Sunday, 1 November 2015

SPLIT : Normal Saline versus balanced solutions - the battle continues...


Paper Link


It has been said that the administration of intravenous fluids is the most commonly prescribed intervention in hospitalised patients.

Its hard to believe then, that the issue remains as controversial as it is.  We really know very little about the use of fluids in clinical practice - best dose, best type, etc - and as usual, when there is a lack of evidence, the void is rapidly filled with mythology.

Nonetheless, years of use has taught us some things, like the fact that when multiple litres of (hardly) normal saline are administered rapidly, the chloride seems to rise and the pH seems to fall.  A number of smaller studies have been performed to explore the impact of these changes, with mixed results.  A before and after study in a tertiary ICU seemed to indicate that chloride restriction (by changing to fluids low in chloride) reduced the risk of renal impairment (1).  A similar study of post-operative patients appears to back this up (2).  However not all studies support this, and a retrospective study of non-surgical sepsis patients found no difference at all (3).  

Have a listen to one of the authors of these papers, Prof Rinaldo Bellomo, discussing the possible pathological mechanisms for this in a podcast I conducted with him a few years ago.

Buffered fluids (such as Ringer's lactate, Hartman's solution or Plasm-Lyte) more closely reflect the chemistry of the plasma, with a strong ion difference that reduces or prevents the acidosis that results from rapid administration.  I explored this topic in more depth in a podcast with Dr John Kellum a couple of years ago, and its well worth a listen.  He points out that while there is evidence that buffered fluids do in fact reduce acidosis, we still don't really know whether this makes any difference to the patient in real terms.

In recent years, major research groups have demonstrated that albumin, the starches and the gelatins have little or no place in resuscitation.  But the question of whether or not balanced solutions out-perform normal saline remains unanswered.  And that's where the SPLIT trial comes in.

At a glance

  • Randomised - cluster randomised by unit
  • Controlled - yes, Plasma-Lyte v 0.9% normal saline
  • Placebo used - no
  • Blinded - yes
  • Population / Sample size calculation - 2278 adult patients in 4 units, mostly post-operative
  • Groups same at randomisation - yes
  • Groups treated the same after randomisation - yes
  • Intention to treat analysis - yes
  • Follow-up - almost complete at 90 days, but 7.5% missing data rate for primary outcome
  • Significant primary endpoint - no difference in AKI rate at 90 days
  • Harms analysed - not specified

Key Question

SPLIT (the Saline v PLasmalye for Intensive care fluid Therapy) trial was designed to examine a pragmatic question : Does the use of Plasmalyte instead of Saline for all indications for the use of a crystalloid solution in adult ICU patients reduce the risk of acute kidney injury.

The Study

The SPLIT trial randomised 2278 adult ICU patients in 4 intensive care units, including 3 mixed medical/surgical units, and one predominantly cardiothoracic unit in New Zealand.  Rather than randomise the patients individually, the units were randomly assigned to either "fluid A" or "fluid B".  Each unit then used the study fluid for 7 weeks, then changed to the other fluid, alternating for 7 weeks at a time for 28 weeks (each group used the each fluid twice).  In this regard, the study was not truly randomised in nature.

Patients were included if they required crystalloid therapy, and thus include a very broad range of patients.  Only patients with established renal failure (about to have or already on RRT), potential organ donors or palliative patients were excluded.  Interestingly, the investigators use the "opt-out" consent method.

The volume and rate of fluid administration was entirely at the discretion of the treating team.  No guidelines were provided on the indications or end points for crystalloid use, strengthening the value of the trial in the real world.

The primary end-point was the occurrence of acute kidney injury within the 90 day follow up period, as defined by a doubling of baseline serum creatinine, or a rise of creatinine to greater than 3.96 mg/dL with an increase of at least 0.5mg/dL.  This is effectively the "injury" classification in the RIFLE criteria (note that the urine output component of the criteria was NOT used).  Secondary endpoints included need for RRT and death in hospital, among others.  Subgroup analyses were planned in for sepsis, trauma, trauma with brain injury, cardiothoracic surgery and severe illness (APACHE 2 score >24)

The Patients

Understanding the patient group in the study is very important to interpreting the results.  

The 2278 patients included in the study were from 4 ICUs in New Zealand.  On average they were 60 years old, and about two-thirds were male.  65% were of european background and a further 20% were indigenous.  

The majority were elective post operative admissions, largely due to cardiothoracic surgery.  By and large, the patients were not overly sick, with a mean APACHE 2 score of 14 and relatively few co-morbidities.  While about two thirds were ventilated, the average ventilation time was only 15 hours, and ICU stay was less than 2 days.  Most patients had received some fluid prior to admission, most of which was Plasmalyte.

Overall, the patients received an average of about 2000ml of fluid during their ICU admission in both groups, and most of this was delivered in the first day of their admission.

The Results

Overall, the risk of acute renal injury (as defined by the Creatinine component of the RIFLE criteria) at 90 days was insignificantly different (9.6% v 9.2), a result reflected across the subgroups.  No other secondary outcome was significantly different.  

Unfortunately, data related to the primary endpoint was missing in over 7.5% of patients.  The analysis was repeated assuming all patients with data missing had renal failure, and then again assuming they all did not.  Neither analysis altered the results significantly

Strangely, Chloride concentrations were not reported, making it difficult to determine whether the patients had been exposed sufficiently to the intervention in question compared with control (known as separation of the groups).  

Additionally, given the premise of the study was that the intervention prevented renal injury, it seems somewhat odd that no imaging or biochemical markers of renal failure were presented.  Its likely though that the authors will argue that this is in keeping with the pragmatic nature of the trial.

Interestingly, at the end of the trial, the authors surveyed clinicians at all 4 hospitals and asked them to guess which fluid was which, and 66% guessed correctly.  Does this call the issue of "blinding" into question?

The rub

So what does all this mean?

The study is an extremely robust examination of the intervention in the population involved - i.e.  it supports the notion that the routine use of saline in small volumes in relatively well post-operative ICU patients is not inferior to the use of a more expensive product like Plasmalyte.  But it seems to stall at this point when the study had the potential for so much more.

The wider application of the results is compromised by the study group.  Acidosis and hyperchloraemia, the purported potential causes of renal impairment, are associated with much larger volumes of normal saline than were given in the study.

As Kellum and Shaw note in the associated editorial, the volume of saline given was likely insufficient, and the study group were likely too well, to generate a "hazard" that can be prevented by the use of a buffered solution.   Together these factors appear to conspire against the authors in their efforts to progress the debate significantly.

Additionally, the pathology of renal failure varies dependent on cause (for example, in sepsis), and the relatively low numbers of these patients in the trial limits interpretation

The authors suggest that this paves the way for a future trial to explore the issue in more depth.  And indeed, this appears necessary.  It's worth noting that no sample size calculation was performed because prior to SPLIT, there was little evidence on which this can be based.  SPLIT essentially now proves the feasibility of future trials in the area.

Finally, the study only compared normal saline with Plasmalyte, not all balanced solutions, and it may not be appropriate to assume the same results would have occurred had Hartmann's solution been used.

Until further studies are conducted in this subject, little more can be gleaned from SPLIT.


  1. Yunos NM, Bellomo R, Hegart C et al.  Association between a chloride liberal vrs chloride restrictive intravenous fluid administration strategy and kidney injury in critically ill adults.  JAMA 2012;30:1566-72
  2. Shaw AD, Bagshaw SM, Goldstein SL et al.  Major complications, mortality and resource utilisation after open abdominal surgery : 0.9% saline compared to Plasma-Lyte.  Ann Surg.  2012; 255:821-29
  3. Raghunathan K, Shaw A, Nathanson B et al.  Association between the choice of IV crystalloid and in-hospital mortality among critically ill adults with sepsis.  Crit Care Med 2014; 42:1585-91

Reviewed by : Dr Todd Fraser

Reviews and Editorials
Additional Resources

Social Media
  • Twitter : 
    • #SPLITtrial
    • Paul Young : @dogICUma

No comments:

Post a Comment