Role of End-Tidal Carbon Dioxide as a Non-invasive Predictor of Fluid Responsiveness in Mechanically Ventilated Patients with Shock
Abstract
Background: Circulatory shock remains a leading cause of mortality in the intensive care unit, with inadequate tissue perfusion and impaired oxygen delivery as its defining pathophysiological features. Fluid resuscitation is central to early haemodynamic management, as volume expansion increases venous return and augments cardiac output through the Frank–Starling mechanism. However, only approximately half of haemodynamically unstable patients demonstrate a meaningful rise in cardiac output following a fluid bolus, and indiscriminate fluid administration in non-responsive patients carries well-recognised risks including pulmonary oedema, prolonged mechanical ventilation, and increased mortality. Accurate identification of fluid responsiveness prior to fluid administration has therefore become a fundamental component of modern critical care practice. Static haemodynamic parameters such as central venous pressure have been shown by meta-analysis to be unreliable predictors of fluid responsiveness, and dynamic indices such as pulse pressure variation and stroke volume variation, while more accurate, require invasive arterial monitoring and strictly controlled mechanical ventilation. End-tidal carbon dioxide (EtCO2), can be measured continuously by capnography in all mechanically ventilated patients, reflects pulmonary blood flow and hence cardiac output under stable ventilatory conditions, and may offer a simple, non-invasive bedside alternative for predicting fluid responsiveness. Methods: This prospective observational study was conducted in the Intensive Care Unit of Dhaka Medical College Hospital, Dhaka, Bangladesh, between November 2019 and March 2021. One hundred adult patients receiving controlled mechanical ventilation with clinical evidence of shock requiring fluid challenge were enrolled. Patients with significant pulmonary disease, unstable ventilator settings, severe arrhythmias, or contraindications to fluid challenge were excluded. Each patient received a standardized fluid challenge of 500 mL isotonic crystalloid. Cardiac output was measured by transthoracic echocardiography using left ventricular outflow tract velocity-time integral (LVOT-VTI) before and 15 minutes after the fluid challenge. EtCO2 was recorded simultaneously using mainstream capnography. Patients demonstrating a ≥15% increase in cardiac output were classified as fluid responders, consistent with established thresholds in fluid responsiveness research. Pearson correlation and receiver operating characteristic (ROC) curve analysis were used to assess the relationship between EtCO2 change and cardiac output change, and to determine the diagnostic performance of EtCO2. Results: Of 100 patients, 45 (45%) were classified as fluid responders and 55 (55%) as non-responders. Fluid challenge produced significant increases in LVOT-VTI, stroke volume, cardiac output, and EtCO2 across the whole cohort (all p < 0.001). Responders demonstrated a significantly greater percentage rise in cardiac output compared with non-responders (20.27 ± 3.90% vs 7.18 ± 3.48%; p < 0.001), with a corresponding difference in percentage change in EtCO2 (8.87 ± 2.93% vs 3.58 ± 2.03%; p < 0.001). Pearson correlation demonstrated a statistically significant moderate positive correlation between percentage change in cardiac output and percentage change in EtCO2 (r = +0.678; p < 0.001). ROC curve analysis identified a ≥4.5% increase in EtCO2 as the optimal cut-off for predicting fluid responsiveness, yielding a sensitivity of 97.8%, specificity of 72.7%, positive predictive value (PPV) of 74.6%, negative predictive value (NPV) of 97.6%, and overall diagnostic accuracy of 84%. The area under the ROC curve (AUC) was 0.949 (95% CI: 0.908–0.989; p < 0.001), indicating excellent discriminatory performance. Conclusion: A ≥4.5% rise in EtCO2 following a standardised fluid challenge reliably identifies fluid-responsive mechanically ventilated patients with shock, achieving high sensitivity and an excellent AUC of 0.949. The outstanding NPV of 97.6% supports using a flat EtCO2 response as a firm clinical signal to withhold further fluid. EtCO2 monitoring by capnography represents a practical, non-invasive, and readily available bedside tool for guiding fluid therapy, particularly in resource-limited settings where advanced haemodynamic monitoring may be unavailable. Bangladesh Crit Care J September 2026; 14 (2): 90-97