Abstract Purpose It has been suggested that, by recruiting lung regions and enlarging the distribution volume of the cold indicator, increasing the positive end-expiratory pressure (PEEP) may lead to an artefactual overestimation of extravascular lung water (EVLW) by transpulmonary thermodilution (TPTD). Methods In 60 ARDS patients, we measured EVLW (PiCCO2 device) at a PEEP level set to reach a plateau pressure of 30 cmH2O (HighPEEPstart) and 15 and 45 min after decreasing PEEP to 5 cmH2O (LowPEEP15′ and LowPEEP45′, respectively). Then, we increased PEEP back to the baseline level (HighPEEPend). Between HighPEEPstart and LowPEEP15′, we estimated the degree of lung derecruitment either by measuring changes in the compliance of the respiratory system (Crs) in the whole population, or by measuring the lung derecruited volume in 30 patients. We defined patients with a large derecruitment from the other ones as patients in whom the Crs changes and the measured derecruited volume were larger than the median of these variables observed in the whole population. Results Reducing PEEP from HighPEEPstart (14 ± 2 cmH2O) to LowPEEP15′ significantly decreased EVLW from 20 ± 4 to 18 ± 4 mL/kg, central venous pressure (CVP) from 15 ± 4 to 12 ± 4 mmHg, the arterial oxygen tension over inspired oxygen fraction (PaO2/FiO2) ratio from 184 ± 76 to 150 ± 69 mmHg and lung volume by 144 [68–420] mL. The EVLW decrease was similar in “large derecruiters” and the other patients. When PEEP was re-increased to HighPEEPend, CVP, PaO2/FiO2 and EVLW significantly re-increased. At linear mixed effect model, EVLW changes were significantly determined only by changes in PEEP and CVP (p