A Description of Intraoperative Ventilator Management in Patients with Acute Lung Injury and the Use of Lung Protective Ventilation Strategies

• Anesthesiologists typically increase Fio²and positive end-expiratory pressure intraoperatively when patients become hypoxemic during anesthesia.

• For patients with a preoperative diagnosis of acute lung injury anesthesiologists still increased Fio², tolerated higher peak inspiratory pressures, and used lower positive end-expiratory pressure than guidelines recommend intraoperatively. Anesthesiologists do not appear to use low tidal volume ventilation intraoperatively.

RELATIVE hypoxia and decreasing Pao²/Fio²(P/F) ratios are common conditions during general anesthesia.1Because atelectasis is a common reason for relative hypoxemia in the surgical setting, tidal volumes of 10–12 ml/kg have been traditionally advocated.2Acute lung injury (ALI) is a clinical syndrome that is defined as the rapid onset of hypoxia with a P/F ratio ≤ 300 and bilateral pulmonary infiltrates in the absence of left atrial hypertension.3The overall in-hospital 90-day mortality of patients with ALI is more than 30%.4There are data suggesting ventilation strategies may influence the postoperative incidence of ALI after pneumonectomy, but to date, the incidence of ALI in hypoxic patients undergoing surgical anesthesia has not been described.5–7Intensive care unit (ICU) ventilator management of patients with ALI involves minimizing ventilator-associated lung injury by using lower tidal volumes (V^t) of 6 ml/kg predicted body weight (PBW) and limiting plateau pressures to less than 30 cm H²O combined with moderate positive end-expiratory pressure (PEEP) in a lung protective ventilation strategy (LPVS).8–9In addition, aggressive PEEP and recruitment maneuvers have been successfully used to improve oxygenation while reducing potential oxygen toxicity and alveolar shear stress, although they have not been found to provide mortality benefit.10–11 

In our previous work, we demonstrated that patients with a low intraoperative P/F ratio were typically managed using increased Fio²and peak airway pressures.12However, this study did not take into account the preoperative hypoxic status of the patient, the etiology of the low P/F ratio, the incidence of ALI in the population with a low intraoperative P/F ratio, or the mortality of the population. Therefore, we sought to determine the incidence of preexisting ALI in patients undergoing surgery and examine the intraoperative management of preoperatively hypoxemic patients using a separate database of patients with known ALI. Furthermore, we sought to determine if the management of patients with a low P/F ratio because of ALI was fundamentally different from those who had a low P/F ratio because of some other etiology and to examine the mortality rates between these two groups. Using a large, multiyear set of automatically collected intraoperative data, we examined preoperative arterial blood gas (ABG) values along with the intraoperative V^t, PEEP, Fio², and peak inspiratory pressures (PIP) that were used in the surgical setting. Specifically, we explored if there were differences in mechanical ventilation strategies in patients with known ALI compared with those of patients with hypoxia from other causes. We hypothesized that anesthesia providers would not routinely implement LPVS in hypoxemic patients or in patients with known ALI.

Institutional Review Board approval was obtained for this cohort study at The University of Michigan Medical Center in Ann Arbor, a large, tertiary care facility. All adult operations performed between January 1, 2005 and July 1, 2009 with at least one ABG value in the 24 h before proceeding to the operating room (OR) were examined. In the event multiple ABG values were obtained, the average values of all validated blood gases were used to generate an average P/F ratio. Patients with average P/F ratios less than 300 were included in the study.

Preoperative data were collected from routine clinical documentation that was entered into the institutional anesthesia information system (Centricity, General Electric Healthcare, Waukesha, WI). The record includes a structured preoperative history and physical examination allowing for coded entry and free text where required. Data obtained from the preoperative history included age, sex, American Society of Anesthesiologists classification, and height. From the height variable, PBW was calculated using the formula 50 + 2.3 (height [in]– 60) for men and 45.5 + 2.3 (height [in]– 60) for women.

The values for the intraoperative blood gases examined were manually entered by the anesthetic team into the structured electronic anesthesia information system (Centricity, General Electric Healthcare). Preoperative ABGs were obtained from an automated laboratory interface that connected to the anesthesia information system. Intraoperative physiologic and ventilator data were acquired using an automated, validated electronic interface from the anesthesia machine (Aisys, General Electric Healthcare) and physiologic monitors (Solar 9500; General Electric Healthcare). Fio², PIP, V^t, PEEP, oxyhemoglobin saturation (Spo²), and end-tidal carbon dioxide (Etco²) values from the time of incision to the end of anesthesia were obtained and analyzed for median values to eliminate spurious and isolated values. Preoperative ABG values were excluded if 0.21 > Fio²> 1.00. From the recorded intraoperative Pao²values and median Fio²the P/F ratio was calculated for each blood gas value. The cc/kg PBW was calculated from the median V^tand the PBW. The values for intraoperative Pao², Paco², and pH were analyzed if available. Values with 45 mmHg > Pao²> 600 mmHg, 20 mmHg > Paco²> 110 mmHg, 6.79 > pH > 7.99, 0.21 > Fio²> 1.00 or associated with a median 20 mmHg > EtCO²> 110 mmHg, P/F ratio > 600 were excluded because of the high probability of venous origin of the blood sample or erroneous data entry. Blood gases with abnormal characters were also removed from the dataset. Cases with patients younger than 18 yr, a documented height of less than 55 inches, or a documented height greater than 80 inches were also excluded. Case times were validated as having started and ended by electronically documented heart rate from electrocardiogram or electronically documented start, incision, and end times in the event electrocardiogram data were not available. Cases with negative or undocumented times were excluded. Cases from patients graded as American Society of Anesthesiologists classification 6 were also excluded because the ventilation strategy implemented may have been designed to favor perfusion to other organs.

To identify the subpopulation of hypoxemic patients who actually had ALI compared with those with other causes for hypoxia, patients in this dataset were merged with a dedicated preexisting research dataset of all patients on ventilators at the University of Michigan Medical Center who were screened for entry into ALI studies. In this research dataset, it was ultimately determined if a patient had ALI by analyzing the patient's ventilator status, ABG, chest x-ray, and clinical documentation. Patients were included in this dataset if they were on a ventilator, had bilateral infiltrates on chest x-ray as determined by a clinician, had a P/F ratio ≤ 300, and had minimal evidence of fluid overload. If identical onset and OR dates were present for a patient, the chest x-ray, ABG, and anesthesia start time were analyzed by one of the authors to determine if the criteria for ALI were met before or after administration of anesthesia. Patients were assigned as either having or not having ALI before anesthesia was administered. Hence, only patients who were receiving mechanical ventilation before anesthesia were considered to possibly have ALI.

To determine the percentage of patients receiving mechanical ventilation and the associated settings before their anesthetic was administered, a respiratory therapy random check database was merged with the dataset. The ventilator settings from this dataset were obtained within the 24 h before the patient's procedural start time. Patients without any ventilator settings (V^t, PIP, plateau pressure, or PEEP), but with a recorded Fio²were considered to be spontaneously ventilated. Comparisons were made between the preoperative and intraoperative ventilator setting from this dataset.

This combined dataset was then merged with an institutional death registry. Patients were categorized as being alive or dead at 90 days. In order to avoid the effect of multiple surgeries and anesthetics on patient outcome, patients who had more than one surgical procedure were excluded from the dataset for univariate analysis for mortality at 90 days.

Statistical analysis was performed using SPSS version 17 (SPSS Inc., Chicago, IL) and R version 2.12 (R Foundation for Statistical Computing, Vienna, Austria). Patients were divided into those with ALI and those without ALI. The mean value and SD for each group was calculated for cc/kg PBW, PIP, PEEP, Fio², Pao², Paco², pH, age, and American Society of Anesthesiologists classification. The groups were compared using either an independent sample Student t  test for parametric distributions or the Mann–Whitney U test for nonparametric distributions. Comparisons between preoperative and intraoperative ventilator settings were made using the paired Student t -test for parametric distributions, and the Wilcoxon signed-rank test for nonparametric distributions. Boolean analysis was completed using chi-square analysis. Statistically significant values were considered to have P  values ≤0.05 using two-tailed tests. Univariate analysis was completed on predictors of mortality at 90 days.

A total of 1,286 surgical cases with preoperative hypoxemia were identified and met inclusion criteria. Surgical cases that were excluded are shown in figure 1. Of the patients, 64% were male and the average age was 55.3 yr. Preoperative ventilation data were available on 730 cases (56.8%). Fifteen of the 730 cases (2.1%) were found not to be undergoing mechanical ventilation. Six of 155 cases in the ALI group were spontaneously ventilated at the time of the spot check, and 9 of 575 not in the ALI group were spontaneously ventilated at the time of the spot check. Table 1shows the demographic, preoperative, and average preoperative arterial blood gas data obtained for the cases included in the analysis. In an effort to determine if there was a statistical difference between the groups of patients with preoperative ventilation data compared with those without these data, table 2was generated. There were statistically different yet clinically insignificant values found for the weight, preoperative pH, and preoperative Paco². This is validated by the fact that the body mass index and preoperative P/F ratios are statistically insignificant, and the pH values differ by 0.01.

Of the 1,286 cases meeting criteria, 242 (19%, 95% CI 16.7–21.0) were found to have preoperative ALI. Table 3displays the average and SD of the ventilator parameters for each group of patients while in the OR. Boxplots were generated displaying the average PIP, PEEP, Fio², and V^tfor patients with and without ALI (fig. 2). There was a statistically significant difference in the ventilatory parameters between the two groups with the cc/kg PBW, PEEP, and PIP administered; however, these differences were clinically insignificant.

In order to determine if patients were indeed being managed differently in the OR than in the ICU, we compared the preoperative and intraoperative ventilator settings. The results of this comparison are shown in table 4. There were statistically significant differences in the Fio², PIP, Vt, and PEEP administered. To examine only the patients with ALI, preoperative and intraoperative ventilator settings were compared with patients with a preoperative diagnosis of ALI. Differences were found in the Fio², PIP, and PEEP applied. There were increased intraoperative tidal volumes in both total V^tand cc/kg PBW, but these did not meet statistical significance. The results of this analysis are shown in table 5. We then examined the preoperative management of patients who were receiving less than 7 ml/kg PBW preoperatively with known ALI. Thirty-four patients were identified with an average V^tof 5.92 ml/kg PBW preoperatively. Intraoperatively, these patients received 7.5 ml/kg PBW (P = 0.002) with increased PIP (25.52 cm H²O versus  28.7 cm H²O, P = 0.026).

Univariate analysis was then completed on the predictors of 90-day mortality on the reduced dataset that consisted of patients who underwent only one surgery. A total of 791 patients were included in this dataset, of whom 17% had a diagnosis of ALI before anesthetic was administered. The overall 90-day mortality for the population was 21%, and the ALI group mortality was 32% compared to 19% for the non-ALI population. Results of the univariate analysis are shown in table 4. Of note, significant predictors for mortality preoperatively included pH and P/F ratio, body mass index, American Society of Anesthesiologists status, and age. Intraoperative ventilatory predictors of mortality included PEEP and PIP.

We have previously demonstrated that patients who became hypoxic intraoperatively were typically managed using increased Fio²and increased PIP, and had a marginal increase in the amount of PEEP delivered from 2.86 cm H²O to 5.48 cm H²O.12In the most severely hypoxic patient population, this was accompanied by a clinically trivial yet statistically significant reduction in V^ton average from 9.05 to 8.64 ml/hg PBW. However, this previous study did not take into account the preoperative P/F ratio of patients or the preexisting incidence of ALI.

In this new series, approximately 20% of patients presenting to the OR with a preexisting P/F ratio ≤ 300 were identified as meeting criteria for a diagnosis of ALI. From the data collected, it appears the typical management of a preexisting hypoxic patient by the anesthesiologists is to use increased Fio², tolerate high PIP, and use lower amounts of PEEP than are prescribed by the ARDSnet (ARMA and ALVEOLI) trials regardless of the etiology of the hypoxia. The average Vt observed was 8.7 (2) cc/kg PBW with an average PEEP of approximately 5 cm H²O. It did appear that anesthesiologists limited PIP to less than 30 cm H²O on average. The Vt delivered in the OR was very much in line with the Vt provided in the ICU before the patient's procedure. These data suggest that most anesthesiologists do not implement strict ARDSnet LPVS when hypoxic patients come to the OR regardless if they have ALI or some other cause of hypoxia, and they tend to use preexisting ventilator settings that are similar to those found in the ICU. However, when looking at a subset receiving ≤ 76 ml/kg PBW tidal volumes preoperatively with known ALI, the volumes were increased intraoperatively. This finding is of interest because there is considerable evidence suggesting ALI outcomes are improved if patients are managed using LPVS while in the ICU.9,13Further investigation is required to determine whether this practice is more common in other medical centers, as well as whether the evidence for improved ALI outcomes from LPVS from the ICU generalizes to mechanical ventilation practices in the operating room.

In the ARDSnet ARMA trial, 861 patients were randomized to receive ventilation with either 6 ml/kg PBW or 12 ml/kg PBW of ventilation when a diagnosis of ALI was made. Overall the actual mortality difference was 9% in favor of the lower tidal volume group, despite the fact that oxygenation was better in the higher tidal volume group. In the ALVEOLI trial, patients with ALI received 6 ml/kg PBW ventilation with either an aggressive PEEP, “open lung” strategy or the conventional PEEP dictated by ARMA. This study did not show a mortality benefit, but it did show slightly reduced ventilator days. These two well-conducted studies provide evidence that increased oxygenation does not lead to reduced mortality in the ALI population. However, from our study, it appears anesthesiologists are favoring increased oxygenation over a proven modality in the treatment of ALI. This is shown by the increased intraoperative Pao²compared with the preoperative Pao².

There has been a concern in the anesthesia community about the potential downsides of using LPVS in the operating room because these ventilator settings potentially result in increased atelectasis and hence decreased oxygenation. It has been demonstrated that LPVS does not result in increased atelectasis when used in patients undergoing general anesthesia. Cai et al.  performed computed tomography scans on healthy volunteer patients before induction, after induction, and after mechanical ventilation with patients treated with either low V^Tventilation or normal V^T.2In this study there was no difference in the amount of atelectatic lung, regardless of the V^Tprovided, based on computed tomography.

There may be some concerns by anesthesiologists that LPVS would be harmful because of potentially increased cytokines from increased PEEP and potential hypercapnia in patients who do not have ALI. It has been shown that LPVS does not increase biologic markers of lung injury in normal lung. Wrigge et al.  have shown that LPVS or a high Vt/low PEEP strategy showed equivalent or minimal elevation in the amounts of pulmonary and systemic levels of inflammatory markers after thoracic, intraabdominal, or cardiac surgery in patients with a normal lung.14–15These data may support the decision of providers to avoid using LPVS in routine patients because it is equivocal and better oxygenation may be provided using non-LPVS ventilator settings. However, in patients with previously injured lung, there appeared to be an association with increased plasma cytokines after high Vt mechanical ventilation.16As the incidence or development of ALI in a particular patient may not be known before an OR course, it would seem reasonable for the anesthesiologist to consider the use of LPVS in the hypoxic patient if it was believed that there was a high likelihood of the patient having or developing ALI.

Some may contend that the short duration of high pressure/high volume ventilation endured during a surgery will not lead to worse outcome in the overall care of a critically ill patient. Licker et al.  described the intraoperative management and postoperative incidence of ALI in a cohort of 879 patients undergoing lung resection for cancer and found that postoperative ALI was diagnosed in 37 patients (4.2%) and acute respiratory distress syndrome in 17 patients (1.5%).5After multivariate analysis, it was concluded that the ventilatory hyperpressure index was an independent predictor of development of ALI with an odds ratio of 3.2. Fernandez-Perez et al.  described the occurrence of ALI after elective surgery as 1.8%.17In the multivariate analysis, increased PIP was a significant predictor of postoperative ALI with an odd ratio of 1.07. These data suggest that short intervals of high pressure do indeed have a small but demonstrable correlation with the development of ALI. In our data, we found the PIP to be associated with increased mortality. This finding may be simply a marker of increased illness, or it may be indicative of active injury to the lung. The current data do not provide an answer to this question.

This study has several limitations. First, the data were collected as part of the clinical care delivered and from a research study screening database. As a result, the data reflects the electronic anesthetic record, and no additional detail is available. There were no rigorous processes to validate the entry of data, although the use of automated collection of physiologic data has been accepted in many previous studies.18–20However, this methodology is devoid of the Hawthorne effect and provides insight into the actual modalities that are used by anesthesiologists for patients with preexisting low P/F ratios. Second, the data are from a single tertiary care center and may not serve as a representative sample of patients throughout the world. Next, because of the data resources available, some patients who were not ventilated preoperatively may indeed have ALI. Hence, it is possible there were patients in the non-ALI group who should have been included in the ALI registry and the incidence may indeed be even higher than what we identified. To address this concern, we attempted to examine a defined set of patients of which their preoperative ventilatory status was known from our respiratory therapy database. However, data were not available on all patients and it is possible that this population was different from the entire population studied. To determine if there was a difference between groups with and without preoperative ventilation data, we conducted five rounds of multiple imputations on missing values, understanding there are limitations on the utility of imputing values that are determined by clinicians. The results showed new statistically significant differences in preoperative absolute Vt, plateau pressure, FiO2, and intraoperative Paco²because of increased power. However, these imputed values were not substantially changed from the actual values reported in table 1, and there was no statistical difference in the Vt in cc/Kg PBW or P/F ratio, suggesting that the patients with complete data reasonably represent the preoperative ventilation and comorbidities of all patients. Next, our dataset focused on PIP rather than plateau pressures. Although these values differ in their nature, from the data presented, the trend is consistent in that increasing PIP correlated with increasing plateau pressures. Finally, there is the possibility of type-1 error in the analysis because we conducted more than 50 inferences with a significant P  value <0.05. We realize that we have a small sample size and the number of inferences performed may in fact lead to the possibility that some inferences will be significant.

In conclusion, the incidence of ALI in patients undergoing anesthesia with preoperative P/F ratios ≤ 300 in our institution is approximately 20%. These patients are managed in a similar format that does not use LPVS regardless of the etiology of their hypoxia, and the Vt settings appear to mirror the ventilator settings provided to patients in the ICU. Overall, it appears risk factors for mortality at 90 days include increased PIP and preexisting ALI. Further study is required to determine if the use of LPVS intraoperatively has a substantive impact in reducing mortality of hypoxic patients who undergo surgery both with and without ALI.