1367

Effectiveness of a Human Monoclonal Anti-Endotoxin Antibody (HA-lA) in Gram-Negative Sepsis: Relationship to Endotoxin and Cytokine Levels Cornelis H. Wortel, * Marijke A. M. von der Mohlen, Sander J. H. van Deventer, Charles L. Sprung," Michael Jastremski, Maarten J. Lubbers, Craig R. Smith, I. Elaine Allen,* and Jan W. ten Cate

Center for Hemostasis, Thrombosis. and Atherosclerosis Research. Department of Gastroenterology. and Department of Intensive Care. Academic Medical Center, University ofAmsterdam. Netherlands; VA Medical Center. University of Miami School of Medicine, Florida; State University ofNew York. Syracuse; Research and Development Division. Centocor, Malvern, Pennsylvania

Gram-negative sepsis is a clinical syndrome frequently associated with serious sequelae such as multiple organ failure and death [I]. Endotoxins have been implicated as an important causal factor in the development of this syndrome [2]. Several lines of evidence indicate that the biologic effects of endotoxin are mediated by tumor necrosis factor (TNF) and interleukins (IL )-1 and -6 [3-8]. These cytokines are released from macrophages and endothelial cells after interaction with endotoxin. The prospect of inhibiting the biologic effects of endotoxin with antibodies directed against core determinants of the lipopolysaccharide molecule has been studied in animal models of endotoxemia and gram-negative bacteremia. These studies demonstrated that human polyclonal anti-

Received 31 July 1991; revised 23 July 1992. Presented in part: International Congress for Infectious Diseases. Montreal. July 1990 (abstract 495); 30th Interscience Conference on Antimicrobial Agents and Chemotherapy. Atlanta, October 1990 (abstract 239). Financial support: Centocor, Inc.: S.J.H.v.D. is a Fellow of the Royal Dutch Academy of Arts and Sciences. The study protocol was approved by the Institutional Review Boards of the participating study centers. and human experimentation guidelines were followed. Informed consent was obtained from the patients or their parents or guardians before entry into the study. Reprints or correspondence: Dr. C Wortel, Department of Critical Care and Emergency Medicine. Ben Taub General Hospital. 1504 Taub Loop, Houston, TX 77030. * Present affiliations: Department of Clinical Research. Centocor, Malvern, Pennsylvania, and Department of Critical Care Emergency Medicine. Baylor College of Medicine, Houston (CH. W. adjunct appointments); Hadassah Hebrew University Medical Center, Jerusalem (CL.S.); Ariad Pharmaceuticals. Cambridge. Massachusetts (I.E.A.). The Journal of Infectious Diseases 1992;166:1367-74 © 1992 by The University of Chicago. All rightsreserved. 0022-1899/92/6606-0024$01.00

serum to endotoxin produces a protective effect [9]. Clinical trials in patients with gram-negative sepsis showed that human polyclonal antiserum directed against endotoxin can reduce mortality in septic patients and prevent the development of sepsis in high-risk surgical patients [10, II]. On the basis of this work, a human monoclonal IgM (HA-IA) was developed that binds specifically to the lipid A domain of endotoxin [12]. We hypothesized that HA-I A should decrease mortality in septic patients with endotoxemia and reduce serum levels ofTNF and IL-6 during the first 24 h after therapy, To test this hypothesis, we studied 82 septic patients enrolled in a large, multicenter, placebo-controlled clinical trial of HAIA [13] with the objective of determining the efficacy of HA-I A in patients with endotoxemia and delineating the effects of HA-I A on serum TNF and IL-6 levels.

Methods Consecutive adult patients with suspected gram-negative sepsiswho participated in a larger prospective, randomized, doubleblind, multicenter, placebo-controlled trial ofHA-IA were eligible for this study. The criteria for diagnosis of sepsis had to be met within 24 h before enrollment and are described in detail elsewhere [13]. All participating centers in the overall HA-IA trial were invited to participate in this substudy when the amendment to the original protocol concerning these measurements was written. Only three centers could dedicate personnel for the sampling required in the protocol. In Amsterdam, 52, in Miami, 30, and in Syracuse, 8 consecutive patients were entered from November 1988 until the end of the study. Since the patients were randomized per center in blocks of 4, the patients entered in this substudy were randomized as well.

Downloaded from http://jid.oxfordjournals.org/ at Indiana University Library on July 13, 2015

Gram-negative sepsis is caused by endotoxin-induced release of tumor necrosis factor (TNF) and other cytokines. HA-1A is a human monoclonal antibody that binds specifically to endotoxin. HA-1A should prevent death in endotoxemic patients and reduce serum levelsofTNF and interleukin-6 (lL-6). This hypothesis was tested in 82 septic patients who were randomly allocated to receive a single intravenous 100-mg dose ofHA-1A or placebo. Pretreatment endotoxemia was detected in 27 patients (33%). Death occurred within 28 days of treatment in 8 (73%) of 11 placebo recipients and in 5 (31%) of 16 HA-1A recipients (P = .02). The median decrease in serum TNF level 24 h after treatment was 12 nglL in patients given HA-1A and 0 ng/L in placebo recipients (n = 65; P = .04). For IL-6, this was 204 ng/L in patients given HA-1A and 44 ng/L in placebo recipients (n = 67; P = .4). Thus, HA-1A reduces mortality in septic patients with endotoxemia and lowers serum TNF levels.

1368

Wortel et al.

Table 1. Baseline characteristics of patients treated with HA-I A or placebo. Treatmentgroup

Mean age (years) Gram-negative bacteremia Gram-negative infection Endotoxemia Severity of disease Mean APACHE II score Hypotension and/or vasopressors Disseminated intravascular coagulation Adult respiratory distress syndrome Acute hepaticfailure Acute renal failure Underlying conditions Gastrointestinal disease Recent surgery Recent trauma

Placebo

HA-IA

(n = 41)

(n = 41)

57.6 ± 14.0

57.0 ± 16.7

13 (32)

19 (46)

32 (78) II (41)'

30 (73) 16(59)*

30.2 ± 10.4

25.6 ± 9.0 t

31 (76)

28 (68)

14 (34)

6 (15),

3 (7) 14 (34) 21 (51)

2 (5) 9 (22) 13(32)

15 (37) 17 (41) 4 (10)

17 (41) 17 (41) I (2)

NOTE. Unlessindicated, data are no. (%). Meansare given ± SD. 'n = 27. t P = .03 compared with placebo recipients. t p = .04 compared with placebo recipients.

TNF and IL-6 levels were calculated by subtracting pretreatment values from the values obtained I and 24 h after treatment. The changes in pre- and postinfusion TNF and IL-6levels were compared in the two treatment groups using the Wilcoxon rank sum test. P < .05 (two-tailed) was used as the criterion for statistical significance.

Results

Clinical data. Ninety consecutive septic patients at three sites participating in the multicenter HA-I A study were enrolled. In 8 patients (4 in each treatment group), the samples obtained for endotoxin determination were contaminated. These patients were excluded before data analysis. Of the 82 remaining patients, 41 each were treated with HA-I A and placebo. There were no statistically significant differences between groups in the presence of hypotension, the prevalence of gram-negative bacteremia, and the frequency of underlying diseases at study entry (table 1). However, the mean APACHE II score was 30.2 ± 10.4 in the placebo group and 25.6 ± 9.0 in the HA-IA-treated group (P = .03), and the incidence of disseminated intravascular coagulation was 34% in the placebo group and 15% in the HA-I A group (P = .04). Since the APACHE II score is the strongest predictor for mortality in this study, we adjusted for the difference

Downloaded from http://jid.oxfordjournals.org/ at Indiana University Library on July 13, 2015

Patients were randomly allocated to receive either 100 mg of HA-I A (Centocor) or a human serum albumin placebo. Patients were followed for 28 days or until death. The severity of the acute illness was assessed by APACHE II score [14]. Blood and samples of all other suspected foci of infection were obtained for culture within the 24 h before enrollment. Gramnegative bacteremia was defined as isolation of pathogenic gram-negative bacteria from at least one blood culture. Blood samples were obtained either from arterial catheters that had been in situ ~36 h or by direct puncture ofa peripheral vein. Immediately before infusion of the study medication and I and 24 h after infusion, blood samples for the endotoxin, TNF, and IL-6 assays were collected. Blood samples for the endotoxin assay were collected in monovettes (Sarsted, Numbrecht, Germany) prefilled with endotoxin-free heparin so as to give a final concentration of 50 IU/mL (Organon, Oss, Netherlands) and immediately immersed in melting ice. Platelet-rich plasma was prepared within I h by centrifugation for 10 min at 180 g and 4°C. Each blood sample was divided into two aliquots in polystyrene tubes (Falcon 2063, Oxnard, CA) and stored at -20°C or lower until being assessed in batches. The endotoxin assay was done with a previously described and validated chromogenic method [15, 16]. To avoid misclassification, aliquots were each assayed in duplicate. Assay results were taken as the mean value of quadruplicate tests. The sensitivity of this assay is 3 ng/L. Endotoxemia was defined as a plasma endotoxin level> 5 ng/L in both aliquots [16, 17]. The coefficient of variation for the assay was 14%. Contamination with exogenous endotoxin was suspected when levels> 100 ng/L were detected in consecutive samples. After review ofthe sampling procedure and assessment of additional samples when available, a sample was designated contaminated before the randomization code was broken. Within 2 h after blood sampling, serum aliquots for TNF and IL-6 assays were prepared by centrifugation at 1200 g and room temperature. The aliquots were immediately frozen and stored at -20°C or lower in polystyrene tubes until assessment. All TNF levels were determined in a single run using a commercial immunoradiometric assay (Medgenix, Brussels). Assay results were the mean values of duplicate tests. The sensitivity of this assay was 5 ng/L, and normal values are below the detection limit. All IL-6 levels were determined batchwise (by L. Aarden, Central Laboratory of the Red Cross Blood Transfusion Service, Amsterdam) using the B9 bioassay as described previously [18]. Assay results were the mean values of duplicate tests. Sensitivity of this assay is 0.5 ng/L, and normal values are < I0 ng/L (Aarden L, personal communication). All assays for endotoxin, TNF, and IL-6 were interpreted blindly, without knowledge of the clinical condition, infection status, or treatment regimen. The baseline characteristics of the patients were compared using Student's t test or the Wilcoxon rank sum test for continuous variables and the x 2 test for categorical variables. Patients were grouped by the presence or absence of pretreatment endotoxemia, and survival was estimated using the Kaplan-Meier method. The effects ofHA-IA or placebo on survival were compared using the Wilcoxon statistic. A Cox proportional hazards model was fitted to the survival data to estimate the treatment effect of HA-I A after adjustment for differences in APACHE II scores in the placebo and HA-I A treatment groups. Changes in

JID 1992; 166 (December)

HA-IA in Endotoxemic Patients

JID 1992; 166 (December)

Table 2. Baseline characteristics of patients in this study and in the larger study [13] of which this group is a part.

1369

Table 3. Underlying infection and survival in patients with endotoxemia.

Study

(n

58.2 ± 16.5

Present study (n

=

82)

59.0 ± 14.1

200 (37)

32 (39)

401 (74)

62 (77)

25.0 ± 8.6

Underlying infection

No. endotoxernic/total

Placebo

HA-IA

18/32 5/62 2/4

2/8 1/1 0/1

7/10 3/4

2/16

0/1

1/1

Gram-negative bacteremia Gram-ngeative infection Gram-positive infection No gram-negative or -positive infection

l/l

27.9 ± 9.9

283 (52)

59 (72)

88 (16)

20 (24)

78 (14) 114 (21) 207 (38)

5 (6) 23 (28) 34 (41)

191 (35) 181 (33) 37 (7)

32 (39) 34 (41) 5 (6)

NOTE. Unless indicated, data are no. (%). Means are given ± SD.

in APACHE II scores between treatment groups at study entry when assessing the effect of HA-I A on survival. Comparison of this subset of patients with the larger HAI A study population [13] shows that in the subgroup sepsis was more severe (table 2). A statistical comparison could not be done since the subset is part of the larger patient population.

Efficacy of HA-IA in patients with pretreatment endotoxemia. Endotoxemia was detected before treatment in 27 (33%) of the 82 septic patients studied. Endotoxemia was present in 18 (56%) of the 32 with gram-negative bacteremia and in 9 (18%) of the 50 without gram-negative bacteremia (P = .00 I). Endotoxemia was furthermore detected in 5 of the 62 patients with gram-negative infection, 2 of the 16 without any detectable gram-negative or -positive infection, and 2 of the 4 with a pure gram-positive infection. In the latter patients, however, the levels were very close to the upper or lower bound for endotoxemia: 6 ng/L (subsequent levels .5). At I h after infusion, the median change in TNF level was a 5-ng/L decrease in the HA-I A-treated patients and a l-ng/L decrease in the placebo recipients (P = .13). At 24 h after infusion, the median change was a 12-ng/L decrease (interquartile range, -131 to -I) in HA-IA-treated patients and 0 ng/L (interquartile range, -26 to 6) in the placebo group (P = .04) (figure 3). In the 30 placebo recipients, we evaluated the relationship between a change in TNF levels and survival. The 24 placebo recipients that were alive at day 3 showed a decrease of 21 ± 26 ng/L and the 6 nonsurvivors showed an increase of 84 ± 52 ng/L (mean ± SD, P = .08). Of the patients who showed a decrease in TNF levels within 24 h, 12 (86%) of 14 were alive at day 3. The median serum concentrations of IL-6 before infusion were 330 ng/L (range, 17-57,000) in the HA-IA-treated patients and 342 ng/L (range, 1-31,900) in the placebo recipients (P> .5). At I h after infusion, the median change in serum IL-6 level was a 112-ng/L decrease in the HA-I A recipients and an 18-ng/L decrease in placebo recipients (P = 1.0

..,r--------------------,

0.8

HA-1A (ij

>

0.6

.~

:::I (/l

'0

0.4

Placebo

~

..

:a J:l 0

rr.

P = 0.13

0.2

0.0 0

4

6

8

10

12

14

16

18

20

22

24

26

28

Days Since Infusion

Figure 2. Probability of survival in patients without endotoxemia treated with HA-IA (n = 25) or placebo (n = 30); 37% reduction in mortalitywas seen with HA-IA treatment.

Downloaded from http://jid.oxfordjournals.org/ at Indiana University Library on July 13, 2015

0.0

lID 1992; 166 (December)

HA-IA in Endotoxemic Patients

JID 1992; 166 (December)

Discussion Recent evidence has demonstrated that immunotherapy with a human monoclonal IgM antibody (HA-IA) directed

--..J

5

~

l:

U.

Z I-

f-

0

eI)

III

as eI)

...u

l-

f-

C)

--

Placebo

Placebo

--

-5 -

eI)

C

e ·10 as

=s eI)

::iii: -15

-

--

HA-1A

*P

= 0.04 1

HA-1A* 24

Hours Following Infusion Figure 3.

Median decrease in TNF (tumor necrosis factor) level I and 24 h after infusion ofHA-IA (n = 33) or placebo (n = 32).

Table 5.

Relationship of pretreatment interleukin-6 level to 28day mortality. Treatment group

Pretreatment interleukin-6 level (ng/L) 0-99 100-1000 >1000

Placebo

HA-IA

3/11 (27) 8/10 (80) 11/12 (92)

1/9 (II) 7/16 (44) 6/9 (67)

NOTE. Data are no. of patients dying within 28 days/total (%). P < .016, Cochran-Mantel-Haenszel test, across all strata.

against endotoxin reduces mortality in septic patients with gram-negative bacteremia [13]. However, in previous studies no data were obtained on endotoxin and cytokine levels after anti-endotoxin immunotherapy. We used recently developed reliable methods for detection of endotoxin, TNF, and IL-6 in blood to study the effects of HA-I A treatment on mortality in patients with and without endotoxemia and the ability of HA-I A to reduce serum levels of proinflammatory endogenous mediators in patients with gram-negative sepsis. The results of this study show that treatment with HA-IA significantly reduces mortality during 28 days after infusion in septic patients with endotoxemia. Adjustment oftheHAIA effect for differences in APACHE II score or IL-6 levels before treatment did not alter the significance of this finding. In three subgroups of patients with endotoxemia alone, gram-negative bacteremia alone, or endotoxemia and gramnegative bacteremia, we observed a decrease in mortality when treated with HA-IA (respectively from 67% to 33%, P = .524; 80% to 44%, P = .30 I; 75% to 30%, P = .058). In the absence of gram-negative bacteremia, HA-I A decreased the mortality from 54% to 32% (F: = .124). However, 16% of these patients were endotoxemic at the time of infusion. Given the extremely small number of patients, not much importance can be given to these analyses. Our results are consistent with the hypothesis that HA-I A specifically interferes with the biologic effects of endotoxin in septic patients. Furthermore, since HA-I A binds specifically to the lipid A domain of endotoxin, these data suggest that endotoxin is a causal factor in the mortality associated with gram-negative sepsis [12, 13]. The determination of endotoxemia would therefore be helpful in identifying a gram-negative patient population. Our results show that the limulus assay detects endotoxemia more frequently in patients with gram-negative bacteremia (56%) than in patients with gram-negative infection (8%) or undetectable bacterial infection (13%). In two patients with pure gram-positive infection we observed borderline endotoxin levels, which could represent false-positive results. At present, routine use ofthe limulus assay does not seem practical. Experienced analytic personnel should be available at all times. False-positive results due to contamination with

Downloaded from http://jid.oxfordjournals.org/ at Indiana University Library on July 13, 2015

.02). At 24 h after infusion, the median change was a 204ngiL decrease (interquartile range, -1584 to -12) in the HA-I A recipients and a 44-ng/L decrease (interquartile range, -884 to 25) in the placebo recipients (P = .4). To determine the relationship between pretreatment IL-6 levels and mortality, the IL-6 levels were stratified into three groups in placebo-treated patients « I00, 100-1000, > 1000 ng/L) because the values had a log-normal distribution. The correlation between pretreatment IL-6 levels and mortality was highly significant (P = .005) in placebo-treated patients. HA-I A treatment significantly reduced mortality overall after adjusting for pretreatment IL-6 level by the CochranMantel-Haenszel test (P = .016) (table 5). Although the numbers are small and the patients that died within 24 h after infusion cannot be assessed, we observed in 46 (77%) of 60 patients a decrease in IL-6 levels within the 24-h period. Of those patients, 22 (48%) survived and 24 (52%) died within 28 days. An increase in IL-6 levels was observed in 12 (20%) of60 patients, ofwhom 4 survived and 8 died (P = .37 for the comparison between surviving patients with a decrease or increase in IL-6 levels, x2 ) . Blood for endotoxin measurement was obtained 24 h after treatment in 21 endotoxemic patients. Samples were not obtained in 6 patients: 2 had contaminated samples and 2 were not sampled at 24 h in the HA-I A group, and 2 endo toxemic patients in the placebo group died within 24 h. Twenty-four hours after treatment, endotoxemia was no longer present in 8 of 12 given HA-IA and 5 of9 given placebo (P = .673). A very strong relationship was observed between the concomitant presence of TNF and IL-6 in the serum. No such relationship was observed for the presence of endotoxin and TNF or IL-6.

1371

Wortel et al.

1372

Table 6. Relationship between endotoxemia and gram-negative bacteremia in septic patients in various studies. Incidence of endotoxemia, no.jtotal (%)

Year

[28) [29) [30) [31) [32) [33) [34)' Present study!

1970 1972 1973 1988 1988 1989 1991 1992

10/15 (67) 17/31 (55) 28/65 (43) 9/10 (90) 16/19(84) 24/35 (69) 11/19(58) 18/32 (56)

Without gram-negative bacteremia 6/80 (6) 16/184 (9) 11/43 (26) 3/16(19) 1/4 (25)

0/7 32/81 (39) 9/50(18)

p

Effectiveness of a human monoclonal anti-endotoxin antibody (HA-1A) in gram-negative sepsis: relationship to endotoxin and cytokine levels.

Gram-negative sepsis is caused by endotoxin-induced release of tumor necrosis factor (TNF) and other cytokines. HA-1A is a human monoclonal antibody t...
734KB Sizes 0 Downloads 0 Views