This prospective observational cohort study was carried out in a single 6-bed RICU at the University Hospital of Modena (Italy) over a 24-month period (January 2016–January 2018). Written informed consent to participate was obtained from all patients enrolled or their relatives. Approval was obtained from the local ethics committee of Modena (protocol 839/C.E. and 266/16 C.E.), and the trial was registered at clinicaltrials.gov (NCT03851822).

Eligible patients >18 years of age were those consecutively admitted to the RICU due to ARF (hypoxemia or hypoxemia with hypercapnia) requiring MV, and with a stay >8 day. Exclusion criteria were: 1) refractory shock, 2) patient goals of care not consistent with aggressive management, 3) end-stage chronic obstructive pulmonary disease (COPD) requiring home oxygen and/or ventilatory support, 4) interstitial lung disease, 5) neuromuscular disease, 6) chest wall deformities, 7) pregnancy, 8) chemotherapy or radiotherapy during the past 30 days, 9) brain injury on CT scan and/or Glasgow Coma Scale (GCS) score <8, 10) tracheostomy. All patients were treated according to the best current clinical practice by the attending staff, unaware of the study purpose. All patients were provided with mobilization and rotation on a daily basis.

Demographics, severity scores, namely the Kelly-Matthay Scale, the Acute Physiology and Chronic Health Evaluation II (APACHE II), the Simplified Acute Physiology Score (SAPS II), and the Subsequent Organ Failure Assessment (SOFA) were recorded at admission. Presence of ARDS, pneumonia or septic shock was assessed at baseline whereas pre-existing comorbidities were reported through the Charlson Comorbidity Index; concomitant COPD had to be confirmed by both smoking habit, past medical history, non-reversible obstruction at spirometry performed in the year before. Hospitalizations in the previous 6-month were also reported.

Arterial blood gases (PaO2, PaCO2), pH, and PaO2/FiO2 ratio were recorded at baseline; blood procalcitonin, glucose, albumin, creatinine and lactate were measured at baseline and after 1, 2 and 7 days following admission. Multidrug-resistant (MDR) microorganism colonization and infection, and the onset of a second infection during RICU stay were recorded.

A subset of patients under MV as divided into two groups according to the ventilatory mode (assisted spontaneous breathing — ASB, and controlled ventilation-CMV) was subjected to measurement of lung mechanics.

Patients in the ASB group underwent esophageal pressure (Pes) and transpulmonary pressure (PL) monitoring performed through a nasogastric tube (NutriVent nasogastric polyfunctional catheter; SIDAM, Mirandola, Italy) with a pressure transducer (OptiVent monitor, SIDAM) and according to a standard procedure.15,16 PL was calculated as airway pressure (Paw) – Pes. In order to avoid using absolute values for Pes and PL, we always referred to ΔPes and ΔPL from the end-expiratory level, respectively. The waveforms of Paw, Pes and airflow were continuously recorded using a data acquisition system (PowerLab; AD Instruments, Colorado Springs, CO, USA) at a sampling frequency of 100 Hz for offline data analysis.17

Patients in the CMV group, underwent ventilation set in control mode with a Vt = 6 ml/kg and positive end-expiratory pressure (PEEP) adjusted on the basis of the incremental FiO2/PEEP combinations.18 Measurements of static respiratory mechanics were performed after 30 min of constant flow during MV. The values of pressure were obtained during baseline ventilation with an airway occlusion at the end of inspiration pressing the end-inspiratory hold, until reaching a plateau pressure, and subsequently performing a similar procedure with an end-expiratory occlusion. During the procedure, occlusions that did not produce a clear plateau were discarded. Airway driving pressure (ΔPaw) was defined as the end-inspiratory plateau pressure (Pplat) — positive end-expiratory airway pressure (PEEPtot). All these measurements were performed within 24 h from MV start.

Ultrasound (US) of the diaphragm was assessed on both sides at admission during spontaneous breathing. Motility of the diaphragm was assessed at the bedside in the semi-recumbent position by a B-mode US device (GE Vivid 7; GE Healthcare Life Sciences, Helsinki, Finland) connected to a 7–12 MHz linear probe as previously reported.19 Thickening fraction (TF) during spontaneous breathing19 was calculated as: ΔTdi = (end-inspiration Tdi – end-expiration Tdi/ end-expiration Tdi) × 100. The best value of three measurements was taken as representative of the diaphragm function and then recorded for analysis.

Diaphragm dysfunction (DD) was defined according to the presence of TF bilaterally lower than 20%.19

Survivors with both hospital stay >8 days and with tracheostomy due to a need of MV > 21 consecutive days for at least 6 h/day were defined as having chronic CI.1 The prevalence of chronic CI was the primary aim of this study.

According to the Berlin definition, patients were considered as having ARDS if they had: [1] acute respiratory failure not fully explained by cardiac failure or fluid overload; [2] bilateral opacities consistent with pulmonary edema on the chest radiograph or the computed tomography scan; and [3] onset within 1 week after a known clinical insult or new/worsening respiratory symptoms.20 Septic shock was present according to the onset of sepsis with requirement of vasopressor to maintain a mean arterial pressure ≥65 mmHg and serum lactate level >2 mmol/L (>18 mg/dL) in the absence of hypovolemia.21 Multi-Drug-Resistant (MDR) infection was defined as the presence of sepsis with blood isolation of a MDR bacterial agent. The onset of a second infection was defined by the occurrence of clinical, radiological and microbiological signs of infection during RICU stay (after at least 2 days of stability when recovering from the first episode). The association between chronic CI and demographics, inflammatory and metabolic blood parameters (C-reactive protein-CRP, sepsis and septic shock, infection and MDR bacterial infection), and respiratory mechanics (transpulmonary and driving pressure) on admission and/or their trends between days 1 and 7, were considered as secondary outcomes of the study.

The statistical package GraphPad Prism 7.0 (GraphPad Software, Inc., La Jolla, CA, USA) was used for statistical analysis. Initially, we performed a power test (α = 0.05, power 80%) considering a prevalence of up to 20% of patients undergoing MV for >48 h and/or tracheostomy in the critical care area,7 and an average annual mortality rate of 24% among patients admitted to RICU and requiring MV. A sample size of 127 patients was required to confidently perform analysis on the pre-specified primary scope of our study.

The considered variables were investigated through univariate analysis and contingency table analysis for relative risks to detect the possible risk factors associated with chronic CI once in RICU.

Significant variables then entered a multivariate regression analysis, with a backward stepwise method to exclude non-significant variables from the model. Independent variables able to predict chronic CI at the multivariate test were then entered intoa ROC analysis.

A p-value lower than 0.05 was considered to be statistically significant.

Given the lack of a univocal definition of chronic CI, the exact prevalence of this condition remains unknown. As a whole in patients admitted to ICUs, the prevalence of chronic CI defined as the need to prolong MV is around 10%.1 Moreover, chronic CI represents a predominant clinical feature in up to 50% of patients who survive an episode of sepsis.10 In our study performed in a medical specialized ICU in patients with de novo ARF, onset of chronic CI was lower than that reported in a surgical ICU.10 However, in the latter study10 the definition of chronic CI referred to the length of stay in ICU for ≥14 days with persistent organ dysfunction as assessed by SOFA score. In our study, chronic CI was defined as the need for respiratory intensive care for ≥8 days associated with the decision to perform tracheostomy and to prolong MV. Prolonged mechanical ventilation and tracheostomy are factors that increase per se the risk of bed-lying, thus requiring combined strategies to avoid complications and speed-up recovery in specialized intensive care settings.22

During stay in ICU, persistence of inflammatory state is the pathophysiological key which may favor chronic CI development.23 The term “persistent inflammation, immunosuppression and catabolism syndrome” (PICS) has been coined to describe a phenotype of chronic CI in surgical ICU patients who suffered from major pro-inflammatory event (blunt traumatic injury/sepsis).24,25 A self-perpetuating cycle of inflammation constitutes the pathobiological hypothesis underlying PICS development.26 Patients who develop chronic CI after sepsis exhibit persistent elevations of systemic inflammatory cytokines (IL-6 and IL-8) even at 28 days following onset of ARF,21 which is the case of the inflammatory response during ARDS.27 Persistence of elevated levels of inflammatory cytokines and chemokines in circulating fluids and bronchoalveolar lavage, could lead patients with ARF to chronic CI.28 In patients with ARDS, longitudinal sampling has proved that local and systemic inflammation may persist for several weeks, even after the resolution of the respiratory syndrome.29 Recently, both hyperinflammatory and hypoinflammatory subphenotypes were identified analyzing biomarkers plasma levels from two trials in patients with ARDS.30 The hyperinflammatory subphenotype is characterized by fewer days free from ventilatory and/or organ failure, conditions which constitute a risk factor for chronic CI. In our study, the measurement during the first week after admission of CRP as a marker of inflammation was able to distinguish the group of patients who would develop chronic CI from those who did not and would recover (Fig. 2).

To study the underlying mechanisms that drive chronic CI progression over the course of sepsis, different animal models and clinical studies in survivors have been conducted.30–32 Data show that immunological dysfunction and repeated infections have a role in maintaining inflammation and, finally, trigger chronic CI.33,34 In line with these findings, our study confirms that patients with ARF presenting a second infection during stay in RICU or having a MDR bacterial infection are more likely to develop chronic CI. Overall, this suggests that immunological dysregulation and susceptibility to infection constitute a mechanism favoring the development of chronic CI in these individuals. Interestingly, septic shock, a condition characterized by profound circulatory, cellular and metabolic alterations, was the strongest predictor of chronic CI development in our study (Fig. 3).

In patients suffering from acute critical illness of various etiologies, respiratory muscle dysfunction is quite frequent and may develop early. Demoule et al. have shown that 64% of patients admitted to ICU have DD at the acute onset, sepsis being the major independent risk factor.35 The mechanisms underlying sepsis-induced DD (SIDD) are complex.36 At the molecular level, there is evidence that cytokines (TNF-alpha, IL-1alpha, IL-1beta and IL-6) and inflammatory signaling (NF-kappaB pathways) are involved,37 and may promote mitochondrial dysfunction, reactive oxygen species (ROS) production, and breakdown of sarcomere proteins during infection.38 In the present study, we have reported that DD (i.e. TF < 20%) by noninvasive assessment early at the time of admission is a risk factor in the specific population of difficult-to-wean patients. To the best of our knowledge, this result is new and opens further research perspectives.

There is evidence in literature that indicates that MV plays a role in determining ventilator-induced lung injury.39 In particular, non-physiological stress/strain could act as a trigger and amplifier for local and systemic inflammation.40 In our study, we measured ΔPaw as a surrogate marker for cyclic lung strain in patients receiving ventilator support41 and ΔPL as a marker of lung stress in patients with ASB16; however, we failed to find any correlation with chronic CI development. Therefore, even though lung mechanics have been used with a small group of patients, data suggest that mechanical stress/strain on lung parenchyma might play only a minor role in the physiopathology behind the onset of chronic CI.

Despite intriguing findings our observational study presents some major problems which limit generalizability. First, the evidence of persistent inflammation as a potential risk for chronic CI development lacks measurement of specific inflammatory cytokines. Second, immunological assessment has not been investigated in our population. Third, the definition of chronic CI as applied here excludes those patients with persistent signs of organ dysfunction (i.e. renal insufficiency), or without any need for tracheostomy and prolonged MV.