Intended for healthcare professionals

Clinical Review

Lesson of the week: Lumbar puncture still has an important role in diagnosing subarachnoid haemorrhage

BMJ 1997; 315 doi: (Published 13 December 1997) Cite this as: BMJ 1997;315:1598
  1. Jonathan Wasserberg, senior registrara (jwasserberg{at},
  2. Philip Barlow, consultant neurosurgeona
  1. a Department of Neurosurgery, Institute of Neurological Sciences, Southern General Hospital, Glasgow G51 4TF
  1. Correspondence to: Mr Wasserberg
  • Accepted 26 August 1997


An early diagnosis of subarachnoid haemorrhage is essential if patients are to undergo appropriate investigation and treatment. It is increasingly common for patients with suspected subarachnoid haemorrhage to initially undergo computed tomography. Although the procedure is sensitive and non-invasive, it does not detect every instance of subarachnoid haemorrhage. Lumbar puncture should be considered in patients who have an acute onset of headache and who are fully conscious but have a normal scan or when the hospital does not have access to a computed tomographic scanner. We report on two patients in whom an early diagnosis of subarachnoid haemorrhage was missed because they did not undergo lumbar puncture.

Case reports

Case 1—A 58 year old man had lost consciousness for one minute. On recovery he had a severe headache and bloodstained vomitus. He was admitted to hospital and given intramuscular opiates for pain relief. Haematemesis was provisionally diagnosed. His initial symptoms were attributed to possible alcohol withdrawal and treatment with diazepam was started. Two days later his headache had not improved. Subarachnoid haemorrhage was suspected, and he was booked in for a routine computed tomography, the earliest appointment being two days later. On the next day, however, he became unconscious and had a fixed, dilated left pupil. An emergency scan showed extensive subarachnoid haemorrhage, and he was transferred to a neurosurgical unit. Before further investigations and treatment could be performed he died, presumably of a rebleed.

Case 2—A 17 year old woman attended her doctor because of a history of headache associated with nausea and vomiting for one week. She was referred to hospital and underwent computed tomography to rule out subarachnoid haemorrhage. The scan was normal, and she was reassured and discharged home. Two days later she awoke with a sudden severe headache, vomited, and then collapsed. On arrival at hospital she was in a coma but reacting to pain. She was intubated and ventilated and transferred to a neurosurgical unit. Computed tomography showed a subarachnoid haemorrhage and a cerebral angiogram showed a terminal carotid artery aneurysm. During angiography her condition deteriorated, both pupils became fixed and dilated, and she subsequently died. The original tomogram was reviewed and did not show any evidence of subarachnoid haemorrhage.


Subarachnoid haemorrhage should be suspected in all cases of severe headache of sudden onset. In the first case diagnosis was delayed while the patient awaited computed tomography. In the second case a normal scan was considered sufficient to exclude subarachnoid haemorrhage. In both cases an early diagnosis would have been possible had a lumbar puncture been performed. Although an earlier diagnosis would probably not have altered the outcome in these two patients, it could be important in other cases.

In a large prospective study scans were normal in 5% of patients subsequently shown to have had a subarachnoid haemorrhage.1 In a recent study in which computed tomography was performed with third generation scanners within 12 hours of the onset of headache, two out of 119 patients with subarachnoid haemorrhage (2%, 95% confidence interval 0.25% to 6%) had a normal scan.2 As extravasated blood can disperse quickly it may be undetectable by computed tomography after 12 hours. Thus if a scan was delayed there would be more chance of a subarachnoid haemorrhage being missed. In a series of 181 patients investigated for suspected subarachnoid haemorrhage the sensitivity of computed tomography performed within 12 hours of the onset of headache was 93.1% and after 12 hours was 83.8%, giving an overall sensitivity of 91.2%.3 The decreasing sensitivity of computed tomography with increasing time from the onset of headache was shown in a series of 2940 patients presenting with subarachnoid haemorrhage. The sensitivity of scanning decreased from 92.1% on the day of the bleed to 57.6% on day 5.4 Thus the sensitivity of computed tomography cannot reliably exclude lumbar puncture in patients who have symptoms suggestive of subarachnoid haemorrhage.

Lumbar puncture is necessary in cases of suspected subarachnoid haemorrhage when a computed tomogram is normal or unavailable. Subarachnoid haemorrhage can be detected by the presence of xanthochromia between six and 12 hours after haemorrhage.5 To ensure that xanthochromia is detected after subarachnoid haemorrhage, lumbar puncture should be deferred until 12 hours after the onset of headache, necessitating admission in most cases. In cases of late presentation after subarachnoid haemorrhage all patients have xanthochromia up to two weeks after haemorrhage, 70% after three weeks, and 40% after four weeks.5

The risk of neurological deterioration after lumbar puncture is disputable. In 1982 seven cases of neurological deterioration after lumbar puncture were reported in 55 patients investigated for suspected subarachnoid haemorrhage.6 All seven patients had moderate to severe headache and neck stiffness and two had a focal deficit and confusion. As lumbar puncture is contraindicated in the presence of papillo-oedema, focal deficit, or reduced consciousness it may have been inappropriate in some if not all of these patients. In a retrospective review of 123 cases of subarachnoid haemorrhage, 91 patients underwent lumbar puncture with no evidence of deterioration from the procedure. In the same study 22 out of 24 patients with an intracerebral haematoma on computed tomography had focal neurological signs and severe impairment of consciousness. The authors concluded that lumbar puncture in patients without these signs was safe.7

The Society of British Neurological Surgeons has recently circulated guidelines to its members on the initial management of subarachnoid haemorrhage. The guidelines emphasise that computed tomography detects subarachnoid haemorrhage in some but not all patients. When a computed tomography scanner is available computed tomography should be performed and lumbar puncture carried out if the results are normal. A lumbar puncture should be performed if a diagnosis is in doubt and computed tomography is unavailable but the patient is oriented and obeying commands. Lumbar puncture should not be performed in patients with papillo-oedema or focal neurological signs.

One study proposed that reports of neurological deterioration after lumbar puncture might lead physicians to abandon the procedure in cases of suspected subarachnoid haemorrhage.8 Even with the increasing availability of computed tomography, lumbar puncture should still be performed in appropriate cases.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.