INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Isocapnic hyperventilation has been proposed as a possible "new" therapy for carbon monoxide (CO) poisoning (1). Early studies of hyperventilation for CO poisoning used unassisted ventilation with a fixed fraction of carbon dioxide (CO₂) in oxygen (O₂) (2), whereas Fisher and coworkers more recently used a passive apparatus that introduced CO₂ at rates proportional to minute ventilation (E) (5). The benefit of unassisted isocapnic hyperventilation, as practiced before the onset of mechanical ventilation, was limited because increased CO elimination rates depend on spontaneous or voluntary increases in E. Takeuchi and coworkers have proven that 90 min of spontaneous hyperventilation, 2 to 6 times baseline E, is possible in conscious volunteers with low levels of carboxyhemoglobin (HbCO) (6), but these subjects may not adequately model the typical CO-poisoned patient. As predicted by mathematical models (7) and confirmed in dogs (1) and humans (6), large increases in E are required to significantly increase rates of CO elimination. An optimal level of hyperventilation may not be spontaneously sustainable in clinical practice because the time required to decrease HbCO levels to 25% of their initial value is up to 160 min for normobaric O₂ (NBO) at baseline E (11).

Hyperventilation therapy using mechanical ventilatory support is a viable option but cardiac output may be compromised by increased intrathoracic pressure and reduced venous blood return to the heart. If this were the case, O₂ delivery to tissue (DO₂) would decline acutely during therapy when tissue PO₂ is at low levels, potentially worsening the clinical condition of the patients. Fisher and coworkers demonstrated a dramatic increase in CO elimination in dogs mechanically ventilated with a tidal volume (T) of 50 ml · kg¹ when isocapnic hyperventilation was compared with normal ventilation with NBO or room air (RA) ventilation. Because the order of each treatment was fixed and each therapy arm was performed at different mean HbCO levels (51% for RA, 25% for NBO, and 6% for hyperventilation), this study was not able to address the issue of DO₂ during hyperventilation at high HbCO levels.

We used an animal model of multiple severe CO poisonings to assess the effect of isocapnic hyperventilation on CO elimination and DO₂ at high HbCO levels. Because large increases in T may result in increased alveolar ventilation (A) but decreased cardiac output compared with increases in respiratory rate (RR), we also examined the effect of respiratory pattern on CO elimination and DO₂.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Animal Model

This study was approved by the University of Washington (Seattle, WA) Institutional Animal Care and Use Committee and the National Institutes of Health guidelines for animal use and care were followed throughout. Five adult sheep (27.5-36 kg) of either sex were studied. The animals were premedicated by intramuscular injection of xylazine (1.0 mg · kg¹), anesthetized by intravenous administration of thiopental sodium (20 mg · kg¹), and intubated. Anesthesia was maintained throughout the study, using a constant intravenous infusion of thiopental sodium, titrated to suppress hemodynamic and motor responses to noxious stimuli. The animals were initially ventilated with a piston animal ventilator (Harvard Apparatus, Hollison, MA) with a VT from 10 to 15 ml · kg¹, zero positive end-expiratory pressure (PEEP), and a respiratory rate (RR) set to maintain the arterial PCO₂ between 35 and 40 mm Hg. A Swan-Ganz pulmonary artery catheter was placed via the right external jugular vein and arterial and venous catheters were placed in the left groin. The animal was then placed in the prone posture for the remainder of the study. Animals were killed after completion of the study, using a concentrated pentobarbital injection. Data collected include hemoglobin (Hb) concentration, heart rate (HR), arterial pressure (Pa), pulmonary arterial pressure (Ppa), pulmonary capillary wedge pressure (Pcw), thermodilution cardiac output, airway pressures, inspired and exhaled CO₂ concentrations, arterial and mixed venous blood gases (pH, PO₂, PCO₂), and arterial HbCO levels. All exhaust gas from the ventilators was scavenged and released outside the building with ambient air tested continuously for CO.

Treatment Trials

Each animal was exposed to six cycles of CO poisoning and treatment. Poisoning was accomplished by inhalation of 0.6% CO gas in air for 20 to 30 min, delivered via ventilator, until arterial [HbCO] reached 65 to 75%. Ventilation treatment trials were performed with a Servo 900C ventilator (Seimens-Elema, Solna, Sweden) at varying VT, RR, and FI_co2 (fraction of inspired carbon dioxide; balance O₂). After the first poisoning, the animals were ventilated with the Servo ventilator at the baseline ventilatory pattern (RR · VT) with an RR of 10 breaths · min¹ and a VT adjusted to maintain PCO₂ between 35 and 40 mm Hg. The initial HbCO measurement and DO₂ calculation for each trial were made immediately after switchover to CO₂:O₂ ventilation and continued every 5 min until [HbCO] fell below 20%, at which time poisoning was reinstituted. The remaining five trials, performed in varied order, were used to test different ventilatory treatments and are listed in Table 1. For all hyperventilation trials, the amount of inspired CO₂ was adjusted to maintain arterial PCO₂ between 30 and 40 mm Hg.

Data Analysis

CO elimination rates were compared across treatment trials in each animal, allowing each sheep to serve as its own control, increasing the power of the study and limiting the number of animals required. The HbCO washout half-time (t_1/2) of each of the therapies was determined graphically for each animal studied by plotting HbCO levels over time for each treatment trial. DO₂ (ml O₂ · min¹ · kg¹) is the product of blood O₂ content and the cardiac index (CI). O₂ saturation (%) = 100%  HbCO (%), because FI_O2 remained above 550 mm Hg throughout treatment and all heme units were saturated with O₂ or CO. All comparisons used analysis of variance (ANOVA), with p < 0.05 indicating significance. Because larger values of HbCO and DO₂ show more variability than smaller values, statistical tests that assume constant variance will not yield accurate p values. Log(HbCO) and log(DO₂) showed similar variability across their ranges of values; therefore, statistical tests were performed with the logarithms of these variables. To avoid multiple comparison effects, the Fisher protected least significant difference was used. The first trial for all animals, always performed at normal ventilation, functioned only to "load" extravascular tissues with CO and was not included in any of the analyses.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Increases in E are associated with decreasing HbCO t_1/2 (Figure 1). In the five sheep studied, the mean HbCO t_1/2 was 14.3 ± 1.6 min for baseline ventilation (RR · VT), 9.5 ± 0.9 min when RR was doubled (2 · RR), 8.0 ± 0.5 min when VT was doubled (2 · VT), 6.2 ± 0.5 min when RR was quadrupled (4 · RR), and 5.2 ± 0.5 min when VT was quadrupled (4 · VT). The HbCO half-lives for each ventilatory pattern are significantly different from the HbCO half-lives for all other ventilatory patterns (p < 0.05).

View larger version (72K): [in this window] [in a new window]   Figure 1.   Mean HbCO t1/2 for various ventilatory patterns. The HbCO t1/2 for each ventilatory pattern is significantly different from the t1/2 of all other ventilatory patterns (p < 0.05). Vertical bars indicate 1 standard deviation (SD). See Table 1 for description of ventilatory patterns.

A can be estimated for each level of E by assuming a fixed deadspace (Vds) equal to 35% of baseline VT (unpublished data for normal prone sheep). Figure 2 demonstrates the relationship between HbCO t_1/2 and A, described by an exponential function: HbCO t_1/2 =  +  exp(A), where:  = 5.20 (± 0.33),  = 21.2 (± 1.84), and  = 0.279 (± 0.031). This exponential function fit the data better (r² = 0.961) than an exponential function with an asymptote at zero (r² = 0.860) or a quadratic polynomial (r² = 0.925).

View larger version (10K): [in this window] [in a new window]   Figure 2.   HbCO t1/2 versus alveolar ventilation ( A). A is calculated using measured E and an estimated fixed deadspace fraction equal to 35% of baseline VT. HbCO half-lives for trials RR · VT, 2 · RR, 2 · VT, 4 · RR, and 4 · VT for all sheep are plotted. The fitted curve is represented by HbCO t1/2 =  +  exp( A), where  = 5.20 (± 0.33),  = 21.2 (± 1.84), and  = 0.279 (± 0.031). See Table 1 for a description of ventilatory patterns.

DO₂ was increased during treatment with all patterns of hyperventilation compared with baseline ventilation (Figure 3). The difference in DO₂ reached significance for 2 · VT, 4 · RR, and 4 · VT groups compared with RR · VT at 10 and 15 min.

View larger version (18K): [in this window] [in a new window]   Figure 3.   Oxygen delivery during treatment with various levels of isocapnic ventilation. Differences are noted when p < 0.05: *RR · VT significantly different from 4 · VT,  RR · VT significantly different from 4 · RR, # RR · VT significantly different from 2 · VT. Open triangles, RR · VT; open circles, 2 · RR; open squares, 2 · VT; open diamonds, 4 · RR; open inverted triangles, 4 · VT. Vertical bars indicate 1 standard deviation (SD). See Table 1 for a description of ventilatory patterns.

Hemodynamic and gas exchange parameters for all five patterns of ventilation averaged over the duration of each treatment period in all five animals are listed in Table 2. Mean Pa, mean Ppa, Pcw, CI, and arterial pH were not different (p > 0.05) across all ventilation patterns. Arterial PCO₂ was less for the 2 · VT, 4 · RR, and 4 · VT groups than for the RR · VT group (p < 0.05). The few differences in HR and arterial PO₂ between ventilatory patterns that reached significance are noted in Table 2.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Our study adds to the findings of Fisher and colleagues and Takeuchi and colleagues, that isocapnic hyperventilation increases the rate of CO elimination. Our study differs from previous work in four areas. (1) We used a multiple poisoning model, with each animal subject exposed to all treatment permutations in varied order, greatly enhancing the statistical power of the study. The multiple poisoning model developed in our laboratory has been shown to be hemodynamically stable and yields highly reproducible HbCO t_1/2 for up to seven repeat poisonings (12, 13); (2) we studied the effects of isocapnic hyperventilation at clinically relevant HbCO levels. Takeuchi and coworkers used low HbCO levels in human volunteers, with a peak [HbCO] of 12%. Fisher and coworkers studied dogs with severe CO poisoning, but each treatment was performed at different HbCO levels, the order of the treatments was not varied, and isocapnic hyperventilation was performed on each animal at the lowest HbCO levels (mean [HbCO] approximately 11%). While HbCO t_1/2 is relatively independent of HbCO levels, other clinically relevant variables may depend heavily on HbCO levels. For example, DO₂ measurements are highly related to HbCO levels. DO₂ will be higher when HbCO is lower, particularly under high FI_O2 conditions, because heme sites not occupied by CO would be occupied by O₂. In addition, tissue oxygen debt and acidosis, CO₂ production, myocardial contractility, and the rate of venous blood return may also vary significantly with HbCO levels; (3) we measured DO₂ during all treatment trials, whereas previous studies had not; and (4) we studied the effects of RR and VT increases on HbCO t_1/2 and DO₂ by varying RR and VT independently over a 4-fold range.

Isocapnic hyperventilation increased CO elimination in a dose-dependent manner, with a 4-fold increase in VT resulting in a 2.75-fold increase in the CO elimination rate (Figure 1). This finding is consistent with the findings of increased CO elimination with isocapnic hyperventilation by Fisher and colleagues in dogs (2.33-fold) and by Takeuchi and colleagues in human volunteers (2.52-fold), using comparable levels of hyperventilation. A similar relationship between hyperventilation and CO elimination is predicted by the CFK model of CO kinetics described by Coburn, Forster, and Kane (7).

Ventilatory pattern affects CO elimination. A higher VT and lower RR results in faster CO elimination than a higher RR and lower VT at the same overall E (Figure 1). Presumably, high VT ventilation results in a lower deadspace fraction (Vds/VT) and higher A than low VT ventilation. The relationship between HbCO t_1/2 and A (estimated) suggests an upper limit of effective ventilation because HbCO t_1/2 does not change significantly when A is increased from 15 to 20 L/min (Figure 2). The relationship between CO elimination and A in our study, diminishing effect at higher A, is consistent with the CFK mathematical model, the findings of Takeuchi and coworkers in humans, and a diffusion limitation to CO transport in the lung.

DO₂ is likely to be at least as relevant an end point as HbCO t_1/2 because a therapy that results in a decreased DO₂ while increasing CO elimination could possibly result in greater morbidity than more conservative treatment. We observed a dose-dependent increase in DO₂ during isocapnic hyperventilation in our prone sheep model; for increases in E we demonstrate increases in DO₂ (Figure 3). Cardiac index was not attenuated by any of the patterns of hyperventilation in this study, even with VT as high as 60 ml · kg¹ (Table 2). HbCO levels fell faster during the hyperventilation trials, resulting in increased DO₂ compared with control.

Isocapnic hyperventilation did not cause overt changes in hemodynamics, gas exchange, or lung mechanics. Mean Pa did not decrease with hyperventilation. Likewise, Ppa and Pcw were not altered. Arterial PO₂ increased during three of the four hyperventilation patterns even under conditions of "high stretch ventilation" (VT = 60 ml · kg¹) and baseline PO₂ values were unchanged between trials, even after multiple exposures to high VT ventilation. Airway pressures and arterial PCO₂ returned to normal after each treatment trial. It should be noted that the lack of significant hemodynamic or pulmonary alterations during hyperventilation is demonstrated in prone sheep; the hemodynamic effects of positive-pressure hyperventilation in CO-poisoned supine humans remain unknown. The size of VT in this and a previous animal study (dog) of isocapnic hyperventilation for CO poisoning is far greater than what is currently recommended for patients with acute respiratory distress syndrome (ARDS) (14). However, we are not aware of any studies that demonstrate lung damage from large E or VT ventilation in normal lungs. In fact, an abstract demonstrated no lung damage in burn patients ventilated at 2- to 10-fold baseline E (15). In addition, sustained spontaneous isocapnic hyperventilation in normal human volunteers was well tolerated (6). Future studies of human victims of CO poisoning will be needed to address the effect of mechanically induced hyperventilation on lung function.

Although the 2 · VT, 4 · RR, and 4 · VT trials had significantly lower mean arterial PCO₂ than the baseline ventilation trial, the PCO₂ was not significantly different between any of the hyperventilation trials. The difference in PCO₂ between the highest and lowest E was only 5 mm Hg and not clinically significant. In this study, FI_CO2 was adjusted manually for each of the five levels of ventilation. It is possible that a device such as the one used by Fisher and coworkers could have been used to keep PCO₂ within a tighter range.

We used a sheep model of severe CO poisoning for several reasons. First, sheep have been used by multiple investigators in the past to study physiologic effects of CO and CO elimination under a variety of conditions, and the kinetics of CO association/dissociation from sheep Hb have been well studied (16). Second, because sheep Hb has a lower affinity (M) for CO (140-150) than most other species (~ 250) they lend themselves more readily to the study of multiple repeat poisoning (21); a similar study in dogs would have required nearly 24 h. Finally, previous studies in our laboratory have observed stable hemodynamic variables and gas exchange parameters in sheep with multiple poisonings (12, 13), findings confirmed in this study. Because the treatment effect of hyperventilation is normalized to tidal breathing of pure O₂ in each animal we believe the fractional improvement in CO elimination rates would apply across species. This appears to be the case because previous studies of hyperventilation in dogs and humans report similar results (1, 6).

Hyperventilation with exogenous CO₂ and O₂ may result in hypocapnia or hypercapnia if E and FI_CO2 are not well matched. Hypocapnia is a potentially dangerous complication resulting in lower tissue PO₂ due to further increase in the affinity of Hb for O₂ (in addition to the increased affinity caused by CO) as well as cerebral vasoconstriction (22). Hypercapnia, on the other hand, is associated with mild increases in cardiac output and systolic Pa in critically ill patients (23), increases in DO₂ in animals (24), and a right shift in the oxyhemoglobin dissociation curve, which would counter the CO-induced increase in Hb-O₂ affinity. Hyperventilation with excessive levels of CO₂ may be safer than hyperventilation without adequate levels of exogenous CO₂. Considering the risks of hypocapnia and the decreasing effect of hyperventilation for CO removal at very high E, it may be advisable in clinical practice to use a fixed FI_CO2 gas source such as Carbogen (5% CO₂ + 95% O₂) and an initial recommended E (perhaps 20 L/min). If isocapnic hyperventilation for CO poisoning is attempted in humans, whether a fixed or variable source of CO₂ is used, care must be taken to ensure that eucapnea is maintained. Because mild hypercapnia does not appear to be detrimental a fixed FI_CO2 gas source may be safer than a mechanism with varying FI_CO2 and may be more quickly and easily implemented.

While the increases in CO elimination rates during hyperbaric oxygen therapy (HBO) are incontrovertible, the clinical benefit of HBO remains uncertain. HBO (at 3 ATA [atmosphere absolute]) increases the rate of CO elimination 3.5-fold compared with NBO (8), but a prospective randomized sham study of HBO versus NBO demonstrated no neurological or survival benefit of HBO (25). In addition, the observation that HbCO levels at presentation are poorly correlated with clinical manifestations suggests that the rate of CO elimination is not the only factor associated with clinical effect (26). The lack of consistently observed clinical benefit of HBO may be due solely to delays in initiating therapy (30). Therefore hyperventilation therapy, which eliminates CO nearly as quickly as HBO and is more readily accessible, could have a significant clinical benefit.

In conclusion, we have demonstrated up to a 2.75-fold increase in CO elimination and increases in DO₂ with isocapnic ventilation in an animal model of severe CO poisoning. These findings, the findings of others in animals and human volunteers, and the potentially universal access of isocapnic hyperventilation warrant further study of this therapy in CO-poisoned patients.