Author + information
- John A. Bittl, MD∗ ()
- ↵∗Reprint requests and correspondence:
Dr. John A. Bittl, 1221 SE 5th Street, Ocala, Florida 34471.
“There’s something about death going on here.”
—Poet Laureate Billy Collins, in “Workshop” (1)
The use of antithrombotic therapy is associated with an inherent tradeoff between increased bleeding and reduced ischemic events. How do clinicians decide which antithrombotic strategy will maximize net clinical benefit?
In this issue of JACC: Cardiovascular Interventions, Secemsky et al. (2) touch on this issue by analyzing what happened in clinical practice when practitioners used bivalirudin in place of unfractionated heparin (UFH) to treat ST-segment elevation myocardial infarction with percutaneous coronary intervention. Although the study did not capture information about ischemic endpoints, like acute stent thrombosis, the investigators found in unadjusted analyses that bivalirudin was associated with decreased bleeding (risk difference: -4.2%; p < 0.0001) and mortality (risk difference: -0.84%; p < 0.001). After statistical adjustment, the investigators found that bivalirudin remained associated with reduced bleeding (risk difference: -3.75%; p < 0.001), but not reduced mortality (risk difference: -0.10%; p < 0.280). Omitting platelet glycoprotein inhibitors with UFH seemed to decrease the bleeding advantage of bivalirudin.
A reviewer faced with such observations could invoke Billy Collins (1) and say, “I wonder what we have here is really two [studies], or three, or four, or possibly none.” But, in fact, the report from Secemsky et al. (2) contains a critical new clue that can be used, perhaps not in the way the investigators intended, to answer the central unsettled question about antithrombotic therapy: What is more grave, bleeding or ischemic events?
To complete their engaging analysis, Secemsky et al. (2) used a sophisticated statistical tool called instrumental variable analysis.† Because statistical analyses in general involve an element of art and different approaches may yield different results (4), the present analysis could give the impression of trying to obscure a mortality advantage of bivalirudin. However, some type of adjustment seems reasonable, because an observational study is subject to confounders caused by imbalances in baseline characteristics that could have given the bivalirudin-treated patients a prognostic advantage.
Randomization of subjects in a trial is a better way to eliminate imbalance than to perform statistical adjustments. It is, therefore, relevant to recall that a well-executed randomized controlled trial, namely the HORIZONS-AMI (Harmonizing Outcomes with Revascularization and Stents in Acute Myocardial Infarction) trial (5), was the setting in which the use of bivalirudin was first found to have lower 30-day mortality than use of UFH (2.1% vs. 3.1%; p = 0.047) in the treatment of ST-segment elevation myocardial infarction with percutaneous coronary intervention. This was associated with reduced bleeding (4.9% vs. 8.3%; p < 0.001), despite higher rates of acute stent thrombosis within the first 24 h with bivalirudin (1.3% vs. 0.3%; p < 0.001). As if to emphasize that replication is the highest form of evidence in clinical investigation, the MATRIX (Minimizing Adverse Hemorrhagic Events by Transradial Access Site and Systemic Implementation of Angiox) trial (6) recently reported lower 30-day mortality with bivalirudin (1.7% vs. 2.3%; p = 0.04), associated with reduced bleeding (1.4% vs. 2.5%; p < 0.001), again despite higher rates of acute stent thrombosis (1.0% vs. 0.6%; p = 0.048). For clinicians trying to choose between the lesser of 2 evils—increased acute stent thrombosis with bivalirudin or increased bleeding with UFH—the mortality signals seen in the broad range of study settings (2,5,6) favor the use of bivalirudin over UFH during percutaneous coronary intervention for ST-segment elevation myocardial infarction.
In a different but related debate, the use of prolonged dual antiplatelet therapy (DAPT) after newer generation drug-eluting implantation has been associated with a similar dilemma: What is worse, increased bleeding with prolonged DAPT or increased ischemic events with shorter DAPT? The tradeoff associated with prolonged DAPT emerged most clearly in the DAPT trial (7), which found that longer courses of therapy were associated with lower rates of stent thrombosis (0.4% vs. 1.4%; p < 0.001), higher rates of bleeding (2.5% vs. 1.6%; p = 0.001), and borderline higher mortality rates (2.0% vs. 1.5%; p = 0.05). Putting the DAPT results (7) into the context of 10 other trials, a systematic analysis confirmed the tradeoff between increased bleeding and reduced ischemic events with prolonged DAPT (8), but using the present line of thinking, identified a persistent mortality signal with prolonged DAPT that tips the balance in favor of shorter courses of DAPT.
Extending the present line of thinking to a another similar debate (9,10) provides a fresh perspective on using extended antiplatelet therapy in patients with prior myocardial infarction. The PEGASUS-TIMI 54 (Prevention of Cardiovascular Events in Patients with Prior Heart Attack Using Ticagrelor Compared with Placebo on a Background of Aspirin–Thrombolysis In Myocardial Infarction 54) trial identified a tradeoff after using the 60 mg of ticagrelor twice daily, which had more desirable benefit–risk profile than the 90-mg regimen for secondary prevention (11). The PEGASUS investigators estimated that for every 1,000 patients treated with the 60-mg regimen of ticagrelor, there would be 4 (95% confidence interval [CI]: 1 to 8) fewer ischemic events but 3 (95% CI: 2 to 4) more major bleeding events/year, as compared with control subjects (11). Faced with such a tradeoff, clinicians can point to the favorable but nonsignificant mortality signal of 2 (95% CI: 0 to 3) fewer deaths/year with the 60-mg regimen as weak evidence for recommending extended DAPT in patients with prior myocardial infarction.
Is mortality a useful outcome for studies not powered for such a measurement? The short answer is “yes.” The “hard” endpoint of all-cause mortality is arguably more precise than the measurement of other endpoints. All-cause mortality is an endpoint less likely to be influenced during endpoint committee adjudication by the lack of blinding than the other “primary” endpoints, and mortality means a lot to patients and families. The present observational study (2) and a raft of randomized, controlled trials (5–7,11), all of which show a tradeoff from antithrombotic therapy, suggest that when a treatment decision is associated with a compromise between bleeding and ischemic events, the primacy of all-cause mortality can help clinicians and patients to choose the best approach.
Every study tells a story. Reviewing the present report (2) inspired me to consider using mortality signals as the tie breaker in a tradeoff. Poet Laureate Billy Collins has captured this subjective sense of discovery by writing (1), “But I don't know if anyone else was feeling that. Maybe that was just me. Maybe that’s just the way I read it.”
↵∗ Editorials published in JACC: Cardiovascular Interventions reflect the views of the authors and do not necessarily represent the views of JACC: Cardiovascular Interventions or the American College of Cardiology.
↵† Such an approach can be summarized by stating: if X and Y are the exposure and outcomes of interest in an observational study, the relation of X and Y can be mathematically related to a third variable Z, called an instrumental variable or instrument, which is associated with X but not associated with Y except through its association with X (3). In the present analysis, the investigators compared X (use of bivalirudin) with Y (in-hospital mortality or bleeding) by using the instrument Z (operator preference for using bivalirudin), but this led to several questions. Is X (use of bivalirudin) in the present analysis actually equivalent to Z (operator preference for bivalirudin)? Is the present analysis simply a comparison of outcomes for patients treated by operators in the highest quartile versus outcomes for patients treated by operators in the lowest quartile of bivalirudin use? The authors responded to these questions by writing that “the present analysis … generates an estimate of the effect of 100% use of bivalirudin vs. 0% bivalirudin use, derived from effect in the top versus bottom quartile of bivalirudin users.” How this transformation occurs remains a mystery.
Dr. Bittl has reported that he has no relationships relevant to the contents of this paper to disclose.
- 2016 American College of Cardiology Foundation
- Collins B.
- Secemsky E.A.,
- Kirtane A.,
- Bangalore S.,
- et al.
- Greenland S.
- Bertsekas D.P.,
- Tsitsiklas J.N.
- Bittl J.A.,
- Baber U.,
- Bradley S.M.,
- Wijeysundera D.N.
- Bittl J.A.,
- Bates E.R.,
- Kaul S.
- Bonaca M.P.,
- Sabatine M.S.