Comparison of transpulmonary thermodilution measurements of global end-diastolic volume index, extravascular lung water index and central venous pressure with radiographic estimation of these parameters using computed tomography

Introduction

In critical illness, optimization of the fluid status is of central importance. This study aims to investigate whether radiographic estimation of the global end-diastolic volume index (GEDI), the extravascular lung water index (ELWI) and the central venous pressure (CVP) using computed tomography (CT) is able to contribute to an early, noninvasive evaluation of fluid status.

Methods

Thirty-two CT scans, 26 ICU patients. Estimation of GEDI, ELWI and CVP using CT. Transpulmonary thermodilution using the PiCCO within at most 6 hours (mean 2.25 hours) before or after CT.

Results

The diagnostic accuracy of CT-based estimation of the volume status (GEDI <680, 680 to 800, >800) was 25%. Sensitivity and specificity at diagnosis of hypovolemia were 0% and 100%, respectively. The positive predictive value (PPV) for hypovolemia was 0%. The negative predictive value (NPV) was 74%. In prediction of a hypervolemia, radiographic estimation had a sensitivity of 89% and a specificity of 62% (PPV 36%, NPV 96%). CT-estimated and PiCCO-assessed GEDI values were significantly different (P < 0.005; overestimation of CT-estimated GEDI in 91%). CT-based estimation of ELWI (<7 or >7 ml/kg) had a diagnostic accuracy of 72%. The sensitivity for the prediction of elevated ELWI was 92% (specificity only 13%, PPV 76%, NPV 33%). The ELWI estimated using CT and the PiCCO-assessed ELWI were significantly different (P = 0.029; underestimation of CT-estimated GEDI in only 6%). In prediction of CVP (1 to 9 or >9 mmHg) the estimation using CT had an accuracy of only 53%. Sensitivity for prediction of hypervolemia was only 48% (sensitivity of 80%, PPV 93%, NPV 22%).

Conclusion

Estimation of hemodynamic parameters using CT is difficult. Estimation of the GEDI is not accurate, sensitive or specific for prediction of hypovolemia. At prediction of hypervolemia, radiographic estimation of the GEDI is suitable. At assessment of elevated ELWI, the radiologist overestimates ELWI values (high sensitivity 92%, poor specificity, accuracy 72%). At estimation of CVP, the radiographic estimation is not sufficiently accurate, sensitive or specific. Values for the GEDI and ELWI assessed using CT or PiCCO are significantly different. Diagnostic sensitivity/specificity of radiographic estimation probably can be improved by interdisciplinary training.