Introduction

Top Abstract Introduction Materials & Methods Results Discussion References

Despite the use of effective antibiotics, corrective surgical procedures, and life support measures, the mortality rate from sepsis and septic shock has remained high (35-50%) over the last three decades (Natanson, 1997; Zeni et al., 1997). Trials of anti-inflammatory agents [antitumor necrosis factor (TNF) antibodies, interleukin-1 receptor antagonist, antiprostaglandins, platelet-activating factor antagonists, bradykinin antagonists] have, at best, produced only small consistent, nonsignificant beneficial effects on survival rates, or, in specific cases, have been harmful (in high doses glucocorticoids and soluble TNF receptors) (Natanson, 1997; Zeni et al., 1997). Thus, new approaches and/or therapeutic agents for sepsis and septic shock are needed.

One such agent under investigation is pentoxifylline, a methylxanthine derivative that inhibits cyclic nucleotide phosphodiesterases and increases intracellular cyclic nucleotide levels (Ward and Clissold, 1987). Potentially important to septic patients, pentoxifylline has been shown, both in vitro and in vivo, to inhibit TNF production (Han et al., 1990; Doherty et al., 1991), a major mediator of toxicity during sepsis. Based on this, multiple effects of pentoxifylline have been reported in septic patients (Castañon-Gonzalez et al., 1995; Lauterbach and Zembala, 1996; Zeni et al., 1996; Bacher et al., 1997; Staubach et al., 1998). Two double-blind clinical trials examined pentoxifylline's effect on survival rates (Zeni et al., 1996; Staubach et al., 1998). In one trial, five of eight pentoxifylline-treated (1.5 mg · kg¹ · h¹ per 24 h) septic patients and four of eight placebo-treated controls died at 28 days (Zeni et al., 1996). In the other, 8 of 27 pentoxifylline-treated septic patients (1.0 mg · kg¹ · h¹ per 28 days) and 8 of 24 placebo-treated controls died at 28 days (Staubach et al., 1998).

Studies of pentoxifylline's effects on survival rates in animal models of sepsis have also been inconclusive. Different studies have shown pentoxifylline to either improve (Puranapanda et al., 1987; Noel et al., 1990; Hadjiminas et al., 1994; Lundblad et al., 1995; Del Moral et al., 1996), worsen (Fletcher et al., 1992; Ridings et al., 1994), or have no significant effect (Netea et al., 1995) on survival rates. The reason for these disparate outcomes is unknown. Furthermore, no animal study has examined the effects of pentoxifylline administered as a long-term continuous infusion (the regimen most often used in septic patients) in addition to standard therapy with cardiovascular support and antibiotics.

We investigated whether long-term continuous infusions of pentoxifylline are beneficial in an animal model of septic shock. The effects of a wide range of pentoxifylline doses given as a 36-h continuous infusion were studied in addition to therapy with antibiotics and fluid resuscitation. We used a canine model of Gram-negative peritonitis, which reproduces some of the cardiovascular abnormalities of human septic shock and has been shown to be useful for the study of standard and new therapies for septic shock (Natanson et al., 1986; Eichacker et al., 1994; Sevransky et al., 1997).

	Materials and Methods

Top Abstract Introduction Materials & Methods Results Discussion References

Experimental Design and Treatment Groups. Two-year-old (10-12 kg) purpose-bred beagles (male or female; Covance, Vienna, VA, and Marshal Farms, North Rose, NY) were randomly assigned to treatment groups in Table 1. Each week for 14 weeks four of these animals (including at least one control) were infected and studied concomitantly for 28 days or until death. Three sequential experiments were performed. Initially (study 1), in a dose-range study, animals were surgically implanted with an i.p. fibrin-clot infected with Escherichia coli 0111:B4 [1.5 × 10¹⁰ colony-forming units (CFU) per kilogram b.wt. and then randomized to receive a 36-h infusion of one of five doses of pentoxifylline (0.5, 1.0, 5.0, 10, or 20 mg · kg¹ · h¹), or control therapy. Because animals receiving high doses of pentoxifylline died rapidly in the first experiment, only low doses were studied further. In the second experiment (study 2), animals were infected with E. coli 0111:B4, (1.2 × 10¹⁰ CFU per kilogram b.wt.) and then received a 36-h infusion of 5.0, 1.0, or 0.5 mg · kg¹ · h¹ of pentoxifylline or control therapy. Subsequently, in a third experiment (study 3), animals were infected with E. coli 0111:B4 (1.5 · 10¹⁰ CFU per kilogram b.wt.) and received only 0.5 mg · kg¹ · h¹ of pentoxifylline for 36 h, to examine whether a low dose was beneficial. Animals were observed continuously only for the first 36 h after clot implantation. Any unobserved death after the first 36 h was considered to have occurred when the animal was found. Animals sacrificed to relieve suffering by the veterinary staff, blinded to animals' treatment assignment, were considered to be nonsurvivors.

View larger version (24K): [in this window] [in a new window]   Fig. 1.   Outline of treatments and laboratory evaluations.

1

1

1

1

1

Pentoxifylline Preparation. Pentoxifylline (Hoechst-Roussel Pharmaceuticals, Inc., Somerville, NJ), was provided as a white, crystalline powder. Using sterile techniques and pyrogen-free equipment, pentoxifylline was reconstituted with sterile 0.9% saline solution to a concentration of 25 mg × ml¹ and refrigerated at 4°C. Reconstituted pentoxifylline had less than 0.025 endotoxin units per milliliter as measured by the Limulus amebocyte lysate assay.

Temperature, Hemodynamics, and Laboratory Measurements. Hemodynamic data and core temperature were obtained from femoral and balloon flotation thermodilution pulmonary arterial catheters as described. For a baseline evaluation, on day 7, we measured temperature (°C) and hemodynamic and laboratory values before and after fluid challenge. To determine serial effects of sepsis, we repeated the baseline measurements at the times shown in Fig. 1. In addition, on day 0 we obtained additional laboratory studies including pentoxifylline, TNF, quantitative blood culture, and endotoxin levels at the times shown in Fig. 1.

Endotoxin, TNF, and Pentoxifylline Levels. Endotoxin concentrations (endotoxin units per milliliter) were determined from heparinized plasma, which was diluted, heat-treated, and then assayed using a kinetic modification of the chromogenic Limulus amebocyte lysate assay (Whittaker M.A. Bioproducts, Walkersville, MD). Serum TNF levels were determined by a quantitative cytotoxicity assay using WEHI-164 (American Type Cell Collection, Rockville, MD) as described previously (Eichacker et al., 1994; Sevransky et al., 1997).

1

1

In Vitro Study. We evaluated the effects of pentoxifylline on TNF production in both canine and human blood stimulated with endotoxin. Whole blood from six human volunteers and six beagles was drawn in heparinized syringes (100 U of heparin/ml of blood). Aliquots (1 ml) of animal and human blood were inoculated with endotoxin from E. coli 0111:B4 (Sigma Chemical Co., St. Louis, MO) in concentrations of 1 µg · ml¹. Activated blood was incubated with pentoxifylline (0, 0.045, 0.45, or 4.5 µg · ml¹) at 37°C in a 5% CO₂ incubator for 4, 8, or 24 h. After incubation, blood samples were centrifuged at 1200 rpm for 10 min at 15°C and the supernatant was stored at 70°C for determination of TNF levels as previously described (Eichacker et al., 1994; Sevransky et al., 1997). The doses of pentoxifylline and endotoxin used in the in vitro study were based on serum levels of pentoxifylline and endotoxin found in the in vivo study.

Statistical Methods. Survival data were analyzed by a Cox proportional hazards model (Cox, 1972) for pentoxifylline treatment effects. Relative risk and the 95% confidence interval are reported. For hemodynamic, temperature, and laboratory parameters, an analysis of variance (ANOVA) (Scheffe, 1959) was performed. A four-way ANOVA was constructed, with effects for treatment, dog (nested within treatment), time, and bacterial load as the main effects. In addition, two- and three-way interactions were included in the model, with primary attention given to the treatment-time interaction. Higher order interaction that includes dog were pooled to form the error term for the ANOVA. Additionally, the treatment-time interactions were not significantly altered by bacterial load, so only the pooled interactions are reported. The levels of pentoxifylline and its first metabolite were tested at each time point using a Kruskal-Wallis test, modified to test for ordered effects (Siegel, 1956). The resulting p values were adjusted using a Bonferroni procedure to control for the multiple time points analyzed. For statistical analysis, log transformation of endotoxin and TNF levels was performed because these variables had a skewed distribution. For presentation in Table 2 and Appendices 1 and 2, these data were transformed back to the arithmetic scale by taking the antilogs (geometric means). To analyze the in vitro experiment of TNF levels, a four-way ANOVA was constructed, with effects for species, subject (nested within species), dose of pentoxifylline, and time as main effects, and with relevant 2- and 3-way interactions included in the model. The in vivo analysis of TNF levels also used an ANOVA, with a variable to control for week-to-week variation. Mean effects of pentoxifylline on TNF levels, along with a S.E., were computed for both the in vitro and in vivo experiments and were compared using a t test.

Animal Care. This protocol was approved by the Animal Care and Use Committee of the Clinical Center of the National Institutes of Health. Animals were housed according to guidelines of the Veterinary Resources Branch of the National Institutes of Health in temperature- (23°C) and humidity- (50%) controlled runs. During the experiments, animals were kept in standard stainless steel individual large animal cages or specially constructed slings, otherwise, animals were unrestrained. Animals had unrestricted access to food (dog food) and water throughout the study except for 12 h before surgery. All efforts were undertaken to minimize animal pain and suffering. If the independent veterinarian staff determined that an animal was unduly suffering, it was sacrificed.

	Results

Top Abstract Introduction Materials & Methods Results Discussion References

Clinical Manifestations and Survival. During the 4 days after infected clot implantation, all animals had typical signs of sepsis (weakness, lethargy, and anorexia). The relative risk of death for these animals was increased with pentoxifylline therapy at doses of 5.0, 10, and 20 mg · kg¹ · h¹ (Fig. 2). Moreover, this increase in relative risk of death was dose ordered (pentoxifylline 20  10  5.0  1.0  0.5 mg · kg¹ · h¹, p = .008).

View larger version (31K): [in this window] [in a new window]   Fig. 2.   Proportion of animals surviving versus time.

Temperature and Hemodynamics. At baseline, temperature (38.05 ± 0.05°C, mean ± S.E.) and all hemodynamic values measured were similar in all animals treated with pentoxifylline and controls (p = N.S.). At 6 h after clot implantation, there were no significant differences in temperature and hemodynamic parameters among animals treated with pentoxifylline and controls (p = N.S.). In addition, by 24 h, only animals treated with pentoxifylline at 1.0 and 0.5 mg · kg¹ · h¹ and controls were alive. Furthermore, temperature and hemodynamic variables were remarkably similar in animals treated with pentoxifylline at 0.5 mg · kg¹ · h¹ and controls (Fig. 3, p > .20). Thus, to increase our ability to find significant effects, we combined the pentoxifylline 0.5 mg · kg¹ · h¹ group and controls and appropriately corrected the p values for the number of ways the data could be combined.

View larger version (26K): [in this window] [in a new window]   Fig. 3.   Serial mean ± S.E. changes in temperature (A), left ventricular ejection traction (B), systemic vascular resistance (SVRI) (C), and cardiac index (CI) (D) in animals treated with pentoxifylline (0, 0.5, 1.0, 5, 10, and 20 mg · kg1 · h1 per 36 h).

1

1

1

1

Endotoxin Levels and Quantitative Blood Cultures. Pentoxifylline after infected clot implantation had a different effect on endotoxin levels at early (0-24 h) compared with later (24-672 h) time points (p = .0014; Fig. 4 and Appendix 1). From 0 to 24 h after infected clot implantation, pentoxifylline at all doses studied increased mean endotoxin levels compared with controls (p = .003). In contrast, from 24 to 672 h, pentoxifylline at all doses decreased mean endotoxin levels compared with controls (p = .049). After clot implantation, there were no significant differences throughout when comparing the increases in numbers of CFU in blood of animals treated with pentoxifylline and controls (data not shown, p = N.S.).

View larger version (21K): [in this window] [in a new window]   Fig. 4.   Serial mean ± S.E. changes in endotoxemia. For effect of individual pentoxifylline doses, see Appendix 1

TNF Levels (In Vitro Experiments in Animals versus Humans). In canine and human whole blood assays, pentoxifylline had similar effects on mean TNF levels (Table 2, p = .9). In the absence of endotoxin in both canine and human whole blood assays, pentoxifylline at all doses and time points studied had no significant effect on mean TNF levels (p = .8). In endotoxin-stimulated canine and human blood assays, pentoxifylline had no significant dose-dependent effect on TNF production; however, at all doses and time points studied, pentoxifylline produced a borderline significant decrease in mean TNF levels (p = 0.09; Table 2).

TNF Levels (In Vivo Versus in Vitro Experiments in Animals). Pentoxifylline had a significantly different and opposite effect on mean TNF levels in the in vivo versus in vitro animal experiments (p = .04). When all doses and time points studied were combined, pentoxifylline increased mean TNF levels in vivo after infected clot implantation compared with controls (mean ± S.D., percent change in TNF levels, 126 ^× 1.09%, p = .025; Appendix 2); whereas pentoxifylline decreased mean TNF levels in vitro in endotoxin-stimulated whole blood (mean ± S.D., percent change in TNF levels, 70 ^× 1.20%, p = .09, Table 2).

Pentoxifylline Levels. At 15 min and 3, 6, and 24 h after the start of the pentoxifylline infusion, pentoxifylline and its first metabolite's levels were dose ordered (5.0  1.0  0.5  0 mg · kg¹ · h¹, all p < .001; Fig. 5). By 12 h after discontinuation of continuous infusion of pentoxifylline, levels of pentoxifylline and its first metabolite were undetectable.

View larger version (25K): [in this window] [in a new window]   Fig. 5.   Serial mean × S.E. changes in pentoxifylline levels and its first metabolite (insert) are plotted on the y-axis in animals treated with pentoxifylline (0, 0.5, 1.0, and 5.0 mg · kg1 · h1 per 36 h).

Other Laboratory Measures. There were no other significant differences associated with pentoxifylline therapy at any dose studied in the other laboratory parameters measured (p = N.S.). Furthermore, there were no significant differences in any laboratory parameter measured at baseline (7 days) when compared with groups that would receive pentoxifylline and controls (p = N.S.).

	Discussion

Top Abstract Introduction Materials & Methods Results Discussion References

In an antibiotic-treated, volume-resuscitated canine model of Gram-negative peritonitis, pentoxifylline at doses of 0.5 to 20 mg · kg¹ · h¹ given as a 36-h continuous i.v. infusion increased mortality rates in a dose-dependent fashion. This increase in mortality rates was associated with altered endotoxin clearance from the blood, increased TNF levels, and worsened cardiac function. One hypothesis consistent with these data is that high pentoxifylline levels in the first 24 h after infection slowed endotoxin clearance from the blood, resulting in high levels of endotoxemia, increased release of proinflammatory mediators such as TNF, worsened cardiac function, and increased mortality rates.

There were no significant dose-dependent increases in TNF or endotoxin levels to explain these dose-dependent significant increases in mortality rates with pentoxifylline therapy. This could be because of variability in endotoxin and TNF measures combined with the small numbers of animals studied at the high doses of pentoxifylline. Alternatively, in septic animals receiving high doses (5, 10, and 20 mg · kg¹) of pentoxifylline, death could be related to some rapidly lethal process associated with marked increases in intracellular nucleotides independent of TNF and endotoxin levels. The increased but lower mortality rate in animals receiving lower doses of pentoxifylline (1.0 mg · kg¹ · h¹) was associated with hypothermia, worsened cardiac dysfunction, and increases in peripheral resistance. These hemodynamic changes have been associated with increased TNF levels as found in this study (Eichenholz et al., 1992). However, very low doses of pentoxifylline (0.5 mg · kg¹ · h¹) were not associated with temperature or hemodynamic changes, but were associated with worsened outcome. It is possible that pentoxifylline, by increasing intracellular nucleotides (Bessler et al., 1986) levels during severe infection, depending on the doses used, could have multiple mechanisms to produce harm.

Previous animal studies examining the effects of pentoxifylline on mortality rates during sepsis and septic shock have yielded conflicting results. Pentoxifylline has been shown to increase survival rates in canines (Puranapanda et al., 1987) and rats (Noel et al., 1990) challenged with endotoxin, rats (Lundblad et al., 1995) and mice (Hadjiminas et al., 1994) after cecal ligation and puncture, and piglets (Del Moral et al., 1996) challenged with group B streptococcus. Pentoxifylline has also been shown to increase mortality rates in other sepsis models (Fletcher et al., 1992; Ridings et al., 1994). These disparate results could potentially be due to differences in the models studied, as well as dose, timing, and route of pentoxifylline administration. In support of this, in endotoxin-challenged rats (Fletcher et al., 1992), pentoxifylline given as a single bolus (20 mg · kg¹) improved survival rates, whereas multiple pentoxifylline injections at this same dose worsened survival rates. In this rat study, a lower dose of pentoxifylline (12 mg · kg¹), whether administered in single or repeated doses, had no significant impact on survival rates (Fletcher et al., 1992). Swine challenged with i.v. Pseudomonas aeruginosa and given pentoxifylline before the onset of infection had less sepsis-induced organ dysfunction (Ridings et al., 1994). In contrast, pentoxifylline given 2 h after onset of infection caused profound hemodynamic instability and death. Taken together, these studies suggest that pentoxifylline may have a narrow therapeutic index and high risk-benefit ratio, such that depending on species studied, timing of administration, and doses used, its impact on outcome may markedly differ. Of note, in severe asthma, continuous infusions of methylxanthines have also been shown to have a narrow therapeutic range and high risk-benefit ratio, limiting their use (Strauss et al., 1994).

The reported effects of pentoxifylline on hemodynamics have also been variable. During sepsis in humans pentoxifylline has been documented to either have no hemodynamic effect (Zeni et al., 1996) or to increase cardiac index (Castañon-Gonzalez et al., 1995; Bacher et al., 1997) and decrease systemic vascular resistance (Castañon-Gonzalez et al., 1995; Bacher et al., 1997). In septic models, pentoxifylline has been shown to improve (Law et al., 1992; Sigurdsson and Youssef, 1993; Del Moral et al., 1996), to worsen (Ridings et al., 1994), or to have no effect (Flamand et al., 1995) on various hemodynamic parameters. Among these studies there are marked differences in doses of pentoxifylline used and timing of administration. Our data suggest pentoxifylline infusions worsen cardiac function, but this hemodynamic effect varies with the dose. Taken together, these data further indicate that during sepsis pentoxifylline may have a narrow therapeutic index and high risk-benefit ratio for hemodynamics, as well as survival.

This is the first study to examine serially the effects of pentoxifylline on endotoxin clearance during sepsis. Interestingly, alterations in endotoxin clearance were not associated with changes in bacterial clearance or numbers of circulating neutrophils. However, in vitro studies have shown that pentoxifylline has several effects on leukocyte function that could adversely impact on endotoxin clearance. Pentoxifylline has been shown to inhibit neutrophil chemotaxis (Sullivan et al., 1984; Hammerschmidt et al., 1988), phagocytosis (Bessler et al., 1986), degranulation (Sullivan et al., 1988), and generation of superoxide (Bessler et al., 1986; Hammerschmidt et al., 1988; Sullivan et al., 1988; Carletto et al., 1997). It is possible that these inhibitory effects on leukocyte function with high pentoxifylline levels could have an adverse impact on the clearance of endotoxin from the blood. Once the pentoxifylline infusion was stopped, high endotoxin levels led to activation of compensatory mechanisms such that endotoxin clearance was subsequently increased.

This study also demonstrates for the first time that pentoxifylline in vivo can actually increase TNF levels during bacterial sepsis. This finding is in contrast to other investigations that have shown pentoxifylline decreases TNF levels during human sepsis (Zeni et al., 1996) and in animal models (LeMay et al., 1990; Noel et al., 1990; van Leenen et al., 1993; Lundblad et al., 1995). Although these disparate results could be related to species studied, we have shown in vitro that pentoxifylline reduces endotoxin-stimulated TNF production in canine whole blood the same as it does in whole blood assays of other animals as well as humans (Barton and Moore, 1994; Bienvenu et al., 1995; D'Hellencourt et al., 1996). In vitro, reduction of endotoxin-induced TNF production is related to down-regulation of TNF gene transcription (Han et al., 1990; Doherty et al., 1991). Perhaps under the conditions produced in vivo in our canine model, the pentoxifylline-associated delayed clearance of endotoxin from the blood resulted in markedly increased stimulus for TNF production. This increased stimulus for TNF production during Gram-negative peritonitis-induced endotoxemia may have had a greater effect on TNF production than pentoxifylline's known inhibitory effect on TNF gene transcription. Of note, human studies reporting TNF levels (Lauterbach and Zembala, 1996; Zeni et al., 1996) included patients with many types of bacterial infection that may have impacted on levels of endotoxemia and subsequent effects of pentoxifylline on TNF production. In addition, endotoxin given as a bolus or infusion may not be associated with as prolonged endotoxemia as found in our study during Gram-negative peritonitis.

A number of points are worth noting regarding endotoxemia, TNF production, and survival in the setting of live bacterial infection in our model. First, therapies that have decreased TNF levels in our canine model such as tyrosine kinase inhibitors (Sevransky et al., 1997) and granulocyte colony-stimulating factor (Eichacker et al., 1997) have significantly improved survival rates. Second, increases in endotoxin levels in canines that are less than those seen in this study in the absence of significant increases in serum TNF levels have been associated with worsened cardiac dysfunction and/or increased mortality rates (Eichacker et al., 1993). Therefore, in the setting of live bacterial infection, the association between endotoxin and TNF levels and survival can be rather complex and is not entirely understood. Nevertheless, previous studies using this canine model support its ability to determine the effects of therapies that alter endotoxin clearance and TNF production on survival of Gram-negative septic shock.

In summary, in a canine model of Gram-negative peritonitis, pentoxifylline, over a wide range of doses given as a continuous i.v. infusion, altered endotoxin clearance, increased TNF levels, and worsened cardiac function and survival rates. Previous animal studies of pentoxifylline have yielded conflicting results and have not clearly defined therapeutic mechanisms. Therefore, the conditions under which pentoxifylline has beneficial, neutral, or deleterious effects needs to be established before clinical use.