New Tool Helps Balance Bleeding and Thrombotic Risks After PCI

Researchers say their two models are being incorporated into a mobile application that should launch in the next few weeks.

New Tool Helps Balance Bleeding and Thrombotic Risks After PCI

A new tool is in the works to help clinicians balance the risk of bleeds against potential thrombotic events when planning the post-PCI management of patients at high risk for bleeding.

Derived specifically from this high-risk group, two new models—focused on ischemic and bleeding risks, respectively—are being integrated into a soon-to-be-launched application that will allow physicians to input patient information and receive both the patient’s bleeding and thrombotic risk at 1 year, lead investigator Philip Urban, MD (Hôpital de la Tour, Geneva, Switzerland), told TCTMD. “Once [they] see the numbers, it will help clinicians educate [themselves] and also make some decisions in terms of intensity and duration of antiplatelet agents.”

While a number of other risk tools have evolved over the years most of these, the authors note, have been derived from patients at low-to-moderate risk of bleeding.

Urban and colleagues published an analysis supporting their models last week in JAMA Cardiology.

Patients at Higher Post-PCI Risk for Bleeding

For the study, Urban and colleagues included 6,641 PCI patients from six studies conducted at more than 200 centers in Europe, the United States, and Asia between July 2009 and September 2017. All patients were deemed to be high bleeding risk (HBR) per the Academic Research Consortium (ARC) criteria.

At 1 year, 5.3% reported nonperiprocedural MI and/or stent thrombosis and 5.7% reported BARC 3-5 bleeding. The risk of death at 1 year was increased after bleeding (HR 3.7; 95% CI 2.9-4.8) and MI and/or stent thrombosis (HR 6.1; 95% CI 4.8-7.7) compared with patients with no events.

Modelling then allowed the researchers to identify four unique predictors for each event, along with an additional four associated with both endpoints.

Event Predictors

MI and/or Stent Thrombosis

BARC Types 3-5 Bleeding


Prior MI

Age ≥ 65 Years


Medical Treatment of Diabetes

Chronic Obstructive Pulmonary Disease

Kidney Insufficiency


Cancer, Severe Liver Disease, or Planned Surgery

Current Smoking

Use of Bare-Metal Stents

Planned Oral Anticoagulation

Complex PCI Procedure

Predicted rates of the two key endpoints were then compared with the predicted rates using the model and “showed a good model fit,” investigators said. For both the bleeding and the ischemic endpoints, patients in the top quintile by score had more than five times the risk of those in the bottom quintile, and the same relative difference was seen among for observed events in both groups.

Both models showed moderate discrimination, with C-statistics of 0.68 and 0.69 for predicting BARC types 3-5 bleeding and MI and/or stent thrombosis, respectively.

Urban et al then validated their models using 1,498 HBR patients from the Onyx ONE trial. Here, they say, their models discriminated “with a similar strength” but tended, on the one hand, to overestimate the risk of BARC 5- 5 bleeding and, on the other, to underestimate the risk of MI and/or stent thrombosis. Still, when the top and bottom quintiles were compared in the Onyx ONE high-risk cohort, for both outcomes those in the highest quintile had a more than a five times greater risk than patients in the lowest quintile for BARC 3-5 bleeds and for MI and/or stent thrombosis.

‘Not a Stand-alone Model’

As the authors note in the paper, there are other risk tools that have been used to predict bleeding and ischemic risk, including PARIS and PRECISE DAPT. Neither addressed the specific HBR patients studied here, however, and neither performed as well as the model proposed by Urban and colleagues, at least in the validation cohort used.

To TCTMD, Urban acknowledged that in any given patient, there may be too many clinical factors for a doctor to be able to readily calculate a patient’s risks, Urban said. “It's slightly complex, so without an application that does the math for you, you can't really get a feeling as a clinician. But with an application, which we're working on, I think it could be quite useful to try and quantify things a bit.” He expects the app to be available within the next few weeks.

Notably, he added, this tool offers insights on prognosis, though not necessarily what can be done in terms of dual antiplatelet therapy (DAPT) to mitigate risks. “We can't make any definite recommendations for DAPT adjustment based on this because of the patients who are included. Some of them had a DAPT duration that was driven by the protocol of the study [and] some of them had guideline-based DAPT duration, so it's a slight leap of faith to say that we can then improve the outcome by adjusting DAPT based on the numbers we're providing,” Urban explained.

That information may come in time. For now, clinicians still need to make the call as to how to incorporate the model’s findings in a given patient. “Also, there are some quite rare but very important parameters that, because of statistics, don't get into the model,” Urban said. “For instance, if you have severe thrombocytopenia, everyone would agree that that increases your bleeding risk but it is not part of the model because it's rare. . . . It's not a stand-alone model, unfortunately. I honestly don't think it can ever become one and it probably shouldn't.”

  • Urban reports receiving personal fees from Biosensors and Edwards Lifesciences and being a shareholder of MedAlliance and CERC (Center for European Research in Cardiovascular Medicine) outside the submitted work.