Identification of Core Biomarkers and Temporal Mechanism Investigation in Cardiovascular Disease: A Machine Learning and Cross-Lagged Path Model Approach
This study utilized the CHARLS database and machine learning to identify the TyG-BMI index and cystatin C as core biomarkers for predicting cardiovascular disease in Chinese residents, revealing a bidirectional causal relationship between CVD and the TyG-BMI index and a unidirectional link with cystatin C.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Heart disease is a silent, relentless force that claims more lives in China than any other cause, often striking without warning. For decades, doctors have relied on a standard set of clues to guess who might be at risk: blood pressure, cholesterol levels, and sugar in the blood. These markers are useful, but they often tell only part of the story, missing the subtle, complex ways our bodies signal trouble long before a heart attack or stroke occurs. As the population ages and metabolic conditions like diabetes become more common, the need for a sharper, more accurate way to predict who will develop heart disease has become urgent. Researchers are now turning to powerful computer tools to sift through vast amounts of health data, looking for patterns that human eyes might miss, hoping to find the specific biological signals that act as the earliest warning lights for cardiovascular trouble.
A team of researchers from Xinjiang Medical University and the Urumqi Center for Disease Control and Prevention decided to tackle this challenge by analyzing the health records of over 10,000 Chinese adults. They used a massive, nationally representative survey called the China Health and Retirement Longitudinal Study, which tracks the health of people across the country over many years. The goal was not just to list risk factors, but to build a sophisticated prediction model that could identify the most critical biomarkers—measurable substances in the blood and body—that forecast heart disease. To do this, they employed four different types of machine learning algorithms, which are computer programs designed to learn from data and make predictions. These programs acted like four different experts, each using a unique method to analyze the same set of information, which included age, gender, education, and a wide array of blood tests ranging from white blood cell counts to complex metabolic indices.
The researchers fed data from 2015 into these computer models, asking them to distinguish between people who already had heart disease and those who did not. The models were trained on a large portion of the data and then tested on a separate group to see how well they could generalize their findings. Among the four approaches, one algorithm, known as XGBoost, proved to be the most effective at sorting through the noise and pinpointing the true predictors. While the computer models evaluated dozens of variables, two specific indicators stood out as the most powerful predictors of all. The first was a composite score called the TyG-BMI index, which combines measurements of triglycerides (a type of fat in the blood), blood sugar, and body mass index. The second was cystatin C, a protein that serves as a sensitive marker for how well the kidneys are filtering waste from the blood. These two factors consistently rose to the top of the list, outperforming traditional markers like standard cholesterol levels in their ability to predict who would have heart disease.
To understand not just which factors were important, but how they relate to heart disease over time, the researchers took a deeper look at a subset of participants who had been tracked for four years, from 2011 to 2015. They used a statistical method called cross-lagged path analysis to determine the direction of the relationship between these biomarkers and heart disease. This approach allowed them to see if high levels of a biomarker came before the disease, or if the disease caused the biomarker levels to change. The results revealed a fascinating and complex dynamic. For the TyG-BMI index, the relationship was bidirectional, meaning it was a two-way street. High levels of this metabolic marker increased the risk of developing heart disease, and conversely, having heart disease seemed to drive the levels of this marker even higher, creating a vicious cycle where the two conditions fueled each other.
In contrast, the relationship between cystatin C and heart disease moved in only one direction. The analysis showed that having heart disease led to higher levels of cystatin C, but high levels of cystatin C did not predict the future onset of heart disease in the same way. This suggests that while kidney function, as measured by cystatin C, is a strong indicator of existing heart problems, it may be more of a consequence of the strain heart disease places on the body rather than a primary cause. The study found that nearly half of the participants in the survey already had some form of cardiovascular disease, a rate significantly higher than in many Western populations, highlighting the severity of the issue in China. The researchers concluded that by focusing on these two core markers—the metabolic TyG-BMI index and the kidney-related cystatin C—doctors could potentially identify high-risk individuals earlier and intervene before a catastrophic event occurs.
The findings offer a new perspective on how we view the connection between the heart, metabolism, and the kidneys. The bidirectional link between the TyG-BMI index and heart disease suggests that managing body weight, blood sugar, and fat levels is not just about preventing heart disease, but also about breaking a cycle that, once started, becomes self-perpetuating. The unidirectional link with cystatin C underscores the importance of monitoring kidney health in patients who already have heart conditions, as the strain of the disease can quickly impact renal function. While the study focused on middle-aged and older adults, and the researchers noted that further validation is needed in younger populations, the work provides a clear, data-driven roadmap for the future of heart disease prevention. By moving beyond single numbers like cholesterol and embracing these complex, interconnected biological signals, the medical community may finally have the tools needed to stay one step ahead of a disease that has long been the leading cause of death.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.