We already know that patients taking warfarin are at very high risk of intracranial haemorrhage (ICH) after head injury, and that delayed bleeds aren’t so uncommon either. In fact, we’ve even discussed whether such patients routinely require not only an initial CT scan but also a delayed scan 24 hours later. The NICE guideline tells us that any head injured patient who loses consciousness or has any amnesia and has a coagulopathy should have a CT scan. Most of us would be even more conservative than this – and would go as far as to suggest that any anticoagulated patient with a minor head injury should have a CT scan. Some would suggest that all of these patients should then be admitted for neurological observation followed by a repeat scan.
So the issue is fairly clear when it comes to warfarin…. Warfarin + head injury = HIGH RISK = CT scan.
But what about antiplatelet agents? In the era of PCI (percutaneous coronary intervention) and improved secondary stroke prevention, many of our patients will be taking powerful antiplatelet agents like clopidogrel. If these patient sustain a minor head injury, does clopidogrel increase the risk of ICH? If so, how much does it increase the risk by?
— JC_StE (@JC_StE) December 14, 2012
At JC this week, we looked at a paper on just this topic. The authors wanted to find out the risk of ICH in patients with head injury who were taking clopidogrel, compared to the risk in patients taking warfarin. So, they ran a prospective cohort study at 6 US centres.
There are lots of different types of cohort study. In a standard diagnostic cohort study, the aim should be to include all patients suspected of having a certain condition, subject them to the diagnostic test (+/- a comparison) and then subject them to a reference standard for the condition (or outcome) of interest. This study differs, in that the authors wanted to include all patients suspected of having the condition (traumatic ICH) providing that they were taking either warfarin or clopidogrel. All patients ought to have had the reference standard (a CT scan, with robust arrangements for reporting and clear definitions of what constitutes ICH). The authors then wanted to compare the two subgroups (warfarin vs. clopidogrel).
A prospective observation multicentre study to answer this question! Loads of attracting terms here #JC_StE
— JC_StE (@JC_StE) December 14, 2012
In a cohort study of this nature, in my opinion the really key questions are…
(1) Is this the right study to answer the question?
(2) Is this the right patient group to answer the question?
(3) Did all the patients have the initial ‘interventions’ (in this case, either taking warfarin or clopidogrel pre-head injury)?
(4) Is the primary outcome appropriate and was it assessed robustly?
So, first things first, is this the right study? Well, I do have an issue with the aim to compare warfarin and clopidogrel groups. These are observational data – this isn’t a comparison of two different treatments. The patients taking warfarin are different to the patients taking clopidogrel – otherwise they wouldn’t be on different treatments. That means we’re comparing apples and oranges here. So, we’ll have to accept (from the outset) that direct comparisons between the group are flawed – because, regardless of how well matched the baseline characteristics may look, there will inevitably be hidden confounders.
Are these the right patients? Arguably, we should be looking only at those patients with mild head injury. Patients with GCS < 15 are going to be having a CT scan anyway. (At least we would hope so!) We really need to know whether, in patients with no other indication for CT, the fact that a patient is taking clopidogrel ought to mean that a CT scan is necessary. In this study, the authors included all comers.
97.3% of the patients did have a GCS of 13-15 (64.3% had a GCS of 15), only 4.2% had vomiting and 18% lost consciousness or had amnesia. So, on the whole, it’s a low risk group. But it’s not the ideal sample to answer the question, let’s face it.
Third, did all patients have the initial ‘interventions’? This question is most important when a diagnostic test is being investigated. It’s a little less important here. It would be nice to know how compliant the patients had been with their clopidogrel/warfarin and what doses they were taking. Being pragmatic, though, it’s enough for us to know that the patient reported that they were taking one of these medications – fair enough, although it might have been nice to know exactly how this history was confirmed.
Lastly, is the primary outcome appropriate? Well, ICH was defined as any ICH or contusion on initial CT scan, as reported by the faculty radiologist. And here’s an important question… Why do we want to get a CT scan on head injured patients? Is it because it might show some blood? No. It’s because it might tell us something that has clinical significance for the patient – e.g. that they need neurosurgery to survive; or that they’re going to get neuro-cognitive defects as a result of the head injury. I.e. the most important thing is the patient’s outcome – not the CT appearance.
So, we could critique the primary outcome by suggesting that, in actual fact, we really want outcome data (Glasgow Outcome Scores, neurosurgery, etc). But we haven’t got that in this study, so we’ll have to live with it.
Even accepting the primary outcome as it is, we need to know whether it was assessed appropriately. Who reported the scans? (A single faculty radiologist of unknown seniority). Was interobserver reliability assessed and/or accounted for? (No). Did all patients undergo the reference standard? (No – 6% didn’t have a CT scan, although patients were also followed up after 14 days by telephone).
Clearly, there are some limitations. There’s a reason why we call this process ‘critical appraisal’ – it’s hard to do it without picking some holes and sounding critical, even for some of the best papers. But there’s actually a lot we can take away from this one…
12% of the patients taking clopidogrel had ICH. That’s quite some figure – 1 in 8 patients! And these were, predominantly, well patients with normal GCS, no loss of consciousness, and no amnesia. The warfarin group was also at high risk – 5% had ICH. If you believe that the comparison is valid, the patients taking clopidogrel were significantly more likely to have ICH than the patients on warfarin. The Relative Risk was 2.31 (95% confidence intervals 1.48 to 3.63).
What about the incidence of late bleeding?
We know, as we’ve previously discussed in St. Emlyn’s, that patients taking warfarin are not only at risk of ICH when they arrive (justifying an immediate CT scan) but they are at risk of developing late-onset ICH. The big question is whether this happens with clopidogrel too.
As a secondary objective, the authors tried to answer this question too. They found that 0.6% of the patients taking warfarin developed late ICH compared to none (0.0%) of the patients taking clopidogrel. Unfortunately, we don’t know how many patients had repeat CT scan and we don’t know Glasgow Outcome Scores of these patients – but the fact that no patients taking clopidogrel had developed late ICH suggests that this isn’t a big problem in the clopidogrel group.
What’s the bottom line?
Even accepting the limitations, this evidence strongly suggests that patients taking clopidogrel are at substantial risk of ICH after head injury. Scan them. Even in the absence of other indications.
— Katrin Hruska (@Akutdok) December 14, 2012
— katherine potier (@kazpotier) December 14, 2012
— Gareth Hardy (@DrGDH) December 14, 2012
Any comments or alternative opinions are most welcome!
Before you go please don’t forget to…