Introduction

Acute respiratory failure (ARF) is frequently characterized by hypoxemia. However, little is known about the effect of oxygenation on dyspnea and respiratory variables in patients undergoing mechanical ventilation for ARF. This may be of particular interest when patients are ventilated in pressure support ventilation (PSV). When properly chosen, PSV is generally used to ameliorate the patient-ventilator interaction: this mode of ventilation allows the patient to change respiratory frequency, inspiratory to expiratory time ratio, and tidal volume on his own respiratory drive and mechanics. To satisfy the needs of patients the physician can directly control four ventilatory parameters: the level of pressure support, the extrinsic positive end expiratory pressure (PEEPe), the expiratory trigger cycle off, and the fraction of inspired oxygen (FiO2). The level of PEEPe can improve spontaneously breathing patterns, decreasing respiratory frequency and increasing the tidal volume (V T) [1, 2, 3, 4]; on the other hand, the role played by different inspiratory oxygen fractions (FIO2) has not been determined with certainty. Pesenti et al. [5] demonstrated that in patients with acute respiratory distress syndrome (ARDS) modifying oxygenation there are significant changes in respiratory drive, but the most valuable result is that an important hypoxic drive is still present even at a SpO2 level thought to be adequate (92.6±2.7%). Hence the hypothesis could be made that the respiratory drive can be strongly affected by arterial oxygenation to a point at which by varying FIO2 it should be possible to modulate various respiratory variables, such as the respiratory frequency (f) [6]. Until now, to reduce f during PSV a physician can mainly increase (a) the level of pressure support, which could put the patients at risk of volo-biotrauma or, in presence of expiratory flow limitation, of an increase in the end expiratory lung volume, which has been previously advocated as a major reason for hemodynamic instability and wasted effort [7, 8]; or (b) the level of sedation, although it has been previously demonstrated that heavily sedation should be avoided due to its deleterious effects on weaning [9]. In this connection the respiratory drive modulation obtained by varying FIO2 could help to reduce the need for sedation. Hence the aim of this study was to evaluate whether different levels of arterial oxygenation affect (a) the patient's comfort, (b) the respiratory variables, or (c) the gas exchange, since high levels of FIO2 can reduce the respiratory drive or alter the VA/Q distribution within the lung [10] to a point that PaCO2 can rise.

Materials and Methods

Patients

The study enrolled 13 semirecumbent hypoxemic patients undergoing mechanical ventilation for ARF of various causes (Table 1). At the time of the study all patients were in stable clinical condition and were ventilated in PSV mode, the setting of which was established and adjusted according to clinical criteria by the treating physician, who was blinded to the study. The exclusion criteria at the time of the study were: acute ischemic heart disease, emodynamic instability, rrhythmia, resence of metabolic disturbances, and need of sedation. The investigation was approved by the Institutional Ethics Committee and informed consent was obtained from each patient. The patients' clinical characteristics and the baseline ventilator settings are shown in Table 1. All of them had a variable degree of hypoxemia, as indicated by a PaO2/FIO2 ratio lower than 300. At the beginning of the study none of them exhibited any pH disturbance (mean pH 7.39 ± 0.01), and the lactate level in the blood was within the normal range.

Table 1 Patient clinical characteristics and baseline ventilator settings

Monitoring and instrumentation

Eleven patients were ventilated by using the Evita 4 (Dräger). During pressure support ventilation the following variables were read from the ventilator screens: tidal volume (V T), f, and airways pressure (Pao). The airway pressure drop developed in the first 100 ms of an occluded inspiration (P0.1), a useful index of neuromuscular inspiratory drive was manually calculated by using the cursor that appears on the Evita 4 ventilator screen after the occlusion maneuver. The mean value of three consecutive determinations was used as the real value of P0.1.

The remaining two patients were ventilated by Servo 900C (Siemens-Elema,Berlin, Germany), and hence other equipment was used to monitor the respiratory variables. Flow (V′) was measured with a heated pneumotachograph (3700, 0–160 l/min, Hans Rudolf, Kansas City, Mo., USA) connected to a differential pressure transducer (DP55±3.5 cmH2O Raytech Instruments, Vancouver, BC, Canada). The response of the pneumotachograph was linear over the experimental range of flow. Changes in lung volume were obtained by digital integration of the flow signal. Pao was measured through a side port on the connector between the endotracheal tube and the pneumotachograph using a differential pressure transducer (DP55 ± 100 cmH2O Raytech Instruments). The transducers were calibrated before and after each study. The V′, V T, and Pao signals were amplified, low-pass filtered at 50 Hz and digitized at 100 Hz by a 16-bit analog-to-digital converter (Direc Digital Recording System Software version 3.3; Raytech Instruments;. The digitized data were stored on a computer hard disk for subsequent analysis using Anadat software (Anadat 5.1; RHT-InfoDat; Montreal, QC, Canada).

P0.1 was measured by using a rapid valve (opening time < 10 ms) connected to the expiratory line of the ventilator. The airway occlusion lasting for 100 ms at the beginning of inspiration was controlled by computer software (Raytech Instruments). Five resting breaths elapsed between each P0.1 determination. As for the P0.1 calculation based on the Evita 4 screen the mean value of three consecutive determinations was used.

Severity of dyspnea was evaluated by a resident physician not involved in the study, by using the visual analogue scale (VAS) [11]. Patients were asked to place a horizontal mark on a printed 10-cm horizontal scale in response to the question: “How short of breath are you right now?” [12]. The line had descriptors below the extreme ends. On the left was the word “none,”indicating no breathlessness at all, while on the right was the opposite response “extremely severe.” For each condition tested patients placed a vertical mark on the line that best represented the intensity of their dyspnea. The latter was measured as the distance in cm from the left side of the horizontal line (corresponding to no dyspnea) to the mark placed by the patients. A new scale was used each time the measurements of patients' dyspnea were assessed [13]. The patients were carefully instructed on the appropriate use of the scale before the beginning of the protocol. The sensitivity and the reproducibility of the VAS has been validated with other measures of dyspnea [12]

The electrocardiogram, heart rate, systemic arterial blood pressure, pulse oximetry (SpO2), and end-tidal carbon dioxide (ETCO2) were continuously monitored. All patients had the radial artery catheterized with a 20-G plastic catheter to measure arterial pressure and to sample arterial blood, which was analyzed for pH, blood gases, and hemoglobin saturation (ABL 700, Radiometer, Copenhagen, Denmark).

Protocol

The respiratory variables were firstly measured at the level of FIO2 chosen by the treating physician. While the level of both pressure support and PEEPe were kept constant for the entire duration of the study, FIO2 was changed randomly by about 10% of the initial value, from 21% to 60% (Fig. 1). Severity of dyspnea was rated by using the VAS [11] 15–20 min after FIO2 variation. The respiratory variables were collected immediately before the assessment of the severity of dyspnea and before any FIO2 change. To ascertain whether a higher level of hyperoxemia has any further effect on the respiratory variables, measurements were also carried out with a level FIO2 of 80%. Randomization was obtained by using a sequentially numbered, opaque, sealed envelope.

figure 1

Relationships of proportional changes in dyspnea, P0.1, and respiratory frequency (f) from baseline to corresponding variation in inspiratory oxygen fraction (FIO 2). Values are expressed as mean ± standard error. p < 0.05 vs. baseline, ∗∗ p < 0.05 vs. previous significantly different value

Statistical analysis

Data are presented as mean ±SD. The approximate degree of normal distribution was calculated for each parameter by the Kolmogorov-Smirnov test. In presence of normally distributed data values obtained at different levels of FIO2 were compared using Friedman's repeated-measures analysis of variance on ranks, and Tukey's test was used for all pairwise multiple comparison procedures. When the normality test failed, the Kruskal–Wallis one-way analysis of variance on ranks was used, and the Student–Newman–Keuls Method was used for all pairwise multiple comparison procedures. Regression analysis was performed by the least squares method. A value of p < 0.05 was accepted as statistically significant.

Results

Modulation of FIO2 was able to vary consistently and significantly the respiratory variables since a FIO2 decrease was associated with an increase in dyspnea, P0.1, f, and V E, whereas a FIO2 increase caused a decrease of these variables (Table 2; Figs. 123). This was evident also when values of FIO2 higher than those thought to satisfy the patient's needs were used. Indeed, the mean FIO2 chosen by the treating physician was 40% (Table 1), but a significant improvement in the respiratory variables was detected at FIO2 60% (Figs. 123). In detail, hyperoxemia was very effective in reducing respiratory discomfort which was statistically related to the reduction in the respiratory drive (P0.1; R 2 = 0.89). Moreover, V E also decreases and this was entirely due to a reduction of f since V T did not vary. This result was associated with a concomitant increase of PaCO2, which was not so large to induce respiratory acidosis.

Table 2 Respiratory variables at different levels of inspiratory oxygen fraction (FIO2)
figure 2

Relationship between changes in dyspnea and P0.1, and corresponding values of inspiratory oxygen fraction (FIO 2). Values are expressed as mean ± standard deviation. p < 0.05 vs. baseline, ∗∗ p < 0.05 vs. previous significantly different value

figure 3

Relationship between changes in respiratory frequency (f), minute volume (V E), and PaCO2, and corresponding values of inspiratory oxygen fraction (FIO 2). Values are expressed as mean ± standard deviation. p < 0.05, vs. baseline, ∗∗ p < 0.05 vs. previous significantly different value

Values of FIO2 higher than 60% were associated with a further amelioration of dyspnea (Figs. 123) and to a further decrease in P0.1. In this connection it should be noted that the dyspnea was reduced by more than 50% by simply increasing the FIO2 level from 21% up to 80% (Figs. 123)

Discussion

The most important finding of the present study is that during PSV the respiratory drive can be modulated by varying the FIO2. Since the decrease in respiratory frequency obtained by increasing the FIO2 reflects a concomitant decrease in dyspnea, it may be inferred that the patient'scomfort is strongly influenced by using different levels of FIO2. Surprisingly, P0.1 was affected by inspiratory oxygen fraction even when the SaO2 was higher than 98%, suggesting an influence of PaO2 on respiratory drive. This was proposed in 1958 by Lloyd and coworkers [14]. These authors found that the simple increase in the arterial PaO2 from about 90 mmHg to 120 mmHg was able to reduce minute ventilation by about 1 l while a further increase in PaO2 to 600 mmHg decreases V E by about 3.5 liter. In our patients V E decreased by 3 l (Table 2), mainly because of f reduction. The implication of these findings is that the comfort of a patient undergoing pressure support ventilation can be influenced by oxygenation even in presence of an SpO2 of 98% (Table 2). This is of clinical interest because patients are continuously monitored by SpO2, which has been shown to differ consistently from SaO2. Indeed Van de Louw et al. [15] suggested that an SpO2 above 94% is necessary to ensure an SaO2 of 90%. This implies that patients monitored only by SpO2 are at risk of an increased hypoxic respiratory drive, and hence the effect of FIO2 should be tested before using other therapeutic approaches, such as an increased level of sedation. Although dyspnea and respiratory variables continued to improve to FIO2 of 80% (Table 2; Figs. 123), greater statistical variation was observed when the PaO2 reached 184 ± 24 mmHg at an FIO2 of 60%. This could be of clinical interest because it implies that a modulation of the respiratory drive can be achieved at an FIO2 level which can be used safely in the ICU setting [16]. In this connection it should be noted that no chronic obstructive pulmonary disease (COPD) patients with acute respiratory failure were enrolled in the present study. It is well known that excessive O2 administration in COPD patients can lead to hypercapnia mainly due to increased inhomogeneity of VA/Q distribution within the lungs [10]. Hence it is likely possible that the results obtained in the present study are not applicable to COPD patients. The same reasoning can be applied to patients with more severe hypoxemia than that evidenced in our study, although Pesenti et al. [5] showed similar results in ARDS patients undergoing pressure support ventilation. They demonstrated that at an SpO2 level thought to be adequate (94%), an important hypoxic drive was still present and an FIO2 increase was able to reduce both VE and P0.1. In this case V E decrease was due to a drop in f, as in the present study. Interestingly, although the values of f at low and high level of FIO2 were similar to those obtained in our study, the values of P0.1 were higher in the present study, probably reflecting a different patient population or different levels of sedation and hence a different pharmacological drive modulation. Indeed, our patients did not receive sedative drugs for 24 h before the investigation, while the patients enrolled in the study of Pesenti et al. were given a certain amount of benzodiazepines or fentanyl.

As previously demonstrated in a population of ambulatory COPD patients, acute hyperoxemia at rest induces a significant reduction in ventilation and dynamic hyperinflation with concurrent improvement in dyspnea sensation [18, 19, 20], which in our study was strictly connected to the amelioration of P0.1(R 2=0.89; Fig. 2). While in our patients the reduction in dyspnea was probably due to the decrease in the respiratory drive during hyperoxia rather than to a decrease in dynamic hyperinflation, as it is for COPD patients; we cannot exclude that the amelioration of dyspnea obtained by increasing the FIO2 levels was due to the progressive decrease in the intrinsic positive end expiratory pressure since the latter has been shown not only in patients with COPD but also in those with ALI/ARDS [21].

Another interesting aspect of hyperoxemia is its effect on well known weaning parameters, such as P0.1 and f/V T [22, 23], which varied greatly with oxygenation (Table 2). This suggests that FIO2 affects the ability of the considered test to discriminate successful weaning and weaning failure, although it is never mentioned the level of oxygenation that should be used during the determination of the weaning indices. Conti et al. [17] have recently shown that none of the predictors of weaning is powerful enough to predict weaning success. It is then possible that one of the reasons for theirs results could be the absence of FIO2 standardization during the weaning trial.

The results of the present study can probably be explained by the ability of the peripheral fast responding (1–3 s) chemoreceptors to react to a fall in PaO2 or to a rise in PaCO2. Simple elicitation of the ventilatory response to hypoxemia will normally result in hypocapnia and the response is then a combination of the hypoxic drive and the resultant hypocapnic depression of breathing. Hence under isocapnia or nearly so both PO.1 and f changes would have been amplified compared with those in this study. Our aim, however, was to investigate a practical aspect of the daily care of ICU patients i.e., in which ways one can modulate the respiratory drive of patients undergoing PSV. Finally, it is true that the peripheral chemoreceptors are under nervous control and also under the influence of a fall of their perfusion rate. While the latter should have played a minor role since no hypotension was detected for the entire duration of the study, we cannot exclude a nervous influence.

Criticism: possible side effects of O2 administration

It has previously been demonstrated that pulmonary oxygen toxicity is present in laboratory animals such as the rat. However, humans seem to be far less sensitive, probably because of different level of provision of defenses against free radicals [16]. Although the production of oxygen-derived free radicals is increased at high levels of PaO2, it seems that the tissue defenses against free radicals are usually effective up to a tissue oxygen partial pressure (PO2) of about 450 mmHg [16], a value much higher than that normally obtained in the clinical setting (Table 2). Moreover, there is compelling evidence that prolonged exposure to FIO2 of 100% does not result in demonstrable pulmonary oxygen toxicity if the PaO2 is lower than 255 mmHg [16] underlying the hypothesis that the oxygen toxicity is related to tissue PO2 and not to FIO2. This partially explains why no studies have until now demonstrated adverse effects of oxygen administration in patients. While Gilbe et al. [24] concluded that adverse effects of oxygen on the alveolar epithelium are rarely of practical importance in hypoxemic patients, Capellier et al. [25] provided evidence that the lungs of patients with acute respiratory failure exhibit some relative resistance to prolonged oxygen exposure. Hence it seems that an FIO2 level responsible of a PaO2 less than 250 mmHg can be probably used safety in the clinical setting.

Conclusion

During PSV the respiratory drive can be modulated by varying the FIO2. Interestingly, this effect is independent of the SpO2 achieved, since even at FIO2 greater than 60%, which corresponded to an SpO2 above 98%, dyspnea and P0.1 continuing to improve. Our results suggest the clinical relevance of having an appropriate oxygenation in patients undergoing PSV for acute respiratory failure, bearing in mind that even moderate levels of hypoxemia, not always detected by SpO2 monitoring, are responsible for acute dyspnea and patient discomfort.