Simon has already covered the recent SMART trial1, regarding the comparison of balanced crystalloid solutions versus normal saline for resuscitation in critically ill adults. However, there was another paper in the same edition of the NEJM looking at a similar comparison in Emergency department patients admitted to hospital, to a non-critical care bed. Appropriately titled the SALT-ED study2, this one makes for further interesting reading and is a good edition to the evidence base.
As always – read the paper yourself. The NEJM is being very good with open access articles like this one at the moment and so there is no reason not to avail yourself of the raw data and stroke your chin thoughtfully in academic contemplation while you pore over the statistical methodology. Have a beer while you’re at it.
Who was studied
The trial population consisted of adults (≥18 years old) who received at least 500 ml of intravenous isotonic crystalloids in the emergency department and were subsequently hospitalized outside an ICU. If you got <500mls in total, you were excluded. If you were <18 or discharged from ED, you were excluded. Otherwise you were in. They used a waiver of consent, meaning they got ethical approval to recruit everyone in the department without approaching any individual for informed consent. They also used attendance as the unique identifier – thus the same patient could come back multiple times and be recruited each time. This is what they mean when they describe the trial as ‘pragmatic’ and they are right; this study looks at the effect of clinicians using normal saline or balanced crystalloid solutions across a large group of heterogenous presentations, assigned randomly. As such, the findings should be valid, potentially generalisable and applicable to routine emergency department patients.
What was the intervention?
Patients received normal saline or balanced crystalloids (Ringer’s lactate or Plasmalyte-A) as resuscitation fluid during their ED stay. Once admitted to a hospital floor, data collection regarding type and volume of fluid ceased. Randomisation is interesting here – the authors used block randomisation, without blinding, by alternate month. So everyone in January got balanced crystalloids, February saline etc… This has the advantage of limited resource use regarding trial personnel – individual recruits do not require any additional research input/activity. However, it does mean that clinicians were aware of the fluid they were giving and when they would be giving it, so could allow the influence of conscious and subconscious bias to affect their overall management strategy.
What were the endpoints?
The primary outcome was hospital free days after 30 days from admission – the number of days each participant spent alive, and out of hospital. I don’t think you can argue against this as a patient related outcome measure; if I was admitted to hospital for any reason I would be interested in a treatment that would increase the chances of me being alive and out of hospital. However, the validity of this outcome depends a little on how you collect the data – if I was discharged and went home, then off to see my friend in another state, and I developed an acute kidney injury as a result of the previous treatment given, I would be upset. Would the study capture this event? Or would it be considered that I was alive and out of hospital?
To counteract some of these concerns, the authors included several secondary endpoints; Major adverse kidney events within 30 days (defined as death, new need for renal replacement therapy or persistent renal dysfunction) stage 2 KDIGO3 AKI, and in hospital death. I wondered if these primary and secondary outcomes were the wrong way around. In a trial looking at such an innocuous intervention across such a heterogenous population, surely the rationale suggests we should be looking to compare major adverse kidney events rather than all cause death/hospital free survival.
Tell me about the methodology
Data extraction was performed by electronic health record review. This is really interesting. It means no challenging nurse follow up, interviews and awkward telephone calls. However it does mean that some of the granularity would not be captured – for instance if a patient had experienced an adverse event but never been asked about it, it would not be in the electronic record. It also means that the results are heavily reliant on the accuracy and completeness of routine clinical entry. For the outcomes of mortality and acute kidney injury, that is probably OK but still needs to be understood. The authors describe this as a pragmatic strength.
Their power calculation is again interesting – the study was designed to show that use of balanced crystalloid could increase hospital free survival by 0.5 days as a primary outcome. A superiority trial. Given their pragmatism around design and absence of challenges to recruitment, they hypothesised that they would be able to include 14,000 patients, which would give them 90% power and an alpha value of <0.5, meaning that they could be fairly confident that their results would be valid. This is an interesting point to power your study on though. Again, physiological rationale has us questioning the use of normal saline due to perceived risks of hyperchloraemic metabolic acidosis, and adverse renal outcomes. I am more interested in these differences between groups, than I am of whether a patient is discharged home in the morning or afternoon. We all know how subjective these discharge decisions are, and how they often fail to reflect the underlying severity of disease. Half a day discharge wise can occur as a result of transport, medication, social or sometimes political interest.
Intention to treat and per-protocol analyses were pre-specified which, allows us to assess these trial results in the true context, where we know that some people will use what they are given and some will be real sticklers for demanding their favourite fluid, irrespective of randomisation month.
And the results?
13,347 patients were recruited in 16 months. That deserves a round of applause in itself I think. The split was pretty much 50:50 as you would expect with this randomisation technique and baseline characteristics were pretty equal between groups, as you would hope for.
Patients got a median of 1L during their stay in the emergency department. But the mean was significantly higher at >1.6L. This is a good example of a population that likely had a skewed distribution, with a small number of patients getting a load of fluid and thus significantly affecting the overall mean value. It is interesting also to see that only 83% of patients randomised to balanced crystalloids got them, whereas >90% of those randomised to saline received it. This means that 1 in every 5 cases, a clinician was kicking off about the idea of using a balanced crystalloid solution and insisting on using something different (saline presumably). This speaks volumes to the variability in international practice – I would have intuitively thought this to be the exact opposite. Just goes to show why this research is needed.
But onwards to the primary outcome – There was no difference in the number of hospital-free days between patients in the balanced-crystalloids and saline groups (median, 25 days in each group; adjusted odds ratio with balanced crystalloids, 0.98; 95% confidence interval [CI], 0.92 to 1.04; P=0.41. This means the median length of stay was 5 days in each group. No difference. Nada. Zip. Zilch.
Secondary outcomes are again, perhaps where the meat of the discussion lies. Patients in the balanced-crystalloids group had a lower incidence of major adverse kidney events within 30 days than those in the saline group (4.7% vs. 5.6%; adjusted odds ratio, 0.82; 95% CI, 0.70 to 0.95; P=0.01). Now if we were in support of balanced crystalloids all along, we would rejoice at this and highlight the significance testing, as well as the large numbers of recruited patients implying that this is highly unlikely to be a type 1 error. If we were saline fans, then we might point out that this was only a secondary outcome, a composite (shudder) and that absolute risk reduction is marginal (0.9%). As such you would need to prescribe balanced crystalloid to over 100 patients in the ED, instead of saline, in order to avoid 1 major adverse kidney event. We would also point out the limitations in blinding and imply that patients prescribed crystalloid may have had differences in care, driven by a desire to prove the benefit of this regime.
The per-protocol analysis essentially generates the same results, which is reassuring in some ways. But perhaps more interesting is the heterogeneity analysis of treatment effect. This figure essentially provides a forest plot broken down by pre-specified subgroups, to provide a nice visual representation of whether balanced crystalloids were more effective in preventing primary and secondary outcomes in particular subgroups. The results that jump off the page here are of the subgroups with either a high baseline chloride or a degree of baseline AKI. Among patients who presented to the emergency department meeting KDIGO criteria for stage 2 or higher acute kidney injury (1274 patients), resolution of acute kidney injury during hospitalization was more common with balanced crystalloids, as shown by a lower incidence of major adverse kidney events within 30 days in the balanced-crystalloids group (28.0%) than in the saline group (37.6%) (P<0.001). An absolute risk reduction of 9.6%? A number needed to treat of 10? Now that’s a result that I can get on board with…
I see – is that the argument locked up?
Well…….there are limitations as pointed out previously. In addition, this is a single centre study and only 5% of balanced crystalloid patients received Plasmalyte-A, so if this is your particular brand of poison then I am not sure you can consider this study robustly generalisable. The authors also mention the outcome ascertainment in the limitations section and this talks to the concerns mentioned earlier regarding use of electronic health records. Essentially, the possibility remains that they could have missed relevant outcomes in this study. Fluid provided after admission was not recorded within the study – as such patients could have had litres of anything after admission, all of which could very realistically impact on outcome. There is no data on this, so we cannot reassure ourselves about the baseline total (and type of) fluid administration.
It needs to be remembered that as a superiority trial, these results mean we essentially have to accept the null hypothesis and report that the authors were unable to prove that balanced crystalloids in the emergency department are superior to saline. However, the data on major adverse kidney events is interesting, and appears to have real validity when you look at the subgroup analysis.
And the take-home?
What can we read into this? Well, we can tell our pharmacists that although balanced crystalloids don’t seem to reduce length of stay for patients admitted from the ED, they seem overall to not cause any adverse events. They also have sound clinical rationale, and in the event that these poor unsuspecting ED patients have an AKI, then use of balanced crystalloids potentially reduces the risk of major adverse renal outcome within the next 30 days by a decent amount. And the cost is roughly the same. So actually, what’s not to like about balanced crystalloids?
Do we need more research? Always. As a single centre trial with multiple limitations, this cannot be considered definitive enough evidence to leave saline on the shelf. However, it provides some granularity to the concept of when it may not matter that much what patients get, and when it potentially does. Also, the design of this trial is relatively groundbreaking. There are great lessons here for those of us in research about delivery of a large scale pragmatic randomised controlled trial with all interventions embedded within routine care.
All in all? A result that does not leave a bad (salty) taste in the mouth.
That last line sounds a bit dodgy now I read it back.
I think I’ll stop now.
As always, let us know what you think.
Before you go please don’t forget to…
- Subscribe to the blog (look top right for the link)
- Subscribe to our PODCAST on iTunes
- Follow us on twitter @stemlyns
- PLEASE Like us on Facebook
- Find out more about the St.Emlyn’s team