Sensitivity of computed tomography performed within six hours of onset of headache for diagnosis of subarachnoid haemorrhage: prospective cohort study BMJ 2011;343:d4277

25 Aug

Perry JJ, Stiell IG et al Sensitivity of computed tomography performed within six hours of onset of headache for diagnosis of subarachnoid haemorrhage: prospective cohort study. BMJ 2011 Jul.;343(jul18 1):d4277–d4277. PMCID 21768192

This paper has rightly received a lot of attention in the blogosphere so far.

For some background check out one of the older SMART EM podcasts which discussed another paper (free full-text) by the same authors which looks at the same data set.

Cliff Reid has already talked about this paper too.

And Dreapadoir

Also check out some concerns of the idea of CT only in SAH by Stuart Swadron (via

So if that’s not enough for you here’s my under-educated, unwarranted opinions.


  • prospective data on patients being worked up for SAH. No specific criteria, just if the doc suspected and was going to work up SAH you got in
  • question was how good is the CT and importantly a priori they were interested in less than 6 hours to see how good it was then

This is different than doing the study and then going back to find a time cut off that suits what you want to say – data dredging. These guys designed the study with CT in less than 6 hours in mind

  • decision to have an LP or not was down to the doc looking after the pt

This leads to the charge of work-up bias. If only some of the subjects get the gold-standard work-up then you need to be cautious about the accuracy of the results. They try to deal with this by follow-up

  • excluded patients who had something obviously catastrophic going on in their head; ie obtunded in any way or any focal neuro findings

this is really important as previous studies of CT in SAH included people who you didn’t need a CT to diagnose! This would make the numbers look a bit better than perhaps they were. The patients in this study are the ones we’re interested in – conscious and looking pretty well but we need a CT to be sure.


  • those who got both the gold-standard tests (CT plus LP) weren’t followed-up. Fair enough
  • those who only got the CT were followed by phone, then mailed survey, then chart review and death record review at 6/12

In studies like this follow-up is key so look out for it in the results


  • SAH defined as: blood on CT;  xanthocromia and +ve angio
  • SAH on follow-up

If you practice in the UK or Ireland you should note that this didn’t include spectrophotometry, what I thought was the much better test. Remember +ve xanthocromia means someone in the lab holding a tube of CSF against a white page and deciding whether it looks a bit yellow or not.

So what was the reason for not using spectrophotometry? To quote the authors:

there is no strong evidence to demonstrate the superiority of xanthochromia by spectrophotometer versus visual detection of xanthochromia plus red blood cell analysis. The specificity of spectrophotometry defined xanthochromia is poor

This was news to me so I very briefly pulled some of the papers on this (some by Perry himself) and found numbers from 29% to 100%. I couldn’t bring myself quite yet to look at all the papers so this is still an open question to me for now…

From their point of view this was the best available definition.


  • 3100 pts
  • 7% rule in rate on CT

They predicted this before hand and is useful information in itself. 1 in 15 times we order a CT looking for SAH in an awake patient it’ll be +ve. I can tell patients that and I can tell radiologists that

  • only half got the LP if the CT was -ve

so only half of the population got both the gold-standard tests. So for half the patients we’re calculating sens/spec off results of follow-up

  • 2% lost to follow-up (2% of those who had a -ve CT but no LP)
  • having said that by their definitions CT was 100% sensitive if done before 6 hrs (based on about 115 +ve out of 953 CT scans)
  • dropped to 85.7% after 6 hours
So how were the other 14.3% picked up? 13 of 17 were diagnosed by xanthocromia and 2 by blood in the final tube. the other 2 were a bit more complex.
I can’t find the number of pts after 6 hrs who had -ve CT and an LP – if I could then we could work out some idea of how often an LP might be +ve. My guess using assumptions from their figures is somewhere between once and twice every 100 LPs it will be +ve

It’s worth noting that over half of the SAH that were diagnosed on LP were actually non-aneurysmal bleeds. If you didn’t know it already I’ll point out that people with non-aneurysmal SAH do great most of the time, and there’s no surgical intervention.

  • in terms of their follow-up
    • they followed 98% of those with -ve CT but no LP and not a single one was found to have an SAH
    • 6 died but these were found to be other causes
    • 2% (50 patients) were lost to follow-up but didn’t show up at a neurosurgical centre or in the death records

This could mean a few things: The docs (who decided on whether to do the LP or not) chose perfectly and did all the appropriate LPs. Or some had really had SAHs and they did fine despite us missing it. Or some of the 50 lost to follow-up went home and died somewhere.

I suspect the answer is closer to the first but I can’t prove that.

My Thoughts

This is some of the best evidence we have on this as far as I can see. The CTs have got better since the original (deeply flawed) studies.

The controversy and mess surrounding spectrophotometry and xanthocromia is still pretty opaque to me but I suspect it’s not crucial.

So this is a game changer for me. By this I mean if you ask my opinion on a -ve CT in a patient who’s alert with a normal neuro exam I’ll say that I think they’re -ve for SAH

So will I do this in practice and send them home? Well… perhaps not yet. For a start I work under supervision and in a department not just as a lone ranger.

I already discuss whether or not to do the LP with patients, though I’m still not confident on what specific numbers to tell them.

I’m now back to teaching anatomy for most of the next year so I’ll not have to make that call too much anyhow!

Is anyone else telling patients that they’re fine after a -ve 6hr scan?

3 Replies to “Sensitivity of computed tomography performed within six hours of onset of headache for diagnosis of subarachnoid haemorrhage: prospective cohort study BMJ 2011;343:d4277

  1. Here’s the real question. Why start with the CT (expensive, ionizing radiation) and not the LP (sometimes a headache)?

    • Good question.

      – part the renowned risk of herniation from an lp in someone with raised icp (vastly orerated I think) – partly that ct picks up lots of thongs that might look that sah that aren’t (and would be missed on lp only) – partly patient choice and preference. – post lp headache is probably in the region of 10% and it’s not insignificant.

      Lp is much feared by patients and often shouldn’t be. I’ve done quite a few where the patient thought that getting the bloods taken was worse than the lp. It’s a core skill that a lot of people are poor at.

  2. Pingback: Age Adjusted D-dimer Cut Offs - Emergency Medicine Ireland

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.