In Dec 2014 our Australian colleagues Yunos et al. published an extended analysis of their original study looking at chloride-containing fluids and the incidence of AKI in ICU. The original study considered two consecutive 6-month study periods, the first being a period of unrestricted fluids as a control, and a second when chloride-containing fluids were restricted (0.9% saline, 4% albumin, gelatin). The extended analysis used similar methodology but the time periods were each 12 months. Although these studies have each been subject to some criticism, both papers concluded that there was a higher incidence of AKI and RRT in the control groups. The extended study also suggested that there were other unidentified confounders which may have contributed to the incidence of AKI.
This study (the SPLIT trial), published in October’s JAMA, aimed to provide further insight into the development of AKI after administration of 0.9% saline and buffered crystalloid solutions in a broad range of ICU patients.
What and where:
The authors published their proposed study protocol in ‘Intensive Care’ last year. The prospective, double-blinded cluster randomised (i.e. they looked at groups of patients rather than individuals) double cross-over (so each group received both the control and intervention, twice) study was conducted between Apr-Oct 2014.
- Objective: To determine the effect of a buffered crystalloid compared with saline on renal complications in patients admitted to the ICU
- 4 tertiary ICUs in New Zealand; 3 general, 1 cardiothoracic/vascular (in Australasia CT surgeons also do aortic work).
- 90-day follow-up.
- Inclusions; all receiving crystalloid fluid therapy.
- Exclusions; current RRT for end-stage CKD, RRT within 6 hours of admission, admission for organ donation or palliation, previous enrolment in the study.
- Crystalloids; 0.9% saline vs Plasma-Lyte . The dose was determined by the treating clinician.
- Double-blinded (both investigator and clinician blinded). Computer randomisation.
- Double-crossover; Used one fluid for 7 week interval and then switched, This process was repeated (hence 'double') to give a total 28 week study period.
- No restriction on other fluids or therapies.
- Primary; Proportion of AKI (≥ ‘I’ in RIFLE classification).
- Secondary; ∆ creatinine (pre-admission to peak creatinine), cumulative incidence of AKI defined by RIFLE and serum creatinine thresholds, use of RRT in ICU plus indications, RRT post-hospital discharge, need for and duration of mechanical ventilation, ICU LOS, proportion of patients needing readmission to ICU, ICU and hospital all-cause mortality, ICU and hospital cause-specific mortality.
- Primary outcome and hospital mortality examined in pre-defined subgroups; admission diagnosis of sepsis, admission diagnosis of trauma +/- TBI, cardiac surgical admission, pre-enrolment APACHE II ≥ 25.
- No fixed sample size (defined study period and there is no established statistical methodology for prospective cluster randomised, double-crossover studies with binary outcome variables - therefore this was a feasibility study).
- Intention-to-treat analysis.
- Binary outcomes compared using relative risk, 95% confidence intervals and chi squared tests.
- Continuous outcomes analysed using mixed linear modelling with results as differences or ratios with 95% confidence intervals.
- Missing primary outcome data in 5%, their data was analysed using extreme case scenarios i.e. both best and worst case scenarios)
- p-value ≤ 0.05 statistically significant.
- Pre-defined subgroups.
- Nearly all eligible included, 2278 patients enrolled - 10 Plasma-Lyte and 6 saline patients ‘opted out’
- Baseline characteristics were similar (although with crossover studies this is slightly less important).
- 57% post-op (majority were elective cardiac surgery unsurprisingly given the ICUs involved) 14% of patients came from the ED.
- Primary outcome: 102 in buffered crystalloid group developed AKI, 94 in 0.9% saline group (p values for individual RIFLE classes not significant). Total AKI rate 9%, with RRT rate 4%.
- Secondary outcomes: No statistical significance in any of the secondary outcomes.
- One reported adverse event from a patient admitted post-renal transplant who received buffered crystalloid. They developed lactic acidosis > MOF > death.
- 66% clinicians who responded to the questionnaire ‘guessed’ the fluid being administered, questioning the effectiveness of the blinding.
Paper conclusions and authors discussion:
- There is no statistically significant difference in the incidence of AKI within 90 days of ICU admission when giving 0.9% saline or buffered crystalloid.
- There is no statistically significant difference in key secondary outcomes.
- Use of colloids in other studies could account for the increased incidence of AKI and therefore account for lack of difference seen with this study.
- The study design reduced bias compared to previous observational studies and a primary end point using serum creatinine eliminates observer bias.
An interesting interview podcast with the author can be found here, or you can download it via iTunes (search for intensive care network podcasts)
- Large multi-centre randomised study (and New Zealand is not too different from the UK)
- Included a broad spectrum of presentations
- Structured and adhered-to protocol which was published pre-trial.
- Information from this study can be useful to guide the design of an RCT
- Despite blinding 2/3 of clinicians were able to identify the fluid they were giving
- 90% had fluids administered prior to ICU admission; mainly Plasma-Lyte
- Low volumes of fluid administered (median 2L), with the majority < 24hrs post-admission, therefore limited exposure in a low risk population
- Serum chloride levels were not included and so we don’t know if any patients developed hyperchloraemic acidosis and AKI. Hence, doesn’t help clarify if hyperchloraemic acidosis leads to AKI
- High risk groups under represented and therefore unknown benefit/harm for this population (mean APACHE II 14 and majority elective post-op)
- Low background mortality in the study population reduces the power for a clinically important change in mortality.
- Unable to consider a sub-clinical effect.
- Not an effectiveness trail but a feasibility study for future work.
The take home message from this paper is that in small volumes (approximately 2L) of 0.9% saline and Plasma-Lyte do not cause AKI or increase mortality in low-moderate risk patients. This is great, but the discussion points below need to be considered:
- Plasma-Lyte was chosen as the buffered crystalloid in this study due the preference of this over Hartmann’s and CSL in the ICU’s studied. In our ICU we use CSL, which although similar, does have significant differences to Plasma-Lyte. Firstly, it should be pointed out that no solution is ‘physiological’, however Plasma-Lyte has the benefit of being more isotonic and containing less chloride. However, Plasma-Lyte does contain acetate, which has been shown to cause vasodilatation and is pro-inflammatory. Also, sodium gluconate may act as an osmotic diuretic. It is not an absolute given to assume the study can be translated to Hartmann’s solution or CSL.
- Sepsis is the main reason for admission to ICU in the UK but few patients with sepsis were included in the study. Sepsis-induced AKI has a different pathophysiology to other causes of renal injury and therefore potentially changes the susceptibility to AKI.
- Many of the patients we admit have either received large volumes of IVT on the wards/ED or require fluid resuscitation greater than 2L of crystalloid.
- The typical Sunderland patient often has multiple significant co-morbidities, unlike the study population, again potentially increasing their susceptibility to AKI.