New Equations Help ID Patients at Low Risk for Bleeding With Aspirin

A series of randomized clinical trials have taken the shine off aspirin for the primary prevention of cardiovascular disease, but the creators of new sex-specific bleeding-risk equations believe they could be used to identify patients most likely to benefit from treatment.

The models, developed by Vanessa Selak, MBChB, PhD (University of Auckland, New Zealand), and colleagues, utilize risk factors and patient characteristics known to predict bleeding in men and women, including age, smoking status, and diabetes, among others, and can be used to assess the relative hazards of aspirin therapy, their study suggests.

“In order to determine the balance of benefits and harms of aspirin in primary prevention, there’s a need to know an individual’s risk of CVD and their risk of a major bleed without aspirin,” Selak told TCTMD. “We have lots of equations than can be used to determine, among people considering aspirin for primary prevention, an individual’s risk of CVD, but few bleeding-risk equations that can be used to determine their risk of a major bleed.”

Late in 2018, three randomized trials showed disappointing results for prophylactic aspirin use in various patient populations: ASPREE in elderly patients, ARRIVE in those at moderate risk for disease, and ASCEND in patients with diabetes. Just a few weeks ago, a large meta-analysis incorporating these three trials showed aspirin does reduce the risk of cardiovascular events but that the benefit is largely canceled out by the increased risk of bleeding.

While widespread aspirin use is not recommended—even the United States Preventive Services Task Force (USPTF) tightened up their recommendations to state aspirin should only be use in primary-prevention patients at high risk of CVD—experts have stated that the decision to start aspirin needs to be tailored to each individual. This is where the problem lies, according to the researchers, given that there are few prognostic models for bleeding risk available to physicians for estimating its likely harms.  

“The question that remains is: given the benefits and harms of aspirin, in whom might the benefits of aspirin outweigh its harms?” said Selak.   

PREDICT: Primary Prevention in New Zealand

Published February 25, 2019, in the Annals of Internal Medicine, the sex-specific models for risk prediction are based on major bleeding events in the large, prospective PREDICT study of adults receiving primary care in New Zealand between 2002 and 2016. The cohort included 169,053 women (average age 56 years) and 216,138 men (average age 51 years); 55% identified as European, 13% as Māori, 12% as Pacific, 9% as Indian, and 11% as Chinese or other Asian ethnicity.

There were 4,442 major bleeding events, of which 7% were fatal, during 1,619,846 person-years of follow-up. The majority of events were gastrointestinal bleeds (69%). Of the fatal bleeding events, 57% were intracerebral hemorrhages.

The adjusted hazard ratios for major bleeding were similar in men and women, except among men who identified as Chinese or other Asian ethnicity, who had a higher risk of major bleeding compared with European men (there was no difference among women). Older age was associated with a higher bleeding risk, as was socioeconomic deprivation and smoking. Māori and Pacific people had a higher risk of bleeding compared with Europeans, but Indians did not. Other established bleeding risk factors, such as cancer, prior bleeding, alcohol-related conditions, and chronic liver disease, were also associated with a higher bleeding risk. Nonsteroidal anti-inflammatory use was associated with a higher bleeding risk in men, while peptic ulcer disease was linked with more bleeding in women.

Overall, the main models predicted a median 5-year bleeding risk of 1.0% in women and 1.1% in men. According to the investigators, “plots of predicted against actual event rates showed good calibration throughout the risk range in the models for women and men.” In terms of performance, the C-statistic, which measures the concordance between a model-based risk estimate and observed events, was 0.68 for women and 0.70 for men, which is considered a good fit. The under- and overprediction of bleeding risk did not exceed 0.2% at any decile of predicted risk.

To TCTMD, Selak said their next step is the development of a tool than can integrate the benefits and harms of aspirin for primary prevention in a single risk calculator, which they hope can be used to identify patients in whom the benefits of treatment are likely to outweigh the risks. Additionally, the risk equations need to be validated in patients outside New Zealand. In the meantime, though, Selak believes the models provide an estimate of risk among New Zealanders in whom aspirin is being considered.

J. William McEvoy, MBBCh (National University of Ireland, Galway), who was not involved in the study, said that while the US guidelines have previously recommended aspirin for primary prevention in patients with sufficiently elevated risk, the European guidelines no longer recommend its use, advice that is supported by the ASPREE, ARRIVE, and ASCEND trials. However, while bleeding remains a problem, the totality of data still show aspirin reduces nonfatal CVD events, including MI, suggesting there might be a role in selected high-risk patients.

For that reason, the development of a risk score for bleeding, something that has been tried in the past, might be “extremely useful,” said McEvoy.

The general problem “is that many of the risk factors for bleeding on aspirin are also the same risk factors for CVD events (eg, age, smoking, diabetes), so these scores have not been easy to apply clinically and have not been widely used because they often don’t distinguish the group that are higher risk for bleeding and lower risk for CVD (those who wouldn’t need aspirin),” McEvoy told TCTMD in an email.

Instead, the risk scores tend to identify patients at high risk for bleeding and CVD. “It’s more complicated as to whether or not aspirin might have a role in these folks, as most patients would prefer to avoid a stroke or MI than to avoid a bleed,” said McEvoy.

Confirming Previously Identified Risk Factors

In an editorial, Evelyn Whitlock, MD (Patient-Centered Outcomes Research Institute, Washington, DC), and Eric Johnson, PhD (Kaiser Permanente, Portland, OR), note that the models developed by the researchers confirm risk factors previously shown to increase bleeding risk in men and women. They also believe one of the next steps will be external validation in relevant populations outside New Zealand to assess the models’ “transportability to different clinical case mixes and data sources.”

“However, any model that requires clinicians to determine 19 patient characteristics and enter them into a risk calculator presents a barrier to routine clinical use in primary care,” caution Whitlock and Johnson. Such a barrier could be reduced by integrating the models with a population-relevant assessment of CVD risk and autoscoring through electronic medical records, they say.

As the bleeding-risk equations are further improved to ensure clinical utility and applicability, the PREDICT cohort data represent a chance for the USPSTF and others to update treatment decision models, according to the editorialists.