Pulmonary artery catheterization and mortality in critically ill patients{dagger}

S. D. Murdoch, A. T. Cohen and M. C. Bellamy*

Department of Anaesthesia and Intensive Care Medicine, St James’s University Hospital, Beckett Street, Leeds LS9 7TF, UK

{dagger}This article is accompanied by Editorial I.

Accepted for publication: May 2, 2000


    Abstract
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Pulmonary artery catheters are widely used in intensive care, but evidence to support their widespread use is sparse. Some published data suggest that greater mortality is associated with use of these catheters. The largest study to date looked at >5500 patients in several centres in America and found a greater 30 day mortality in those patients receiving a pulmonary artery catheter. We tested the hypothesis that, on our intensive care unit, mortality was greater for those patients receiving a pulmonary artery catheter. Using a propensity score to account for severity of illness, the odds ratio for mortality in those patients receiving a pulmonary artery catheter was 1.08 (95% confidence interval 0.87–1.33). We believe that continued use of the pulmonary artery catheter is safe; a large randomized controlled trial examining outcome is unlikely to provide an adequate answer.

Br J Anaesth 2000; 85: 611–5

Keywords: lung, pulmonary artery catheter; intensive care, mortality


    Introduction
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Since its introduction in 1970, the pulmonary artery catheter1 has become commonplace in intensive care practice. Pulmonary artery catheterization was first described in 19452 and from then until the 1970s was used in the catheter laboratory as a diagnostic tool. When it became possible to perform the procedure at the bedside in critically ill patients in the intensive care unit, the balloon-tipped pulmonary artery catheter became widely used as a guide to treatment. In 1996, over two million catheters were sold worldwide.3

The effectiveness of new techniques and treatments should ideally be evaluated before they are introduced,4 but this is not always done. The pulmonary artery catheter was not evaluated in this way before being introduced. Despite this, many physicians believe it can guide treatment and provide useful physiological data. Several well conducted studies have shown an improvement in outcome when acting on data provided by the pulmonary artery catheter.5–7 However, greater mortality has been demonstrated in some patients, most notably in those who have had a myocardial infarction.8 9 It has proved difficult to conduct a randomized, controlled trial of pulmonary artery catheter use.10

In September 1996, Connors and colleagues published a paper examining the use of the pulmonary catheter in 5735 intensive care patients.11 This multicentre study found greater mortality in patients receiving a pulmonary artery catheter. The paper was based on data collected as part of the SUPPORT12 study, which was designed to look at decision-making and outcomes in patients with an estimated 6 month mortality of 50%. The study took place between 1989 and 1994 in five American teaching hospitals with patients in one or more of nine disease categories: acute respiratory failure, chronic obstructive pulmonary disease, congestive heart failure, cirrhosis, non-traumatic coma, colon cancer metastatic to the liver, non-small-cell cancer of the lung (stage III or IV) and multiple organ system failure with malignancy or sepsis. Not all patients with these conditions would commonly be given pulmonary catheters in European practice. The statistical techniques used by Connors and colleagues included the propensity score13 to allow comparison between patients receiving or not receiving a pulmonary artery catheter while compensating for severity of illness and treatment selection bias.

We examined the effect of the pulmonary artery catheter on intensive care mortality in all adult patients admitted over 7 yr to an intensive care unit in a large British teaching hospital. A propensity score was used to compensate for severity of illness between groups.


    Materials and methods
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Data have been collected prospectively on all patients admitted to the intensive care unit at St James’s University Hospital, Leeds, since 1984. The unit is based in a major teaching hospital and accepts patients from all specialties except neurosurgery and cardiothoracic surgery. The unit provides a regional tertiary referral service for patients with respiratory and renal failure, and a supraregional liver failure and transplant programme.

We examined patient characteristics, physiological, treatment and outcome data for 4182 patients aged >16 yr admitted between January 1990 and December 1996. Patient characteristics available included age, gender, race (Caucasian or non-Caucasian), length of intensive care stay, the need for readmission, APACHE II (Acute Physiology and Chronic Health Evaluation) score,14 presence of acute renal failure, elective/emergency admission and physician’s prediction of death. Treatment variables included insertion of a central venous catheter, insertion of an arterial cannula, treatment with epinephrine, norepinephrine, dopamine or dobutamine (separate variables) and the need for ventilation of the lungs. Outcome data were discharge from the intensive care unit or death on the intensive care unit. APACHE II data were based on the first 24 h of intensive care admission. The physician’s prediction of death was a dichotomous variable, with patients being predicted to survive or predicted to die.

Following a univariate analysis, 15 variables were entered in a backward stepwise logistic regression analysis. Statistically significant variables, identified from the regression analysis, were used to calculate the propensity score. Variables in the final equation were acute renal failure, need for treatment with epinephrine, norepinephrine, dopamine and dobutamine, monitoring of arterial pressure and central venous pressure, physician’s prediction of death and patient age. Variables excluded from the equation included gender, race, APACHE II score, duration of intensive care stay, need for elective (as opposed to emergency) admission and the need for subsequent readmission to the intensive care unit. The significant variables were entered into a logistic regression analysis with pulmonary artery catheter insertion during intensive care stay as the dependent variable. This gave the probability (from 0 to 1) of pulmonary artery catheter insertion occurring during a patient’s intensive care stay, the ‘propensity score’. Statistical analysis was performed using SPSS for Windows version 6, the receiver operator characteristic (ROC) plot was produced by Astute (University of Leeds) on Microsoft Excel 4, all running on Microsoft Windows 98 on a Dan 120 MHz Pentium computer.

The accuracy with which a propensity score predicts a treatment option can be tested using the ROC plot. The area under the ROC curve corresponds to the probability of the propensity correctly predicting if a patient receives a treatment or not (in this case a pulmonary artery catheter).

Bias in the propensity score can be investigated by examining subclasses of data based on the propensity score. If the score is valid it compensates for the covariates; within each class there should be only random differences in the distribution of covariates, which cancel each other out. This means that within each stratum, no factor retains a relationship with the exposure.15 It is sufficient to look at only five equal groups, known as quintiles.16 Quintiles based on the propensity score were examined to determine the distribution of the predictive variables for pulmonary artery catheter insertion.

The predictive value of the propensity in predicting the insertion of the pulmonary artery catheter was tested using a ROC curve.16 The calculated propensity score and pulmonary artery catheterization were then investigated as the two potential predictors of mortality in intensive care, in a logistic regression model.

Subgroup analysis of the effect of the pulmonary artery catheter on mortality was performed based on (i) patients’ APACHE II score on admission and on (ii) elective compared with emergency admission.


    Results
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
There were 2456 male patients and 1726 females; their median age was 56.3 yr (interquartile range 42.8–70.5 yr), median APACHE II score 19 (interquartile range 12–27) and median length of stay 2 days (interquartile range 1–4 days). Further details of the patients are given in Table 1.


View this table:
[in this window]
[in a new window]
 
Table 1 Characteristics of 4182 intensive care patients
 
Patients who had a pulmonary artery catheter inserted were more likely to be younger, have renal failure, be treated with inotropes, have acute renal failure and have a prolonged stay in intensive care stay. Patients without a pulmonary artery catheter were more likely to be older, admitted electively and have a lower APACHE II score. There was no difference in the two groups with respect to gender, racial origin or previous intensive care unit admission.

There was an overall mortality of 1087/4181 (26%), comprising a mortality of 293/2333 (12.6%) in those without a pulmonary artery catheter and 794/1849 (42.9%) in those with a pulmonary artery catheter (chi-squared P<0.0001). The pulmonary artery catheter was, therefore, strongly associated with mortality when no adjustment had been made for severity of illness.

The propensity score had an area under the ROC curve of 0.88, indicating good predictive value for insertion of the pulmonary artery catheter (Connors and colleagues obtained a propensity score of 0.83).

The median propensity score was 0.797 (interquartile range 0.456–0.947) for patients who had a pulmonary artery catheter inserted and 0.176 (interquartile range 0.086–0.337) for those who did not have a pulmonary artery catheter inserted.

The calculated propensity score and pulmonary artery catheterization were then investigated as the two potential predictors of mortality, in intensive care, in a logistic regression model. The cut-off value for the propensity score was 0.5. The propensity score was predictive of death, (odds ratio (OR) 45; 95% confidence interval (CI) 34.7–58.3). Use of the pulmonary artery catheter was not predictive of death (OR 1.08; 95% CI 0.87–1.33).

Analysis of patients according to APACHE II score did not demonstrate a deleterious effect of pulmonary artery catheter use in any group.

In patients with an APACHE II score of <20, the presence of a pulmonary artery catheter was not predictive of death (OR 1.14; 95% CI 0.7–2.5), but the propensity score was (OR 75.8; 95% CI 34.8–164.7). In patients with an APACHE II score of 20–25, pulmonary artery catheter insertion was not predictive of death (OR 1.3; 95% CI 0.8–2.15) but the propensity score was (OR 11.27; 95% CI 5.3–24). In patients with an APACHE II score of >25, the pulmonary artery catheter was not predictive of death (OR 0.89; 95% CI 0.64–1.25) but again propensity score was predictive (OR 7.43; 95% CI 4.48–12.3).

In patients admitted electively to the intensive care unit, the pulmonary artery catheter was not predictive of death (OR 1.9; 95% CI 0.57–6.14), but propensity score was predictive (OR 330; 95% CI 54.2–2006).

Quintile analysis showed that the predictor variables were evenly distributed between interventional groups, with the exception of central venous and arterial cannulae in the lowest quintile. This is unlikely to be important in the absence of a treatment effect from the pulmonary artery catheter.


    Discussion
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
The data we have examined show no increase in mortality in those patients receiving a pulmonary artery catheter, nor do we demonstrate a beneficial effect of its use. These results were observed after correction for treatment bias using a propensity score. In observational studies, patients are not assigned to treatment groups randomly. It is, therefore, inevitable that some patient variables are over-represented in one group. Consequently, it is difficult to interpret direct comparison between the two groups.

Previous observational studies found an increase in mortality in patients receiving a pulmonary artery catheter. It has been claimed that patients who were sicker and more likely to die received the pulmonary artery catheter and this underlying severity of disease led to the increase in mortality, rather than the pulmonary artery catheter itself.9 The ideal is a randomized controlled trial, where all variables are represented equally in each treatment group. Where such a trial is not possible, a technique used to overcome this problem is the propensity score, which describes the probability of assignment to a given treatment given a vector of observed covariates. Propensity score is calculated using logistic regression analysis of the variables that lead to a given treatment. It is dependent on identifying variables leading to exposure and is liable to bias owing to failure to include important variables that lead to exposure. In effect, the propensity score represents the probability that a patient will receive a given treatment, taking into account the individual patient’s variables that are believed to lead to that treatment. The propensity score can then be used with the treatment it predicts in order to investigate the effect of the treatment, independently of the variables leading to that treatment.

The pulmonary artery catheter is a monitoring device; it is generally used to obtain haemodynamic data on patients which may be used to guide treatment. Any possible effect on mortality could result from a direct effect or from the way in which the information obtained from the catheter is used to guide treatment.

Direct effects can be divided into the immediate problems of central line insertion (damage to vessels,17 infection,18 cardiac arrhythmias, etc.) and the later risk of infective endocarditis. It is unlikely that these complications are responsible for the increased mortality previously reported with pulmonary artery catheters. The reported increase in mortality with the use of pulmonary artery catheters is probably, therefore, an indirect effect of their use or caused by problems with studying their use.

The therapy received by a patient plays a major role in determining the outcome. In therapies based on data obtained from the pulmonary artery catheter, the catheter may act as an indicator of excess mortality of that therapy. It is easier to blame the catheter than the treatment for which it is being used. The use of supranormal oxygen delivery was once an aim of treatment guided by data obtained from the pulmonary artery catheter. A recent study has demonstrated that this therapeutic regimen may be harmful to patients.19 As the Connors study was multicentre and observational, it is possible that, in some patients in that study, the pulmonary artery catheter was used to obtain supranormal oxygen delivery.

The use of variables in our propensity score indicating the use of inotropes allows compensation to be made in the final analysis reflecting therapeutic intervention. This use of inotropes was recorded at any point during a patient’s illness on the intensive care unit, and so reflected the change in a patient’s condition over time and their response to treatment. The variables used in the construction of the propensity score in the Connors paper did not reflect therapeutic intervention but predominantly reflected the patient’s physiological status on admission to the intensive care unit.

Interpretation of the data obtained from the pulmonary artery catheter has been shown to be poor in studies in America20 and Europe.21 There are differences between intensive care practice in France and America,22 and probably also between the UK and USA. It is likely that some patients in the American study would not have been admitted to intensive care in the UK. The SUPPORT study showed differences in medical practice between the five hospitals studied and between different physicians. The results we present are obtained from one intensive care unit, where practice is consultant led and protocol driven. We use pulmonary artery catheters extensively.23

We studied all adult patients admitted to an intensive care unit, classed all patients who had a pulmonary artery catheter inserted at any time in the pulmonary artery group and examined patients’ need for cardiovascular and renal support throughout their stay. This provides an overall examination of the effects of the pulmonary artery catheter. Connors and colleagues produced a subgroup analysis based on diagnosis but we did not attempt this because we could not match our groups with those of Connors and colleagues. Smaller groups would have a different propensity score for pulmonary artery catheter insertion (leading to lack of power in the analysis). Modelling treatment effects would also be more difficult in small groups of patients.

While we have not demonstrated an increase in mortality attributable to the use of pulmonary artery catheters, we have failed to demonstrate a beneficial effect. This may be because there is no beneficial effect. It could also be because the benefit accruing from additional clinical information has little effect on individual patients, but rather informs practice as a whole. Would the results in patients who did not have pulmonary artery catheters have been as good without the accumulated experience resulting from years of using pulmonary artery catheters and familiarity with the likely haemodynamics in a given clinical situation? If we were to abandon the pulmonary artery catheter, would this skill base decay and, if so, over how long? It is unlikely that a simple head-to-head trial could answer such questions. A randomized, controlled trial of pulmonary artery catheters is required to assess their benefit, but our results suggest that such a study, even with very large numbers of patients, would have insufficient power to demonstrate an effect on mortality. Organizational differences between intensive care units, together with the change in practice that might occur as a result of participation in such a trial, would need to be taken into account in the analysis of such a study.


    Footnotes
 
* Corresponding author Back


    References
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
1 Swan HJC, Ganz W, Forester J, Marcus H, Diamond G, Chonette D. Catheterizaton of the heart in man using a flow directed balloon tipped catheter. New Engl J Med 1970; 283: 447–51[ISI][Medline]

2 Dexter L, Haynes FW, Burwell CS, Eppinger EC, Seibel RE, Evans JM. Studies of congenital heart disease. I. Techniques of venous catheterization as a diagnostic procedure. J Clin Invest 1947; 26: 547–53[ISI]

3 Ginosar Y, Sprung CL. The Swan Ganz catheter: twenty five years of monitoring. Crit Care Clin 1996; 12: 771–6[ISI][Medline]

4 Cook DJ, Sibbald WJ, Vincent J-L,Cerra FB. Evidence based critical care medicine: what is it and what can it do for us? Crit Care Med 1996; 24: 334–7[ISI][Medline]

5 Boyd O, Grounds RM, Bennett ED. A randomized clinical trial of the effect of deliberate perioperative increase of oxygen delivery on mortality in high risk surgical patients. J Am Med Assoc 1993; 270: 2699–707[Abstract]

6 Wilson J, Woods I, Fawcett J, Whall R, Dibb W, Morris C, McManus E. Reducing the risk of major elective surgery: randomised controlled trial of preoperative optimisation of oxygen delivery Br Med J 1999; 318: 1099–103[Abstract/Free Full Text]

7 Mimoz O, Rauss A, Rekik N, Brun-Bruission C, Lemaire F, Brochard L. Pulmonary artery catheterization in critically ill patients: a prospective analysis of outcome changes associated with catheter prompted changes in therapy. Crit Care Med 1994: 22; 573–9

8 Gore JM, Goldberg JM, Spodick DH, Alpert JS, Dalen JE. A communitywide assessment of the use of pulmonary artery catheters in patients with myocardial infarction. Chest 1987; 92: 721–7[Abstract]

9 Zion MM, Balkin MM, Rosenmann D et al. Use of the pulmonary artery catheter in patients with acute myocardial infarction. Chest 1990; 98: 1331–5[Abstract]

10 Guyatt G, Ontario Intensive Care Study Group. A randomized controlled trial of right heart catheterization in critically ill patients. J Intensive Care Med 1991; 6: 91–5[Medline]

11 Connors AF, Speroff T, Dawson NV et al. The effectiveness of right heart catheterisation in the initial care of critically ill patients. J Am Med Assoc 1996; 276: 889–97[Abstract]

12 SUPPORT Principal Investigators. A controlled trial to improve care for seriously ill hospitalised patients: the Study to Understand Prognoses and Preferences for Outcome and Risks of Treatments (SUPPORT). J Am Med Assoc 1995; 274: 1591–8[Abstract]

13 Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika 1983; 70: 41–55[ISI]

14 Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a severity of disease classification system. Crit Care Med 1985; 13: 818–29[ISI][Medline]

15 Cook EF, Goldman L. Performance of tests of significance based on stratification by a mulitvariate confounder score or by a propensity score. J Clin Epidemiol 1989; 42: 317–24[ISI][Medline]

16 Lusted LB. Decision making studies in patient management. New Engl J Med 1971; 284: 416–24[ISI][Medline]

17 Sznajder JI, Zveibil FR, Bitterman H, Weiner P, Bursztein S. Central vein catheterization. Failure and complication rates by three percutaneous approaches. Arch Intern Med 1986; 146: 259–61[Abstract]

18 Mermel LA, Maki DG. Infectious complications of Swan Ganz pulmonary artery catheters: pathogenesis, epidemiology, prevention and management. Am J Respir Crit Care Med 1994; 149: 1020–36[ISI][Medline]

19 Hayes MA, Timmins AC, Yau EHS, Palazzo M, Hinds CJ, Watson D. Elevation of systemic oxygen delivery in the treatment of critically ill patients. New Engl J Med 1994; 330: 1717–22[Abstract/Free Full Text]

20 Iberti TJ, Fischer EP, Leibowitz AB, Panacek EA, Silverstein JH, Albertson TE. A multicenter study of physicians knowledge of the pulmonary artery catheter. J Am Med Assoc 1990; 264: 2928–32[Abstract]

21 Gnaegi A, Feihl F, Perret C. Intensive care physicians insufficient knowledge of right heart catheterization at the bedside: time to act? Crit Care Med 1997; 25: 213–20[ISI][Medline]

22 Knaus WA, Wagner DP, Loirat P. A comparison of intensive care in the USA and France. Lancet 1982; ii: 642–6

23 Vincent JL, Bihari DJ, Suter PM et al. The prevalence of nosocomial infection in intensive care units in Europe. Results of the European Prevalence of Infection in Intensive Care (EPIC) study. J Am Med Assoc 1995; 274: 639–44[Abstract]