Pediatric Radiology

, Volume 42, Issue 6, pp 685–691

Comparison of standard, prone and cine MRI in the evaluation of tethered cord

Authors

  • Sukhjinder Singh
    • Department of RadiologyCohen Children’s Medical Center
  • Beth Kline-Fath
    • Department of RadiologyCincinnati Children’s Hospital Medical Center
  • Karin Bierbrauer
    • Department of NeurosurgeryCincinnati Children’s Hospital Medical Center
  • Judy M. Racadio
    • Department of RadiologyCincinnati Children’s Hospital Medical Center
  • Shelia Salisbury
    • Division of Biostatistics & EpidemiologyCincinnati Children’s Hospital Medical Center
  • Maurizio Macaluso
    • Division of Biostatistics & EpidemiologyCincinnati Children’s Hospital Medical Center
  • Elizabeth C. Jackson
    • Department of Pediatrics, Division of NephrologyCincinnati Children’s Hospital Medical Center
    • Department of RadiologyCincinnati Children’s Hospital Medical Center
    • Department of RadiologyPhoenix Children’s Hospital
Original Article

DOI: 10.1007/s00247-011-2308-8

Cite this article as:
Singh, S., Kline-Fath, B., Bierbrauer, K. et al. Pediatr Radiol (2012) 42: 685. doi:10.1007/s00247-011-2308-8

Abstract

Background

Tethered cord syndrome (TCS) is defined by abnormal traction on the spinal cord that confines its movement. Surgical cord release usually stops neurological deterioration; therefore, early and accurate neuroradiological diagnosis is important. Supine MRI is the imaging modality of choice, but prone MRI and cine MRI can demonstrate cord movement.

Objective

We compared the diagnostic accuracies of standard MRI, prone MRI and cine MRI in patients with clinical suspicion of TCS and evaluated inter-reader reliability for MR imaging.

Materials and methods

Children who underwent MRI for suspicion of TCS were retrospectively identified. Supine, prone and cine MRI studies were re-read by two pediatric neuroradiologists. Conus level, filum appearance and cord movement were documented.

Results

Thirteen of 49 children had tethered cord documented at surgery. Conus level had the highest diagnostic accuracy (sensitivity 69–77%, specificity 94%, positive predictive value 82–83%, negative predictive value 89–92%, correct diagnosis 88–90%) and highest between-reader concordance (98%). Prone and cine MRI did not add to the accuracy of the supine imaging.

Conclusion

Conus level provides the highest diagnostic accuracy and inter-reader reliability in TCS. Until a larger series is evaluated, it remains questionable whether prone or cine MRI provides enough additional diagnostic information to warrant routine use.

Keywords

Tethered cord syndromeMRIProne MRICine MRI

Introduction

Tethered cord syndrome (TCS) is a neurosurgical disorder caused by abnormal attachment or tethering of the spinal cord that confines the movement of the cord within the spinal canal. Causes of tethering include spinal dysraphisms such as myelomeningocele, lipomyelomeningocele and split cord malformation [1]. In addition, thickened and/or fatty filum terminale, prior surgery/trauma or tumor may lead to tethering of the cord. In TCS, the conus medullaris is frequently low in position (tip located below the L2-L3 interspace) [24]. Tethering of the spinal cord causes abnormal stretching and tension, which results in a variety of neurological symptoms. Natural history without treatment is usually characterized by deterioration of neurological function with increasing age [2, 5]. Surgical release of the tethered cord frequently stops the clinical deterioration and may reverse symptoms [1, 5]. Therefore, accurate neuroradiological diagnosis is very important for the clinician to confirm diagnosis and define treatment.

MR imaging is the imaging study of choice to confirm a diagnosis of tethered cord. Supine MR can delineate the level of the conus as well as the size and presence of fatty infiltration in the filum terminale. Additionally, any obvious tethering causes such as intraspinal tumor can be identified (Fig. 1). MR imaging is also useful for surgical planning. Prone MRI has been used to define movement of the conus within the spinal canal [3] (Fig. 2); in children with normal supine imaging, lack of movement on prone MRI indicates probability of TCS. Cine MRI is a technique in which movement of the conus in cephalocaudad and anterior-posterior planes can be subjectively evaluated. Therefore, it may provide additional dynamic information about spinal cord movement to aid in diagnosis.
https://static-content.springer.com/image/art%3A10.1007%2Fs00247-011-2308-8/MediaObjects/247_2011_2308_Fig1_HTML.gif
Fig. 1

Images in a 6-month-old with sacral dimple; MR findings considered positive for tethering. a Supine sagittal FSE T2-W image demonstrates tip of conus lying in the dorsal 1/3 of the spinal canal at the L2-L3 level with abnormal hypointense signal in the filum (arrow). b Prone sagittal single-shot FSE image demonstrates minimal anterior translation of conus (arrow), still lying in the dorsal 1/3 of the canal. c Axial T1-W image confirms a distal filar lipoma

https://static-content.springer.com/image/art%3A10.1007%2Fs00247-011-2308-8/MediaObjects/247_2011_2308_Fig2_HTML.gif
Fig. 2

Images in an 8-year-old with urinary and bladder incontinence; MR findings considered negative for tethering. a Supine sagittal fast spin-echo (FSE) T2-W image shows a normally positioned cord with tip at T12-L1 level lying in the dorsal 1/3 of the spinal canal. b Prone sagittal single-shot FSE image demonstrates anterior translation of conus and cauda equina to the anterior 1/3 of the spinal canal

In this study, we compared the predictive value of standard supine MRI, MRI in prone position and cine MRI in a group of children with clinical suspicion of TCS. We also evaluated inter-reader reliability for the MR imaging.

Materials and methods

After institutional review board approval, a keyword search of the internal hospital medical notes system was conducted to identify children who underwent radiological evaluation for possible tethered spinal cord between August 2002 and July 2008. A retrospective review of the imaging and clinical database of these children was performed. To evaluate the utility of different types of MR imaging in children in whom the diagnosis is uncertain, those with definite dysraphic lesions were excluded. In addition, because children with anorectal malformations may have signs and symptoms that overlap with those associated with tethered cord, those children were excluded. Therefore, children were excluded for (1) history of open spinal dysraphisms or prior spinal surgery, (2) dysraphic lesions other than a fatty or thickened filum (e.g., lipomyelomeningocele, split cord malformation, myelocystocele), (3) presence of an anorectal malformation, or (4) no clinical follow-up information available. Patient demographics, clinical findings such as cutaneous stigmata, bowel/bladder incontinence or gait disturbances, and urodynamic testing results were recorded.

The initial search identified 400 children who underwent imaging for possible TCS; 350 patients met exclusion criteria. This study comprised 50 patients, including 32 girls and 18 boys, age range 3 days to 18 years (median 6 years).

All of the children underwent MRI with a tethered cord protocol, which included sagittal T1 and FSE T2 and axial T1 sequences through the lumbar spine in the supine position, with sagittal T2 EPI sequence in the prone position. Phase-contrast cine MR images were also performed in some cases. The sensitivities, specificities and predictive values of each imaging modality were calculated. Each imaging study was re-read by two pediatric neuroradiologists who were blinded to clinical information, the original radiology report and each other’s findings, with their findings compared to determine inter-reader reliability.

On supine images, the level of the conus, thickness of the filum at L5-S1, presence of fatty infiltration of the filum and presence of any obvious tethering mass were documented. A conus level above L2-L3 in children older than 3 months was considered normal; because of the variability of conus level in young infants, it was not evaluated in children younger than 3 months of age. Filum thickness less than 2 mm at L5-S1 was considered normal. Based on these findings, the MRI examinations were reported as “positive,” “negative” or “indeterminate” for tethered cord.

For comparison purposes, the anterior-to-posterior distance of the spinal canal on a sagittal view was divided into anterior, middle and posterior thirds. The position of the conus was documented on supine images and then compared to its position on prone images. Movement of at least 1/3 of the distance (e.g., from the posterior 1/3 of the canal to the middle 1/3 of the canal) was considered “negative” for tethered cord. Examinations with some movement, but less than 1/3 of the AP dimension, were considered “indeterminate.” No movement was considered “positive” for tethered cord. On cine MR imaging, presence of AP and/or cephalocaudal movement of the cord was noted; presence of movement was reported as “negative” for tethering, absence of movement as “positive” and others “indeterminate.”

Children were considered to have a clinical diagnosis of tethered cord based on findings of skin abnormalities, gait abnormalities, abnormal muscle tone, back pain and/or urological abnormalities, often confirmed with urodynamics. Those children underwent surgical untethering. Follow-up data were obtained from clinical and imaging records. Clinical course, surgical interventions and surgical findings were documented.

Measures of diagnostic accuracy were computed by comparing each reader’s assessment of a diagnostic criterion with the conclusions based on clinical judgment and follow-up. Sensitivity was computed as the proportion of subjects with confirmed TCS who were classified as positive by a reader using a specific criterion. Specificity was computed as the proportion of subjects without TCS who were classified as negative by a reader using a specific criterion. Positive predictive value (PPV) was computed as the proportion of subjects classified as positive by a reader using a specific criterion who had confirmed TCS. Negative predictive value (NPV) was computed as the proportion of subjects classified as negative by a reader using a specific criterion who were confirmed to be without TCS.

Inter-reader reliability was examined using Cohen’s simple kappa coefficient for nominal variables. By convention, values between 0.21 and 0.40 indicated fair agreement, between 0.41 and 0.60 indicated moderate agreement, between 0.61 and 0.80 indicated substantial agreement and between 0.81 and 1.00 indicated almost perfect agreement. The 95% confidence interval of the kappa coefficient was computed as a measure of its precision.

The number of subjects correctly diagnosed (either positive or negative) was also computed to compare the overall performance of the readers and of the diagnostic criteria. To evaluate the within-reader relative performance of two tests, each subject was classified as correctly or incorrectly diagnosed according to each test, and the statistical significance of the discordance between tests was evaluated with McNemar chi-square test for paired samples.

Results

Clinical indications for MR imaging are summarized in Table 1. All 50 children underwent both supine and prone MR imaging; in addition, 26 children underwent cine MR imaging. Thirteen underwent surgical intervention for suspected tethered cord based on clinical findings; all of those children were confirmed to have tethered cord. Thirty-six of the remaining 37 children were not thought to have tethered cord based on clinical findings; they were followed clinically. One child was clinically suspected of having tethered cord and surgery was recommended; however, the child was lost to follow-up and therefore excluded. Five children were younger than 3 months of age and were therefore excluded from the evaluation of conus level.
Table 1

Clinical indications for MR imaging

Clinical indication

Number of patients

Sacral cutaneous abnormality (e.g., sacral dimple, hair tuft)

21

Voiding dysfunction (e.g., incontinence, urgency, multiple UTIs)

21

Back pain

4

Lower extremity neuro-orthopedic abnormalities (e.g., toe walking, increased or decreased tone, weakness)

11

Rubinstein-Taybi syndrome, Kabuki syndrome

3

Supine MRI was not particularly sensitive (23–46%) or specific (69–72%) in identifying children with tethered cord (Table 2). However, a lack of findings related to tethered cord (low level of conus medullaris, fatty filum) on supine MRI had a negative predictive value of 87–93%. Prone MRI and cine MRI, used to evaluate the presence of cord movement, did not have any higher sensitivity, specificity or predictive value than supine MRI (Tables 3 and 4). When used individually, the filum appearance (thick, fatty infiltration, lipoma) had very little association with the presence of tethered cord (Table 5), although there was excellent between-reader agreement. The level of the conus medullaris appeared to offer the best combination of sensitivity and specificity, resulting in high diagnostic accuracy (88–90% of the children were correctly diagnosed) as well as very high reliability, with 98% concordance in diagnostic assessments and a high kappa: 0.94 (95% CI: 0.83–1.0) (Table 6). The kappa statistic is influenced by disease prevalence and by the balance of discordant assignments made by the raters (bias). In general, the value of kappa is lower for extreme prevalence levels and for low levels of bias. Prevalence of tethered cord was similar across Tables 2, 3, 4, 5 and 6 (27% in Tables 2, 3 and 5, 6, 23% in Table 4), and could not affect the comparison of kappa estimates. Bias was more variable, but it was lowest for conus level ratings, suggesting that the high degree of concordance between raters may have been underestimated in this case. Kappa estimates computed for other tables were low because of high frequencies of discordant ratings between raters even in the presence of larger bias. Thus, bias could only have little impact on the interpretation of the findings pertaining to inter-rater agreement. Statistical comparisons of the relative accuracy of the five assessments yielded a very coherent pattern, confirming the superiority of the assessment of the conus medullaris level as compared to all other tests evaluated in this study (Table 7).
Table 2

Measures of diagnostic accuracy and inter-reader agreement of supine MRI, by reader

 

Reader 1

Reader 2

Patients with tethered cord (n = 13)

No tethered cord (n = 36)

Patients with tethered cord (n = 13)

No tethered cord (n = 36)

Positive supine MRI

6

2

3

0

Negative supine MRI

4

26

2

25

Indeterminate supine MRI

3

8

8

11

 

Sensitivity: 46% Specificity: 72%

Sensitivity: 23% Specificity: 69%

NPV: 87% PPV: 75%

NPV: 93% PPV: 100%

Correct diagnosis: 65%

Correct diagnosis: 57%

Between-reader concordance: 63%; kappa: 0.35 (95% CI: 0.13–0.57)

Table 3

Measures of diagnostic accuracy and inter-reader agreement of prone MRI, by reader

 

Reader 1

Reader 2

Patients with tethered cord (n = 13)

No tethered cord (n = 36)

Patients with tethered cord (n = 13)

No tethered cord (n = 36)

Positive prone MRI

8

10

8

3

Negative prone MRI

4

22

3

24

Indeterminate prone MRI

1

4

2

9

 

Sensitivity: 62% Specificity: 61%

Sensitivity: 62% Specificity: 67%

NPV: 85% PPV: 44%

NPV: 89% PPV: 73%

Correct diagnosis: 61%

Correct diagnosis: 63%

Between-reader concordance: 69%; kappa: 0.49 (95% CI: 0.29–0.69)

Table 4

Measures of diagnostic accuracy and inter-reader agreement of cine MRI, by reader

 

Reader 1

Reader 2

Patients with tethered cord (n = 6)

No evidence of tethered cord (n = 20)

Patients with tethered cord (n = 6)

No evidence of tethered cord (n = 18)

Positive cine MRI

3

1

4

2

Negative cine MRI

1

9

0

11

Indeterminate cine MRI

2

10

2

5

 

Sensitivity: 50% Specificity: 45%

Sensitivity: 67% Specificity: 61%

NPV: 90% PPV: 75%

NPV: 100% PPV: 67%

Correct diagnosis: 46%

Correct diagnosis: 62%

Between-reader concordance: 57%; kappa: 0.34 (95% CI: 0.06–0.63)

Table 5

Measures of diagnostic accuracy and inter-reader agreement of filum characteristics, by reader

 

Reader 1

Reader 2

Patients with tethered cord (n = 13)

No evidence of tethered cord (n = 36)

Patients with tethered cord (n = 13)

No evidence of tethered cord (n = 36)

Fatty/lipoma

6

11

7

12

Normal

7

25

6

24

 

Sensitivity: 46% Specificity: 69%

Sensitivity: 54% Specificity: 67%

NPV: 78% PPV: 35%

NPV: 80% PPV: 37%

Correct diagnosis: 63%

Correct diagnosis: 63%

Between-reader concordance: 96%; kappa: 0.91 (95% CI: 0.79–1.00)

Table 6

Measures of diagnostic accuracy and inter-reader agreement of conus medullaris level, by reader

 

Reader 1

Reader 2

Patients with tethered cord (n = 13)

No evidence of tethered cord (n = 36)

Patients with tethered cord (n = 13)

No evidence of tethered cord (n = 36)

Below the L2-L3 interspace

9

2

10

2

At or above the interspace

4

34

3

34

 

Sensitivity: 69% Specificity: 94%

Sensitivity: 77% Specificity: 94%

NPV: 89% PPV: 82%

NPV: 92% PPV: 83%

Correct diagnosis: 88%

Correct diagnosis: 90%

Between-reader concordance: 98%; kappa: 0.94 (95% CI: 0.83–1.0)

Table 7

Selected differences in diagnostic accuracy between tests, by reader

 

Reader 1

Reader 2

Comparisons between testsa

Correctly diagnosed

P-valueb

Correctly diagnosed

P-valueb

Conus vs. supine MRI

88% vs. 65%

0.001

90% vs. 57%

<0.0001

Conus vs. prone MRI

88% vs. 61%

0.003

90% vs. 65%

0.003

Conus vs. cine MRIc

92% vs. 46%

0.001

92% vs. 63%

0.03

Conus vs. filum

88% vs. 63%

0.01

90% vs. 63%

0.01

Cine MRI vs. supinec

46% vs. 77%

0.01

63% vs. 50%

NS

aAll other between-test comparisons were not significant

bBased on McNemar chi-square test

cBased on 26 subjects

Discussion

Tethered cord syndrome results from traction on the conus medullaris; this occurs most commonly with spinal dysraphism [2]. The clinical diagnosis of TCS remains complex, with a variety of clinical presentations and tethering lesions. The age at presentation, underlying cause, specific symptoms, and the severity and progression of TCS vary from one individual to another. While the presence of cutaneous manifestations of spina bifida occulta may be the only symptom in infants, toddlers and children commonly have both motor and sensory dysfunction [2, 6]. In some children, abnormal skeletal or skin findings and/or neurological deficits including scoliosis, lower extremity spasticity or deformity, gait abnormality and/or characteristic back or leg pain are highly suggestive of TCS. Bowel and bladder symptoms also occur, with the most common finding being detrusor hyperreflexia [2, 6]. Urodynamic testing can be helpful in confirming sphincter dysfunction and documenting clinical course during observation or following a detethering procedure [2, 6]. Thus, the clinical diagnosis may be based on a number of factors. Because of this, the accuracy of the clinical diagnosis is difficult to determine. The clinical diagnosis of TCS is supported by imaging findings in virtually all patients. Presurgical imaging is used to confirm the diagnosis of a low-lying conus or tethering lesion, unless one is operating on a child with proven progressive neurogenic bladder [7, 8]. At surgery, one would encounter and identify the tethering lesion, such as a thickened or fat-infiltrated filum terminale. Direct visualization of such a structure would occur after making the durotomy, with the abnormally taut appearance of the cord and tethering lesion usually located more dorsally in the dural sac.

US of the spine has been described since the early 1990s. This imaging modality is most useful in children up to 6 months of age, after which there is loss of the acoustic window in the presence of ossification of posterior spine elements [9]. Gray-scale (B mode) imaging has been shown to be comparable to MRI in the assessment of TCS [10, 11]. M-mode imaging, which has been utilized to analyze motion of the conus and cauda equina, has proven to be an asset in documenting TCS by demonstrating impaired pulsations or low oscillation amplitude [12, 13]. However, US has inherent limitations related to soft-tissue contrast, smaller field of view and potential loss of acoustic window.

MRI remains the preferred imaging modality in assessing most children for suspected TCS. MR imaging may demonstrate the level of the conus, show the cause of tethering and provide anatomical detail for surgical planning [6]. However, “minimal” or “occult” TCS has also been described, in which children have neurogenic voiding dysfunction and back pain, without a low conus on MR imaging. In those cases in particular, it may be beneficial to have additional imaging information beyond standard sequences to aid in the diagnosis of TCS. Although controversial, recent trends favoring more aggressive surgical approaches for children with minimal or no imaging abnormalities, but clinical symptoms of tethering such as neurogenic bladder, have produced encouraging results in some series. A thorough history and physical examination is crucial in evaluating young patients who present with neurological or urological signs or symptoms suggestive of tethering, even with unremarkable imaging studies [7, 8, 14, 15].

In this study, we examined standard supine imaging, including level of the conus medullaris and filum characteristics, as well as the potential utility of prone MRI and cine MRI in a group of children without obvious spinal dysraphism who had clinical symptoms that could be consistent with TCS. In those children, the finding that had the highest diagnostic accuracy was the level of the conus medullaris. The filum appearance did not correlate well with the presence of tethered cord; the level of the conus was significantly more accurate than the overall supine MRI reading that took into account both conus level and filum appearance. Prone MRI and cine MRI did not increase the diagnostic accuracy of the supine imaging. In addition, significant inter-reader variability was noted in the imaging of tethered cord. Documentation of the level of the conus medullaris, however, had very high concordance between readers.

Additional techniques, such as prone MRI and cine MRI, have been utilized to try to provide surgeons with as much diagnostic information as possible to aid in clinical decision-making. Although it has been suggested in a small study that prone MR imaging can provide additional information in children clinically suspected of having tethered cord and in whom supine MR imaging depicted no abnormalities [3], our study indicates that this technique did not provide significant additional information beyond what is gained by supine MRI alone. Prone imaging also requires repositioning of the child, a relocalization sequence, additional imaging time, and added anesthesia time or possibly additional sedation. Prone positioning was not evaluated in children with a clinical concern of re-tethering, but this could be a topic for future investigation. While cine MRI was not found to be particularly useful in the initial evaluation of children for tethering, it may also be considered as an adjunct to clinical decision-making in the challenging subgroup of children with postoperative re-tethering.

However, as related to the postoperative child with re-tethering, both prone and cine MRI sequences remain open to further investigation.

This study was limited by the small number of patients, as well as the subjective method of evaluation of prone and cine MRI. Although the reviewers were blind to the diagnostic conclusion, the diagnosis was not independent from the MRI results, which introduced some bias into the results.

Conclusion

In children with clinical suspicion of tethered cord syndrome, measurement of the level of the conus medullaris on supine MRI provides the highest diagnostic accuracy as well as the best inter-reader reliability. Prone MRI and cine MRI, while potentially yielding information the clinician may find helpful in individual cases, have not proved to be statistically reliable between interpreters, nor have their reliability and validity been confirmed in determining which patients with clinical suspicion of tethered cord would benefit from surgical untethering. Therefore, until a larger series can be collected, it remains an open question whether these additional sequences provide enough additional diagnostic information to warrant their routine use.

Copyright information

© Springer-Verlag 2011