The end of the road for colloids in sepsis??? St.Emlyn’s
So it looks as though Cochrane have published another update on the use of colloids in the management of severe sepsis. This is something we have looked at on St.Emlyn’s before when we looked at the CRYSTMAS trial and the Perner study in the NEJM . These showed that mortality was increased in patients with sepsis who received colloids and as a result we have locally seen their use almost dissapear.
These trials though convincing were just a small part of the evidence based though, so it is reassuring to see a wider systematic review come to similar conclusions.
There are no apparent benefits to the use of colloids in sepsis.
But don’t take my word for it. You should read the paper yourself as this is an important area for all of us involved in resuscitation and as I will argue later, some of the important detail is lost if you just read the author’s conclusions.
The methodology of the Cochrane collaboration is well described and this study is unlikely to miss out any significant papers and looking at the list of 70 papers included in the paper I can’t see any glaring exceptions.
What does this tell us?
Analysis 1 - Albumin vs Crystalloid
Overall 922/4951 (18.6%) patients died in the crystalloid group
Overall 914/4969 (18.3%) patients died in the albumin group
This was not statistically significant.
Analysis 2 Hydroxyethylstarch vs crystalloid
In the Cochrane review they have looked at a total of 25 trials.
- Overall they found a mortality of 1002/4615 (21.7%)( in the starch group
- Overall they found a mortality of 912/4521 (20.2%)
That’s a number needed to harm (by giving starch) of 66.6
Analysis 3 Modified gelatin vs crystalloid
- Overall deaths were 13/224 (5.8%) in gelatin group
- Overall deaths were 15/282 (5.3%) in crystalloid group. (Ed – pretty low death rate for sepsis here hmmm)
A pretty low event rate to be honest and not much data here.
They also looked at Dextran, but since no-one I know uses it I’m not going to talk about it!
So the bottom line is???
“Authors’ conclusions: There is no evidence from randomised controlled trials that resuscitation with colloids reduces the risk of death, compared to resuscitation with crystalloids, in patients with trauma, burns or following surgery. Furthermore, the use of hydroxyethyl starch might increase mortality. As colloids are not associated with an improvement in survival and are considerably more expensive than crystalloids, it is hard to see how their continued use in clinical practice can be justified.”
Now I am fairly happy with that as it does concur with my own feelings, but I am still left with a few questions about this kind of study when we look at the complexities of a condition such as sepsis.
Go on then.....
This is not the case as it is quite possible to do a bad systematic review. So is that something that has been done here?
In general no. The work is explicit in its methods and the authors have clearly described how they have sought, analysed and transcribed their data. The analysis that I have simplified above is the correct model and with most Cochrane reviews it is well presented.
I do think that it is worth looking at the collation of trials in studies like this. In order for the pooling of data I believe that it is very important to look to see if the patient populations are similar. Whilst there are statistical methods to pool data from trials with different event rates it is never quite so pure as using data from trials with similar methods and patient groups. This may well account for the difference in the number needed to harm for starches we found previously (NNH – 13) as compared to the pooled data in the Cochrane review (NNH-66.6). If you download the full version of the paper then you will see that the event rate (death) varies widely betweem studies and this is usually a fair indicator that we are looking at different populations where the effect of colloids may subsequently also be different.
Learning points please
- Colloids do not appear to have any benefit in sepsis
- Converting numbers reported in meta-analysis to NNH or NNT is SO much more intuitive
- Look at the event rates for individual studies in the meta-analysis to get an idea about the variation in patient populations.
Obviously there is always more to talk about, but we are all busy people and I must pack to travel to the other side of the world at SMACC2013. Look out Sydney, here comes Virchester.
And as Chump says….