Since Caffey first described the association between fractures of the appendicular skeleton and subdural haematomas (SDHs), there has been increasing awareness of the problem of child abuse. Non-accidental head injury (NAHI) is a major cause of mortality in the first year of life, and a significant proportion of survivors have severe developmental disability.1^–3 The diagnosis is based on a combination of clinical, social and radiological features. The radiological findings are often critical to the diagnosis and the litigation process.

The purpose of our study was to examine the incidence of spinal pathology detected by MRI in infants with NAHI.

MATERIALS AND METHODS

All infants in the study were admitted directly or after transfer from regional hospitals to our tertiary neurosciences centre, which has a catchment population of 2.8 million. The radiological images and clinical notes were examined of all cases of NAHI between 2000 and 2007 for which MRI of the spine was performed. Eighteen infants were included in the study (11 male, seven female; age 1–12 months (mean age 3 months)). All infants were referred to the local child protection services and proceeded through care or criminal legal proceedings. All the infants were subsequently made subjects of care orders, and the diagnosis of NAHI confirmed. This cohort of infants formed part of a larger study that had local ethics committee approval. Cases that were initially thought to be suspicious but subsequently not proven to be non-accidental in nature were excluded. In all cases, the MRI of the spine was performed before a lumbar puncture.

Between 2000 and the end of 2004, MRI of the spine was performed only when there were features suspicious of spinal injury (seven cases). After 2005, MRI of the whole spine was routinely performed in all children with suspected NAHI (11 infants).

All 18 had a CT of the head at presentation, followed by MRI of the head and spine within the next 1–3 days. MRI was undertaken on 1.5 T clinical imaging systems (GE Excite, Milwaukee, Wisconsin, USA; Philips Intera, Eindhoven, The Netherlands), using a four-channel small paediatric array coil. MRI of the spine included T₁-weighted and T₂-weighted sagittal imaging accompanied by axial T₁ and T₂ images where appropriate. Follow-up MRI was performed according to the clinical situation within 1–3 months.

RESULTS

Eight of 18 infants (44%) had spinal SDHs (sSDHs). Two of seven patients who were imaged up to 2004 had spinal subdural collections, and six of 11 from the later cohort. All had supratentorial and infratentorial intracranial SDHs.

Of the eight cases of sSDH, six were characterised as large, extending from the sacral thecal cul-de-sac as high as the cervical spine in two cases, up to the thoracic spine in three cases, and up to the upper lumbar spine in one. Two were characterised as small, being streak-like in nature, one tracking along the posterior aspect of the cervical spine, and one similarly in the thoracolumbar region (figs 1 and 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 T₁-weighted sagittal image. Large hyperintense spinal subdural haematoma in the thoracolumbar region. Note the subdural collection of the same intensity in the posterior fossa (arrow).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 T₂-weighted sagittal image. Small T₂ hyperintense spinal subdural haematoma at the thoracolumbar spine posteriorly (black arrows). Note the slightly reduced height of the body of T10 accompanied by mildly increased signal intensity (white arrow).

In all cases, the signal intensity of the sSDH on T₁-weighted and T₂-weighted sequences was homogeneous and identical with that of the posterior fossa subdural collections (fig 3). In contrast, the supratentorial collections often had mixed-signal characteristics. These were interpreted as being either of mixed age or variable maturation of single-aged SDH, or acute traumatic subdural effusions; it was therefore not possible to correlate the signal intensity of supratentorial SDHs with the sSDHs. In two cases, anatomical continuity was present between the posterior fossa and the spinal collections (fig 4). No spinal cord abnormality was demonstrated in any of the infants.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 T₁-weighted images. Slightly hyperintense spinal subdural haematoma tracking the intensity of the posterior fossa subdural haematoma (arrows).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4 T₁ hyperintense subdural haematoma of the posterior fossa tracking into the cervical spine.

Three of the eight children with sSDHs (all large) had skull fractures. One had a large V-shaped skull defect and two had non-depressed parietal fractures. Five of the 10 children without sSDHs had skull fractures.

Two of the 18 children had fractures of the thoracic spine, both of which were depicted on plain radiography. One of these two vertebral body fractures had a small associated epidural haematoma. No paraspinal hematomas or extraosseous injuries of the vertebral column were identified; however, fat-suppressed sequences were not performed.

In six of the eight cases of sSDH, follow-up MRI of the spine was performed. In three cases, the MRI was performed within the first month after presentation and showed reduction in the size of the collection. In the other three cases, the MRI was performed 3 months after the initial admission. In one case, there was complete resolution of the sSDH, and in the other two there was marked contraction and T₂ hypointensity of the thecal sac (fig 5).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5 (A) T₁-weighted and (B) T₂-weighted images of a lumbar spinal subdural haematoma. (C) T₁-weighted and (D) T₂-weighted images 3 months later show marked contraction of the thecal sac.

DISCUSSION

In the context of suspected NAHI, imaging of the spine in most centres, and according to the American College of Radiology, comprises lateral views of the spine taken as part of a skeletal radiographic survey.4^–6 The aim in such cases is to assess the osseous component of the vertebral column. Spinal fractures are a well-recognised feature of NAHI,7^–10 although it is generally reported as an uncommon manifestation, and usually associated with other injuries.11^–13

Extraosseous spinal pathology has also, but less commonly, been reported in association with child abuse.8 14 15 Spinal pathology in a brain-injured child is difficult to recognise clinically.15 16 The incidence of clinically significant cord injury is uncertain, as there are only isolated case reports or small case series. It may be more common than currently postulated. Hadley et al17 in 1989 reported the post-mortem findings in six children with NAHI and intracranial SDHs and found that five of them had extra-axial cervical spinal haematomas and four had spinal contusions. Feldman et al18 in 1997 compared the MRI results for the cervical spine and post-mortem findings in five children with NAHI. At autopsy one had a thin SDH and three had subarachnoid haemorrhage at the level of the cervical spine. None of these findings was visible on MRI, which in four of the five cases where undertaken was performed post mortem. Rutty et al19 performed post-mortem examination of the spine in eight children, two of whom were victims of NAHI, the other six having no evidence of trauma. sSDHs were seen in both the cases of NAHI but in none of the second group. In the series of Soto-Ares et al,20 eight of the children with suspected NAHI and bilateral SDHs had spinal imaging. One infant had a definite, and a second a probable, perimedullary haematoma in the thoracolumbar spine. Although described as epidural haematoma in the English translation of the text, we suspect that the collections were in fact subdural in location.

In our case series, there was a high incidence of clinically occult sSDH. In most of these cases, the haematoma pooled in the thoracic or lumbar spine and would not have been detected during an MRI examination of the brain except that, fortuitously, a sufficient extent of coverage of the spine was achieved on sagittal imaging of the brain coupled with heightened awareness of what may be subtle intraspinal pathology. Except for the series of Soto-Ares et al, there is limited literature specifically addressing MRI of spinal pathology in abused children with associated intracranial SDHs.

The origin of the sSDH in NAHI is unclear. There are reports in adults of simultaneous occurrence of spinal and intracranial SDHs.21^–25 The spinal subdural space lacks bridging veins as an origin for the haemorrhage.26 Several theories have been proposed, none of which has been proven.21 24 An appealing theory is that sSDHs are extensions of their intracranial counterparts. The signal characteristics of the sSDHs tracked that of the posterior fossa collections. Lecouvet et al23 demonstrated by MRI the anatomical continuity of an sSDH in the lumbar region with cranial extension that reached the posterior fossa. The fact that we also noticed anatomical continuity in two of eight cases shows that the above theory is anatomically feasible. It is possible that anatomical continuity was not seen in the rest of the cases because blood tends to pool in the most dependent portion of the spine, the thoracolumbar region in preference to the cervical region, where there is a natural convexity in the supine position.

The consequences of our observation have not been elucidated. In the adult, sSDHs are associated with neurological symptoms and have been managed surgically in many cases.21 27 28 Many infants with NAHI have severe brain injury, which can mask signs of spinal cord compromise. None of the sSDHs in our study were evacuated surgically, and, in the cases that were followed up in the first month, there was resolution of the collections and no radiologically depicted cord abnormalities. It appears that sSDHs are more common in the thoracic and lumbar spine than in the cervical region. sSDH in the neonate is an uncommon entity which we have not come across in our institution in a setting other than non-accidental trauma. Although it would be interesting to see if sSDHs occur in traumatic intracranial collections or other medical conditions, the design of such a study would raise issues of ethical propriety. Morphological changes in the thecal sac were seen in the longer term in two of our cases, both associated with large sSDHs, but again the significance of this finding is not clear. However, it is important to recognise the potentially common and occult occurrence of sSDHs in NAHI in order to be able to evaluate their long-term clinical implications.

In conclusion, our series has identified a high incidence of occult and previously unsuspected spinal pathology in children with NAHI. Further work is required to evaluate the clinical implications.