Randomized trials may provide answers to troubling questions and outcomes regarding the pulmonary artery catheter.

The pulmonary artery (PA) catheter is a mystery. Inserted via a neck vein and floated through the right side of the heart into a branch of the pulmonary artery, the tiny, balloon-tipped device provides direct information on a patient’s volume status, cardiac output, and pulmonary vascular resistance. Critical care practitioners rely on PA catheters to guide fluid management and administration of inotropic drugs in their sickest, most unstable patients—-those with shock, severe heart failure, and acute respiratory distress syndrome. More than 1.2 million devices are inserted in critical care units each year. Up to 40% of patients in intensive care units (ICUs) receive them. “Clinicians tend to rely on them because they are easily available, especially at larger institutions,” says Lynn Kelso, RN, ACNP-CS, assistant professor of nursing at the University of Kentucky, Lexington, and an experienced teacher of many nursing aspects of critical care.

Yet, in 1996, a study of PA catheter use published in the Journal of the American Medical Association reported a disturbing finding: in a retrospective analysis of more than 5,000 patients treated in ICUs in five hospitals, including 1,000 pairs of patients closely matched for severity of illness and intensity of interventions, lead author Alfred F. Connors, Jr, MD, and his coinvestigators found increased mortality among the patients who had received pulmonary artery catheters. Their risk of dying was 24% higher than the risk in comparable patients whose treatment had been managed without PA catheters.
Was it simply that sicker patients were the ones who had received catheters and sicker patients were also the ones more likely to die? Experts did not think so. In an accompanying editorial, the late Roger C. Bone, MD, and James E. Dalen, MD, MPH, evaluated the statistical methods used to control for severity in the Connors study and concluded that there was a distinct possibility that the PA catheter itself had contributed in some way to the increased deaths. They called for a randomized trial to resolve the question. Failing that—if the clinical community was not willing to submit the PA catheter to such a test—Dalen and Bone called for a moratorium on its use.

In 1997, however, a consensus conference convened by the National Institutes of Health and the Food and Drug Administration found sufficient basis for continuing to use the PA catheter as before. Published studies were limited. But with some exceptions, participants agreed that the available literature justified the continued use of the PA catheter in many specific situations, including respiratory failure, sepsis, and myocardial infarction complicated by progressive hypotension or cardiogenic shock. There would be no moratorium. Participants also agreed that training in the use of the PA catheter should be standardized, since errors in interpreting the tracing, or errors in implementing the information, might contribute to poor outcomes. They decided to develop training materials to be used throughout the field of critical care. Finally, they agreed that randomized trials should be conducted.

Like many at the conference, Alfred Connors believed this was long overdue. “I had been involved in critical care for more than 20 years, and it had troubled me for a long time that there had never been a randomized trial to determine whether the PA catheter was beneficial. It seemed to me that if we are using this device so broadly, and we are so dependent on it, then we ought to have better evidence that it is beneficial—not just be operating on our feelings,” says Connors, who is professor and director of the Department of Health Evaluation Sciences at the University of Virginia School of Medicine, Charlottesville.
Connors had followed the literature closely and had noted several attempts to investigate the effects of right heart catheterization in observational studies. “Those studies consistently showed a harmful effect,” Connors says. The studies were criticized—and justifiably so, according to Connors—for not adjusting completely for case mix and severity of illness. “As a result, people in the critical care community believed they did not have to pay attention to those studies. But I believed this was not sufficient reason to dismiss them entirely,” Connors says.

In his own study, Connors used detailed data that enabled him to match patients in terms of 85 factors involving severity of illness and the intensity of interventions used. He assigned each patient a propensity score reflecting the likelihood that clinicians might have inserted a catheter. In this way he was able to compare outcomes in equivalent patients who did and did not receive catheters. The comprehensive analysis made it unlikely that Connors and his colleagues failed to identify any way in which the patients who received catheters were actually sicker than their matched counterparts. “Still, this was not a randomized trial in which patients would be assigned to receive catheters or not independently of any physician intuitions,” Connors says.

How might the PA catheter cause harm?
As Connors points out, although PA catheters are considered quite safe, complications do occur. These include infection, thrombosis, cardiac dysrhythmia, and injury to tissues within the heart chamber. The incidence of each is low, but may still contribute to overall mortality. For example, Connors found an incidence of line sepsis of 4% to 6%. “People dismiss this, but there are a number of studies showing that when you get line sepsis, you increase the risk of death 28% to 31%,” Connors says.

He reasons that a 5% incidence of line sepsis might therefore increase mortality by 1.5%. “A clinician who puts in 10 PA catheters a month—using them a fair amount—is never going to be able to observe an increase that small,” Connors says. Similarly, the incidence of cardiac thrombosis is about 25% among patients with PA catheters in place. “But you take the catheter out, the patient gets better, and you don’t realize the larger implications,” Connors says.

“But if infection increases the risk of dying 1% or 2%, and thrombosis increases the risk of dying 1% or 2%, and having the catheter in place increases the risk of dysrhythmias a little bit, and that increases the risk of dying a fraction of 1%, the cumulative effect could be a 3% to 4% increase in mortality—and that’s just from the direct effects,” Connors says.

Indirectly, the catheter provides a basis for the use of more drugs and more changes in treatment as well. “We respond to what the catheter tells us, and our treatment has risks,” Connors explains. “For example, let’s say the catheter tells us that a patient’s cardiac output is at the low end of normal. I can give dopamine, which is a drug that increases cardiac contractility. But that drug also increases the risk of fatal arrhythmias. So if increasing the cardiac output does not improve the chances of survival very much, but does add to the risk of dying—on top of the risk of complications from the catheter itself—the net effect may be harmful,” Connors says.

He notes that in his own study, patients with a better initial prognosis experienced a greater increase in mortality associated with the PA catheter than patients at high risk of dying. In patients initially at high risk, the catheter made little difference. “But my study can’t distinguish direct from indirect effects,” says Connors, who would like to see what a randomized trial reveals.

This, however, has proven difficult. “If a physician truly believes that a procedure is beneficial, then it would not be ethical to allow patients to be assigned to go without it,” says J. Dean Sandham, MD, FRCPC, FACP, professor and head of the Regional Division of Critical Care at the University of Calgary, Alberta, Canada, and medical director of critical care for the Calgary Health Authority. According to Sandham, who has successfully randomized almost 2,000 patients into just such a trial, it has been extremely difficult. “First of all, it has been difficult because of the preconceived ideas clinicians have about the impact of the catheter; second, because of the preconceived ideas they have about whether or not it is ethical to do a study; and third, because of the fact that critical care involves patients who come from other disciplines and so there are always a large number of physicians involved,” Sandham says.

His trial tested the catheter in 1,996 surgical patients aged 60 and older who were undergoing major abdominal, thoracic, vascular, or hip surgery. Eligible patients were admitted to the ICU within 24 hours of surgery. For patients randomly assigned to receive a catheter, the researchers adopted supernormal hemodynamic goals and a sequence in which therapies would be offered. Patients randomized to the control group received standard therapy, without a specific algorithm. Sandham is presently engaged in analyzing the data and hopes to submit the study for publication in the spring.

As he points out, before asking doctors to let their patients be randomized, it is necessary to establish “clinical equipoise,” that is, a conviction that it is equally reasonable to use or not use the tool being tested. He notes that in 1991, a trial of the PA catheter in Ontario had to be aborted because physicians were ultimately unwilling to allow their patients to be randomized. Sandham originally drafted his protocol in 1989, after several observational studies suggested that the PA catheter might have adverse effects and Canadian doctors were persuaded to take the possibility seriously. “Another thing we did to establish equipoise was to show that there was absolutely no evidence out there that PA catheters were beneficial,” Sandham says.

How do doctors decide whether to insert a PA catheter?
“That is part of the difficulty. Not only does it vary between physicians, but each physician may be inconsistent as well,” Sandham says. He finds it significant that Connors was able to identify 1,000 pairs of patients with equal propensity for getting a catheter, but only one member of each pair received one. “That implies a great deal of small-area variation among physicians in the decision to use a catheter,” Sandham says.

Is it possible that physicians have an intuitive sense about some patients that would not be reflected in a propensity score? “There is some of that, but a lot of the variation is situational. We all tend to be more likely to use a catheter if we don’t have confidence in the staff who are going to be on duty that night. Nobody is going to talk about that. You can’t measure it. But the decision may depend on what else is happening in the rest of the unit. If you have three or four emergent situations going on, you may feel you can’t clinically assess as frequently as you would like, and so you put a catheter in. It is a very complex decision,” Sandham says.

“There is a great deal of practice variation among physicians in our network,” agrees Herbert Wiedemann, MD, chairman of the Department of Pulmonary and Critical Care Medicine at the Cleveland Clinic Foundation and member of a network of 10 institutions involved in studying treatments for acute respiratory distress syndrome. As chairman of the protocol committee for a randomized trial of the PA catheter in acute lung injury, Wiedemann has given a lot of thought to the implications of this variability. “Currently, in one hospital, a patient might have a high chance of getting a catheter and in another hospital a very low chance, depending on institutional philosophies and the particular physician taking care of the patient,” he observes. In his view, it does not make sense that in good centers there can be such wide practice variation. “It just points to the fact that there is not a lot of data; therefore, physicians are working on their own personal assumptions,” Wiedemann concludes.

However, the variation has made physicians in the network more willing to commit themselves to a randomized trial. Under the factorial design Wiedemann has helped to develop, patients will be randomly assigned to receive either a PA catheter or a central venous pressure (CVP) line, and at the same time they will be randomized to one of two fluid management strategies: a fluid-liberal strategy or a fluid-conservative strategy. “If we simply put a catheter in and do not link it to any fluid management strategy, the outcome of the trial might be hard to interpret,” Wiedemann says. He hopes to study 1,000 patients, 250 in each cell, with 28-day survival as the primary endpoint.

Statistically speaking, will the study be large enough to show whether the PA catheter is beneficial? “It is possible that we would miss a small effect one way or the other, but we will be able to detect a relatively large effect. As a matter of fact, we will probably be able to come close to ruling out an adverse effect of the size the Connors study seemed to show,” Wiedemann says.

How does he account for the finding of increased mortality in the Connors study? “Some people believe that the finding was the result of the fact that the Connors study was retrospective, and that patients with greater mortality have subtle complexities that lead physicians to put catheters in,” Wiedemann says. To be fair, he admits that this is just speculation. “It is important to prove or disprove that finding in a randomized trial,” he says. How does he, as a clinician, decide whether to put in a PA catheter? “That is a hard question to answer. If a patient is very complex in terms of the underlying pathophysiology, and especially if he is tending in the wrong direction, based on what we think is reasonable using the information we have without the catheter, then we would insert one,” Wiedemann says. This puts him in the middle of the road. “There are centers in our network that almost always use a catheter and there are centers that almost never use one,” he says.

He hopes his trial will establish the safety of the PA catheter more precisely. “We do know that there is a very low rate of complications that can occur—for example, pulmonary artery rupture, infections, heart damage, and arrhythmias. These are things that can occur with a PA catheter that wouldn’t occur with a central venous line. However, we believe that the incidence of these events will be very low,” he says.

According to Kelso, although the CVP line is considered less invasive than the PA catheter, it can be associated with complications as well. Both bleeding and pneumothorax can occur. At the 1997 consensus conference, Kelso’s role was to review the evidence for direct complications of the PA catheter. Major complications were rare. In her view, the device provides certain information that cannot be obtained any other way—or at least not as quickly. It enables clinicians to administer fluid replacement and inotropic drugs much sooner. “But does it really improve outcomes in the long run? That’s what we have to find out,” Kelso says. “Maybe it improves survival in the short run, but really doesn’t change the final outcome at all.”

According to Kelso, there are many training issues with the device. Once it is correctly placed, reading a normal waveform is straightforward. But if the patient is experiencing backflow in one of the heart valves, or if the tracing contains large artifacts, it can be confusing. “You should still be looking at the waveform that shows the systolic pressure and the diastolic pressure,” Kelso says. It is crucial to read the wedge pressure at the end of expiration. “And when you are reading it based on the ECG tracing, you also have to make sure that you are reading it at end-expiratory,” she adds. If the patient is on a ventilator, that can help to identify the right point in the respiratory cycle, but again the practitioner must be alert to artifact.

“Like any tool, the more you use it, the better you get with it,” she says. Her main concern is that many centers rely on the PA catheter so heavily. This will make it difficult to do a randomized trial. “If you think the catheter helps you do a better job, and then your patient is randomized to go without it, you may believe you are not giving that patient your best,” Kelso says.

Planning a Trial
According to George Sopko, MD, project officer for a randomized trial of the PA catheter in congestive heart failure (CHF) now being planned at the National Heart, Lung, and Blood Institute, it is important for centers participating in the trial to have systematic training as well as systematic algorithms for managing patients with and without the PA catheter. If the training materials developed at the behest of the consensus conference are ready, centers participating in the trial will be asked to use them.

Sopko expects the role of nurses to be even greater in randomized trials of the catheter than in routine clinical use. “Don’t forget that the nurses are the ones who maintain the equipment and frequently are the ones who tell the physicians what the values are,” says Sopko, who has issued a request for proposals to recruit potential participating centers for his study. “Truthfully, nurses are the ones who are there consistently and who can recognize the problems—not just the benefits, but also potential complications. We are going to be monitoring those, too,” Sopko says.

“Our study will be conducted in 500 patients with Class IV CHF, because that is the group with high mortality and also the group in which management with the PA catheter often plays a significant role,” Sopko says. Patients will be randomized to receive a catheter or to be managed without one. “Our trial is going to address the question as to whether the invasive strategy, and the information gained from the invasive strategy, significantly alter the treatment plan or the outcome,” Sopko says.
Will the Canadian study and the two planned for the United States solve the mystery? “Let’s hope so,” Sopko says. “But no one trial will have all the answers.”

India Smith is a contributing writer for RT Magazine.