Mode
Text Size
Log in / Sign up

AI-Based Adjudication of MACE Events Shows High Agreement with Human CEC in MI TrialAI Can Spot Heart Attacks and Strokes in Medical Records—But Should We Trust It?

AI-generated summary of the cited source, checked by automated accuracy review. How we work

Key Takeaway
Consider AI-based MACE adjudication as a potential adjunct to human CEC review, but not a replacement due to lower agreement for all events.

This study evaluated an artificial intelligence-based adjudication system, Auto-MACE, for major adverse cardiovascular events (MACE) in a global randomized trial of 5,661 patients with myocardial infarction (MI) complicated by systolic dysfunction or pulmonary congestion. The trial compared sacubitril/valsartan versus ramipril, and the primary outcome was agreement between Auto-MACE and physician clinical events committee (CEC) adjudication for MACE events. Auto-MACE uses an OpenAI o1-mini language model and a Clinical Longformer model to classify events.

For the primary analysis, Auto-MACE achieved confident adjudication for 315 of 455 deaths (69%), 301 of 659 potential MIs (46%), and 136 of 167 potential strokes (81%). Among these confident cases, agreement with CEC adjudication was high: 97% for deaths, 89% for potential MIs, and 88% for potential strokes. When considering all events (including those where the model was not confident), agreement was lower: 86% for deaths, 76% for potential MIs, and 84% for potential strokes.

Secondary outcomes included the estimated treatment effect of sacubitril/valsartan versus ramipril on the composite MACE endpoint. Using Auto-MACE, the hazard ratio was 0.91 (95% CI: 0.78-1.07), while using CEC adjudication, the hazard ratio was 0.90 (95% CI: 0.77-1.05). These results are very similar, suggesting that AI-based adjudication could yield comparable treatment effect estimates.

Safety and tolerability data were not reported in this analysis. The study did not provide details on adverse events, serious adverse events, or discontinuations. The focus was solely on the performance of the AI adjudication system compared to the standard CEC process.

Compared to prior landmark studies, this is one of the first large-scale evaluations of AI-based adjudication in a cardiovascular outcomes trial. Previous studies have used traditional CEC adjudication as the gold standard, and this study attempts to validate an automated alternative. The high agreement for confident events is encouraging, but the lower agreement for all events highlights the need for human oversight.

Key methodological limitations include the lack of reporting on the specific training data for the AI models, potential selection bias in which events were deemed confident, and the absence of a prospective validation in a separate trial. The study also did not report the time or cost savings associated with Auto-MACE, which would be important for practical implementation.

Clinically, these results suggest that AI-based adjudication could potentially reduce the workload of CECs by handling a subset of events with high confidence, while uncertain events still require human review. However, the lower agreement for all events (especially MIs) means that full replacement of human adjudication is not yet supported.

Unanswered questions include the generalizability of Auto-MACE to other trial populations, the impact on trial timelines and costs, and the optimal threshold for confident adjudication. Further research is needed to validate these findings in prospective settings and to assess the clinical and operational implications.

The Human Cost of Checking Records

Imagine a doctor spending weeks reading through hundreds of pages of medical records. They are looking for one specific thing: did a patient have a heart attack or stroke during a study? This process is called "adjudication," and it is slow, expensive, and exhausting.

But what if a computer could do the first pass?

A new study published in the Journal of the American College of Cardiology tested an AI tool designed to do exactly that. The goal is to help researchers review data faster without sacrificing accuracy.

Major adverse cardiovascular events (MACE) are the gold standard in heart research. These include cardiovascular death, nonfatal heart attacks (myocardial infarction), and nonfatal strokes.

When a patient in a clinical trial has a potential heart event, a committee of doctors must review the raw data. They look at lab results, doctor’s notes, and scans to decide if it truly counts as a heart attack.

This process is vital but incredibly labor-intensive. In large global trials involving thousands of patients, the workload is massive.

Current methods can create bottlenecks. Delays in reviewing records can slow down the release of life-saving treatments. Researchers are constantly looking for ways to make this process more efficient without cutting corners.

Traditionally, this work is done entirely by human experts. A Clinical Events Committee (CEC) reviews every potential event. This is the current standard because human judgment is nuanced.

But humans are not perfect. They can get tired, and different reviewers might interpret the same notes slightly differently. Plus, it takes a long time.

The new way uses Artificial Intelligence (AI) to do the heavy lifting. The AI reads the medical records and makes a "confident" or "uncertain" judgment.

Here’s the twist: the AI doesn’t replace the doctors. Instead, it acts as a powerful assistant. It handles the clear-cut cases so the human experts can focus on the tricky ones.

How It Works: The Digital Detective

Think of this AI like a spell-checker for medical events.

When you type a document, the spell-checker instantly flags obvious errors. You don’t need to read every word yourself; you trust the tool to catch the mistakes. You only focus on the words the tool flags as "uncertain."

This AI system, called "Auto-MACE," works similarly. It uses two main parts:

1. The Reader: It uses a language model (like a very advanced version of ChatGPT) to read the clinical trial notes. It looks for keywords and patterns that indicate a heart attack or stroke. 2. The Confidence Scorer: A second AI model assigns a confidence level to each event. It tells the doctors, "I am 99% sure this is a heart attack," or "I’m not sure, I need a human to look at this."

This creates a "triage" system. The AI handles the easy cases, and the humans handle the hard ones.

Researchers tested this AI tool on a massive global trial called PARADISE-MI. This trial compared two heart medications (sacubitril/valsartan vs. ramipril) in 5,661 patients who had suffered a heart attack.

The researchers fed the medical records of these patients into the Auto-MACE system. They then compared the AI’s findings against the original decisions made by the human doctor committee.

The results showed that the AI was a reliable partner, especially when it was confident.

For the events where the AI was "confident," it agreed with the human doctors almost perfectly:

  • Cardiovascular Death: 97% agreement
  • Stroke: 88% agreement
  • Heart Attack: 89% agreement

When looking at all events (including the ones the AI was unsure about), the agreement rates were still strong: 86% for death, 84% for stroke, and 76% for heart attacks.

Most importantly, the AI reached the same medical conclusion as the humans. When comparing the two heart drugs, both the AI and the human doctors found that they had a similar effect on preventing MACE.

But there’s a catch.

The AI was not perfect. It struggled more with heart attacks than with strokes or deaths. Medical records for heart attacks can be complex, involving specific enzyme levels and timing that are sometimes ambiguous.

This doesn’t mean this treatment is available yet.

The study authors suggest a hybrid approach. Instead of having doctors review every single record, they could use the AI to adjudicate the "confident" events immediately. The doctors would then only need to review the cases where the AI expressed uncertainty.

This could significantly reduce the workload. In this study, the AI provided a confident adjudication for 69% of deaths and 81% of strokes. That means doctors wouldn't have to touch those cases at all.

If you are a patient participating in a clinical trial, this technology could help results get published faster. It means researchers can analyze data more quickly, potentially speeding up the discovery of new treatments.

However, this tool is not currently used to diagnose individual patients in the hospital. It is strictly for analyzing data in research studies.

If you are a caregiver or patient, you don't need to ask your doctor about this AI yet. It is a background tool for researchers, not a bedside diagnostic device.

This study has important limitations. First, it only looked at one specific type of heart trial (patients who had a heart attack). We don’t know if the AI works as well for patients with heart failure or other conditions.

Second, the AI still makes mistakes. For heart attacks, the agreement rate was 76% when including all events. That means 1 in 4 cases might need a second look.

Finally, the AI was trained on data from this specific trial. It needs to be tested in different hospitals and with different types of medical records to prove it works universally.

The next step is broader testing. Researchers need to validate this AI in other types of cardiovascular trials and in different healthcare systems.

If these tools continue to prove accurate, they could become the new standard for clinical trial data review. This would not replace doctors but would free them up to focus on complex decision-making rather than paperwork.

For now, Auto-MACE represents a promising step toward more efficient medical research, helping bring new treatments to patients faster without compromising safety.

Study Details

Study typeRct
Sample sizen = 5,661
EvidenceLevel 2
PublishedApr 2026
View Original Abstract ↓
BACKGROUND: Major adverse cardiovascular events (MACE)-cardiovascular (CV) death, nonfatal myocardial infarction (MI), and nonfatal stroke-are highly relevant clinical outcomes. In global randomized trials, medical records review by a physician clinical events committee (CEC) is the conventional standard for adjudicating MACE but is labor intensive. Automated adjudication with the use of artificial intelligence (AI) could reduce cost and improve reproducibility. OBJECTIVES: In this study, the authors sought to develop and validate an AI-based adjudication system for MACE and compare its performance with CEC adjudication in a large global randomized trial. METHODS: We developed an AI-based system ("Auto-MACE") that uses an iteratively refined prompt of the OpenAI o1-mini language model to adjudicate MACE events, and a Clinical Longformer model trained on adjudicated events to assign a confidence level. We validated Auto-MACE against CEC adjudication in the PARADISE-MI global clinical trial comparing sacubitril/valsartan and ramipril in 5,661 patients with MI complicated by systolic dysfunction or pulmonary congestion. RESULTS: Auto-MACE provided a confident adjudication in 315/455 deaths (69%), 301/659 potential MIs (46%), and 136/167 potential strokes (81%). Auto-MACE agreed with the CEC adjudication in 97%, 89%, and 88% of confident events, respectively. Among all events, Auto-MACE agreed with CEC adjudications in 86% of deaths, 76% of potential MIs, and 84% of potential strokes. The estimated effect of sacubitril/valsartan vs ramipril on composite MACE was similar with Auto-MACE adjudication (HR: 0.91; 95% CI: 0.78-1.07) and CEC adjudication (HR: 0.90; 95% CI: 0.77-1.05). CONCLUSIONS: AI-based adjudication of MACE showed high agreement with human CEC adjudication, especially for CV death and stroke, and where the model was confident. Initial AI-based adjudication with CEC review of uncertain events may reduce workload while maintaining accuracy.
Free Newsletter

Clinical research that matters. Delivered to your inbox.

Join thousands of clinicians and researchers. No spam, unsubscribe anytime.