Stable redundancy and variable ranking of metabolic surrogate indicators for incident ischemic heart disease

Imagine your body as a bustling city where energy (sugar) and fuel (fats) are constantly being delivered, used, and stored. Sometimes, the traffic gets jammed, or the delivery trucks get stuck, leading to a buildup of "gunk" in the pipes. In the medical world, doctors use routine blood tests to check for these traffic jams. They look at things like how much sugar is in your blood, how much fat (triglycerides) is floating around, and how much "good" fat (HDL) you have. To make sense of all these numbers, scientists have invented "surrogate indicators"—clever math formulas that mix these numbers together into a single score. Think of these formulas like different apps on your phone that all try to predict the weather. Some use wind speed, some use humidity, and some use barometric pressure. Even though they all use different combinations of the same basic data, they often give you a very similar forecast.

The big question is: Which app is the best? If one app says "sunny" and another says "stormy," which one should you trust? And more importantly, if an app works perfectly in your neighborhood, will it still work if you move to a different city? This is exactly the puzzle researchers tackled in a new study. They wanted to know if these metabolic "weather apps" are truly superior tools for predicting heart trouble, or if they are just fancy ways of saying the same thing over and over again. They also wanted to see if a simple, compact formula could actually do a better job than just looking at the raw ingredients (the individual blood numbers) it was made from.

---

The Great Metabolic Score-Off

In this study, a team of researchers acted like a panel of judges at a talent show, but instead of singers, they were evaluating ten different "metabolic formulas." These formulas are designed to predict the risk of developing ischemic heart disease—a condition where the heart doesn't get enough blood, often leading to a heart attack or angina. The researchers didn't just look at one group of people; they used data from two massive, long-running studies: one in England (ELSA) and one in the United States (HRS). They treated the English data as the "training camp" to build their models and then sent those same models to the US to see if they could perform well in a completely different crowd. This is like testing a new video game strategy on a practice server and then seeing if it still works when you play against real opponents in a different country.

The researchers compared ten different formulas, including popular ones like the TyG index (which mixes triglycerides and glucose) and the AIP (Atherogenic Index of Plasma, which mixes triglycerides and HDL). They pitted these formulas against a standard "clinical model" that already uses basic facts like age, sex, smoking status, and blood pressure.

The Results: A Tie for First Place?

Here is the twist: No single formula won the whole tournament.

• The "All-Rounder": The AIP formula came out on top in the overall rankings. It had the best balance of predicting heart disease and adding new information to the standard clinical model. If you had to pick one "champion" based on the total score, it was AIP. It had a 39.4% chance of being the absolute best in any given scenario.
• The "Traveler": However, the TyG index was the most reliable traveler. While it didn't win the overall title, it performed the most consistently when the researchers moved the data from the English study to the American study. It didn't get confused by the change in location.
• The "Raw Ingredients" Test: The researchers also asked a critical question: Does mixing these numbers into a fancy formula actually help, or is it just as good to look at the raw numbers (triglycerides, glucose, and HDL) separately? They built "upper-bound models" using just the raw ingredients. The result? None of the fancy formulas consistently beat the raw ingredients. In fact, the formulas were often just as good, but never significantly better, than simply looking at the individual parts.

The "Weather App" Analogy

Think of these formulas like different ways to describe a storm.

• AIP is like a weather app that says, "It's going to rain because the humidity is high and the wind is blowing."
• TyG is another app that says, "It's going to rain because the temperature is dropping and the pressure is low."
• The Raw Ingredients are just the raw data: "Humidity is 90%, wind is 20mph, temp is 50°F."

The study found that all these apps are actually saying the exact same thing because they are all using the same weather data. The formulas are highly correlated—they move together like a flock of birds. But here is the catch: just because they move together doesn't mean they are equally good at predicting the future. In one city (England), AIP might be the best predictor. In another city (the US), TyG might hold up better. But neither is a magic crystal ball.

How Much Better Are They?

The researchers measured how much these formulas improved the prediction of heart disease compared to the standard clinical model. The gains were surprisingly small, like adding a tiny bit of extra spice to a dish that is already well-seasoned.

• In the English validation, adding the best formulas improved the prediction accuracy by less than 1 percentage point (specifically, +0.87, +0.67, and +0.44 percentage points for the top three).
• In the American validation, the improvements were also small, ranging from +0.64 to +1.44 percentage points.

Furthermore, when the researchers tried to use the English models to predict outcomes in the US, the "calibration" (how well the predicted risk matched the actual risk) wasn't perfect. The US predictions were a bit too extreme, suggesting that while the formulas might point in the right direction, they aren't ready to be used as a standalone tool for making life-or-death medical decisions without careful adjustment.

The Bottom Line

The study concludes that these metabolic formulas are stable redundancies. This means they are structurally very similar to each other (they all use the same ingredients), but their ranking of who is "best" changes depending on where and when you test them.

• AIP is the strongest overall candidate for predicting heart disease in this specific setup.
• TyG is the most consistent when moving between different populations.
• No formula is a "super-biomarker" that is inherently better than the raw blood tests it is made from.

The researchers suggest that we should stop treating these formulas as interchangeable, magical biomarkers. Instead, we should view them as compact summaries—a convenient way to package the same old information about your lipids and sugar into a single number. They are useful for summarizing shared metabolic information, but they don't reveal any new secrets that the individual blood tests didn't already tell us. The "winner" of the race depends entirely on the track you are running on, and for now, the raw ingredients are just as powerful as the fancy recipes.