Introduction

Noninvasive positive pressure ventilation (NPPV) is increasingly used to manage hypoxemic acute respiratory failure (HARF), avoiding endotracheal intubation and additional risks of related complications [1]. Although this is particularly true in immunocompromised patients [2], a multivariate analysis of five randomized studies on immunocompetent patients with HARF of varied etiologies showed NPPV to be independently associated with a lower risk of intubation and a lower 90-day mortality rate [3].

Patient comfort is a key point for administering this ventilation technique without frequent interruptions. Although a lot of attention has been dedicated to the development of new interfaces with increased tolerance, mask intolerance and discomfort still represent a major cause for NPPV failure. In a previous study in immunosuppressed patients with HARF we reported an NPPV failure rate related to interface intolerance of 13% associated with an 80% mortality rate [4].

Of note, patient agitation has been considered a relative contraindication for the use of NPPV [5]. Among the large variety of sedatives and analgesics commonly available, opioids, especially morphine and fentanyl, are widely used in the critically ill patient because of their efficacy in pain control and mitigation of psychological discomfort. The introduction into clinical use of new synthetic opioids with limited adverse effects, particularly on the respiratory system, has offered an option for the analgesia sedation of critically ill patients. In a recent study Conti et al. [6] showed that the continuous infusion of sufentanil may be used as a single sedative agent in patients receiving assisted ventilation, allowing mitigation of patient discomfort and obtaining the desired level of awake sedation, with no significant effects on respiratory drive, minute volume, respiratory frequency, respiratory pattern, blood gases, or hemodynamics. Furthermore, Cavaliere et al. [7] have demonstrated similar results with a low-dose continuous infusion of remifentanil. Analgesia-based sedation has recently been proposed to manage NPPV failure in acute respiratory failure patients in a preliminary study [8].

The purpose of this prospective, uncontrolled study was to assess the feasibility of remifentanil-based sedation in acute respiratory failure patients refusing to continue NPPV for intolerance to two different interfaces—helmet and total face mask.

Patients and methods

From September 2007 to December 2009, 204 consecutive HARF patients were treated with NPPV in our 14-bed general ICU; 151 were ventilated with the helmet and 53 with the total face mask. Thirty-eight patients (18.6%) with persistent HARF (defined as PaO2/FiO2 lower than 200 after a trial of NPPV as first-line intervention with the aim of avoiding endotracheal intubation) who, despite optimal NPPV setting, complained of discomfort and asked for an interruption of the NPPV session, were proposed for enrollment in the present study. Two refused and 36 (18%) accepted and were enrolled.

Patients who entered in the protocol started sedation with remifentanil (0.025 μg kg−1 min−1) infused through a dedicated line of the central catheter. Our goal was to obtain patient comfort, achieving a sedation score between 2 and 3 on the Ramsay scale [10] by increasing the infusion rate by 0.010 μg kg−1 min−1 every minute to a maximum of 0.12 μg kg−1 min−1.

Patients with severe hemodynamic instability such as hypotension (mean arterial pressure lower than 60 mmHg despite fluid challenge and vasoactive drugs) or rhythm disorders, severely decreased consciousness (GCS below 12), chronic obstructive pulmonary disease, age less than 18 years, and pregnant women were excluded from the study.

The simplified acute physiology score (SAPS II) was calculated 24 h after admission to the ICU [9]. All the enrolled patients completed the study.

Noninvasive ventilation

All patient were noninvasively ventilated with a latex-free helmet (CaStar, Starmed, Mirandola, Italy) or a total face mask (Respironics, Monroeville, PA, USA) connected to Drager ventilator (Drager, Lubeck, Germany) in pressure support mode (PSV) with NPPV software. Pressure support ventilation was increased in increments of 2–3 cm H2O to obtain an exhaled tidal volume of 6 mL kg−1 and a respiratory rate (RR) lower than 25 breaths min−1. When the helmet was used, part of the volume delivered to the system was spent to distend the helmet and did not reach the patient, requiring higher PS levels [11]. Positive end-expiratory pressure (PEEP) was increased in increments of 2–3 cm H2O up to 12 cm H2O to ensure a peripheral oxygen saturation of 94% with the lowest FiO2 possible. Ventilator settings were then adjusted on the basis of pulse oximetry and serial measurements of arterial blood gases. Duration of ventilation was standardized according to the protocol of Wysocki et al. [12]. Briefly, during the first 24 h, NPPV was continuously maintained until oxygenation and clinical status improved. After this period and once PEEP requirements decreased to 5 cm H2O, each patient was evaluated daily while breathing supplemental oxygen without ventilatory support for 15 min. NPPV and remifentanil sedation were reduced progressively in accordance with the degree of clinical improvement and were discontinued if the patient stably maintained an RR <25 breaths min−1 and PaO2/FiO2 >200.

Criteria for endotracheal intubation

Criteria for switching from NPPV to endotracheal intubation and conventional ventilation were patient discomfort despite analgosedation, high RR (>25/min) and/or failure to maintain a PaO2/FiO2 ratio above 180 after the first hour of protocol, the development of conditions necessitating endotracheal intubation to protect the airways (coma and seizure disorder) or to manage copious tracheal secretions, hemodynamic or electrocardiographic instability, or patient refusal owing to persistent interface intolerance.

Our institutional review board approved the protocol. Written informed consent was obtained from each study patient.

Statistical analysis

All data are expressed as median and interquartile range. Comparisons of median values in failed and successful groups were made using the nonparametric Mann-Whitney test, while the mortality rate in the failed and successful groups was evaluated with nonparametric two-tailed chi-squared test or Fisher’s exact tests when appropriate. Factors independently associated with endotracheal intubation were identified using a logistic regression model. Age, SAPS II, and remifentanil dosage were dichotomized based on the median values of the distribution, whereas the cut-off values were assessed with the receiver operating characteristic curve. The cut-off values selected were those resulting in the fewest false classifications: for equal sensitivity, the threshold value with the highest specificity is given. This decision was based on the assumption that the disadvantages associated with either a false-positive or false-negative result were equal, since delayed or unnecessary intubations were considered equally deleterious. The appropriateness of the cut-off value was evaluated using a logistic regression model. We did not include in the model those variables that define NPPV failure as RR or PaO2/FiO2.

A p value <0.05 was taken to indicate statistical significance.

Results

Patients’ characteristics during the study period are summarized in Table 1. Twenty patients had pneumonia (4 aspiration pneumonia, 10 bacterial pneumonia, 6 interstitial pneumonia), 13 had post-traumatic lung contusions, and 3 had pancreatitis. None of these patients had septic shock or multiple organ failure at the time of enrollment.

Table 1 Patients’ main characteristics

Twenty-two out of 36 patients [61%, SAPS II 32 (30, 38)] continued the NPPV treatment after the introduction of remifentanil infusion; respiratory rate decreased from 34 (31, 37) to 24 (20, 26) min−1 (p < 0.0001) and the PaO2/FiO2 ratio increased from 156 (144, 176) to 270 (210, 300) mmHg (p < 0.0001) after 1 h of NPPV with remifentanil analgosedation either with helmet or total face mask.

Conversely, 14 patients [39%, SAPS II 39 (35, 42)] failed to continue the noninvasive treatment, requiring ETI after a mean remifentanil infusion of 2.5 ± 2.3 h as shown by the Kaplan-Meier curve (Fig. 1); 12 out of 14 failures were caused by the persistence of discomfort despite remifentanil infusion, probably worsened, at least in part, by the concomitant persistence of dyspnea [respiratory rate from 35 (30, 38) to 27 (25, 35) min−1, p = 0.002] and an inability to increase the PaO2/FiO2 ratio above 180 mmHg in 12 patients; conversely 2 failures were owing to hemodynamic intolerance due to septic shock. Thirteen patients were admitted for chest trauma. Six of 13 trauma patients were treated with helmet and 4 out of 6 failed NPPV and were intubated (67%), while 7 of 13 were treated with the total face mask but just one failed NPPV and was intubated (14%, p < 0.05).

Fig. 1
figure 1

Kaplan-Meier curve of NPPV duration with helmet and mask and remifentanil sedation (y-axis percentage of patients with NPPV). The difference was not significant. p = ns

The ICU mortality rate in the failure group patients was 50% (7 out of 14, all died from septic complications and multiple organ failure) versus 14% in the NPPV success group (3 out of 22) (p < 0.05). The mortality rate in this second subgroup was related to abrupt cardiac arrest occurring well after patient stabilization, due to hemorrhagic alveolitis (two cases) and to ventricular fibrillation (one case).

The mean remifentanil dose administered was 0.07 ± 0.03 μg kg−1 min−1 and the infusion lasted for 52 ± 10 h in the NPPV success group, while in the failure group it was continued also during invasive mechanical ventilation. No patients treated with total face mask had sores or skin breakdown, while two patients treated with helmet had armpit skin lesions.

No patient had respiratory drive or hemodynamic alterations due to remifentanil infusion during the study period.

According to the logistic regression model, only a SAPS II >37 was independently associated with NPPV failure and need for endotracheal intubation (Table 2). We did not include in the model those variables that are well known to predict failure such as RR or PaO2/FiO2.

Table 2 Multivariate analysis of risk factors for endotracheal intubation

Finally, the two patients who refused to enter the study both failed NPPV and required endotracheal intubation; one patient survived and one died from sepsis and multiple organ failure.

Discussion

Our results suggest that remifentanil-based sedation is an important tool to treat noninvasive ventilation failure due to discomfort and interface intolerance.

A cross-sectional Web-based survey [13] carried out on American and European physicians concluded interestingly that most physicians infrequently use sedation and analgesic therapy for acute respiratory failure patients receiving NPPV, but practices differ widely within and among specialties and geographic regions. Sedation was usually administered as an intermittent intravenous bolus, outside of a protocol. A benzodiazepine alone was the most preferred (33%), followed by an opioid alone (29%). Europeans were less likely to use a benzodiazepine alone (25 vs. 39%, p < 0.001) but more likely to use an opioid alone (37 vs. 26%, p < 0.009). North Americans more commonly used sedation, analgesia, and hand restraints than Europeans.

Different physiological studies on the effects of analgesia-based sedation on ventilatory response [6, 7, 14] suggest that a continuous infusion of new generation opioids may be used as a single sedative agent, allowing patient discomfort to be mitigated and the desired level of awake sedation to be obtained with no significant effects on respiratory drive, minute volume, respiratory pattern, blood gases, and hemodynamics. These data are fully confirmed by our experience, where no patient had respiratory drive or hemodynamic alterations during the study period.

Among the new opioids, remifentanil offers several pharmacokinetics advantages such as the steady plasma levels achieved in about 10 min and the half-time for equilibration between plasma and its effect compartment of about 1–1.5 min during a constant-rate infusion. A further advantage of remifentanil is its constant and short context-sensitive plasma half-time, which allows a prompt recovery after stopping the infusion.

In more than 60% of our patients, we were able to continue NPPV avoiding endotracheal intubation. In less than 40% of patients, this strategy did not avoid the discomfort and the patients refused to continue NPPV and were intubated after a mean of 2.5 ± 2.3 h. The mortality rate in this subgroup was about 50%. In our protocol we applied a very rapid switch to invasive ventilation because the patient had already failed a prior NPPV trial and expressed a request to interrupt NPPV. We monitored discomfort and pulse oximetry very closely to avoid delayed intubation and increased morbidity or mortality, as recently demonstrated in a large clinical study [15]. The Kaplan-Meier curve clearly shows that most ETI occurred within the first 2 h from protocol enrolment (Fig. 1). Patients who failed NPPV in our study had a higher SAPS II score, 39 (35, 42), and a PaO2/FiO2 lower than 180 after 1 h of NPPV with remifentanil administration. The multivariate analysis showed that a SAPS II >37 was independently associated with the need for endotracheal intubation. These data are consistent with previous studies [16, 17]. Antonelli and colleagues [16] demonstrated with a multivariate analysis that a SAPS II >34 and a PaO2/FiO2 <175 after 1 h of NPPV were independently associated with the need for endotracheal intubation in a group of hypoxemic patients with ARDS.

In a survey of 42 ICUs [17] evaluating the type of ventilatory support used in patients with acute respiratory failure of various origins (N = 689), multiple regression analysis revealed that the SAPS II and intolerance to NPPV were two independent predictors of the need for mechanical ventilation. Indeed, when patients with ARF are unable to tolerate NPPV therapy because of poor interface tolerance or a persistent high level of dyspnea, the result is the need for an early interruption of this treatment, leading to a decreased treatment efficacy. Remifentanil-based analgesia allowed continuing NPPV in 60% of patients that had previously refused the therapy for discomfort, probably contributing to decreasing the dyspnea perception and the respiratory rate.

It is important to note that both interfaces used in our study are commonly considered to offer greater patient tolerance. In our population the remifentanil dosage needed to obtain a better comfort with helmet ventilation was significantly higher than with full-face mask ventilation (p < 0.05). A possible explanation for this difference could be represented by the high incidence of chest trauma (N = 13) in our population, probably showing more side effects due to the armpit braces. Because the total face mask covers the entire face, there are no pressure points around the nose to cause sores or skin breakdown, no pressure under the armpit braces, and good vision through the lightweight face plate, which helps to minimize the feeling of claustrophobia, as with the helmet.

This study has some limitations: the tolerance of NPPV was evaluated without using an accepted comfort scale; moreover, it was performed in a center that is expert both in NPPV and in remifentanil sedation. In cases where there is little experience with NPPV in HARF and/or remifentanil sedation, adopting remifentanil for this indication might be inappropriate.

In conclusion, the results of this preliminary study suggest that the use of remifentanil-based sedation in these patients is feasible and safe. Further prospective controlled clinical trials are needed to confirm the effects on interface intolerance and decreased NPPV failure.