|Home | About | Journals | Submit | Contact Us | Français|
Evidence shows that it does not improve outcomes and increases the risk of bleeding
Antithrombin III, first described in 1939 as a cofactor of heparin, is one of the most important physiological inhibitors of coagulation.1 Absence of this cofactor is regarded as incompatible with life, and acquired deficiency—for example, in sepsis—is associated with a high risk of venous thrombosis. In the 1960s researchers found a link between coagulation abnormalities and infection,2 and the anti-inflammatory characteristics of antithrombin III were reported more recently.3 These discoveries have helped us understand how sepsis develops. In the past 15 years, several clinical trials have investigated whether giving antithrombin III to patients who are deficient in this factor—such as those with sepsis, pre-eclampsia, and traumatic brain injury—improves outcomes. Overall, it had no effect on mortality, although it did improve secondary end points in some trials.
In their systematic review in this week’s BMJ, Afshari and colleagues assess the effects of giving antithrombin III to critically ill patients.4 They reviewed 20 randomised controlled trials and found no significant difference in mortality between people given antithrombin III and those given placebo or no intervention (relative risk 0.96, 95% confidence interval 0.89 to 1.03). Antithrombin III significantly increased the risk of bleeding events (1.52, 1.30 to 1.78).
Unfortunately, the authors did not define “critically ill” in detail (for example, by the amount of intensive care needed); patients were loosely defined as critically ill in the included trials. This may have led to a heterogeneous group of trials with mortality ranging from zero—for example, in women with pre-eclampsia—to almost 50% in patients with sepsis. Interestingly, trials of patients with myocardial infarction receiving antithrombin III were not included for unspecified reasons. Therefore, the biological plausibility of some of the included trials and the generalisability of the meta-analysis is uncertain.5 Even though statistical heterogeneity between trials may be negligible, confidence intervals of the I2 value of 0%, which are essential to assess the extent of heterogeneity, were not reported.6
Pooling trials of poor methodological quality in meta-analyses may introduce bias. Consequently, the authors classified the methodological quality of included trials into whether they had a low or high risk of bias. Although they accounted for important factors—how the allocation sequence was generated, whether allocation was concealed, whether the trial was blinded, and whether an intent to treat analysis was used—a validated measure such as the Jadad scale would have been helpful.7
The largest trial (2314 patients) contributed a relative weight of 80% to the meta-analysis, dominating the results of the meta-analysis. None of the forest plots deviated from the findings of this single large trial.8 Afshari and colleagues applied the method of trial sequential analysis to extrapolate the sample size needed to demonstrate or reject an a priori effect of the intervention on mortality. Thus, they calculated that 14294 patients would be needed to detect a 5% relative risk reduction (mortality and relative risk reduction in trials with low bias risk), but the meta-analysis included only 3458 patients. Their calculation may be inaccurate, however, because of heterogeneity in the definition of critically ill.
Subgroup analysis in Afshari and colleagues’ review indicated that survival was better when antithrombin III was given without concomitant heparin. Other retrospective analyses have indicated that antithrombin III given without heparin may reduce mortality in patients with severe disseminated intravascular coagulation.9 A randomised controlled trial in patients with severe sepsis and disseminated intravascular coagulation that has adequate power to compare antithrombin III with and without heparin would therefore be useful.
The interaction between antithrombin III and heparin is poorly understood. Although heparin increases anticoagulatory activity when bound to antithrombin III, we know little about the modulation of its anti-inflammatory properties under these circumstances. Before a large clinical trial is started, a more in-depth analysis of the pharmacokinetics of antithrombin III in critically ill patients is needed; this should include an assessment of the dose-response relation between antithrombin III and heparin.10 Unfractionated heparin on its own reduced mortality in sepsis under experimental conditions,11 and it is being evaluated as a single immunomodulatory anticoagulant in ongoing clinical trials.12
Why has antithrombin III failed in clinical trials so far? Its lack of effect may be a true finding. Alternatively, it may be the result of interactions with other drugs, such as heparin, or the result of methodological limitations in terms of patient selection or classification.
Afshari and colleagues’ review4 is currently the most comprehensive summary of the use of antithrombin III in critically ill patients. Despite some minor limitations, the conclusion that antithrombin III cannot be recommended in critically ill patients is sound.
This article was published on bmj.com on 23 November 2007
Competing interests: None declared.
Provenance and peer review: Commissioned; not externally peer reviewed.