative variation of the mJOA (*p* = 0.002). B: Significant correlation between preoperative FA average value and the 1-year postoperative variation of the mJOA (*p* = 0.0002).

Concerning the relationship between FA and MEP values, an inverse correlation was found between TCC values and FA scores in the preoperative period, but this result was not statistically significant (r = −0.33; *p* = 0.08).

### **4. Discussion**

DCM is a complex and potentially disabling condition. The time for surgical intervention is usually dictated by the degree of neurologically severity, with moderate to severely impaired patients recommended for surgery, whereas mildly impaired patients may be offered surgery or careful observation. [28] Although most patients with DCM will improve after surgery, it remains challenging to accurately predict good responders. These difficulties are due, in part, to the different pathophysiological causes of chronic cervical myelopathy such as ischemic degeneration of the neural tissue due to hypoperfusion, loss of motoneurons in the anterior horns [29], membrane damages, and conduction decline [30]. Presently, we are faced with a true *"paradigm shift"*, passing from an era where the goal of surgery for DCM was to stop the disease progression, to an era where surgery seems to be able to improve patients' status in most of the cases, having thus a favorable impact on DCM patients' quality of life [31].

Unfortunately, there is a lack of a standard diagnostic protocol that can provide prognostic information for DCM patients. Past notions about the usefulness of MRI findings in DCM, such as the presence of T2 hyperintensity, have shown to be largely nonspecific with regards to the severity of myelopathy in specific patients and the capacity of postoperative neurological improvement [8,13,32–34]. Having said this, T2 hyperintensity seems to have some potential utility when measured in terms of sagittal extent, or when skip lesions are observed [11].

It remains unclear if age impacts outcome, while some have demonstrated that age is a predictive variable, others have found no such relationship for the postsurgical recovery rate [35,36]. In our study we found that T2 hyperintensity, Tobbaco use, diabetes, and a clinical history >6 months, are not related to a worse outcome. Furthermore, multilevel compression was also not predictive of a poorer neurological outcome. While, these results are consistent with some studies found in literature [8,37–40], other studies, notably those derived from the AOSpine multicenter studies on DCM, have shown contradictory results [11,41]. Our results may partially be due to our relatively small cohort and may have been underpowered to detect statistically significant differences.

In contrast with the results reported by various authors [41–44], we found that surgical outcome was independent from both the age and the spinal canal diameter before surgery. In fact, we observed no statistical difference between the spinal canal midsagittal diameter of "best responders" and "normal responders", both in the preoperative and postoperative periods. This may be partially due to dynamic injury mechanisms that contribute to DCM and that are not necessarily influenced by canal diameter.

New MRI techniques, such as the DTI, in combination with neurophysiological assessment can help identify those patients with higher probabilities of improving after surgical decompression [22]. Similarly to previous research [45], FA in "best responders" were significantly higher than those of the "normal responders", both preoperatively and 1-year follow-up. Moreover, there was also a statistical difference in the FA values of the most stenotic level in all the phases of the study between the two groups, and a significant relationship between a preoperative FA > 0.55 at the most compressed level and a better clinical outcome (RR > 50%) at 1-year. This is consistent with the concept that FA can be used in the assessment of the degree of severity of DCM with greater accuracy. [12–14,34,43] Our results support what has been previously suggested—higher values of preoperative FA can be considered as a positive prognostic factor of functional recovery [13,34,37]. It was interesting to note that, FA values were not different between best and normal responders at 3-months, and this potentially indicates a 3-month FA MRI may be too early to detect changes.

Neurophysiological parameters have been reported as useful diagnostic tools for DCM. It has been previously shown that MEPs have a good diagnostic value in the setting of DCM, perhaps even more so than SSEPs [19]. Our study shows support to this concept in that preoperative abnormal CCT was related to a worse clinical condition and a lower mJOA (*p* = 0.0004). The significant correlation between preoperative FA and N22, N8–N22 latencies in the preoperative period are challenging to interpret—further research needs to be done in this area to better understand this relationship and how it may be useful in diagnosis and outcome prediction, or if this was a spurious finding.

Several authors stated that median SSEPs can be used for both diagnostic and prognostic purposes in DCM patients. Lyu et al. [20] stated that normal median SSEPs are related to a better prognosis; Morishita et al. [21] showed that an early improvement of the N18 was related to a good outcome at 3-months after surgery. Restuccia et al. [19] found that the SSEP improvement was related to a better clinical outcome, especially in those patients with an isolated loss of N13. These results confirm that SSEPs are a useful tool for the diagnosis of a cervical myelopathy. Unfortunately, in our study we were not able to perform a complete SSEP analysis in all our patients in the postoperative period. Nevertheless, we observed in the preoperative data, a correlation between fractional anisotropy and neurophysiological parameters and in particular tibial SSEPs, suggesting the existence of connection a between a radiological information and a clinical data.

### *Limitations*

The greatest limitation of the present study is the relatively small cohort and the retrospective nature of our analysis. However, few studies have assessed the combination of electrophysiology in conjunction with the commonly used clinical and advanced MRI parameters. While our study did not support some of the findings of previous authors that remain supported with low evidence, our findings with regards to FA are in accordance with others [34,45,46] and strongly support the effectiveness of DTI analysis in both assessing the clinical status and predicting the surgical outcome in DCM patients.

Unfortunately, the use of different neurophysiological methods in the postoperative period did not allow us to assess the effectiveness of changes in these parameters in the postoperative setting.

Lastly, patients with preoperative mJOA scores of 17 were included; however, while these patients did not show clear neurological impairment as assessed by the mJOA, they exhibited neurophysiological evidence and/or objective clinical signs such as hyper-reflexia, Hoffman's sign.

### **5. Conclusions**

Our results validate the concept that the current "ordinary" assessment of DCM should be upgraded with new diagnostic techniques. DTI could be considered not only a complementary diagnostic analysis, but rather a crucial tool in order to identify the best candidates to surgery. Neurophysiological parameters, in particular MEPs, correlate with the clinical condition of the patient, and could therefore be considered as an additional diagnostic tool in the preoperative period. The inclusion of DTI sequences in the preoperative study provides also prognostic information, enhancing the presurgical evaluation of DCM patients. Our findings suggest that FA values are most useful preoperatively and at 1-year follow-up, and may not be useful at 3-months postoperatively. The inclusion of electrophysiology and DTI measurements may enhance the diagnostic process and may be effective at augmenting the predictive capacity of previously described prediction models.

**Author Contributions:** Conceptualization and methodology: E.T. Formal analysis, data curation: R.S. Writing—original draft preparation, R.S., E.T. Writing—review and editing: E.T, A.N., R.S. Visualization, supervision: E.T., A.N. All authors have read and agreed to the published version of the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

1. Matz, P.G.; Anderson, P.A.; Groff, M.W.; Heary, R.F.; Holly, L.T.; Kaiser, M.G.; Mummaneni, P.V.; Ryken, T.C.; Choudhri, T.F.; Vresilovic, E.J.; et al. Cervical laminoplasty for the treatment of cervical degenerative myelopathy. *J. Neurosurg. Spine* **2009**, *11*, 157–169. [CrossRef] [PubMed]


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
