The spinal segment involvement encountered in our study (as forementioned in the results above), were comparably concordant with Tawfik et al. [12] and Kommana et al. [10] whom both showed that the LSS was the most commonly affected region in 71.1% and 28.5%, followed by the dorsolumbar spine in 17.8% and 21.4%, and the sacrococcygeal spine in 11.1% and 14%, respectively. Whereas Dhingani et al. [13] showed that the LSS was the most common involved in 52.6% of patients, followed by the sacrococcygeal spine in 34.2%.
On the contrary to the Tawfik et al. [12], Kommana et al. [10], and Dhingani et al. [13] studies which included cases of both open and closed SD, we only included cases of CSD to be more precise and since OSD actually requires no imaging.
The most common subtypes of SD encountered in our study are presented in Table 2. Both Tawfik et al. [12] and Dhingani et al. [13] showed comparable similar prevalences of the forementioned CSD subtypes, apart from the mere 6.6% for tethered cord in Tawfik et al. [12] as opposed to the compelling 60% in our own cohort.
Non-spinal associated anomalies confronted in our study varied in diversity and frequentness compared to other studies as discussed in detail in results section. Ruangtrakool et al. [14] documented fecal incontinence, urinary incontinence, motor symptoms, gait abnormalities, and scoliosis, in order of frequency. Whereas Tawfik et al. [12] documented neurological abnormalities, urinary incontinence, hydrocephalus, and Chiari malformation, in order of frequency.
Many of these associated non-spinal anomalies can probably be attributable to the tethered cord syndrome, whereby the fixation of the filum terminale causes secondary traction injury unto the spinal cord.
Our study, along with Ruangtrakool et al. [14], Oh et al. [15] and Jehangir et al. [16], implemented the Krickenbeck's classification system for ARMs.
In our study, we found that 57.9% of the screened patients with ARMs had coexistent spinal dysraphism. This prevalence ranged 16.2% in Ruangtrakool et al. [14], 22% in Oh et al. [15] and 42% in Esposito et al. [17] studies.
The most common subtypes of ARM encountered in our study (as detailed in Table 1) were similar to ARM subtypes encountered in other studies conducted by Ruangtrakool et al. [14], Jehangir et al. [16], and Minneci et al. [18].
In our study, the most common subtype of ARM associated with coexistent spinal dysraphism are enumerated in Table 3. In comparison with other similar studies [14, 16, 18] the most common ARM subtypes associated with SD were the cloacal malformation and rectourethral fistulas, whereas the least common was the perineal fistula. However, these studies differed from ours, in showing a highly common association between the rectovesical fistula, which was not included in our study population, and SD.
The most common subtypes of SD coexisting in these patients with ARM, in both our study and other studies [14, 18], were tethered cord and segmentation/vertebral spine anomalies.
Spinal dysraphism has been reported to frequently coexist with cutaneous stigmata, ranging from as low as 5% in Resmi et al. [19] and up to 26% and 33% in Shields et al. [20] and O'Neil et al. [21], respectively. However, the wide variation in the morphological appearances of the lumbosacral and coccygeal regions along with the sparsity of the literature regarding the predictive value of these variations, creates a diagnostic dilemma when it comes to screening neonates [9]. On the contrary, a vast majority of patients (about 50–80%) with SD are presented with cutaneous stigmata, as documented by Shields et al. [20] and Choi et al. [22]. This was in concordance with the 55% confronted in our study cohort.
In our study population, 16 patients had cutaneous stigmata, who are further categorized in Table 1. The 6 patients with dimples were further categorized as 2 isolated cases and 4 associated with other lesions (1 with a hair tuft, 1 with skin discoloration, 1 with both hair tuft and skin discoloration, and 1 with a hemangioma). Other studies, such as Resmi et al. [19], included many other subtypes of cutaneous stigmata in their study, including café au lait macules, melanocytic and sebaceous nevi, and aplasia cutis.
Spinal dysraphism was coexistent in 11 out of the 16 forementioned patients in our cohort. All the 6 patients with back swellings were associated with SD (1 meningocele, 2 terminal myelocystoceles, 1 teratoma + myelomeningeocele, 1 lipomyelocemingocele, and 1 lipomyelocele). While only 4 out of the 6 patients with dimples were associated with SD (1 dorsal sinus tract, 1 limited dorsal myeloschisis, 1 limited dorsal myeloschisis + intradural lipoma, 1 limited dorsal myeloschisis + tight filum terminale). The single patient with a skin tag was associated with a spinal lipoma.
Resmi et al. [19] revealed meningocele as the most common spinal anomaly. While Shields et al. [20], O'Neil et al. [21], and Choi et al. [22] revealed that low lying conus medullaris and fatty filum terminale were the most common encountered SD subtypes.
Regarding combined cutaneous stigmata, found in 4 patients of our study cohort (3 of whom had two stigmata and 1 had three), we revealed a significantly higher association with SD in patients with combined stigmata rather than single stigmata. This was in concordance with most of the other studies [19,20,21,22].
With sensitivities comparable to those of MRI, spinal USG is growingly being implemented as an initial screening modality in SD [12].
We evaluated the diagnostic accuracy of USG in screening patients with spinal dysraphism by comparing its results with spinal MRI. This is because spinal MRI is well known to be the gold standard in examining the spine and has been proven by Dhingani al. [13] to show a 100% correlation with operative findings.
We concluded from our study that USG showed comparable accuracy to MRI, with a 100% specificity, 81.4% sensitivity, 100% PPV, and 97% NPV. These diagnostic accuracy metrics were comparably concordant with Tawfik et al. [12] revealing 94.5% specificity, 81.8% sensitivity, 84.3% PPV, 86.7% NPV; Jehangir et al. [16] revealing 91% sensitivity, 75% specificity, 80% PPV, and 88% NPV; and Elmesallamy et al. [23] revealing 100% specificity and 83.3% sensitivity compared to gold standard MRI.
Regarding the degree of agreement (full, partial, and no agreement) between the USG and MRI findings, as detailed in our forementioned results, we revealed comparable results to other studies [10, 12, 13] conducting the same categorization of agreement. Whereby we found full, partial, and no agreement in 84.8%, 9.1%, and 6.1%. While Tawfik [12], Dhingani [13], and Kommana [10] et al. found full agreement in 88%, 79.3%, 62.5%, partial agreement in 12%, 20.7%, 25%, and no agreement in 66.7% (patients > 2 yrs.), 0%, 12.5%, respectively.
Regarding the missed USG findings in cases of partial agreement, we and Dhingani et al. [13] both similarly missed cases of tethered cord and intradural lipoma.
Most of the missed cases in our study, limiting the accuracy of USG, were in patients above 6 months of age. However, Jehangir et al. [16] also mentioned that despite the higher detection rates of spinal anomalies by MRI, this has not reflected on surges in surgical intervention. Hence, the significance of those spinal anomalies missed by USG, is still uncertain.
It is important to note that we revealed from our study relatively lower USG sensitivity levels (57.1%) in cases of segmentation/vertebral spine anomalies, USG showed particularly better diagnostic yield in cases of complete agenesis of the spine, such as sacral agenesis. Whereas USG lagged in cases of segmentation spine anomalies, such as butterfly hemivertebrae, or partial hemi-sacral agenesis.
Interestingly enough, we encountered in our study a diagnostic dilemma when differentiating between DST and LDM on imaging. However, guided by the Lee et al. study [24], we managed to solve this dilemma, using their concluded LDM distinct imaging features "a visible intrathecal tract with dorsal tenting of the cord at the tract-cord union", whereas DST is commonly associated with infection and dermoid or epidermoid tumors. Moreover, we were also guided by the Pang et al. study [25], in the subclassification of LDM into saccular, flat (non-saccular), and saccular with transitional skin lesion. The latter, as illustrated in Fig. 7, showing "a squamous epithelium-lined bubble on an otherwise flat skin surface when the patient strained".
Despite the high diagnostic accuracy of spinal USG, we still recommend performing high resolution MRI with thin sections in all patients with a sinus tract, to delineate the whole tract length, its exact termination, and site of attachment to the cord. This is because we faced many challenges when trying to delineate the sinus tracts using USG as shown in Fig. 7 and when attempting to assess their complications, such as meningitis and intraspinal abscesses.
In agreement with Hughes et al. [11], we reported another challenge when using USG to distinguish between echogenic lesions and surrounding cauda equina nerve roots that appear thick and echogenic in neonates, thus leading to high false-positive and false-negative results in such cases as terminal lipomas and fatty fila. Therefore, MRI yields better detection of fatty lesions due to superior tissue characterization.
On the other hand, USG revealed excellent accuracy in the determination of level of conus termination, as well as in the exclusion of cord tethering by examining conus and nerve roots motility by cine B mode and quantified M mode. Ultrasonography also showed high specificity in differentiating solid and cystic masses with superior characterization of the cystic mass contents (nerve endings, cyst within cyst, multi locules, etc.).
Before we wrap up, we would like to emphasize that screening tests are conducted upon asymptomatic populations to highlight the subgroups at risk who require further assessment. Screening tests are applied unto large populations and hence have to be simple, cost effective, readily available, and non-invasive [16]. Spinal USG is a perfect match that fulfils all the above criteria with reasonable accuracy.
Hence, we finally recommend that the high-risk patients for SD who require appropriate screening, are those patients with ARM, particularly high level and complex types, as well as patients with back cutaneous stigmata, particularly atypical dimples and swellings. Ultrasonography should be used as a front-line screening test. If USG is abnormal, equivocal or technically limited, MRI evaluation is advised.
The main limitations of our study were the small study population, the sparsity of some of the included subtypes of ARMs and back cutaneous stigmata, as well as the inclusion of some older patients with ossified posterior spinal elements.