Ultrasound-assessed diaphragmatic impairment is a predictor of outcomes in patients with acute exacerbation of chronic obstructive pulmonary disease undergoing noninvasive ventilation

Patients admitted to the intensive care unit (ICU) or respiratory intensive care unit (RICU) because of severe episodes of acute exacerbations of chronic obstructive pulmonary disease (AECOPD) have considerably high in-hospital (24%) and 1-year (59%) mortality rates [1‐3]. The use of noninvasive ventilation (NIV) in patients experiencing respiratory failure due to AECOPD is considered a first-line treatment but still has a failure rate between 5% and 40% [4]. In particular, risk of death is much higher in those patients receiving invasive mechanical ventilation (MV) once an NIV strategy has proven to be ineffective [5].

Researchers in previous studies have observed that patients with chronic obstructive pulmonary disease (COPD) might have a higher rate of diaphragmatic dysfunction (DD) than to age- and sex-matched healthy control individuals [6]. During AECOPD, biological factors related to systemic inflammation, prolonged use of steroids, and lung mechanical abnormalities due to hyperinflation may act as synergic mechanisms leading to DD [7]. Dynamic hyperinflation during AECOPD worsens the level of end-expiratory lung volume and residual volume, thus shifting tidal volume (V_t) toward the right side of the pressure-volume curve [8]. As a result, higher intrathoracic pressures are needed to maintain an adequate V_t. Furthermore, the early collapse of terminal airways with air entrapment causes an intrinsic positive end-expiratory pressure (PEEPi) that behaves as an adjunctive load that the respiratory muscles must overcome before generating inspiratory flow [7].

In a recently published pilot study, we reported that DD as assessed by a noninvasive ultrasound (US) technique is present in almost one-fourth of patients with AECOPD and admitted to the RICU for severe hypercapnic respiratory failure [9]. Although it is recognized that lung hyperinflation might play a critical role in DD, the impact of this derangement on AECOPD course and treatment remains incompletely elucidated. Therefore, in the present study, we wanted to investigate the clinical outcomes of patients with AECOPD requiring NIV and presenting with DD detectable by US. NIV failure was the prespecified primary outcome, and secondary outcomes were RICU, in-hospital, and 90-day mortality rates; duration of MV; incidence of tracheostomy; and RICU and in-hospital lengths of stay. Moreover, in these patients, we aimed at correlating the DD as assessed by US with the transdiaphragmatic pressure capacity (Pdi) as measured at maximal inspiration using the sniff maneuver (Pdi sniff).

In the considered period, 75 of 175 patients with COPD admitted to our RICU for acute exacerbation requiring NIV treatment were enrolled and studied. According to the predefined criterion (best ΔTdi < 20%), 24 patients (32%) presented with DD at US (DD+), whereas 51 patients (68%) did not (DD−) (Fig. 1). The general features and clinical characteristics at baseline of the whole population and according the presence/absence of DD are presented in Table 1. All the patients enrolled were classified as GOLD (Global Initiative for Chronic Lung Disease) stage 4D when earlier in stable state [16]. The use of systemic steroids in the year preceding admission was the only variable that significantly differed between DD+ and DD− patients. The pressure-generating properties of the diaphragm (Pdi sniff), as recorded in a subset of ten patients, was highly reduced in DD+ patients as compared with DD− patients (19 mmHg [IQR 6–28] and 82 mmHg [IQR 77–87], respectively).

Table 2 shows the clinical outcomes in the study population. DD+ patients presented a higher risk for NIV failure than DD− patients (risk ratio 4.4; p < 0.001). Six DD+ patients who failed NIV then died, and twelve underwent endotracheal intubation. After intubation, two died while on MV, five underwent tracheostomy (two of whom died), and five were successfully weaned. Among DD− patients for whom NIV failed, four DD− patients for whom NIV failed then died, and four underwent endotracheal intubation (two died while on MV, and two underwent tracheostomy with one of them dying).

Among the secondary outcomes, DD+ correlated with higher short- and long-term mortality, longer stay in the RICU, prolonged MV, and higher tracheostomy rate (Table 2).

In ROC analysis, ΔTdi < 20% showed higher accuracy in predicting NIV failure than baseline pH < 7.25, and both changes in arterial blood pH and PaCO₂ did within 2 h after NIV was started (Fig. 2) (AUCs 0.84, 0.51, 0.56, and 0.54, respectively; p < 0.0001). In addition, ΔTdi showed a very good correlation with Pdi sniff (Pearson’s r = 0.81; p = 0.004) (Fig. 3a); moreover, ΔTdi < 20% demonstrated the same accuracy as Pdi sniff in identifying DD+ (sensitivity 100%; 95% CI 0.6–1; specificity 100% 95% CI 0.51–1; p = 0.0048) (Fig. 3b).

Figure 4 reports ΔTdi and Pdi sniff values in the subgroup of ten patients tested with esophageal and gastric balloons (Fig. 4a) and distribution of patients with ΔTdi < 20% or > 20% according to Pdi sniff (Fig. 4b). Among patients with ΔTdi < 20%, two patients had a Pdi sniff of 0, meaning complete diaphragm paralysis, whereas four presented Pdi sniff values between 24 and 33 cmH₂O. Regarding diaphragm paralysis, ΔTdi demonstrated high sensitivity but low specificity in detecting this condition (sensitivity 100%, 95% CI 0.18–1; specificity 50%, 95% CI 0.22–0.78). Kaplan-Meier curves showed a significant increase in NIV failure and 90-day mortality (Fig. 5a and b, respectively) among DD+ versus DD− patients.

Our study demonstrates a significant prevalence of DD in patients with severe AECOPD (32%). In a previous exploratory study on the same subject [12], we found that DD similarly assessed by US was reported in 22% of total patients; however, that sample had fewer patients and less severity than the present study (APACHE II score 20 vs 22, p = 0.041; and SAPS II score 41 vs 43, p = 0.036).

Demoule et al. reported that the prevalence of DD, as assessed by twitch tracheal pressure in response to bilateral anterior magnetic phrenic nerve stimulation, was 64% in a heterogeneous group of patients admitted to a general medical ICU [5]. However, the two populations seem difficult to compare because the population in our present study is limited to a specific subset of critical patients. Moreover, we used a more restricted definition of DD, the ΔTdi in the zone of apposition less than 20%, thus representing a likely condition of very severe impairment, even confirmed by the transdiaphragmatic Pdi assessment as obtained in ten subjects.

The physiopathological mechanism(s) underlying the onset of DD during AECOPD is multifactorial. However, the synergic effect of mechanical disadvantage, functional (exhaustion) and biological (inflammation) impairment, and pharmacological damage (steroid-induced) could only be postulated.

In patients with COPD with frequent exacerbations the maximum pressure produced by the diaphragm’s contraction is significantly than in individuals without COPD, in the same regard as the volitional (Pdi at total lung capacity and/or Pdi sniff) or nonvolitional (phrenic nerve stimulation) nature of the test applied [17‐19]. This difference has been explained by the diaphragmatic shortening and the mechanical derangement following onset of progressive lung hyperinflation [17]. Indeed, patients with severe COPD produce higher transdiaphragmatic pressures at equivalent lung volumes than healthy control individuals [20]. The reason for this imbalance has been found in increased airway resistance, reduced dynamic lung elastance, and raised PEEPi, which are related to the progressive air trapping [17, 20]. Thus, in a patient with AECOPD, the expiratory flow limitation and the increased respiratory rate with reduced expiratory time favor the onset of hyperinflation with subsequent respiratory muscle shortening and limited excursion. The reduction in muscle strength seems even more inadequate to cover the excessive mechanical load imposed by the increased respiratory rate. In these conditions, the diaphragm soon exhausts its functional reserve, and ultimately mechanical impairment occurs.

In the present study, exposure to steroids in the previous year was the only feature differentiating DD+ from DD− patients. To date, no studies have been conducted in humans to investigate the specific effect of steroid course on the respiratory muscles during AECOPD. Notwithstanding this, studies performed on the skeletal muscles clearly show an increase in proteolysis and a reduction in the expression of insulin-like growth factor (IGF)-1 following the use of steroids [21]. In animal models, there are studies showing damage to the diaphragmatic structure after administration of systemic steroids [22, 23], with proteolysis and downregulation of IGF-1 being the plausible mechanisms responsible for steroid-induced myopathy [24]. Therefore, it is possible that previous or current (ab)use of systemic steroids could have contributed to the development of respiratory muscle weakness, thus leading to DD once AECOPD occurred.

Even despite the lack of a preadmission assessment of diaphragmatic function, some aspects seem to indicate that dysfunction might be related to the current acute condition rather than to a chronic exposure in our study population. First, the use of a very low cutoff limit of ΔTdi (< 20%) to identify DD+ let us individuate patients with a condition close to muscle paralysis, whose presence would already have been evident outside the acute phase [25]. Second, studies conducted with patients with stable severe COPD did not identify a significant difference in diaphragmatic function by US evaluation at V_t compared with healthy individuals [26]. Third, in a study by Orozco-Levi et al., transdiaphragmatic pressure values were similar in patients with stable COPD and normal control subjects, thus indicating a substantial functional preservation, although regional stresses and strains (muscle geometry, increased workload during exertion, mechanical stress, and metabolic factors) might induce or accentuate diaphragmatic injury [27].

It is likely that these potential mechanisms of damage are amplified during AECOPD, when the diaphragm is called on to sustain stresses much greater than during clinical stability. Our study demonstrates a positive and significant correlation between ΔTdi at US assessment and Pdi measurement at maximal inspiration. To our knowledge, this is the first time that US evaluation of the diaphragmatic function has been compared with standard techniques in subjects with AECOPD.

Kim et al. showed that US is reliable in identifying DD during weaning from MV and in estimating the work of breathing during NIV [28]. Gottesman et al. indicated that ΔTdi < 20% is the cutoff limit for identifying diaphragm paralysis [25]. In our study, only two patients had Pdi sniff equal to 0 among those with ΔTdi < 20% (see Fig. 4). The sensitivity and specificity of ΔTdi (< 20%) to detect DD+ dropped to 50% in cases of diaphragm palsy.

Overall, the present study demonstrates that US is able to identify dysfunction but not palsy of the diaphragm, for which invasive maneuvers are needed. Given that these patients usually present with acute hypercapnic respiratory failure, the use of volitional or invasive techniques such as transdiaphragmatic pressure might become extremely challenging, in contrast to US assessment.

The main finding of our study is that DD+ patients, once confirmed by US, present a greater than fourfold risk of NIV failure during severe AECOPD (see Fig. 5). Interestingly, the two slopes start separating significantly 4–6 hours after starting NIV, then both groups gradually display a plateauing distribution over time.

Several studies have consistently demonstrated that the severity of hypercapnia and acidosis are associated with early NIV failure during AECOPD [13, 29‐31]; however, no studies have assessed the impact of DD on NIV outcome. Indeed, our previous exploratory study demonstrated only that DD+ is related to NIV failure in a small population underpowered to test this hypothesis [9].

In the present study, we were able to demonstrate that DD+ as assessed by US may predict NIV failure with higher accuracy than both baseline pH < 7.25 and early change of pH and PaCO₂ (see Fig. 2). Notwithstanding this, the reasons for the exceeding NIV failure rate are not completely clear, whereas severe expiratory flow limitation and excessive hyperinflation might have played a critical role in uncoupling patients’ efforts and MV. Because the use of NIV is increasing in different hospital settings, the present data may be of help to select patients with AECOPD who might be successfully ventilated in a general ward (patients without DD), rather than those who would benefit by being treated with MV in the ICU (patients with DD). Therefore, this finding increases the clinical importance of early identification of patients at greater risk for NIV failure even in the ward [32].

In our patients with AECOPD and DD+ short-term (in the RICU) and 90-day mortality are up to fivefold greater than in DD− patients. This finding is in line with data reported by Demoule et al., who described a higher mortality rate in patients admitted to the ICU with low tracheal pressure in response to bilateral anterior magnetic phrenic nerve stimulation [5]. In that hospital setting, a potential explanation for increasing death rate following DD has been identified in the early impairment of nerve conduction and mitochondrial alterations due to the onset of systemic inflammation [33].

Our present data suggest only that the duration of ventilation and the prolonged stay in the critical area following NIV failure may have resulted in a worse prognosis. Interestingly, we did not find any significant difference in hospital stay between patients with impaired or preserved diaphragmatic function. However, it could be hypothesized that other independent factors might have influenced the length of admission (e.g., home setting, chronic comorbidities, lack of domestic facilities).

In this single-center trial, we observed that patients with severe AECOPD admitted to the RICU for NIV treatment have a significantly higher risk of failure and mortality when DD, as noninvasively assessed by US, occurs. US measurement of DD definitely correlates with transdiaphragmatic pressure-generating capacity in these subjects. Therefore, early evaluation of the diaphragm by US in this setting may help clinicians to identify patients with AECOPD at major risk for a negative prognosis.