Abstract
BACKGROUND: The dead space fraction (V(D)/V(T)) has proven to be a powerful predictor of higher mortality in acute respiratory distress syndrome (ARDS). However, its measurement relies on expired carbon dioxide, limiting its widespread application in clinical practice. Several estimates employing routine variables have been found to be reliable substitutes for direct measurement of V(D)/V(T). In this study, we evaluated the prognostic value of these dead space estimates obtained in the first 7 days following the initiation of ventilation. METHODS: This retrospective observational study was conducted using data from the Chinese database in intensive care (CDIC). Eligible participants were adult ARDS patients receiving invasive mechanical ventilation while in the intensive care unit between 1st January 2014 and 31st March 2021. We collected data during the first 7 days of ventilation to calculate various dead space estimates, including ventilatory ratio (VR), corrected minute ventilation (V˙Ecorr), V(D)/V(T) (Harris-Benedict), V(D)/V(T) (Siddiki estimate), and V(D)/V(T) (Penn State estimate) longitudinally. A time-dependent Cox model was used to handle these time-varying estimates. RESULTS: A total of 392 patients (median age 66 [interquartile range: 55-77] years, median SOFA score 9 [interquartile range: 7-12]) were finally included in our analysis, among whom 132 (33.7%) patients died within 28 days of admission. VR (hazard ratio [HR]=1.04 per 0.1 increase, 95% confidence interval [CI]: 1.01 to 1.06; P=0.013), V˙Ecorr (HR=1.08 per 1 increase, 95% CI: 1.04 to 1.12; P < 0.001), V(D)/V(T) (Harris-Benedict) (HR=1.25 per 0.1 increase, 95% CI: 1.06 to 1.47; P=0.006), and V(D)/V(T) (Penn State estimate) (HR=1.22 per 0.1 increase, 95% CI: 1.04 to 1.44; P=0.017) remained significant after adjustment, while V(D)/V(T) (Siddiki estimate) (HR=1.10 per 0.1 increase, 95% CI: 1.00 to 1.20; P=0.058) did not. Given a large number of negative values, V(D)/V(T) (Siddiki estimate) and V(D)/V(T) (Penn State estimate) were not recommended as reliable substitutes. Long-term exposure to VR >1.3, V˙Ecorr >7.53, and V(D)/V(T) (Harris-Benedict) >0.59 was independently associated with an increased risk of mortality in ARDS patients. These findings were validated in the fluid and catheter treatment trial (FACTT) database. CONCLUSIONS: In cases where V(D)/V(T) cannot be measured directly, early time-varying estimates of V(D)/V(T) such as VR, V˙Ecorr, and V(D)/V(T) (Harris-Benedict) can be considered for predicting mortality in ARDS patients, offering a rapid bedside application.