Nonlinear Association of Serum Sodium with Urine Output in Postoperative Pediatric Cardiac Patients Receiving...

Furosemide resistance and fluid overload remain major challenges in postoperative pediatric cardiac patients. Although adult studies suggest synergistic effects of hypertonic saline with loop diuretics, pediatric postoperative physiology differs substantially. Data describing the relationship between serum sodium and diuretic responsiveness after congenital heart surgery remain limited. We conducted a retrospective observational cohort study of 264 pediatric patients (0-18 years) admitted to a tertiary cardiac intensive care unit following cardiac surgery (2021-2023) who received intravenous furosemide. The analytic window comprised postoperative days 1-2 (POD1-2), yielding 528 patient-day observations. The primary outcome was total urine output normalized to body weight (mL/kg/day); the primary exposure was mean daily serum sodium (mEq/L). Linear mixed-effects models with patient-level random intercepts were used. Serum sodium was modeled using natural cubic splines (3 degrees of freedom) to allow for nonlinearity, adjusting for furosemide dose, log(VIS + 1) as a marker of vasoactive support, albumin, creatinine, fluids administered, postoperative day, and infusion mode (continuous infusion vs. bolus). To complement the spline model with a clinically interpretable summary, we additionally fit a piecewise threshold model with a data-driven changepoint. To address temporal-direction concerns from same-window measurement, we performed prespecified lagged-exposure (POD1 sodium → POD2 urine output and POD2 → POD3) and within-patient first-difference sensitivity analyses. The strength of unmeasured confounding required to attenuate the primary association to the null was quantified via E-values for the point estimate and the lower bound of the 95% confidence interval. Multiple imputation by chained equations (m = 30) was used for modest missingness on laboratory covariates; the primary outcome and furosemide dose were essentially fully observed. Mean age was 2.7 ± 3.7 years (median 1.0 [IQR 3.6]). The sodium-urine-output association followed a steep-then-flat shape: predicted urine output rose with sodium across the hyponatremic range and plateaued near low-normal values; a data-driven changepoint analysis on a single imputed dataset estimated the slope transition at serum sodium ≈ 136.5 mEq/L (95% CI 133.5 to 139.5). The formal test for nonlinearity was not significant (χ²=3.24, df = 2, p = 0.198). Translated to a clinically standard 5 mEq/L contrast, correcting sodium from 135 to 140 mEq/L was associated with + 9.48 mL/kg/day urine output (95% CI 7.24 to 11.72), while no significant change was observed for 140 to 145 mEq/L (+ 0.90 mL/kg/day, 95% CI - 1.29 to 3.09). Daily fluids administered were positively associated with urine output (β = +0.27 mL/kg/day per 1 mL/kg/day intake; 95% CI 0.18 to 0.36; p < 0.001), the most precisely estimated covariate in the model. Vasoactive support, entered as log(VIS + 1), was independently associated with greater urine output (β = +11.11 mL/kg/day per unit log[VIS + 1]; 95% CI 7.13 to 15.10; p < 0.001). Continuous furosemide infusion was associated with + 24.76 mL/kg/day greater urine output than bolus dosing (95% CI 7.15 to 42.37; p = 0.006). Furosemide total daily dose was not independently associated with urine output after adjustment (β = -0.27 mL/kg/day per mg/kg/day; p = 0.85). Serum sodium did not modify furosemide responsiveness (interaction likelihood-ratio test χ²=2.26, df = 3, p = 0.520). The primary association was directionally consistent in lagged analyses (POD2 sodium → POD3 urine output: contrast + 9.22 mL/kg/day, 95% CI 7.45 to 11.00; POD1 → POD2: +6.83 mL/kg/day, 95% CI 5.41 to 8.25). The E-value for the lower 95% confidence bound of the primary contrast was 1.75. In postoperative pediatric cardiac patients receiving intravenous furosemide, higher serum sodium in the hyponatremic range was associated with substantially greater urine output, with the association plateauing near 136 mEq/L. Vasoactive support intensity was an additional independent positive correlate of urine output, while serum sodium and furosemide dose did not interact. These observational findings are most consistent with serum sodium serving as a marker of physiologic recovery and renal perfusion rather than a directly modifiable mediator of diuresis. Prospective studies are needed to test whether sodium-modifying or infusion-mode interventions improve diuretic responsiveness and clinical outcomes.