Early Use of the Pulmonary Artery Catheter and Outcomes in Patients with Shock and Acute Respiratory Distress Syndrome a Randomized Controlled Trial

CARING FOR THE CRITICALLY ILL PATIENT

Early Use of the Pulmonary Artery Catheter and Outcomes in Patients With Shock and Acute Respiratory Distress Syndrome A Randomized Controlled Trial

Christian Richard, MD Context Many physicians believe that the pulmonary artery catheter (PAC) is useful Josiane Warszawski, MD, PhD for the diagnosis and treatment of cardiopulmonary disturbances; however, observa- Nadia Anguel, MD tional studies suggest that its use may be harmful. Objective To determine the effects on outcome of the early use of a PAC in patients Nicolas Deye, MD with shock mainly of septic origin, acute respiratory distress syndrome (ARDS), or both. Alain Combes, MD Design, Setting, and Patients A multicenter randomized controlled study of 676 pa- Didier Barnoud, MD tients aged 18 years or older who fulfilled the standard criteria for shock, ARDS, or both conducted in 36 intensive care units in France from January 30, 1999, to June 29, 2001. Thierry Boulain, MD Intervention Patients were randomly assigned to either receive a PAC (n=335) or Yannick Lefort, MD not (n=341). The treatment was left to the discretion of each individual physician. Muriel Fartoukh, MD Main Outcome Measures The primary end point was mortality at 28 days. The Frederic Baud, MD, PhD principal secondary end points were day 14 and 90 mortality; day 14 organ system, renal support, and vasoactive agents–free days; hospital, intensive care unit, and me- Alexandre Boyer, MD chanical ventilation–free days at day 28. Laurent Brochard, MD, PhD Results The 2 groups were similar at baseline. There were no significant differences in Jean-Louis Teboul, MD, PhD mortality with or without the PAC at day 14: 49.9% vs 51.3% (mortality relative risk [RR], 0.97; 95% confidence interval [CI], 0.84-1.13; P=.70); day 28: 59.4% vs 61.0% for the French Pulmonary Artery (RR, 0.97; 95% CI, 0.86-1.10; P=.67); or day 90: 70.7% vs 72.0% (RR, 0.98; 95% CI, Catheter Study Group 0.89-1.08; P=.71). At day 14, the mean (SD) number of days free of organ system fail- ANY PHYSICIANS BELIEVE ures with or without the PAC (2.3 [3.6] vs 2.4 [3.5]), renal support (7.4 [6.0] vs 7.5 [5.9]), that the pulmonary ar- and vasoactive agents (3.8 [4.8] vs 3.9 [4.9]) did not differ. At day 28, mean (SD) days in hospital with or without the PAC (0.9 [3.6] vs 0.9 [3.3]), in the intensive care unit (3.4 tery catheter (PAC) is [6.8] vs 3.3 [6.9]), or mechanical ventilation use (5.2 [8.5] vs 5.0 [8.5]) did not differ. useful for the diagnosis andM treatment of cardiopulmonary dis- Conclusion Clinical management involving the early use of a PAC in patients with shock, ARDS, or both did not significantly affect mortality and morbidity. turbances and for assessing volume status in critically ill patients.1,2 How- JAMA. 2003;290:2713-2720 www.jama.com ever, investigators have raised doubts Author Affiliations: Service de Re´ animation Me´ di- Re´ animation Me´ dicale, Hoˆ pital Lariboisie` re, Assis- about the safety of the PAC3,4 because cale (Drs Richard, Anguel, and Teboul) and Service de tance Publique, Hoˆ pitaux de Paris, Universite´ Paris Sante´ Publique et Epide´ miologie (Dr Warszawski), Hoˆ- VII, Paris (Dr Baud); Service de Re´ animation Me´ di- its use may be associated with direct pital de Biceˆ tre, Assistance Publique, Hoˆ pitaux de Paris, cale, Hoˆ pital Louis Mourier, Assistance Publique, complications5 or harmful effects re- Universite´ Paris XI, Le Kremlin Biceˆ tre; Service de Re´- Hoˆ pitaux de Paris, Universite´ Paris VII, Colombes animation Me´ dicale, Hoˆ pital Europe´ en Georges Pom- (Dr Boyer), France. lated to inappropriate decisions result- pidou (Dr Deye), Service de Re´ animation Me´ dicale, Members of the French Pulmonary Artery Catheter ing from misinterpretation of the data.6 Groupe Hospitalier Pitie´ - Salpe´ trie` re (Dr Combes), and Study Group are listed at the end of the article. Service de Pneumologie et Re´ animation, Groupe Corresponding Author and Reprints: Christian Richard, The most serious concern was raised by Hospitalier Pitie´ -Salpe´ trie` re (Dr Lefort), Assistance MD, Service de Re´ animation Me´ dicale, Hoˆ pital de Bi- the retrospective study by Connors et Publique, Hoˆ pitaux de Paris, Universite´ Paris VI, Paris; ceˆ tre, 78 rue du Ge´ne´ ral Leclerc, 94275 Le Kremlin 4 Service de Re´ animation Me´ dicale, Centre Hospitalo- Biceˆ tre-Cedex, France (e-mail: christian.richard@bct al, which suggested that the use of PAC Universitaire A. Michallon, Grenoble (Dr Barnoud); Ser- .ap-hop-paris.fr). within 24 hours following admission vice de Re´ animation, Centre Hospitalier Re´ gional Caring for the Critically Ill Patient Section Editor: Debo- Orle´ ans-La Source, Orle´ ans (Dr Boulain); Service de rah J. Cook, MD, Consulting Editor, JAMA. Re´ animation Me´ dicale, Hoˆ pital Henri-Mondor, Assis- Advisory Board: David Bihari, MD; Christian Brun- For editorial comment see p 2732. tance Publique, Hoˆ pitaux de Paris, Universite´ Paris XII, Buisson, MD; Timothy Evans, MD; John Heffner, MD; Cre´ teil (Dr Fartoukh and Brochard); Service de Norman Paradis, MD; Adrienne Randolph, MD.

Downloaded From: https://jamanetwork.com/ on 09/25/2021 EARLY USE OF THE PULMONARY ARTERY CATHETER

may increase mortality, length of stay, METHODS of the medical staff was affiliated with the and health care costs. Either a mora- Enrollment Socie´te´deRe´animation de Langue Fran- torium on its use or the performance Patients were eligible for inclusion if they c¸aise. The protocol was approved by the of randomized clinical trials have been met criteria for shock, ARDS, or both. institutional review board of Hoˆpital de proposed,3,7 particularly in patients with Shock was defined by the presence of less Biceˆtre. Written informed consent was septic shock or acute respiratory dis- than 12 hours of 4 criteria: heart rate of obtained from the patients or surro- tress syndrome (ARDS).8 One of the dif- at least 90/min; a respiratory rate of at gates. As patients could understand, they ficulties in performing such a trial is to least 20/min or a PaCO2 of 32 mm Hg or were informed of their right to with- decide whether a PAC-guided thera- lower or the use of mechanical ventila- draw from the study. The trial was moni- peutic protocol should be used. Most tion; the use of vasopressors to main- tored by an independent data and safety physicians use it as a diagnostic tool but tain a systolic blood pressure of at least monitoring board. Randomization was not all physicians have the same thera- 90 mm Hg despite fluid resuscitation, low conducted centrally by telephone on a peutic approach. However, the impact dose of dopamine (Յ5 µg/kg per 24-hour-a-day, 7-day-a-week basis and of a goal-oriented therapy using PAC minute), or dobutamine; and at least 1 based on a permuted-block algorithm, al- has been evaluated in high-risk surgi- of 3 signs of hypoperfusion (urine out- lowing stratification for each center. cal patients and yielded inconsistent put Ͻ0.5 mL/kg of body weight per hour No standardized protocols for man- results, ranging from decreased mor- for 1 hour or more; neurologic dysfunc- aging patients were proposed; there- tality9 to increased morbidity and tion defined by confusion, psychosis, or fore, each patient was treated by his/ mortality.10 a Glasgow coma scale score of Յ6; her physician. The centers agreed on the To avoid the methodological prob- plasma lactate higher than the upper limit following principles: optimization of cir- lems of these studies, the Canadian of the normal value). Patients with shock culating blood volume; vasoactive sup- Critical Care Clinical Trials Group11 were excluded if more than 12 hours had port if necessary at a mean arterial pres- had recently conducted a randomized elapsed since the presence of the 4 cri- sure of at least 60 mm Hg when fluid trial comparing goal-oriented therapy teria had occurred or if the last 2 crite- balance was optimal; no objective for guided by a PAC with standard care ria were present for more than 12 hours. maximization of oxygen transport; free without the use of a PAC in elderly Acute respiratory distress syn- access to echocardiography; assisted high-risk surgical patients requiring drome was defined by the presence of control ventilation with a maximum pla- intensive care. This study did not find more than 24 hours of 4 criteria: acute teau pressure of 35 cm H2O and SpO2 of any effect of goal-oriented therapy by decrease in PaO2/FIO2 to 200 mm Hg or more than 90%; and prevention of PAC over standard care.11 Moreover, lower, whatever the level of positive thromboembolism with low-molecular- the authors reemphasized the findings end-expiratory pressure; bilateral pul- weight heparin, if not contraindicated. of others9,10,12 when they suggested monary infiltrates or a chest radio- The PAC had to be inserted during the that it is difficult to achieve the physi- graph consistent with edema; no clini- 2-hour period following randomiza- ological objectives of the goal- cal evidence of left atrial hypertension; tion. The type of PAC was decided at the oriented therapy. Applied to critically and requirement for positive pressure discretion of the physician. The type, ill patients, this practice of a goal- ventilation. Patients with ARDS were site, and duration of PAC insertion as oriented therapy has not shown any excluded if less than 24 hours had well as the onset of complications were consistent benefit.12 In addition, no elapsed since the presence of the 4 cri- recorded. The decision to remove the consensus yet exists for such a goal- teria had occurred. PAC or to replace it was at the discre- oriented therapy in septic shock or in Patients were excluded if they were tion of the ICU team. ARDS. For these reasons and because younger than 18 years, experienced the harmful effects of the PAC were hemorrhagic shock, myocardial infarc- Outcome Measures found in observational studies with- tion complicated by cardiogenic shock and Data Collection out formal protocols, we chose not to requiring revascularization, or throm- The primary end point was mortality at use any goal-oriented protocols in a bocytopenia (Յ10.0 ϫ 109/L), partici- 28 days. The secondary end points were study designed to assess catheter pated in other trials within the last 30 mortality at 14 and 90 days, duration of safety. days, were moribund, or if their phy- ICU and hospital stay, ICU and hospi- We report the results of a random- sician refused to agree with the use of tal free days, defined as the number of ized clinical trial performed in 36 in- full life support. days from day 1 to day 28 without ICU tensive care unit (ICU) centers in stay or without hospital stay; ventilator- France, in which we assessed the effect Study Design and free days as the number of days from day of the early insertion of a PAC with- Treatment Protocol 1 to day 28 during which the patient out goal-oriented therapy on mortal- Patients were enrolled from January 30, breathed spontaneously; and renal sup- ity and morbidity in patients with shock 1999, to June 29, 2001, at 36 French cen- port, organ system, and vasoactive mainly of septic origin, ARDS, or both. ters. In each center, at least 1 member agents–free days as the number of days

Downloaded From: https://jamanetwork.com/ on 09/25/2021 EARLY USE OF THE PULMONARY ARTERY CATHETER

from day 1 to day 14 without renal sup- thereafter, we did not change our Figure 1. Study Flow of Screening and port, organ system failure, or use of vaso- sample size calculation. Progress of Patients active agents. The study was scheduled for an 18-

At ICU admission, we recorded age; month period. This limit was chosen to 681 Randomized sex; severity of underlying medical con- avoid changes in practice regarding the dition (McCabe and Jackson)13; the Sim- PAC since the trial could not be blinded. 338 Assigned to Receive 343 Assigned to Control plified Acute Physiology Score II (SAPS A slower rate of enrollment than ex- a PAC Group 14 II) ; the Sepsis-related Organ Failure As- pected was observed because of the se- 327 Received a PAC 326 Remained sessment score15; and the organ dysfunc- verity of the inclusion criteria (pres- as Assigned Assigned to 8 Did Not Receive Control Group 16 tion and infection score. Patients were ence of shock for Ͻ12 hours and a PAC as 15 Received a PAC classified as medical, scheduled, or un- presence of ARDS for Ͼ24 hours). Be- Assigned 2 Withdrew 6 Died Before Consent scheduled surgical patients. Signs of or- cause of this slow rate of inclusions, the PAC Insertion 2 Placement gan failure according to the Brussels score steering committee recommended lim- Not Possible 17 were simultaneously recorded. At base- iting the power of the study to 80% in- 3 Withdrew line, we recorded mechanical ventila- stead of 90%. With regard to the ob- Consent tion, SAPS II, Sepsis-related Organ Fail- served rate of inclusion, this decision ure Assessment score, organ dysfunction required inclusion of 754 patients dur- 0 Lost to Follow-up 0 Lost to Follow-up Before Day 28 Before Day 28 and infection score, presence and cause ing a 30-month period. This decision 1 Lost to Follow-up 2 Lost to Follow-up O IO Between Day 28 and Between Day 28 and of shock and ARDS, Pa 2/F 2, and or- was approved by the data and safety Day 90 Day 90 gan failure according to the Brussels monitoring board. The interim analy- score.17 Patients were monitored daily for sis was conducted according to the pro- 335 Included in Analysis 341 Included in Analysis 14 days for organ failure. Complica- tocol. Finally, the rate of inclusions 3 Excluded From 2 Excluded From Analysis (Withdrew Analysis (Withdrew tions related to the PAC insertion, main- further declined, so by the end of the 30- Consent) Consent) tenance, and removal were recorded. month period, 681 patients were in- Arrhythmia was determined by electro- cluded and the data and safety monitor- PAC indicates pulmonary artery catheter. cardiography. Sepsis from the PAC was ing board decided to cease the study. defined by inflammation at the inser- Under these conditions, the power of the the insertion of a central venous cath- tion site and systemic sepsis by sepsis study to detect a 10% absolute differ- eter at baseline. All the statistical tests plus a positive culture of blood or cath- ence, using the mortality observed in our were performed with the Stata statisti- eter tip that resolved with removal of the control group and taking into account cal software version 6.0 (Stata Corp, catheter. the interim safety analysis, was 78%. College Station, Tex) and PՅ.05 was Since use of the PAC could not be The baseline characteristics, compli- considered statistically significant. blinded, an extensive on-site monitor- cations, and outcomes in the 2 groups ing procedure was used to avoid bias were compared with the use of the ␹2, RESULTS evaluating morbidity. Calculation of Fisher exact, and t tests. All tests were Study Population severity scores was not performed by the 2-sided. The primary analysis consisted From January 30, 1999, to June 29, 2001, investigators but centrally calculated by of a comparison of mortality at day 28 681 patients were enrolled at 36 ICU cen- the statisticians from raw data recorded. by a ␹2 test on an intention-to-treat basis. ters (FIGURE 1). Five patients withdrew The crude mortality relative risk (RR) at their consent, leaving a total of 676 pa- Statistical Analysis day 28 was estimated. Overall survival tients randomized to the PAC group We calculated the sample size to de- curves were obtained using the Kaplan- (n=335) and the control group (n=341). tect a 10% difference in the mortality Meier method and compared using the Among the 36 centers, 15 included fewer at day 28 between the 2 groups with a log-rank test. A Mantel-Haenszel ␹2 test than 10 patients (1-9) and 12 included 2-tailed test, a significance level of 5%, of homogeneity was performed to com- more than 20 patients (21-112). The per- and a power of 90%. For a global mor- pare the mortality RR among all centers centages of patients in the PAC vs con- tality with balanced groups of 40% that included 10 or more patients. trol groups randomized for shock (63.1% (35% vs 45%), we planned to enroll A multivariate analysis was per- and 64.2%, respectively), ARDS (29.5% 1100 patients, taking into account 1 informed, using the Cox proportional and 31.2%), or both conditions (7.4% terim analysis after the inclusion of 500 hazards regression model stratified at and 4.6%) did not differ significantly patients.18 Prior to our study, it was dif- hospital centers, to estimate hazard ra- (P=.32). Admission and baseline char- ficult to have an accurate estimate of the tio (HR) adjusted for age, sex, SAPS II, acteristics did not differ between the 2 expected mortality based on our spe- shock or ARDS at inclusion, the medi- groups with the exception of the classi- cific inclusion criteria. As per our pro- cal origin of the patients (medical, fication of medical, scheduled surgical, tocol, the global mortality was esti- scheduled surgical, or unscheduled sur- or unscheduled surgical patients (P=.02) mated after the first 20 validated deaths; gical), the use of echocardiography, and (TABLE 1).

Downloaded From: https://jamanetwork.com/ on 09/25/2021 EARLY USE OF THE PULMONARY ARTERY CATHETER

Violation of the Protocol Table 1. Patients’ Admission and Baseline Characteristics* and Loss to Follow-Up PAC Group Control Group Characteristics (n = 335) (n = 341) A PAC was inserted in 15 (4.4%) of 341 Admission† patients in the control group. In con- Age, mean (SD), y 62.2 (15.4) 63.2 (14.8) trast, 8 (2.4%) of 335 patients in the Men 224 (66.9) 228 (66.9) PAC group did not receive it; 6 died be- McCabe and Jackson classification fore insertion and in 2 placement was 0 231 (69.6) 220 (65.6) not possible. Three patients were lost 1 82 (24.7) 100 (29.8) to follow-up between day 28 and day 2 19 (5.7) 16 (4.8) 90 (1 patient in the PAC group and 2 Score, mean (SD) in the control group) (Figure 1). SAPS II 54.1 (21.5) 54.8 (22.7) SOFA 10.0 (4.5) 9.9 (4.4) Primary End Point: Odin 3.5 (1.2) 3.3 (1.2) Day 28 Mortality Brussels 2.7 (1.3) 2.7 (1.3) No. of patients Overall mortality at day 28 was 60.2%. Medical 292 (87.4) 279 (81.8) On an intention-to-treat basis, this per- Scheduled surgical 9 (2.7) 25 (7.3) centage did not differ significantly be- Unscheduled surgical 33 (9.9) 37 (10.9) tween the PAC and the control group Admission before inclusion‡ 137 (40.9) 138 (40.5) at day 28: 59.4% vs 61.0% (199 vs 208 Baseline§ deaths, with an absolute difference of Mechanical ventilation 322 (96.7) 330 (97.1) –1.6%; 95% confidence interval [CI], Score, mean (SD) –9.0 to 5.8) and a mortality RR associ- SAPS II 53.3 (15.0) 52.7 (13.9) ated with the PAC compared with the SOFA 9.9 (3.1) 9.9 (2.9) control group of 0.97 (95% CI, 0.86- Odin 3.4 (1.0) 3.3 (0.9) 1.10; P =.67) (TABLE 2). FIGURE 2 Brussels 2.6 (1.0) 2.6 (1.0) shows the Kaplan-Meier estimates of Organ failure survival. The crude HR, estimated us- Cardiovascular 276 (82.6) 286 (84.1) ing the Cox proportional hazards re- Pulmonary 279 (83.3) 285 (83.6) gression model, was 1.0 (95% CI, Renal 147 (44.1) 145 (42.9) 0.82-1.22). Similar results to those Hepatic 83 (24.9) 80 (23.5) Hematologic 78 (23.7) 70 (21.3) shown in Table 2 were obtained when Central nervous system 26 (7.9) 23 (6.9) the 23 patients who did not receive the Shock at inclusion࿣ 266 (79.4) 265 (77.7) allocated treatment were excluded. Septic shock 226 (87.3) 226 (85.9) Results Among Centers Hypovolemic shock 12 (4.6) 13 (4.9) Cardiogenic shock 13 (5.0) 19 (7.2) The mortality RRs for all centers that Acute poisoning 8 (3.1) 5 (1.9) included 10 or more patients did not Shock exclusively 148 (44.0) 167 (49.0) differ significantly (P=.37) and a term ϫ ARDS exclusively 67 (20.0) 76 (22.0) for treatment center interaction was Shock and ARDS 120 (36.0) 98 (29.0) not included in the analysis. In 1 cen- Shock for Ͻ12 hours at inclusion¶ 229 (70.5) 225 (68.8) ter (n=112), a significant difference in ARDS mortality in favor of the PAC group was At inclusion 187 (55.8) 174 (51.0) observed. After adjustment for the se- Primary 139 (78.5) 138 (82.1) verity on admission (SAPS II), this dif- For Ͼ24 hours at inclusion¶ 120 (36.9) 117 (35.8) ference was no longer significant with Renal support 29 (8.8) 28 (8.5) an adjusted HR of 0.69 (95% CI, 0.42- Vasoactive agents 296 (88.6) 307 (90.3) 1.13; P=.14). Abbreviations: ARDS, acute respiratory distress syndrome; Odin, organ dysfunction and/or infection; PAC, pulmonary artery catheter; SAPS II, Simplified Acute Physiology Score II; SOFA, Sepsis-related Organ Failure Assessment. *Data are No. (%) unless otherwise specified. McCabe and Jackson classification: 0, nonfatal underlying disease; 1, Secondary End Points ultimately fatal underlying disease; 2, rapidly fatal underlying disease.13 P values were obtained by Pearson ␹2 test for comparison of percentages or t test for comparison of means. TABLE 3 shows the results of the be- †Scores at admission were calculated from the worst values within the 24 hours following the admission. tween-group comparison concerning ‡Percentage of patients with a delay between the admission in the intensive care unit and the inclusion in the study. §Baseline characteristics correspond to their status at inclusion in the study or to the worst value in the 24 hours pre- the secondary end points. Day 14 mor- ceding it. ࿣Type of shock was omitted for 7 patients in PAC and 2 patients in control. tality was 49.9% in the PAC group vs ¶A total of 24 patients did not strictly meet criteria for inclusion in terms of schedule for ARDS, shock, or both. For more 51.3% in the control group (RR, 0.97; details, see “Methods” section. 95% CI, 0.84-1.13; P=.70) and day 90

Downloaded From: https://jamanetwork.com/ on 09/25/2021 EARLY USE OF THE PULMONARY ARTERY CATHETER

mortality was 70.7% vs 72.0% (RR, bidity among patients with shock, ARDS, misinterpretation of data that could have 0.98; 95% CI, 0.89-1.08; P=.71), re- or both. Severe complications were in- resulted in erroneous decisions, and also spectively. No statistically significant frequent. a possible aggressive therapeutic ap- differences were observed for any of the These findings do not confirm those proach. Previous studies showed that de- secondary end points according to the reported by Connors et al.4 Inclusion of liberate increase in oxygen delivery re- intention-to-treat analysis. a more heterogeneous and less se- sulted in either no benefit12,19 or harmful verely ill population might account for effect.10 However, it is also possible that Multivariate Analysis this discrepancy. To explain these find- the study performed by Connors et al4 The mortality RRs at day 28 were ap- ings, Connors et al4 reported complica- may have overestimated the mortality in proximately equal to 1, whatever the tions related to maintenance of the PAC, the group with a PAC because of the reason for inclusion and the severity as- the role of errors in measurements or limitation of retrospective match- sessed by quartile of SAPS II (Table 2). With adjustment for the other vari- Table 2. Estimates of the Relative Risk of Mortality in Patients With a PAC at Day 28* ables listed in TABLE 4, the adjusted HR Mortality, No. (%) associated with the PAC group was 0.97 (95% CI, 0.79-1.20; P=.78). An in- No. of PAC Control Relative Risk (95% P Group Patients (n = 335) (n = 341) Confidence Interval) Value creased mortality risk was associated Ͻ Total 676 199 (59.4) 208 (61.0) 0.97 (0.86-1.10) .67 with increased SAPS II (P .001) and At inclusion† with shock and ARDS compared with Shock exclusively 313 86 (58.9) 105 (62.9) 0.94 (0.78-1.12) .47 ARDS alone (P = .02). ARDS exclusively 143 31 (46.3) 36 (47.4) 0.98 (0.69-1.39) .90 Shock and ARDS 218 81 (67.5) 67 (68.4) 0.99 (0.82-1.19) .89 Complications Related to PAC All septic shock 452 146 (64.6) 150 (66.4) 0.97 (0.85-1.11) .69 The catheter was left indwelling for a Day 1 echocardiography mean 2.3 days (range, 1-10) in the en- Yes 346 92 (62.6) 124 (62.3) 1.00 (0.85-1.18) .96 tire population and a mean 2.6 days in No 327 106 (57.0) 84 (59.6) 0.96 (0.80-1.15) .64 Day 1 central venous catheter the survivors (range, 1-10). Ninety per- Yes 503 148 (62.2) 166 (62.6) 0.99 (0.87-1.14) .92 cent of survivors had an indwelling PAC No 169 49 (51.6) 41 (55.4) 0.93 (0.70-1.23) .62 for less than 5 days. Complications dur- SAPS II (quartile) ing pulmonary catheter insertion in- Ͻ44 174 33 (35.9) 35 (42.7) 0.84 (0.58-1.21) .36 cluded arterial puncture (n=17), hemo- 44-53 186 45 (55.6) 63 (60.0) 0.93 (0.72-1.19) .54 thorax (n=1), arrhythmias and 54-61 145 51 (66.2) 48 (70.6) 0.94 (0.75-1.17) .57 conduction disturbances (n=60), and Ͼ61 165 69 (82.1) 62 (76.5) 1.07 (0.92-1.25) .37 knots (n=6). No death attributable to Abbreviations: ARDS, acute respiratory distress syndrome; PAC, pulmonary artery catheter; SAPS II, Simplified Acute Physiology Score II. ventricular fibrillation or to arrhyth- *P values were obtained by Pearson ␹2 test. SAPS II scores were recorded at inclusion. Echocardiography and central venous catheter were recorded during the first hours following inclusion (day 1). mia was reported. No pulmonary em- †Whatever the moment of onset. bolism and deep venous thrombosis were recorded. In 2 of 10 patients with Figure 2. Kaplan-Meier Survival Estimates During 3 Months positive PAC culture after insertion, the blood cultures were primary positive 1.0 (once with Enterobacter cloacae and once with Staphylococcus epidermidis) and ad- 0.8 Log-Rank P = .65 equate antibiotic therapy was adminis- tered. For the other 8 patients, local clini- 0.6

cal signs and sepsis resolved with Survival 0.4 PAC Group removal of the PAC without antibiot- Control Group ics. For these patients, the microorgan- 0.2 isms were: 5 S epidermidis,1Pseudomo- nas aeruginosa, 1 Klebsiella pneumoniae, 0 7 14 21 28 35 42 49 56 63 70 7784 91 and 1 Citrobacter freundii. Days COMMENT No. of Patients Alive PAC Group 335 198 168 136 99 The major findings of our trial are that Control Group 341 201 166 133 97 clinical management involving early use of a PAC was not associated with sig- PAC indicates pulmonary artery catheter. The vital status at day 90 was known for all but 3 patients (1 in the PAC group and 2 in the control group). nificant changes in mortality and mor-

Downloaded From: https://jamanetwork.com/ on 09/25/2021 EARLY USE OF THE PULMONARY ARTERY CATHETER

ing of patients. Connors et al4 chose to a higher illness severity in the patients sociated with the use of the PAC disap- use a propensity score but this score did with a PAC. This crucial issue was il- peared when the use of vasoactive agents not take into account the intensity of lustrated in a retrospective cohort study was taken into account in the multivar- treatment used to sustain hemodynam- of patients with ARDS, which showed iate analysis.20 ics. This approach could have masked that the apparent increased mortality as- We chose to enroll a selected population of patients who were not sub- Table 3. Secondary End Points According to the Intention-to-Treat Analysis* mitted to a goal-oriented protocol to define the role of the PAC. Our decision PAC Group Control Group End Points (n = 335) (n = 341) P Value to perform a study without any goal- Mortality, No. (%) oriented therapy was based on the Day 14 167 (49.9) 175 (51.3) .70 harmful results found in observa- Day 90 236 (70.7) 244 (72.0) .71 tional studies without use of formal pro- Duration of stay, d tocols and the lack of consensus re- Intensive care unit 11.6 (10.1) 11.9 (10.0) .72 garding hemodynamic support in septic Hospital 14.0 (11.6) 14.4 (11.3) .67 4,21 Day 28 No. of free days shock or in ARDS. Hospital 0.9 (3.6) 0.9 (3.3) .95 No significant differences in mortal- Intensive care unit 3.4 (6.8) 3.3 (6.9) .88 ity between patients with and without Ventilation 5.2 (8.5) 5.0 (8.5) .72 PAC either at day 14, 28, or 90 were ob- Day 14 No. of free days served. There was also no difference Organ system 2.3 (3.6) 2.4 (3.5) .72 between the 2 groups in day 14 organ Renal support 7.4 (6.0) 7.5 (5.9) .93 system failure, renal support, vasoactive- Vasoactive agents 3.8 (4.8) 3.9 (4.9) .78 free days, or in day 28 ICU hospital or Abbreviation: PAC, pulmonary artery catheter. *Data are presented as mean (SD) unless otherwise specified. Vital status was obtained for all patients until day 28 and ventilator-free days. The overall day 28 for all except 3 patients at day 90. P value was obtained by Pearson ␹2 test for comparison of percentages or t test mortality rate was very high (60.2%), as for comparison of means. For details of the calculation of free days, see “Methods” section. in the PAC group (59.4%) and the control group (61.0%). The higher mortal- Table 4. Cox Proportional Hazards Regression Models of Predictors of Day 28 Mortality* ity than previously planned in our pro- Adjusted Hazard Ratio tocol (60% vs 40%) may be partly (95% Confidence Interval) P Value explained by the following. When we PAC vs control PAC 0.97 (0.79-1.20) planned our study, it was difficult to have .78 Control 1.00 a realistic idea of the mortality rate in pa- Sex tients with such strict inclusion criteria Male 1.23 (0.98-1.55) .07 identifying seriously ill patients. In a re- Female 1.00 cent randomized clinical trial involv- Age, per 1 y 1.00 (1.00-1.01) .15 ing new drugs,22 very careful selection At inclusion† of patients was performed to avoid a po- Shock and ARDS 1.47 (1.04-2.08) tentially inflated rate of adverse effects. Shock exclusively 1.10 (0.78-1.54) .02 In our protocol, we choose to include ARDS exclusively 1.00 patients with an accurate diagnosis of Medical admission vs surgical Medical 1.35 (0.98-1.87) shock, ARDS, or both. Thus, we did not .07 Surgical 1.00 include patients with shock for more SAPS II than 12 hours or patients with ARDS for Ͼ61 2.84 (1.99-4.06) less than 24 hours. This decision pre- 54-61 2.32 (1.63-3.30) Ͻ.001 cluded studying partly resolved shock 44-53 1.49 (1.07-2.09) or respiratory failure. The high inclu- Ͻ44 1.00 sion rate in our study, 87.4% in the PAC Day 1 central venous catheter group and 81.8% in the control group Yes 0.83 (0.63-1.09) .18 (Table 1), could also explain the high No 1.00 mortality because comorbidities are of- Day 1 echocardiography Yes 1.06 (0.85-1.33) ten more frequently observed in medi- .61 No 1.00 cal rather than in surgical patients. The Abbreviations: ARDS, acute respiratory distress syndrome; PAC, pulmonary artery catheter; SAPS II, Simplified Acute calculation of the standardized mortal- Physiology Score II. *Central venous catheter and echocardiography were recorded during the first hours following inclusion (day 1). The ity ratio confirmed the lack of influ- Cox proportional hazards regression model was stratified by hospital centers. ence of PAC on outcome: 1.16, 1.09, and †Whatever the moment of onset. 1.23 for the global population, PAC, and

Downloaded From: https://jamanetwork.com/ on 09/25/2021 EARLY USE OF THE PULMONARY ARTERY CATHETER

control groups, respectively.14 At least to the 5% difference in mortality rate den. The use of heparin-coated cath- 1 other randomized trial is now in reported in the study by Connors et al.4 eter was routinely performed. Prevention progress to define the role of the PAC.23 We did not observe a center ϫ treatment of thromboembolism by low-molecular- Randomization, stratified by center, interaction because most RRs were near weight heparin was used. The removal was conducted centrally by telephone. or equal to 1. The RRs remained near or of the PAC was suggested as soon at it The details of each telephone call were equal to 1 among the different sub- was no longer required. This last rec- recorded in a database by patient eligi- groups of baseline predictors of mortal- ommendation partly explains the rela- bility and randomization group. This ap- ity, such as SAPS II and the type of diag- tively short duration of pulmonary ar- proach was useful to avoid patient pro- nosis at inclusion (Table 2). Taking into tery catheterization in our study (mean, tocol violations. Our study shows that account the use of echocardiography at 2.3 days) compared with observational a majority of physicians were willing to day 1, the PAC was not associated with studies.4 Despite the low incidence of have their patients participate in this type mortality (Table 2). No differences were complications observed, it is possible of controlled trial24 because a PAC was found in morbidity criteria between PAC that some of the complications we ob- inserted in only 15 of 341 patients in the and control groups (Table 3). served, such as ventricular arrhythmia, control group. These data are different We performed a careful review of the catheter-related sepsis, and central ve- from a trial performed a few years ago complications linked to insertion and nous access complications, may have off- in Canada that was stopped after only maintenance of the PAC. The inci- set the potential benefits of the PAC. 33 patients were included.25 Physicians dence of complications was low. Con- Even if the purpose of monitoring may have had less uncertainty about the cerning the major complications, our with PAC is ultimately to save lives, it use of PAC at the time that this initial results are in accordance with previ- would be unrealistic to believe that the Canadian trial was conducted, and thus ous studies reporting rates of 0.1% to prognosis of patients could be im- may have been less likely to allow their 0.5%.5,29,30 We observed PAC-related in- proved by its presence alone. An influ- individual patients to undergo random- fections in 10 patients (2.8%). This in- ence on prognosis without goal- ization.25 In contrast, our results were cidence is also in accordance with the oriented therapy could only be suggested comparable with a recently published most precise reported estimate of PAC when the presence of a PAC results in study.26 The exclusion of patients who and/or introducer sheath–related in- significant changes in treatment with violated the study protocol did not affect fections (5.9% to 29.1%) and with that fluid loading and vasoactive agents.32 our results. of PAC bacteriemia (0% to 4.6%).31 In our study, in both patient groups, Our data, which are in agreement with However, our study may underesti- the physicians were able to obtain rel- those of a recently published single- mate the true incidence of this compli- evant information by using echocardi- center study,26 strongly suggest that PAC cation because efforts to diagnose in- ography, thereby influencing diagnosis is not associated with increased mortal- fection were in response to suggestive and treatment.33 The ICU centers cho- ity or morbidity. The power of our study clinical findings instead of through sen for this multicenter randomized to detect a 10% absolute difference, using screening for possible infection. study have free access to echocardiog- the mortality observed in our control In contrast with the recently re- raphy. This technique was used to com- group and taking into account the ported study in high-risk surgical pa- plete evaluation of cardiovascular sta- interim analysis performed, was 78%. tients, we did not observe a higher in- tus by noninvasive determination of left Our study was underpowered to detect cidence of pulmonary embolism in the ventricle ejection fraction. At least one an absolute difference of 5% correspond- group assigned to PACs.11 Thromboem- echocardiographic examination was per- ing to the odds ratio of 1.24 observed in bolic complications range from less than formed in 64% of the PAC group and the study by Connors et al.4 However, 1% to up to 11% of patients with in- 78% in the control group during the ICU the design-power method does not take dwelling PACs. This incidence is obvi- stay. In this latter group, echocardiog- into account the observed difference in ously difficult to determine in critically raphy was used as a morphologic tool the 2 groups27 once the study was com- ill patients in whom the suggested clini- to assess ejection fraction and also as a pleted. In our study, we found an RR dif- cal signs of pulmonary embolism are not dynamic tool with Doppler analysis to ference in day 28 mortality associated easy to detect. This was the case in our evaluate cardiac output and estimate pul- with a narrow CI (RR, 0.97; 95% CI, study in which we did not anticipate the monary artery pressure and left ven- 0.86-1.10), suggesting an absolute dif- diagnosis of pulmonary embolism in the tricle end-diastolic pressure. Echocar- ference of –1.6% with an upper bound presence of a PAC by scheduled diag- diography can be used instead of a PAC of the 95% CI of less than 6% (95% CI, nostic testing. However, to prevent the but also as a complementary technique –9% to 5.8%). Following the CI onset of this complication, 4 preven- to PAC when available. method,28 we can conclude at an ␣ risk tive measures were suggested in our pro- In conclusion, our multicenter ran- of 5% that the absolute difference in mor- tocol. Persistent wedging of the PAC and domized trial demonstrates that the PAC tality rate between the 2 groups is no prolonged occlusion of a proximal ar- remains a safe procedure for the man- more than 7.8%. This result was close tery by an inflated balloon was forbid- agement of patients with shock, ARDS,

Downloaded From: https://jamanetwork.com/ on 09/25/2021 EARLY USE OF THE PULMONARY ARTERY CATHETER

or both. However, we did not observe a chin: A. Cariou; Hoˆ pital Europeén Georges Pompidou: 13. McCabe WR, Jackson GG. Gram negative bac- JL. Diehl, JY. Fagon, E. Guerot, J. Labrousse, G. Meyer, teremia, I: etiology and ecology. Arch Intern Med. morbidity or mortality benefit. Our re- A. Novara, C. Ract, O. Sanchez; Hoˆ tel Dieu: A. Gaubel, 1962;110:847-855. sults, which do not preclude the poten- G. Huchon, A. Lefebvre, A Rabbat, B. Vincent; Hoˆ pital 14. Le Gall JR, Lemeshow S, Saulnier F. A new Sim- Lae¨nnec: G. Meyer, O. Sanchez; Hoˆ pital Lariboisie`re: plified Acute Physiology Score (SAPS II) based on a tial impact of a goal-oriented therapy F. Baud, A. Delahaye, R. Galliot, D. Goldran-Toledano, European/North American multicenter study. JAMA. with a PAC, strengthen the suggestion P. Gueye, B. Megarbane; Hoˆ pital Pitie´-Salpeˆtrie`re:M. 1993;270:2957-2963. of the consensus statement made by the Fartouk, Y. Lefort, T. Similowski;Hoˆ pital Lae¨nnec, 15. Vincent JL, Moreno R, Takala J, et al, for the Work- Quimper: M. Coloigner, JL. Frances, M. Moriconi; Hoˆ- ing Group on Sepsis-related Problems of the Euro- National Heart, Lung, and Blood Insti- pital de La Fontaine, Saint-Denis: M. Bohus, G. Moret, pean Society of Intensive Care Medicine. The SOFA tute and the Food and Drug Adminis- A. Rhaoui; CHI Poissy-Saint-Germain-en-Laye, Saint- (Sepsis-related Organ Failure Assessment) score to de- Germain-en-Laye: JL. Ricoˆ me; Hoˆ pital Civil, Strasbourg: scribe organ dysfunction/failure. Intensive Care Med. tration that a randomized clinical trial A. Jaeger. 1996;22:707-710. with this design can be ethically per- Funding/Support: This study was supported in part 16. Fagon JY, Chastre J, Novara A, Medioni P, Gib- by De´le´ gation Re´ gionale a` la Recherche Clinique d’Ile ert C. Characterization of intensive care unit patients formed in this population of critically ill de France (AP-HP), Projet Hospitalier de Recherche using a model based on the presence or absence of patients.8 Clinique du Ministe` re de la Sante´ , and Socie´te´deRe´- organ dysfunctions and/or infection: the ODIN model. animation de Langue Française. Intensive Care Med. 1993;19:137-144. Author Contributions: Dr Richard had full access to all Role of the Sponsor: Délégation Régionale à la Recher- 17. The Acute Respiratory Distress Syndrome Net- the data in the study and takes responsibility for the integ- che Clinique d’lle de France (AP-HP) and Société de Ré- work. Ventilation with lower tidal volumes as com- rity of the data and the accuracy of the data analysis. animation de Langue Française did not participate in the pared with traditional tidal volumes for acute lung in- Study concept and design: Richard, Warszawski, design and conduct of the study, in the collection, analy- jury and the acute respiratory distress syndrome. N Engl Brochard, Teboul. sis, and interpretation of the data, or in the prepara- J Med. 2000;342:1301-1308. Acquisition of data: Richard, Warszawski, Anguel, tion, review, or approval of the manuscript. 18. O’Brien PC, Fleming TR. A multiple testing proce- Deye, Combes, Barnoud, Boulain, Lefort, Fartoukh, Acknowledgment: We thank Richard Medeiros, the dure for clinical trials. Biometrics. 1979;35:549-556. Baud, Boyer. medical editor of Rouen University Hospital-Rouen, 19. Alia I, Esteban A, Gordo F, et al. A randomized and Analysis and interpretation of data: Richard, France, for his valuable assistance in editing the manu- controlled trial of the effect of treatment aimed at maxi- Warszawski. script, Maryse Pizzato, for her valuable technical as- mizing oxygen delivery in patients with severe sepsis or Drafting of the manuscript: Richard, Warszawski, sistance in performing the study, and Bea` trice Ducot, septic shock. Chest. 1999;115:453-461. Brochard, Teboul. who performed the interim analysis. 20. Vieillard-Baron A, Girou E, Valente E, et al. Predic- Critical revision of the manuscript for important in- tors of mortality in acute respiratory distress syndrome: tellectual content: Richard, Warszawski, Anguel, Deye, focus on the role of right heart catheterization. Am REFERENCES Combes, Barnoud, Boulain, Lefort, Fartoukh, Baud, J Respir Crit Care Med. 2000;161:1597-1601. Boyer, Brochard, Teboul. 1. European Society of Intensive Care Medicine. Ex- 21. Pulmonary Artery Catheter Consensus confer- Statistical expertise: Warszawski. pert panel: the use of the pulmonary artery catheter. ence: consensus statement. Crit Care Med. 1997;25: Obtained funding: Richard. Intensive Care Med. 1991;17:I-VIII. 910-925. Administrative, technical, or material support: Richard. 2. Trottier SJ, Taylor RW. Physicians’ attitudes to- 22. Bernard GR, Vincent JL, Laterre PF, et al. Efficacy Study supervision: Richard. ward and knowledge of the pulmonary artery cath- and safety of recombinant human activated protein C The French Pulmonary Artery Catheter Study Group: eter: Society of Critical Care Medicine membership sur- for severe sepsis. N Engl J Med. 2001;344:699-709. Steering Committee: C. Richard, L. Brochard, JF. Dhain- vey. New Horiz. 1997;5:201-206. 23. Angus D, Black N. Wider lessons of the pulmo- aut, D. Dreyfuss, JY. Fagon, F. Lemaire, H. Logerot, A. 3. Robin ED. Death by pulmonary artery flow- nary artery catheter trial. BMJ. 2001;322:446. Mercat, F. Saulnier, JL. Teboul, J. Warszawski; Data and directed catheter: time for a moratorium? Chest. 1987; 24. Connors AF. Equipoise, power, and the pulmo- Safety Monitoring Committee: C. Com Nougue, L. 92:727-731. nary artery catheter. Intensive Care Med. 2002;28: Fleury, JR. Legall, C. Melot; Participating centers and 4. Connors AF Jr, Speroff T, Dawson NV, et al, for 225-226. investigators: Hoˆ pital d’Agen, Agen: B. Galiacy; F. Plou- the SUPPORT investigators. The effectiveness of right 25. Guyatt G, for Ontario Intensive Care Study Group. vier;Hoˆ pital d’Amiens Sud, Amiens: B. de Cagny, A. heart catheterization in the initial care of critically ill A randomized control trial of right heart catheteriza- Sarraj, M. Slama; Hoˆ pital d’Amiens Sud, Amiens: C. patients. JAMA. 1996;276:889-897. tion in critically patients. J Intensive Care Med. 1991; Dayen, V. Jounieaux, I. Mayeux; Hoˆ pital-Victor Dup- 5. Ivanov R, Allen J, Calvin JE. The incidence of ma- 6:91-95. ouy, Argenteuil: G. Bleichner, H. Mentec, JP. Sollet; Hoˆ- jor morbidity in critically ill patients managed with pul- 26. Rhodes A, Cusack RJ, Newman PJ, Grounds RM, pital Pellegrin, Bordeaux: JF. Cochard, A. Le´ ger, L. Petit, monary artery catheters: a meta-analysis. Crit Care Bennett ED. A randomized, controlled trial of the pul- C. Pinaquy; HoˆpitalduGeńe´ral Duchenne, Boulogne Med. 2000;28:615-619. monary artery catheter in critically ill patients. Inten- sur Mer: G. Brunin, P. Ducq, R. Pordes; Hoˆ pital Cavale 6. Iberti TJ, Fischer EP, Leibowitz AB, et al, for the Pul- sive Care Med. 2002;28:256-264. Blanche, Brest: E. L’Her, M. Moriconi, A. Renault; Cen- monary Artery Catheter Study Group. A multicenter 27. Hoenig JM, Heisey DM. The abuse of power: the tre Hospitalier, Brive: P. Chevallier, E. Karam; Hoˆ pital- study of physicians’ knowledge of the pulmonary ar- pervasive fallacy of power calculations for data analy- Antoine Bećle`re, Clamart: O. Axler, F. Brivet, B. Mega- tery catheter. JAMA. 1990;264:2928-2932. sis. Am Stat. 2001;55:19-24. rbane; Centre Hospitalier Geńe´ral, Cambrai: P. Ducq, 7. Dalen JE, Bone RC. Is it time to pull the pulmo- 28. Hauck WW, Anderson S. A proposal for inter- Ph. Morel, C. Dupres, Ch. Caulry; Hoˆ pital-Louis Mou- nary artery catheter? JAMA. 1996;276:916-918. preting and reporting negative studies. Stat Med. 1986; rier, Colombes: A. Boyer, D Dreyfuss, E. Pigne, A. Salah; 8. Bernard GR, Sopko G, Cerra F, et al. Pulmonary ar- 5:203-209. Hoˆ pital Henri Mondor, Cre´teil: M. Fartoukh, JC. Lach- tery catheterization and clinical outcomes: National 29. Shah KB, Rao TL, Laughlin S, El-Etr AA. A review erade, L. Brochard, F. Lemaire, C. Brun-Buisson; Hoˆpi- Heart, Lung, and Blood Institute and Food and Drug of pulmonary artery catheterization in 6,245 pa- tal Geńe´ral, Dunkerque: P. Martin, S. Moulront; Hoˆpi- Administration Workshop Report. JAMA. 2000;283: tients. Anesthesiology. 1984;61:271-275. tal Michallon, Grenoble: A. Bonadona, D. Barnoud, R. 2568-2572. 30. Connors AF Jr, Castele RJ, Farhat NZ, Tomashefski Hamidfar-Roy, P. Pavese, C. Schwebel, T. Stouls; Hoˆ- 9. Boyd O, Grounds RM, Bennett ED. A randomized JF Jr. Complications of right heart catheterization: a pro- pital de Biceˆtre, Le Kremlin Biceˆtre: N. Anguel, V. Caste- clinical trial of the effect of deliberate perioperative spective autopsy study. Chest. 1985;88:567-572. lain, A. Mercat, C. Richard, JL. Teboul, D. Wermert; Hoˆ- increase of oxygen delivery on mortality in high-risk 31. Mermel LA, Maki DG. Infectious complications of pital Schaffner, Lens: C. Boulle, C. Poisson, D. Thevenin, surgical patients. JAMA. 1993;270:2699-2707. Swan-Ganz pulmonary artery catheters: pathogen- L. Tronchon; Hoˆ pital Calmette, Lille: S. Beague, T. Oni- 10. Hayes MA, Timmins AC, Yau EH, Palazzo M, Hinds esis, epidemiology, prevention, and management. Am mus, F. Saulnier, S. Nseir; Hoˆ pital Roger Salengro, Lille: CJ, Watson D. Elevation of systemic oxygen delivery J Respir Crit Care Med. 1994;149:1020-1036. C. Chopin, T. Desmettre, M. Jourdain; Hoˆ pital Cen- in the treatment of critically ill patients. N Engl J Med. 32. Mimoz O, Rauss A, Rekik N, Brun-Buisson C, Le- tral, Nancy: PE. Bollaert, B. Levy, P. Gawalkiewicz; Hoˆ- 1994;330:1717-1722. maire F, Brochard L. Pulmonary artery catheterization pital de Niort, Niort: P. Charlot; Hoˆ pital Gaston Doumer- 11. Sandham JD, Hull RD, Brant RF, et al. A random- in critically ill patients: a prospective analysis of out- gues, Nıˆmes: C. Arich, N. Balmes, C. Bengler, JE. De la ized, controlled trial of the use of pulmonary-artery come changes associated with catheter-prompted Coussaye, C. Gervais; Hoˆ pital de la Source, Orleáns: N. catheters in high-risk surgical patients. N Engl J Med. changes in therapy. Crit Care Med. 1994;22:573-579. Bercault, T. Boulain, C. Fleury, K. Kuteifan, I. Runge; Paris: 2003;348:5-14. 33. Slama MA, Novara A, Van de Putte P, et al. Di- Hoˆ pital Bichat: JC. Lacherade, F. Schortgen, JF. Timsit, 12. Gattinoni L, Brazzi L, Pelosi P, et al, for SvO2 Col- agnostic and therapeutic implications of transesoph- M. Wolff, J. Chastre, A. Combes, C. Gibert; Hoˆ pital laborative Group. A trial of goal-oriented hemody- ageal echocardiography in medical ICU patients with Boucicaut: N. Deye, JL. Diehl, E. Gue´ rot, J. Labrousse; namic therapy in critically ill patients. N Engl J Med. unexplained shock, hypoxemia, or suspected endo- Hoˆ pital Broussais: JY. Fagon, A. Novara; Hoˆ pital Co- 1995;333:1025-1032. carditis. Intensive Care Med. 1996;22:916-922.

Downloaded From: https://jamanetwork.com/ on 09/25/2021