I'd like to thank the organisers of the conference for inviting me and giving me the opportunity to present our work on a systematic review and meta-analysis comparing trials looking at balanced crystalloids versus saline in critically ill adults. My name is Simon Finfer and I'm a professorial fellow at the George Institute for Global Health in Australia and adjunct professor at the University of New South Wales and I also hold the Chair of Critical Care in the School of Public Health at University College London. My disclosures in relation to this presentation are that I'm the principal investigator of a PLUS study for which Baxter Healthcare Australia manufactured and supplied the study fluids. My employer, the George Institute for Global Health, has also received research funding from Baxter and CSL both of who manufacture intravenous fluids for which are used in critical care units. So this is the first published report of the use of intravenous fluid resuscitation in critically ill patients. This was reported by Robert Lewins giving what is effectively a buffered or balanced sort solution to patients during a cholera epidemic in London in the 1830s and reporting a most wonderful and satisfactory effect. Something that all of us who work in critical care occasionally see in our patients. We have at the Australian New Zealand Intensive Care Society Clinical Trials Group have a programme of surveying various aspects of clinical management in critical care and we have demonstrated in our population work published by my colleague Naomi Hammond that whilst in 2007 by far the majority of crystalloid solutions used in our intensive care units was normal saline whereas by 2013 this had changed with buffered or balanced sort solutions notably Hartman's, lactated ringers and plasmolyte were used far more commonly than saline. And two international cross-sectional surveys we have done have shown the same shift in practice occurring internationally. The reason for this I think is well known to most people in that rapid administration of large amounts of saline cause a hypochlorimic acidosis and this has been linked to adverse effects notably on kidney function and also possibly increasing the risk of death. Initially these data were methodologically not particularly strong being observational data and this before and after study conducted by Ronaldo Belomo also in Australia. And then subsequently there have been some high quality randomised controlled trials notably the top one the SPLIT study led by Paul Young a four hospital cluster crossover trial in New Zealand. And then the SMART study led by Matthew Semler which compared balanced crystalloids and saline in the five intensive care units of a single United States medical centre. And then more recently a very large study of 11,000 patients published by the Bricknet group the publication here led by Fernando Zampieri of the BASIX trial which compared plasmolyte with saline in critically ill patients. And finally we have just published online in the New England Journal on the 18th of January of this year and in the print version on March the 3rd the plasmolyte versus saline study which was closely aligned with BASIX. At the same time as publishing PLUS we took the opportunity to conduct an updated systematic review with meta-analysis including the BASIX and the PLUS data and this was published in NEJM evidence on the 18th of January at the same time as the PLUS study. This NEJM evidence is a new journal from the NEJM group, editor-in-chief is Geoff Drazen and being a new journal it is not yet available through a lot of libraries and so whilst you can access this paper at NEJM evidence with Professor Drazen's permission it is also available at the George Institute PLUS media site if you are unable to obtain it from NEJM evidence. So the meta-analysis, the question we were seeking to address was whether in adult critically ill patients balance or solutions compared to 0.9% saline given for fluid therapy being both fluid resuscitation and other fluid administration affected 90-day mortality and other pre-specified outcomes. We included in the meta-analysis randomized clinical trials with individual randomization, cluster randomization and cluster crossover randomized trials. We used standard systematic review and meta-analytic techniques, a systematic search of databases and trial registries with contact to known experts and seeking unpublished data and a minimum of two of the authors independently conducted the screening and eligibility assessment data extraction and risk of bias using the Cochrane risk of bias tool. Importantly given that a large number of these authors have been involved in the trials that are included in the meta-analysis no one was allowed to do any assessment of their own trial. The protocol was registered and is available on a pre-print server. Time restraints don't allow me to go into the methodology in a huge amount of detail but it is documented in that on that pre-print paper and in the supplement available at NEJM evidence. So the outcomes we looked at our primary outcome was 90-day mortality or the closest time point to that in the trials that were assessed as low risk of bias. Secondary outcomes as I said relate predominantly to kidney injury but also to organ support in the intensive care unit measured as vasopressor free days and ventilation free days and mortality at the longest time point reported. We looked at some important patient level subgroups notably patients who had sepsis at time of recruitment to their respective trials, patients with traumatic brain injury, trauma with or without brain injury, patients following cardiac surgery and those with diabetic ketoacidosis. The statistical analysis again is detailed in the published protocol and I will not go into that in any great detail here. Importantly though however we did conduct not only a standard frequentist analysis but also a Bayesian meta-analysis using vague priors in order to provide a posterior probability of one fluid being better than the other. We screened 1301 published reports also unpublished data abstracts and looking at trials that were on the clinical trial registries. We sought 109 reports for retrieval which were assessed for eligibility and this led to the inclusion of 13 studies in the systematic review and meta-analysis. We judged six studies to be of low risk of bias and these six studies included 95% of all the participants. The 90-day outcome, the primary outcome 90-day mortality in the low risk of bias trials is shown here as you can see the point estimate for risk of death for balanced salt solutions versus saline was 0.96 favouring balanced salt solution although the upper limit of the 95% confident divorce was 1.01 so just crossed one meaning that in traditional frequentist view this would be a non-statistically significant result. The confident in divorce encompass a 9% relative reduction in the risk of death and a 1% relative increase in the risk of death. Including the trials that were judged to have more than low risk of bias produced a similar result with a point estimate of 0.93 but with much wider confidence intervals. We also looked at the risk of death for balanced salt solutions versus salt solution for balanced salt solutions. We also looked at the risk of death for balanced salt solutions versus salt solution for balanced salt solutions. The sensitivity analyses looking at other methods for the traditional frequentist approach and the Bayesian meta-analysis produced almost exactly the same results. For the Bayesian meta-analysis using vague priors which means that the analysis did not assume one fluid was better than the other prior to the analysis. The relative risk was again 0.96 with 95% credible interval of 0.88 to 1.04. This produces a posterior probability that of lower mortality with balanced salt solution versus saline of 89.5%. To expand on that a little the probability of 89.5% is that there is a reduction in the risk of death by some amount. i.e. the risk of death is less than 1.0 so that would be even a reduction in the risk of death of 0.0001% for instance would be included in that probability. If we were to say that we would like to see balanced salt solutions reduce the relative risk of death by about 5% so if we have a mortality rate of 20% that would be an absolute reduction in risk of death of 1% then the probability of balanced salt solutions producing that benefit would be less than 40%. So this probability changes obviously dramatically depending on what you consider to be and the treatment effect that would cause you to use balanced salt solutions rather than saline. In terms of the secondary outcomes for acute kidney injury and treatment with renal replacement therapy again the point estimate favours balanced salt solutions with neither of these being statistically significant in the traditional frequentist view although for acute kidney injury the upper limit of the confidence interval is 1.02 but our grade assessment was that we could only have moderate certainty in that result. For ventilator-free days and for vasopressor-free days the mean difference in the number of days 3 to 28 days was very small and this was not statistically significant. We did not conduct Bayesian meta-analyses of any of these other outcomes. No trial reported data on mortality beyond 90 days or on quality of life or functional outcomes at the time of our review. For the sepsis subgroup we saw an effect favouring balanced salt solutions but again the 95% confidence intervals just cross one with an upper limit of 1.01. The majority of this effect coming from the SMART trial with also some effect seen in Basics, no effect seen in PLUS. For traumatic brain injury and this was only, the data were only available from three trials. In PLUS we specifically excluded patients with traumatic brain injury because we believed that saline was the indicated fluid for those patients but in the three trials, SPLIT, SMART and Basics that assessed this, they produced an effect favouring saline with a risk ratio of 1.26 although again the confidence interval just crosses one with a lower limit of 0.98. So suggesting that the treatment effect likely resides within balanced salt solutions producing either a 2% reduction in the relative risk of death or a 60% increase in the relative risk of death in patients with traumatic brain injury. So in summary our results suggest that the use of balanced salt solutions compared to saline results in an effect line between a 9% relative reduction to a 1% relative increase in the risk of death. The results were robust using multiple analytic methods and there is an 89.5% probability that compared with saline balanced salt solutions reduce 90-day mortality by some amount and including very very small amounts. The probability of a relative reduction of 5% or more of using balanced salt solution is less than 40%. Patients with sepsis may benefit from balanced salt solutions rather than saline and whereas patients with a traumatic brain injury probably benefit from receiving saline rather than balanced salt solution. These subgroup effects can be investigated further in a patient level meta-analysis which the authors of our paper are working on at the moment. As I said this paper is available at NEJM Evidence and also at the George Institute Plus media site. I would like to thank all my colleagues listed there for their hard work in producing this meta-analysis to coincide with the publication of the PLUS study and thank you again to the conference organizers for giving me this opportunity to present and to you for your attention.

Professor Simon Finfer

balanced crystalloids

saline

intravenous fluid resuscitation

critically ill patients