Diaphragm thickening fraction predicts noninvasive ventilation outcome: a preliminary physiological study

Over the last 2 decades, it has been suggested that noninvasive ventilation (NIV) might be effective in avoiding intubation [1, 2] and improving survival in the acute care setting [3] when compared to conventional oxygen therapy [4].

Nonetheless, its use in de-novo Acute Respiratory Failure (ARF) is debated, since a strict association between the unsuccessful NIV and the poor outcome has been suggested [5]. Given the risks associated with either premature or delayed NIV discontinuation [6], evaluating weaning readiness, as well as correct timing of intubation, is a critical challenge in patients with de-novo ARF.

Up to now, there are few available bedside measurements for prediction of NIV outcome and the decision to discontinue NIV is mainly based on clinical and physiologic parameters [7, 8].

Indeed, although sophisticated methods have been proposed to predict weaning outcome [8], none of these methods has ever achieved a wide bedside use because of their invasiveness, of the inconsistent results, and of the need for trainee personnel and complicated equipment. Consequently, the rapid shallow breathing index is still preferable to these methods because of its simpler application and more immediate interpretation [9].

The diaphragm is the main respiratory muscle, and its dysfunction has been associated with prolonged mechanical ventilation and weaning failure [10, 11].

Recently, ultrasound has been used at the bedside in acutely ill patients both for rapid functional assessment of the skeletal muscles [12] and for evaluation of diaphragmatic function [13] with minimal invasiveness and without any X-ray exposure. Ultrasound can be used to measure the excursions of the diaphragm, its thickness and its speed of contraction [14, 15], yielding information about the muscle function and the respiratory efficiency.

Indeed, diaphragmatic thickness (DT) correlates with the strength and the shortening of the muscle [16, 17]. The volume of muscle mass remains constant during diaphragm contraction. Consequently, as the muscle shortens it becomes thicker, so that DT changes are inversely related to changes in the length of the muscle. Moreover, the magnitude of diaphragm shortening and contraction may predict successful extubation in invasively ventilated patients [18].

Recently, Vivier et al. [15] have conducted a physiological study to assess the accuracy of Diaphragm Thickening Fraction (DTF) and its contribution to the respiratory workload in 12 critically ill patients requiring planned NIV after extubation. The patients were studied either while spontaneously breathing or during NIV at three different levels of pressure support, measuring DTF and diaphragmatic Pressure–Time Product per breath (PTP_di).

These authors found that increases in the level of Pressure Support were associated with a reduction in both PTP_di and DTF, and that there was a significant correlation between PTP_di and DTF (ρ = 0.74, p < 0.001). They concluded that DTF is a noninvasive method that may be useful in evaluating the diaphragm contribution to the respiratory workload in acutely ill patients undergoing NIV treatment.

The aim of this preliminary study is to assess whether measurements of DTF and its relationship with respiratory rate may predict NIV outcome in patients with de-novo ARF admitted to the emergency department.

All patients with de-novo ARF requiring NIV treatment admitted to the emergency department at Fondazione Policlinico Universitario A. Gemelli, IRCCS Rome-Italy were included into this preliminary study.

The protocol was approved by the local Ethics Committee (Prot 20813/16 ID 1200) and informed consent was obtained by all study participants.

The criteria for eligibility were de-novo ARF in the presence of respiratory rate ≥ 35 breaths per minute, a ratio of the PaO₂ to the fraction of inspired oxygen (PaO₂/FiO₂) of less than 200 with oxygen therapy through a Venturi mask or a High Flow Nasal Cannula (50 L/min) and active contraction of the accessory muscles of respiration or paradoxical abdominal motion.

The exclusion criteria were: age < 18 years, pregnancy, diaphragm paralysis, exacerbation of asthma and/or chronic obstructive pulmonary disease, neuromuscular disorders, severe obesity with Body Mass Index (BMI) ≥ 35 kg/m², palliative NIV in patients with malignancy, ineffective cough and/or inability to protect airways.

Noninvasive Pressure Support Ventilation (PSV) with positive end-expiratory pressure (PEEP) was delivered through either a facial mask with an inflatable soft cushion seal (Gibeck, Upplands, Sweden; Vitalsigns, Towota, NJ, USA) or with a clear, latex-free helmet (CaStar, Starmed, Mirandola, Italy), according to a clinical decision.

In the patients with a facial mask, PSV was started at 10 cmH₂O and increased with progressive stepwise increments of 2–3 cmH₂O, to obtain an exhaled tidal volume of 6 mL/kg, respiratory rate ≤ 25 breaths per minute, patient comfort and disappearance of accessory muscle activity or paradoxical abdominal motion. PEEP was increased with stepwise increments of 2–3 cmH₂O up to 12 cmH₂O to ensure peripheral oxygen saturation (SpO₂) of ≥ 90% with the lowest possible FiO₂.

In the patients with a helmet, a soft cushion around the neck provides sealing of the interface, reducing air leaks and allowing high levels of PEEP. PEEP was set to 10–12 cmH₂O to ensure adequate inflation of the interface, and since part of the volume delivered to the system was used to distend the helmet without reaching the patient PS levels were increased with stepwise increments of 2–3 cmH₂O up to 15–18 cmH₂O. Ventilator settings were then adjusted according to SpO₂ and measurements of arterial blood gases. The flow trigger was set at 3 L/min, checking out for the absence of auto-triggering phenomena.

Over a period of 36 months, 53 eligible patients were screened for inclusion. Of these, 32 patients were excluded and 21 were enrolled. Three patients were excluded after inclusion for a concomitant diagnosis of chronic obstructive pulmonary disease unknown at the time of enrolment, so that only18 patients (8 males, 44.5%) were included into the final analysis (Fig. 1).

Mean age (SD) was 66 (19) years, with a mean (SD) SAPS II score of 44 (4). On admission to the emergency department the median [IQR] PaO₂/FiO₂ ratio was 89.4 [77.4–117.5], the median PaCO₂ was 30.6 [29.3–39.2] mmHg, and the median respiratory rate was 38 [36–44.2] breaths per minute. The cause of acute hypoxic respiratory failure on admission was pneumonia in more than 70% of included patients.

Demographic and clinical characteristics of the population are reported in Table 1.

To our knowledge, this is the first study that evaluates DTF as a predictive index of NIV outcome in patients with de-novo ARF admitted to the emergency department.

Our study shows that DTF with a cut-off lower than 36.3% could predict NIV failure in hypoxemic patients, confirming previous results in other population [18, 26, 27]. In a prospective observational study including 63 patients with ARF of various origin, Di Nino et al. [18] found that a cut-off value of DTF ≥ 30% predicted extubation success regardless of the used weaning technique.

In a cohort of chronic patients ventilated in PSV through a tracheostomy tube, Ferrari et al. [26] reported that a cut-off value of DTF ≥ 36% was associated with a successful weaning after spontaneous breathing trial.

In a recent systematic review, Zambon et al. [27] evaluated 20 studies including a total of 875 critically ill patients where ultrasound was performed to detect diaphragmatic dysfunction. The authors used both diaphragmatic excursion and DTF to predict extubation success or failure during weaning and reported that the optimal cut-offs ranged from 10 to 14 mm and 30–36%, respectively.

Our study is the first to focus on hypoxemic patients during NIV treatment. However, the cut-off value of DTF that identified patients at major risk of NIV failure was similar to the cut-off value reported for critically ill patients requiring invasive mechanical ventilation [27], thus suggesting that DTF assessment is reliable to detect diaphragm dysfunction in hypoxemic patients on NIV treatment, as well.

In our study, we also assessed the reproducibility of DTF measurements. The ICC represents the proportion of total variance due to the variation between the subjects [27]. An ICC equal to 1 shows that the total variance is only due to the variation between the subjects, while an ICC equal to 0 indicates that the total variance is attributed to the variation between observers. The two operators performing ultrasound assessments in our study were appropriately trained [24] and all measurements were obtained in blind. We found overall good repeatability of DTF measurements, with ICC above the 0.75 cut-off that is usually considered as an index of good agreement among operators [28, 29]. Given the expertise of our evaluators and the overall good repeatability of DTF assessment, we don’t expect that further skilled operators could affect inter-rater agreement, as reported by other authors [15].

Considering that previous studies established a correlation between DTF and respiratory workload [15, 30‐32], in a preliminary analysis we also explored the possibility that the respiratory rate/DTF ratio could be a valid predictive index of NIV failure.

In our study, the respiratory rate/DTF ratio intended to represent an ultrasound surrogate of rapid shallow breathing index [9], one of the most used and studied weaning predictors in the clinical practice [33]. We did not report the rapid shallow breathing index since it could be not accurate in patients who were spontaneously breathing. We therefore believe that another physiologic study should be performed to compare this index with the values measured by ultrasound.

We found that the cut-off value of respiratory rate/DTF ratio > 0.6 was a reliable predictor of NIV outcome in patients with de-novo ARF admitted to the emergency department.

Recently, several authors have reported the beneficial and detrimental effects of spontaneous breathing in acutely ill patients [34, 35]. The entity of inspiratory effort that is surrogated in the clinical practice by the negative swing in the esophageal pressure (P_es) has been indicated as a possible aggravating mechanism of lung injury in patients with hypoxemic ARF [34, 35]. On the other hand, the active movements of the diaphragm may permit to recruit the dependent lung zones and maintain the end-expiratory lung volume, while the cranial shifts of the diaphragm due to the use of excessive sedation and/or neuromuscular blockade drugs may cause a significant decrease in end-expiratory lung volume [36]. Evidence for beneficial effects of spontaneous breathing has been provided only for patients with normal lungs and less severe forms of ARDS who have mild ventilatory requests [37]. In these patients, spontaneous breathing and NIV may improve gas exchange by the recruiting effect of PEEP, the improvement of hemodynamics and the avoidance of diaphragm atrophy [37].

In a recent observational study conducted on 30 patients with hypoxemic ARF who were candidate for a 24-h NIV trial, Tonelli et al. [38] found that a change in ΔP_es less than 10 cmH₂0 within the first 2 h of NIV treatment was an accurate predictor of NIV outcome at 24 h, when compared to other variables. The authors also reported that in patients with moderate to severe hypoxemic ARF, the reduction in the inspiratory effort clinically translate into a significant improvement of oxygenation and a decrease of both respiratory rate and tidal volume.

Other authors [39] have also reported that the progressive development of diaphragm atrophy as well as the rapid early increases in diaphragm thickening during ventilation are associated with prolonged mechanical ventilation and an increased risk of complications.

Our study has demonstrated that a value of respiratory rate/DTF ratio > 0.6 is accurate to predict the NIV outcome in hypoxemic patients. However, we are far from understanding the mechanisms leading to worsening respiratory failure and intubation (excessive respiratory workload/inadequate unloading of the respiratory muscles versus diaphragm atrophy) since the inspiratory effort was not measured in our study population.

In hypoxemic ARF patients, NIV treatment improves respiratory discomfort [40] and reduces the need of intubation, the rate of infections and the intensive care unit mortality [41]. However, high rates of NIV failure ranging between 30 and 50% [42] are related to an increased mortality [43], most likely due to the prolonged exposure of injured lungs to high swings in P_es and increased tidal volumes.

In our study, the percentage of patients who failed NIV treatment and required intubation in patients with de-novo ARF was similar to the percentage reported by other authors [42], thus confirming that unsuccessful NIV is related to poor outcomes [5, 43].

Both DTF and respiratory rate/DTF ratio may represent valid, feasible and noninvasive tools to predict the NIV outcome in patients with de-novo ARF.

Ultrasound monitoring of diaphragmatic function should be encouraged as an integral part of clinical practice when defining the correct timing of NIV discontinuation and/or the need of intubation.

However, future studies should be conducted on a large group of patients to address the ability of DTF and respiratory rate/DTF ratio to predict the NIV outcome.