Intensive Care Med (2004) 30:437–443 DOI 10.1007/s00134-003-2095-2
Mait Garrouste-Orgeas Gilles Troch Elie Azoulay Antoine Caubel Arnaud de Lassence Christine Cheval Laurent Montesino Marie Thuong Franois Vincent Yves Cohen Jean-Franois Timsit Received: 21 May 2003 Accepted: 5 November 2003 Published online: 6 February 2004 Springer-Verlag 2004 Electronic Supplementary Material Supplementary material is available in the online version of this article at http:// dx.doi.org/10.1007/s00134-003-2095-2. The members of the OUTCOMEREA study group are listed in the appendix. Financial support: OUTCOMEREA is supported by non-exclusive educational grants from Aventis Pharma, France, Wyeth, and the Centre National de la Recherche Scientifique (CNRS). M. Garrouste-Orgeas ()) · A. Caubel Service de Ranimation Polyvalente, Hpital Saint Joseph, 185 rue Raymond Losserand, 75014 Paris, France e-mail:
[email protected] Tel.: +33-1-44123415 Fax: +33-1-44123280 G. Troch Surgical ICU, Antoine Bclre Teaching Hospital, Clamart, France E. Azoulay Medical ICU, Saint Louis Teaching Hospital, 75010 Paris, France A. de Lassence Medical ICU, Louis Mourier Teaching Hospital, 92701 Colombes, France C. Cheval Surgical Vascular ICU, Saint Joseph Hospital, 75014 Paris, France
ORIGINAL
Body mass index An additional prognostic factor in ICU patients
L. Montesino · J.-F. Timsit Medical ICU, Bichat Teaching Hospital, 75018 Paris, France M. Thuong Medical-Surgical ICU, Delafontaine Hospital, Saint Denis, France F. Vincent Renal-ICU, Tenon Teaching Hospital, 75020 Paris, France Y. Cohen Medical ICU, Avicenne Teaching Hospital, 93009 Bobigny, France J.-F. Timsit Epidemiology and Biostatistics Department, Bichat Teaching Hospital, 75010 Paris, France
Abstract Objective: To examine the association between body mass index (BMI) and mortality in adult intensive care unit (ICU) patients. Design: A prospective multi-center study. Interventions: None. Methods: A cohort study (yielding the OUTCOMEREA database) was conducted over 2 years in 6 medicalsurgical ICUs. In each participating ICU, the following were collected daily: demographic information, admission height and weight, comorbidities, severity scores (SAPS II, LOD, and SOFA), ICU and hospital lengths of stay, and ICU and hospital mortality rates. Results: A total of
1,698 patients were examined and divided into 4 groups based on BMI: <18.5, 18.5–24.9, 25–29.9, and >30 kg/m2. These groups differed significantly for age, gender, admission category (medical, scheduled surgery, unscheduled surgery), ICU and hospital lengths of stay, and comorbidities. Severity at admission and within the first 2 days was similar in the 4 groups, except for the SOFA score. Overall hospital mortality was 31.3% (532 out of 1,698 patients). By multivariate analysis, a BMI below 18.5 kg/m2 was independently associated with increased mortality (odds ratio 1.63; 95% confidence intervals 1.11–2.39). None of the other BMI categories were associated with higher mortality and even a BMI>30 kg/m2 was protective of mortality (odds ratio 0.60, 95% confidence intervals 0.40–0.88). Conclusions: A low BMI was independently associated with higher mortality and a high BMI with lower mortality in this large cohort of critically ill patients. Since BMI is absent from currently available scoring systems, further studies are needed to determine whether adding BMI would improve the effectiveness of scores in predicting mortality. Keywords Intensive care unit · Body mass index · Evaluation studies · Severity scores · Mortality
438
Introduction
Data collection
The relationship between body weight and mortality in critically ill patients is unclear. Long-term studies have found higher mortality rates in both obese and thin individuals [1, 2, 3]. There is overwhelming evidence that the prevalence of obesity has been increasing steadily since the 1960s in industrialized countries. In 1980, in the US, 34.9% of individuals were overweight and 2.9% were morbidly obese [4]. A recent study conducted from 1988 to 1994 in the US found that 60% of the general population was overweight (BMI>25, <30) or obese (BMI>30) [5]. In France, lower prevalences have been found, with a BMI between 30 and 39 in 12% of the general population and greater than 40 in only 0.5% [4]. In patients outside the ICU, the relationship between BMI and mortality remains a focus of controversy. The increased risk of mortality reported in patients with a low BMI appears to be a direct consequence of illness-related weight loss [6]. In a recent large prospective cohort study conducted in the US, the risk of all-cause mortality adjusted for smoking status and pre-existing illness increased throughout the range of moderate to severe overweight in both genders and in all age groups [7]. However, a protective effect of obesity compared to normal weight has been reported in older communitydwelling individuals [8]. Two recent studies [9, 10] were available for assessing a potential association between BMI and mortality in ICU patients giving contradictory results. A retrospective study conducted in two medical ICUs [9] found that morbid obesity (40 kg/m2, n=117) was independently associated with longer duration of hospital stay and mechanical ventilation and with higher mortality compared to a group of non-obese patients (BMI<30 kg/m2) with similar APACHE II scores. A large study of 67,646 patients did not identify increased mortality in the overweight, obese or severely obese patient groups [10]. We hypothesized that BMI may add prognostic information of use for predicting mortality. To explore this hypothesis, we conducted a prospective multicenter study to examine the association between BMI and mortality in a large prospective cohort of patients from six ICUs in France.
Data were collected daily on computers by ICU physicians closely involved in establishing the database. All codes and definitions were written before data collection was started. All data were reviewed by a steering committee. A control quality of the data was made by a random check of 2% of the database in each center by a senior ICU physician of another ICU with inter rater correlation coefficients between 0.67 and 1 for clinical variables severity and organ dysfunction scores, and kappa coefficients for qualitative variables ranged between 0.5 and 0.9. The following information was recorded prospectively: demographic characteristics (age, sex, weight, height), underlying diseases using the McCabe score [11] and Knaus classification [12], presence of diabetes mellitus (with the type and complications), admission category (medical, scheduled or unscheduled surgery), admission diagnosis, duration of the ICU stay and acutecare hospital stay, and vital status at ICU and hospital discharge. SAPS II at admission was computed using the worse clinical and biological data of the first 24 h of ICU stay. Organ dysfunction and severity scores using the simplified acute physiologic score (SAPS II) [13], logistic organ dysfunction score (LOD score) [14], and sepsis-related organ failure assessment (SOFA score) [15] were also computed daily using clinical and laboratory data recorded on each calendar day. Day 1 was defined as the interval from the time of admission to 8 am on the next day, all other days were calendar days from 8 am to 8 am. In addition, the intensity of care provided in the ICU was evaluated using the Omega score [16], which comprises 47 diagnostic and therapeutic items weighted from 0 to 10 points according to the corresponding workload. The items were recorded daily (e.g., mechanical ventilatory support), when performed (e.g., dialysis), or once during the ICU stay (e.g., central line or vasoactive drugs). The total Omega score is the sum of points accumulated during the ICU stay. Weight and height were measured at admission in sedated patients and estimated in others. In sedated patients the body length was measured with a tape measure at admission. The accuracy of the method was not tested. BMI (weight in kilograms divided by the square of the height in meters) values were calculated retrospectively and categorized using World Health Organization cut-off points, as follows: normal 18.5–24.9; grade 1 overweight, 25–29.9; grade 2 overweight, 30–39.9; and grade 3 overweight 40 [17]. We added a fifth category, as in the study of Calle et al. [7], for BMI values lower than 18.5. However, only 23 patients (0.01%) had a BMI>40 kg/m2 and we consequently combined the 30–39.9 and 40 BMI groups into a single group, so that we had only 4 BMI groups in all (<18.5, 18.5–24.9, 25–29.9 and >30 kg/m2).
Patients and methods Study population A prospective multicenter cohort study (yielding the OUTCOMEREA database) was conducted over 2 years in 6 ICUs in France. All were situated in the Paris metropolis in university hospitals, two were medical, two surgical and two were medical-surgical ICUs. All patients who were older than 18 years of age and had an ICU stay of at least 48 h were included.
Statistical analysis The results are reported as medians with their inter-quartile ranges or as proportions. Categorical variables were compared using the Fisher exact test and continuous variables using the Wilcoxon rank sum test for unpaired data or the Kruskal-Wallis test, as appropriate. The assumption that quantitative variables were linear in the logit was checked using cubic polynomials and graphical methods. When this assumption was not verified, the variables were transformed into dummy variables based on their median value. Multivariate analysis was performed using stepwise forward logistic regression with hospital mortality as the outcome variable of interest. Variables significantly associated with hospital mortality in the univariate analysis and variables that were not balanced among BMI categories were first introduced into the model. Second, the BMI category was forced into the model as three dummy variables. Two-way interactions were tested in the final model. Odds ratios and their confidence intervals were calculated. Statistical analyses were performed using S-Plus (Mathsoft Inc.,
439
Seattle, WA) and SAS 8.0 (SAS Inc. Cary, NC) software. P values of less than 0.05 were considered significant. A standardized mortality ratio (SMR) and 95% confidence intervals was calculated for each of the BMI categories. The SMR is the ratio of actual to predicted mortality (as predicted by the SAPS II on admission). An SMR less than 1 suggests that fewer patients died than would have been expected and vice versa.
not significantly different among the four groups, except for the SOFA score (Table 1). Similarly, the number of patients transferred to the ICU from wards was not significantly different among the BMI groups (Table 1). Mortality
Results Patients The database included 1,698 patients, of whom 1,072 (63%) were medical patients. About half the patients (836 out of 1,698, 49.2%) were admitted from a ward and half from the emergency department. Nearly half the patients had normal BMI values, slightly over 10% were clinically malnourished, and about 40% were overweight or obese (i.e., Table 1). Tables 1, 2, 3, 4 show that these four groups differed significantly for age, gender, admission category, ICU and hospital lengths of stay, admission diagnosis, lung function abnormalities and other comorbidities (McCabe score), immunosuppression, and diabetes mellitus. In the 250 patients with immunosuppression, the only underlying disease where the rate of occurrence differed significantly across the four weight groups was AIDS, which was more common in the low BMI group (<18.5 kg/m2). Severity of the acute illness, as assessed using several scores at ICU admission and within the first 2 days, was Table 1 Patient characteristics according to body mass index
BMI groups
Weight (kg) Height (cm) BMI (kg/m2) Age (years) Males/females (n) Admission category (%) Medical Scheduled surgery Unscheduled surgery OMEGA score OMEGA 1 OMEGA 2 OMEGA 3 Transfer from ward, n (%) ICU stay (days) Hospital stay (days)
The total hospital mortality was 31.3% (532 out of 1,698), but the mortality rate differed strongly among the 4 BMI groups. ICU and hospital mortality were significantly higher in patients with a BMI<18.5 kg/m2 (28.6% or 54 out of 189 and 43.9%, 83 out of 189, respectively) and lower in patients with a BMI>30 kg/m2 (18%, 41/227 and 25.1%, 57/227, respectively). The mortality rate was significantly higher in the patients with neutropenia (25/ 38, 65%, vs. 507/1,660, 30.5%, P<0.0001), steroid treatment (56/125, 44.8% vs. 476/1,573, 30.2%, P=0.0007), or chemotherapy (46/100, 46% vs. 486/ 1,598, 30.4%, P=0.001). Table S1 displays the factors significantly associated with hospital mortality in the univariate analysis which were included in the logistic regression analysis. As shown in Table 5, the factors independently associated with increased mortality were older age, worse SAPS II at admission, worse SOFA score on day 2, transfer from a ward, and more than one chronic illness. Fatal disease by the McCabe score was significantly associated with mortality. A BMI of less than 18.5 kg/m2 was independently associated with higher mortality (OR, 1.63; 95% confidence interval, 95%CI,
<18.5
18.5-24.9
25–29.9
>30
n=189
n=806
n=476
n=227
48.5 (43–51) 168 (160–172) 17 (16–17.7) 62 (45–73) 118/71
62 (56–69) 77 (71–82) 90 (83–103) 176 (162–175) 170 (161–175) 165 (159–172) 22 (20.7–23.4) 27 (26–28.3) 32 (31–35.6) 65 (48–75) 69 (56–77) 65.5 (57–74) 532/274 305/171 112/115
141 (74) 16 (8.5) 32 (17) 202€248 18.6€14.4 17.5€16 168.6€258 101 (53.5)
570 (70.7) 84 (10.5) 152 (18.8) 207€300 17.2€14.7 16.6€19.2 158€272 401 (49.7)
6 (4–14) 21 (10–40)
6 (4–11) 20 (10–37)
319 (67) 76 (16) 81 (17) 219€291 18.3€13.9 17.7€23.8 165€258 230 (48.3) 6 (4–13) 20 (11–38)
143 (63) 36 (15.8) 48 (21) 299€382 20.7€14.2 20€21 246€356 104 (48.8) 7 (4–19) 25 (14–48)
p-value
0.0001 0.0002 0.0001 0.0001 0.0005
0.003 0.008 0.10 0.005 0.44 0.003 0.0005
Data are means€SD or medians (range) unless otherwise indicated. BMI weight in kilograms divided by the height in meters squared. OMEGA intensity of care scoring system. OMEGA 1 mainly reflects the treatments and treatment procedures (e.g., catheters, vasopressors, blood transfusion or parenteral nutrition). OMEGA 2 surgical and investigational procedures. OMEGA 3 the duration of mechanical ventilation, length of stay, and number of high-intensity treatments such as dialysis.
440
Table 2 Severity and outcome according to body mass index
Variables a
SAPS II at admission Severity on day 1 LOD SAPS II APACHE II SOFA Severity on day 2 LOD SAPS II APACHE II SOFA ICU mortality, n (%) Hospital mortality, n (%) Standardized mortality ratio, 95% CIb
<18.5
18.5–24.9
25–29.9
>30
n=189
n=806
n=476
n=227
p-value
41 (29–59)
38 (28–51)
39 (29–53)
38.5 (28–54)
0.32
4 37 21 5
4 (2–6) 36.5 (27–46) 21 (16–27) 5 (3–8)
4 37 22 6
4 36 21 5
0.52 0.81 0.51 0.03
(2–5) (28–46) (17–26) (3–7)
4 (2–6) 37 (28–47) 20 (15–25) 4 (2–7) 54 (28.5) 83 (43.9) 1.16 (0.97– 1.35)
3 (2–6) 35 (27–46) 20 (14–25) 4 (2–8) 181 (22.4) 257 (31.8) 1.01 (0.91– 1.16)
(2–6) (29–45) (17–27) (4–8)
4 (2–5) 36 (28–46) 20 (14–26) 5 (3–8) 92 (19.3) 135 (29.9) 0.85 (0.73– 0.98)
(2-6) (28–49) (16–28) (3–8)
4 (2–6) 27 (21–35) 20 (15–27) 5 (3–8) 41 (18) 57 (25.1) 0.70 (0.6– 0.948)
0.17 0.55 0.14 0.01 0.03 0.0006
a Refers to the worse clinical and biological data recorded during the first 24 h of ICU stay. Day 1 refers to the data recorded between admission and 8 am on the next day. Day 2 refers to the data collected between 8 am to 8 am thereafter. b As compared to mortality predicted by the SAPS II score at admission. SAPS II Simplified acute physiologic score II. LOD Logistic organ dysfunction. APACHE II Acute physiology and chronic evaluation II. SOFA Sepsis-related organ failure assessment. 95% CI 95% confidence interval.
Table 3 Characteristics of the underlying diseases in the 1,698 patients according to body mass index
BMI groups
McCabe score No fatal disease Ultimately fatal disease Rapidly fatal disease Underlying diseases Pulmonary Cardiac Hepatic Renal Immunosuppression Neutropenia Steroids Chemotherapy HIV AIDS More than one underlying disease More than two underlying diseases Diabetes mellitus Complicated diabetes mellitus
<18.5
18.5–24.9
25–29.9
>30
p-value
n=189
n=806
n=476
n=227
74 (39.2) 76 (40.2) 36 (19)
403 (50) 304 (37.7) 96 (12)
263 (55.3) 166 (34.9) 45 (9.5)
117 (51.5) 96 (42.3) 14 (6.2)
45 15 16 3 40 4 22 12 2 13 81 14 9 1
132 77 49 18 145 22 57 55 8 27 374 50 95 28
89 64 23 7 50 8 33 26 1 4 198 38 91 19
58 29 14 1 15 4 13 7 0 2 104 15 60 15
0.0002
(23.8) (7.9) (8.5) (1.6) (21.2) ( 2.1) (11.6) (6.3) (1.0) (6.9) (42.3) (7.4) (4.8) (0.5)
(16.4) (9.6) (10.3) (2.2) (18) (2.7) (7.0) (6.8) (0.1) (3.4) (46.4) (6.2) (11.8) (3.5)
(18.7) (13.5) (4.5) (1.5) (10.5) (1.6) (6.9) (5.4) (0.8) (41.6) (8) (19.1) (4)
(25.6) (12.8) (6.2) (0.5) (6.7) (1.7) (5.7) (3.0) (0.9) (46) (6.6) (26.4) (6.6)
0.005 0.07 0.36 0.3 0.0001 0.6 0.1 0.2 0.2 0.0001 0.004 0.7 0.0001 0.01
Data are presented as N (%). Underlying diseases are defined using the Knaus classification [10].
1.11–2.39, P=0.01). When patients with AIDS or metastatic cancer were removed from the analysis, the leaner patients (BMI <18.5 kg/m2) were at increased risk of mortality (OR: 1.49, 95%CI: 0.98–2.26, P=0.06) and patients with overweight were associated with a better prognosis (OR:0.62, 95%CI: 0.41–0.93, P=0.02). Grades 1 or 2 overweight were not associated with an increase risk of death and even grade 2 overweight was
associated with a protective effect on mortality (OR: 0.60, 95% CI: 0.40–0.88, P=0.01). The standardized mortality ratio (SMR) was lower in the grades 1 and 2 overweight compared to the normal group (OR: 0.85, 95%CI: 0.73– 0.98 and OR: 0.77, 95%CI: 0.6–0.948, respectively) (Table 2). In the patients who were severely overweight (BMI>40 kg/m2), hospital mortality was 4/23 (17%), as compared to 530/1,645 (30%) in the rest of the study
441
Table 4 Admission diagnoses in the 1,698 patients of the cohort study
BMI groups
Multiple organ failure and shock Acute respiratory failure Exacerbation of COPD Acute renal failure Coma Trauma Monitoring
<18.5
18.5–24.9
25–29.9
>30
n=189
n=806
n=476
n=227
39 61 20 66 31 2 18
126 249 61 41 121 16 81
102 155 39 29 63 7 37
61 59 25 17 25 2 14
(20.6) (32) (10.5) (34) (16) (1) (9.5)
(15.6) (31) (7.6) (5) (15) (2) (10)
(21.4) (32.6) (8.2) (6) (13) (1.5) (7.7)
(26.8) (26) (11) (7.5) (11) (0.8) (6.2)
p-value
0.001 0.4 0.3 0.23 0.4 0.7 0.5
Data are presented as N (%). COPD Chronic obstructive pulmonary disease.
Table 5 Independent predictors of mortality by logistic regression analysis Independent variables
P value
OR (95%CI)
Age
<0.0001
1.022 (1.01–1.031) per additional year 1.036 (1.027–1.046) per point 1.12 (1.08–1.17) per point
SAPS at admission
a
<0.0001
SOFA on day 2b <0.0001 McCabe score No fatal disease Ultimately fatal disease <0.0001 Rapidly fatal disease <0.0001 Transfer from a ward 0.039 At least one underlying 0.004 disease (%) BMI 18.5–24.9 <18.5 0.01 25–29.9 0.053 >30 0.01 Admission diagnosis COPD 0.005
1 2.07 2.41 1.30 1.65
(1.55–2.77) (1.60–3.63) (1.01–1.66) (1.25–2.19)
1 1.63 0.75 0.60 0.51
(1.11–2.39) (0.56–1.004) (0.40–0.88) (0.32–0.82)
SAPS II Simplified acute physiologic score. SOFA Sepsis-related organ failure assessment. COPD Chronic obstructive pulmonary disease. a Refers to the worse clinical and biological data recorded during the first 24 hours of ICU stay. b Day 1 refers to the data recorded between admission and 8 am on the next day. Day 2 refers to the data collected between 8 am to 8 am thereafter. Variables tested in the multivariate model and found non-significant: cirrhosis, chronic cardiac disease (NYHA IV), metastatic cancer, acute cardiovascular failure or multiple organ failure at admission, acute respiratory failure at admission, need for mechanical ventilation within the first 48 h, vasoactive drugs within the first 48 h, and diabetes with or without complications, AIDS, neutropenia. Body mass index was introduced in the model at the last step. Final model: Hosmer Lemeshow c2 7.96, p=0.43, C=0.810.
population (P=0.02). Only 2 of the 23 severely overweight patients (BMI>40 kg/m2) died in the ICU.
Discussion This prospective study assessed the relationship between BMI and mortality in a large cohort of critically ill
patients in France. BMI was independently associated with higher mortality in the leaner patients (BMI <18.5 kg/m2) and was protective in the patients with severe overweight (BMI >30 kg/m2) but not in the patients with moderate overweight (BMI 25–30 kg/m2). BMI is a simple and reproducible method which added prognostic information to the conventional scoring systems in this large database of ICU patients. Few studies of BMI in critically ill patients are available, and most were done in the United States where the prevalence of obesity is 3 times higher than in France and 1.5 times higher than in the United Kingdom [4]. In a retrospective study of 184 patients with blunt trauma divided into 3 groups according to BMI (<27, 27–31, >31 kg/m2), Choban et al. [18] found that mortality was 42.1% in the group of patients with a BMI>31 kg/m2, an 8-fold increase over the 5% mortality rate in the normal group (BMI<27 kg/m2). Respiratory problems explained most of the differences. No difference in mortality was found between the normal group (BMI<27 kg/m2) and the moderately overweight group (BMI 27–31 kg/m2, mortality 8%). Recently, El-Solh et al. [9] reported a retrospective study in patients admitted to the ICUs of two university hospitals in the US. Mortality was nearly twice as high in the morbidly obese group (30%) than in the non-obese group (17%) (P=0.019). This result does not contradict our finding that mortality was not increased in the high BMI groups. In our study, only 23 of our patients were morbidly obese as defined by El-Sohl et al. (BMI40 kg/m2). The mean BMI was 35.4€5 in our patients with a BMI>30 and 45.7€5.8 in those with a BMI>40 kg/m2, as compared to a mean of 51€25.9 in the patients studied by El-Sohl et al. [9]. Our results were similar to the retrospective analysis of a large database of 63,646 ICU patients showing that mortality was increased in underweight patients but not in overweight, obese or severely obese patients [10]. Care must be taken in interpreting our results. Body weight in ICU patients is a continuous variable, modified by fluid therapy. We chose to use body weight at admission to calculate BMI and this may be significantly different to baseline values. For the leaner patients, we do not know whether these patients had lost weight in the last few months or were constitu-
442
tionally thin. Nevertheless, this subgroup had a high rate of immunosuppression, most notably related to AIDS, suggesting that the low BMI may have been a marker for the underlying disease. However, the chronic-related illness did not explain all the effects on mortality as shown by the results after exclusion of the AIDS and metastatic cancer patients. We did not find increased mortality in moderate overweight and, even a negative effect of mortality in the severe overweight patients. The proportion of surgery patients is higher in the BMI>30 kg/ m2 group. They were probably admitted to the ICU earlier in the course of their disease, in particular in the postoperative period of scheduled surgery, with increased likelihood of being managed in ICU without having a substantially increased risk of death. In keeping with other studies [9], obesity was associated with a longer ICU stay in our population. BMI may add to the prognostic information supplied by conventional variables. BMI is not used in the severity scores designed for ICU patients, such as the SAPS II or APACHE II and III. El Solh et al. [9] found that morbid obesity was associated with higher mortality as compared to non-obese patients with similar APACHE scores. Mortality prediction models perform best in populations identical to those used for their development [19]. Customized models for well-defined patient categories, such as patients with sepsis, have been developed to improve prediction [20]. Predictive performance of available scores was poor in patients with longer hospital stays [21] and in those with differences in demographic characteristics. In our cohort, severity at admission within the first 48 h, as assessed by several scores, was similar in the four BMI groups. The differences between observed mortality and mortality predicted based on these scores indicate a pressing need for improving the accuracy of predictive tools by identifying additional risk factors for mortality. Our data suggest that BMI may be a useful component of future scoring systems.
Conclusions This prospective study indicates that a BMI<18.5 kg/m2 in ICU patients is a predictor of mortality and a high BMI appeared to be associated with favorable outcome, independent from conventional predictors. Severity scores, even those determined repeatedly during the first 2 ICU days, failed to predict mortality accurately. Thus, studies are needed in order to determine whether incorporating BMI into severity scores would improve the prediction of mortality. Furthermore, our data invite an investigation of the mechanisms that underlie the association between BMI and mortality in critically ill patients. Acknowledgments We thank A. Wolfe, MD, for help in preparing this manuscript.
Appendix List of members of the OUTCOMEREA study group, including all authors of the present manuscript (in italics). Scientific committee Timsit, Jean-Franois MD, staff physician, ICU, Hpital Bichat, Paris, France. Troch, Gilles MD, staff physician, Hpital A. Bclre, Clamart, France. Moine, Pierre MD, PhD, staff physician, DAR, Hpital Lariboisire, Paris, France. De Lassence, Arnaud MD, staff physician, ICU, Hpital Louis Mourier, Colombes, France. Azoulay, Elie MD, staff physician, ICU, Hpital Saint Louis, Paris, France. Cohen, Yves MD, Professor, ICU, Hpital Avicenne, Bobigny, France. Garrouste-Orgeas, Mat MD, staff physician, ICU, Hpital Saint Joseph, Paris, France. Fosse, Jean-Philippe MD, staff physician, ICU, Hpital Avicenne, Bobigny, France. Soufir, Lilia MD, staff physician, ICU, Hpital saint Joseph, Paris, France. Zahar, Jean-Ralph MD, attending physician, Microbiology department, Hpital Necker, Paris, France. Adrie, Christophe MD, staff physician, ICU, Hpital Delafontaine, Saint Denis, France. Carlet, Jean MD, staff physician, ICU, Hpital Saint Joseph, Paris, France. L’Hriteau, Franois MD, attending physician, ICU, Hpital Bichat, Paris, France. Biostatistical and informatics expertise Chevret, Sylvie MD PhD, Professor, Medical Computer Sciences and Biostatistics Department, Hpital Saint Louis, Paris, France. Alberti, Corinne MD, staff physician, Medical Computer Sciences and Biostatistic Department, Hpital Saint Louis, Paris, France. Lecorre, Frederik, informatician, Supelec, France. Nakache, Didier, informatician, Conservatoire National des Arts et Mtiers (CNAM), Paris, France. Investigators of the OutcomeRea database Bornstain, Caroline MD, staff physician, ICU, Hpital Europen Georges Pompidou, Paris, France. Thuong, Marie MD, staff physician, ICU, Hpital Delafontaine, Saint Denis, France.
443
Costa de Beauregard, Marie-Alliette MD, staff physician, Nephrology, Hpital Tenon, Paris, France. Colin, Jean-Pierre MD, staff physician, ICU, Hpital de Dourdan, Dourdan, France. Le Miere, Eric MD, attending physician, ICU, Hpital Louis Mourier, Colombes, France. Caubel, Antoine MD, attending physician, ICU, Hpital Saint Joseph, Paris, France. Marie, William MD, attending physician, ICU, Hpital Saint Joseph, Paris, France. Cheval, Christine MD, staff physician, SICU, Hpital Saint Joseph, Paris, France. Vantalon, Eric MD, staff physician, SICU, Hpital Saint Joseph, Paris, France. Clec’h, Christophe MD, staff physician, ICU, Hpital Avicenne, Bobigny, France. Vincent, Franois MD, staff physician, Nephrology, Hpital Tenon, Paris, France. Salah, Amar MD, staff physician, ICU, Hpital Louis Mourier, Colombes, France. Montesino, Laurent MD, attending physician, ICU, Hpital Bichat, Paris, France.
Pign, Etienne MD, staff physician, ICU, Hpital Louis Mourier, Colombes, France. Boyer, Alexandre MD, staff physician, ICU, Hpital Pellegrin, Bordeaux, France. Tuil, Olivier MD, staff physician, ICU, Hpital Avicenne, Bobigny, France. Mourvillier, Bruno MD, Staff physician, ICU, Hpital d’Aulnay sous bois, France. Jamali, Samir MD, staff physician, ICU, Hpital de Dourdan, Dourdan, France. Moreau, Delphine MD, staff physician, ICU, Hpital Saint Louis, Paris, France. Thiery, Guillaume MD, staff physician, ICU, Hpital Saint Louis, Paris, France. Duguet, Alexandre MD, attending physician, Hpital Piti-Salptriere, Paris, France. Laplace, Christian MD, attending physician, ICU, Hpital Kremlin-BicÞtre, BicÞtre, France. Lazard, Thierry MD, staff physician, ICU, Hpital de la Croix St Simon, Paris, France.
References 1. Diehr P, Bild DE, Harris TB, Duxbury A, Siscovick D, Rossi M (1998) Bodymass index and mortality in nonsmoking older adults: the cardiovascular health study. Am J Public Health 88:623–629 2. Harris T, Cook EF, Garrison R, Higgins M, Kannel W, Goldman L (1998) Body mass index and mortality among non smoking older persons. The Framingham Heart Study. JAMA 259:1520– 1524 3. Cornani-Huntley JC, Harris TB, Everett DF, Albanes D, Micozzi MS, Miles TP (1991) An overview of body weight of older persons, including the impact on mortality. The National Health and Nutrition Examination Survey. An epidemiological follow up study. J Clin Epidemiol 44:743–753 4. VanItallie TB (1996) Prevalence of obesity. Endocrinol Metab Clin North Am 25:887–905 5. Must A, Spadano J, Coakley ZH, Field AE, Colditz G, Dietz WH (1999) The disease burden associated with overweight and obesity. JAMA 282:1523– 1529 6. Wannamethee SG, Shaper AG, Walker M (2001) Weight change, body weight and mortality: the impact of smoking and ill health. Int J Epidemiol 30:777– 786
7. Calle EE, Thun MJ, Petrelli JM, Rodriguez C, Heath CW (1999) Body weight and mortality in a prospective cohort of U.S. adults. N Engl J Med 341:1097–1105 8. Grabowski DC, Ellis JE (2001) High body mass index does not predict mortality in older people: analysis of the longitudinal study of aging. J Am Geriatr Soc 49:968–979 9. El-Solh A, Sikka P, Bozkanat E, Jaafar W, Davies J (2001) Morbid obesity in the medical ICU. Chest 120:1989–1997 10. Tremblay A, Bandi V (2003) Impact of body mass index on outcomes following critical care. Chest 123:1202–1207 11. McCabe WR, Jackson GG (1962) Gram negative bacteremia: etiology and ecology. Arch Intern Med 110:847–855 12. Knaus WA, Drapper EA, Wagner D, Zimmerman JE (1985) Apache II: a severity of disease classification. Crit Care Med 13:818–829 13. Le Gall JR, Lemeshow S, Saulnier F (1993) A new simplified physiologic score (SAPS II) based on a European/ North American multicenter study. JAMA 270:2957–2963 14. Le Gall JR, Klar J, Lemeshow S, Saulnier F, Alberti C, Artigas A (1996) The logistic organ dysfunction system. A new way to assess organ dysfunction in the intensive care unit. ICU scoring group. JAMA 276:802–810
15. Vincent JL, Moreno R, Takala J et al. (1996) The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ/dysfunction failure. On behalf of the Working Group on Sepsisrelated Problems of the European Society of Intensive Care Medicine. Intensive Care Med 22:707–710 16. Socit de Ranimation et de Langue Franaise: utilisation de l’indice de gravit simplifie et du systme Omega: mise jour 1986 (1986) Ranimation Soins Intensifs Mdecine d’Urgences 2:219–221 17. World Health Organisation (1995) Physical status: the use an interpretation of anthropometry: report of a WHO expert committee. WHO Tech Rep Ser 854, Geneva,pp 1–452 18. Smith Choban P, Weireter LJ, Maynes C (1991) Obesity and increased mortality in blunt trauma. J Trauma 31:1253–1257 19. Lemeshow S, Le Gall JR (1994) Modeling the severity of illness of ICU patients. A systems update. JAMA 272:1049–1055 20. Le Gall JR, Lemeshow S, Leleu G et al. (1995) Customized probability models for early severe sepsis in adult intensive care patients. JAMA 273:644–650 21. Hadorn DC, Draper D, Rogers WH, Keeler EB, Brook RH (1992) Cross validation performance of mortality prediction models. Stat Med 28:475– 489