Fluid Resuscitation Volume After Early Echocardiography and Mortality in Sepsis: Target Trial Emulations With Strictly Pre-Exposure Confounder Measurement in MIMIC-IV
Higher fluid volumes after echocardiography in sepsis patients were associated with increased 28-day mortality, and vasopressin initiation also showed a strong association with mortality.
Where it sits
this study against the rest of the vasopressin corpusSummary and findings
This study examined the relationship between fluid resuscitation volume after early echocardiography and 28-day mortality in Sepsis-3 patients. The analysis included 7,819 patients, with a focus on those receiving ≥30 mL/kg of fluid versus <30 mL/kg. The findings indicated an association between higher fluid volumes and increased mortality, particularly in patients receiving vasopressin.
Abstract
<title>Abstract</title> <p>Background Whether echocardiographic phenotypes modify the association between early fluid resuscitation volume and mortality in sepsis is unknown. We emulated three target trials within a single cohort of Sepsis-3 patients with early transthoracic echocardiography (TTE), measuring every confounder strictly before the start of the exposure window. Methods Using MIMIC-IV v3.1 (2008–2022), we identified Sepsis-3 stays with TTE within ± 24 h of ICU admission (first stay per patient; n = 7,819). Emulation 1 (primary) compared post-TTE 24-h fluid volume ≥ 30 vs < 30 mL/kg with a landmark at TTE + 24 h (n = 7,398). Emulation 2 (secondary) compared dobutamine new-users versus non-users within the LVEF < 40% stratum (n = 1,201; 100 new-users). Emulation 3 (secondary) compared vasopressin new-users versus non-users (n = 6,474; 360 new-users). All adjustment covariates—including SOFA components, vasopressor exposure with norepinephrine dose, Sepsis-3 septic shock status, fluid balance and urine output—were measured strictly before the echocardiogram. Overlap weighting was used throughout; propensity scores were re-estimated within phenotype strata. The outcome was 28-day all-cause mortality from each landmark. Effect modification was assessed with formal interaction tests on multiplicative and additive scales. Bias analyses included E-values with hazard-to-risk-ratio conversion, tipping-point grids, and negative control outcomes restricted to diagnoses from prior admissions. Results Post-echocardiography fluid volume ≥ 30 mL/kg was associated with higher 28-day mortality (HR 1.74, 95% CI 1.54–1.97; weighted risk 38.7% vs 24.8%; risk difference + 13.9%, 95% CI 10.8–17.0). The association was similar across LVEF strata (< 40%: HR 1.63, 1.26–2.10, n = 1,361; 40–54%: 2.02, 1.47–2.79, n = 1,180; 55–69%: 1.88, 1.51–2.34, n = 2,752; ≥70% hyperdynamic: 1.39, 1.02–1.89, n = 896) and across RV strata (normal: 1.67, 1.42–1.97, n = 4,814; dysfunction: 1.77, 1.43–2.18, n = 1,979). No formal interaction test supported effect modification (LVEF: multiplicative p = 0.44, additive p = 0.56; RV: multiplicative p = 0.65, additive p = 0.16; LVEF modelled continuously: p = 0.82). A previously observed protective association in hyperdynamic hearts without cardiac disease was not confirmed when cardiac disease was ascertained from pre-admission diagnoses (HR 1.33, 0.90–1.97). The association persisted within pre-TTE SOFA tertiles and septic-shock strata, at 20 and 40 mL/kg thresholds, per 10 mL/kg as a continuous exposure (HR 1.14, 1.11–1.17), with a 6-h exposure window (HR 1.43, 1.27–1.60), and after excluding pre-ICU echocardiograms (HR 1.64, 1.44–1.87). Dobutamine initiation (Emulation 2) was imprecisely estimated (HR 1.57, 0.95–2.60); vasopressin initiation (Emulation 3) was strongly associated with mortality (HR 2.55, 2.13–3.06) without phenotype interaction (multiplicative p = 0.23). The E-value was 2.60 (2.29 for the confidence bound); prior-admission negative control outcomes indicated residual bias acting, if anything, toward the null. Conclusions After strict temporal alignment of confounder measurement, exposure and follow-up, higher fluid volume in the 24 h after early echocardiography was consistently associated with higher 28-day mortality in sepsis, but we found no statistically supported evidence that LVEF or RV phenotype modifies this association. These results do not support phenotype-guided fluid restriction derived from observational subgroup patterns; adequately powered randomised trials with protocolised echocardiography remain necessary.</p>
Background
This paper investigates whether echocardiographic phenotypes influence the relationship between fluid resuscitation volume and mortality in sepsis. Prior research has indicated variability in outcomes based on fluid management strategies in septic patients. Understanding this relationship is crucial for optimizing treatment protocols in sepsis management.
Methods
The study utilized a cohort from MIMIC-IV v3.1, identifying Sepsis-3 patients with transthoracic echocardiography (TTE) within ±24 hours of ICU admission, totaling n=7,819. The primary analysis compared post-TTE fluid volume ≥30 mL/kg versus <30 mL/kg with a landmark at TTE + 24 hours (n=7,398). Secondary analyses evaluated dobutamine and vasopressin initiation in specific subgroups.
Results
The primary endpoint indicated that post-echocardiography fluid volume ≥30 mL/kg was associated with higher 28-day mortality (HR 1.74, 95% CI 1.54–1.97). This association was consistent across various left ventricular ejection fraction (LVEF) strata and right ventricular (RV) function categories. No significant interaction was found to support effect modification by LVEF or RV phenotype.
Interpretation
The results align with previous findings suggesting that excessive fluid resuscitation may be detrimental in sepsis patients. However, the effect sizes, while statistically significant, may not be clinically meaningful for all subgroups. Limitations include potential confounding from the observational design and the need for randomized trials to validate these findings.
Key findings
- HR 1.74, 95% CI 1.54–1.97; weighted risk 38.7% vs 24.8%; risk difference +13.9%, 95% CI 10.8–17.0.
- HR 1.63, 1.26–2.10, n=1,361 for LVEF <40%.
- HR 2.02, 1.47–2.79, n=1,180 for LVEF 40–54%.
- HR 1.88, 1.51–2.34, n=2,752 for LVEF 55–69%.
- HR 1.39, 1.02–1.89, n=896 for LVEF ≥70% hyperdynamic.
- HR 2.55, 2.13–3.06 for vasopressin initiation.
Limitations
- observational study design may introduce confounding
- results may not be generalizable beyond the studied cohort
- short follow-up period of 28 days
- no randomization to confirm causality
- potential residual bias indicated by negative control outcomes