JC: Plasma-lyte or Saline in Trauma? St. Emlyn’s

We’ve been having debates about what fluids to use in trauma and sepsis for about as long as – well – about as long as we’ve had fluids.  We’ve had quite a bit of dogma going round, and a fair amount of recent dogmalysis – for example with Dan Horner heralding the end of potatoes on the ICU at St. Emlyn’s.

That’s not to mention the groundbreaking FEAST trial, of course!

In fact, most trials of fluid regimens seem to either find no difference or evidence of harm.  So it’s going to be hard work for a new kid on the block to gain acceptance.  This week, at St. Emlyn’s journal club, we appraised a randomised controlled trial comparing Plasma-Lyte to normal saline for patients with major trauma.

Plasmalyte Saline

[DDET What are the theoretical advantages of Plasma-lyte?]

There’s nothing that ‘normal’ about saline, really. It’s isotonic but that’s about as good as it gets. The high chloride concentration may lead to hyperchloraemic acidosis when large volumes are infused. The pH is around 5. And there’s absolutely no potassium or magnesium to be found.

Then we have ‘Plasma-Lyte’.  Wow.  Sounds pretty funky?  It sure is – it’s a balanced electrolyte solution with constituents that are custom designed to closely reflect those of plasma.  I bet you’re itching to know what kind of extortionate price it goes for already, right?  Here’s what the manufacturer says about the electrolyte composition of Plama-Lyte.Blood gas

Of course, in trauma we have this concept of the ‘Triad of Death’hypothermia, coagulopathy and acidosis. If Plasma-Lyte might help to correct or avoid acidosis for patients with major trauma, it could potentially lead to clinical benefit.

 

[/DDET]

[DDET What did these authors do? ]

The authors aimed to determine whether Plasma-Lyte or normal saline led to less acidosis in patients with major trauma. In effect, their research question was:

[In patients with severe acute injury who undergo intubation, blood transfusion, surgery or an interventional radiological procedure within 60 minutes] does [resuscitation with Plasma-Lyte or normal saline] lead to [superior correction of base deficit within 24 hours]?

Their null hypothesis was effectively:

In severely injured patients, resuscitation with Plasma-Lyte does not lead to superior correction of base deficit at 24 hours when compared to resuscitation with normal saline.

They ran a phase 2 prospective randomised, controlled trial to address this.

[/DDET]

[DDET Did the authors ask the right question?]

In answering this question, we need to break it down. We can take each of the three parts of the research question (or 3PQ)…

The patient group

The authors included patients with severe acute injury who had penetrating injury to the neck, chest, abdomen or pelvis; GCS <9 or deteriorating by 2h after injury; systolic BP<90mmHg; injured extremity without a pulse; or need for intubation.  All patients must have undergone intubation, blood transfusion, surgery or interventional radiology within 60 minutes.

Clearly, they were after the most seriously injured patients who are likely to be acidotic and to have a high mortality rate. The trial didn’t include less seriously injured patients and those who didn’t immediately go for intervention within 60 minutes. It’s therefore a slightly narrow population – but that’s perhaps not unreasonable.

What about the intervention?

The patients were randomised to receive their fluid resuscitation with either Plasma-Lyte or saline. This was blinded and the volumes infused were up to the treating clinicians. I think this is pretty reasonable.

What about the primary outcome?

The primary outcome was correction of base deficit at 24 hours. This is where we start to have problems. The most clinically relevant outcome would probably be mortality – and there are other clinical outcomes that may be important. We have to accept that the trial would have to be pretty massive to detect small differences in mortality so sometimes we do need smaller initial trials looking at surrogate outcomes to provide the justification for further work. This is one such trial – a phase 2 trial. So perhaps we can accept the fact that the most clinically relevant outcomes weren’t chosen.

What we have as a primary outcome instead is the change in base deficit at 24 hours. But is that really something that we want to know about? If we infuse a more acidic solution (saline) than Plasma-Lyte, wouldn’t we expect that Plasma-Lyte will perform better?  Is that clinically relevant? This outcome just won’t tell us.

[/DDET]

[DDET Was there any potential for bias?]

Actually, this was a quite nicely designed trial. It was prospectively registered. There was apparent allocation concealment, nice double blinding, decent randomisation, analysis by the intention to treat principle. The trial itself seems like it was pretty well designed. The baseline characteristics of the two groups were well matched.800px-Die_bone

Perhaps the only problem was attrition (or drop outs). 46 patients were included in the final analysis but the authors lost 7 patients in each group as they didn’t provide consent (although there’s not much we can do about that) and 4 patients didn’t follow the protocol sufficiently to be included in the final analysis. To be fair, if that’s our main quibble we’re doing pretty well so far.

[/DDET]

[DDET What were the key findings?]

So, the primary outcome was the change in base deficit at 24 hours compared to the base deficit at randomisation. The Plasma-Lyte group showed a significantly greater change – by 3.1mmol/L (95% CI 0.5–5.6). There was no difference in the most clinically relevant secondary outcomes such as mortality but this study wasn’t powered to detect that. Arterial pH was higher in the Plasma-Lyte group at 24h and we can see (from their graphs) that arterial pH seemed to correct more quickly with Plasma-Lyte. As you might expect, chloride levels were higher in the saline group.

[/DDET]

[DDET What’s the clinical relevance of this?]

saline

That’s the key question, isn’t it? At 24h the mean base excess was -2mmol/L in the saline group and +2.1 in the Plasma-Lyte group. Both are pretty much within the normal range.  (In fact you could argue that the mean was above the normal range in the Plasma-Lyte group!)  Are we really convinced that this difference will make a clinical difference to our patients?

What was the overall incidence of clinically relevant hyperchloraemic acidosis in each group? We don’t know. How many patients (in terms of absolute numbers) had persistent acidosis? We can’t tell.

Lastly, we have to think about the external validity of these findings – i.e. can we generalise them to our own settings? The patients in this trial had a lot of intravenous fluid: 9 to 10 litres in each group. Is that how we practice in trauma? Perhaps not if you’re following the principles of permissive hypotension. Would that have made a difference to the findings? Quite possibly – yes.

[/DDET]

[DDET So would you use Plasma-lyte in practice?  The cynic’s perspective]

Bad Science

There will naturally be some cynicism out there. This trial compares 2 products manufactured by Baxter. Of course, we assume that Baxter would like it if their more expensive product (Plasma-Lyte) was beneficial. So we’d like to know a bit about the authors’ potential conflicts of interest – but we can’t find that here.  There doesn’t seem to be any declaration about potential conflicts or about funding.

At the Virchester journal club when this paper was appraised, Roy Royce asked, “Is this a case of Ben Goldacre’s Bad Science?” Could the manufacturers be sowing seeds of doubt in our mind without going as far as providing clinically relevant evidence? Might those graphs showing a tendency towards earlier correction of acidosis influence us in the heat of a major trauma scenario? Will we move to Plasma-Lyte even though the evidence is not yet convincing?

[/DDET]

[DDET A practical perspective]

We know that we can’t really answer this research question based on these data alone, unless you really think that an apparently pretty minor contribution to overall metabolic acidosis is clinically significant.  The authors knew that too – they set out to run a phase 2 trial after all.  So we still have equipoise based on the data.  But we have 2 solutions that are commercially available, and right now we have to choose one of them for the patient we have in front of us with major trauma.  What should we decide?

Maybe the issue of cost will factor.  We’d expect that the funky one will be more expensive – you wouldn’t expect a fluid called ‘Plasma-Lyte’ to be cheaper than saline, right?  The authors tell us about the price though.  They say that saline costs $0.86 per litre while Plasma-Lyte costs $1.62.  There are two ways of looking at this.  Plasma-Lyte is nearly double the cost.  But, on the other hand, $0.80 per litre of fluid is hardly going to break the bank for our individual patient!   What’s more, according to Baxter’s Australian website, the far less funky Harmann’s + 5% Dextrose costs $8.00 for 12 litres.  Plasma-Lyte costs $6.50 for the same amount.  

[/DDET]

[DDET The bottom line]

Based purely on the science and the hard facts, while the findings of this trial may justify a larger phase 3 trial, they unfortunately don’t provide the robust evidence we would need to say that Plasma-Lyte is better than saline in trauma.  That’s not a criticism of this trial: it did what it was designed to do – evaluate whether Plasma-Lyte may have some kind of efficacy (not necessarily effectiveness).  We now need that larger, phase 3 trial.

In the meantime, it’s perhaps more a case of give less chloride, find less chloride in the blood’.  And so, we’ll need to decide whether to invest an additional $0.80 per litre on Plasma-Lyte in order to avoid the potential danger of saline-induced hyperchloraemic acidosis.  If you’re going to give a lot of IV fluid, this is probably something we should do.  If you’re not (and I don’t in trauma), then the jury is still out.

[/DDET]

Rick

9 Comments

  1. Anand Swaminathan (@EMSwami)

    Rick – thanks for the nice JC type review. Agree that this doesn’t answer any important questions and raises more like why did these patients with severe trauma get so much fluid? Comparing NS to plasmalyte just not that helpful as we move more and more away from giving any fluids in trauma.

    Reply
    1. richardbody (Post author)

      Thanks a lot for the comment, Anand! I agree – they had a *lot* of fluid in the first 24h and that’s not really reflective of current thinking. It may be that saline-induced hyperchloraemic acidosis will become much less of an issue in major trauma (particularly within 24h) as we move away from such liberal use of IV fluids. Maybe they chose the wrong patient group to start with in the emergency setting. Perhaps they should have chosen sepsis – at least until we find out whether FEAST is an indication that we have our fluid strategy wrong for septic adults too!

      Rick

      Reply
      1. Henry Morriss

        Agree sepsis albumin v plasmalyte coming to Virchester.malyte trail much more useful . Only fluid in trauma is blood military have shown us excess mortality with crystalloid.
        of intrest plasmalyte coming to Virchester critical care soon

  2. richardbody (Post author)

    One of the most brilliant things about #FOAMed is the instant interaction you get with such a fantastic wide range of people. Today I was really honoured to see this comment from Simon Finfer on Twitter:

    For anyone who doesn’t yet know, Simon Finfer is a trialist (as you can see from his Twitter profile and the link he provides to his university profile) and he makes a really good point about the intention to treat principle in this trial. Technically, by the intention to treat principle, all patients who were randomised should be included in the final analysis whether they followed the protocol or not. This stops us from over-estimating treatment effects. For example, if a new drug has terrible side-effects that cause many people to stop taking it, we might over-estimate the overall benefit of treatment if we exclude those who stopped the drug due to side-effects (to do so would be called a ‘per protocol’ analysis).

    In this trial, as Simon points out, 29.2% of patients who were initially randomised weren’t included in the final analysis. So why did I suggest that the authors followed the intention to treat principle? Well, there were some interesting quirks to this trial. Because the investigators included patients with major trauma (who were critically unwell) they obtained a consent waiver from their IRB/ethics committee. This meant that patients could be entered into the trial before they’d provided any consent. (It would have been impossible to do the research otherwise). However, they had to obtain consent *after* randomisation in order to take blood samples at 24h to assess the efficacy of the treatment.

    7 patients in each group (most of the drop outs) declined to provide that consent. These patients were excluded from the analysis. I thought it was fairly reasonable to exclude them from the analysis as patients who decline consent would ordinarily be simply thought of as ‘screening failures’ (and thus not analysed); and there’s an ethical question to ask about whether you can continue to use the data of patients who actually declined consent to participate. (They weren’t impartial, incapacitated or not approached – they actually declined).

    On removing those patients, we can see that 1 of the remaining patients didn’t follow the full protocol. This patient was included in the final analysis (just as you should in an intention to treat analysis). 4 patients had some missing data (including 2 who had no data about the primary outcome because they declined the blood test after 24 hours) *but* the authors handled this by imputing the data – so they had at least thought of how they were going to handle missing data.

    That means I think they made some reasonable efforts to follow the intention to treat principle. But was it enough? Should they have imputed data from the patients who didn’t provide consent? Would it even be ethical to do so? Because such a high proportion didn’t consent, should we ignore all the data and repeat the study until a higher proportion of those approached consents?!

    Personally I’m in favour of cutting the authors some slack on this one – but I’d be very interested to hear other opinions. Ultimately, this highlights one of the problems of emergency consent. For the future, if a trial will include patients under a consent waiver and randomise patients before obtaining consent, we need to think about the implications for the analysis while the trial is still being designed. How many patients might be lost at a later stage and how will this affect the value of the trial?

    Lastly, this does also raise the question about whether the investigators did everything they could to minimise the potential that patients wouldn’t consent. Arguably, doing 6-hourly arterial blood gases simply for research is a big ask for anyone, let along a patient who already has multiple injuries. Perhaps they should have made the protocol easier on the participants – and used venous sampling rather than arterial.

    Many thanks to Simon Finfer for asking such a thought-provoking question! I’d really love to know what you think about this, too.

    Rick

    Reply
    1. James French

      I wonder if plasma lyte has the same affect on thromboelastography that ringers lactate does…..

      Reply
    2. Point Dexter

      Thanks for a good post Rick.

      The misuse of the term ‘intention to treat’ is surprisingly widespread – in fact the CONSORT guidelines recommend the term is no longer used to avoid confusion.

      With over a quarter of patients excluded from the primary outcome in the presented analysis there must be some concerns regarding selection bias (however understandable the reasons for loss the drop outs are!).

      The authors used multiple imputation in their sensitivity analysis of consented patients without ABG results, and a scenario analysis assuming favourable and unfavourable ABG values for the declined consent patients could have demonstrated the robustness of results.

      One further minor thought: although the reported variables showed good balance, what about unreported, unmeasured or unknown case mix variables (e.g. proportion with head injury, other comorbidities, shock index etc). With the relatively low sample size could there be a risk of confounding from chance imbalances?

      Keep up the good work!

      Reply
  3. Simon Finfer

    From their consort diagram they excluded one patient who met eligibility criterion (should have been included) four who did not complete protocol (did not get 24h blood draw) and 14 in who consent not obtained.
    Emergency consent is a major issue in acute care trials and must be considered in design. We use delayed consent widely in our major trials (SAFE, CHEST, NICE SUGAR etc.) In SAFE delayed consent was withheld or prior consent was withdrawn in 57 of 6997 patients = 0.8%. We aim to keep loss of primary outcome to 5% taken to indicate high risk of bias or lower quality trial.
    Young et al set out to do difficult trial in difficult population. I would probably have done not much better with that trial in that population which is why I would probably have settled on a different design. The process of obtaining consent is a potential source of bias, especially if consent is obtained after investigators know treatment allocation – may cause same problems as randomisation without allocation concealment (does not appear to be case in this trial).
    Young et al clearly inform themselves and others that different trial design needed for a phase III trial. Another “Young” – Paul Young @dogICUma is doing SPLIT using cluster RCT design which overcomes consent issue.
    Clearly a large effectiveness resuscitation trial of “balanced” salt versus saline is needed and we (@icuresearch, @jamyburgh, @dogICUma) would like to do it. Careful design needed to minimise loss to follow up and loss of primary outcome. Primary outcome should be mortality and trial will need to be like SAFE and CHEST and recruit 7000 participants.
    (That is likely my final word as I do have a lot of work related to my day job)

    Reply
  4. Tom Elic

    So to summarise, we have a methodologically flawed trial that demonstrates non-clinical relevant differences for a product i don’t have access to (plasma lyte) against a product I very rarely use (0.9% saline), in quantities I wouldn’t use, and are both worse than an alternative (blood, evidenced by PROMMT). Where’s the study comparing initial low volume use of blood vs help vs saline vs Hartmans (which seems far more widely used than Plasma lyte)? All those guys trying to stick blood in helicopters should be interested in that study.

    Reply
  5. Pingback: The LITFL Review 136 - LITFL

Thanks so much for following. Viva la #FOAMed

Translate »