Introduction

One of the most frequent reasons for intensive care unit (ICU) admissions is hypercapnic respiratory failure [1, 2]. Minute ventilation decreases from wakefulness through various stages of sleep, and it may also decrease in the supine position in obese patients [3,4,5]. Standard pressure support modes may not adapt to the varying needs of a patient within a given daytime, position or respiratory mechanics. To increase the efficacy of noninvasive ventilation (NIV), volume-assured modes such as average volume-assured pressure support [AVAPSTM (Philips Respironics Inc., Murrysville, Pennsylvania, USA)] has been used in hypercapnic patients [6,7,8]. While some of the previous studies find this mode superior to standard bilevel modes, others do not find any advantage [9,10,11,12].

Possible explanation of the lower performance of volume-assured modes than expected may be inability to open the upper airway and eliminate obstructive apnea [13, 14]. Recently developed auto-titrating NIV mode, i.e., average volume-assured pressure support-automatic expiratory positive airway pressure (EPAP) [AVAPS-AETM (Philips Respironics Inc., Murrysville, Pennsylvania, USA)], while automatically titrating pressure support level according to targeted tidal volume, also measures the upper airway resistance, and responds by adjusting EPAP to maintain upper airway patency.

We hypothesize that this new auto-titrating mode (AVAPS-AE) will decrease a higher amount of PaCO2 within a shorter period of time than a volume-assured mode (AVAPS) in hypercapnic ICU patients.

Methods

Patient selection

Patients who were admitted to a university hospital medical ICU, between October 2012 and September 2014, during an episode of acute hypercapnic respiratory failure (PaCO2 > 55 mmHg) were screened from the ICU database for the study. The prospectively collected and recorded data according to standardized noninvasive ventilation and medical treatment protocols were evaluated retrospectively. Exclusion criteria were; patients < 18 yrs of age, having terminal disease, and intubation indication. The Local Ethics Committee approved this study (No: 894-2015), and exempted it from the informed consent requirement because researchers accessed an anonymized database retrospectively for this analysis.

Primary outcome of the study is the amount of reduction in PaCO2 levels and the timing of this reduction. Secondary outcome is the duration of ICU stay.

Included patients were evaluated in two groups: group 1 (AVAPS-AETM) and group 2 (AVAPSTM). The most recent stable pulmonary function tests (PFT) and echocardiography values before admission to hospital were recorded. Obesity hypoventilation syndrome (OHS) was defined as the presence of obesity [body mass index (BMI) > 30 kg/m2]; the absence of other causes of hypoventilation and daytime hypercapnia (PaCO2 45 mm Hg) [15]. Both groups were paired and matched with two matching methods; individual matching (matched pairs) and group (frequency) matching (comparing statistically). Our privileged matching parameters were baseline PaCO2, diagnosis of the patients (chronic obstructive lung disease, OHS, heart failure), BMI and daily usage time. First, we sorted our 73 patients according to their baseline PaCO2 level from highest to the lowest for both groups. AVAPS AE group had more patients with higher baseline PaCO2 levels (for example > 90 mmHg), and the AVAPS group had more patients with lower PaCO2 levels (45–55 mmHg).We excluded patients with highest and lowest PaCO2 levels first. Then, we tried to pair the groups according to their diagnosis, daily NIV usage time and BMI as much as possible. After this process, we used group matching for other parameters such as age, gender, and PFT for making statistical comparisons. Most frequent reasons of exclusion in 23 patients were PaCO2 levels (very high or low), daily usage time (less than 10 h) and diagnosis. Another exclusion reason was switching of the starting mode to other modes in case of persistent hypercapnia or intolerance within longer than 1 h at the first 6 h, and longer than 6 h at the first 4 days.

The PaCO2 response was evaluated in different aspects as 10% decrease, 10 mmHg decrease, decrease within the 1st 6 h, and decrease within the first 4 days according to baseline.

Noninvasive mechanical ventilation application protocol

The same type oro-nasal mask was used for all patients. Both modes were delivered by Trilogy 100 ventilators (Philips-Respironics, Murrysville, Pennsylvania, USA). According to our ICU patient follow-up rules, NIV parameters of all our patients were recorded in different memory cards and transmitted to a database in ICU computer. The following data given in Table 4 were recorded from the ventilators’ software (Directview®); ventilator settings, daily usage time, mean targeted tidal volume, leaks, respiratory rate, tidal volume (VT), minute ventilation and percentage of respiratory cycles triggered by the patient.

Arterial blood gas (ABG) samples were obtained from each patient before the first application of NIV at admission, and just before deconnecting patient at the end of every single NIV session.

Noninvasive ventilation protocol and settings for modes were as follows:

AVAPS: maximum inspiratory positive pressure: 30 cmH2O and minimum inspiratory pressure: 4 cmH2O higher than EPAP value. EPAP: 5–8 cmH2O. According to snoring of the patients or apnea alarms of the ventilators or if obstructive apneas were witnessed during NIV, the EPAP level was raised 1–2 cm H2O by the ICU physician.

AVAPS-AE: Maximum pressure: 25–35 cmH2O, PS maximum: 15–25 cmH2O, PS minimum: 5 cmH2O, EPAP max: 15 cmH2O, EPAP minimum: 5 cmH2O, Rise time: 300 ms, AVAPS rate: 2 cmH2O/min.

In both modes, the trigger was adjusted as auto-track, and the target tidal volume (VT) was set as 6–8 ml/kg (ideal body weight), and the respiratory rate was set as 14 breaths per minute.

Oxygen therapy was given so as to keep oxygen saturation measured by pulse oximeter higher than 92%, or PaO2 higher than 60 mm Hg. All patients were treated at least for 10 h/day with both modes. If there was no improvement in PaCO2, and pH values or patient could not tolerate the mode, then the mode was switched to the other modes (BiPAP-S, BiPAP-ST, AVAPS-S, AVAPS-ST or AVAPS-AE). All patients received standard medical therapy (antibiotics, corticosteroids, bronchodilatators as needed) according to their underlying diagnosis unless there was a contraindication. Since diuretics may influence serum HCO3 and pH levels, the diuretic usage of the patients were recorded.

Statistical analysis

We performed a power analysis to calculate sample size, and 35 patients were found to be necessary for a significance level of 0.05. The difference between the mean PaCO2 levels of the two groups at baseline and at first 6 h [(baseline PaCO2 − first 6 h PaCO2) of group 1] [(baseline PaCO2 − first 6 h PaCO2) of group 2] and at baseline and at 4th day [(baseline PaCO2 − first 4th day PaCO2) of group 1] [(baseline PaCO2 − first 4th day PaCO2) of group 2] were compared with independent t test. The mean values between the 1st and 4th days were compared with repeated measures of ANOVA method. To compare the response rates and means, Chi-square test, t test and Mann–Whitney U tests were used. The Kaplan–Meier estimate-of-survival curve was used to determine the cumulative PaCO2 reduction rates with time; curves of the two groups were compared using the log-rank test.

Results

Seventy-three patients were screened, and according to the inclusion criteria, 50 were assigned for the study. Twenty-eight of them were selected for group 1 (AVAPS-AETM) and 22 for group 2 (AVAPSTM). There were no significant difference between the demographic parameters, diagnosis, pulmonary and cardiac functions and admission ABG values across the groups (Table 1). Mean target VT settings (501 ± 42 ml in group 1 and 497 ± 37 ml in group 2, p = 0,701), daily duration of NIV use (11 ± 2 h in group 1 and 10 ± 3 h in group 2, p = 0,456) (Table 4) and diuretic usage rates (32% in group 1 and 44% in group 2, p = 0.405) were also similar across the groups.

Table 1 Demographics and baseline characteristics of the patients
Full size table

At the first 6 h, there was a significant difference between the amount of decreasing PaCO2 across the groups (7 ± 7 mmHg in group 1 and 2 ± 5 mmHg in group 2, p = 0.025). The PaCO2 decreased more than 5 mmHg in 93% of group 1 and in 60% of group 2 (p = 0.004). Although 10 mmHg or 10% reduction in PaCO2 levels occurred in a similar number of patients in both groups, the decrease occured within a shorter time period in group 1 (p < 0.05) (Tables 2, 3). Figure 1 shows the timing of at least 10% reduction in PaCO2 levels of group 1 and group 2 patients with NIV therapy. There was a significant difference between the groups’ response in log rank test (p = 0.007). The HCO3− and PaCO2 levels decreased significantly more in group 1 than group 2 at 4th day (p = 0.007 and p = 0.008, respectively) (Table 3).

Table 2 Comparison of the PaCO2 reduction responses across the Group 1 and Group 2
Full size table
Table 3 Comparison of mean daily ABG values across the groups
Full size table
Fig. 1
figure 1

Ten percent reduction rate in PaCO2 levels of the two groups after NIV therapy during ICU stay

Full size image

While maximum and mean inspiratory pressures were similar, maximum and minimum EPAP levels in group 1 were significantly different from group 2. Mean VT and leak were significantly higher in group 1 (Table 4).

Table 4 Comparison of the course of ventilation parameters between the groups
Full size table

There were two intubations in group 1, and no intubation in group 2. There was no significant difference in the length of ICU stay of the two groups (7.4 ± 2.8 days vs. 8.2 ± 2.8 days p = 0.354).

Discussion

In this study, the PaCO2 responses of acute hypercapnic respiratory failure patients to new auto-titrating NIV mode and volume-assured pressure support mode are evaluated. We assess this response from different aspects such as 10 mmHg or 10% reduction rates of PaCO2 levels, timing of this reduction, the amount of decrease within the first 6 h and the first 4 days. We find that higher amount of PaCO2 levels decreases within a shorter period of time in patients treated with auto-titrating NIV (AVAPS-AE) mode. This response with AVAPS-AE mode is more pronounced within the first 6 h, and significantly greater reductions in PaCO2 and plasma HCO3 levels take place within the first 4 days. There is no significant difference between the lengths of ICU stay. With AVAPS-AE mode, significantly higher maximum EPAP levels, mean VT and leaks are achieved. This study is the first to investigate the efficiency of AVAPS-AE mode in hypercapnic respiratory failure critical care patients, and this efficient automatic titration might be important for an ICU environment.

Auto-titrating modes (automatic EPAP or CPAP) have been used for years in sleep medicine to treat obstructive sleep apnea syndrome (OSAS) [16, 17]. Combined with servoventilation, automated EPAP has been used in patients who have both Cheyne Stokes ventilation and OSAS [18, 19]. Similar to these usages, they may also have potential advantages in hypercapnic respiratory failure in ICUs or respiratory wards. Recently, two studies report this ventilation mode as effective and safe in patients with chronic obstructive pulmonary disease (COPD) and overlap syndrome [20, 21]. On the other hand, the most severe forms of these diseases are usually treated in ICUs or respiratory wards, and the physicians who care for them are faced with many difficulties in adjusting appropriate pressure and volume parameters. First, it is difficult to predict which patient has obstructive apnea and adjust EPAP accordingly; because, not only obese patients but also nonobese patients with COPD, neuromuscular disease and kyphoscoliosis may have obstructive apneas [22, 23]. Second, in contrast to patients with mild or moderate sleep-related breathing disorders, critically ill patients with severe blood gas abnormalities might need NIV therapy nearly the entire day in ICUs. On the other hand, obstructive apneas occur during sleep, and disappear during wakefulness, and EPAP levels should continuously be changed during sleep and wakefulness. Third, physicians usually don’t have time for these frequent adjustments, and cannot predict the exact amount of inspiratory and expiratory pressures required by the patient. Our results suggest that AVAPS-AE may meet these expectations in part, and reduce PaCO2 more efficiently than AVAPS. The more efficient PaCO2 reduction in group 1 may be attributed to the higher VT that was achieved automatically by the mode. In fact, target VT of the two groups was set similarly as 6–8 ml/ideal kg of the patient; but, actual VT was significantly higher in group 1 in spite of similar actual maximum and mean inspiratory pressure levels. This might be explained by better elimination of upper airway resistance, and therefore, obstructive events with higher EPAP levels in group 1. AVAPS-AE can treat hypercapnia in patients with hypoventilation syndromes not just by delivering targeted tidal volume but also by identifying and treating obstructive apneas. If EPAP levels are suboptimal, obstructive events may reduce NIV efficacy. Another risk during NIV is to set expiratory pressure levels higher than the requirements of the patients who have normal upper airway resistance. A meta-analysis performed about obstructive sleep apnea disorder supports our results; showing that compared to fixed CPAP, the use of auto-CPAP reduces mean therapeutic pressures [16, 17].

In 64% of the patients of each group, the BMI was ≥ 30 kg/m2 suggesting that the upper airway resistance might be similarly higher in both groups. In this situation, one can expect that most of our patients required higher EPAP levels, and AVAPS-AE met this need [24]. This is also an important point to see the differences in efficacy of two modes. There is no significant difference between the reduction of mean amounts of PaCO2 levels across the groups (Table 3 and Fig. 1) during the first 4 days, but similar amounts of PaCO2 reduction was recorded within a shorter time period in group 1. This finding made us think that AVAPS-AE might be more efficient in different patient populations such as in more obese ones. This also supports our hypothesis that in an ICU setting, and in most obese patients, manual expiratory pressure titration may not be adequate. But due to their low number, it would be impossible to compare the obese and nonobese patients in this study.

When we look at the bicarbonate (HCO3−) levels during the first 4 days, we saw that it decreased significantly within this time period in group 1 while it did not show any change in group 2. This finding also supports a more efficient reduction in PaCO2 levels with the auto-titrating mode. On the other hand, diuretic usage may also influence HCO3− kinetics independently from existing hypercapnia [25]. That is why we compared the diuretic usage rates of our patients but could not find any significant difference across the groups.

We can explain higher levels of air leak in group 1 with higher VT and EPAP levels. There may be other reasons for high leak like inappropriate mask adjustments, yet as another limitation, due to retrospective nature of the study, those cannot be analyzed in this study.

In our study, length of ICU stays were similar in both groups. This is not unexpected since length of ICU stay is not related only to PaCO2 levels. Most of our ICU patients had acute exacerbations of underlying diseases such as COPD and heart failure, or had infections; and treatment of these conditions prolongs the length of ICU stay independent from PaCO2 levels. On the other hand, although the length of ICU stay does not differ with these two modes, rapid PaCO2 response is still quite important in these patients. The rapid improvement in acidosis and consciousness may relieve hypercapnic encephalopathy, may protect those hypercapnic patients from intubation and the many related complications of invasive mechanical ventilation. In context with our hypothesis, Claudett and coworkers find that AVAPS facilitates rapid recovery of consciousness when compared to traditional noninvasive pressure support mode in patients with COPD and hypercapnic encephalopathy [7].

We encountered no safety problems during the treatment period with this new auto-titrating NIV mode. There were two intubations in group 1, but those were not thought to be related to NIV therapy. In those patients, indications for NIV therapy were postextubation hypercapnia and kyphoscoliosis, and both patients showed very good response to NIV therapy at the first week of their admission. But later on, they developed severe pulmonary infection and sepsis, and then had to be intubated. Supporting our results, in a recent study performed in patients with overlap syndrome at their home settings; authors also report the auto-titrating mode as safe and effective [20].

Limitations

One of the important limitations of this study is its retrospective nature, and due to small number of patients and the heterogenity of the study group, the results cannot be generalized. We used daily recorded ventilatory data to understand mechanical parameters and gas exchange relationship. The value of this type of information has not yet been validated [26, 27]. Theoretically, auto-titrating NIV mode (AVAPS-AE) should be more useful in patients with OSAS or OHS, but in our study population since PSG could not be performed, we do not know whether our patients had any obstructive events, nor whether or not the mode provided the patients’ optimal inspiratory and expiratory pressures and opened their upper airway properly. Additionally, we do not know if the mode gave higher EPAP levels just for obstructive events, or whether it increased expiratory pressure levels unnecessarily.

Conclusion

This study gives preliminary data with which to design future prospective controlled studies to compare the efficacy of AVAPS-AE and AVAPS in ameliorating clinical outcome in acute hypercapnic patients by suggesting that ventilation with both NIV modes, provides a similar control of PaCO2, but the auto-titrating NIV mode may provide additional benefits on volume-assured mode, resulting in more efficient and rapid decrease in PaCO2.