Generalizability of DAPT score to real-world populations uncertain

MUNICH — The dual antiplatelet therapy score used to guide decision-making for DAPT duration after PCI poorly discriminated ischemic and bleeding risks in a real-world Swedish population, according to data presented at the European Society of Cardiology Congress.

To investigate the generalizability of the DAPT score in determining appropriate DAPT duration after PCI, Peter Ueda, MD, PhD, from the Karolinska Institutet in Stockholm, and colleagues evaluated the score’s performance in 41,101 patients from the SWEDEHEART registry who continued DAPT beyond 12 months after PCI from 2006 to 2014.

Ischemic outcomes, including MI or stent thrombosis, and MACCE, defined as MI, stroke and all-cause death, were similar in this study and in the overall DAPT trial, from which the DAPT score was derived. However, fatal or major bleeding, which was based on an algorithm developed and validated in Swedish cohort studies, corresponded roughly to Global Use of Strategies to Open Occluded Coronary Arteries (GUSTO) moderate or severe bleeding used in the DAPT study.

When looking at cumulative incidence between 12 and 30 months of DAPT, risk for MI or stent thrombosis was largely similar for patients with DAPT scores from –2 to 2 and significantly increased only in the approximately 20% of patients with scores of 3 or higher. In terms of MACCE, the pattern was J-shaped, with a significantly increased risk in only the approximately 7% of patients with scores of 4 or higher, Ueda said during a presentation at the ESC Congress.

Additionally, the absolute differences in the risk for bleeding were small across levels of DAPT score, he noted, although the group with the highest scores had the numerically highest bleeding risk.

“Provided these data, it is not surprising that the DAPT score had poor discrimination for ischemic outcomes and did not discriminate the bleeding risk in the SWEDEHEART population,” Ueda said.

Poor discrimination, calibration

The DAPT score had Harrell’s C statistics, which was used to determine discrimination, of 0.58 (95% CI, 0.56-0.6) for MI or stent thrombosis, 0.54 (95% CI, 0.53-0.55) for MACCE, 0.49 (95% CI, 0.45-0.53) for fatal or major bleeding events and 0.48 (95% CI, 0.46-0.51) for fatal or major bleeding or bleeding requiring hospitalization, according to the data.

The researchers also assessed calibration by comparing predicted risk and observed risk in groups of patients stratified by DAPT score. As the DAPT score does not produce an absolute risk estimate, Ueda noted, the researchers compared the risk in high- and low-score patients to the placebo arm of the DAPT trial.

In patients with high scores, rates of MI or stent thrombosis were lower in the SWEDHEART population vs. the DAPT trial placebo group (cumulative incidence, 4.5% vs. 5.7%). However, in patients with low scores, the rates were higher in the SWEDHEART population vs. the DAPT trial placebo group (cumulative incidence, 3% vs. 2.3%).

Although rates of MACCE were similar between the SWEDEHEART population and DAPT trial placebo group in patients with high DAPT scores (7.1% vs. 7.6%), rates of MACCE were significantly higher in the SWEDEHEART population than in the DAPT trial placebo group in patients with low scores (5.8% vs. 3.8%).

Ueda said the rates of fatal or major bleeding in the SWEDEHEART population was roughly half that of GUSTO moderate or severe bleeding in the DAPT trial placebo group in both patients with high scores (0.7% vs. 1.4%) and those with low scores (0.8% vs. 1.4%).

Unanswered questions

Although guidelines recommend the DAPT score as a decision-making tool for determining DAPT duration, Ueda said that these findings indicate that the DAPT score and its decision rule have questionable generalizability, at least with regard to the SWEDEHEART population.

In general, there is still uncertainty around reporting and interpretation of risk scores overall, particularly in terms of calibration, although discrimination receives much more attention, he noted.

“But it also has to be said that it is easier to show that something doesn’t work than it is to develop something that actually does, so questions remain,” he said. “Provided the issues highlighted here, is the DAPT score still better than clinical judgment? Can we develop better scores? Is it even possible to develop scores that are generalizable across populations worldwide or, given what we know about CV risk prediction today, should we develop local scores for each population? Should we devise scores that are intentionally designed to be recalibrated? Perhaps we should change our approach and focus more on modeling effect modification rather than risk without treatment.”

John A. Bittl, MD, interventional cardiologist at Munroe Regional Medical Center in Ocala, Florida, discussed potential reasons why the DAPT score performed less well in the real-world population in this external validation study and how the DAPT score fits in with other statistical models. The new data represent “a swing and a miss for the DAPT score,” according to Bittl.

Bittl noted that the findings of this study are “a reminder that clinical decision rules must be held to a high standard. Doctors should rest assured that even the best mathematical models and prediction rules cannot replace clinical judgement.” – by Melissa Foster

References:

Ueda P, et al. Advances in Science 1400. Presented at: European Society of Cardiology Congress; Aug. 25-29, 2018; Munich.

Bittl JA. J Am Coll Cardiol. 2018;doi:10.1016/j.jacc.2018.06.024.

Ueda P, et al. J Am Coll Cardiol. 2018;doi:10.1016/j.jacc.2018.06.023.

Disclosures: Ueda and Bittl report no relevant financial disclosures.