Sodium chloride or plasmalyte‐148 for patients presenting to emergency departments with diabetic ketoacidosis: A nested cohort study within a multicentre, cluster, crossover, randomised, controlled trial

To test the hypothesis that fluid resuscitation in the ED with plasmalyte‐148 (PL) compared with 0.9% sodium chloride (SC) would result in a lower proportion of patients with diabetic ketoacidosis (DKA) requiring intensive care unit (ICU) admission.


Introduction
Diabetic ketoacidosis (DKA) is a common acute complication of diabetes mellitus (DM) which leads to ED presentations, hospitalisations and sometimes intensive care unit (ICU) admission. The presentation to hospitals in Australia with DKA has increased over the last decade mainly in outer metropolitan and • DKA patients treated in ED with plasmalyte-148 compared with 0.9% sodium chloride had similar rates of requiring intensive care unit admission after adjusting for admission pH, diabetes type and blood glucose. • Adherence with allocated fluid in the plasmayte-148 group was low compared with 0.9% sodium chloride, raising the possibility that the benefits of plasmalyte-148 may have been underestimated in the study.
regional ICUs. 1 The cornerstones of DKA management are correction of underlying cause, rehydration with intravenous fluids, insulin therapy and electrolyte replacement. 2 While there is broad consensus with regards to insulin and electrolyte replacement, the choice of the optimal fluid for management of DKA remains unknown. Current international guidance 3,4 and majority of local hospital protocols in Australia 5,6 recommend 0.9% sodium chloride (SC) as the initial fluid of choice. The use of SC, a crystalloid fluid with high chloride content, has been associated with several complications including normal anion gap hyperchloremic acidosis which in turn perpetuates metabolic acidosis and may delay resolution of DKA. 7 Plasmalyte-148 (PL), an isotonic crystalloid fluid, can be safely used as a resuscitation and maintenance fluid and may be a better alternative for patients with DKA. 7 The Sodium Chloride or Plasmalyte-148 Evaluation in Diabetic KetoAcidosis (SCOPE-DKA) trial, 8 a phase 2 cluster crossover randomised controlled trial, tested the hypothesis that PL use in patients with severe DKA requiring ICU admission may result in a faster resolution of DKA than SC. The SCOPE-DKA trial conducted across seven metropolitan and regional sites in Queensland, Australia, randomised 93 patients with severe DKA to receive PL or SC and reported that PL may result in faster resolution of DKA with similar or lesser adverse events. In this trial, the fluid treatment with the study fluid was stipulated to start as soon as an eligible patient was identified in ED, but only patients who were ultimately admitted to an ICU were included in the trial. Therefore, it is possible that if PL is a superior fluid to SC for DKA, then fewer patients treated with PL would have been admitted to ICU from the ED and hence not reported in the SCOPE-DKA trial. Thus, we pre-specified the conduct of this nested cohort study within the SCOPE-DKA trial.
We hypothesised that fluid resuscitation in ED with PL, compared with SC, would result in a lower proportion of patients presenting to ED with DKA requiring ICU admission.

Methods
Ethics approval was obtained from the responsible human research ethics committee (HREC/2018/QRBW/43868) with a waiver of informed consent.
This nested cohort study was conducted within the SCOPE-DKA trial at Caboolture and Rockhampton Hospitals in Queensland, Australia. These two sites have similar ED and ICU characteristics and clinical practice. Additionally, as this was a unfunded study it relied on the extended support by the site investigators, hence it was decided to limit the nested cohort study to these two sites. ICU at both sites and majority of Australia works on a closed ICU model with primary responsibility for clinical care residing with Intensive Care Consultants who hold a Fellowship of the College of Intensive Care Medicine. 9 The ICU at both sites provides both ICU level and HDU level of services for the hospital (ICU level or ICU1 has with 1:1 nursing to patient staffing ratio and HDU level or ICU2 has with 1:2 nursing to patient staffing ratio). 9 The inclusion criteria for SCOPE-DKA were adult patients with age 16 years and above, presentation to the ED or ICU at a participating site during either intervention period with DKA, and requirement for ICU admission as judged by the treating clinician. Severe DKA was defined as arterial pH ≤7.25 (or serum bicarbonate ≤15 mmol/L) and blood glucose level (BGL) ≥14 mmol/L. Exclusion criteria were: patients aged less than 16 years, primary diagnosis of hyperglycaemic, hyperosmolar state and previous inclusion in the SCOPE-DKA trial. Full methodological details are included in the SCOPE-DKA publication and are not repeated here.
Briefly, in the SCOPE-DKA trial, a phase 2 feasibility cluster crossover open label randomised controlled trial, sites were randomly allocated in a 1:1 ratio to use PL or SC for all eligible DKA patients for a 6-month period, followed by a 1-month washout period, and a second 6-month study period where all the sites crossed over to the alternate fluid. Study fluid allocation to sites was revealed 1 week before first study period. All eligible patients with DKA received either PL or SC according to their site allocation for all fluid boluses and maintenance fluids in ED and up to 72 h in ICU, or ICU discharge, whichever occurred earlier. All other aspects of DKA management, including fluid volume and rate, insulin and electrolytes were at the discretion of the treating clinician.
For this nested cohort study within the main SCOPE DKA trial, the inclusion criteria were: adult patients with age 16 years and above presenting to one of the two participating EDs with DKA (defined as presence of blood glucose level >14.0 mmol/L, pH <7.30 and or bicarbonate <18.0 mmol/L and ketosis). Exclusion criteria were age <16 years, primary diagnosis of hyperglycaemic, hyperosmolar state and previous inclusion in this study. Essentially, we included all the patients from both sites who were prospectively enrolled in the SCOPE-DKA trial. In addition to this, we identified all other DKA patients who presented to the participating EDs, who did not require ICU admission, through a retrospective review of EDIS (ED information system used in Queensland Health). This would capture all patients presenting to both EDs'.
The primary outcomes measured were the proportion of patients admitted to ICU. Secondary outcomes measured proportion of patients discharged home from ED, ED length of stay (LOS) and hospital mortality.
Compliance with allocated study fluid was defined using the same definition as in the SCOPE-DKA trial, "the proportion of the total fluid received by each patient that was consistent with the allocation." For instance, if a patient in the PL group received 1000 mL of fluid overall, and 700 mL of this was PL, compliance would be calculated as 70%.

Statistical analysis
Continuous data are presented as mean and standard deviation or median and interquartile range (IQR) depending on their distribution, and categorical data are presented as frequencies and proportions. Univariate analyses were performed using the Mann-Whitney U test for continuous variables and Fisher's exact test for categorical variables. The effect of fluid allocation (PL compared with SC) on the primary outcome: need for ICU admission, was evaluated using multivariable logistic regression with adjustment for type of diabetes and pH at presentation. Sensitivity analyses for primary and secondary outcomes were conducted using a modified as-treated approach. This was performed as it was known from the results of SCOPE-DKA that compliance with fluid allocation was low, particularly in the PL group. To test whether greater exposure to PL had an effect on outcomes, we performed a modified as-treated approach, comparing only patients who had >50% compliance with the allocated fluid. All analyses were performed in Stata 17.0 (StataCorp, College Station, TX, USA).

Patients
There were 84 patients enrolled in the study, 38 (45%) in the SC group and 46 (55%) in the PL group (Fig. 1). The baseline characteristics of both groups are presented in Table 1. The SC group had a significantly higher blood glucose concentration at presentation (median 27.4 mmol/L, IQR 22-37.3) than the PL group (median 20, IQR 13.3-31.1, P = 0.01). The median pH at presentation in the SC group was 7.09 (IQR 7.01-7.21) compared with 7.17 (IQR 6.99-7.26) in the PL group, but this difference was not statistically significant (P = 0.32). The other biochemical parameters at presentation were similar between the two groups ( Table 1). The physiological parameters at presentation were similar between both groups (Table 1).

Fluid volumes
Thirteen patients (34%) in the SC and 22 (48%) in PL group had fluid resuscitation commenced pre-hospital or at the referring hospital. Among those who received prehospital fluids the median volume administered was 1000 (IQR 450-2000) mL of fluids in SC group and 1000 (IQR 535-2100) mL in PL group. The pre-hospital fluids administered were SC.
A median fluid volume of 2150 (IQR 2000-3200) mL was administered during ED stay for SC group compared with 2200 (IQR 2000-3450) mL in the PL group achieving a net fluid balance of 2104 (IQR 1501-2825) mL and 2850 (IQR 1680-4200) mL, respectively, prior to being transferred from ED to the hospital ward or ICU or being discharged. The fluid treatment in ED is summarised in Table 2.
The compliance with allocated study fluid in the PL group was 40% compared with 95% in SC group. The need for dextrose-containing fluids while in ED was minimalmedian 0 mL (IQR 0-312 mL), mean 197.1 mL (SD 356.4 mL) in SC group and median 0 mL (IQR 0-500 mL), mean 259.2 mL (SD 404.7 mL) in PL group.

Other therapies
The dose of insulin was similar in both SC and PL groups with a median use of 18 units (IQR 12-27.5) and 21 units (IQR 14.3-31), respectively, while in ED. The median potassium replacement while in ED was 10 mmol (IQR 0-20) mmol in both groups (    Primary outcome: ICU admission ICU admission was required in 18 (39%) patients in the PL group compared with 19 (50%) in the SC group. After adjustment for pH and blood glucose concentration at presentation and diabetes type in a multivariable logistic regression model, the PL group did not have a significantly different rate of ICU admission compared with the SC group (odds ratio for ICU admission 0.73, 95% confidence interval [CI] 0.13-4.16, P = 0.72). Sensitivity analyses for ICU admission and hospital LOS were conducted using a modified as-treated approach. There were no significant differences in either the rate of ICU admission or hospital LOS in the sensitivity analyses.

Secondary outcomes
Among the secondary outcomes, there were no patients discharged home from ED during the study period, and neither were there any deaths in either group.
The ED LOS was similar for both groups 5.8 h (IQR 2.8-8.2) in SC group compared with 3.9 h (IQR 3-7.6; P = 0.54) in the PL group. Among those who were not admitted to ICU from ED, the median ED LOS was 7.8 h (IQR 4-10.6) in SC group and 5.7 h (IQR 3.9-8.9) in PL group (P = 0.23; Table 2). After adjustment for fluid allocation, pH, blood glucose concentration at presentation and diabetes type, ED LOS was not significantly associated with ICU admission rates (Table S1).
The median hospital LOS among all patients was 3.3 (IQR 2-5) days in the SC group compared with 3 (IQR 3-7.6) days in the PL group (P = 0.62). There was no statistically significant difference in hospital LOS among those who were admitted to ICU from ED and those who were not. Table 2 tabulates the stratified outcomes in both intention-to-treat analysis and as-treated analysis.

Discussion
The present study did not show a significant reduction in ICU admission rate or hospital LOS for patients presenting with DKA when treated with PL compared with SC as fluid therapy. However, the point estimate for ICU admission was in favour of PL and the confidence interval included the possibility of clinically relevant benefits. Further investigation, with adequately powered randomised trials of the early use of PL for patients with DKA presenting to ED, is required.
Between the two study sites, we were able to recruit an 84-patient cohort with similar baseline characteristic in the two groups. There was a trend towards lower ICU admission rates in the PL group 18 (39%) compared with the SC group 19 (50%). This trend, however, did not reach statistical significance. Thus, these results give us some confidence that the observed trends towards better clinical outcomes in SCOPE-DKA (including shorter ICU and hospital stay) with plasmalyte for patient admitted to ICU were probably not influenced by differential ICU admission rates between the plasmalyte and saline groups.
The overall compliance with the study fluid was 95% in the SC group and 40% in PL group, in part because the recommendation for the use of SC for patients with DKA in the Queensland Health DKA guidelines. However, low compliance with study fluid in the PL group would be more likely to bias the findings towards the null hypothesis, as most of the fluid used instead of PL was SC. Thus, this would more likely lead to a Type-2 error, that is, incorrect acceptance of the null hypothesis, rather than a Type-1 error. Even with "dilution" of PL treatment in the PL group, some signal toward reduced ICU admission rate was observed in this study. This warrants further investigation with larger randomised trials.
The rates of adverse outcomes are similar among both groups suggesting that the use of PL in management of DKA is safe compared with SC. There was no hospital mortality among this cohort, which is consistent with expectations for a DKA cohort.
The main strengths of this study were that it was a pre-specified nested cohort study within a randomised controlled trial. The median volume of resuscitation fluid administered in ED is keeping in with published literature. For example, in the DKA subgroup analysis of SALT-ED 10 and SMART 11 trials, patients with DKA received 2.1-2.2 L of fluid while in ED. 12 The study was conducted at two sites with similar characteristics for ED and ICU management for DKA. The utilisation of ICU admission as an outcome is important from a resource utilisation perspective, especially in regional and outer metropolitan hospitals, where ICU beds may be particularly scarce, noting also that these centres have had rising DKA admissions to their ICUs in the last decade. 1 Nonetheless, adequately powered trials in the future should also include patientcentred outcomes such as hospitalfree days.
There were, however, several limitations, mainly poor compliance with study fluid in the PL group. The lack of strict adherence to study fluid administration on presentation may have been because of a proportion of patients being less sick and perhaps considered unlikely to require ICU admission and therefore not recruited into the SCOPE-DKA trial until later in their stay in ED. There were some important baseline imbalances between the two groups because of small sample size. There was a lead time to ED presentation for some patients and substantial volumes of prehospital fluids may have been administered to these patients. While there was no statistically significant difference in ED LOS, it could have been a confounder. The results may not be generalisable to tertiary centres, where there may be a dedicated Endocrinology HDU, which however is uncommon in the Australian regional and outer metropolitan setting.
The poor compliance to allocated study fluid is a limitation which limits the generalisability of results. This should be improved in future research.

Conclusion
Patients with DKA treated with PL compared with SC in the ED had similar rates of ICU admission and hospital LOS. A larger, definitive, randomised trial of fluid therapy for DKA patients is required to elucidate the optimal fluid choice.