INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Right heart catheterization (RHC) is commonly used in the setting of acute lung injury (ALI) (1, 2). Common indications for catheter insertion include diagnostic separation of cardiogenic and noncardiogenic pulmonary edema and therapeutic titration of intravascular fluids, vasoactive drugs, and positive end-expiratory pressure (1). However, universally accepted indications for RHC in ALI are unavailable, and there is considerable controversy regarding the use of RHC in the intensive care unit (ICU) (3). Furthermore, the use of RHC places demands on limited resources in terms of both cost and time devoted to training, technical aspects involved in performing the procedure, and interpretation of results. Unfortunately, limited information exists regarding which patients with ALI are receiving RHC, the factors that influence the decision to perform catheterization, or the degree of variation in RHC use between different medical centers.

In the present study, we examined the early (< 72 h) use of RHC in patients with ALI who were prospectively identified in the University of Minnesota Acute Lung Injury Specialized Center of Research (SCOR) database. We hypothesized that the decision to perform RHC in ALI is influenced by both pulmonary and nonpulmonary parameters that reflected the overall severity of illness. Identifying factors strongly associated with catheter insertion might offer insights into the contemporary use of RHC in ALI and may contribute to the design of future prospective studies and their acceptance by the critical care community. In addition, we assessed the clinical impact of RHC in our study population by examining whether RHC was associated with a change in therapy.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

Study subjects were identified on a daily basis by reviewing the census in the adult ICUs at three teaching hospitals affiliated with the University of Minnesota Medical School. Enrollment of study patients was completed on March 1, 1997, having begun at Fairview-University Medical Center (Minneapolis) on June 1, 1994, Regions Hospital (St. Paul) on October 1, 1994, and Hennepin County Medical Center (Minneapolis) on April 1, 1995.

The American-European Consensus Criteria for the diagnosis of ALI were used as the criteria for SCOR database entry, including: acute onset, ratio of arterial oxygen pressure to fraction of inspired oxygen (Pa_O2/FI_O2) < 300 mm Hg, bilateral infiltrates, and pulmonary artery occlusion pressure < 18 mm Hg or the absence of clinical evidence of left atrial hypertension (4). Exclusion criteria included the presence of significant underlying chronic lung disease, organ transplantation, pregnancy, age < 18 yr, and a lung injury score (LIS) < 1.5 (5). Significant underlying lung disease was defined as chronic obstructive pulmonary disease with moderate to severe airflow obstruction, interstitial lung disease, pulmonary vascular disease, and pulmonary neoplasm.

Data Collection and Analysis

This study was approved by the institutional review boards at the three participating sites. All patient management decisions, including the decision to perform RHC, were left to the discretion of the treating physicians. Data were collected by the SCOR research nurse, using specific data collection forms. The clinical record was reviewed, including the clinical course preceding the diagnosis of ALI, admitting history, and physical examination. Demographic information, vital signs, laboratory values, net fluid balance, LIS, and ventilator parameters were assessed on Day 1 of ALI, defined as the day study entry criteria were met. Variables of the Acute Physiology and Chronic Health Evaluation (APACHE) III score were collected on the day of admission to the ICU and on Day 1 of ALI. In addition to the APACHE III and LIS scores, study parameters included: age; sex; ALI risk factor; mean arterial pressure (); heart rate; urine output; net fluid balance; Pa_O2/FI_O2; Pa_CO2; arterial pH; applied positive end-expiratory pressure (PEEP); peak and plateau airway pressures; total respiratory system compliance (Cst,rs); hematocrit; and serum values for albumin, bilirubin, blood urea nitrogen (BUN), creatinine, and lactate.

The primary predisposing condition or risk factor was determined during the initial 72 h after the diagnosis of ALI, after review of the patient medical record, laboratory results, and discussion with the attending staff. Risk factors were later separated into four categories: aspiration, pneumonia, sepsis, and other. Aspiration was defined as an observed inhalation of gastric contents, or the presence of a high clinical suspicion for aspiration and the absence of an alternative explanation for ALI. Pneumonia was defined as the presence of a new focal infiltrate, absence of a noninfectious explanation, and either signs of systemic inflammation (fever, leukocytosis) or purulent sputum production with an identifiable pathogen. Sepsis was defined according to the consensus recommendations from the American College of Chest Physicians and the Society of Critical Care Medicine (6). Other risk factors included acute pancreatitis, cardiopulmonary bypass, disseminated coagulation, drug overdose, head injury, hypertransfusion, major burn, major fracture, major trauma, near drowning, pulmonary contusion, and toxic inhalation.

The APACHE III score was calculated by methods detailed in the APACHE Data Collection Manual (7). The values of the four LIS components were summed and the final score was obtained by dividing the aggregate sum by the number of individual components used. Components of the LIS were recorded from parameters obtained in proximity to 8:00 A.M. on Day 1 of ALI. The radiographic score was determined by a chest radiologist. The Cst,rs was calculated by dividing the effective tidal volume by the plateau airway pressure minus the applied PEEP.

In patients who received RHC, the following parameters were examined: indication for RHC, central venous access site for RHC, complications associated with RHC, initial hemodynamic profile from RHC, and an assessment of whether RHC was associated with a change in therapy. Patients were considered to have received RHC if catheter-derived parameters were recorded on their intensive care flowsheets. The time of RHC placement was defined as the time when the first set of hemodynamic parameters was recorded on the flowsheet. The indication for RHC was determined by review of the medical record and procedure note. Patients could have more than one of the following indications: hypotension, severe hypoxia, oliguria, other, and unknown. Complications of RHC were determined by review of the medical record, procedure note, chest radiograph after RHC, blood and catheter tip culture results, and autopsy data when available. Complications were grouped into the following categories: pneumothorax, hemorrhage, arterial puncture, arrhythmia, infectious, and other. An infectious complication related to RHC was defined as the presence of both positive blood and catheter tip cultures with similar bacteriologic or fungal isolates. The initial hemodynamic values recorded after RHC included: right atrial pressure (Pra); pulmonary artery systolic, diastolic, and mean pressures (Ppas, Ppad, , respectively); pulmonary artery occlusion pressure (Ppao); cardiac output (CO) by thermodilution; and calculated systemic vascular resistance (SVR).

To determine whether RHC was temporally associated with a change in therapy, we reviewed the intensive care flowsheets and medication records of all patients who received RHC. The total amount of intravascular fluids, the dose of vasoactive agents, and whether or not diuretics were given during the 6 h before and after RHC were recorded. Patients were excluded from the change in therapy analysis if there were less than 2 h of data available before or after RHC. A change in intravascular fluid management was defined as a 50% or greater increase or decrease in the mean volume per hour of all fluids combined (crystalloids, colloids, blood products) and an absolute change of at least 100 ml/h. Crystalloids were separately examined using the same criteria for defining a change in therapy. A change in vasoactive therapy was defined as a 50% or greater increase or decrease in the mean infusion rate of a vasoactive agent, elimination of an agent, or addition of a new agent. For purposes of our study, a dopamine infusion rate of 3 µg/kg/min or less was not considered to be vasoactive therapy. A change in diuretic therapy was defined as one of the following observations: the administration of a diuretic before RHC but not after RHC or the administration of a diuretic after RHC but not before RHC.

Comparisons were made between patients with and without RHC using t tests, chi-squared tests, and Mann-Whitney U tests, depending on the level of measurement of the variable considered. Nonchance findings were determined based on a 0.05 level for statistical significance. Comparisons among patient groups defined by timing of RHC (none, early, or late) were made using analysis of variance followed by Tukey's Studentized range test for pairwise comparisons or Kruskal-Wallis tests followed by pairwise Mann-Whitney U tests, with a Bonferroni adjusted significance level of 0.017. No further adjustments for multiple comparisons were used. The p values are reported according to the following conventions: p < 0.001, if less than 0.001; an exact value if between 0.001 and 0.20; and not reported if exceeding 0.20. The standard deviations of all variables are provided. The relationships between variables were analyzed by Pearson product-moment correlation.

The use of RHC was modeled by multiple logistic regression with the Logistic Procedure of SAS (8). The predictor variables used in multiple logistical regression modeling were factors found to be associated with the use of RHC in univariate logistic regression, or were included because of possible clinical relevance as determined by the investigators' judgment. The best subsets selection procedure was used to identify the most informative subsets of predictors. The quality and fit of several parsimonious models were compared. A final model was selected on the basis of its improvement in fit and adequacy as assessed by likelihood ratio tests, adjusted R² values, and goodness of fit tests (9). The impact of multivariate adjustment on the relative risk estimates from this model was assessed by comparing the univariate log odds ratios to the coefficients in the model.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A total of 135 patients with ALI were entered into the study database. Demographic characteristics of the study patients are shown in Table 1. The study subjects had a mean Pa_O2/FI_O2 ratio of 130 ± 65 and a mean Day 1 APACHE III score of 73 ± 32. Fifty-four percent of patients survived to ICU discharge. Seventy (52%) of the 135 study patients underwent RHC within 3 d of meeting ALI criteria. Of these 70 patients, 57% were catheterized on Day 1, 37% on Day 2, and 6% on Day 3 of ALI. The indications for RHC were the following: hypotension (63%; n = 44), severe hypoxia (31%; n = 22), oliguria (9%; n = 6), unknown (9%; n = 6), and other (20%; n = 14). Other indications included entry into a nitric oxide protocol, PEEP trials, assessment of volume status, hemodynamic monitoring, and the exclusion of congestive heart failure. Combined indications of hypotension and hypoxia were noted in 21% (n = 15) of patients. Central venous access sites for RHC included the internal jugular vein in 54% (n = 38), subclavian vein in 33% (n = 23), femoral vein in 9% (n = 6), and unknown in 4% (n = 3). The mean duration of RHC was 6.3 d.

Clinical Parameters and Use of RHC: Univariate Analysis

RHC was associated with sepsis-related ALI, higher APACHE III scores on both ICU Day 1 and ALI Day 1, a lower , a positive net fluid balance, higher BUN and creatinine values, a lower Pa_O2/FI_O2 ratio, and use of a higher level of applied PEEP (Tables 2 and 3). Although the Pa_O2/FI_O2 ratio and level of PEEP represent two of the four parameters used to calculate the LIS, the latter was not statistically associated with RHC. Patients who received RHC had a lower survival rate than those who did not undergo catheterization (40% versus 69%; p < 0.001). Right heart catheterization was performed in four of the seven patients who died during the first 3 d after meeting ALI criteria.

Logistic Regression Analysis

The parameters examined by multiple logistic regression were first analyzed by univariate logistic regression (Table 4). In the multiple logistic regression analysis used to model predictors of RHC, the final model contained the following parameters: sepsis as the risk factor for ALI, Pa_O2/FI_O2 ratio, BUN ( 30), and (< 65 mm Hg) (Table 5). This model had an adjusted R² value of 0.299, and correctly predicted RHC use in 72% of the patients. The addition of two-way interaction terms or the APACHE III score did not significantly improve the final model. The odds ratio estimates for sepsis and based on this model were similar to those from the univariate analysis. Patients with  < 65 mm Hg at any time during the first day of ALI were 4.3 times more likely to receive RHC. Patients with sepsis as their primary risk factor for ALI were 3.1 times more likely to receive RHC compared with nonseptic patients. Patients with a Pa_O2/FI_O2 ratio < 200 on ALI Day 1 were 3.8 times more likely to receive RHC during the initial 3 d compared with patients with Pa_O2/FI_O2 ratios  200. After adjustment for the other factors, the odds ratio of an elevated BUN value was slightly less than in the univariate analysis; patients with a BUN value  30 on ALI Day 1 were 2.6 times more likely to receive RHC than patients with lower values.

Timing of RHC and Severity of Illness

To investigate the timing of RHC, study patients were separated into three groups: (1) patients who did not receive RHC during the initial 3 d after meeting ALI criteria (n = 65); (2) patients who received RHC during the initial 24 h after meeting ALI criteria ("early RHC"; n = 40); and (3) patients who received RHC between 24 and 72 h after meeting ALI criteria ("delayed RHC"; n = 30) (Table 6). The APACHE III scores on both ICU Day 1 and ALI Day 1 were highest in patients who received early RHC, intermediate in those who received delayed RHC, and lowest in those who did not receive RHC. A similar pattern was identified for the PEEP and BUN variables, with the highest values in patients receiving early RHC and the lowest values in patients who did not receive RHC. The values for Pa_O2/FI_O2 were lowest in patients who received early RHC, intermediate in those receiving delayed RHC, and highest in those managed without RHC. The difference in survival between patients who received early RHC and those who received delayed RHC was not statistically significant.

Hemodynamic Parameters from RHC

Mild to moderate pulmonary hypertension was identified in the study sample (Ppas = 39 ± 11 mm Hg; Ppad = 22 ± 6 mm Hg; = 28 ± 8 mm Hg), with Ppao = 15 ± 4 mm Hg and Pra = 13 ± 5 mm Hg. Although the initial Ppao value measured was > 18 mm Hg in 17% of patients (11 of 65 patients with Ppao values recorded immediatly after RHC placement), the Ppao values were  18 mm Hg during the first day of ALI in all study patients, meeting the American-European Consensus Criteria. In 58 patients who had simultaneous recordings of the initial Pra, Ppad, and Ppao values, the correlation of Pra and Ppao measurements (r = 0.72; p < 0.001) was stronger than the correlation between Ppad and Ppao values (r = 0.59; p < 0.001) (Figure 1). Ppao values were greater than simultaneous Pra values, with a mean difference (Ppao  Pra) of 2.4 ± 3.3 mm Hg. Cardiac output was increased (8.5 ± 3 L/ min) and SVR was reduced (587 ± 275 dynes s/cm⁵) compared with normal values.

View larger version (14K): [in this window] [in a new window]   Figure 1.   Relationship between Pra and Ppao in patients with ALI. Solid circles represent a single patient; open diamonds represent two patients.

Clinical Impact of Right Heart Catheterization

There were sufficient data on 60 patients to determine whether a change in therapy occurred within 6 h after RHC placement. A significant change in therapy occurred with regard to total intravascular fluid management in 39% (n = 23), crystalloid management in 42% (n = 25), vasoactive drug use in 33% (n = 20), and diuretic use in 38% (n = 23). Overall, a change in one or more of these interventions occurred in 78% (n = 47) of these 60 patients. Crystalloid fluids comprised 80% of the total volume of intravascular fluid administered to these patients during the 6 h before and after RHC; colloids were administered to 38% and blood products to 34%.

Complications associated with RHC included the following: pneumothorax in three patients; arrhythmia during catheter placement in one patient; central vein thrombosis in one patient; hypoxia and carotid puncture in one patient; and bacteremia with a positive catheter tip culture in three patients (4.3%). Autopsies were obtained in one-third of the patients who received RHC and died (n = 14); none of the autopsy results demonstrated any pathological condition attributed to RHC.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The clinical factors that influence the decision to insert a pulmonary artery catheter in patients with ALI are not well defined. The principal finding of the present study is that patients who underwent RHC during the first 3 d of ALI were more severely ill than those who did not undergo RHC, as reflected by higher APACHE III scores, and by greater impairments in oxygenation, blood pressure, and renal function. In addition, the overall severity of illness on Day 1 of ALI appeared to be greatest in those who received early RHC, intermediate for those who received delayed RHC, and least for those who were managed without RHC. The association between the timing of RHC and severity of illness is not unexpected, because sicker patients may be more likely to receive invasive interventions early in their hospital course. In contrast, in patients who are less ill, the decision to place an RHC may be deferred, with the use of invasive interventions guided by the subsequent clinical course.

Sepsis was the most common risk factor for the development of ALI and was strongly associated with the use of RHC. Other nonpulmonary parameters associated with the use of RHC included a reduction in , increased net fluid balance, and the presence of azotemia, each of which is a common feature of sepsis. In light of the cardiovascular instability and renal impairment that often result from severe sepsis, the association of sepsis with RHC was not surprising. With regard to pulmonary parameters, use of RHC was associated with the presence of a lower Pa_O2/FI_O2 ratio and use of higher levels of applied PEEP. Again, the latter finding was not unexpected, as severe hypoxemia may well have encouraged RHC-guided efforts to reduce lung water, and high levels of PEEP have well known effects on hemodynamic function. Interestingly, the aggregate LIS on ALI Day 1 was not associated with the use of RHC, even though the Pa_O2/FI_O2 ratio and level of PEEP comprise two of the four components of the LIS.

Information regarding the frequency with which patients with ALI undergo RHC is limited (1). Right heart catheterization was performed in 52% of our patients during the first 3 d of ALI. Of those who received RHC, 94% underwent catheterization during the first 2 d after meeting ALI criteria. In a study from the University of Utah, Suchyta and colleagues reported that RHC was used in 75% of their patients with ALI, but information regarding the timing of RHC or the indications for the procedure was not reported (10). In the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatment (SUPPORT), one-third of the patients with acute respiratory failure received RHC during the first 24 h of intensive care; however, the percentage of patients with ALI was not reported (3). In a multicenter study from Finland, 11% of the patients with pulmonary edema received RHC, but the proportion of patients with ALI was not given (11).

The use of RHC in ALI could be influenced by factors unrelated to severity of illness. Diagnostic uncertainty would likely encourage catheterization. Indeed, a number of epidemiologic and therapeutic studies used specific Ppao criteria in their definitions of ALI (12). More recent definitions of ALI have not required the measurement of Ppao, allowing for the clinical exclusion of congestive heart failure (4, 5, 17). A recent study compared three definitions of ALI that did not require placement of RHC with a stricter definition of ALI requiring a Ppao < 18 (17). The definitions examined in this study included the LIS (5), the American-European Consensus Conference criteria (4), and a modified LIS that required a Pa_O2/FI_O2 ratio of < 175 and bilateral infiltrates on chest radiographs (17). The investigators concluded that RHC did not improve diagnostic accuracy. In our study, hemodynamic parameters from RHC were available in only 40 (30%) patients on the day that ALI criteria were met.

The therapeutic strategy used to manage patients with ALI may also influence the frequency of RHC. A retrospective study found that reduction of Ppao in patients with ALI was associated with improved survival (18). In another study, achievement of a lower positive fluid balance by diuresis and fluid restriction led to a reduction in ventilator-supported and intensive care days, although a benefit in mortality was not observed (19). The authors of the latter study suggested that the principal benefit of RHC in patients with ALI may be to avoid complications of therapy related to excessive reductions or elevations of Ppao, and that limiting positive fluid balance may reduce morbidity. A different strategy for management of critically ill patients, including those with ALI, is to maximize oxygen delivery by volume expansion, red blood cell transfusion, and use of inotropic agents (20). Although two randomized studies failed to show a survival benefit in patients who were treated with RHC-guided optimization of oxygen delivery (21, 22), this approach is still used in some centers. Clearly, the adoption of rigorous therapeutic strategies seeking to achieve either the lowest Ppao consistent with adequate tissue perfusion or the highest possible oxygen delivery would dictate almost routine use of RHC in patients with ALI. Although individual differences in management among treating physicians undoubtedly existed in our study, neither aggressive diuresis to the limits of hemodynamic tolerance nor augmentation of oxygen delivery to predetermined end points is routinely used in the management of our patients with ALI.

Clinical estimates of Ppao and cardiac output are frequently inaccurate in critically ill patients (23). Furthermore, hemodynamic data obtained from RHC in the critically ill commonly lead to a change in therapy (23). For instance, Mimoz and colleagues reported that physicians correctly predicted hemodynamic profiles in critically ill patients in only 56% of cases (25). The investigators noted that information obtained from RHC prompted changes in therapy in 58% of their patients, and in 63% of patients who were unresponsive to empiric therapy. We noted a similar association between RHC and change in therapy, with 78% of our patients receiving a change in therapy after RHC. However, our ability to determine that a change in therapy was prompted by the RHC data is limited by the retrospective nature of our study design.

Central venous pressure (CVP), or Pra, is highly correlated with Ppao in normal individuals (26). However, this relationship may be markedly impaired in some patients with critical illness (26, 27). For instance, compared with the strong correlation between Pra and Ppao values found in normal individuals (r = 0.92), Civetta and Gabel demonstrated a weak correlation between Pra and Ppao values in patients with multiple injury (r = 0.27) (26). In addition, Forrester and colleagues observed a weak correlation of Pra and Ppao values in patients with acute myocardial infarction (r = 0.45) (27). We observed a moderately strong correlation between Pra and Ppao measurements (r = 0.72) (Figure 1), with a mean difference (Ppao  Pra) of 2.4 ± 3.3 mm Hg. Thus, a central venous catheter pressure measurement may provide a reasonable approximation of Ppao in many patients with early ALI and mild to moderate pulmonary hypertension.

Complications associated with RHC may be classified as those that occur during catheter insertion and those that develop after catheter insertion. Our patients experienced a spectrum and incidence of insertion-related complications that were similar to complications reported by other centers, including arterial puncture, arrhythmia, and pneumothorax (28). Complications identified after catheter insertion included a catheter-associated venous thrombosis in one patient, and bacteremia with positive catheter tip cultures in three patients (4.3%). Recognizing the limitations of our observational study design, the true incidence of infectious complications from RHC may be underestimated by our data. For comparison, other centers have reported the incidence of RHC-associated bacteremia to be 0.0 to 5.3% of patients (29). Of interest, previous studies have revealed an increase in the risk of catheter-related infections when the duration of catheterization exceeds 3 d (29). Thus, the duration of RHC noted in our study (mean, 6.3 d) may have increased our risk of catheter-related infections.

Controversy regarding the use of RHC in critically ill patients has been generated by the results of recent retrospective studies (3, 30, 31). A study by Connors and colleagues (3) generated considerable controversy within the critical care community (32). They reported RHC to be associated with increased mortality, even after apparent correction for severity of illness. In our study, the survival rate of patients who underwent RHC was significantly lower than those who were not catheterized (40% versus 69%; p < 0.001). Although we cannot exclude the possibility that RHC contributed to this adverse outcome, it seems more likely that the increased mortality rate was caused by the greater severity of illness in patients who received RHC.

The major limitation of our study is its retrospective-observational design, preventing a clear cause-and-effect conclusion. In addition, our analysis was limited to the demographic and clinical parameters available in our SCOR database, and the availability of medical records for review. The decision to perform RHC was made by the treating physician, which included multiple physicians at three institutions with primary training in either medicine, surgery, or anesthesiology. This heterogeneity in clinical institution, type of ICU (medical, surgical, coronary), and training background of treating physicians increases the generalizability of the conclusions.

In summary, patients with ALI who underwent RHC within 72 h of diagnosis were more severely ill than ALI patients who were not catheterized, and the timing of RHC was related to the severity of ALI. Furthermore, RHC was associated with a change in therapy in the majority of our patients. Of interest, we also observed a moderately strong correlation between the initial Pra and Ppao measurements in our patient population with early ALI and mild to moderate pulmonary hypertension. Ultimately, well-designed multicenter, randomized clinical trials are needed to define the proper role of RHC in the management of patients with ALI. However, we recognize that designing a prospective trial of RHC that is acceptable to the intensive care community will be challenging (1, 2, 33). Hopefully, the information presented in the present study may shed some light on the contemporary use of RHC in patients with ALI, and may contribute to the design of randomized clinical trials in this patient population.