ARDS is among the most devastating postoperative complications and is associated with significant mortality.5, 25
The main results of this study are as follows: 1) In a general surgical population presenting to the operating room without ARDS, the overall incidence of new onset postoperative ARDS was approximately 0.2%; 2) The risk of ARDS was extremely low in ASA Class 1-2 patients; 3) Patients with ASA Class 3-5, emergency surgery, renal failure, COPD, and multiple anesthetics are at an increased risk of ARDS; 4) After controlling for patient comorbidities and risk factors, intraoperative ΔP, increased FiO2, volume of crystalloid, and transfusion are associated with the development of ARDS; 5) Development of ARDS greatly increases patient mortality, regardless of preoperative comorbidities.
Considerable research has helped ascertain preoperative risk factors for postoperative respiratory complications.26-30
Although there is some literature investigating the development of ARDS postoperatively, most of this work has focused on high-risk elective surgeries such as cardiac and thoracic interventions.3-7, 31
This study focused on a relatively common surgical population, a group of patients seen in hospitals throughout the world with a low incidence of ARDS that has precluded prospective study.
Preoperative univariate associations with the development of ARDS in our cohort were similar to prior studies, reflecting a greater number of comorbidities among the patients who developed ARDS.5, 7
In multivariate analysis, only ASA status, emergent surgical procedure, renal failure, COPD, and number of anesthetics were significant predictors of future ARDS. A history of being an active smoker and ethanol abuser did not meet statistical significance for predicting ARDS, although they have been associated with ARDS risk in other populations.7
This may be because such conditions are common in the surgical populations which prior have focused upon. The great importance of ASA status in predicting postoperative ARDS suggests that the actual comorbidities are less important, except for renal failure and COPD, than the preoperative medical optimization of comorbidities. From this data, ASA 1 and 2 patients are at low risk of the development of ARDS and patients undergoing emergency surgery may present a unique cohort deserving special attention.
The data from this study suggests asthmatic patients may be protected from ARDS. While a type I error is possible given the low incidence of asthma in our population, several asthma treatments have been suggested to be potentially effective in ARDS, despite sparse data.32-36
Prior studies have focused on the administration of agents after the onset of ARDS, whereas the patients in our study would have received treatment prophylactically. Although beyond the scope of the present investigation, this is an issue that we feel is worthy of future study.
After matching, it appears that intraoperative ventilator management may impact the development of postoperative ARDS. The finding that exposure to higher ΔP appears to increase the odds of ARDS development is consistent with the findings by other investigators in higher risk populations demonstrating that increased PIPs were similarly predictive of ARDS.5, 7
However, our data suggest that exposure to elevated PIP with a lack of PEEP is a component of the development of ARDS. This may be because such ventilator settings introduce barotrauma and atelectrauma, potentially compounding lung injury. The fact increased pressures but not increased Vt were predictive of future ARDS is interesting. Much of the work in the critical care community in preventing ARDS mortality has focused on the use of low Vt ventilation.2, 37-41
Previous perioperative literature suggests both pressure and volume may be important in preventing ARDS. Increased pressures appear to be consistent predictors of future ARDS,4-7, 31
while a protocol that focused on low Vt, and consequently, lower plateau pressures has shown a specific association with a reduction in the incidence of ARDS in thoracic surgery patients.8
Our current study appears to support that increased ΔP with low PEEP is associated with the development of ARDS to a greater degree than elevated Vt or PIP alone.
Our analysis also showed intraoperative fluid and transfusion of blood products to be significantly associated with the onset of ARDS. Transfusion related acute lung injury has been described in the literature.42
The use of blood products in our population may be an indicator of more aggressive resuscitation, and as such, this would support works that showed that increased volume resuscitation was predictive of future ARDS.5, 8, 9, 43
Given the observational nature of the current study, no recommendations can be made about the use of blood products or volume resuscitation intraoperatively for the prevention of ARDS, but this area certainly warrants further investigation.
ARDS is known to be associated with approximately 30 to 40% mortality at 90 days.2, 25, 44
In our previous work, patients that underwent an anesthetic at our institution with a preoperative diagnosis of ARDS had a 32% 90 day mortality, compared to 19% mortality for those who underwent an anesthetic with a P/F ratio < 300 without meeting criteria for ARDS.11
Fernandez-Perez found a 60-day mortality of 27% compared to 1% for patients that did not have pulmonary complications.5
In this study, after matching, we have shown 27% mortality at 90 days for those patients that develop ARDS after their anesthetics, compared to 12% for those that do not. The development of ARDS in this cohort is associated with an increase in mortality and is a worthwhile target for efforts at prevention.
This study has several limitations. First, the data were collected as part of routine clinical care and were not subject to the validation processes used in prospective trials. Although data is typically entered through a predefined selection process for each variable, there was no formal training on the definitions for each variable (appendix 1
). In addition, free text is allowed in all fields and was to interpretation by the research team. Models from such data have become common in the literature and have correlated with models based on prospectively collected data by dedicated research staff.13, 45
Additionally, items that may be associated with the development of ARDS that are not provided with specific coded entry boxes such as sepsis and aspiration were not included in the model. Next, the data are from a single, large, tertiary care center, collected over several years. The patient population may have changed over time and may not represent the typical patient seen in other locations. The database used to determine if patients had ARDS required mechanical ventilation and hence may underestimate the true frequency of ARDS. The etiology of the ARDS, whether it was primary or secondary to another insult was not determined and may itself be predictive of outcome. Additionally, we cannot be completely confident the elevated ΔP and increased FiO2 requirements were not signs of ARDS that had developed since the last datapoint was obtained preoperatively. Intraoperatively, there was not collection of plateau pressures and hence analysis was limited to PIP. While PIP has been used in several intraoperative ARDS manuscripts, it is limited and can be considerably higher than plateau pressures based on a variety of factors. Finally, the mortality data is based on an internal death registry and may not capture potential mortality of patients that were discharged to another long-term facility for ongoing care.
Due to the observational nature of this study, we did not have specific protocols for the intraoperative and postoperative management of the patient population other than routine clinical care. There was no mandated ventilator or fluid protocol at any point during the care of the patients involved in this study, including the postoperative period. However, despite this, there were statistically significant predictors of future ARDS determined both preoperatively and intraoperatively. Furthermore, in observational studies using clinical databases, the problem of model overfitting can be particularly vexing. We used several different bootstrapping techniques to reduce the variance in our models and select robust predictors, thereby minimizing the risk of type I statistical error. However, no amount of statistical testing can replace prospective validation of the predictive model and further study of interventions to mitigate risk. Ideally, the best evidence supporting the role of high pressure ventilation and development of ARDS in the surgical population would be a randomized controlled trial. Using the preoperative predictors described, one could develop criteria for enrollment in such a trial. However, careful consideration must be given to the design and ethical implications of such a study.
Despite the limitations, this investigation provides new evidence for a poorly studied population. We again demonstrated the high attributable mortality associated with the diagnosis of postoperative ARDS. We documented an exceptionally low incidence of postoperative ARDS in a general surgical population, particularly in the ASA 1 and 2 patients. Postoperative ARDS was predicted with high reliability using a model based on preoperative ASA, emergent surgical case, COPD, renal failure, and multiple surgeries. After matching, additional intraoperative risk factors for the development of ARDS(ΔP, increased FiO2, fluid administration, and transfusion) were identified, potentially offering clinicians opportunities to reduce the risk of postoperative ARDS. However, further investigation is still required before causation can be firmly established.