Circulation 75,648 • Volume 20, No. 1 • Spring 2005   Issue PDF

Report of APSF Workshop on Postoperative Long-Term Outcome

David M. Gaba, MD; Robert C. Morell, MD

Anesthesiologists have long been pioneers in seeking ways to improve the safety of patients. To date, most of our efforts have been directed toward reducing the likelihood of adverse events in the immediate perioperative period. Over the last few years several threads of information have been coalescing that suggest there is another patient safety issue involving adverse outcomes occurring long after the traditional perioperative period. These outcomes are not directly tied to a specific complication of the surgery and may be referred to as “long-term outcomes” of anesthesia and surgery.

The APSF has begun to take an active interest in this topic. In the APSF Newsletter (Fall 2003 and Spring 2004), a number of these issues were introduced, along with the hypothesis that the underlying biological mechanism to explain the occurrence of such long-term outcomes might have to do with inflammatory processes triggered in the perioperative period. As detailed in those articles there are a variety of reasons to think that mortality, and presumably morbidity, can be affected by perioperative events, and that inflammation could be a key element in such occurrences.

Because the threads were disparate, from different medical domains, and still uncertain, the APSF decided to convene a multidisciplinary experts’ workshop. Following is a brief report of some of the highlights from that workshop. Dr. David Gaba was selected as the principal investigator for the workshop because of his extensive research experience, expertise in patient safety, role as secretary of the APSF, and the fact that he does not conduct research in this particular area. Thirty experts attended the conference in Boston, MA, on September 21-22, 2004. The names of attendees and their biographies are available on the APSF website. The full program for the workshop is also available on the site. The program was divided into 4 sessions including epidemiology of long-term outcomes following anesthesia and surgery, 2 sessions on inflammatory processes and other underlying biological mechanisms, and a final session reviewing the issues discussed and the pitfalls of conducting outcomes research, and debating the best course of action for future research.

The goals for the conference were to:

  • Define the problem(s) of long-term outcomes after anesthesia and surgery
  • Estimate the scope and nature of the problem(s)
  • Assess the state of the science of the putative inflammatory mechanisms
  • Identify the most important research questions, and the key gaps between what is being studied and what needs to be studied
  • Develop an agenda for possible research
  • Determine if existing interventions require greater attention for evidence-based guidelines/practice parameters
  • Identify new interventions in need of study
  • Develop a plan for dissemination and follow-up.

Session 1: Epidemiology of and Risk Factors for Long-Term Outcomes

Robert Lagasse, MD, began by reviewing some of the history of research on short-term, adverse outcomes in anesthesiology, illuminated issues in studying long-term outcomes, and suggested that the advent of Anesthesia Information Management Systems (AIMS) could make it easier to link intraoperative events to both short- and long-term outcomes. Next, Shukri Khuri, MD, described the Veterans Health Administration’s National Surgical Quality Improvement Program (NSQIP). Dr. Khuri is a cardiac surgeon who has led the NSQIP project for many years. NSQIP tracks surgical outcomes (typically out to 30 days post-op) for a number of major surgical procedures, in a variety of surgical specialties, across all VA facilities that perform them. The NSQIP database depends on data entered by a specially trained nurse at each site, who is able to leverage the VA’s electronic medical record system. Data are collected on a large number of preoperative variables, on a few intraoperative variables (e.g., duration of surgery, surgical procedure), and then on a suite of postoperative outcome variables. The NSQIP database now has more than 1.2 million records and for selected patient groups has assessed mortality out to 10 years. The NSQIP program has expanded to the private sector, first in a set of studies and now in a joint program with the American College of Surgeons–ACS-NSQIP. Finally in this session, Terri Monk, MD, discussed the history of studies of long-term mortality following anesthesia and surgery in different patient populations. She provided an overview of intervention studies such as those using beta-blockade in the perioperative period. She also presented preliminary but provocative data from her original research examining perioperative cognitive dysfunction in patients having general anesthesia. Data on intraoperative vital signs and blinded bispectral index (BIS) data were recorded for these patients as were various postoperative outcomes, including mortality at 1 year. The data on mortality were then subjected to multiple regression modeling to determine which underlying preoperative and intraoperative factors correlated with death at 1 year. Not surprisingly, underlying medical problems were the major risk factors (odds ratio 16.1), as was time spent with a systolic blood pressure less than 80 mmHg (odds ratio 1.04/minute <80). A BIS value <40 was also identified as a significant risk factor. Dr. Monk also described results of a similar study performed on a larger number of patients in Sweden by Lennmarken et al. that also showed that BIS <45 was a statistically significant risk factor. Dr. Monk described all of these results as surprising, since there was no obvious mechanism to account for these findings. In addition, approximately one-half of the deaths in her study were due to cancer, and the correlation of time with hypotension or with BIS <40 to death due to cancer was even harder to explain.

Discussion of Session 1

Robert Stoelting, MD, APSF President, chaired the discussion of Session 1, at which the panel reviewed the epidemiology. There was particular focus on the data from Weldon and Monk’s study and from Lennmarken et al. because these results were unexpected. It was acknowledged that these data are preliminary and are strictly observational. Nonetheless, they are provocative data, as they suggest that some factor occurring during the brief period of the anesthetic may be linked to mortality remote from the perioperative time frame. There was spirited debate about these data and what they could mean if they are confirmed with further studies. One suggestion was that the occurrence of BIS <45 represents merely a marker for patients who are particularly vulnerable for some reason. Another possibility discussed was that these patients may have a higher adrenergic state and are subsequently treated with higher levels of hypnotics or volatile anesthetics. It was also suggested that the patients with the low BIS values might have a greater or more extended inflammatory response to anesthesia and surgery, although there are no studies yet that have investigated such a putative mechanism. It was widely agreed that these questions will need to be examined in larger studies that specifically investigate these issues.

There was additional discussion of other perioperative factors and treatments that have an impact on long-term outcomes. In particular, the data on perioperative beta-blockers are complex. While there have been multiple randomized trials, and there is a consensus for beta-blockade in patients with known cardiac disease having vascular surgery, whether this is beneficial for broader use is still open to considerable debate.

Sessions 2 and 3 – Inflammation

Potential biological mechanisms for such occurrences were the topics of Sessions 2 and 3. At the beginning of Session 2, Dr. Steffen Meiler proposed a set of hypotheses concerning the perioperative inflammatory/immune response as a potential biological link to long-term outcomes after anesthesia and surgery. Dr. Meiler proposed a “two-hit” model, which states that the inflammatory response to surgery may amplify pro-inflammatory cell mechanisms of certain disease states, such as coronary artery disease, hence contributing to disease acceleration and adverse postoperative events. The evidence that inflammatory processes are critical for the progression of atherosclerosis is undeniable. Similarly, inflammatory processes and infections are known to play a key role in cancer biology (e.g., hepatitis leading to hepatocellular carcinoma). The role of inflammation in the degenerative central nervous system diseases, such as Alzheimers, is more tentative, but a growing body of evidence definitely points in this direction.

Furthermore, Dr. Meiler proposed that certain patients or patient populations may exhibit an exaggerated inflammatory response to surgery and/or delayed resolution to the preoperative immune status. Limited human data are in support of this notion. If true, these patients may be at even greater risk to experience postoperative complications. What would cause an abnormal inflammatory response to surgery is not known, nor are all the factors that might be triggers beyond the surgical procedure itself. Whether anesthetic drugs, other aspects of anesthetic technique, or physiologic occurrences during surgery could be potent triggers for abnormal inflammation is not well established. There are threads of evidence that anxiety, fear, and pain can trigger inflammation.

This led Dr. Meiler to propose that perioperative care is a key nexus for affecting both short- and long-term outcomes. A common goal between anesthesiologists, surgeons, internists, and others will be to identify which patients are at risk, to define adjuvant treatments and modified patient care processes to prevent the negative outcomes, and to apply them throughout the continuum of the perioperative period. According to these models, taking an inflammation/ immune-based approach to dissecting the biological interactions between anesthesia, surgery, and postoperative complications therefore promises to yield important insights.

Last in the session, Rod Eckenhoff, MD, discussed the effects of volatile anesthetics on the oligomerization of brain proteins. This line of research was triggered by the speculation that many neurodegenerative diseases may be caused by the aggregation of normal and abnormal proteins (similar to what occurs in mad cow disease). Halothane and other volatile anesthetics do cause oligomer formation in amyloid precursor protein at clinically relevant levels, and this process lasts a long time. A single exposure of desflurane caused 3 days of differences in protein expression. Other proteins could be affected similarly. One example cited is ferritin, which binds volatile anesthetics at low concentrations. New animal models are being established in rats, which are thought to be a good model system. Finally, although these mechanisms suggest that exposure to volatile anesthetics might be linked to the occurrence of dementia, this conclusion is very speculative at this point.

Discussion of Session 2

The discussion was led by Dr. Carl Rosow. There was extensive discussion about what is meant by “stress,” and whether “anxiety,” “stress,” or other terms really describe the same state. There was further discussion of the diverse responses and timeframes of inflammatory responses that were being discussed. Some inflammatory response is critical for wound healing and warding off surgical infection, yet too much or too prolonged a response might be deleterious. The panel tried to determine whether any existing studies show a definitive link between the kinds of inflammatory mechanisms suggested and postoperative complications. Some threads were drawn from studies of patients having cardiac surgery and cardiopulmonary bypass, but it was acknowledged that no studies to date demonstrate this specific putative linkage. There was discussion of whether one could study the long-term outcome of animals (rodents in particular) that had undergone anesthesia and surgery, while studying their inflammatory responses. There was wide agreement that understanding the subtle issues of long-term outcomes would require both animal and human studies.

Session 3: Inflammatory Mechanisms Redux

Session 3 continued the theme of looking at the basic biological aspects of inflammatory mechanisms. First, Charles Serhan, MD, presented a fascinating description of the active processes that resolve inflammation after it has been triggered. This resolution is not just a “burnout” of the pro-inflammatory functions, but rather has a set of resolution functions that involve resolving mediators. Resolution is thus different than mere “anti-inflammation,” and the resolution processes offer another potential target for therapeutic manipulation. Furthermore, anti-inflammatory therapies may sometimes also inhibit the natural pro-resolution pathways, thus delaying or blunting their beneficial effects. Many of the mediators of the resolution phase are lipid mediators, among which are lipoxins, resolvins, and neuroprotectins. Resolvins may be the active ingredients of the beneficial effects of omega3 fish oil and other dietary elements. Similar molecules in the nervous system are called neuroprotectins. Dr. Serhan summarized the potential promise of this line of investigation as offering ways to mitigate the negative effects of inflammation by turning on resolution rather than by attempting to inhibit the pro-inflammatory phase itself. However, turning this basic science into therapies ready for clinical trials will take some time.

Discussion of Session 3

Don Stanski, MD, chaired the discussion of Session 3. There was extensive discussion of how to find the middle ground between the elegant basic science work and the interface to the clinical issues of long-term, post-surgical outcomes. In this regard, there is some work underway to develop more easily run assays for some of the resolution molecules. Another thread was trying to better link the seemingly beneficial effects of drugs like beta-blockers, clonidine, statins, and the underlying bioactive mediators.

An interesting question was raised as to the effects of discontinuing patients’ aspirin or non-steroidal anti-inflammatory therapies prior to surgery so as to minimize their effects on platelet aggregation, and thus on perioperative blood loss. A side effect of stopping these drugs could be to promote or unmask a more extreme inflammatory response.

In addition, there was considerable discussion about whether it is beginning to be possible to tease out which patients are most susceptible to short and long-term negative outcomes on the basis of their genotype or biochemical markers, in addition to the traditional risk factor analysis.

Session 4: Wrap Up

Dr. Jeff Cooper chaired this session and reiterated the interest of the APSF in the concept of long-term, postoperative outcome. Active participants in this session included Drs. Dan Sessler and Lee Fleisher who shared their experience, observations, and recommendations concerning future investigations. Dr. Karl Hammermeister proposed a “straw man” sequence of investigations to be considered by the panel. This sequence would be to:

  • Confirm excess late adverse outcomes (e.g., mortality)
  • Identify predictors of late adverse outcomes
  • Evaluate mechanisms of excess late adverse outcomes
  • Conduct small-scale trials of interventions for excess late adverse outcomes
  • Conduct large-scale comparative randomized clinical trials of interventions.

Following this discussion Dr. Cooper called for 3 mini-votes of the participants:

  • Do you believe that there IS some relationship between inflammatory processes in the perioperative period and long-term survival?

A majority of participants voted yes.

  • Should an appropriate study be done to measure “excess” mortality resulting from identifiable factors of anesthesia and surgery?

Again, a majority of participants voted yes.

  • Should such a study demonstrating “excess mortality” be completed before any other studies–such as those suggested in the “straw man”–are begun?

Only a few participants voted yes.

Marcel Durieux, MD, PhD, advocated conducting campaigns of basic research and clinical research in parallel, since useful clinical investigations can be done even before the underlying mechanisms are fully defined in the laboratory. Among the clinical questions that currently seem ready for investigation, he listed:

  • Relationship of intraoperative EEG measures to long-term mortality. The panel largely agreed that the data discussed so far are intriguing but very preliminary. He emphasized that the issue is not necessarily “deep anesthesia,” but that this might be a marker for patients who have different underlying physiology. We need to find out what causes this relationship, increased requirement, or exaggerated response.
  • The development of chronic pain after surgery, and use of “preventive analgesia” to prevent postoperative chronic pain
  • Perioperative transfusion and its effect on long-term outcome
  • The development of a perioperative hyper-coagulable state, which may result in thrombo-embolism and/or myocardial ischemia.

Some comments were made about possible follow-up activities stemming from this conference. Dr. Khuri indicated that NSQIP is very interested about incorporating AIMS data in NSQIP. Dr. Hunt from CMS indicated that CMS might well be interested in adding additional variables to the database analysis projects that are currently in design.

Dr. Thomas Russell, the Executive Director of the American College of Surgeons, said that it was very clear that the entire team caring for a surgical patient—surgeon, anesthesiologist, primary care physician, cardiologists, oncologists, and many others—needed to engage in serious new efforts to share and coordinate their knowledge, perspectives, and clinical efforts in order to optimize outcomes for the patients. Never again should they satisfy themselves to work solely within their own silos, no matter how expertly.


In summary, the group arrived at a number of threads of agreement and observation, which will now serve as the basis for future analysis and action. These include

  1. Historically, surgeons and anesthesiologists have largely felt their actions only have immediate or near-term consequences.Things not directly related to the surgical procedure that occurred “way down the line” (the “long-term outcomes”) had to do with the patient’s underlying medical conditions and were just bad luck. But the group thought it was distinctly possible that there are things that happen during surgery that have lasting effects and may have a long-term impact on morbidity and mortality.
  2. The group acknowledged that there may in fact be excess mortality over the long-term linked to the process of anesthesia and surgery. But the data are extremely sparse, complicated, and have many limitations and pitfalls. The question should be pursued further to find more definitive answers.
  3. There should be more studies of large numbers of patients to better identify risk factors for the occurrence of adverse long-term outcomes as well as for short-term complications.
  4. Inflammation has been implicated in many disease processes and it is definitely possible that there exists a relationship between inflammation and the long-term outcomes associated with surgery and anesthesia. Much remains to be determined to see if this linkage is present, and if so, its strength and what can be done about it. Studies are needed both on the basic biology of inflammation, and on the specifics of this biology in the setting of anesthesia and surgery.
  5. As we collect better data about the nature of postoperative outcome, studies are needed to evaluate the mechanisms, and define possible interventions. This may happen first in small-scale trials, but ultimately large-scale studies, with thousands of patients, will be needed. The benefits of extrapolating existing therapies, such as perioperative beta-blockade, to broader groups of patients requires further study as well.

Dr. Gaba is a Professor in the Department of Anesthesia and the Associate Dean for Immersive and Simulation Based Learning at Stanford University, Stanford, CA. He is also Secretary of the APSF.