INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Gastric intramucosal pH (pHi) is considered to be an indirect estimation of the mesenteric perfusion and oxygenation (1). Gastric tonometry measures the partial pressure of carbon dioxide (CO₂) in the interstitium of gastric or intestinal mucosa, and pHi can be calculated from these data (4). Gastric mucosa acidosis is associated with a poor prognosis and increased mortality in patients with septic or anaphylactic shock, gastrointestinal bleeding, and other severe medical conditions (5- 7). Recent reports suggest that more profound changes in pHi might occur in patients who are later unsuccessfully weaned from the ventilator compared with those with weaning success (8, 9). Mohsenifar and coworkers (8) found significant differences only 20 min into the pressure support weaning period but not during mechanical ventilation. In addition, they estimated pHi from intraluminal pH measurements, which may be inaccurate in patients receiving drugs that alter intraluminal pH (e.g., histamine-2-receptor blockers). In a second study, Bouachour and coworkers (9) found patients with weaning failure to have lower pHi before the weaning period while the patients were still on ventilatory support. In this study tonometry with an intragastric latex balloon was employed circumventing some of the problems associated with direct measurements of pHi (9). Both studies looked at pHi either before mechanical ventilation was stopped or very early during weaning in order to employ measurements of pHi as a predictor of weaning outcome. However, no study so far has looked at pHi measurements during weaning, especially at the changes after the ventilatory support has been withdrawn to verify that the differences observed are attributable to the weaning process.

We therefore measured pHi in patients on mechanical ventilation and during the end of the weaning period and compared values and changes between patients with weaning success and failure. In addition, we assessed the predictive value of gastric tonometry obtained during ventilatory support for the weaning outcome.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This prospective study was conducted from December 1, 1997 to June 30, 1998 in the Hospital Clinic, an 850-bed university tertiary care hospital. All intubated and mechanically ventilated patients ( 3 d) of the respiratory intensive care unit (RICU) were eligible for the study when the following criteria were fulfilled: improvement or resolution of the underlying causes of acute respiratory failure; temperature < 38° C; hemoglobin  9 g/dl; absence of vasoactive drugs; Glasgow coma scale  13; Pa_O2 > 60 mm Hg with fraction inspired oxygen (FI_O2)  0.4 and positive end-expiratory pressure (PEEP)  5 cm H₂O (10). Exclusion criteria for this study were: abdominal surgery; gastrointestinal bleeding; and clinically documented gastroesophageal reflux. The study was approved by the ethical committee of our institution and conducted in accordance with its guidelines. In all cases informed consent was obtained from the next of kin.

Data Obtained and Study Design

Demographic characteristics (age, gender, history) and clinical data (e.g., body temperature, mean arterial blood pressure, and general laboratory parameters) were recorded on the day of the study. The study protocol consisted of three phases: mechanical ventilation, weaning period, and postextubation period.

Mechanical ventilation. All eligible patients were on assist-control ventilatory support for at least 6 h before the planned weaning trial. The tonometry catheter (NGS catheter; Tonometrics, Inc., Worcester, MA) was placed and the correct position was confirmed radiographically. One measurement of tonometry and arterial blood gases was performed in all patients 30 min before the weaning trial. The pHi and functional indices during mechanical ventilation were calculated using these measurements only.

Weaning period. At the beginning of the weaning trial, mechanical ventilation was discontinued and patients were breathing spontaneously for 3 min without supplemental oxygen. To proceed with the weaning trial patients had to fulfill the following conditions: respiratory rate (f) < 35 breaths/min; tidal volume (VT) > 5 ml/kg body weight; and maximal negative inspiratory pressure (PI_max) < 20 cm H₂O. The f was averaged over 1 min, VT was calculated from minute ventilation (E); PL_max was measured with a pressure transducer (CP-100 Pulmonary Monitor; BICORE Monitoring Systems, Irvine, CA) attached to the end of the endotracheal tube, and the most negative value of three trials was used. Patients who did not fulfill these criteria were reevaluated on a daily basis (11). No patient needed more than two weaning trials for liberation from the ventilator. The f/VT ratio reported in this study was calculated from measurements obtained at this time only (12).

Included patients were breathing spontaneously through a T-piece supported by supplemental oxygen and gas humidification for the remaining period of the weaning trial. Weaning failure was defined as one or more of the following criteria: (1) arterial oxygen saturation (Sa_O2) < 90% (< 80% in patients with chronic obstructive pulmonary disease [COPD]); (2) f > 35 breaths/min; (3) systolic blood pressure < 80 mm Hg or > 200 mm Hg (± 20% of the value in mechanically ventilated patients with documented hypotension or hypertension); (4) heart rate > 140 beats/min (± 20% of the value in mechanically ventilated patients with previous tachycardia); or (5) clinical signs of spontaneous breathing intolerance (diaphoresis; excessive use of accessory respiratory muscles or thoracic-abdominal discoordination; decrease of consciousness; and agitation) (10).

Arterial blood gases and gastric tonometry were performed at fixed time points of the weaning trial (30 and 60 min). Arterial blood gases could be obtained from all patients during the weaning trial, because no patients failed before 30 min of weaning. For patients with weaning failure, these measurements were used for the calculation of pHi and functional indices. In patients with weaning success, or extubation success, measurement of tonometry and arterial blood gases at 60 min was used for analysis. Results of arterial blood gas analysis but not of gastric tonometry were available to the physician responsible for the evaluation of the weaning trial.

Postextubation period. Extubation was performed after 60 min of T-piece trial in the absence of weaning failure criteria. Patients were followed for at least 24 h and patients who needed ventilatory support were classified as extubation failure. The possible cause of extubation failure was noted and patients were analyzed separately.

Regional gastric intramucosal PCO₂ (Pr_CO2) was measured with a gastrointestinal pump tonometer (Tonocap; Tonometrics, Inc., Worcester, MA) (13). Blood samples of arterial blood gases were drawn via an intra-arterial line immersed in ice and processed within 5 min using a blood gas analyzer (IL-1306; Instrumentation Laboratories, Milan, Italy).

Calculation of Variables

Clinical data were used to calculate the Acute Physiology and Chronic Health Evaluation II score (APACHE II) (14) and the Pa_O2/FI_O2 ratio was computed for corresponding time points. The pHi was calculated from corresponding measures of HCO₃ and Pr_CO2 using a modified Henderson-Hasselbach equation {[pHi = (6.1 + log₁₀ arterial HCO₃)/s × Pr_CO2] with a solubility coefficient (s) for CO₂ of 0.03 mmol/L/mm Hg}. Two additional variables (arterial pH [pHa]  pHi) and (Pr_CO2  Pa_CO2), were computed during mechanical ventilation and the weaning period.

Statistical Analysis

The primary objective of this study was to compare changes in Pr_CO2 between patients with and without successful weaning. The predictive power for weaning outcome of gastric tonometry data during mechanical ventilation was compared with the f/VT ratio obtained early during the weaning period.

Means were compared between two groups by unpaired Student's t test. Measurements of tonometry were compared between the two groups and over time (mechanical ventilation and weaning period) with a repeated-measures analysis of variance (ANOVA). For comparison of changes between groups of patients, the interaction term of the ANOVA is reported. To identify the optimal cutoff value of tonometry measurements [Pr_CO2, pHi (Pr_CO2  Pa_CO2), and (pHa  pHi)] between patients with weaning success and failure, receiver operating characteristics (ROC) curve analyses were employed. We analyzed f/VT ratio (cutoff  105) pairwise together with the categorized tonometry measurements in a conditional stepwise forward logistic regression model, to test whether the predictive power for weaning outcome could be improved (15). Results of this analysis are reported with the adjusted odds ratio (OR) and 95% confidence interval (95% CI). All data, except ROC curve analyses (MedCalc 4.2 for Windows 95), were processed with SPSS version 8.01 on a Windows 95 operating system. The level of significance was set at 5% (all two-tailed) and data are reported as counts or mean ± standard deviation.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

During the study period, 169 patients were admitted to our unit and 87 of 169 (52%) were intubated. Thirty-five patients died (35 of 87, 40%) and eight of 87 were weaned with tracheostomy (9%). Of the 44 survivors without tracheostomy, 29 of 44 patients were intubated and ventilated for 3 or more days (66%). Exclusion criteria applied to five of 29 patients (17%). A total of 24 patients and 31 weaning trials were finally included in this study with 6.3 ± 4.3 d of mechanical ventilation before the weaning trial. Of those, 15 of 31 (48%) weaning trials were successful, but four of 15 patients (27%) were excluded because of extubation failure. The patients were on average 73 ± 13 yr old and the cause of acute respiratory failure was exacerbation of COPD in nine of 20 (45%), congestive heart failure in four (20%), pneumonia in two (10%), postoperative in two (10%), and others in three (15%). The mean f/VT ratio was 84 ± 22 breaths/min/L in patients with weaning success and 144 ± 50 breaths/min/L in those with weaning failure (p < 0.001). Table 1 summarizes the findings of arterial blood gases and gastric tonometry during mechanical ventilation and the weaning period for the trials with weaning success and failure.

Arterial Blood Gases

Arterial blood gases were comparable between the two subsets of the population studied (weaning success and failure) during mechanical ventilation. The patients who could not be weaned successfully tended to have a lower Pa_O2 and Pa_O2/FI_O2 ratio before the weaning trial, but this difference was not significant. No clinically significant differences were shown between groups for Pa_CO2 or pHa during mechanical ventilation. The observed differences in Pa_O2 and Pa_O2/FI_O2 ratio became significant during the weaning trial, possibly because arterial oxygenation (Sa_O2 < 90%) was included, at least indirectly, in the definition of weaning trial failure. Neither the interaction term for Pa_O2 nor for the Pa_O2/FI_O2 ratio was significant, indicating that the changes occurring from mechanical ventilation to weaning were comparable for patients with weaning failure and success. Arterial PCO₂ and pHa were not different between groups during weaning nor were the changes (interaction analysis).

Capnometric Recirculation Gas Tonometry (CRGT)

Neither measured (Pr_CO2) nor calculated parameters of gastric tonometry [pHi (Pr_CO2  Pa_CO2) and (pHa  pHi)] were significantly different between patients with weaning success and failure not during mechanical ventilation nor during weaning (Table 1). However, the interaction analysis of all four variables showed that the changes observed were associated with the weaning outcome. Whereas the Pr_CO2 decreased in patients with weaning success during the weaning trial in comparison to mechanical ventilation (on average from 61.5 ± 15.0 to 56.3 ± 16.7 mm Hg), it rose in patients with weaning failure (on average from 60.4 ± 15.0 to 67.4 ± 21.0 mm Hg, p = 0.046 interaction analysis). This was also true when the difference between Pr_CO2 and Pa_CO2 was analyzed (Table 1). This difference increased in patients with weaning failure (on average from 15.5 ± 9.8 to 17.4 ± 10.6 mm Hg), whereas patients with weaning success showed a decrease in the (Pr_CO2  Pa_CO2) (on average from 17.6 ± 14.9 to 8.5 ± 14.0 mm Hg, p = 0.042 interaction analysis). Corresponding variations were observed for pHi and the (pHa  pHi) (Table 1).

ROC for Weaning Outcome of Gastric Tonometry Parameters Measured During Mechanical Ventilation

The results of the ROC curve analyses are summarized in Table 2. The largest area under the curve was computed for the f/VT ratio measured immediately at inception of the weaning trial. The cutoff value of 105 breaths/min/L generated a sensitivity of 93% with a specificity of 75%. Among gastric tonometry parameters, the pHi had the largest area under the curve and the cutoff value 7.138 was 20% sensitive and 100% specific.

Multivariate Analyses of Weaning Outcome Prediction

None of the gastric tonometry parameters tested [Pr_CO2  45 mm Hg, pHi  7.138 (Pr_CO2  Pa_CO2)  18 mm Hg, and (pHa  pHi)  0.084] was selected by the logistic regression analysis to improve the predictive power of the f/ VT ratio  105 breaths/ min/L (adjusted OR: 42.0; 95% CI 3.8 to 469.1, p = 0.002).

Description of Patients with Extubation Failure

Ventilator support had to be reestablished within 24 h after extubation in four patients (three were reintubated, one received noninvasive ventilation). Three patients developed severe bronchospasm and worsened hypoxemia and one patient had to be intubated again because of upper airway obstruction. Gastric tissue hypercarbia or acidosis was present in three of four patients (75%) at the end of the weaning trial. The values were: Patient 1: Pr_CO2 73 mm Hg, pHi 7.16; Patient 2: Pr_CO2 81 mm Hg, pHi 7.21; Patient 3: Pr_CO2 43 mm Hg, pHi 7.49; and Patient 4: Pr_CO2 69 mm Hg, pHi 7.16. Neither average Pr_CO2 (extubation failure, 79.0 ± 31.2 versus extubation success, 56.3 ± 16.7 mm Hg, p = 0.060) nor pHi (extubation failure, 7.26 ± 0.16 versus extubation success, 7.35 ± 0.12, p = 0.207) during weaning was significantly different between the four patients with extubation failure and those who had been extubated successfully.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The main results of this study in patients with acute respiratory failure were as follows. Pr_CO2 increased significantly more during weaning in patients with weaning failure compared with those who could be weaned successfully. The Pr_CO2  Pa_CO2 decreased significantly more during weaning in patients with weaning success compared with those with weaning failure. A f/VT ratio  105 breaths/min/L measured during the first 3 min of the weaning trial was 93% sensitive and 75% specific for the prediction of weaning success in our population. These operational characteristics could not be improved by gastric tonometry measured during mechanical ventilation.

Principal Findings and Literature Review

Gastric intramucosal Pr_CO2 and consequently also pHi calculated from Pr_CO2 are altered during regional hypoperfusion states (16). Dysoxia as a result of decreased mesenteric oxygen supply leads to local tissue acidosis and increased intramucosal Pr_CO2 (1, 17). The high respiratory workload and the adrenergic response during weaning can induce intestinal mucosa hypoperfusion due to the commensurate increase of blood flow to the respiratory muscles (18, 19). This leads to an accumulation of CO₂ in the splanchnic region after ventilator support is withdrawn and patients have to breathe spontaneously during the weaning trial.

Our results confirm those of Mohsenifar and coworkers (8) who also found that Pr_CO2 increased significantly more during weaning in patients who failed the weaning trial (Table 3). However, the magnitude of the changes we observed when patients proceeded from mechanical ventilation to weaning was smaller in comparison to the previous study. Mohsenifar and coworkers (8) estimated intramural gastric pH by measurements of gastric juice acidity. Notwithstanding, this method is likely prone to errors induced by enteral feeding or medication altering the pH of the gastric content (20). In addition, to interpret the results of gastric juice samples for the estimation of Pr_CO2 correctly, the assumption has to be made that the PCO₂ in the lumen of the stomach and the surrounding gastric wall tissue is similar. Yet, this assumption might not always be valid, especially in patients with uneven gastric blood flow distribution (3).

Bouachour and coworkers (9) did not describe changes in Pr_CO2 when patients with and without successful weaning were compared. Yet, they observed a higher Pr_CO2 in patients with weaning failure compared with weaning success during assist-control mechanical ventilation and weaning (Table 3). This is, at least at first sight somewhat arbitrary, because respiratory muscle demand should be similar during mechanical ventilation for all patients. However, we suggest three possible explanations for this observation: First, because exclusively patients with COPD were included in the study (9), changes during the transition from mechanical ventilation to weaning may have been obscured. Patients with COPD undergo different levels of severity of alveolar hypoventilation during weaning to reestablish the habitual acid base status (21). It has been shown that pHi is not independent from Pa_CO2 levels so that changes in Pr_CO2 might not have become evident (22). Second, patients were on assist-control ventilation for 24 h before the weaning trial (9). One might therefore argue that the oxygen cost of breathing was already increased during assist-control ventilation in those patients who subsequently developed weaning failure during the T-piece trial. However, significant differences in Pr_CO2 between patients with weaning failure and success were also present at the beginning of the assist-control ventilation. Therefore, it has to be assumed that the mesenteric acidosis was caused not only by increased respiratory muscle oxygen demands but also by other conditions (e.g., increased sympathetic activity) (18). Third, Bouachour and coworkers (9) used gastric tonometry that requires a latex balloon filled with saline solution as CO₂ solvent in the stomach to measure Pr_CO2 (23). To ensure a stable equilibrium between the PCO₂ in mucosa and in the balloon, at least 2 h of contact are recommended and values obtained earlier have to be corrected mathematically. The short equilibration time allowed in their study (20 min) might be another reason no significant changes were seen when the patients proceeded from mechanical ventilation to weaning.

In accordance with the study by Mohsenifar and coworkers (8) we also found significant changes in the PCO₂ gap (Pr_CO2  Pa_CO2) in patients with weaning success and failure (Table 3). An increase in the PCO₂ gap indicates an accumulation of CO₂ in the splanchnic region that exceeds changes in Pa_CO2 and may therefore well be attributed to splanchnic dysoxia. A decrease in the PCO₂ gap, as was observed in patients with weaning success in our study, may indicate the successful clearance of mesenteric PCO₂. In the study by Bouachour and coworkers (9) the opposite was reported, although the differences were not significant (Table 3). This, however, indicates that in the majority of patients the Pa_CO2 increased to a greater extent during weaning than the Pr_CO2 resulting in a smaller PCO₂ gap in patients with weaning failure. This can also be explained by the changes in Pa_CO2 that occur during weaning in patients with COPD.

Critique of Methods

Our results indicate smaller absolute changes and differences in Pr_CO2 in patients with weaning failure and success than in the two previous studies. One might argue that this discordance is due to the CRGT methodology. A previous comparative study reported clinically significant differences for gastric intramucosal PCO₂ measured with air or saline tonometry (24). However, these results are conflicting and should affect only the absolute values of Pr_CO2 but not the comparison between patient groups (25). CRGT requires also an intragastric balloon with contact to the mucosa but uses gas instead of saline as a solvent. This ensures shorter equilibration times and mathematical corrections are not necessary (13, 26). CRGT therefore allows the assumption that also short-term changes of Pr_CO2, as they are expected during weaning, can be measured. The validity of both, absolute values and changes, has been proven for CRGT in vivo and in vitro using conventional tonometry catheters (13). We therefore suggest that the difference in Pr_CO2 between patients with weaning failure and success has been overestimated in the previous two studies. This in turn has consequences for the clinical applicability of gastric tonometry in terms of prediction of weaning outcome.

Prediction of Weaning Outcome

The f/VT ratio parameter was the strongest single predictor for weaning outcome in our study. The f/VT ratio had been first described by Tobin and coworkers (27) and was investigated prospectively by others thereafter (12, 28). A high sensitivity of 93% could be confirmed in our study using a threshold value of 105 breaths/min/L. By contrast, the high specificity in our study (75%) was only comparable to the prospective validation of Yang and Tobin (64%) whereas others reported only a moderate to poor specificity (12, 32). One principal defect of the f/VT ratio seems to be the highly variable false-positive rate between 11 and 64% (patients with f/VT ratio  105 who fail) (28, 29). A high false-positive rate is more likely in populations with a high percentage of patients with a f/VT ratio below the cutoff value, as was the case in the two previous studies alluded to testing the clinical value of Pr_CO2 for weaning outcome prediction (Table 3) (8, 9). In addition, Mohsenifar and coworkers (8) measured the f/VT ratio during pressure support rather than during a T-piece trial, which likely contributes to a high false-positive rate. The fact that both Mohsenifar and Bouachour favored pHi over the f/VT ratio for the prediction of weaning outcome therefore reflects rather the poor performance of the f/VT ratio in their study populations (8, 9). In our current study, only 63% of the patients had a f/VT ratio  105 and the operative characteristics were good and could not be improved by gastric tonometry.

Analysis of Extubation Failures and Limitations of the Study

We observed four patients with extubation failure and three showed gastric hypercarbia at the end of the weaning trial. This may indicate that gastric tonometry could be a useful tool to predict extubation failure. However, average values were not significantly different between patients with extubation success and those who failed, and a larger sample size may be needed to further investigate this issue.

A possible limitation of this study was the short duration of the weaning trial which did not allow us to analyze the predictive value of gastric tonometry data obtained during the weaning trial. However, even during weaning the differences in Pr_CO2 between patients with weaning success and failure were small and a good predictive power therefore was very unlikely. Our definition of weaning failure was slightly different from that of the two previous studies (8, 9), because we excluded patients with extubation failure. This should affect, however, only the comparison to the study of Mohsenifar and coworkers (8) since no extubation failures occurred in the study of Bouachour and associates (9).

In conclusion, we found that an increase in Pr_CO2 was associated with unsuccessful weaning from mechanical ventilation in a group of nonselected patients with acute respiratory failure. In addition, the widening of the Pr_CO2  Pa_CO2 difference after proceeding to weaning was more pronounced in patients with unsuccessful weaning outcome. Differences in Pr_CO2 absolute values between patients with successful and unsuccessful weaning were small even during the weaning trial. Finally, the f / VT ratio remained the best predictor of weaning outcome in this study and gastric tonometry could not replace or improve this bedside outcome variable.