Investigators used the MIMIC-IV intensive care database to assemble a retrospective cohort of ICU stays that contained timing-validated vancomycin trough measurements. The primary analysis included N = 255 encounters. The study required at least 80% capture of electronically recorded medication administration records (EMAR) during the prior 24 hours, a delta serum creatinine ≤0.2 mg/dL across the 48 hours before the trough, and a trough measured within 72 hours of the first intravenous vancomycin dose. These filters were applied to limit confounding from recent large changes in renal function and incomplete dosing documentation.
The exposure variable was net positive fluid balance in the 24 hours before the trough, modeled per +5 L increments. The primary pharmacokinetic endpoint was the natural-log transformed vancomycin trough concentration. Multivariable models adjusted for contemporaneous serum creatinine, the administered vancomycin dose in the prior 24 hours, and the number of administrations in that same window. The authors performed sensitivity analyses expanding the allowable time from first dose to trough (within 48 hours and within 96 hours) to examine early-therapy window effects.
Across the primary cohort the median trough was 14.00 mg/L with an interquartile range of 10.15 to 18.15 mg/L. Sixty-two of 255 troughs met a low-target threshold, defined in the report as below 10 mg/L, representing 24.3% of the primary sample. Those descriptive data provide context for why determinants of low early troughs matter in critically ill populations where achieving therapeutic exposure can be challenging.
The adjusted point estimate suggested lower early trough concentrations with greater net positive fluid balance: specifically, an 11.7% lower trough per +5 L (β = -0.1247; SE = 0.079; p = 0.115). The reported 95% confidence interval ranged from a 24.4% decrease to a 3.1% increase, and therefore included no association by conventional significance thresholds. Sensitivity analyses using troughs measured within 48 hours (N = 162) and within 96 hours (N = 289) produced similar point estimates; however, all of those confidence intervals likewise included no association. The study team explicitly described these results as hypothesis-generating rather than definitive.
Serum creatinine had a positive coefficient in adjusted models (β = 0.2995; p = 0.061), indicating higher creatinine associated with higher log-transformed troughs in this dataset, though this estimate did not meet the conventional threshold for statistical significance. The investigators explored interaction by hypoalbuminemia (albumin < 3 g/dL) in a measured subset. That exploratory interaction analysis was limited by substantial missingness of albumin measurements and small subgroup counts, producing imprecise effect estimates and preventing firm conclusions about effect modification by albumin status.
This analysis used deidentified MIMIC-IV electronic health record data and applied strict inclusion criteria to reduce confounding by rapidly changing renal function and incomplete dosing records. The authors reported point estimates that imply lower early vancomycin troughs with higher 24-hour net positive fluid balance, but the confidence intervals included the null. The albumin interaction remained imprecise because of missing data and small subgroups. The report did not provide prospective AUC measurements or detailed dosing algorithms beyond EMAR capture criteria. The authors concluded these findings are hypothesis-generating and stated that prospective validation is required to clarify whether fluid balance associates with early trough concentrations and with AUC exposure. Future work should prospectively measure fluid status, vancomycin exposure (including AUC), and albumin to determine whether fluid accumulation causes clinically meaningful dilutional or distributional changes in early vancomycin pharmacokinetics.