A Randomized Controlled Trial of the Use of Pulmonary-Artery  Catheters in High-Risk Surgical Patients

 

Authors: Sandham JD, and others for the Canadian Critical Care Trial Group

Settings: Multi-center trial in Canada

Source: New England Journal of Medicine 348:5-14, 2003

Summary

Background

Patients treated with PAC-guided therapy may have increased mortality. There is controversial evidence surrounding the benefits of PAC-guided therapy in surgical patients.

 Methods

·    Subjects: High-risk surgical patients undergoing either elective or emergency surgical procedures including abdominal, thoracic, major vascular, and orthopedic operations. Patients eligibility criteria included age ³ 60 with American Society of Anesthesiologists (ASA) class III or IV in patients scheduled to undergo elective or emergency major abdominal, vascular, thoracic or hip-fracture surgery.

·        Intervention: 1994 out of 3803 patients screened were randomized into two groups: a standard-care group  (997 patients) and a PAC group (997 patients). Patients in the standard-care group were treated without a PAC, but were allowed to have measurement of central venous pressure. Patients in the catheter group had a PAC inserted before surgery and had treatment guided by physiological end-points, namely, an oxygen delivery index of 550 to 600 ml per minute per square meter and a cardiac index between 3.5 and 4.5 liters per minute per square meter with a mean arterial pressure of 70 mm Hg, a PCWP of 18 mm Hg, a heart rate less than 120 beats per minute, and a hematocrit greater than 27%. The strategy to reach the specific end-points in the PAC group included, in order of priority, volume loading, inotropic therapy, vasodilator therapy, and blood transfusions, if Hct was less than 27%. Clinical data acquired in both groups included New York Heart Association (NYHA) functional class, Goldman Cardiac Risk Index, vital capacity, and forced expiratory volume in one second. Vital status at 6 and 12 months was established by telephone contact in both groups

·    Outcome analysis: The primary outcome was in-hospital mortality from any cause (crude mortality). Secondary outcomes were 6 and 12-months mortality, and in-hospital morbidity. In-hospital morbidity was defined by the occurrence of one of the following events:

¨      Myocardial infarction defined by the presence of a new Q-wave on ECG or ECG changes with increased CPK-MB fraction or troponin

¨      CHF assessed on the basis of CXR interpreted by an observer unaware of protocol assignment

¨      Arrhythmia based on ECG or rhythm strip

¨      Pneumonia according to CDC criteria

¨      Pulmonary embolism documented at autopsy or by pulmonary angiography, positive CTPA, high-probability V/Q scan or positive Duplex scan of legs

¨      Renal failure defined by a 50% increase in creatinine concentration or the need for dialysis

¨      Hepatic failure defined by a serum bilirubin concentration greater than 34 mmol per liter and an increase in PT > 3 seconds

¨      Sepsis from the CV or PAC defined by inflammation at the catheter site in association with systemic sepsis and positive blood cultures and positive catheter cultures

 

Statistical Analysis

The sample size was calculated to provide a > 90% power to detect a difference in mortality rates of 10% and 15% in the two groups, allowing a two-sided alpha of 0.05 (if I understand correctly this assumption, the authors were postulating a treatment effect of 33% because a reduction in mortality from 15% to 10% corresponds to a relative risk reduction of 33.3%). Additional calculations (certainly done after the observation of a mortality of 7.7% in the standard-care group) confirmed a 78% power to distinguish between mortality rates of 5% and 8%  (I have checked this value and in fact is correct; it assumes a treatment effect of 37.5%. However if one wished to maintain the 90% power with a two-tailed alpha of 0.05 used in the initial calculation of sample size to detect differences between these two values [8% and 5%] then one would need 1488 patients in each arm of the study). I wonder why the authors were happy to accept this extreme decrease in the power to detect a difference!

No crossover was allowed. Analysis, done with unpaired t-test or Wilcoxon rank-sum test for continuous variables and Fisher’s exact test or chi-square test for proportions, was on an intention-to-treat basis.

 

Results

 

¨      6-month survival: 88.1%  [95% CI: 86.0 to 90.1] in the standard-care group as opposed to 88.7%  [95% CI 85.3 to 89.5] in the PAC group (p >0.05)

¨      12-month survival: 83.9% [95% CI: 81.6 to 86.2] in the standard-care group as opposed to 83% [95% CI: 80.6 to 85.4] in the PAC group (p>0.05)

¨      Pulmonary embolism: zero in the standard-care group vs. 8 (0.8%) in the PAC group (p=0.004)

¨      No difference in the other secondary outcome including MI, CHF, arrhythmia, renal failure, hepatic failure, sepsis

Conclusions by the authors

In elderly high-risk surgical patients who undergo elective or emergency major surgery there is no advantage to therapy guided by a PAC as compared to standard care

 

. 

Comments:

This study will be used by many physicians as the study that proves that PAC-guided therapy does not benefit high-risk surgical patients. Is this study valid? Are the conclusions of the authors generalizable?  Should we abandon the use of PAC-guided therapy based on this study?  In fact, an in depth analysis of this study shows that the conclusions are not valid and that the authors, while claiming not to be biased by inserting an Avoidance of Bias paragraph, were biased by accepting the possibility of a type II error, in addition to using eligibility criteria that identified surgical patients who would not be considered at high-risk by most surgeons. Based on the control event rate of 7.7%, assuming the same treatment effect, they would have required 1488 patients to identify a difference between the groups; why did they not expand the sample size after the safety analysis done after the enrollment of the first 800 patients? Did the safety board that conducted the interim analysis inform the authors of the control event rate?  If so, why did they not continue to enroll patients until the appropriate sample size of 1488 patients in each arm of the study was reached?  Are the patients entered in this study, truly high-risk surgical patients? Previous studies that have attempted to answer the same question have used an expected mortality in excess of 30% by POSSUM to identify high-risk surgical patients [1,2]. Can surgical patients who have a mortality of 7.7% with standard care be considered high-risk?  Has the ASA risk classification developed in the 60s been validated as an in-hospital mortality prediction tool?  The ASA risk classification has never been validated as a prediction tool for 28-day hospital mortality. Additionally, while the eligibility criteria required patients to have an age ³ 60 with an ASA class III or IV, only 123/997 (13.4%) and 113/997 (12.54%) of patients in the standard-care and PAC group, respectively, met these criteria. Of note, only 9 patients in the standard-care group and 13 in the PAC group were class IV ASA.  Since the authors admit to having treated 327 (32.8%) patients in the standard-care group with inotropes, are they suggesting that inotropes should be used without information on preload or afterload? Did they use inotropes in patients who had only a peripheral IV line? Is the standard of care in Canada to use inotropes without load analysis?  Since 228 patients, deemed to be high-risk surgical patients, had only a peripheral IV, why were they admitted to the ICU? Did these patients have an arterial line? How can the author suggest a causal relationship between PEs and the use of PACs when they do not provide us with the information concerning the work up of these patients with respect to DVT?  DVT prophylaxis was started after surgery in 52.1% of the patients in the standard-care group and in 53.7% of patients in the PAC group; this represents sub-optimal DVT prophylaxis.  The conclusions of this study are not generalizable, and it is my opinion that the conclusions of the study are not valid.

1. Boyd :  A deliberate increase in oxygen delivery to decrease mortality and morbidity in high-risk surgical patients. JAMA 270:2699-2707, 1993

2. Lobo: Effects of maximizing oxygen delivery on morbidity and mortality in high-risk surgical patient. Crit Care Med. 28: 3396-404, 2000

 

Reviewed on 3/25/2003 by:

Corrado P. Marini, M.D.

Long Island Jewish Medical Center

 

marini@lij.edu