0.9% saline solution has a supraphysiological chloride concentration and has been associated observationally with acute kidney injury and mortality. How randomized fluid type changes serum chloride early in critical illness, and what determines this change, remain uncertain. Focusing on the first three days after randomization, we estimated the randomized saline-versus-Plasma-Lyte difference in serum chloride and its variation by renal function, the observational association with study-saline volume, and the sources of variation in serum chloride.

This secondary analysis of BaSICS (75 Brazilian ICUs; Plasma-Lyte 148 vs 0.9% saline) included participants with baseline serum chloride and at least one measurement on days 1-3. A Bayesian hierarchical longitudinal model estimated the treatment effect over time and its variation by baseline estimated glomerular filtration rate (eGFR). Group-level analyses used day 2, which balanced accumulated exposure against attrition of measurements. We normalized randomized between-group differences, estimated an inverse-probability-weighted within-saline volume association, and partitioned modeled variation in serum chloride.

Among 3,482 patients (33% of the trial population), serum chloride rose 2.15 mmol/L (95% confidence interval, 1.76 to 2.53) more with saline than with Plasma-Lyte by day 2, when saline recipients had received on average 145.7 mmol more study-fluid chloride, corresponding to 1.47 mmol/L per 100 mmol of additional chloride (1.17 to 1.81). Within saline recipients, the weighted association was +0.31 mmol/L per 100 mmol (0.22 to 0.44). The day-2 between-group difference decreased by 0.29 mmol/L (95% credible interval, 0.12 to 0.46) per 15 mL/min/1.73 m² higher baseline eGFR (+3.0 mmol/L at 30 vs +1.8 mmol/L at 90 mL/min/1.73 m²). Of modeled variation in serum chloride, the fluid component accounted for 0.7% (0.4 to 1.0; 1.8% with randomized arm included); measured baseline characteristics, almost entirely baseline serum chloride, for 48.1%; persistent patient-specific differences not explained by measured characteristics for 24.5%; and day-to-day within-patient fluctuation for 26.9%.

Over the first three ICU days, randomization to saline increased serum chloride relative to Plasma-Lyte, and the between-group difference was larger at lower baseline eGFR. Most modeled variation in serum chloride reflected each patient's baseline chloride and persistent patient-specific differences that measured characteristics did not explain; the fluid component was small.

ClinicalTrials.gov, NCT02875873.