New ABCD-GENE Score Taps Clinical, Genetic Factors to ID Clopidogrel Nonresponders

Researchers have developed and validated a novel score combining both genetic and clinical factors to identify patients with high platelet reactivity (HPR) on clopidogrel and who are at increased risk for adverse ischemic events following acute MI.

The ABCD-GENE (age, body mass index, chronic kidney disease, diabetes, and genotyping) score incorporates five independent predictors of HPR as identified through data from 714 patients enrolled in the GRAVITAS study who had CYP2C19 genetic testing performed. Researchers led by Dominick Angiolillo, MD, PhD (University of Florida College of Medicine, Jacksonville), validated the score using the dual antiplatelet therapy (DAPT)-treated populations from the POPULAR and FAST-MI registries.

Although the field has struggled to show the value of genetic testing for this purpose in the past, the goal of having a score such as this would be to ultimately use it to tailor therapy for post-MI patients, especially in those who may be clopidogrel nonresponders, Angiolillo explained to TCTMD.

“One of the problems with genotyping is that genotypes don’t explain all the reasons for why a patient may be a clopidogrel nonresponder,” he said. “There are genetic factors, but there are also nongenetic factors such as clinical factors that contribute to clopidogrel response. So on this background, we thought: why not come up with a scoring system that can improve the accuracy with which we define clopidogrel nonresponse by using both clinical and genetic factors?”

Two recent positive trials, TROPICAL-ACS and POPular Genetics, have lent encouragement to those who have been advocating for the potential of platelet function and genetic testing to lead de-escalation of P2Y12 inhibitors in selected patients. Yet many have been awaiting results from the prospective, randomized TAILOR-PCI trial—scheduled to be presented in a late-breaking clinical trial session at ACC 2020 later this month—to have a more definitive answer.

ABCD-GENE Score Creation

Published in the March 9, 2020, issue of JACC: Cardiovascular Interventions, the paper by Angiolillo and colleagues shows a C-statistic for the combined clinical factors included in the ABCD-GENE score alone to be 0.63. However, after inclusion of the genetic factors, this figure rose to 0.71 and 0.64 using the derivation and validation data sets, respectively.

Moreover, the score showed a gradient in the distribution of HPR (P < 0.001 for trend). Using a score cutoff of ≥ 10—determined to have the best sensitivity and specificity—more patients had HPR with scores above the threshold at 30 days than those below in both the derivation cohort (75.4% vs 43.2%; P < 0.001) and the validation cohort (73.7% vs 49.9%; P < 0.001).

Again for both cohorts, the score showed a high specificity (81% and 92%) and positive predictive value (75.4% and 73.7%) but a lower sensitivity (47.9% and 18.3%) for the ABCD-GENE score.

C-statistics for the score were 0.67 and 0.66 for the respective endpoints of all-cause death and the composite endpoint of all-cause death, stroke, or MI at 1 year. The ABCD-GENE score was independently associated with death and the composite endpoint at 1 year on all multivariate models tested with the exception of the analysis restricted to 2-day survivors, which trended toward significance (P = 0.08).

Finally, the score did not predict bleeding either as a continuous variable or using the cutoff value of ≥ 10.

Highlighting the Utility

Angiolillo acknowledged that while the ABCD-GENE score has similar limitations to the PRECISE-DAPT or DAPT scores, it’s important to highlight its utility. “The utility of this score is to argue against de-escalation if you have a high score, [where] you should consider agents other than clopidogrel,” he said. “The starting point is that, for example, if you come in with an acute coronary syndrome, you should follow a guideline-recommended therapy and be using one of the more potent P2Y12 inhibitors. Since de-escalation is something that is happening in real-world clinical practice, [often] in a nonguided fashion, we may be exposing patients to increased risk of ischemic events, including mortality, if patients are not optimally responding to clopidogrel. Here we have, in my opinion, a very simple tool that can help identify who are these patients.”

Ultimately, patients with a high ABCD-GENE score should not undergo de-escalation, Angiolillo argued.

Because of the availability of rapid bedside assays for the CYP2C19 loss-of-function alleles, calculating a score like this with available clinical factors is practical and reasonable for physicians, he pointed out.

“Now that we have identified the score and made it user-friendly, it would be nice to validate this score in other patient cohorts but also test this score prospectively,” Angiolillo said. “The score may have more relevance as we move forward with data supporting that using platelet function testing, genetic testing, or genetic testing integrated with clinical variables to help tailor antiplatelet therapies.”

Why not come up with a scoring system that can improve the accuracy with which we define clopidogrel nonresponse by using both clinical and genetic factors? Dominick Angiolillo

Looking forward to the TAILOR-PCI data, he guessed that the findings will have implications regardless. “If TAILOR-PCI were to be a positive study, it further supports that we should be tailoring therapy, and maybe our score can further refine those patients in whom we are tailoring,” Angiolillo said. “If TAILOR-PCI were to be negative, one would say that maybe the tool that was used did not allow to select patients who could benefit most from a specific antiplatelet agent. So, I could see it both ways.”

Additionally, if the TAILOR-PCI data are in fact positive, he added that it would be “impossible for guidelines not to at this point acknowledge the wealth of data that have accumulated with three trials.”

Prospective Evaluation Needed

In an accompanying editorial, John A. Bittl, MD (AdventHealth Ocala, FL), writes that “the bad news is that most experts believe that C-statistics in the range of 0.51 to 0.70 represent inadequate discrimination. To its credit, however, the ABCD-GENE score was tested more rigorously than most of the 350 other prediction rules in cardiovascular medicine.”

He calls the proposal to use the score to de-escalate treatment “propitious” and examined its potential utility. “Patients prefer clopidogrel for several reasons. A recent study found that cost and side effects were the most common reasons that adherence was higher for clopidogrel than for prasugrel and ticagrelor. In another analysis, investigators found that patients ≥ 75 years of age with acute coronary syndromes switched from clopidogrel only rarely but switched from ticagrelor 42% of the time, mainly because of dyspnea,” Bittl writes.

Additionally, he notes that switching medications “can be hard” for patients and de-escalation protocols require vigilant maintenance for both patients and physicians, not to mention extra cost.

“Testing genotypes and creating risk models have been hot areas of cardiovascular investigation,” Bittl allows. “The ABCD-GENE score used both approaches to reassure doctors that it is safe to use clopidogrel, but the test is not ready for widespread use. On a practical level, genotype testing is not routinely performed before intervention to provide guidance of P2Y12 inhibitor therapy. From a scientific viewpoint, the findings in the present study prove again that a genotype does not map one to one with a trait.”

More proof is needed that the ABCD-GENE score can accurately make treatment decisions in individual patients,” he concludes. “However, there is a clear way forward for the ABCD-GENE score. The next step will be prospective validation by an independent research group in a separate cohort of patients.”

TAILOR-PCI PI Perspective

Commenting on the study for TCTMD, TAILOR-PCI lead investigator Naveen Pereira, MD (Mayo Clinic, Rochester, MN), said the ABCD-GENE score shows the “incremental value” of combining both genetic and clinical factors. “My impression overall is the score helps risk stratify patients better and it makes sense,” he said.

Notably, “even for high platelet reactivity, it appears that the score has a low sensitivity, so it can miss people, and a high specificity,” Pereira said. “So the actual application, as the authors suggest, would be limited to not determining upfront as to who should get what antiplatelet drug, but perhaps use it in a strategy of de-escalating antiplatelet drug therapy.”

He agreed that the ABCD-GENE score should be validated in a prospective clinical trial. One idea would be to use the TAILOR-PCI cohort, given that nearly all patients had CYP2C19 genotyping performed. “We won't have high platelet reactivity information, but ultimately, there is some controversy about high platelet reactivity and clinical outcomes,” Pereira said. “What's important to the patient and practitioner is clinical outcomes, and so TAILOR PCI will be able to provide the clinical outcomes, the genotype information, the clinical variables that were used in the score and would be useful to validate this score in a prospective clinical trial.”

Another important outstanding issue will be seeing if changing treatment based on a high score will change clinical outcomes, he said. Once the TAILOR-PCI results are out, a post hoc analysis could be conducted to determine if “attenuating therapy made a difference in patients with a high score in one arm versus not attenuating therapy—that's continuing clopidogrel—in patients with a high score in the other arm,” Pereira suggested.

Going forward, “determining high risk patients is going to be key,” according to Pereira. With today’s low event rates, “stratification becomes really important in deciding whether you really want to start a more powerful alternative P2Y12 inhibitor like ticagrelor or prasugrel. I think this is an effort to stratify patients and we need to continue in this effort rather than just consider one treatment for all. . . . We hope eventually to use machine learning techniques using clinical and biological data to risk stratify patients. I think that's where the future lies, ultimately.”