The Pulmonary Artery Catheter, 1967–2007: Title and subTitle BreakRest in Peace?

In this issue of JAMA, an investigation using a nationally representative administrative database reported a marked decline in the use of pulmonary artery (PA) catheters from 5.66 per 1000 medical admissions in 1993 to 1.99 per 1000 medical admissions in 2004.¹ These significant declines in PA catheter utilization were most prominent for patients with myocardial infarction (81% decrease), but also were significant for surgical patients (63% decrease) and for patients with septicemia (54% decrease).

These national data are consistent with trends at our institution, an academic public hospital and level 1 trauma center with 75 intensive care unit (ICU) beds with a relatively low volume of patients with acute myocardial infarction. For example, from July 2002 to May 2003, the hospital billed patients for 871 PA catheters. Although the ICU census has increased, the use of PA catheters has declined to 262 catheters from July 2006 to May 2007. Recently, nurses and residents gathered around the bedside of the sole patient in the medical ICU with a PA catheter so they could actually observe one in use. If the demise of the PA catheter is more than a rumor, why has this occurred and what are the implications for clinical care and training?

Forty years have passed since the afternoon in 1967 when Jeremy Swan watched boats from a Santa Monica beach and conceived of a bedside procedure that would use cardiac output to sail a catheter into the pulmonary arteries.² PA catheterization was initially used to assess patients with acute myocardial infarction, but use of this procedure spread rapidly to the operating department and from there to a broad range of patients in the ICU. The addition of mixed venous oximetry and cardiac output measurement to central venous and pulmonary arterial pressure monitoring provided clinicians with detailed feedback about physiological response to therapy. This information, coupled with clinical evaluation, allowed clinicians to titrate fluids, inotropes, vasopressors, and vasodilators to optimize oxygen delivery to tissues. Twenty years after its invention (in 1987), more than 2 million PA catheters had been sold worldwide annually.³ However, enthusiasm was not universal, and in the late 1980s concerns were raised about the unknown benefits of PA catheter–guided therapy in the face of potential risks from an invasive procedure.⁴

Subsequent research on PA catheterization supported these concerns. Clinicians, even experts, disagreed on the interpretation of data from the device and demonstrated poor knowledge of many basic concepts.⁵ Observational studies, limited by their ability to account for confounding variables, suggested that PA catheterization was associated with increased mortality.^{6 - 7} None of the 5 large randomized trials of PA catheterization have demonstrated a significant clinical benefit to patients. These randomized trials are particularly compelling given the consistent results across an extraordinarily broad spectrum of diseases and study designs, including patients with severe congestive heart failure,⁸ high-risk surgical patients using specified physiological targets,⁹ patients with shock or acute respiratory distress syndrome (ARDS) managed at the physicians' discretion,¹⁰ a heterogeneous ICU population stratified for use of noninvasive cardiac monitoring,¹¹ and patients with ARDS treated by clinicians at select academic ICUs who were trained in interpreting hemodynamic data and who were following a very specific management protocol.¹² Given these data, it is not surprising that a recent guideline on hemodynamic monitoring recommends against routine PA catheterization in shock and does not specifically recommend it for any clinical scenario.¹³

Different fields of medicine often provide useful cross-disciplinary perspectives on clinical practice. For example, although intended for entirely different patients, electronic fetal monitoring devices and the PA catheter have taken nearly identical paths. These technologies were both developed in the late 1960s and were disseminated rapidly. Both were based on reasonable but untested assumptions that intervening on abnormal physiology detected by the monitor would improve patient outcome. Interpretation of electronic fetal monitoring waveforms had the same interobserver variability and standardization problems as hemodynamic data from the PA catheter.¹⁴ Clinical trials of electronic fetal monitoring failed to show a reduction in the clinically significant end points of cerebral palsy or perinatal death, but led to an increase in cesarean delivery and operative vaginal delivery rates.¹⁵ Moreover, no professional society guideline has recommended electronic fetal monitoring over clinical assessment.¹⁶

Here the parallel paths diverge. In 2004, the last year data were available, electronic fetal monitoring devices were used in 85% of births and this number had remained relatively stable during the preceding 4 years.¹⁷ In contrast, PA catheter use has declined. Although a number of explanations for this variation in practice seem plausible including the risk of litigation after adverse outcomes in childbirth, an important difference between the devices is that electronic fetal monitoring is noninvasive and there is no less invasive or less expensive alternative to it.¹⁶ In the ICU, rather than abandoning hemodynamic monitoring, clinicians may simply be shifting away from the PA catheter to increasingly available, less invasive options for obtaining similar data including oximetric central venous pressure monitoring, echocardiography, pulse contour and pulse pressure variation analysis, and transesophageal Doppler.^{18 - 19} Some evidence supports this speculation. In a recent randomized trial in the United Kingdom,¹¹ 72% of participating ICUs preferred to use alternate cardiac output monitoring technology to care for patients who would not receive a PA catheter. It remains a question for future studies whether the apparent decline in PA catheterization reflects a shift in philosophy or simply a substitution of technology.

The declining use of the PA catheter has important implications for training and clinical care. Some critical care physicians actually view the current negative trials on PA catheterization as supporting their use because the results challenge previous studies of increased mortality associated with the device. The lack of efficacy in the clinical trials is attributed to poorly defined or selected physiological targets, incorrect use, or inappropriate patient selection. Clinicians with these views may find the decline in PA catheterization unwarranted and potentially even dangerous. Regardless of the confidence individual clinicians have in their personal ability to improve outcome guided by a PA catheter, the entire field must confront the current reality of its infrequent use. Patients with a broad range of diseases, including critically ill patients, benefit from care at high-volume centers.²⁰ Procedures like PA catheterization often demonstrate worse outcomes when they are performed infrequently. In the ESCAPE trial,⁸ a National Institutes of Health–funded study evaluating PA catheterization in patients with congestive heart failure, there was no overall benefit from invasive monitoring. However, there was a trend toward worse outcomes at sites that enrolled fewer patients and may have had less experience with the procedure. Hospitals should consider several options to address the increasingly rare procedure of PA catheterization.

First, centers performing only a few PA catheterization procedures should consider not doing any. Because it is likely that centers that use PA catheters infrequently will derive even less benefit than the published studies that show no benefit, low-volume centers should carefully consider whether there is greater risk from performing only a few PA catheterizations than doing none. Little evidence is available to guide selecting a specific threshold, although more than a decade ago, Swan recommended a minimum of 50 procedures per year per physician to maintain competency.²¹ Because the experience of the physician placing the catheter is only 1 component of a team that uses data from hemodynamic monitoring, we believe that ICUs should evaluate their procedure volume at the unit level. This advice may lead to a self-fulfilling prophecy: the fewer PA catheters used, the less useful they are likely to be in the future. If true, this certainly does not justify continuing PA catheterization to maintain the skill for the theoretic benefit of future patients.

Second, if PA catheterization use is continued, the procedure should be limited to a small number of skilled clinicians. Specialized teams of clinicians devoted to specific tasks have demonstrated value in critical care. With decreasing volumes of PA catheterization procedures, it may be necessary and feasible to assign 1 nurse on each shift to be responsible for setting up and recording data from all PA catheters in a hospital and to restrict the use of the devices to a highly select group of physicians. This suggestion may not be well accepted by those clinicians who view PA catheterization as part of the basic skill set of all critical care clinicians. However, the evidence for and declining use of PA catheters no longer support this claim. Although recent studies cast significant doubt that any patient “requires” a PA catheter, if clinicians perceive a need for this level of monitoring at a hospital that performs few of these procedures, it might be safer to transfer these patients to a more experienced center than to perform them locally.

Third, consider alternate hemodynamic monitoring tools, but consider them skeptically until convincing outcome data are available. Randomized trials of PA catheters were not simply experiments of a piece of plastic tubing, but rather tested the ability of clinicians to use hemodynamic information from the PA catheter to improve outcome. The overall disappointing results of these trials may reflect a mixed signal; that is, a benefit from hemodynamic monitoring but harm from the PA catheter. If true, less invasive alternatives might yield better outcomes by providing the same information with less risk. However, the harm and cost of hemodynamic monitoring may not result from the device itself but from treatment decisions based on the information. There is also a potential risk from focusing clinicians' limited time away from aspects of intensive care known to improve outcome to learning how to use and interpret data from new, untested devices.

Some preliminary evidence exists that resuscitation guided by central venous pressure and oximetry in patients with sepsis¹⁹ or transesophageal cardiac output measurement in patients after cardiac surgery can improve outcome.²² However, there was no benefit of access to alternate cardiac output monitoring in a heterogeneous group of critically ill patients in the control group of the PAC-Man trial¹¹ who were treated without a PA catheter. Similarly, patients with ARDS or shock did not appear to benefit from the information provided by an echocardiogram on day 1 of their illness in the control group of a randomized trial that did not receive a PA catheter.¹⁰

Therefore, until persuasive studies demonstrate that patient outcome can be improved by care guided by alternate hemodynamic monitoring options, clinicians should view this technology skeptically. Critical care clinicians certainly know how to measure hemodynamic variables but have not yet determined which interventions based on this information improve patient outcome. It would be unfortunate if the last 40 years of experience with PA catheterization is repeated with a set of new devices.

Fourth, continued use of PA catheterization demands intensive evaluation and education. Even when PA catheter use was high, studies repeatedly demonstrated that clinicians around the world had poor ability to use the technology. Many ICUs rely on nurses to set up, troubleshoot, and record hemodynamic data from the PA catheter. With an increasing nursing shortage and turnover in critical care, collective experience with the PA catheter will decline rapidly. In the waning years of PA catheterization when the risks of infrequent use may outweigh any unproven benefit, it is more important than ever to remain vigilant for common errors in PA catheter placement and data interpretation. Passive education may not be sufficient and hospitals that choose to continue to use PA catheters will need to invest in simulation experiences, hands-on training, and quality monitoring to maintain a vanishing skill.

No one would consider using a new drug without some evidence that it improves patient outcome. It is tempting to apply a different set of standards to diagnostic, screening, and monitoring technology, which are only expected to provide accurate information or to change the choice of treatments.²³ Unfortunately, more information, even when accompanied by a change in management, does not necessarily translate into better patient outcome and, in fact, more information, even when it is accurate, may lead to ineffective, expensive, and even harmful treatments.^{24 - 25}

The 40-year story of the PA catheter is nearing its end. It is a cautionary tale of rapid adoption and slow evaluation of a monitoring device that, when used correctly, provides exquisitely detailed physiological data that, regrettably, does not appear to benefit patients. Older clinicians will look back wistfully on the hours spent placing, troubleshooting, and debating the data from the PA catheter. Younger colleagues will just wonder what all the fuss was about.