So, as it’s the middle of August in Virchester our thoughts turn to Christmas and the joys of short days, long nights and general inebriation (a very popular hobby in central Virchester over the festive period).
Hang on though, this has nothing to do with SANTA (check that link…Ed), but rather we are back to looking at the use of starch solutions in the management of critically ill patients. The CRYSTMAS study is an RCT comparing 6% Hydroxyethylstarch 130/0.4% vs. 0.9% NaCl in patients with severe sepsis….HANG ON you say! Did we not do this a few weeks ago you cry, and yes you are right, we did and the conclusion from the trial was that this was not a therapy that we could justify any longer. However, as one swallow does not make a summer, one trial does not always end a discussion so let’s see if there is anything new we can pick out in this different approach.
First – ideally read the paper and give it a mark out of 10. If you don’t have time at least scan the abstract below toget an idea of what the authors are saying.
Sooo, what did we think, and in particular what did we think about this having read the paper? Did it make us question the conclusions of the Perner trial in the NEJM that we should abandon the use of starches in sepsis
This was a very, very different paper, but where to start? Perhaps at the beginning. In any great paper we want a clear, unambiguous and clinically important question to be asked, but here we find a trial that seeks to demonstrate that patients receiving HES achieve haemodynamic stability (HDS) quicker than if given Saline. OK, but is that something that really bothers us as clinicians or is it just a way-point on a more important journey? We think the latter, and particularly in severe sepsis where the mortality is about 40%. I’m sure patients don’t give a monkeys whether they achieve HDS quickly, they just want to live, and live well. So, the first thing to say is that the primary outcome measure of this study is arguably not relevant and if you like you can stop reading now. Seriously, if it’s not an important question bin it, move on, read a book, press a flower….do something more interesting instead.
You’re still here? OK,then you are hard-core or looking to improve your skills in critical appraisal so let’s move on.
The next thing we noticed is in the analysis of the data. Something weird goes on here as they do not analyse their data on an Intention to Treat basis. They use something called a Full Analysis Set (FAS) which only included patients who responded to therapy….yep, they excluded patients who did not reach HDS. This is odd as at the time of randomisation they could not have known who would and would not survive. In RCTs you should analyse by allocation not by what seems convenient. This worried us a lot.
We also looked at what they found. The main outcome was a statistically significant difference in the amount of fluid given to patients in order to achieve HDS…., but the difference was an average of 331mls (p=0.0185). This is a triumph of statistics over clinical relevance which is the wrong way round. Do we think that a difference of that fluid is important (especially when you read the next paragraph)? We don’t.
Then we started to think about outcomes again. Seriously folks, any trial in severe sepsis that does not have mortality as a major outcome worries me. The mortality of sepsis is huge, and in this trial it was 30% in the HES group and 25.3% in the NaCl group. Now because the numbers were small this was not statistically significant, but the difference is similar to that found in the Perner trial, and on the basis of that trial we decided to stop using Starches in sepsis. So, arguably this is even more evidence in that direction.
So, what shall I do with my spuds after reading this paper? A few weeks ago @thegreathornero decided to turn them into Rumblethumps, this week I think I’ll find some Spinach and put together a rather nice Sag Aloo. I will most certainly not be giving HES to any of my Severe Sepsis patients.
Right, next week we need a game-changer of a paper. This one was great for critical appraisal training but I want something I can put into practice.
OK, so it’s FCEM style question time. Having read this paper answer the following questions.
[DDET This study states in the statistical analysis section of the methods that it was designed to have 90% power. What does that mean?]Power relates to the ability of a study to avoid a type 2 error. A type 2 error is when the study is too small to show a difference statistically even if one truly exists. The more participants in the study the less likely you will have a type 2 error. A 90% power means that the numbers in this study design (86 in each group) would detect a difference of 400ml of fluid given on 90% of occasions a study like this was performed. So, 10% of the time a true difference may exist but they might not find it. Conventionally a power of 80 or 90% is quoted in sample size calculations.[/DDET]
[DDET If 31% of patients in the HES group died and 25.3% of the NaCl patients died. What is the Number needed to Harm?]This is a really useful thing to be able to work out. The number needed to harm NNH is the same as the number needed to treat NNT, it just depends on which way you are thinking. For this trial let’s think about NNH in giving HEC as the mortality is higher. So to work out the NNH… 1. Work out the ARR – Absolute Risk Reduction…Eh??? Simple – what’s the difference between the death rates in percentages. So that’s 31% – 25.3% which is 5.7%. So the ARR is 5.7% 2. Convert ARR to NNH. Again easy. it’s just 100/ARR So 100/5.7 = 17.54 So the NNH is 18 (I round up because I struggle with the concept of 0.54 of a patient). For every patient you treat with saline as opposed to HES you get one extra survivor. That’s a really impressive number…but beware it’s not statistically significant in this study so we don’t know for sure.[/DDET]