Practice Guidelines for Pulmonary Artery Catheterization: An Updated Report by the American Society of Anesthesiologists Task Force on Pulmonary Artery Catheterization

Task Force documents and primary evidence are available at the Wood Library, American Society of Anesthesiologists, Park Ridge, Illinois.

The members of the Task Force were originally selected by the American Society of Anesthesiologists Committee on Practice Parameters and by the Task Force chair. Members were invited to reprise their roles on the Task Force for the current update. The nine-member Task Force included four university-based anesthesiologists, four community-practice anesthesiologists, and a methodologist. Median experience with PA catheter use among the eight anesthesiologists was 21 yr (range, 18–29 yr). Task force members used an average of nine PA catheters per month, primarily for cardiovascular surgery cases. Current monthly averages were 0–3 for four panel members, 10–12 for three panel members, and 31 for one panel member.

Reliable national data are lacking to assess whether these practice patterns are representative of PAC use among most anesthesiologists, even for those who regularly care for high-risk patients. A survey of 1,000 critical care physicians revealed that 12% inserted no catheters, 60% inserted one to five catheters per month, and 27% inserted six or more catheters per month. 13An anonymous cross-sectional survey of 214 anesthesiologists in Canada and the United States reported that physicians practicing in sites with transesophageal echocardiography (TEE) availability averaged 17.6 cardiovascular cases per month, whereas those practicing in sites without TEE availability anesthetized 4.6 cardiovascular cases per month. 14As a proportion of the total cases, fewer PACs were inserted when TEE was available (11.4/17.6, or 65%) as compared to when in it was not available (3.2/4.6, or 70%). In a survey of 170 attendees at the May 2001 meeting of the Society of Cardiovascular Anesthesiologists in Vancouver, British Columbia, Canada, 25% reported that they placed 0–5 PA catheters per month, 22% placed 5–10, 36% placed 10–25, and 17% placed more than 25.

The objectives of the Task Force, as in its original effort, were to develop guidelines on the appropriate indications for PA catheter use. The purpose was not to describe how to perform the procedure or to interpret results. The Task Force sought to develop guidelines based on scientific evidence supplemented by expert opinion, and to follow a systematic methodology for reviewing evidence and developing recommendations. The guidelines were intended for situations encountered by anesthesiologists and the perioperative care team. They do not address the use of PA catheters in nonsurgical settings or by other medical specialists.

In its original effort, the Task Force developed its recommendations following a systematic review of the clinical benefits and harms of PA catheterization. Benefits and harms were evaluated by reviewing relevant scientific evidence and incorporating expert opinion about effectiveness. The review was guided by an evidence model developed by the Task Force. In addition to clinical effects, the Task Force also considered public policy issues, such as costs and implementation issues and strategies. The current effort began by revisiting its earlier process and determining which aspects of the methodology to revise before proceeding with updating. Among the changes was a refinement of the evidence model (fig. 1).

Both the original review of scientific evidence in 1991–1992 and the update in 2000–2002 included a detailed literature search, a critical appraisal of individual studies to assess methodologic quality, and a synthesis of the results.

The original computerized and manual literature search was conducted in November 1991 and was updated in May 1992. It sought all relevant English-language articles or abstracts published after 1972. A total of 860 clinical trials, controlled observational studies, uncontrolled case series reports, and individual case reports were considered. The update in 2000–2002 involved three computerized literature searches of the MEDLINE database (conducted in September 2000, March 2001, and May 2002) for articles published between 1992 and 2002. The search sought all English-language articles or abstracts indexed under the Medical Subject Heading “Catheterization, Swan-Ganz.” That search strategy retrieved 665 articles, 71 of which met inclusion criteria. A manual search (review of bibliographies, consultation with Task Force members) identified 19 additional articles, for a total of 90 new studies meeting inclusion criteria.

Detailed exclusion criteria are described in table 1. The Task Force focused its review on evidence of effectiveness based on clinical outcomes. PAC use was interpreted as including its diagnostic applications (in measuring PA pressures, cardiac output, mixed venous oxygen saturation, and other indices) and selected therapeutic uses (e.g. , pacing, PA venting). Systematic reviews and meta-analyses were included. Editorials, review articles, and letters were not systematically reviewed. The Task Force did not directly examine the accuracy of PA catheter monitoring, value of PA catheter data as predictors of morbidity and mortality, or evidence of the effectiveness of treatment for PA catheter-detectable conditions. The Task Force did not evaluate the effectiveness of alternate hemodynamic monitoring technologies (e.g. , TEE), although it recognizes that in settings in which TEE is available and appropriate it may supplant the need for PAC. 15Issues related to the performance of PA catheterization, such as rates of utilization, practitioner skill, resource constraints imposed by staff and equipment availability, medicolegal concerns, and reimbursement, were not a specific focus of the literature review. A focused review of the “learning curve” literature was performed to examine what is known about the number of procedures physicians must perform to acquire and to maintain cognitive and technical skills.

The methodologic quality of individual studies was assessed in a systematic manner by considering study design category (e.g. , observational vs.  experimental design) and the quality of the research methods (e.g. , statistical power, selection bias, measurement error, confounding variables, internal and external validity). The Task Force recognized the general superiority of randomized controlled trials over observational studies in evaluating the effect of interventions on outcomes.

The synthesis was narrative and utilized traditional evidence tables. Evidence of effectiveness was not suitable for formal meta-analysis.

The expert opinion of the Task Force was assessed in its original guideline by informal consensus development. It was later suggested that such opinions should be assessed according to a more formal group consensus process. 16,17In this update, the Task Force used a confidential voting scheme to assess the appropriateness and necessity of PAC, specifying its views for 27 clinical scenarios. The 27 scenarios considered each potential combination of patients, surgical procedures, and practice settings in low-, moderate-, and high-risk categories, as defined in table 2.

After considering the findings of the updated systematic review, seven Task Force members (one panel member was absent and one was not an anesthesiologist) completed an anonymous questionnaire at the second Task Force meeting, assigning scores for the appropriateness and necessity of PAC in each of the 27 scenarios. Task Force members were unaware of the votes taken by other Task Force members. A 1–9 scale was used, with 1 representing the most inappropriate (or unnecessary) indications and 9 indicating the most appropriate (or necessary) indications. The definition for appropriate  set more liberal boundaries (“may or may not be necessary, but doing it is not wrong”) than that for necessary  (“should be performed”). The distinction allowed for circumstances in which catheterization is appropriate but not mandatory (“necessary”) and, conversely, unnecessary but not inappropriate. The median (and distribution) of the scores for each of the 27 scenarios was reviewed by the group (without disclosing the votes taken by individual members) and was used as a basis for formulating recommendations. PAC was considered appropriate (or necessary) when median scores were in the range of 7 through 9 and inappropriate (or unnecessary) when in the range of 1 through 3. The definitions of low-, moderate-, and high-risk were not expounded beyond the level of detail provided in table 2(e.g. , giving examples of specific operations that are high-risk) because these judgments depend on local circumstances, but this decision was made with a conscious recognition that the lack of specificity creates some ambiguity in how the categorizations might be interpreted and applied.

Costs and implementation issues were considered by the Task Force only after the clinical benefits and harms of PA catheterization were studied. The Task Force's recommendations were based on perceived clinical benefits, harms, and cost-effectiveness of PA catheterization. Cost information provided in published clinical research was reviewed, but the Task Force did not seek out cost data from other sources (e.g. , payers, manufacturers).

The Task Force's first guideline was informed by a widely publicized open forum on the proposed guidelines, which was held in San Francisco in March 1992 at the annual meeting of the International Anesthesia Research Society. A similar vetting of the update occurred at open forums held at the March 2001 meeting of the International Anesthesia Research Society in Fort Lauderdale, Florida, and at the May 2001 meeting of the Society for Cardiovascular Anesthesiologists in Vancouver, British Columbia, Canada.

Attendees at both sessions were invited to complete the same survey that panel members used to vote on the appropriateness and necessity of PA catheterization in 27 clinical scenarios. Seven attendees at the International Anesthesia Research Society meeting completed the surveys before the panel votes were presented. The five attendees at the Society for Cardiovascular Anesthesiologists who completed the survey were given clinical examples to illustrate the risk categories and were asked to answer questions about their clinical background. More than half indicated that they had used PA catheters for longer than 15 yr; 50% used the catheters 5–10 times per month, and 40% used them 11 or more times per month. The proportion that used them more than 30% of the time was 100% for open-heart chamber cardiac cases, 60% for closed-heart chamber cardiac cases, 73% for off-pump bypass cases, 75% for aortic valve cases, and 100% for peripheral leg vascular cases.

The guideline underwent peer review by experts in PA catheterization and by relevant specialty societies and organizations. Reviewers are listed in table 3.

Clinical effectiveness was judged by considering the benefits and harms of PA catheterization. Clinical benefits and harms were evaluated by reviewing relevant scientific evidence and expert opinions of effectiveness held by the Task Force and reviewers.

The results and design of individual studies are reviewed next. The results of the controlled studies are summarized in table 4 

Survey studies in postoperative and intensive care units have demonstrated that PA catheter data provide new information or seem to change therapy in 30–62% of cases. 18–25The clinical significance of these changes is uncertain. Treatment modifications were judged important in 25% of adults 21and in 10% of children 26monitored by PA catheter. Studies have found no association with mortality among patients whose therapy is altered based on PA catheter data. 22–24In one study, the added information from mixed venous oxygen monitoring changed the therapy in 57% of patients undergoing cardiac surgery, but the changes were not associated with improved outcomes. 27A subgroup analysis in one cohort study reported that patients with shock unresponsive to standard therapy had significantly lower mortality rates when PAC hemodynamic data led to a change in therapy than when therapy was unmodified. 24 

The quality of this evidence is poor. The conclusions are based largely on self-reported data in questionnaires, which are subject to measurement and recall biases. Most studies were unblinded and, in many cases, judgments about whether treatment was altered based on PA catheter data were made subjectively. Sample sizes were inadequate to conclude that alterations in treatment had no effect on mortality, and clinical outcomes beside mortality were generally not examined.

Uncontrolled case series reports have shown that preoperative PA catheterization is associated with cancellation or modification of surgical procedures and with altered hemodynamic management, and investigators have concluded that it therefore prevents morbidity and mortality. 28,29Controlled trials attempting to confirm this hypothesis have yielded mixed results and suffer from design limitations. Post hoc  data analysis in one trial demonstrated a lower mortality rate in patients who were monitored preoperatively than in patients who were first monitored after surgery, but control for confounding was limited. 4Another trial suggested that preoperative catheterization reduced intraoperative complications and graft thrombosis in patients undergoing peripheral vascular surgery, 30but it also was vulnerable to confounding. 31 

Conversely, a trial examining routine catheterization before elective vascular surgery reported no significant differences in mortality or complications, but the sample was small and excluded high-risk patients. 32An observational study found no difference in outcomes between elderly patients who did not undergo preoperative catheterization and unmatched patients who were admitted to the hospital during the same time period for other reasons. 33The similar outcomes may have been due to selection biases. In summary, no high-quality evidence exists to infer that routine, or even selective, preoperative catheterization improves outcomes regarding hemodynamic optimization).

For many years, the principal evidence regarding the benefits of hemodynamic monitoring in the surgical setting was limited to nonrandomized observational studies. For example, a historical control study of 733 patients found that patients with previous myocardial infarction who underwent noncardiac surgery during a period when invasive hemodynamic monitoring was common had lower reinfarction and mortality rates than patients in previous years during which invasive monitoring was less common. 34It is unclear from this study whether hemodynamic monitoring or other temporal factors were responsible for the improved outcomes and whether the two groups of nonrandomly selected patients were comparable in terms of case mix and severity of illness.

A subsequent randomized controlled trial involving 146 patients found no difference in intraoperative mortality, length of hospital stay, length of intensive care unit (ICU) stay, ventilator use, or postoperative mortality when surgical patients monitored by PA catheter were compared with central venous catheter (CVC)-monitored patients. 35However, sample size may have been inadequate to reveal a true benefit. Patients monitored by PA catheter who were managed with the goal of achieving supranormal metabolic goals seemed to have significantly lower postoperative mortality, length of ICU stay, and ventilator use. 35Because of uncertain methodology and inconsistent data reporting, it is unclear whether this group of patients differed in case mix from other patients monitored by PA catheter, but the findings stimulated research efforts to confirm the benefits of goal-directed therapy.

Most studies of goal-directed therapy have been conducted in the ICU and have included, primarily or exclusively, surgical patients monitored by PAC. One randomized trial involving 107 high-risk surgical patients found that a pre- and postoperative protocol to achieve high oxygen delivery rates was associated with significantly lower 28-day mortality and complication rates. 36Although a smaller trial with design flaws also reported encouraging results—a higher cardiac index was associated with lower mortality in patients with septic shock 37—subsequent randomized trials of goal-directed therapy have been less encouraging.

Two trials from the same center reported no reduction in mortality, organ dysfunction, or length of stay, although they failed to achieve a difference in oxygen delivery between the two groups. 38,39,Post hoc  data comparisons suggested a significant mortality reduction in patients who achieved high oxygen delivery levels, but this disregarded intention-to-treat analysis. Another trial examined the benefits of increased oxygen delivery in 100 high-risk, critically ill patients who did not reach target values with volume expansion. The authors reported increased in-hospital mortality rates and no significant improvements in surrogate outcomes (days of ventilation, length of stay). 40The protocol was unsuccessful in significantly improving oxygen consumption, however, and PAC insertion often did not occur until after surgery when complications occurred on the ward. The largest trial involved 762 high-risk patients admitted to 56 ICUs; surgical patients comprised only one of five subgroups. The trial reported no difference in mortality, organ dysfunction, or length of stay when the goals of normal cardiac index, supranormal cardiac index, and normal mixed venous oxygen saturation were compared. 41A trial in the setting of aortic surgery compared patients receiving PAC with preoperative and intraoperative goal-directed therapy and patients receiving no PAC. The PAC group had significantly more intraoperative (but not postoperative) complications, but complication rates even in the control group exceeded current norms, raising questions about the external validity of the study. 42 

A randomized trial in an urban emergency department of patients admitted with severe sepsis or septic shock compared 130 patients receiving 6 h of early goal-directed therapy with 133 patients receiving standard therapy before admission to the ICU. 43Twenty-eight day mortality was significantly lower for the early goal-directed therapy group, with a relative risk of 0.58 (95% CI, 0.39–0.87). A special mixed venous oxygen sensing central venous catheter was used in this study; therefore, this article does not fit the evidence model as described above, nor does it describe the effect of PAC guided therapy. However, it confirms a benefit to “goal-directed therapy” for this specific subgroup of patients and suggests that other monitoring techniques to guide goal-directed therapy (including PAC) may also lead to beneficial outcomes.

A trial in Britain involving 138 patients undergoing major elective surgery reported positive results, including a significant reduction in mortality, for patients who underwent PAC-guided preoperative optimization as part of a protocol that included return to a high-dependency or intensive care unit following surgery. 44Because 16 patients in the control group did not receive postoperative intensive care, it is unclear to what extent the observed outcomes were due to the latter and not optimization. 45 

A Finnish trial involving 393 patients undergoing cardiac surgery reported that length of stay was slightly shorter (median of 6 rather than 7 days) and hospital discharge faster for the group randomized to receive goal-oriented therapy postoperatively. 46Although it was not a primary outcome measure, the proportion of patients with organ dysfunctions on the day of discharge was significantly lower (1%vs.  6%) in the intervention group. Other outcomes, including prolonged stays in intensive care and mortality rates, did not differ between groups.

Hemodynamic disturbances, for which PAC is often useful in the perioperative setting, have been the context for studies of PAC involving medical patients who did not necessarily undergo surgery. Some of these studies are reviewed here because the hemodynamic disorders that prompted PA catheter use (e.g. , myocardial infarction, sepsis, pulmonary edema) include the principal risk factors for which surgical patients often undergo PA catheterization.

Uncontrolled studies have produced inconsistent findings. 47,48A study with historical controls reported that mortality in patients with septic shock decreased during a period in which use of PA catheters had increased, but the study design provided weak evidence that PA catheters had a causal role. 49Studies with control groups have generally found that patients with myocardial infarction and other hemodynamic disorders who are monitored by PA catheters have a higher in-hospital mortality, longer hospital stay, and shorter long-term survival than patients who do not undergo PA catheterization. 50–52Although these studies have included large samples (300–6,000 patients), their designs were retrospective and failed to control adequately for differences in severity of illness, selection bias, and variation in catheter use. The data therefore do not clarify whether patients who underwent PA catheterization were sicker than unmonitored patients.

The most careful effort to control for this form of confounding was a cohort study of the effects of PAC in 5,735 critically ill medical patients who entered a trial during the initial 24 h of their ICU stay. 9The investigators developed a “propensity score” to adjust for over 30 demographic, prognostic, and physiologic variables known to be associated with PAC use. In an analysis that paired patients by propensity scores (to eliminate confounding variables other than whether patients received PAC), the investigators observed that patients who received a PAC were 24–27% more likely to die within 1–6 months and had longer admissions and required more intensive care. The hazard rate for PAC use was even higher for the subset of patients receiving postoperative care—their risk of death in association with PAC was increased by 58% (95% CI, 1.17–2.13)—but the latter accounted for only 708 (12%) of the 5,735 patients in the study.

The authors noted the limitations of the study. Although the propensity score held up under several sensitivity analyses, it retained wide confidence intervals and the authors could not exclude the possibility of a missing covariate. They noted, however, that for PAC to have a true hazard ratio of 0.8 (20% reduction in risk), a missing covariate would have to independently increase the risk of death and likelihood of PAC sixfold to explain the observed increase in mortality. Some contend that subtle clinical variables not addressed in the propensity score, such as physiologic trends over time and response to treatment, do have such an effect. 53Others critique the magnitude of the estimated influence of a missing covariate. 10Moreover, the study did not capture data on causes of death or on how patients were managed to know whether observed outcomes were due to the direct effects of PAC, level of skill in insertion or catheter use, therapies directed by the catheter, or other interventions associated with PAC. 54Finally, the generalizability of the study to conventional perioperative monitoring is limited because surgical and trauma patients were poorly represented, the study cohort was severely ill (30-day mortality of 31–38%), and most patients had noncardiac disease.

Uncontrolled observational studies that examined outcomes in cardiac surgery patients who were monitored by PA catheter 55–61have been limited by the lack of comparison data from unmonitored patients. A small study that included historical controls found that 28 patients who underwent PA catheterization for repair of the left main coronary artery stenosis had lower rates of perioperative myocardial infarction, ventricular fibrillation, and deaths than 20 patients from the previous year who were monitored by CVC. 62However, the historical design left unanswered questions about whether the results were attributable to temporal factors or differences in case mix.

Studies that included internal controls found little benefit. A controlled prospective observational study of 1,094 patients found no difference in measured outcomes (e.g. , mortality, cardiac ischemia, postoperative myocardial infarction) among coronary artery bypass graft patients who were monitored by CVC, elective PA catheterization, or emergency PA catheterization. 63High-risk patients who underwent elective PA catheterization had longer ICU stays and were more likely to receive vasopressor agents.

Other observational studies have suggested an adverse association between PAC and health outcomes. A retrospective study compared outcomes among 194 patients who met predetermined criteria for CVC, a third of whom were monitored by PAC to accommodate physician preference; the PAC-monitored patients had increased complications and duration of mechanical ventilation. 64These differences and the lack of observed differences in other outcomes may have been attributable to the case mix, because patient assignment to groups was made selectively; for example, PAC-monitored patients were older and more likely to have congestive heart failure. Another study retrospectively reviewed administrative data on 13,907 patients who underwent nonemergent coronary artery bypass surgery at 56 hospitals in the United States. The 8,064 patients who received PAC, even after adjustment for case mix and other covariables, had significantly higher in-hospital mortality rates and length of stay than those who did not receive PAC. The association was strongest for hospitals that inserted fewer PACs per year. 65 

Whether such associations are due to PAC or other confounding variables is difficult to ascertain without conducting a randomized controlled trial. One such trial involving 226 patients found no difference in measured outcomes (deaths, length of ICU stay, use of vasopressors) between coronary artery bypass graft patients monitored by PA catheter and CVC, but this may have been attributable to the small sample size and selection biases. Fully 64% of patients in the control group were removed selectively after randomization if the anesthesiologist thought that they required PA catheterization. 66In summary, there is conflicting evidence from controlled studies regarding the benefit that cardiac surgery patients receive from PA catheterization.

A randomized controlled trial found that patients undergoing peripheral vascular surgery were less likely to experience intraoperative disorders (tachycardia, hypotension, arrhythmia) if PA catheters were placed preoperatively and if hemodynamic status was optimized. 30The overall incidence of postoperative complications (renal failure, congestive heart failure, myocardial infarction, graft thrombosis, death) seemed to be lower in the patients monitored by PA catheter. This was due mainly to a higher incidence of postoperative graft thrombosis in the control group, which was attributed to poor cardiac output. Postoperative morbidity and mortality otherwise did not differ between groups. The reported data do not support the authors’ conclusion that postoperative morbidity was reduced. The study was limited by discrepancies in data reporting and by uncertain methods for group assignment.

Although some case series reports have noted that outcomes in patients monitored by PA catheter were better than adjusted rates for the general population, 67the absence of control groups limits the value of the data. A study with historical controls found that mortality, perioperative hypotensive episodes, and renal failure were less common in patients who received an aggressive fluid management protocol that included PA catheterization than in previous patients who did not receive the protocol. 68The study was limited by its historical design, inconsistent data reporting, unexplained attrition, and the fact that PA catheterization was only one component of the protocol. Uncontrolled case series have suggested that low-risk cases can be safely managed without PAC. 69One prospective study of two hospitals with notably different mortality rates for nonelective aortic repair, despite comparable acuity, noted that PAC (and colloid and inotrope) use was much more common in the hospital with higher mortality rates. 70The only two randomized controlled trials examining the benefits of PAC over CVC in patients undergoing abdominal aortic reconstruction found no difference in outcomes. 71,72 

An uncontrolled case series and an observational study with a comparison group have examined the effectiveness of PA catheterization in patients undergoing neurosurgical procedures. 73,74The studies only addressed the ability of PA catheters to detect air embolism and did not measure clinical outcomes. An uncontrolled observational study of pediatric head trauma patients who underwent monitoring that included PA catheterization reported lower mortality rates than published rates for patients with similar trauma scores. 75The study lacked internal controls, and PA catheterization was only one of the multiple interventions offered.

Some studies of limited quality have suggested that hemodynamic monitoring of trauma patients, often including PAC, improves outcomes. For example, a retrospective analysis concluded that patients with life-threatening burns who were monitored by PAC had lower mortality and organ-failure rates than a historical control group in which PAC was used irregularly; patients managed with hyperdynamic endpoints had substantially better outcomes. 76A small observational study of critically injured patients reported a significantly lower incidence of organ dysfunction, but not of organ failure, in the group monitored by PAC to achieve optimal left ventricular performance. 77Retrospective biases and failure to adjust for confounding variables limit the generalizability of the findings.

Randomized controlled trials examining goal-directed therapy in trauma patients offer encouraging results but suffer from design limitations. A trial involving 67 primarily surgical trauma patients reported significantly fewer organ failures per patient among those receiving PACs in the operating room or postoperatively as part of a protocol to achieve supranormal values. 78The number of patients with organ failure was not compared, however, and the study suffered from failure to conceal allocation, nonblinding, and other design flaws. A subsequent trial from the same group reported similar findings, as well as lower mortality rates, but it suffered from additional design limitations. 79Another trial involving 58 patients with critical illness due primarily due to trauma reported no significant benefit from PAC and goal-directed therapy. 80It also suffered from numerous design limitations.

Evidence regarding the effectiveness of PA catheterization in obstetrics and gynecology is lacking. PA catheterization has been recommended for severe preeclampsia, 81case reports have supported its value, 82and its use in critical illness seems common, 83but controlled clinical outcome studies have not been reported. Case reports have also described PA catheter use in uncommon obstetric settings, 84–86whereas others have documented its lack of necessity in myocardial infarction during pregnancy. 87Uncontrolled case series have examined mortality and other outcomes in series of obstetric and gynecologic surgical patients that underwent PA catheterization, but they did not include control groups for comparison. 88,89 

PA catheterization is performed in critically ill newborns, infants, and children, 90but its effect on clinical outcomes is poorly studied. Uncontrolled case series reports have shown that it is useful in clarifying diagnoses, but these studies have not examined clinical outcomes or included unmonitored controls. 91,92 

A meta-analysis of 16 randomized controlled trials involving PAC, which examined results from studies published between 1970 and 1996 and which used a random effects model to compensate for study heterogeneity, yielded a relative risk ratio of 0.81 (95% CI, 0.60–1.10) for mortality in patients treated with PAC. 93Patients from surgical series had a relative risk ratio of 0.58 (95% CI, 0.36–0.94). The same research team later extended the analysis to include morbidity data from 12 of the 16 trials, calculating a relative risk ratio of 0.78 (0.65–0.94) for the incidence of organ failure. 94Although the estimates were statistically robust under sensitivity analysis, their validity is arguable, given the disparate patient populations and protocols and the numerous design flaws of the studies collated in the analysis. Another meta-analysis of PAC, which examined results from four homogeneous randomized controlled trials published between 1991 and 1998 involving vascular surgery patients yielded an odds ratio of 1.198 for mortality when PAC was used (P = 0.60; confidence intervals were not reported). 95The odds ratio for postoperative complications between PA catheter patients and controls was not significant.

Evidence regarding the adverse effects of PA catheterization comes from studies that examined multiple complications 96–108and from studies of specific complications. The Task Force considered only potentially life-threatening complications. The reported incidence of complications is summarized in table 5.

In summary, complications from PA catheterization can occur with the establishment of central venous access, the catheterization procedure, and catheter residence. Complications of establishing central venous access include malpositioning, 109unintentional puncture of nearby arteries, (e.g. , carotid or subclavian artery), 110,111bleeding, neuropathy, air embolism, and pneumothorax. 109,112,113Technologies, such as ultrasound guided venous cannulation, are available and may reduce the risk of catheter misplacement. 114–119 

Dysrhythmias are the primary complication of the catheterization procedure. Minor dysrhythmias, such as premature ventricular and atrial contractions, occur often with catheter insertion or withdrawal but usually resolve spontaneously after the catheter is advanced or withdrawn through the right heart chambers. 120–123Ventricular tachycardia or fibrillation occurs occasionally and can usually be cardioverted with antiarrhythmic agents or electrical defibrillation. 124,125Catheter advancement can produce a right bundle-branch block, and in patients with a preexisting left bundle-branch block it can precipitate a complete heart block. 126–128A minor increase in tricuspid regurgitation has been shown to occur 17% of the time after catheter placement. 129This study did not show that PAC resulted in new severe tricuspid regurgitation.

Complications of catheter residence include venous thrombosis, 130thrombophlebitis, and pulmonary embolism and infarction. 131Autopsy studies have revealed evidence of mural thrombi, endocarditis, and valvular injury in patients with indwelling PA catheters. 132–136Catheter-related venous thrombosis can be reduced with heparin, 137but the latter carries its own risks. 138A serious complication of PA catheters is PA rupture, which occurs in an estimated 0.03–0.2% of cases. 139–143Mortality from this complication has been estimated to be 41–70%, 142–144a rate that can be influenced by a variety of factors (e.g. , age, pulmonary hypertension, coagulopathy, heparinization).

Sepsis is a potential complication of PA catheter residence, but its exact incidence is uncertain. Positive cultures of indwelling PA catheter tips are common 145–151and are often due to incidental contamination from skin flora, but in the context of septic patients it is frequently unclear whether positive cultures reflect colonization from another source or the primary nidus of infection. Incidence rates for culture-positive catheters, positive blood cultures, and catheter-related sepsis, therefore, vary considerably in the literature. In general, the incidence of infection increases with the duration of placement 152,153; the risk seems to increase significantly when catheters are in place for more than 72–96 h. 145,146,152–154Although the complication therefore has largely postoperative rather than intraoperative origins, poor technique at the time of insertion can introduce contamination. Skin flora are a common source of infection. 154–156Evidence regarding long-term infectious complications is limited. For example, invasive hemodynamic monitoring during placement of prosthetic aortic grafts does not seem to be associated with graft infection after 4 yr. 157 

The rate of iatrogenic deaths from PA catheterization is uncertain. Patients who are monitored by PA catheter have high mortality rates, but in few cases is it possible to attribute their deaths specifically to PA catheterization and not the underlying illness.

Currently available evidence from published research provides incomplete information about the effectiveness of PA catheter monitoring and the incidence of complications. Gaps in the evidence occur at several levels. First, surgical procedures that have been examined in studies of PA catheterization (e.g. , cardiac surgery, abdominal aortic reconstruction, neurosurgery) represent only a subset of clinical settings for which PA catheters are used. Second, available studies generally suffer from poor design and lack the statistical power to demonstrate benefit. In its 1993 review, the Task Force called for randomized controlled trials to offer more compelling evidence about the effectiveness of PAC. Although a number of such trials have since been published, they have not settled the issue, in part because of inadequate documentation and design flaws; Ivanov et al.  calculated a mean Chalmers quality score of 40 (on a scale of 1–100) for trials examining PAC. 93Third, studies without randomized designs generally do not control for differences in case mix and practitioner skill and, therefore, may not be generalizable to typical practice conditions. In one analysis, severity of illness was 7% greater for cardiac surgery patients receiving a PAC. 65Failure to account for the conditions with which PAC is associated often limits study validity.

Because of deficiencies in the evidence, it is difficult to draw meaningful conclusions about the effectiveness and safety of PA catheterization based on currently available data. In general, the evidence suggests that the routine use of PA catheters in low-risk patients does not reduce mortality, length of stay, or other surrogate markers for severity of illness. In some settings, the risks from the procedure may outweigh its benefits. The evidence does not exclude the possibility that PA catheterization improves outcomes in select clinical circumstances. Post hoc  data analyses in the above literature, although an inadequate evidence base for drawing firm conclusions, suggest that patient subgroups do benefit from interventions that include or require PA catheter monitoring.

There is a great need for additional research to provide this information. In the meantime, important insights about the benefits and harms of PA catheterization can be obtained from clinical experience. The Task Force acknowledges the limitations of expert opinion, which include subjectivity, recall bias, nonuniformity of measures, and confounding. With these caveats in mind, the following observations about the benefits and harms of PA catheterization are offered:

Clinical experience suggests that PA catheter monitoring of selected surgical patients can reduce the incidence of perioperative complications, primarily by providing immediate access to critical hemodynamic data. The expert opinion of the Task Force is that access to these data for selected indications and settings, coupled with accurate interpretation and appropriate treatment tailored to hemodynamic status, can reduce perioperative mortality and morbidity through reduced cardiac complications (e.g. , myocardial ischemia, congestive heart failure, arrhythmias), renal insufficiency, brain injury, and pulmonary complications. The Task Force also believes that, in selected cases and settings, PA catheter use may reduce length of stay in the hospital and ICU, enhance postoperative functional status, and reduce the need for transfused blood products through optimization of fluid therapy. The Task Force believes the use of PA catheters in selected obstetric patients may reduce maternal and neonatal morbidity and mortality.

Although these benefits may not be realized in every surgical patient who is catheterized, the Task Force believes that having immediate access to PA catheter data allows important preemptive measures for that subset of patients who encounter hemodynamic disturbances that require immediate and precise decisions about fluid management and drug treatment. The exact proportion of patients for whom this applies and the magnitude of benefit from PA catheterization are uncertain. However, the Task Force believes that reliance on clinical assessment or alternative devices (e.g. , CVC monitoring) is inadequate and that TEE, which can provide similar and important additional information, may be unavailable or impractical. The delay associated with PA catheter placement after complications have developed may endanger certain patients and may increase the risk of complications from insertion. Emergency catheterization under hastily prepared conditions may increase the risk of vascular injury and catheter-related sepsis. Prospective studies have found that the relative risk for catheter-related sepsis is 2.1 when PA catheters are inserted in the operating room using a less stringent sterile technique. 154 

Numerous studies have shown that PA catheter data are more accurate than clinical assessment in evaluating the hemodynamic status of complicated patients. 18,19,23–25For postoperative patients, who have less immediate access to physicians after leaving the operating room, PA catheter data and trained nurses may provide an important means of rapidly communicating precise information about hemodynamic status to physicians not at the bedside. This enables immediate execution of treatment decisions that are tailored to the patient's physiologic state.

The Task Force believes that these benefits have not been demonstrated in currently available research because of deficiencies in study design and performance. It is suggested that a properly designed randomized controlled trial with adequate sample size, well-trained physicians and nurses, well-defined interventions, and meaningful outcome measures would reveal the benefits observed in practice. To do this, trials must be carefully designed to distinguish between the incremental benefits offered by PAC and those attributable to other treatment interventions with which PAC is associated (e.g. , intensive care). It is recognized that the performance of such a study might be difficult because of logistical and ethical considerations, but calls for such research have intensified in recent years. 10,12 

The generalizability of research findings to practice conditions is often limited. For example, the Task Force believes that outcomes for experienced clinicians may differ significantly from published rates, because catheterizations in clinical studies were often performed under conditions in which interventions were not adequately standardized or implemented according to protocol, or in early years, when outmoded techniques and materials were used. The Task Force believes that experience and understanding are major determinants of PA catheter effectiveness. Experienced PA catheter users can achieve better outcomes and encounter fewer complications because of their enhanced skill in the interpretation of PA catheter data, in the prompt design of rational treatment strategies, and in the use of safe techniques of catheter insertion and management. Characteristics of the practice setting, such as the skill of nurses and their attentiveness to PA catheter tracings, also influence significantly the outcome of PA catheter use.

There is compelling evidence that PA catheterization can result in serious and potentially fatal complications. Catheterization is, therefore, inappropriate as a routine practice in surgical patients and should be limited to cases in which the anticipated benefits of catheterization outweigh the potential risks. The actual complication rate is uncertain, in part because of variability in the designs and results of studies of complications and in part because only a subset of reported complications is attributable uniquely to PA catheters. Serious complications also occur with the principal alternatives to PA catheterization, such as CVC (e.g. , unintentional arteriotomy, pneumothorax, line sepsis) or less frequently with transesophageal echocardiographic monitoring (e.g. , esophageal injury, vocal cord paralysis, dysrhythmias). Published rates for certain complications are based on old studies that may not reflect recent advances. For example, venous thrombosis has been reduced by the use of prophylactic heparin and heparin-bonded catheters. 137 

The opinion of the Task Force, based on clinical experience, is that serious complications due specifically to PA catheterization (PA rupture, serious ventricular dysrhythmias, endocardial lesions) occur in 0.1–0.5% of PAC-monitored surgical patients. Physicians dealing with other patient populations have made different estimates. In one survey, American cardiologists and other internists, when asked to estimate the probability of severe morbidity or death from PAC, gave higher estimates (2–5% and 0.5–1.0%, respectively) that seemed to differ by specialty. 158A study by the same group, comparing American and British physician perspectives on PAC, yielded estimates of 5% and 1%, respectively. 159 

The Task Force believes that the risk of PAC is both appropriate and necessary in selected surgical patients undergoing procedures associated with complications from hemodynamic changes (e.g. , cardiac surgery, aortic reconstruction) or entering surgery with preexisting risk factors for hemodynamic disturbances (e.g. , advanced cardiopulmonary disease). The level of hemodynamic risk should be assessed as a function of three interrelated variables: the health status of the patient, level of risk associated with the specific surgical procedure, and characteristics of the practice setting (e.g. , physician skills, technical support related to PAC). Consideration of the interrelationships between these three variables (fig. 2) aids in the accurate assessment of hemodynamic risk.

Although the recommendations in this report are based primarily on clinical benefits and harms, the following resource and implementation issues were considered by the Task Force in structuring its recommendations:

The costs of PA catheterization include the costs of equipment (e.g. , PA catheters, pressure transducers, electronic monitoring devices, solutions) and personnel (e.g. , physician costs for insertion and interpretation, nurses, technicians). There is limited information from published literature about the actual costs of PA catheterization. Published estimates of charges for the procedure range widely, from $300 to $1,649. 66,160,161Yet another study estimated the PAC cost to be $667 on the first day of catheterization and $541 for each additional day. 162An analysis of 13,907 patients who underwent nonemergent coronary artery graft surgery in 1997 found that even after regression analysis for case mix and other covariables, total hospital costs were significantly higher for PAC-monitored patients than for patients not receiving PAC, a difference of $1,402. 65 

In the previously mentioned analysis of 5,735 critically ill patients during the first 24 h of admission to an ICU, PAC use was associated with significantly higher total hospital costs ($7,900; SE, $3,900) even after multivariate adjustment for confounding variables. 9A logarithmic formula was used to adjust costs for the portion of the hospital stay spent in the ICU. The authors speculated that a large part of the cost related to the association between PAC use and other expensive technologies, and to increased nursing care. In their study, patients with PAC spent 2 days longer in the ICU, and the average intensity of care was four to seven points higher on the Therapeutic Intervention Scoring System, 163but the generalizability of such findings to perioperative settings is limited.

Perhaps a more pertinent issue is the incremental cost of PAC compared with alternatives. A study of 194 patients who underwent coronary artery bypass surgery found that the difference in total hospital charges (1996 dollars) for patients monitored by PAC or CVC ($31,300 and $28,900, respectively) lacked statistical significance, which for demonstration would require a sample size of 778. 64 

More important than whether PAC increases hospital costs is its cost-effectiveness, which considers the health benefits of the procedure to determine whether resources are being spent wisely. There have been few published cost-benefit or cost-effectiveness analyses of PA catheter monitoring, and none regarding its use in the perioperative setting. An analysis of PAC use in patients with chronic obstructive pulmonary disease estimated that its incremental cost-effectiveness was $77,407 per quality-adjusted life year saved (if changes based on the information improve survival by 5%). 162A procedure's cost-effectiveness cannot be properly ascertained without establishing its clinical effectiveness, a more fundamental uncertainty with PAC, 164and until the latter is resolved estimations of cost-effectiveness can be based only on speculative assumptions.

In addition to clinical benefits, harms, and costs, the Task Force also considered patient and provider concerns that could influence implementation of the guideline:

Disadvantaged patients who meet criteria for PA catheterization may lack access to hospitals with facilities and qualified personnel to perform the procedure. This is especially true for patients in minority or disadvantaged populations, who have experienced documented disparities in care, and for those with inadequate insurance coverage. 165Although in most cases the urgency of conditions requiring PA catheterization does not lend itself to discussions beyond the basic requirements for informed consent, patients undergoing procedures for which the use of PAC is elective may benefit from the opportunity to review the indications for using the device and to decide, based on personal preferences, whether to proceed.

Preoperative catheterization, which as noted earlier lacks compelling evidence of benefit, may also be less desirable to patients. Insertion of the PA catheter in the operating room may be more comfortable, less anxiety-provoking, and induce less physiologic stress than preoperative insertion. 166 

The appropriateness of PA catheterization and the determination of whether benefits exceed risks hinge on the competence of physicians and nurses in catheter use. This competence encompasses both technical and cognitive skills, which are first acquired in residency or postresidency training. Maintenance of skills following training often requires regular catheter use, but there is disturbing evidence that skill levels are inadequate. A 1990 study by Iberti et al. , in which a 31-item examination on PA catheters was completed by 496 North American physicians, found that only 67% of the answers were correct. 167The instrument yielded similar results in Europe. 168A 1996 survey of more than 1,000 critical care physicians found that, although 83% of questions were answered correctly, a third of the respondents could not correctly identify PA occlusion pressure on a clear tracing and could not identify the major components of oxygen transport. 13 

Similar problems have been identified among critical care nurses. Exposure to the subject in nursing school is limited, 169and surveys of practicing nurses demonstrate knowledge deficits. A 31-item examination of critical care nurses in California found that only 57% of the responses were correct. 170Only 39% of respondents correctly identified a PA wedge measurement value from a waveform recording. Most of the nurses (95%) had more than 1 yr of experience in critical care, and 99% used the PAC more than once per month. Scores were better for nurses with CCRN certification, attendance at a PA catheter class, more years of critical care experience, and frequent PA catheter use.

Because PA catheterization by persons who have not maintained these skills is potentially harmful to patients and could threaten the acceptability of the procedure, it is important for the profession to periodically assess technical and cognitive performance. Recognition of the need to strengthen quality control and competency has grown in recent years. 133,171The best measure of competence is clinical outcome, but long periods of observation and careful data analysis may be necessary to obtain meaningful information. Surrogate measures such as the frequency of catheter use or the results of proficiency examinations may be the best alternative, but they are imperfect measures of competence.

Reimbursement policies play a role in the ability of providers to offer PAC. In addition to other factors, the evidence that PAC is only appropriate for certain indications and should therefore not be used as a matter of routine in the perioperative setting makes it inappropriate to assume that PAC is part of the surgical procedure or for payers to bundle it in reimbursement. 172 

Because of limitations in scientific evidence about the limits of appropriateness for PA catheterization, guidelines based on expert opinion should not be used as standards of care or to define cases of unnecessary catheterization.

The evidence reviewed to date clearly does not support the routine use of PA catheters when there is a low risk of hemodynamic complications. In its 1993 report, the Task Force indicated that such risk is a function of three interdependent variables: the patient, procedure, and practice setting (fig. 2). When these conditions culminate in a high-risk situation is both subjective and variable and is influenced by interpretations of the scientific evidence, individual circumstances of the case, and local conditions.

To provide further guidance on when these conditions arise, the Task Force turned to expert opinion because current data are inadequate to establish firm evidence-based criteria. The Task Force's expert opinion, assessed using the voting process and the criteria for appropriateness and necessity described in Methodology, represents its best judgment on when PA catheterization is appropriate or necessary. The votes on which the recommendations are based are tabulated in tables 6 and 7, and the definitions for terms used in the ballot are in table 2. In certain cases, the alternative to PA catheterization should be CVC, TEE, or other less invasive monitoring methods (e.g. , esophageal Doppler, pulse wave analysis, bioimpedance, carbon dioxide Fick, lithium dilution), rather than no hemodynamic monitoring. The degree of appropriateness of PA catheterization may differ in circumstances when these alternatives are available.

The votes by the Task Force (tables 6 and 7) demonstrated that the appropriateness of routine PA catheterization depends on the combination of risks associated with the (a) patient , (b) surgery , and (c) practice setting  (the latter referring to the risks from PA catheterization introduced by practice conditions and staff circumstances). The votes are depicted graphically in figure 3. With some exceptions, routine catheterization is generally inappropriate for low- or moderate-risk patients . The three variables are defined in greater detail below.

Patients at increased risk for hemodynamic disturbances are those with clinical evidence of significant cardiovascular disease, pulmonary dysfunction, hypoxia, renal insufficiency, or other conditions associated with hemodynamic instability (e.g. , advanced age, endocrine disorders, sepsis, trauma, burns). Patients at low risk  include those with American Society of Anesthesiologists (ASA) physical status 173of 1 or 2 or those with hemodynamic disturbances unlikely to cause organ dysfunction. Those at moderate risk  are in category ASA 3 or have hemodynamic disturbances that occasionally cause organ dysfunction. Those at high risk  are in category ASA 4 or 5 and have hemodynamic disturbances with a great chance of causing organ dysfunction or death. The assessment of risk should be based on a thorough analysis of the medical history and physical examination findings, rather than on exclusive consideration of specific laboratory results or other quantitative criteria.

Surgical procedures associated with an increased risk of complications from hemodynamic changes, including damage to the heart, vascular tree, kidneys, liver, lungs, or brain, may increase the chance of benefiting from PA catheterization. This report does not provide a list of indicated procedures and disease states for catheterization because the Task Force believes that catheterization decisions should be based on the hemodynamic risk characteristics of the individual case rather than on the type of procedure. The Task Force defines low-risk  procedures as those carrying a small probability of fluid changes or hemodynamic disturbances and having low perioperative morbidity or mortality. Moderate-risk  procedures have a moderate chance of fluid changes, hemodynamic disturbances, or infection that could cause morbidity or mortality. High-risk  procedures have a predictably large chance of fluid changes or hemodynamic disturbances or other factors with high risk of morbidity and mortality.

Patients undergoing procedures that usually lack hemodynamic complications may need PA catheterization if circumstances pose a special risk. The clinician should therefore assess hemodynamic risks based on the case at hand and not on generic criteria.

The setting for the procedure may increase the risk of complications from hemodynamic changes. Factors that should be considered in assessing perioperative risk include catheter use skills and technical support. Factors affecting postoperative risk include the level of training and experience of nursing staff in the recovery room and ICU, technical support for ancillary services, and the availability of specialists and equipment to manage potential complications detected by the PA catheter.

To assess whether panel votes differed distinctly from the views that might be held by others with expertise in PA catheterization, the same questionnaire used by the panel was administered to 12 anesthesiologists attending an open forum about the guideline at the 2001 meetings of the International Anesthesia Research Society and the Society for Cardiovascular Anesthesiologists. The results suggested significant concordance in views.

In judging appropriateness , the median scores of the attendees were often identical and within the same range as the panel's vote in 22 out of 27 scenarios. When their results were pooled with the panel votes, conclusions about the appropriateness of routine PA catheterization shifted for only three scenarios. Specifically, they shifted for low-risk patients undergoing high-risk surgery in a moderate-risk setting, moderate-risk patients undergoing moderate-risk surgery in a moderate-risk setting, and high-risk patients undergoing moderate-risk surgery in a high-risk setting. In these cases the panel had voted that routine PA catheterization was inappropriate, but the combined votes that included open-forum attendees placed these circumstances in the “possibly appropriate” range (median score of 4–6) but the median score did not exceed 4.

In judging necessity , the votes taken by the attendees fell within the same range as the panel's median scores for 17 out of 27 scenarios. When these were pooled with the panel votes, conclusions about necessity shifted definitively for only four scenarios. In two scenarios for which the panel considered routine PA catheterization unnecessary (median score, 1–3)—moderate-risk patients undergoing high-risk surgery in a moderate-risk setting and high-risk patients undergoing high-risk surgery in a high-risk setting—the median combined vote was 5, suggesting that necessity was uncertain (range, 4–6). In two circumstances for which the panel considered necessity uncertain—high-risk patients undergoing either moderate-risk surgery in a low-risk setting or high-risk surgery in a moderate-risk setting—the median combined vote was 7 and 8, respectively, suggesting that PA catheterization was necessary (range, 7–9).

There is insufficient evidence to support preoperative PA catheter monitoring (e.g. , the evening before surgery) in patients who are hemodynamically stable. The risk of catheter-related sepsis and other complications increases with the duration of catheter residence. It is recognized that preoperative PA catheterization is often under the control of other medical specialists, such as cardiologists, pulmonologists, and surgeons.

Because of the risk of complications from PA catheterization, the procedure should not be performed by clinicians who lack competence in safe insertion or in the accurate interpretation of results. Competence in catheter placement is related to experience and is a function of the quality of the initial training and regular performance of the procedure after training.

All persons who use PA catheters should undergo high-quality, supervised training to establish competency. Across specialties, however, there is little agreement about the minimum number of procedures that must be performed to acquire cognitive and technical skills. The American College of Physicians, American College of Cardiology, and American Heart Association hold that at least 25 procedures must be performed to acquire technical and cognitive skills in hemodynamic monitoring (defined as pressure measurement in the PA, central venous system, and arterial system and the measurement of cardiac output). In a national survey of critical care physicians, 73% believed that 20–50 supervised PA catheterizations were required to establish competency. 13 

However, learning curve research, conducted primarily in gastroenterology, has found that different and variable levels of experience are required to establish competency, depending on the type of procedure: 25–30 procedures for flexible sigmoidoscopy, 17530 for laparoscopic hernia repair 176and pediatric caudal epidural blocks, 17740 for endoscopic sphincterectomy, 178130 for upper endoscopies, 179100–200 for colonoscopy, 179–182and 40–200 for endoscopic retrograde cholangiopancreatography. 183–185Citing such evidence, a 1998 report by the American Society for Gastrointestinal Endoscopy (Oak Brook, Illinois) acknowledged that its previous criteria on the minimum number of procedures required to attain competency (which ranged from 5 to 100, depending on the procedure) “are not adequate for most trainees.”186 

Moreover, even groups that require a minimum number of procedures acknowledge that more or less than a nominal training requirement may be needed, depending on case severity and other individual training circumstances. The use of simulators and software-based decision aids 187may improve knowledge and reduce experiential requirements in actual cases. In recognition of these variables, although the aforementioned minimum requirements for establishing competency in PA catheterization can help in decision-making about training requirements, it is important that other factors be considered.

A quality improvement program must be in place at all sites where PA catheters are used. Ideally, maintenance of knowledge and skills should be evaluated based on clinical outcomes. It is widely held that such skills cannot be maintained without performing some minimum number of procedures per year. For example, in a survey of 1,095 critical care physicians, 69% of the respondents believed that physicians must perform 10–25 PA catheterizations per year to maintain competence. 13Similarly, four out of six attendees at the open forum convened by this panel at the Society for Cardiovascular Anesthesiologists indicated that 10–25 PA catheter insertions per year were necessary to maintain competence. Swan has suggested that 50 procedures must be performed annually. 188 

Allowances for smaller case loads have been made by other groups. The American College of Physicians, American College of Cardiology, and American Heart Association state that “physicians with extensive experience who work in hospitals where these procedures are done infrequently may be able to maintain competence with a minimum number of continuing procedures, perhaps as few as 12 per year. The same groups, however, recommend that 100 echocardiograms must be performed annually to maintain competence. 189,190Requirements for credentialing at individual hospitals vary widely. One survey found that the required number of procedures per year required for cardiologists to be recredentialed in angioplasty and coronary angiography were 10–150 and 12–300, respectively. 191 

The wide variation in requirements reflects, in part, the absence of good data to inform such policies. 191Although studies have demonstrated correlations between anesthesiologists’ self-reported comfort level in giving anesthetics and the number of procedures they perform each year, 192there is no scientific proof for the belief that procedural competence, once established in initial training, decays significantly over time if the procedure is performed infrequently. 193Studies at an institutional level consistently demonstrate an inverse relationship between complication and volume rates×a pattern confirmed in the administration of anesthesia 193×but factors other than the procedural competence of physicians (e.g. , hospital characteristics, population profiles, referral preferences) may account for such observations. 194 

Competence in the interpretation of catheter data should be based on cognitive and technical requirements, such as those outlined by the American College of Physicians/American College of Cardiology/American Heart Association Task Force on Clinical Privileges in Cardiology (table 8).

Individuals with limited experience (e.g. , house staff, nurse anesthetists) should not insert PA catheters or make clinical decisions based on the data without qualified supervision. Nurses who provide care to patients who are monitored by PA catheters should be required to meet minimum training requirements and continuing education for catheter maintenance and for the interpretation and communication of PA catheter data. The nursing literature describes a number of programs to evaluate and maintain PA catheter skills among critical care nurses. 195,196 

Additional research is needed to demonstrate the effectiveness of PA catheterization. Deficiencies in current evidence suggest that future studies should emphasize certain design features to provide more meaningful evidence. Future studies should use relevant clinical outcome measures to judge effectiveness, including measures of morbidity (e.g. , cardiac, pulmonary, renal, and neurologic disease; patient functional status after discharge) and mortality (e.g. , in-hospital fatality rates, 5-yr survival, quality-adjusted life years).

Future studies should include adequate sample sizes to demonstrate statistically significant effects. Investigators should perform power calculations before initiating studies to confirm the adequacy of the sample size. Researchers should collect adequate information about severity of illness and comorbid variables to help separate the influence of these factors from the effects of PA catheterization on observed clinical outcomes. Studies must be clear about the endpoints of treatment, specifying the hemodynamic indices that clinicians use to make decisions based on PA catheter data. Published reports from studies should provide complete information about methodology and complete data reporting to minimize confusion about statistical calculations.

Many authors have emphasized the need for well-designed randomized controlled trials to test the effectiveness of PA catheterization. 197,198The Task Force continues to support the need for such research, as it did in 1993, but lessons learned from trials published in recent years suggest that formidable design challenges must be overcome if the findings from such trials are to gain acceptance and influence policy. There are fundamental logistical and ethical problems with randomizing the allocation of PA catheters in critically ill patients. Of particular concern is the potential contamination of control groups. This crossover effect has occurred frequently in past research when patients assigned to receive no monitoring or CVC monitoring were subsequently reassigned by their physicians to undergo PA catheterization because of clinical deterioration or because the physicians were uncomfortable with the group assignment. In a Canadian randomized trial, 52 of 148 potentially eligible patients were excluded because the attending physician believed that right heart catheterization was “ethically mandated,” leaving only 33 patients available for randomization. 199 

Ethicists argue that patients should be subjected to randomization only under conditions of clinical equipoise , 200in which the profession has genuine uncertainty regarding the relative merits of each therapeutic arm in the trial, 201rather than under conditions of a prevailing belief of benefit, ineffectiveness, or harm. To meet this ethical standard and to minimize crossover, the Task Force therefore recommends that future randomized controlled trials should be undertaken in settings where the profession has split opinions about the appropriateness of PA catheterization.

To determine what these conditions are, the Task Force examined the spectrum of clinical scenarios considered in its voting exercise and identified those conditions that elicited the greatest heterogeneity in views, producing a median vote of 4–6. This occurred in five scenarios: Assuming that risks from the practice setting (e.g. , catheter-use skills) are low, views about appropriateness differed substantially for (1) low-risk patients undergoing high-risk surgery, (2) moderate-risk patients undergoing moderate-risk surgery, and (3) high-risk patients undergoing low-risk surgery (fig. 4). The Task Force also expressed heterogeneous votes when (4) the practice setting imposes moderate risk (when moderate-risk patients undergo high-risk surgery in moderate-risk settings), and when (5) high-risk patients undergo high-risk surgery in high-risk settings (fig. 5). The equipoise observed for these five clinical settings offer potential parameters for conducting an ethically justifiable randomized trial of PA catheterization.

Such a trial should probably be restricted to a specific type of surgery (e.g. , coronary artery bypass graft). Physicians participating in such studies should adhere closely to the protocol and should not violate the design by reassigning patients following randomization. Studies that use low-risk patients may require large sample sizes to demonstrate statistically significant results, depending on power calculations; therefore, multicenter involvement may be necessary. The study should be performed in a setting attuned to the management of postoperative hemodynamic status, and the control groups should be monitored by CVC or other appropriate alternate methods (e.g. , TEE).

Although a number of randomized trials involving PA catheterization have been published in recent years, deficiencies in their design and conduct have limited their internal and external validity. For example, few trials have included concealment of allocation, the most important source of bias in trial design. 202Blinding, although infeasible for patients or clinicians at the bedside, is possible and would reduce bias among those who measure outcomes, but this has rarely been attempted in PAC trials. Violation of intention-to-treat analysis and overindulgent post hoc  data analysis have been commonplace.

Moreover, the questions that trials in the past decade have sought to answer have lacked relevance to perioperative care. Most trials have focused on patients in ICUs and have yielded data of only tangential relevance to conditions in the operating room. A 1997 consensus conference 12was successful in convincing the National Institutes of Health to fund multicenter trials of PA catheterization, but the trials that were ultimately launched focus on adult respiratory distress syndrome 203and congestive heart failure. 204–213The results of these trials will have limited bearing on perioperative care, in which 55% of PA catheters are used in the United States. 12 

A further limitation of past trials has been ambiguity about the intervention. Treatment groups have often received interventions related to, but distinct from, PAC (e.g. , admission to ICU), making it difficult to draw firm conclusions about the role PA catheterization plays in observed benefits or harms. For future trials to overcome this limitation, patients must be assigned to groups in a way that disentangles the incremental influence of component interventions. To know whether the treatments initiated by PAC use are themselves responsible for observed benefits and harms, those treatments must be described explicitly, replacing opaque descriptions of types of treatments with detailed explication of treatment targets (e.g. , physiologic endpoints) and how well they were achieved.

The optimal randomized controlled trial will provide evidence from only one class of patients and one type of surgery. To provide more comprehensive data about the effects of PA catheterization, the Task Force recommends the performance of a large-scale multicenter observational study that (1) includes adequate numbers of patients for each indication; (2) collects adequate data to generate a comprehensive database regarding past medical history, comorbidity, severity of illness, hospital course, and immediate and long-term clinical outcomes; and (3) provides details about catheter insertion techniques, practitioner skills, and data interpretation. For example, rather than stating that PA catheter monitoring was performed for 48 h, the investigators should document the type of device, the specific hemodynamic variables that were measured, and the manner in which PA catheter data were used to make treatment decisions.

Ultimately, the state of knowledge will need to move beyond expert opinion to provide objective evidence about the benefits and harms of PA catheterization.

After acceptance of these guidelines by the ASA House of Delegates, an important article concerning the use of pulmonary artery catheters in surgical patients was published (Sandham JD, Hull RD, Brant RF, Knox L, Pineo, GF, Doig CJ, Laporta DP, Viner S, Passerini L, Devitt H, Kirby A, Jacka M: A randomized, controlled trial of the use of pulmonary-artery catheters in high-risk surgical patients. N Engl J Med 2003; 348:5–14). The article was accompanied by an editorial (Parsons PE: Progress in research on pulmonary-artery catheters. N Engl J Med 2003; 348:66–8). Because the article was published after the guidelines were written and approved, the authors feel it is inappropriate to comment on this work other than to express the opinion that its findings are important and are consistent with the guidelines.