Seven studies and just shy of 2500 patients were included, but only 50 of the patients were non-elective cases. Overall, the pooled rate of leakage was 9.6%, which sounds a bit high, although there was a marked variation between the studies (95% CI for rate 6.9-12.6%).
The analysis considered CRP values on the 3rd (5 studies), 4th (5 studies) and 5th (3 studies) post-operative day. The cut off values were 172, 124 and 144mg/dl respectively. For each of the days the PPV was approx. 20%, and the NPV approx. 97%. In other words; patient A has a high CRP so their chance of having an anastomotic breakdown is approx. 20%, whereas patient B has a low CRP; their chance is approx. 3% (It’s worth at this stage pointing out that in the real world you’d never consider the CRP in isolation, and the above data offers more of a description of the population rather than a probability for the individual patient).
Even in the elective setting, the use of CRP as a rule in doesn’t look viable. Although not given in the paper, I estimate that about 50% of the patients included had a CRP above the cut-off value based on rough sums, so even as a rule out it doesn’t look very sensitive (in that you will only have the opportunity to be reassured in half of the patients you test despite only 10% having leaked).
In comparison, the high risk patient would generally have a higher chance of anastomotic leak, but also be more likely to have a high CRP. Extrapolating this data would make a low CRP potentially more useful as a rule out (more sensitive) in this group, but less useful as a rule in (less specific).
The only data I could find in the critical care setting came from a poster presentation so details are lacking. The Dutch authors measured CRP daily for 174 general surgery patients admitted to their ICU, also calculating daily SOFA scores. The outcomes were ‘events’ (CT, USS, endoscopy or re-laparotomy/ thoracotomy), as well as complications. The results were interesting, in that the group who underwent an event had a higher CRP than those that did not, and that a rise in CRP was correlated with an increasing likelihood of an event. Adjusting for SOFA scores (so very roughly for level of support), the impact of CRP on likelihood for an event was statistically (but I’d argue not clinically) still present with an OR of only 1.033. The rise in, and level of, CRP between those with complications and those without was however the same. So in this group at least, CRP was leading to more events but added nothing by way of prediction of complications.
So where does all this leave us in critical care, and how best should we interpret the data from a different population? Stating the obvious first, looking at the whole picture, clinical judgement and team working are the best and overriding steps (I do not for example believe that performing a negative CT on the basis of a clinicians ‘strong gut feeling’ is bad medicine).
In my view, if the CRP is less than 200, I think that’s reassuring. If it’s up I wouldn’t dive in and do anything about it – I’d treat it a bit like delirium; as a hint that I need to just look a bit more closely to reassure myself. I also don’t think routine post-op CRPs will add much without repeating the observational methodology described above to assess utility for our own patient group. But that’s just my opinion – feel free to disagree!
I’m afraid I’ve ended up writing this without suggesting any areas for discussion, but if you can think of anything, if you disagree with me, or if you know of any other relevant data please leave a comment below. I’d be particularly keen for any comment from our surgical colleagues, or if you know whether other units use CRP and how useful they find it….