Researchers have developed and validated a new clinical tool to better identify patients who are at risk of bleeding after transcatheter aortic valve replacement (TAVR).
PREDICT-TAVR is the first model specifically developed and validated to identify patients at risk of bleeding after TAVR, said the authors, led by Eliano Pio Navarese, MD, PhD, from the Interventional Cardiology and Cardiovascular Medicine Research Center at Nicolaus Copernicus University, Poland.
Published online Monday, ahead of its appearance in the June 14 issue of JACC: Cardiovascular Interventions, the development and validation of the new model is a joint effort of the Italian and Polish Cardiac Interventional Societies.
Navarese and colleagues noted that while recent advances in TAVR technology have improved outcomes for patients with severe aortic valve stenosis, bleeding events after the procedure remain common in this elderly and comorbid population.
“Importantly, such events may be preventable,” said the team, adding that the rate of bleeding complications is reported to be 3% to 11% within the first year of TAVR, with most episodes occurring early.
“Bleeding events after TAVR influence prognosis and quality of life and may be preventable,” they noted. However, to date, only a few relatively small studies have applied pre-existing bleeding risk scores to TAVR patients.
Using machine learning and multivariate regression, Navarese and colleagues analyzed more than 100 clinical variables from 5,185 consecutive patients undergoing TAVR in the prospective multicenter RISPEVA registry. The algorithm created through this machine learning and multivariate regression process was then tested and externally validated in 5,043 TAVR patients from the prospective, multicenter POL-TAVI database.
The process generated a six-item score – subsequently called PREDICT-TAVR – which is composed of blood hemoglobin and serum iron concentrations, oral anticoagulation and dual antiplatelet therapy (DAPT), common femoral artery diameter, and creatinine clearance.
The overall PREDICT-TAVR Score assigns various points to each of these six criteria, with a total of 25 points available. The team reported that scores of 8 points or lower were classified as “Low” risk, while those of 10 or lower were “Moderate” risk, 12 and lower were at “High” risk, and those with a score above 12 were classed as “Very High” risk.
Bleeding events across these risk groups were 0.8%, 1.1%, 2.5%, and 8.5%, respectively, with an exponential rise for patients assigned to the top quartile (overall p < 0.001).
Compared with the lowest quartile, bleeding risk increased numerically for the second quartile (odds ratio [OR]: 1.75; 95% confidence interval [CI]: 0.73–4.19; p = 0.20) and significantly doubled, or more, for the third-quartile (OR: 2.0; 95% CI: 1.39–3.02; p < 0.001) and fourth-quartile (OR: 2.49; 95% CI: 1.96–3.17; p <0.001) patients, said the team.
The team concluded that PREDICT-TAVR is a practical, validated, six-item tool to identify patients at risk of bleeding post-TAVR that can assist in decision making and event prevention, adding that the web-based calculator tool can be used at an early stage both to predict bleeding risk if DAPT is chosen, and to prompt actionable measures to mitigate risk.
“The PREDICT-TAVR score may have an impact on clinical practice, with high-risk scores prompting the choice of single- rather than dual antithrombotic therapy, as well as renal preservation strategies, and raising the important question of whether treating pre-TAVR anemia and iron deficiencies may affect outcomes,” said the authors.
A Step Forward
Writing in an accompanying editorial, Sunil Rao, MD, and Zachary Wegermann, MD, from Duke University Medical Center, noted that the unique use of robust machine learning techniques to supplement statistical modeling has helped to create a “simplified and parsimonious” bleeding risk prediction tool.
The editorialists added that although the PREDICT-TAVR model performs well in identifying patients at risk of post-TAVR bleeding events, further research is needed to determine whether the tool can be implemented prospectively to identify and target modifiable factors to reduce post-TAVR bleeding events, as has been successfully done with post-PCI bleeding risk scores.
“PREDICT-TAVR risk model is an important first step toward predicting post-TAVR bleeding events, but work remains to define the path toward patient-centered and individualized bleeding reduction programs in TAVR,” they concluded.
Navarese EP, Zhang Z, Kubica J, et al. Development and Validation of a Practical Model to Identify Patients at Risk of Bleeding After TAVR. JACC Cardiovasc Interv 2021;14:1196–206.
Rao SV, Wegermann ZK. Quo Vadis, Bleeding Risk Models?. JACC Cardiovasc Interv 2021;14:1207-8.