The seven patients whose cases are reported here all suffered damage to more than a single nerve root, three after single-shot spinals and four after CSEs. Their ages ranged from teens to fifties and their weights from 47 to 102 kg. That six of them were obstetric patients probably reflects the author's sphere of practice. In all cases an atraumatic spinal needle, usually a 25 or 27 G Whitacre, was used. In all cases the anaesthetist believed the needle was being inserted at L2−3 or thereabouts. In three cases there was free flow of CSF without any need for needle adjustment, yet all patients reported pain on insertion of the needle, and in only one was there also pain on injection of the anaesthetic. All patients received bupivacaine rather than lidocaine, the former having considerably less propensity to neurotoxicity [3, 4], while some but not all received fentanyl. In all but one case, spinal anaesthesia produced a block such as would be expected from the dose of bupivacaine that was given (Table 1). Case 4 might have been expected to suffer the most severe symptoms as the only case in which the local anaesthetic appeared to have been injected into the cord, with a resulting incomplete block. Indeed, sensory symptoms were the most extensive, but there was no sphincter disturbance and motor loss was short-lived. Symptoms were believed to relate to the spinothalamic tract, although such damage would be unlikely from a spinal needle as the spinothalamic tract is of course on the anterolateral aspect of the spinal cord. Uniquely in this case, no changes were said to be visible on MRI.
In all cases sensory and motor deficit of lower motor neurone distribution relating to one leg, usually with unilateral MRI changes in the conus, followed pain on spinal needle insertion on the same side (Table 2). In no case did the spinal cord appear to be unduly long.
Aetiology of conus lesions
When major neurological symptoms follow neuraxial blockade, the worried clinician turns to imaging techniques to exclude an epidural space-occupying lesion which would require urgent decompression. Epidural haematoma, abscess, prolapsed disc and spinal stenosis have all been reported in such circumstances [5, 6]. Subdural haematoma is also reported after spinal anaesthesia . Compression in the lumbar region, as it is often exerted below the cord, usually gives rise to cauda equina syndrome. No such features were visible on MRI in any of the present cases.
Signs and symptoms relating to the conus medullaris may also result from compression by tumours [8, 9], syrinx , congenital cysts [11, 12] or vascular malformations, or from trauma [13–15], infarction [16–19] and tethered cord [20, 21].
Mathew & Todd  analysed the presentation in 62 patients with tumours in either the cauda equina or the conus. The commonest symptom in both groups was back pain, with bilateral leg pain being more common with conus and unilateral with cauda equina tumours. Unilateral or bilateral leg weakness, usually lower motor neurone, could occur in both groups, with bladder involvement present in 36% of conus and 26% of cauda equina tumours. Anal sphincter involvement was uncommon.
A congenital cyst differs from a syrinx in that the former is a congenital dilatation of the ventriculus terminalis (the terminal portion of the central canal, not normally visible on MRI) and is lined with ependyma . Presenting symptoms are said to be non-specific and include low back pain, sciatica, leg weakness and bladder dysfunction [11, 12]. The MRI appearance of such a cyst is much larger than in the present series, and suggests that it virtually fills the vertebral canal, causing considerable compression of nervous tissue and accounting presumably for the common occurrence of back pain, a symptom that was prominent in case 6 in the present series.
Trauma may result from vertebral fracture , a shearing injury (as seen in leg injuries from motor cycle accidents) causing nerve root avulsion , or occasionally even manipulation  causing vascular damage. Pain and bladder symptoms are prominent .
Infarction of the conus medullaris is uncommon, because the blood supply to the conus is normally secure, being derived from the artery of Adamkiewicz with generous anastomoses. Occlusion of this artery may result in paraplegia, not cauda equina or conus syndrome. However, in a minority of individuals, perhaps 20%, the blood supply to the conus comes from sacral radicular arteries with fewer anastomoses, and may be more vulnerable. The usual victims are the elderly with arterial pathology [16–18]. Pre-eclampsia might have been a possible contributory factor in cases 3 and 5.
Tethered cord syndrome results from a congenital anomaly; it may occur in isolation, in some forms of scoliosis and typically diastematomyelia. It commonly presents with urinary symptoms [20, 21] and may be exacerbated by the lithotomy position. Although typically the conus is abnormally low, it may occasionally be at a normal level but tethered by a tight filum terminale. Tethered cord should represent a contraindication to neuraxial anaesthesia. None of the cases reported here showed evidence of it.
Intrathecal injection of a neurotoxic substance, such as irritant agents that were once used for spinal anaesthesia, can produce cauda equina syndrome, because the sacral roots are poorly myelinated and particularly vulnerable to chemical damage. Although Waters et al.  reported ‘conus medullaris injury’ following spinal anaesthesia using tetracaine and lidocaine in sequence, the cauda equina was the more likely site of such injury. Cauda equina syndrome is indeed described following continuous spinal anaesthesia using lidocaine . Tedeschi et al.  reported a case of true conus injury following spinal anaesthesia in a 62-year-old diabetic woman, with the MRI appearance of gas within the cord and oedema surrounding it. Katz & Hurley  described the case of a parturient who was given repeated painful top-ups via an epidural catheter embedded within the conus. The resulting syrinx appeared to fill the vertebral canal, an MRI appearance similar to that of a congenital cyst. Greaves  reported a case of haematomyelia following attempted spinal anaesthesia believed to be at L3−4. The patient, another elderly woman, had a hip replacement under general anaesthesia following failed spinal anaesthesia. She had suffered severe pain reportedly in the left leg on injection of an estimated 0.3 ml of hyperbaric bupivacaine. An atraumatic spinal needle had not been used. Postoperatively, she had sensory deficit from T12 to L3 and a dense motor paralysis in the right leg. She died on the tenth postoperative day of pulmonary embolus, and at autopsy the needle track, identified at T12–L1, penetrated the conus and was associated with a haemorrhage within the cord extending 4.5 cm cephalad, on the right of the midline. The anomalous side of the initial pain was apparently unnoticed by the author, editor and all referees of this paper. This case is important because, although there was no MRI, the nature of lesion was verified at postmortem examination, and also because of the proven error of perhaps three segments in the level of spinal insertion.
Three anatomical factors
1 Although the cord commonly ends opposite the lower border of L1 or the L1−2 interspace, it may extend as low as L3. Thomson  found that it reached L2 in 43% of women but only 27% of men. The frequency distribution of the segmental level at which the spinal cord ends was assessed by Reimann & Anson  in 129 cadavers, and by Saifuddin et al.  in a more recent MRI study of 504 adults. Their results are summarised in Fig. 4. Considering the angle of entry of a spinal needle, if inserted at L1−2 it might reach a conus that ended at the lower border of L1, which according to Fig. 4 would encompass between 63 and 78% of individuals. Using the same argument, if inserted at L2−3 it might be possible to reach the conus in 4–20% of people.
Figure 4. Proportion of adults in whom the cord would be present at each spinal level. Cumulative data derived from Reimann & Anson  (▪) and Saifuddin et al.  (◆); L2 lower = the lower third  or half  of the body of L2; L2 mid = middle of the body of L2 (interpolated in upper curve); L2 upper = upper third  or half  of the body of L2; L1/2 = L1−2 interspace; L1 lower = the lower third  or half  of the body of L1.
2 Tuffier's line, that joining the iliac crests, while commonly used to identify lumbar interspaces, does not bear a constant relationship to them. Although the mode is the lower border of L4 to the L4−5 interspace [30, 31], the level may vary from L3−4 to L5–S1, hence a major source of error.
3 At the level of the conus the nerve roots form a highly organised overlapping pattern in close proximity to the cord, and bound to it by an intricate web of arachnoid membrane .
Possible mechanisms in the cases reported here
The consistent histories in these seven cases, with damage to more than one root, strongly suggest that the needle-tip alone can cause conus damage. In all cases the symptoms were mainly unilateral, unlike with more severe conus lesions, but this is perhaps not surprising because the MRI lesions appeared small and unilateral (Figs 1–3). The MRI changes that were observed are consistent with fluid collection, intramedullary haemorrhage or a small infarct. Yet most of those involved did not believe the needle could have reached the spinal cord. This may be a misapprehension (see below) but it is possible, perhaps, that a needle inserted among the tightly knit terminal roots could tear the surrounding membrane causing a small haemorrhage.
Why this cluster of cases? Can they be attributed to the current fashion for atraumatic needles? It is true that an atraumatic spinal needle has at least 1 mm of blind tip beyond the hole and there may be a tendency to insert it further into the subarachnoid space than is necessary with a Quincke needle. Yet this 1 mm is unlikely to be the whole answer.
For many years spinal anaesthesia was in the doldrums, while epidural blockade flourished, and anaesthetists learnt to site epidural needles at all manner of levels. Then, with the re-introduction of atraumatic needles and the use of less noxious local anaesthetic solution, the practice of spinal anaesthesia made a comeback, particularly in obstetric practice, and it is now the most popular form of anaesthesia for Caesarean section . So now, many anaesthetists with liberal attitudes to lumbar interspaces that are, moreover, condoned by many textbooks , have taken to using spinal and CSE anaesthesia. It should be emphasised, however, the cephalad tilt given to the spinal needle at CSE using the needle-through-needle technique is not the only problem, as three of the cases reported here related to single-shot spinal anaesthesia.
Although many anaesthetists are confident that they can identify lumbar interspaces accurately, van Gessell et al.  demonstrated that 59% of dural punctures were performed one or two spaces higher than assumed. More recently, Broadbent et al.  found that when a group of experienced anaesthetists believed they had identified L3−4, in 85% of observations the space selected was one to four segments higher than this. Given the inaccuracy of methods of identifying lumbar interspaces, and the variability of the position of the conus, it cannot be logical to aim to insert a needle intrathecally above the spinous process of L3.
If a spinal needle causes pain, it is obviously correct to avoid injection although it may be too late to prevent nerve damage. It would also be wise to establish by radiological means the interspace that has been used. The moral of these stories is not to avoid atraumatic needles but to avoid upper lumbar interspace at all times, exercising particular care in women.