Sensitivity of computed tomography performed within six hours of onset of headache for diagnosis of subarachnoid haemorrhage: prospective cohort study

BMJ 2011; 343 doi: (Published 18 July 2011)
Cite this as: BMJ 2011;343:d4277

Recent rapid responses

Rapid responses are electronic letters to the editor. They enable our users to debate issues raised in articles published on Although a selection of rapid responses will be included as edited readers' letters in the weekly print issue of the BMJ, their first appearance online means that they are published articles. If you need the url (web address) of an individual response, perhaps for citation purposes, simply click on the response headline and copy the url from the browser window.

Displaying 1-3 out of 3 published

Dear Sir,

We appreciate that this study was a reflection of 'true clinical practice' and a thorough follow-up of those patients discharged back into the community, which revealed a very high negative predictive value for a negative CT. Most importantly, we felt it showed what we are doing is safe, but perhaps we now feel a bit more insecure about diagnosis in that > 6-hour group. In our discussions of this paper, we had a few questions about the analysis. Most critically, would you be able to provide a two-by-two table for overall and the <6-hour group? We include our own two-by-two table interpretation of the text.

Our first question is about the identification of false-positives, which we feel might not be the most critical question in emergency practice, but is a critical aspect of diagnosis and ongoing care. You report specificity and positive-predictive values (PPV) but it is not clear in the results how many were identified or how you isolated true- positives from false-positives. It appears that a subset of participants had angiograms but it is not clear how many of these were positive and it is not clear what the neurosurgical outcome for the positive CT group.

There are two reasons for enquiring. The first is that if there was no follow-up to identify 'false-positive' results then you would be risking a Type I error and testing a CT against itself as a gold-standard - and thus, of course, your specificity and positive-predictive value is going to near 100%. We appreciate that false-positive CT reports are likely a rare event, but without defining it, the validity of the PPV and specificity is questionable. In clinical practice, we have seen cases where a 'blush' on the CT reported by the radiologist reveals no further findings on neurosurgical investigation. If you were searching for false- negatives with morbidity/mortality criteria, we thought it scientifically valid to identify false-positives with similar treatment?

The second reason for enquiring: if you do have data on the 'false-positives' we thought it would be interesting, if your data allows it, to publish an ROC analysis on what time since onset of headache CT provides it's greatest clinical utility? We appreciate your informed selection of a cut-off of 6 hours, but it would be interesting to see at what time-point since the onset of headache the sensitivity of CT significantly starts to fall.

Our next question is regarding the establishment of false-negatives. It is not explicitly stated in the text whether there were no false negatives for the group with a CT scan within 6 hours (outside the 6 hour limit, it is reported in the text as 17). We assume that it was zero from the strength of the results? We were wondering how many of the 1546 participants having LP's were within the <6-hour group. We only ask, as it has some importance in determining the value of lumbar puncture (LP) in that time-group. As stated in the introduction, it is this question of utility of LP that seems to be clinical hurdle once a CT is reported as negative. As reported, it appears that the 'true negatives' were established by a 6 month follow-up of morbidity and mortality rather than LP alone. This follow-up in our minds was clinically more important, as LP as a diagnostic study for SAH is well known to be suboptimal. However, it would be interesting to see how many LP's were performed and how often they confused or confirmed the diagnosis.

Competing interests: None declared

Cynthia Bierl, ED Registrar

Will Sargent, ED Specialist

Royal Darwin Hospital, NT Australia

Click to like:

Professor Jeffrey J. Perry and colleagues show that multislice CT in patients imaged within 6 hours of headache onset is extremely sensitive for detecting subarachnoid haemorrhage.1 With a specificity, sensitivity, positive and negative predictive values of 100% in these patients, their results challenge current practice guidelines which recommend lumbar puncture in patients with a negative CT scan.1

However, I wonder whether these results could have come to light much earlier? The authors calculated that they needed 2860 patients for their study. It took nine years to collect data on 3132 patients, from among 5424 potentially eligible patients. They state that a lack of data collection meant that 2292 potentially eligible patients could not be enrolled.1 These 2292 patients potentially could have contributed 80% of the sample needed.

I would be interested to know what methods the authors tried to improve data collection rates by emergency department staff. Multi- centre, long term studies could benefit from having an online presence to encourage, enable and reward clinical staff who collect data. For example, IST-3 uses its website2 to connect with multiple trial centres worldwide. Updated tables show how many patients have been recruited from each centre, and forms can be completed online. In addition, updates at milestone recruitment marks are announced to the trial community through monthly newsletters, for example, when the 3000th patient was randomised by Professor Martin Brown's team in London.3

A local study nurse was able to identify 2292 potentially eligible patients who had been missed. Perhaps these misses needed to be communicated to the clinical staff, online or in person, so that data collection rates could have been improved and this study could have been completed sooner.

1.Perry JJ, Stiell IG, Sivilotti MLA, Bullard MJ, Emond M, Symington C et al. Sensitivity of computed tomography performed within six hours of onset of headache for diagnosis of subarachnoid haemorrhage: prospective cohort study. BMJ 2011; 343:d4277

2.IST-3 Trial Co-ordinating Unit. IST-3 trial. The third international stroke trial (thrombolysis) Last updated 16th May 2011. Accessed 1st August 2011.

3.IST-3 Trial Co-ordinating Unit. NewsJuly11.pdf. The IST-3 Times. Last updated July 2011. Accessed 1st August 2011. URL:

Competing interests: None declared

Lorna M. Gibson, Doctor

NHS Fife

Click to like:

This study is very interesting.Emergency physicians must be vigilant in evaluating patients with severe acute headache in the emergency room. This evaluation must take place with an understanding of the diagnostic tests used. Previous studies document that misdiagnosis of SAH occurs approximately 25% of the time (12-50%), even in the era of ready access to cranial computed tomography (CT) scanning. Failure to obtain a CT scan was the most common error.The first decision to be made when evaluating patients with headache is whether or not to pursue any diagnostic studies. CT Scan is available in many big cities around the developing countries. The limitations of the CT scan is accuracy decays with time. Aggressive evaluation of patients with acute-onset severe headache should reduce the phenomenon of delayed or missed diagnosis.

Competing interests: None declared

Rizaldy Pinzon, Neurologist

Bethesda Hospital Yogyakarta Indonesia 55224

Click to like: