James O'Brien

Excess Body Weight Is Not Independently Associated with Outcome in Mechanically Ventilated Patients with Acute Lung Injury

Context Although obesity poses many health risks, clinicians have been uncertain whether excess body weight adversely affects the outcomes of severe illnesses such as acute lung injury requiring mechanical ventilation. Contribution Among patients in a trial of mechanical ventilation strategies, obese patients and lean patients had similar mortality and ventilation outcomes. Implications Physicians should not assume that intubated obese patients fare worse than those who are of normal weight. Whether excess body weight puts patients at risk for poor outcomes in other types of critical illness is a subject for future study. The Editors Sixty-four percent of U.S. adults are overweight or obese, and this trend is accelerating (1, 2). Despite the well-described chronic health consequences of excess weight (3), we know little about the effect of obesity on outcomes from acute illnesses, particularly those requiring admission to the intensive care unit. Obese patients have a greater prevalence of comorbid conditions that may affect outcome (3), and they experience physiologic changes (4, 5) that may impair their ability to compensate for the stresses of critical illness. Because of these findings, conventional wisdom holds that obesity increases mortality and morbidity for patients in the intensive care unit. However, an independent effect of obesity on outcome from critical illness has never been conclusively demonstrated. If, in fact, obese persons are at risk, investigators should determine the mechanism of this increased risk and target interventions to this group. Acute lung injury is an inflammatory pulmonary condition associated with a variety of initiating insults. Acute lung injury is a frequent cause of respiratory failure requiring mechanical ventilation and a common indication for admission to the intensive care unit. The reported mortality rate is 40% to 60% (6). We performed a secondary analysis of a randomized trial of ventilator management in patients with acute lung injury (7) to better describe the influence of excess body weight on the outcome of critical illness. In that trial, patients randomly assigned to low tidal volume had better outcomes than patients assigned to high tidal volume. The experimental protocols for this trial required measurement of height to determine assigned tidal volume. This measurement also allowed calculation of body mass index (BMI) for each patient, a variable not often recorded for critically ill patients. Some argue that larger tidal volumes are beneficial for obese patients requiring mechanical ventilation (8). This raises concern that patients with different BMIs may require different ventilator strategies. By evaluating the interaction between the assigned ventilator protocol and BMI, we were able to determine whether the beneficial effect of lower tidal volume extends to obese patients with acute lung injury. Methods Setting and Sample We examined data on patients who participated in the National Heart, Lung, and Blood Institutes multicenter, randomized trials of the Acute Respiratory Distress Syndrome Network (7, 9, 10). Of the 902 patients in these studies, the first 861 participated in a randomized trial of mechanical ventilation that compared lower tidal volume with higher tidal volume (6 mL/kg of predicted body weight vs. 12 mL/kg, respectively). In a factorial design, 2 other trials evaluated ketoconazole versus placebo (234 patients) or lisofylline versus placebo (194 patients). After the ventilator trial ended because it showed a significant benefit associated with lower tidal volumes, an additional 41 patients received lisofylline or placebo plus the lower tidal volume strategy. Neither lisofylline (9) nor ketoconazole (10) affected outcomes of acute lung injury. Details of these studies and inclusion and exclusion criteria are described elsewhere (7, 9, 10). In brief, patients were eligible if they required mechanical ventilation and met diagnostic criteria for acute lung injury. Patients with a weight-to-height ratio (kilograms divided by centimeters) of 1.0 or greater were excluded. Analysis was done on an intention-to-treat basis. Measures of Excess Body Weight We used BMI as a measure of the degree of excess body weight. We calculated BMI from data in enrollment documents by dividing the patients body weight in kilograms by the square of his or her height in meters. Ventilator and Weaning Protocols The protocol for mechanical ventilator management is described elsewhere (7). The major difference between the two study groups was the selected tidal volume. Investigators calculated predicted body weight from the patients height and sex and used this predicted weight to determine the initial tidal volume for each patient. In the group assigned to higher tidal volumes, the initial tidal volume was 12 mL/kg of predicted body weight. In the group treated with lower tidal volumes, the initial tidal volume was 6 mL/kg. Investigators performed a daily weaning screen on every patient in an attempt to standardize the process of liberation from mechanical ventilation. Outcome Measures The primary outcome measure was survival to 28 days after study enrollment. Secondary dependent variables included achievement of unassisted ventilation by day 28, survival to discharge to home or to 180 days (the duration of follow-up in the primary studies), and the number of ventilator-free days. Unassisted ventilation was defined as liberation from mechanical ventilation for 48 or more consecutive hours. The number of ventilator-free days is the number of days of unassisted ventilation from day 1 to day 28. Statistical Analysis We performed unadjusted analyses by comparing values for patients across the 3 BMI categories (normal vs. overweight vs. obese) for outcome variables of interest and for other predictors. Unadjusted associations between other predictors and the outcomes were also explored. We used a 2-sided Fisher exact test for dichotomous variables; a 2-sided likelihood ratio chi-square test for nondichotomous categorical variables; and a KruskalWallis test, analysis of variance, or Wilcoxon rank-sum test for continuous variables, as appropriate. We constructed correlation matrices to guide regression estimation. We used logistic regression for the dichotomous outcome variables and linear regression for the continuous outcome variables. To estimate the base regressions, we selected variables for inclusion on the basis of several considerations, including significant differences in unadjusted analyses and clinical relevance. Among variables with a correlation greater than 0.50, only 1 was considered for inclusion to minimize multicollinearity. Variables that were thought to be strongly clinically relevant to the outcome and those found to have a statistically significant unadjusted effect (P < 0.05) were ultimately included in the base model. Variables in addition to those in Table 1 that were evaluated for inclusion were study site, ethnicity, diagnosis of diabetes, peak glucose level within 24 hours of enrollment, nonpulmonary organ failures, use of vasopressors, fluid balance in the 24 hours before study entry, and pneumonia as primary cause of lung injury. Unless otherwise stated, variables reflected the patients clinical state at the time of study enrollment. Table 1. Characteristics of the Sample After estimation of the base regressions, we forced the indicators of excess body weight into the model and determined their predictive values. We performed analyses in several different ways. We used the National Heart, Lung, and Blood Institute divisions of BMI to categorize patients as normal weight (BMI of 18.5 to 24.9 kg/m2), overweight (BMI of 25.0 to 29.9 kg/m2), or obese (BMI 30 kg/m2). To test any effect across BMI category, we used a categorical variable with 2 degrees of freedom in the regression. Because of concern that we would not be able to detect an effect that was nonincremental, we compared the overweight BMI group with the normal BMI group and the obese BMI group with the normal BMI group. To examine whether the efficacy of lower tidal volume ventilation varied by degree of excess body weight, we estimated the interaction effects between BMI group and assignment to the higher tidal volume protocol. Because the interaction effects between BMI category and treatment assignment were not significant (as tested by using a likelihood ratio test with 2 degrees of freedom), a main effects model was fit. This likelihood-ratio test was also used to test the significance of the 3-category BMI variable. To examine the patients with extreme excess body weight, patients were divided into 4 BMI categories (normal, overweight, obese [BMI of 30 to 39.9 kg/m2], and severely obese [BMI 40 kg/m2]). This categorical variable with 3 degrees of freedom was also tested in the regression. In addition to these analyses, we also used BMI as a continuous variable. Because critically ill patients often receive fluid resuscitation or diuresis, we recalculated BMI as adjusted for the net fluid balance for each patient over the 24 hours before study entry (fluid-adjusted BMI). Negative fluid balances were added to the patients body weight and positive fluid balances were subtracted from his or her weight to calculate BMI. We substituted the median fluid balance for the patients study site if the individual fluid balance was unavailable (51 records). We used a MantelHaenszel chi-square test for the ordinally categorical variables and a Wilcoxon rank-sum test for continuous variables. Pearson chi-square test produced results similar (P > 0.2) to those of the MantelHaenszel test. We used SAS software, version 8.02 (SAS Institute, Inc., Cary, North Carolina), for all analyses. A P value less than 0.05 was considered statistically significant. Protection of Human Subjects The institutional review boards of each participating center approved the primary studies. Patients or their sur

New approaches to the treatment of sepsis

The clinical spectrum of sepsis, severe sepsis, and septic shock is responsible for a growing number of deaths and excessive health care expenditures. Until recently, despite multiple clinical trials, no intervention provided a beneficial outcome in septic patients. Within the last 2 years, studies that involved drotrecogin alfa (activated), corticosteroid therapy, and early goal-directed therapy showed efficacy in those with severe sepsis and septic shock. These results have provided optimism for reducing sepsis-related mortality.

The Influence of Treatment Effect Size on Willingness to Adopt a Therapy

Background. Physicians are slow to adopt novel therapies, and the reasons for this are poorly understood. The authors sought to determine if the size of the treatment effect of a novel therapy influences willingness to adopt it. Methods. We developed 2 experimental vignette pairs describing a trial of a therapy for a hypothetical disease that showed a statistically significant mortality benefit. The size of the mortality effect was varied in vignettes of a pair (3% v. 10%). The 2 experimental vignette pairs differed in whether study enrollment was reported. Vignettes were mailed to a random sample of physicians using an intersubject design. The main study outcome was respondents’ willingness to adopt the hypothetical therapy, based on the results of the hypothetical trial. Results. There were 124 and 89 respondents to vignette pairs 1 and 2, respectively. In vignette pair 1, 91% versus 71% of respondents adopted the therapy when it reduced mortality by 10% and 3%, respectively (P = 0.0058). For vignette pair 2, 88% versus 51% of respondents adopted the therapy when it reduced mortality by 10% and 3%, respectively (P = 0.0002). In both vignette pairs, nonadopters were more likely than adopters to report side effects of the therapy as a principal reason for their decision. Conclusions. In this study, respondents were less likely to adopt a lifesaving therapy if its associated mortality reduction was 3% compared to 10%. Because most therapies for major medical conditions reduce mortality within or below this range, and because there were no opportunity costs associated with the adoption of the therapy, we believe that this effect represents a bias. Further investigation will be required to determine its prevalence and mechanism.

Recombinant human activated protein C sentenced to the death of a thousand cuts

I n this issue of Critical Care Medicine, Dr. Laterre and colleagues (1) present longer term follow-up data to the clinical trial of recombinant human activated protein C (rhAPC) in severe septic patients with lower severity of illness (ADDRESS) (2). There was no statistically significant difference in outcomes between subjects in the treatment and placebo groups. The report provides reassurance that there is no late effect of rhAPC. With data available for approximately 90% of subjects, the overall 1-yr mortality rate was 34.1%. Because most deaths occurred within 90 days, this is a reasonable mortality landmark for determining a benefit in clinical trials of severe sepsis. Although this study adds to a multitude of studies regarding rhAPC, we seem no closer to a consensus about the drug. PROWESS included patients with severe sepsis and an a priori plan for a primary analysis stratified by a variety of clinical factors (3). Presumably, this would guide subsequent studies, should there be no overall effect. However, there was a mortality benefit for the whole cohort (absolute risk reduction, 6.1%; 95% confidence interval [CI], 1.9–10.3%). After analyzing subgroups, regulatory bodies concluded that the sickest patients benefited the most, and rhAPC was approved for those at a high risk for death. Statistically, an Acute Physiology and Chronic Health Evaluation (APACHE) II score 25 provided the best discrimination of those likely to benefit (absolute risk reduction, 12.8%; 95% CI, 6.2–19.4%) (4). ADDRESS was designed to answer the question about lower-risk patients for whom rhAPC was not already indicated (2). This study was stopped early because of a projected lack of effect. Among ADDRESS patients with an APACHE II score 25 (12% of those enrolled), those treated with rhAPC had numerically higher mortality than those assigned to placebo (absolute risk increase, 4.8%; 95% CI, 4.9% decrease to 14.2% increase). This raised questions about PROWESS’ conclusions, despite the clear differences in studies, as shown by the dissimilarity in mortality among placebo patients (APACHE II 25: PROWESS 43.7%, ADDRESS 24.7%; APACHE II 25: PROWESS 19.0%, ADDRESS 16.0%). The sequence in which we received data produced some of the uncertainty about rhAPC. PROWESS demonstrated a benefit, but only for half of the subjects. Then, ADDRESS showed no effect. We tend to simplify these results as contradictory: one was positive, and one was negative. If the sequence was changed, would the same conclusions have been reached? If PROWESS was without overall effect but the subgroup at highest risk showed a benefit, there would have been a second study examining only those at a higher risk. If that study demonstrated rhAPC was beneficial, we would have reached a consensus that rhAPC should be used for severe septic patients at a high risk for death. Instead, rhAPC has become contentious (5–8). Given the heterogeneity of severe sepsis, it would be surprising if any single agent were uniformly effective. We certainly would not expect a single chemotherapeutic agent to be useful in all cancers. If a single drug were to be successful, it would have to address pathways common to the majority of patients. Alternatively, it may work only in a subgroup of patients with the targeted derangement. rhAPC may fit such a description. Considering the acceptance of other therapies without definitive mechanisms of action (e.g., lower tidal volume ventilation [9]), it is peculiar that similar uncertainty causes a reluctance to use rhAPC. The increased risk of bleeding is important when considering rhAPC use. This is a complex issue because we cannot confidently predict who will benefit from treatment and who will bleed. We must determine the net effect for the entire study population and imperfectly extrapolate this as a probability of benefit for an individual patient. When rhAPC is given, there is no immediate feedback that a clinician made the correct decision. When thrombolytics are administered for acute myocardial infarction, chest pain resolves and the electrocardiogram normalizes. With rhAPC, the patient lives, dies, or bleeds. The clinician only credits rhAPC with bleeding. This favors an omission bias–omitting action with a net benefit to avoid the possibility of harm (10, 11). Another rhAPC study is planned to “help clinicians better identify severe sepsis patients at high risk of death who are more likely to benefit (12).” Although it may not be intended as a “tie-breaking” study, some might view it as such. Dichotomizing studies as either positive or negative contributes to the confusion about rhAPC. PROWESS and ADDRESS did not demonstrate opposite results of equal magnitude. Among high-risk patients, PROWESS had very convincing results for a benefit and ADDRESS was too underpowered to show anything definitive. If we combine the high-risk patients from ADDRESS and PROWESS, the composite absolute risk reduction is 7.8% (95% CI, 2.3–13.3%; p .006). However, because both studies were halted before target enrollment, it is difficult to determine the influence that early stopping had on the common estimate of effect. The upcoming study should be designed with consideration of prior results. Early stopping should be discouraged. Suppose 500 subjects are randomized and the outcomes are identical to the high-risk group in ADDRESS. Enrollment in this study could be halted because of a trend toward harm with p .23 and without definitive results. If we then combine data from all three studies, there is a composite 3.9% absolute risk reduction (95% CI, 0.6% risk increase to 8.4% risk reduction) *See also p. 1457.