INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Although antithrombin (AT) concentrates are commercially available in several countries for patients with acquired AT deficiency, this indication has not yet been supported by conclusive evidence from well-designed studies (1, 2). Published controlled trials of AT concentrate have been few, small, and mostly limited to laboratory findings rather than to clinical end points (2). This report describes the largest randomized trial to date of the clinical utility of AT concentrate in a homogenous group of patients with acquired AT deficiency.

Premature infants with respiratory distress syndrome (RDS) have low plasma activities of AT (3) as well as increased markers of thrombin formation such as thrombin-antithrombin (TAT) complexes (7). There is also an abundance of intrapulmonary fibrin deposition in the form of hyaline membranes (10). Experimental evidence suggests that thrombin and fibrin in the lung contribute to characteristic features of acute lung injury such as pulmonary hypertension (14), endothelial permeability (15), surfactant dysfunction (16, 17), inflammation (18, 19), and fibrosis (20, 21).

We therefore hypothesized that premature infants with RDS would benefit from antithrombotic therapy with AT concentrate. A placebo-controlled randomized trial was designed to test whether AT concentrate reduces thrombin formation, improves gas exchange, and decreases the duration of mechanical ventilation and oxygen therapy. The safety of AT therapy was evaluated primarily by comparing the incidence of severe (Grade 3) intraventricular hemorrhage and of periventricular echodensities in the two treatment groups.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Selection of Patients and Randomization

This trial took place in the Neonatal Intensive Care Unit (NICU) at McMaster University Medical Centre from November 1992 to February 1996. This 33-bed NICU serves a regional population of 1.8 million people. Approximately 1,000 newborns are admitted annually, of whom 80% are born in the center; the remainder are referred from surrounding hospitals.

Premature infants were eligible for enrollment if they fulfilled all of the following inclusion criteria: birth weight, 750 to 1,900 g; postnatal age between 2 and 12 h; endotracheal intubation and positive pressure ventilation for RDS; indwelling arterial catheter; ratio of arterial to alveolar oxygen pressure [(a/A)PO₂] less than 0.3 after the first dose of exogenous surfactant. Infants were excluded for any of the following reasons: congenital infection, congenital malformation(s), hydrops, pulmonary hypoplasia, clinically apparent bleeding disorder, or thrombocytopenia (platelet count of 50 × 10⁹/L or less). Infants considered to be moribund were also excluded.

The study protocol was approved by the Research Advisory Group at McMaster University and by the Health Protection Branch of Health and Welfare Canada under an Investigational New Drug submission. Parents of eligible infants were informed in detail about the trial, and written consent was obtained.

Before randomization, each infant received cranial ultrasonography to document any periventricular and intraventricular pathology before the administration of study treatments. Patient study medication packages containing AT or placebo were prepared and sequentially numbered before the start of the trial. The sequence of treatments was determined by random assignment with an allocation ratio of 1:1, using a computer program developed by Hoechst AG (Frankfurt, Germany). Within each of four strata (male or female; birth weight, 750 to 1,249 g or 1,250 to 1,900 g), randomly chosen blocks of 2, 4, 6, or 8 were used; after the first few blocks, the probability for the block size of 2 was increased in order to achieve balance toward the end of the trial. Packages containing lyophilized study medication labeled with the unique patient number were provided to the study center by the manufacturer. Before and after reconstitution, all drug vials and their contents were of identical appearance. Provisions were made for emergency unblinding during the trial; however, no unblinding occurred during the entire study period. The allocation code was released to the study statistician (R.R.) by the company in December 1996, after the database was declared closed.

Intervention

Antithrombin (Kybernin; Behringwerke AG, Marburg, Germany) or placebo (Human Albumin 1%; Behringwerke AG) was reconstituted with sterile water for injection before use. A loading dose of 2 ml per kilogram (equivalent to 100 U/kg of AT) was given intravenously, followed by 1 ml/kg (equivalent to 50 U/kg) every 6 h for 48 h. All study infants received the same exogenous surfactant (Exosurf; Burroughs Wellcome Inc., Montreal, Canada). Arterial catheters were perfused with 1 ml/h of 5% dextrose containing 1 IU/ml of unfractionated heparin. No additional administration of anticoagulants was permitted by the study protocol.

Outcome Assessments

Plasma AT activity, TAT, and F₁₊₂ (Berichrom, Enzygnost TAT, and Enzygnost F₁₊₂; Behringwerke AG) were measured before the first, third, fifth, seventh, and last dose of the study medication. Batch-assaying was performed under the supervision of L.M. and M.A. These two coinvestigators were not involved in any other aspect of the trial. To maintain allocation concealment, laboratory results were not disclosed by these two investigators to anyone until the database was declared closed. The ventilator efficiency index (VEI) and the (a/A)PO₂ were computed every 6 h during the treatment phase and then once daily until Day 7 of life in those infants who remained intubated and who still had an indwelling arterial catheter. The VEI was originally devised by Notter and colleagues (22) as a measure that relates alveolar ventilation to respirator input in the absence of spontaneous breathing. In the present study, VEI was computed as suggested by Kwong and colleagues (23) who used this measure in spontaneously breathing infants during a controlled trial of surfactant replacement. All values that exceeded 1 ml/mm Hg/kg were defined as 1 (23). The duration of mechanical ventilation and of supplemental oxygen were recorded when each therapy had been discontinued for at least 24 h.

Cranial ultrasonography was performed at baseline and again on Day 7 of life in all surviving patients. For the purpose of this study, all images were reviewed by a single radiologist who was also masked to treatment allocations at the time of his report (C.C.).

Statistical Considerations

The mechanistic outcomes of AT, TAT, and F₁₊₂, and the primary efficacy outcomes of (a/A)PO₂ and VEI were each measured immediately before randomization and at multiple time points after randomization. Both types of outcomes were analyzed by, firstly, computing the difference (i.e., change) between the prerandomization level and the mean of postrandomization observations for each patient. F₁₊₂ and TAT were analyzed under a logarithmic transformation because of distinct right-tail distributional skewing. For AT, TAT, and F₁₊₂ the postrandomization means were computed over the 48 h of treatment. For the (a/A)PO₂ and the VEI, means were computed separately for the 48 h treatment period and for the full 7-d postrandomization period. The mean prerandomization to postrandomization change was then compared between active and placebo groups. Because the randomization was stratified into four sex/birth weight categories, the comparisons of mean change were achieved via two-way analysis of variance (24), including treatment and stratum as factors. Patients who died during the study period were included in these analyses by computing the post-treatment mean up to the time of death.

The time to cessation of mechanical ventilation and supplemental oxygen were compared between treatment groups via Kaplan-Meier survival curves (25) and a Mantel-Haenszel test (26), again stratified for sex/birth weight category. Patients who died were included as censored observations at the time of death. All randomized patients were included in the analyses (whenever possible) according to the intention-to-treat principle.

The postrandomization incidence of intraventricular hemorrhage (IVH), periventricular echodensity (PVED), and death were compared between treatment groups via logistic function regression incorporating adjustment for stratum (27).

The External Safety and Efficacy Monitoring Committee conducted a single formal interim analysis of efficacy, based on (a/A)PO₂, using an alpha of 0.002, thus retaining an alpha of 0.048 for the final analysis. The interim analysis was conducted when outcome data were complete for 70 patients. Limited data were available for the determination of sample size. We estimated that a study of 64 patients per group would yield 80% power to detect an increase in postrandomization (a/A)Po₂ of about 0.17 with AT. However, the estimate of the standard deviation to be anticipated in this trial was quite speculative.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

All infants admitted to the NICU with birthweights of 750 to 1900 g were considered for entry into the study and recorded on a screening log (Figure 1). By February 1996, 224 infants had been identified who met the inclusion criteria, 198 of whom had no exclusion criteria present and were thus eligible. Sixty-two percent (n = 122) of these eligible subjects were randomized. The most frequent reason for not randomizing an eligible patient was lack of informed parental consent (50%); some patients were not approached (usually because of an administrative oversight). Randomization was also suspended for a short period in 1994 at the request of the Health Protection Branch (the equivalent of the FDA in Canada) while they investigated a perceived problem in the antithrombin production process.

View larger version (42K): [in this window] [in a new window]   Figure 1.   Trial profile.

After 122 patients had been recruited, randomization was halted, at the request of the External Safety and Efficacy Monitoring Committee, six patients short of the planned sample size. Having observed for some time an apparent imbalance in deaths between the treatment groups, the committee decided to break the code and make its recommendation because seven deaths had occurred in patients receiving AT and two with placebo. The final rates of death before discharge were 7/61 (11.5%) with AT and 3/61 (4.9%) with placebo (adjusted odds ratio, 2.65; 95% CI, 0.64 to 11.02; p = 0.18).

Baseline Comparability

The baseline status of the patients is summarized in Table 1, which shows the two treatment groups to be closely balanced for most important prognostic features. Of particular note is that the therapy was initiated on average at 7.2 h of life in each group.

AT, TAT, and F₁₊₂

The administration of exogenous AT clearly increased plasma AT activity (Figure 2). Against a background of essentially no change in placebo patients, the AT level was already substantially increased at 12 h and continued to rise during the remainder of the 48 h of treatment. Because of the statistical requirement to analyse F₁₊₂ and TAT under a logarithmic transformation, these measures are expressed as a percentage of the baseline level. Both F₁₊₂ and TAT levels decreased during the first treatment day in each group and then maintained a more constant level during the second day (Figure 2). Declines in F₁₊₂ were consistently larger with AT (Figure 2) than with placebo. In contrast, TAT levels did not differ consistently between the two comparison groups (Figure 2). The size and statistical significance of the treatment effect on AT, F₁₊₂, and TAT are summarized in Table 2.

View larger version (19K): [in this window] [in a new window]   Figure 2.   Plasma antithrombin activity (upper panel ), prothrombin fragment (F1+2) (middle panel ), and thrombin-antithrombin complex (TAT) (lower panel ). Values shown are means and 95% CI. Antithrombin (solid line); placebo (dashed line).

Efficacy Outcomes

Mean (a/A)PO₂ increased steadily during treatment in both groups and continued to rise in the remainder of the observation period. After allowing for the slight baseline difference in mean (a/A)PO₂, the two curves in Figure 3 are quite similar and the numerical estimates of treatment effect in Table 2 are close to zero and statistically nonsignificant. VEI exhibited a similar pattern (Figure 3), but early postrandomization increases were not sustained beyond the first treatment day, after which VEI remained basically constant to Day 7. As for the (a/A)PO₂, there was no evidence of a treatment effect on VEI (Table 2). Note that two patients were not included in these efficacy analyses because they had no postrandomization values for (a/A)PO₂ and VEI. One AT patient died very early in the treatment period; the other, a placebo patient, had the arterial line removed just after randomization, and attempts to reestablish arterial access were unsuccessful.

View larger version (24K): [in this window] [in a new window]   Figure 3.   a/A ratio = a/A PO2 (upper panel ), and ventilator efficiency index (VEI) (lower panel ). Values shown are means and 95% CI. Antithrombin (solid line); placebo (dashed line).

On the basis of the observed standard deviations in the data, we estimate that this study would have yielded 80% power to detect a true mean increase of 0.055 for (a/A)PO₂ and 0.10 for VEI. These treatment effects, if present, would have represented increases of 14 and 16%, respectively, in relation to typical placebo values of (a/A)PO₂ and VEI.

The median times to cessation of mechanical ventilation and supplemental oxygen were both significantly longer with AT than with placebo (Table 2). The rates of pulmonary air leaks and patent ductus arteriosus were comparable in the two groups.

Safety Outcomes

The rates of severe (Grade 3) IVH and PVED were comparable in the two groups (Table 3). There was, however, a trend toward a higher combined rate of any IVH or PVED on Day 7, even after adjustment for the presence or absence of intraventricular and periventricular lesions at baseline (Table 3). The odds ratio for IVH or PVED on Day 7, adjusted for birth weight/sex stratum, and the presence/absence of IVH/PVED at baseline was 2.30 (95% CI, 0.86 to 6.16; p = 0.06). Clinically apparent bleeding from puncture sites, the umbilicus, nasogastric tubes, or endotracheal tubes was documented during the first week of life in 37 of 61 treated infants compared with 30 of 61 control infants (adjusted odds ratio, 1.64; 95% CI, 0.77 to 3.50; p = 0.20). Deaths in study infants are summarized in Table 4.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study shows conclusively that the administration of AT concentrate during the first 2 d of life does not improve gas exchange in premature infants with RDS. Moreover, our results suggest that AT therapy may be harmful in this patient population. The duration of mechanical ventilation and oxygen therapy were significantly prolonged, whereas deaths, periventricular and intraventricular hemorrhages, and clinically apparent bleeding episodes, all tended to be more common in infants randomized to AT concentrate. The certain lack of benefit together with the potential harm strongly argue against the routine use of AT concentrate in sick premature infants. The results of this placebo-controlled trial cast doubt on the previous claim (based on nonradomized comparisons) that AT concentrate is effective in inhibiting the progression of intracranial hemorrhage in premature infants (28, 29).

Antithrombin levels in healthy preterm infants are about 40% of normal adult values (30). Even lower levels have been reported in sick premature infants with RDS (3). This acquired AT deficiency may contribute to the activation of the plasma coagulation system in RDS (5) that was also observed in this trial. Study infants had markedly elevated plasma concentrations of two biochemical markers of thrombin formation, F₁₊₂, and TAT. Although a strong association between this hypercoagulability and the severity of RDS does not necessarily establish a pathogenetic role for intrapulmonary thrombin formation and fibrin deposition, evidence from in vitro studies and experimental animals supports the hypothesis that regulation of unopposed thrombin activity may be beneficial during acute lung injury (14, 31). Because acquired AT deficiency is a consistent and prognostically important finding in infants with RDS (5), AT concentrate was an obvious choice for a controlled trial of antithrombotic therapy in neonatal RDS (32).

To the best of our knowledge, only one other controlled trial of antithrombin concentrate has been performed in preterm infants with RDS. This particular trial of antithrombin prophylaxis was published as an abstract only (33), and it was not placebo-controlled. No benefit of AT administration was found. Of note, although the AT dose used in this study was considerably lower than in the present trial, there was a similar trend towards more frequent deaths in the experimental group: nine of 45 infants died in the experimental group compared with eight of 53 in the control group (33).

There are two possible interpretations for the treatment failure of AT concentrate in this patient population. The most radical conclusion might be that thrombin inhibition does not alter the course of neonatal RDS, and, further, that acquired AT deficiency and increased thrombin formation are merely coincidental markers of the disease process. Alternatively, thrombin may indeed play a pathogenetic role in neonatal acute lung injury, but AT concentrate was not sufficiently effective in suppressing unopposed thrombin activity. The laboratory findings in this trial appear to support the latter interpretation. Although the decline of F₁₊₂ levels was greater during treatment with AT than with placebo, mean F₁₊₂ levels remained abnormally high in both groups, even at the end of the 48-h treatment phase, suggesting that AT concentrate did not entirely abolish thrombin formation. Similarly, although improving over time, mean TAT levels were still substantially elevated after 2 d of AT therapy, again reflecting ongoing thrombin formation since the plasma elimination half-life of the TAT complex is only a few minutes (34).

Before, however, future studies with more powerful antithrombotic agents are entertained, possible adverse effects have to be carefully considered in this high-risk population. In contrast to our study hypothesis, we found that the duration of mechanical ventilation and oxygen therapy were prolonged in infants who received AT. The reasons for this surprising result are entirely speculative; the extremely small p values would support the conclusion that this is truly an adverse treatment effect, rather than a chance finding. Although the rates of clinically apparent bleeding from the endotracheal tube were identical in the two groups, we cannot rule out the possibility that alveolar hemorrhage was increased by AT therapy. This hypothesis is supported by the observed trend toward more intracranial bleeding in infants randomized to AT concentrate. It is important to stress that this trial did not have sufficient power to detect clinically important differences in the rates of intracranial hemorrhage. However, in the context of the adverse effects of AT on duration of mechanical ventilation and oxygen therapy, the trend toward increased bleeding is concerning and raises doubts about the safety of more potent thrombin inhibitors in sick premature infants.

Antithrombin concentrates have been commercially available in some countries for more than 15 yr, yet the evidence to justify their use in acquired AT deficiency remains weak (1, 2). The present trial has clearly demonstrated that treatment of acquired AT deficiency associated with neonatal RDS does not improve clinical outcomes. Well-designed controlled trials are now required to determine the safety and efficacy of AT concentrate in other groups of critically ill patients with acquired AT deficiency.