Driving pressure and long-term outcomes in moderate/severe acute respiratory distress syndrome

We conducted a prospective longitudinal cohort study of 22 survivors of moderate/severe ARDS, recruited from six different ICUs located in Hospital das Clínicas, São Paulo, Brazil, from November 2008 to January 2012.

Patients were enrolled in this study in conjunction with a clinical trial in mechanical ventilation, the “ARDSnet Protocol versus Open Lung Approach in ARDS” trial (NCT 00431158) [7]. The institutional review committee approved the study that included the follow-up, and informed consent was obtained from each patient or legal representative. Briefly, this was a randomized controlled trial in which patients were ventilated with a ARDSnet protocol, which uses low tidal volumes, relatively high respiratory rates, with oxygenation managed according to PEEP and FIO2 relationships as defined in a table, or with an open lung approach strategy, which uses a technique to recruit collapsed lung areas and then uses the lowest PEEP level that prevents recollapse of recruited lung units, being the best PEEP level determined by a decremental PEEP trial involving a series of pressure measurements taken after the recruitment maneuver. Both the ARDSnet protocol and the open lung approach require low tidal volumes and plateau pressures. In conclusion, open lung approach improved oxygenation and driving pressure, without detrimental effects on mortality, ventilator-free days or barotrauma.

The inclusion criteria for the study were as follows: Patients intubated and mechanically ventilated, with diagnosis of ARDS using American–European Consensus Criteria and enrollment in study < 48 h since diagnosis of ARDS. For 12–36 h (ideally 12–24 h), after diagnosis of ARDS, patient must be ventilated as follows: volume A/C, tidal volume of 4–8 mL/kg PBW, plateau pressure ≤ 30 cmH₂O, PEEP/FIO₂ adjustments using ARDSnet table, and ventilator rate to keep PaCO₂ = 35–60 mmHg. During the 12–36-h (ideally 12–24-h) period, PaO₂/FIO₂ must remain < 200 mm Hg for an ABG obtained 30 min after placement on the following specific ventilator settings: volume A/C, tidal volume = 6 mL/kg PBW, plateau pressure ≤ 30 cmH₂O, inspiratory time ≤ 1 s, PEEP ≥ 10 cmH₂O, FIO₂ ≥ 0.5, ventilator rate to keep PaCO₂ = 35–60 mmHg. No lung recruitment maneuvers or adjunct therapy. Total time on mechanical ventilation < 96 h at time of randomization.

Patients were excluded if they presented one of the following criteria: age < 18 years or > 80 years, weight < 35 kg PBW, body mass index > 60, intubated 2° to acute exacerbation of a chronic pulmonary disease, acute brain injury (ICP > 18 mmHg), immunosuppression 2° to chemo- or radiation therapy, severe cardiac disease (one of the following): New York Heart Association Class 3 or 4, acute coronary syndrome or persistent ventricular tachyarrhythmias, positive laboratory pregnancy test, sickle cell disease, neuromuscular disease, high risk of mortality within 3 months from cause other than ARDS, e.g., cancer, more than 2 organ failures (not including pulmonary system), documented lung barotrauma, i.e., chest tube placement other than for fluid drainage, persistent hemodynamic instability or intractable shock, penetrating chest trauma, enrollment in another interventional study. Randomization in the pivotal study was stratified by center, age and APACHE II scores.

Baseline data collected at enrollment included age, sex, height, severity of illness measured by the Acute Physiology and Chronic Health Evaluation (APACHE) II score, ratio between arterial oxygen tension and fraction of inspired oxygen (PaO₂/F_IO₂) with a positive end-expiratory pressure (PEEP) of at least 10 cmH₂O and FIO₂ > 0.5, and static compliance (respiratory system) 30 min before protocol enrollment. Twenty-four hours after patient enrollment, we collected respiratory variables, including tidal volume (mL/kg of predicted body weight), PEEP and plateau pressure. Airway driving pressure was defined as plateau pressure minus total PEEP. To measure the plateau pressure we used neuromuscular blocking agents and volume-controlled ventilation.

Blood samples were obtained on the day of randomization (day 0) and on days 1, 3, 7, and the day of extubation. Blood samples were centrifuged, and plasma was stored at 70 C. NT-PCP-III was assayed by a sandwich ELISA method according to the methodology specifications of the manufacturer (Elabscience, Texas, USA). The normal range of serum levels of nonsmoking individuals was determined to be 1.6–4.0 ng/L.

Pulmonary function testing (PFT) was performed at 1 and 6 months after the onset of ARDS using the MedGraphics Cardiorespiratory Diagnostic System (Medical Graphics Corporation, USA). All tests were done according to Brazilian guidelines [8]. Reference ranges were calculated based on statistics formulated from the Brazilian population [9‐11].

High-resolution computed tomography (HRCT) scan of the lungs was also performed 1 and 6 months after the onset of ARDS in supine position during inspiration, close to total lung capacity. All CT scans were segmented by applying the region growing algorithm to select the lung parenchyma using OsiriX (OsiriX 64-bits, Pixmeo Sarl, Geneva, Switzerland). After the segmentation, the original images (DICOM files) as well as each respective ROI were exported and analyzed with a purpose-built routine (QALI-DV software) written in MATLAB (MathWorks, USA). Manual correction was applied to the segmented images containing peripheral atelectasis. Total lung volume (TLV), total air volume (TAV) and total lung mass (TLM) were extracted from the segmented whole lung in 3D [11, 12]. The percentages of hyperaerated (− 1000 to − 900 HU), normally aerated (− 900 to − 500 HU), poorly aerated (− 500 to − 100 HU) and nonaerated (− 100 to + 100 HU) compartments of the lung parenchyma were calculated [12, 13]. To assess the occurrence and progression of emphysema in this longitudinal study we used the percentile point [14‐17] using a threshold of 15% (P15). The sensitivity of the percentile point method has been shown to be similar within a broad range of percentiles from the 10th to the 30th [14].

Six months after ARDS onset, we also performed a standardized 6MWT [18] and fulfilled the Medical Outcomes Study 36-item Short-Form General Health Survey (SF-36), which measures the HRQoL [19]. The SF-36 includes eight multiple-item scales that assess physical functioning, social functioning, physical role, emotional role, mental health, pain, vitality and general health. Scores for each aspect can range from 0 (worst) to 100 (best).

Categorical values are described as frequency and percentages, and continuous variables as the mean and SD, or median and interquartile range. The Fisher exact test was used to compare independent categorical variables. Continuous variables were compared with the Student t test or the Mann–Whitney test for dependent or independent data. The Friedman test was used for one-way repeated-measures analysis. The strength of the association between two variables was measured using correlation coefficient (r). We used Pearson correlation to parametric variable and Spearman correlation to nonparametric variable. We used linear regression to get r² in parametric variable. In order to determine clinical variables independently associated with lung function, we performed a multivariable linear regression. For NT-PCP-III regression we used log10 transformations as is commonly performed for biomarkers with a right-tailed distribution. Statistical significance was set at a two-tailed p value of ≤ 0.05, and analyses were performed with R, version 3.0.2 (http://​www.​r-project.​org).

Over the 58-month (2007–2012) recruitment period, we enrolled 33 patients. The patients in cohort were predominantly male. Mortality at 28th day was 33% (Fig. 1).

Of those enrolled, we lost the follow-up of 5 patients at month-1 and 2 between month-1 and month-6. Reasons for exclusion are outlined in Fig. 1. There was no difference in terms of severity score, age and static compliance measurement between patients followed up at month-6 and patients that were lost.

The descriptive baseline and hospital data for all randomized patients and for survivors are described in Table 1. Monitored variables during mechanical ventilation at baseline and 24 h after randomization are also shown in Table 1. Driving pressure at 24 h was related to driving pressure at 48 h (r = 0.58, P = 0.006) and 72 h (r = 0.56, P = 0.009) after randomization.

Pulmonary function tests showed a mildly reduced FVC and a moderately reduced DL_CO after 6 months (Table 2). The FVC was below normal (< 80% of predicted) in eleven (65%) patients at month-1, and in five (33%) patients at month-6. Twelve patients (70%) showed a reduced DL_CO at month-1, whereas four (29%) patients showed a reduced DL_CO after 6 months of follow-up (Table 2).

Driving pressure was the only ventilation variable significantly correlated with FVC at 1- (r = 0.65) and 6-month (r = 0.67) follow-up (Fig. 2). Driving pressure (r = 0.51) and APACHE II (r = 0.59) were correlated with DL_CO at month-1. Of note, tidal volume and respiratory system compliance were very weakly correlated with pulmonary function tests.

After testing for the confounding effects of tidal volumes, plateau pressures and baseline static respiratory compliance (R² = 0.51, F(4,10) = 4.66, p = 0.022), the association between driving pressures and FVC at month-6 remained the only statistically significant one (β = − 4.62, IC95% (− 7.11 to − 2.13), P = 0.002).

When comparing the effects of the randomized treatments (OLA vs. ARDSnet) on FVC at month-6, there was a marginal difference between arms, with FVC of 4.54 ± 0.93 L versus 3.41 ± 1.04 L (OLA vs. ARDSnet, respectively, P = 0.06), representing 96 ± 20% versus 86 ± 15% of predicted, respectively (P = 0.32).

Twenty-one patients were performed a HRCT at month-1 of follow-up. The median and interquartile range of the total lung volume (TLV) was 3.54 L (2.29–4.37); the total lung weight (TLW) was 1194 g (861–1873), and the mean pulmonary density (MPD) was 451 g/L (243–610). The P₁₅ (percentile 15% for lowest lung densities) was 72 g/L (42–117).

At month-6, 16 patients were performed a HRCT. The median and interquartile range of TLV increased (P = 0.01) to 4.99 L (3.73–6.20), the TLW decreased (P = 0.01) to 762 g (652–934 g), the MPD decreased (P < 0.001) to 164 g/L (133–183), and P₁₅ decreased (P = 0.02) to 49 g/L (23–65). Those longitudinal reductions in MPD, TLM and percent of nonaerated/poorly aerated lung volumes were all significant, as well as the increase in TLV.

Lung image parameters were significantly correlated with driving pressure and pulmonary function. After 6 months, individual MPD was significantly correlated with individual driving pressure (r = 0.53, P = 0.03), but not with respiratory system compliance (P = 0.95). Consistently, FVC at month-6 strongly correlated with CT parameters: MPD (r = − 0.86, P = 0.00005), % of nonaerated/poorly aerated volume (r = − 0.70, P = 0.003), TLV (r = 0.65, P = 0.008) and air/tissue volume ratio (r = 0.70, P = 0.003).

Dividing patients based on median into high (HDP) or low (LDP) driving pressure groups (≥ 13 and < 13 cmH₂O, respectively), MDP was higher in HDP group (P = 0.04; Fig. 3).

When comparing the effects of the randomized treatments (OLA vs. ARDSnet) on MPD at month-6, there was no difference, with MPD of 175 (131–176) g/L versus 171 (138–195) g/L (OLA vs. ARDSnet, respectively, P = 0.92). There was no difference in terms of TLV (P = 0.38) and TLW (P = 0.35).

Individual increments in log₁₀NT-PCP-III levels, from baseline till extubation, correlated with individual values of driving pressures at 24 h (β = 0.006, IC95% (0.001–0.011), r² = 0.35, F(1,13) = 8.55, p = 0.011; Fig. 4): the higher the driving pressure, the higher the increment in log₁₀NT-PCP-III difference.

Mean pulmonary density (MPD) at month-6 was related to log₁₀NT-PCP-III level during the first week of mechanical ventilation: at day 0 (β = 0.005, IC95% (0.001–0.010), r² = 0.35, F(1,12) = 8.06, p = 0.014), day 1 (β = 0.006, IC95% (0.002–0.010), r² = 0.34, F(1,14) = 8.84, p = 0.010), day 3 (β = 0.005, IC95% (0.0003–0.009), r² = 0.25, F(1,12) = 5.38, p = 0.038).

Dividing patients into high (HDP) or low (LDP) driving pressure groups (≥ 13 and < 13 cmH₂O, respectively), LDP group did not change levels over time (P = 0.15), while HDP group increased (P = 0.03; Fig. 5).

When comparing the effects of the randomized treatments (OLA vs. ARDSnet) on log₁₀NT-PCP-III, there was no statistic significant difference at day 0 (P = 0.89), day 1 (P = 0.55), day 3 (P = 0.19) and on extubation (P = 0.65). There was a difference in terms of log₁₀NT-PCP-III at day 7 (median log₁₀NT-PCP-III 1.69 in OLA arm vs. 1.99 in ARDSnet, P = 0.03). There was no difference in the change between day 0 to day 1 (p = 0.69), to day 3 (p = 0.23), to day 7 (p = 0.31) and to extubation (p = 0.18).

A standardized 6MWT was performed in 11 patients, but we could not observe any relationship between the walked distance and the ventilation variables selected. In terms of HRQoL, 10 patients completed SF-36 survey and we found a correlation between driving pressure (24 h after randomization) and the general health domain (r = − 0.69, p = 0.02).