Increased Intestinal Permeability Is Associated with the ... - ATS Journals

171 downloads 7625 Views 143KB Size Report
course. These patients, as a group, had significantly worse IP at admission than did a non-MODS .... dialysis, plasmapheresis, or other physiologic support requiring extra- ... ars was made by one technician blinded to the clinical condition and.
Increased Intestinal Permeability Is Associated with the Development of Multiple Organ Dysfunction Syndrome in Critically Ill ICU Patients CHRISTOPHER J. DOIG, LLOYD R. SUTHERLAND, J. DEAN SANDHAM, GORDON H. FICK, MARJA VERHOEF, and JON B. MEDDINGS Departments of Medicine, and Community Health Science, University of Calgary, Calgary, Alberta, Canada

We conducted a prospective, observational cohort study designed to compare intestinal permeability (IP) and development of multiple organ dysfunction syndrome (MODS) in a subset of critically ill patients in an intensive care unit (ICU). All patients with an expected ICU stay of 72 h or more were entered into the study, and IP was determined on a daily basis whenever possible from the urinary fractional excretion of orally administered lactulose and mannitol (LMR). Forty-seven consecutive patients were studied, and 28 developed MODS either at the time of admission or during their ICU course. These patients, as a group, had significantly worse IP at admission than did a non-MODS cohort (LnLMR: 22.10 6 1.10 versus 23.26 6 0.83). Those patients who developed MODS following admission also had a significantly greater admission IP than did the non-MODS group (22.51 6 0.85). Differences in IP between cohorts could not be explained by differences in the incidence of systemic inflammatory response syndrome (SIRS)/sepsis or shock. With multivariate regression analysis, the only parameter present on admission that was predictive of subsequent MODS was IP. Differences in IP and the severity of organ dysfunction were also present (MODS severity mild: 23.01 6 0.72; moderate: 21.97 6 0.69; and severe: 21.12 6 0.96). Patients who developed MODS had a persistently abnormal IP during their ICU stay, and a significantly delayed improvement in their IP compared with the non-MODS cohort. We conclude that the development of MODS is associated with an abnormal and severe derangement of IP that is detectable prior to the onset of the syndrome. This observation lends credence to the premise that gastrointestinal (GI) dysfunction may be causally associated with the development of MODS in the critically ill patient. Doig CJ, Sutherland LR, Sandham JD, Fick GH, Verhoef M, Meddings JB. Increased intestinal permeability is associated with the development of multiple organ dysfunction syndrome in critically ill ICU patients. AM J RESPIR CRIT CARE MED 1998;158:444–451.

Evolution of critical care medicine has resulted in improved management of single-organ-system failure and increased survival of intensive care unit (ICU) patients. However, as the critically ill patient survives longer, a new syndrome has emerged as the major cause of death in the ICU: multiple organ dysfunction syndrome (MODS) (1–8). MODS is characterized by the progressive deterioration and subsequent failure of the body’s basic physiologic systems, and has been described as death in slow motion. Whether this is a separate disease or a final common pathway to death remains uncertain. The epidemiology, basic risk factors, and underlying physiologic processes in MODS remain poorly understood (3). Patients in the ICU are heterogeneous and are admitted with varied and often unrelated conditions, yet MODS is the final common path(Received in original form October 28, 1997 and in revised form March 5, 1998) Supported in part by the Alberta Heritage Foundation for Medical Research, the Medical Research Council of Canada, and by Searle Canada Inc. Correspondence and requests for reprints should be addressed to Dr. Jon Meddings, 1705 Health Sciences Center, University of Calgary, 3330 Hospital Drive NW, Calgary, AB, T2N 4N1 Canada. E-mail: [email protected] Am J Respir Crit Care Med Vol 158. pp 444–451, 1998 Internet address: www.atsjournals.org

way to death in a large proportion of these patients. This final common pathway is believed to be caused by a hyperinflammatory response with resultant activation and/or release of common endogenous cellular mediators or enzymatic cascades (2, 5, 7). It is hypothesized that the development of MODS is due to a loss of autoregulation of the normal inflammatory response. What initiates the cascade of events leading to MODS is a matter of debate; however, it has been suggested that the gastrointestinal (GI) tract may play a central role. Under normal conditions, the intestinal epithelial barrier acts as a selective route of entry, allowing movement of necessary molecules through the epithelium but at the same time preventing the entry of potentially pathogenic organisms or their products. This barrier function of the intestinal epithelium can be measured with the use of monosaccharide and disaccharide probes (9–12). In the theoretical framework of the uncontrolled inflammatory response as the cause of MODS, it has been postulated that ongoing activation of the inflammatory response is due to translocation of bacteria, or their products, through the epithelial-cell barrier of the GI tract (2, 5, 7, 13). Abnormal intestinal permeability (IP) has been identified as a marker of clinical disease in a number of chronic and

445

Doig, Sutherland, Sandham, et al.: MODS in Critically Ill ICU Patients

acute inflammatory intestinal disorders, including inflammatory bowel disease, gluten enteropathy, and tropical sprue (14–19). In animal models that mimic pathophysiologic processes present in many critically ill patients, mucosal ischemia, the extent of pathologic injury to the intestinal mucosa, the severity of bacterial translocation, and the severity of abnormal IP have been correlated (20–27). Clinical studies have been performed to assess IP in subsets of critically ill patients with severe burns, patients requiring cardiopulmonary bypass, and patients who are critically ill during the postoperative period (28–36). These studies have shown a trend toward increased intestinal permeability, but have demonstrated inconsistent associations between IP and outcome, IP and severity of illness, and IP and the development of septic complications. Preliminary work by our group suggested an association of IP with MODS. The objectives of the current study were twofold. First, we wished to determine whether patients with MODS or who were developing the syndrome had increased IP on admission to the ICU. Second, we asked whether patients who developed MODS more than 24 h after admission had greater intestinal permeability on admission than did patients who never developed this syndrome.

METHODS Setting The Foothills Provincial General Hospital is an academic, regional tertiary-care hospital serving an estimated 1.4 million adult citizens of Calgary, Alberta; southeastern British Columbia; and southwestern Saskatchewan. The hospital has approximately 700 acute-care beds. Tertiary critical-care services include an 18-bed multisystem ICU admitting critically ill patients from medical, surgical, and trauma disciplines. The ICU is a “closed” unit staffed by physicians who are specialized in critical care.

Study Design The study was a prospective observational cohort study.

Study Population The study population consisted of consecutive patients admitted to the Foothills Hospital multisystem ICU who were identified a priori as having an expected duration of stay exceeding 72 h. These patients represent a subset of all of the hospital’s ICU patients, and the criterion for an expected duration of stay exceeding 72 h identified patients who were expected to require prolonged physiologic support, and who were therefore theoretically at increased risk for MODS. The expectation of a stay exceeding 72 h eliminated patients who were admitted to the ICU for monitoring, nursing care, or short-term temporary physiologic support.

Study Sample The study sample was drawn from the study population, excluding patients with anuric renal failure; an expected intervention involving dialysis, plasmapheresis, or other physiologic support requiring extracorporeal blood removal; known causes of increased bowel permeability (e.g., known primary small-bowel disease, peptic ulcer disease, mesenteric ischemia, or inflammatory bowel disease); documented use of nonsteroidal antiinflammatory drugs (NSAIDs) within the prior 30 d; expected use of mannitol or lactulose as part of therapy; and entry into other clinical studies. The study was approved by the local institutional review board with a priori patient or appropriate proxy consent obtained prior to participants’ entry into the study.

Analytic Technique Intestinal permeability was measured in a standardized fashion (19). Five grams of lactulose with 2 g of mannitol in a suspension of 20 ml of

distilled water was administered daily when possible during the ICU stay. Twenty milliliters of tap water was given to rinse the feeding tube after administration of the sugar solution. Feeding with isosmotic enteral preparations was temporarily interrupted during administration of the sugar solution, but was immediately resumed following the rinse solution. The excreted portion of each sugar marker was collected for 6 h in urine via a standard urinary catheter collecting system to which 80 mg of gentamicin had been added. All urine collected was placed in a collection bottle containing 5 ml of 10% thymol. The collection bottle was then refrigerated at 48 C, and all samples were processed within 24 h. Measurement of the urinary concentration of sugars was made by one technician blinded to the clinical condition and identity of the patient. Ten-milliliter samples of urine were decanted for analysis. The samples were deionized by adding 1 g of a 1:1.5 (wt/wt) mixture of Amberlite 1R-120 and IRA-400 resin (BDH Chemicals, Toronto, Canada). The supernatant was filtered through a 45-m Millipore filter (Millipore, Inc., Bedford, MA). Cellobiose was added as an internal standard. Urine samples were separated on a Dionex Carbopac MA-1 anion-exchange column (Dionex, ON, Canada) in a Dionex DX500 high-pressure liquid chromatography (HPLC) system using 520 mM NaOH as the isocratic mobile phase. Peak identification was done with pulsed amperometric electrochemical detection on a gold electrode. Quantitation was done with known standards at multiple concentrations, with linear interpolations between peaks. Samples were diluted after addition of the internal standard. Lactulose samples were initially analyzed at a 1:3 dilution and mannitol at a 1:20 dilution. If these dilutions were not satisfactory for proper analysis, adjustments to the concentrations were made so that the mannnitol and lactulose concentrations fell within the range of the cellobiose internal standard. The fractional excretion of lactulose and mannitol was calculated from the urinary concentrations of these sugars; the lactulose/ mannitol ratio (LMR) is reported. LMR results were converted to their natural log (ln) values to normalize the distribution for analysis. The abbreviation used for reporting purposes is ln(LMR). The upper limit of normal for the LMR in our laboratory has been defined as the mean 1 3 SD of several hundred normal volunteers, and is 0.030. Therefore, an ln(LMR) of 23.50 represents the upper limit of normal for this study.

Outcome Measures The primary outcome measure was the presence of multiple organ dysfunction. The presence of MODS was based on the published criteria of Knaus and colleagues (3). Because this is a dichotomous classification for MODS, the severity of organ dysfunction was also classified, on the basis of the criteria of Marshall and coworkers (37). The determination of organ failure and evaluation of the severity of organ dysfunction was made by one individual blinded to the results of the calculation of LMR (C.D.). Because a subset of patients entering the ICU developed MODS early in their course, prior to having well-documented permeability studies, we elected to stratify patients who developed MODS into two groups. The first group we denoted as a primary MODS group (n 5 10), and comprised patients who developed MODS less than 24 h after the initial permeability study that was done within 24 h following arrival in the ICU. We felt that it was impossible to exclude the possibility that some of these patients either had or were developing MODS upon arrival in the unit. All patients with secondary MODS (n 5 17) developed the syndrome more than 24 h after the initial permeability measurement. Systemic inflammatory response syndrome (SIRS)/sepsis, and shock were considered a priori as potential effect modifiers or confounders of any association between abnormal permeability and the development of MODS. SIRS was defined according to the criteria of the Society of Critical Care Medicine/American College of Chest Physicians (SCCM/ACCP) Consensus Conference (38). The criteria for the presence of shock were a change in blood pressure (BP) consisting of at least one of the following: a mean arterial pressure (MAP) , 60 mm Hg, a systolic BP , 90 mm Hg, an absolute decrease in systolic BP of 40 mm Hg, or the use of adrenergic agents to maintain MAP . 60; with at least one of the following other signs: oliguria (0.3 ml/kg/h for two consecutive hours), a change in level of consciousness, serum

446

AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE

lactate . 2.5 mmol/L, or an absolute mixed venous oxygen saturation , 60%. Clinical data were collected for all patients, and included primary admitting diagnosis, acute physiology and chronic health evaluation II (APACHE II) score, daily acute physiology score (APS), and therapeutic intervention severity (TISS) score, an inventory of the use of therapy, and technique of administration of nutrition. The determinants for the calculation of APACHE II score, APS, and TISS were in accordance with previously published criteria (39, 40).

Data Analysis All data were analyzed with the statistical packages SPSS 6.0 (Windows) (SPSS, Chicago, IL) and Splus 3.1 (Unix; Statistical Science Inc., Seattle, WA). The cohorts for comparison consisted of patients who did and did not develop MODS, with the latter subdivided into primary and secondary MODS, as defined earlier. Comparisons between cohorts of continuous variables having a normal distribution were done with Student’s t test or with analysis of variance (ANOVA). Comparisons of dichotomous variables were done with Fisher’s exact test. Cohorts were compared for daily changes in permeability through the use of a linear mixed-effects (LME) model. The LME is an accepted technique in the analysis of repeated measures. The LME technique allows comparisons between the means of cohorts, comparisons of unit changes in permeability per unit change per day, inclusion of effects of daily changes in the global severity of physiologic dysfunction, and accounting for different numbers of permeability results between patients. The LME model demonstrating a daily difference in permeability between MODS cohorts was: ln p ij = ( β 00 + β 01 MODS i + b 10 ) + ( β 10 + β 11 MODS i + b 11 ) day + ε ij

(1)

where lnpij is the natural log of permeability on the ith individual on Day j; MODSi is an indicator variable, with one group assuming a value of 0 for the ith individual and the second group assuming a value of 1 for the ith individual, and b00 and b10 are the population average intercept in the group assigned a value of 0 as the indicator in MODSi. If b11 is not 0, then there is evidence of a difference in the slope (or daily change in permeability) between groups. TABLE 1 DEMOGRAPHICS OF STUDY PARTICIPANTS BY COHORT

Age, mean 6 SD Gender, M/F Reason for ICU admission Respiratory failure, n 5 21 Trauma, n 5 9 Shock, n 5 7 Postoperative, n 5 7 Coma, n 5 2 Post-cardiac arrest, n 5 1 APACHE II, mean 6 SD ICU LOS (median, minimum, maximum) Enteral nutrition, days to start (median, minimum, maximum) Percentage of days on enteral nutrition (median, minimum, maximum) Survival ICU Hospital

Non-MODS (n 5 19)

MODS (n 5 28)

Significance

57.9 6 17.34 12/7

56.5 6 20.5 16/12

p 5 NS p 5 NS

8 5 2 4 0 0 17.9 6 8.7

13 4 5 3 2 1 22.9 6 7.2

0.04

5 (3, 21)

12 (3, 57)

VOL 158

1998

From preliminary work, we estimated a difference in permeability of 25% between our two cohorts, and that 60% of our selected patients would develop MODS. A sample size of 50 was estimated to demonstrate a difference between cohorts, based on a two-tailed a value of 0.05 and a power of 80%.

RESULTS Among screened patients with an expected stay exceeding 3 d, 62 consecutive patients were identified over a 7-mo period. Fifteen of these 62 patients were excluded; six because of primary GI pathology, three because of the use of intravenous mannitol, and six because of enrollment in other clinical studies. Comparisons of the characteristics of the patients with MODS and those who did not develop MODS are presented in Table 1 (a description of individual patients), and Table 2 (a description by study cohort). Of note was the lack of a difference in age or gender distributions between cohorts. As would be expected, the MODS cohort had an APACHE II score that was greater than the APACHE II score in the non-MODS cohort. The survival of the MODS cohort was significantly less than that of the non-MODS cohort. No difference was found in the time to starting enteral nutrition in the two cohorts. One patient did not receive a permeability test within 24 h of admission, and the analysis based on admission permeability therefore, includes only the remaining 46 patients. IP (lnLMR) on admission was 22.10 6 1.10 (mean 6 SD) in the cohort of patients with MODS, as compared with 23.26 6 0.83 in the cohort of patients without MODS (95% confidence interval [CI] of the difference: 20.59 to 21.73, p , 0.001). Therefore, the cohort with MODS had greater IP on admission than did the cohort that did not develop MODS. A significantly increased IP on admission was also observed in those patients who developed MODS more than 24 h after the initial permeability test (secondary MODS) (22.51 6 0.85, p , 0.01). No intracohort effect was present to explain SIRS/sepsis or shock as a possible alternative reason for the observed association between development of MODS and IP (Table 2, Figure 1). We further pursued the interrelationships between the development of MODS, admission permeability, shock, and sepsis, using a multivariate logistic-regression model. In the patients who developed MODS more than 24 h after completion of their first IP test, the criteria for MODS were met at a median of 6 d (interquartile range: Days 5 to 9) after ICU admission. A difference in IP was found between the MODS subcohorts. The ln(LMR) for the subcohort that had MODS within the first 24 h was 21.42 6 1.11, compared with an ln(LMR) of 22.51 6 0.88 for the subcohort that developed MODS more than 24 h after the first IP measurement (p , 0.03). These data are illustrated in Figure 2. A comparison of admission IP

p 5 NS

TABLE 2 2 (1, 4)

2 (1, 5)

p 5 NS

60 (0, 80)

66 (0, 78)

p 5 NS

18/19 15/19

16/28 14/28

p 5 0.005 0.04

Definition of abbreviations: APACHE II 5 acute physiology and chronic health evaluation II; ICU 5 intensive care unit; MODS 5 multiple organ dysfunction syndrome; LOS 5 length of stay. The cohorts demonstrated no difference in gender or age distribution, frequency, or timing of enteral nutrition, or in the etiology of their reason for ICU admission. The cohort that developed MODS demonstrated a worse admission APACHE II score, had a longer ICU stay, and had an overall lower survival. All of these features are consistent with our current understanding of MODS.

MODS AND EFFECT OF SIRS/SEPSIS OR SHOCK ON IP, LN(LMR)

All patients Patients without SIRS/sepsis or shock SIRS/sepsis Shock

All MODS (mean 6 SD)

Secondary MODS (mean 6 SD)

Non-MODS (mean 6 SD)

22.10 6 1.10

22.51 6 0.85

23.26 6 0.83

21.89 6 1.13 22.13 6 1.11 22.10 6 1.23

21.78 22.56 6 0.86 22.71 6 0.84

23.09 6 1.14 23.31 6 0.77 23.15 6 0.86

Definition of abbreviations: ANOVA 5 analysis of variance; MODS 5 multiple organ dysfunction syndrome; SIRS 5 systemic inflammatory response syndrome. ANOVA for effects of MODS and covariates SIRS/sepsis and shock. F 5 0.009 for covariates within cohorts (p 5 NS). F 5 33.29 for main effects between cohorts (p , 0.001).

Doig, Sutherland, Sandham, et al.: MODS in Critically Ill ICU Patients

447

Figure 1. Intestinal permeability in MODS and non-MODS cohorts. Natural logarithm of the ratio of the fractional urinary excretion of lactulose and mannitol (ln[LMR]: a measure of intestinal permeability). SIRS 5 systemic inflammatory response; MODS 5 multiple organ dysfunction syndrome. The figure shows the box plots for the study populations by outcome, stratified by the presence or absence of SIRS/sepsis, and shock. A significant difference in IP at admission was present between the MODS group and the entire MODS cohort (all MODS), mean 6 SD of ln(LMR) 5 22.10 6 0.89 versus 23.26 6 0.83. Furthermore, the same significant difference was apparent in those patients who developed MODS more than 24 h after the first IP measurement (secondary MODS). As shown, these significant differences were not secondary to the presence of SIRS/sepsis or shock. However, only a single patient with secondary MODS had neither SIRS/sepsis or shock. A more detailed statistical analysis of this is presented in the text.

in the group that did not develop MODS and the group that developed MODS more than 24 h after the first IP measurement showed a significant difference (ln[LMR] 5 23.26 6 0.83 versus 22.51 6 0.88, p 5 0.006). Multivariate modeling using logistic regression was done to assess the relationship between the development of secondary MODS during ICU stay (MODS developing . 24 h after first permeability test) and independent variables of IP on admission, presence of shock or sepsis/SIRS at admission, admission APACHE II score, and age. Sequential elimination of variables was done with the likelihood ratio method. Standard diagnostics were performed in all analyses, with no evidence of violation of mathematical assumptions. Using this analysis, admission IP was the only variable statistically associated with the development of MODS. The final relationship was defined by the following equation: log (MODS) 5 2.89 1 1.04(lnLMR) (p 5 0.02). Increased IP on admission correlated with both the development and severity of MODS. Figure 3 illustrates the corre-

Figure 2. Intestinal permeability and MODS category. Natural logarithm of the ratio of the fractional urinary excretion of lactulose and mannitol (ln[LMR]: a measure of intestinal permeability). MODS 5 multiple organ dysfunction syndrome. The figure presents box plots of the ln(LMR) of the non-MODS cohort and the

MODS cohort, stratified as primary or secondary MODS. MODS developing within 24 h after the first permeability measurement was arbitrarily defined as primary MODS, and MODS developing later than this as secondary MODS. *p , 0.05 versus secondary MODS; **p , 0.05 versus the primary MODS group; ***p , 0.05 versus the non-MODS group.

448

AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE

VOL 158

1998

Figure 3. IP and the Marshall MODS score. Natural logarithm of the ratio of the fractional urinary excretion of lactulose and mannitol (ln[LMR]: a measure of intestinal permeability). MODS 5 multiple organ dysfunction syndrome. The scatter plot demonstrates the relationship between IP and severity of MODS (both primary and secondary MODS). Data points represent individual values for each patient. There is an association between admission IP and subsequent worst daily Marshall score for MODS for each of 46 patients (Patient 31 did not have a test available within 24 h of admission).

lation between the worst Marshall MODS score during the ICU stay and admission IP. A significant correlation was apparent (r2 5 0.37), with the 95% CIs of the slope shown by the dotted curves. To further assess this, the Marshall MODS score was arbitrarily characterized into four levels, based on the cumulative score in at least two dysfunctional organ systems: mild (5 to 8), moderate (9 to 12), and severe (. 12). No patient characterized as having MODS by the Knaus criterion had a MODS score of less than 5. Permeability in patients with severe or moderate organ dysfunction (ln[LMR] 5 21.12 6 0.96 and 21.97 6 0.69, respectively) was significantly greater than that observed in patients with either mild organ dysfunction or no evidence of MODS (ln[LMR] 5 23.01 6 0.72 and 23.26 6 0.83, respectively). These data are illustrated in Figure 4A. Figure 4B illustrates the same relationship, restricted to patients with secondary MODS, and the same relationships can be observed, although the sample size is smaller. The LME method was used to analyze changes in permeability over the course of the ICU stay of individuals in each cohort: the change in IP (LnLMR), was represented as a function of their ICU stay. By definition, the expression used in the LME method describes a linear relationship whose slope represents the daily change in permeability for each individual. A difference in the daily change in permeability between defined groups is available for testing. The daily change in permeability between individuals who developed MODS at least 24 h after the first IP measurement (secondary MODS), was compared with that in the non-MODS cohort. The initial LME expression included patient-specific variables such as

Figure 4. IP and severity of MODS. Natural logarithm of the ratio of the fractional urinary excretion of lactulose and mannitol (ln[LMR]: a measure of intestinal permeability). MODS 5 multiple organ dysfunction syndrome. Horizontal lines represent mean permeability for each group. Data points represent individual values for each patient. The categories of MODS severity were based on the Marshall score, and were arbitrarily defined as mild: 5 to 8, moderate: 9 to 12, and severe: . 12. (A) Results for all cases of MODS. (B) Analysis restricted to patients with secondary MODS (clearly developed within the ICU). The IP on admission for patients in the moderate and severe category was significantly different than the IP for patients in the mild or no MODS category (p , 0.05) for all comparisons except severe MODS in (B). Too few patients were available in this group for a reliable comparison.

Doig, Sutherland, Sandham, et al.: MODS in Critically Ill ICU Patients

Figure 5. Daily IP measurements. Natural logarithm of the ratio of the fractional urinary excretion of lactulose and mannitol (ln[LMR]: a measure of intestinal permeability). MODS 5 multiple organ dysfunction syndrome. Dashed line and open circles represent patients who developed MODS more than 24 h after ICU admission (i.e., secondary MODS) (median: Day 6, interquartile range: Days 5 and 9). Continuous line and closed circles represent patients who did not develop MODS during their ICU stay. The shaded rectangle represents the normal range of permeability. Data points represent mean 6 SE. Significant differences were present in daily permeability measurements between cohorts. Improvement in permeability was at a significantly slower rate for individuals who developed MODS (p 5 0.03).

age, physiologic variables such as the APS from the APACHE II score, and a maximum of two-level interactions. Variables were removed from the expression by sequential elimination, using the likelihood-ratio method. The final relationship showed that the cohort of individuals who developed MODS following admission had a significantly slower daily improvement in IP than did the non-MODS cohort (p , 0.03, Figure 5). Inclusion of the APS did not show evidence of improving the discrimination of the relationship (log-likelihood of expression with APS 5 2228.86, versus 2229.33, without APS, p 5 NS). Therefore, the difference in the daily rate of change in IP could not be explained by a difference in daily changes in the global severity of physiologic derangement.

DISCUSSION MODS is a major cause of morbidity and mortality in ICU patients. Since its original suggestion by Fine in the 1960s, the clinical hypothesis for the source of secondary organ dysfunction and development of MODS has been an abnormality in the GI tract that precedes and may be a nidus for ongoing dysfunction in remote organs (4, 13). The basis for explanation of the development of MODS involves hypoperfusion of the gut in shock, the importance of enteral nutrition in preventing mucosal atrophy, the overgrowth of pathogenic nosocomial bacteria in the proximal GI tract, and the occurrence of increased circulating inflammatory mediators (7, 13, 41). If the

449

clinical hypothesis for the development of MODS is correct, the sine qua non step in its development must be the translocation of bacteria, or their products, and such translocation should occur only when the intestinal epithelial barrier is abnormal. Animal studies have documented bacterial translocation across the GI epithelium following intestinal hypoperfusion (19, 20, 22). In these studies, the quantity of translocating bacteria correlated with the severity of mucosal damage and the duration of hypoperfusion (22). However, despite the plausibility of the “GI tract” hypothesis for development of MODS, studies of heterogeneous ICU-patient samples involving the use of selective decontamination of the digestive tract or the use of antibodies to inflammatory mediators have not consistently shown an effect on survival or the outcome of organ dysfunction (42). Demonstrating alterations in the intrinsic epithelial barrier in patients who develop MODS would seem to be the necessary first step in considering the role of the GI tract in the pathophysiology of MODS. Additionally, developing a technique to determine which patients are at risk for MODS may facilitate the selection of patients at high risk for inclusion in clinical studies of MODS therapy. The primary hypothesis for this study was that the subset of critically ill ICU patients either with or developing MODS would have significantly greater IP than patients who never developed MODS. It is clear from the data that this hypothesis was correct. Abnormal permeability at the time of admission was present in most of the study sample. This finding is similar to the results of our pilot study and to the results of the work by Harris and colleagues (34). The second question addressed by this study was whether increased IP at admission was associated with an increased risk of developing MODS. We defined this group of individuals as developing MODS more than 24 h after their first IP measurement. All IP measurements at admission were made within 24 h of ICU admission, with the exception of one patient who was excluded from this analysis. Therefore, this group developed the clinical criteria for MODS at least 24 h after ICU admission. In fact, the median time to developing MODS in this group was 6 d. Although this distinction between primary and secondary MODS was arbitrary, we feel that it was justified for two reasons. First, we are unaware of an accepted definition for temporally differentiating primary from secondary MODS. Second, we created this definition prior to the start of the study, on the basis of the best evidence available at the time. Our data clearly showed that the only variable statistically associated with the development of secondary MODS was increased IP on admission. As illustrated in Figure 1, the presence of shock or sepsis was randomly distributed between the three study groups. Furthermore, using multivariate logistic regression techniques, we assessed the impact of multiple variables on the development of secondary MODS. These variables included patient age, IP at admission, the presence of shock or SIRS/sepsis at admission, and the admission APACHE II score. With the exception of IP at admission, the impact of all variables on the subsequent development of MODS was insignificant. These data strongly suggest that increased IP is temporally related to the development of MODS in the ICU, which provides support, but not proof, for an etiologic role of altered barrier function in MODS. During their ICU admission, patients who subsequently developed MODS had a persistently greater impairment and slower overall improvement in IP than did patients who did not develop MODS (Figure 5). No difference observed at admission or in the daily change in permeability between cohorts

450

AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE

could be explained by differences in the APS. Therefore, the association of abnormal IP at admission with the development of MODS cannot be explained by expected confounding variables or by differences in severity of global physiologic dysfunction. The severity of organ dysfunction, as measured by the Marshall MODS score, also correlated with IP at admission (Figure 3). Despite the small sample size in the study, the statistical association was strong, supporting a strong association of impaired IP with MODS and making a type 1 error unlikely. Is increased IP causally associated with the development of MODS? In determining causality, a number of factors must be considered, including biologic plausibility, strength of association, dose–response, temporal sequence of change, reproducibility, absence of confounding or effect-modifying variables, and the exclusion of bias. The findings of increased IP at admission, and the persistence of abnormal permeability throughout the stay in the ICU, would meet the criterion of temporal sequence, with bacterial/endotoxin translocation occurring prior to the development of MODS. The finding that greater IP was associated with a higher MODS score may be analogous to a dose–response effect. Major potential confounders or effect modifiers that alter the association between IP and MODS include the severity of a patient’s physiologic dysfunction or differences in the handling of sugar-marker probes within the body. No relationship between the APS and IP was found that could explain the difference seen between cohorts in our study. No difference was seen in the occurrence of shock or SIRS/ sepsis that might explain the association between IP at admission and MODS. No difference was seen between cohorts in the start and use of standard enteral nutrition. The four areas of handling of the sugar probes include the delivery, intestinal permeation, disposal, and analysis of the probe molecules. The use of two sugars ensures that within individuals, the marker for “normal” permeation is handled in the same way as the marker for “abnormal” permeation. The only point at which the handling of sugars may be different is in their permeation through the mucosal wall (9, 10). Therefore, differences in the LMR between subjects will occur only as a result of differences in permeation, and not as a result of differences between subjects in gastric emptying, rate of intestinal transit, metabolism, or excretion. Since the administration of all sugar probes and the collection of all urines were done in a standardized manner, differences in techniques between cohorts are unlikely to account for the differences in IP observed between cohorts. The collection of all data and the analysis and interpretation of all permeability tests were blinded in order to minimize any observer bias. Therefore, observer or measurement bias is not a likely source of error in accounting for the differences in IP results between the cohorts in our study. The absence of another explanation for the observed association between abnormal IP and development of MODS supports the association of the first with the second as causal. In conclusion, the development of MODS is associated with increased IP at admission. Although this study did not assume a causal relationship between abnormal IP (and, by definition, disruption of the epithelial barrier) and MODS, the observations in the study lend credence to the premise that GI dysfunction may be a stimulus for development of MODS. Acknowledgment : The authors thank Ms. Kim Tran for assistance with analysis of urine samples.

References 1. Beal, A. L., and F. B. Cerra. 1994. Multiple organ failure syndrome in the 1990s: systemic inflammatory response and organ dysfunction.

VOL 158

1998

J.A.M.A. 271:226–233. 2. Cipolle, M. D., M. D. Pasquale, and F. B. Cerra. 1993. Secondary organ dysfunction: from clinical perspectives to molecular mediators. Crit. Care Clin. 9:261–298. 3. Knaus, W. A., and D. P. Wagner. 1989. Multiple systems organ failure: epidemiology and prognosis. Crit. Care Clin. 5:221–232. 4. Carrico, C. J., J. L. Meakins, J. C. Marshall, D. Fry, and R. V. Maier. 1986. Multiple organ failure syndrome. Arch. Surg. 121:196–208. 5. Cerra, F. B. 1992. Multiple organ failure syndrome. Disease-A-Month 38:843–947. 6. Poole, G. V., F. F. Muakkassa, and J. A. Griswold. 1993. The role of infection in outcome of multiple organ failure. Am. Surg. 59:727–732. 7. Haglund, U. 1993. Systemic mediators released from the gut in critical illness. Crit. Care Med. 21:S15–S18. 8. Macho, J. R., and J. M. Luce. 1989. Rational approach to the management of multiple systems organ failure. Crit. Care Clin. 5:379–392. 9. Travis, S., and I. Menzies. 1992. Intestinal permeability: functional assessment and significance. Clin. Sci. 82:471–488. 10. Bjarnason, I., A. Macpherson, and D. Hollander. 1995. Intestinal permeability: an overview. Gastroenterology 108:1566–1581. 11. Kynaston, J. A., S. C. Fleming, M. F. Laker, and A. D. J. Pearson. 1993. Simultaneous quantification of mannitol, 3-O-methyl glucose, and lactulose in urine by HPLC with pulsed electrochemical detection, for use in studies of intestinal permeability. Clin. Chem. 39:453–456. 12. Sorensen, S. H., F. J. Proud, A. Adam, H. C. Rutgers, and R. M. Batt. 1993. A novel HPLC method for the simultaneous quantification of monosaccharides and disaccharides used in tests of intestinal function and permeability. Clin. Chim. Acta 221:115–125. 13. Mainous, M. R., P. Tso, R. D. Berg, and E. A. Deitch. 1991. Studies of the route, magnitude, and time course of bacterial translocation in a model of systemic inflammation. Arch. Surg. 126:33–37. 14. Madara, J. L. 1990. Warner-Lambert/Parke-Davis Award lecture: pathobiology of the intestinal epithelial barrier. Am. J. Pathol. 137:1273– 1281. 15. Bjarnason, I., D. Maxton, A. P. Reynolds, S. Catt, T. J. Peters, and I. S. Menzies. 1994. Comparison of four markers of intestinal permeability in control subjects and patients with coeliac disease. Scand. J. Gastroenterol. 29:630–639. 16. Bjarnason, I., P. Williams, A. So, G. D. Zanelli, A. J. Levi, J. M. Gumpel, T. J. Peters, and B. Ansell. 1984. Intestinal permeability and inflammation in rheumatoid arthritis: effects of non-steroidal anti-inflammatory drugs. Lancet 2:1171–1174. 17. Ukabam, S. O., J. R. Clamp, and B. T. Cooper. 1983. Abnormal small intestinal permeability to sugars in patients with Crohn’s disease of the terminal ileum and colon. Digestion 27:70–74. 18. Van Elburg, R. M., J. J. Uil, J. G. R. De Monchy, and H. S. A. Heymans. 1992. Intestinal permeability in pediatric gastroenterology. Scand. J. Gastroenterol. 27(Suppl. 194):19–24. 19. May, G. R., L. R. Sutherland, and J. B. Meddings. 1993. Is small intestinal permeability really increased in relatives of patients with Crohn’s disease? Gastroenterology 104:1627–1632. 20. Ryan, C. M., J. Schmidt, K. Lewandrowski, C. C. Compton, D. W. Rattner, A. L. Warshaw, and R. G. Tompkins. 1993. Gut macromolecular permeability in pancreatitis correlates with severity of disease in rats. Gastroenterology 104:890–895. 21. Deitch, E. A., W. Bridges, R. Berg, R. D. Specian, and D. N. Granger. 1990. Hemorrhagic shock-induced bacterial translocation: the role of neutrophils and hydroxyl radicals. J. Trauma 30:942–951. 22. Otamiri, T., R. Sjodahl, and C. Tagesson. 1987. An experimental model for studying reversible intestinal ischemia. Acta Chir. Scand. 153:51– 56. 23. Fink, M. P., J. B. Antonsson, H. L. Wang, and H. R. Rothschild. 1991. Increased intestinal permeability in endotoxic pigs: mesenteric hypoperfusion as an etiologic factor. Arch. Surg. 126:211–218. 24. Deitch, E. A., J. Morrison, R. Berg, and R. D. Specian. 1990. Effect of hemorrhagic shock on bacterial translocation, intestinal morphology, and intestinal permeability in conventional and antibiotic-decontaminated rats. Crit. Care Med. 18:529–536. 25. Deitch, E. A., R. D. Specian, and R. D. Berg. 1991. Endotoxin-induced bacterial translocation and mucosal permeability: role of xanthine oxidase, complement activation, and macrophage products. Crit. Care Med. 19:785–791. 26. Deitch, E. A., R. Berg, and R. Specian. 1987. Endotoxin promotes the translocation of bacteria from the gut. Arch. Surg. 122:185–190. 27. Horton, J. W., and P. B. Walker. 1993. Oxygen radicals, lipid peroxidation, and permeability changes after intestinal ischemia and reperfusion. J. Appl. Physiol. 74:1515–1520.

451

Doig, Sutherland, Sandham, et al.: MODS in Critically Ill ICU Patients 28. Ziegler, T. R., R. J. Smith, S. T. O’Dwyer, R. H. Demling, and D. W. Wilmore. 1988. Increased intestinal permeability associated with infection in burn patients. Arch. Surg. 123:1313–1319. 29. Deitch, E. A. 1990. Intestinal permeability is increased in burn patients shortly after injury. Surgery 107411:411–416. 30. Ryan, C. M., M. L. Yarmush, J. F. Burke, and R. G. Tompkins. 1992. Increased gut permeability early after burns correlates with the extent of burn injury. Crit. Care Med. 20:1508–1512. 31. Ohri, S. K., I. Bjarnason, V. Pathi, S. Somasundaram, C. T. Bowles, B. E. Keogh, A. Khaghani, I. Menzies, M. H. Yacoub, and K. M. Taylor. 1993. Cardiopulmonary bypass impairs small intestinal transport and increases gut permeability. Ann. Thorac. Surg. 55:1080–1086. 32. Ohri, S. K., S. Somasundaram, Y. Koak, A. Macpherson, B. E. Keogh, K. M. Taylor, I. S. Menzies, and I. Bjarnason. 1994. The effect of intestinal hypoperfusion on intestinal absorption and permeability during cardiopulmonary bypass. Gastroenterology 106:318–323. 33. Riddington, D. W., B. Venkatesh, C. M. Boivin, R. S. Bonser, T. S. J. Elliot, T. Marshall, P. J. Mountford, and J. F. Bion. 1996. Intestinal permeability, gastric intramucosal pH, and systemic endotoxemia in patients undergoing cardiopulmonary bypass. J.A.M.A. 275:1007–1012. 34. Harris, C. E., R. D. Griffiths, N. Freestone, D. Billington, S. T. Atherton, and R. R. Macmillan. 1992. Intestinal permeability in the critically ill. Intensive Care Med. 18:38–41. 35. Roumen, R. M. H., T. Hendriks, R. A. Wevers, and R. J. A. Goris. 1993. Intestinal permeability after severe trauma and hemorrhagic shock is

36.

37.

38.

39. 40.

41. 42.

increased without relation to septic complications. Arch. Surg. 128: 453–457. Johnston, J. D., C. J. Harvey, I. S. Menzies, and D. F. Treacher. 1996. Gastrointestinal permeability and absorptive capacity in sepsis. Crit. Care Med. 24:1144–1149. Marshall, J. C., D. J. Cook, N. V. Christou, G. R. Bernard, C. L. Sprung, and W. J. Sibbald. 1995. Multiple organ dysfunction score: a reliable descriptor of a complex clinical outcome. Crit. Care Med. 23:1638– 1652. Bone, R. C., R. A. Balk, F. B. Cerra, R. P. Dellinger, A. M. Fein, W. A. Knaus, R. M. Schein, and W. J. Sibbald. 1992. Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine. Chest 101:1644–1655. Keene, A. R., and D. J. Cullen. 1983. Therapeutic Intervention Scoring System: update 1983. Crit. Care Med. 11:1–3. Knaus, W. A., E. A. Draper, D. P. Wagner, and J. E. Zimmerman. 1985. APACHE II: a severity of disease classification system. Crit. Care Med. 13:818–829. Fiddian-Green, R. G. 1988. Splanchnic ischaemia and multiple organ failure in the critically ill. Ann. R. Coll. Surg. Engl. 70(3):128–134. Zeni, F., B. Freeman, and C. Natanson. 1997. Anti-inflammatory therapies to treat sepsis and septic shock: a reassessment. Crit. Care Med. 25:1095–1100.