Gastric Fluid Volume

In this issue, Schwartz et al. [1] allude to the link between the risk of pulmonary aspiration and residual gastric fluid volume (GFV) at the time of induction of anesthesia. It has been nearly 25 yr since Roberts and Shirley [2] published their classic article in the pages of this Journal, in which they first advocated using an "arbitrarily defined" surrogate, specifically GFV >0.4 mL/kg, to be indicative of an increased risk of pulmonary aspiration. Since publication, these criteria have been widely accepted as indicating an increase in risk of pulmonary aspiration of gastric contents. Perhaps the most appropriate response to the celebration of the 25th anniversary of the publication of this highly influential article would be to finally lay to rest the myth it created. Since publication, the practice of medicine has been undergoing a paradigm shift; decision making based on experience and pathophysiologic rationale is giving way to reliance on evidence-based medicine [3]. The MacMaster University Evidence-Based Medicine Working Group's publication in Journal of the American Medical Association of the ongoing series titled "Users' Guide to the Medical Literature" has further focused attention on the careful evaluation of the medical literature [4]. Specifically, each article in the series asks the following questions: What is the validity of the data, what are the results, and what is the relevance of the results to care of patients? The validity of the data relates to the study methodology, the results speak to the size of the treatment effect or strength of association between exposure and the outcome of interest, and relevance to patients corresponds to the magnitude of the effect and impact on practice. Using an evidence-based approach, let's evaluate the evidence that supports the use of GFV as a surrogate for risk of pulmonary aspiration. What type of evidence would persuasively link a suspected risk factor or exposure to an untoward outcome such as the development of pulmonary aspiration of gastric contents [5]? When practical and feasible from both an ethical and financial perspective, the strongest evidence linking cause and effect comes from a randomized, controlled trial (RCT). The strength of the RCT arises from the process of randomization, whereby all known and unknown risk factors are distributed with equal probability among the experimental groups, and the process of blinding, which prevents bias from influencing the assessment of outcome during the conduct of the trial [6,7]. Although the RCT is the ideal design to demonstrate potential harm, conducting a RCT is frequently not feasible. When this is the case, an observational experiment, such as a cohort study, provides the next best alternative to the RCT. In a cohort study, patients with or without exposure to a putative risk factor are observed for a period of time to determine whether those with the exposure are at an increased risk of developing the outcome of interest. However, because randomization is not involved in the selection process, one cannot presume baseline equivalence of the groups. Even after statistical adjustments for confounding variables, equivalence is not assured. Furthermore, blinding of participants and study personnel may be impossible and can thereby introduce bias into outcome assessment. The results from a cohort study allow one to estimate the prevalence and the relative risk of developing the outcome of interest given the presence or absence of exposure. When the time from exposure until the outcome of interest is prolonged or the incidence of the problem is relatively infrequent, a case-control study, although a weaker form of evidence, will likely prove more efficient and less costly than a cohort study [8]. When conducting a case-control study, the investigator identifies cases by finding subjects who already have the outcome of interest and selects controls from subjects without the outcome. Using this method, one looks backward in time to establish patient baseline characteristics and potential risk factors. Numerous problems and biases can affect the outcome of case-control studies, including selection bias for cases and controls, recall bias, interviewer bias, and inadequate recording of data in preexisting medical records. Case-control studies cannot prove cause and effect, but rather, may suggest hypotheses. Despite the many advances in medicine, evidence gained from experimental or observational study designs is not always available for every clinical scenario. In this situation, one is forced to rely on case reports, case series, personal experience, pathophysiologic rationale, and expert opinion. When interpreting nonexperimental reports, it is imperative to recognize the limitations of reliance on these sources. What types of results prove most persuasive? First, the temporal relationship between the exposure and the outcome should be in the appropriate order. Second, there should be a dose-response gradient among levels of exposure and the likelihood of an adverse outcome. Finally, the primary end points should be clinically meaningful; i.e., the end point must matter to the patient. The fundamental dimensions that matter to patients include improved survival, diminished morbidity, or improved quality of life. As previously mentioned, GFV >0.4 mL/kg at the time of induction of anesthesia has been used as a surrogate marker of risk of pulmonary aspiration. A surrogate is a substitute for a clinically meaningful end point. Surrogates are useful only if they fully capture the entire effect of the pathway between the disease and clinical outcome [9,10]. To do this, the surrogate must be the sole mechanism by which the disease affects outcome. Surrogates may fail to link the disease to the clinical outcome under the following conditions: 1) the surrogate is correlated with the disease but is not part of the pathway that produces the clinical outcome; 2) other pathways or mechanisms are operative (altering the surrogate will not influence or only partially influence outcome); 3) the intervention affects other pathways but not the surrogate (measurements that track the surrogate will fail to correlate with outcome); or 4) the intervention affects the surrogate and other pathways but has its own independent negative affects on outcome [9]. How does GFV hold up as a surrogate? Unfortunately, there is no evidence from a RCT, cohort study, or case-control study that demonstrates a linkage between GFV and an increase in risk of pulmonary aspiration. There are no trials comparing outcome for patients with minimal GFV with those with greater GFV, and, just as importantly, there is no evidence to support the existence of a dose-response relationship between the quantity of GFV and the risk for pulmonary aspiration. The reason for the paucity of data is easy to understand; until noninvasive measurement of GFV before, during, and after surgery is possible, demonstrating a relationship between GFV and risk of pulmonary aspiration is unlikely, if not impossible. At the time of anesthetic induction, GFV is actually quite variable in normal people. In healthy children, the median GFV is approximately 0.3 mL/kg, the mean is approximately 0.4 mL/kg, and the upper limit extends to approximately 4.5 mL/kg [11]. Using the values for GFV proposed by Roberts and Shirley [2], 30%-60% of patients would be at risk. However, aspiration pneumonia is a rare event [12]: the incidence has been estimated at approximately 1 in every 3,000 patients [13]. Patients with comorbid conditions and those undergoing emergency surgery were at higher risk of experiencing an aspiration event. Symptoms from aspiration developed after approximately one third of events; intensive care unit admission was necessary for approximately 1 in every 12,000 anesthetized patients; and mortality as a result of pulmonary aspiration occurred in approximately 1 in every 72,000 anesthetized patients [13]. Because GFVs >0.4 mL/kg are common and aspiration pneumonia is not, it is possible to conclude that other pathways or mechanisms that are not captured by the measurement of GFV are important. Logically, interventions that focus only on reducing GFV or altering gastric pH will not necessarily influence the incidence or severity of aspiration pneumonia unless they also affect the other causative risk factors and mechanisms. A dose-response gradient for GFV and risk of pulmonary aspiration has not been established, and, as a surrogate end point, it has failed to prove its relevancy to outcomes that matter to patients. Because of the rarity of clinical pulmonary aspiration and the difficulties in measuring GFV before the induction of anesthesia, conducting either a RCT or a cohort study is unlikely to prove feasible. Well performed case-control studies and predictive modeling will probably be the most fruitful designs to help to establish and to allow estimation of the magnitude of true risk factors for the development of pulmonary aspiration of gastric contents [14]. It is time to put an end to using the criteria of Roberts and Shirley [2], a surrogate measure with no proven clinical basis, and to put an end to discussions of risk based on studies whose designs fail to meet basic criteria for adequate evidence. The discussion should be shifted from a focus on the GFV at the time of induction of anesthesia to the patient's characteristics and comorbid conditions, as well as to the anesthetic practices that place patients at risk of pulmonary aspiration of gastric and intestinal contents.