Unclear benefits vs. harms
The numbers presented are from the best studies that are currently available. Some of these studies will NEVER be repeated and so this is all we'll ever have to go on. There will be continued study in some areas and we aim to incorporate this forward into our site. We're constantly monitoring the literature for updates (if you think there is something we've missed, email us!). The conclusions we draw are a best estimate, folks. We've presented what we think is the closest thing to the truth about this intervention, but our data is only as good as the studies that underlie it — and often, the studies aren't as complete or as good as we'd like. We present one number here for the NNT, but please realize this is an estimate and there is a range for what this intervention can offer a given person. That range will depend upon the person's spectrum of disease (mild/moderate/severe), their demographic, their subtype of disease, the setting of the intervention, their general health, and literally thousands of other variables. Using these numbers in practice means taking a number of large leaps about all of these variables, and also about the veracity of the underlying research. Therefore, as with any 'high quality' data, the application of data requires a doctor's expertise and deliberate consideration.
In Summary, for those who got the anticoagulation:
- 100% saw no benefit
- 0% were helped by being saved from death
- 0.9% were harmed by a major bleeding event
- 0.9% were harmed by dying due to the bleeding
In Other Words:
- None were helped (life saved, preventing pulmonary embolism)
- 1 in 50-111 were harmed (major bleeding event)
- 1 in >50 were harmed (death from bleeding event)
Where We Get The Numbers:
Efficacy Endpoints: Death, Recurrent PE, Recurrent/Extension of DVT
Harm Endpoints: Major hemorrhage (intracerebral hemorrhage, GI bleed). Risk varies depending on population. In one Cochrane review of trial data major bleeding occurred in 0.9% to 2% of subjects, a NNH of 50 to 111 (Hutten BA), and there were no major hemorrhages in the Nielsen study of 90 ambulatory outpatients. However subjects eligible for randomized trials tend to be at lower risk for bleeding than typical coumadin patients in the community. In one population-based cohort of coumadin patients 2% experienced a fatal bleeding event (NNH=50) and the overall major bleeding rate was 12% (NNH=8). (Landefeld CS et al. Am J Med 1989). It is estimated that 15000-52000 patients die annually in the U.S. of brain bleeding caused by coumadin.
Narrative: There has been one adequately designed study addressing whether anticoagulation improves important clinical outcomes in patients with VTE (Nielson et al., 1994). Subjects were enrolled if they were ambulatory with symptomatic DVT. Half of the 90 subjects also had silent pulmonary emboli. Subjects were randomized to anticoagulation or NSAIDs and there was 1 fatal PE during the study period (anticoagulation group).
In 1960, Barritt and Jordan performed a randomized trial evaluating anticoagulation for presumed PE. While the study is often cited as evidence of the benefit of anticoagulation, methodologic weaknesses render it unreliable (the Cochrane group excludes the study). Patients with suspected PE were included and no confirmatory testing was performed; randomization was not described; no blinding and no placebo were used; all subjects were inpatients for other reasons and all were critically ill at enrollment; the study was stopped inappropriately early. The results of this study appear to have led the medical community to erroneously conclude that anticoagulation is proven beneficial for VTE.
Caveats: While the Nielson study does not prove that anticoagulation is unhelpful for VTE it is our best and only methodologically reliable source. Large, properly designed trials are needed.
Author: Ashley Shreves, MD
Published/Updated: November 28, 2009
