Older adults whose elevated blood marker of heart stress had returned to normal three years later went on to have almost the same eight-year risk of heart disease and death as people who never had an elevated level at all, according to an observational study of 8,454 adults aged 70 and older published August 18, 2026 in Annals of Internal Medicine. People whose level was newly elevated at year 3 had a 7.4 percentage-point higher absolute risk of cardiovascular disease by year 8.
The marker is NT-proBNP, a protein released by heart tissue under strain. What the study tested was not the number itself but the direction it moved.
Key takeaways
- Two measurements three years apart predicted risk better than one, sorting people into persistently free of heart stress, remission, newly elevated, and sustained elevation.
- Newly elevated levels carried a 7.4 percentage-point higher eight-year risk of cardiovascular disease; sustained elevation, 6.7 points. Death risk rose 6.1 and 7.5 points.
- People in remission had adjusted absolute risks similar to those who never had elevation, but nobody was randomized to lower their number.
What the study found
Researchers from Monash University, Shanghai Jiao Tong University School of Medicine and other centers used data from the ASPREE aspirin trial and its observational extension. All 8,454 participants had no history of cardiovascular disease. Mean age was 78.0 years at year 3, and 52.9 percent were women. NT-proBNP was measured at enrollment and again at year 3.
Over a median follow-up of 8.0 years, there were 818 cardiovascular events and 1,584 deaths. Cardiovascular events counted nonfatal heart attack, fatal or nonfatal stroke, coronary heart disease death, and hospitalization for heart failure.
Compared with people who stayed free of heart stress at both visits, adjusted absolute risk of cardiovascular disease at year 8 was higher by 7.4 percentage points for newly elevated levels (95% CI, 4.8% to 10.0%) and by 6.7 points for sustained elevation (CI, 4.3% to 9.0%). In natural numbers, that is roughly 74 and 67 extra cases per 1,000 people over eight years. The ranges mean the true gap is very likely somewhere between about 5 and 10 extra cases per 100 for newly elevated levels, and between about 4 and 9 per 100 for sustained elevation.
Death from any cause rose by 6.1 points for newly elevated levels (CI, 3.7% to 8.5%) and 7.5 points for sustained elevation (CI, 5.2% to 9.8%), or about 61 and 75 extra deaths per 1,000 over eight years.
The remission group, elevated at enrollment but not at year 3, had adjusted absolute risks similar to the persistently unstressed group.
Dr. Kumar’s take
This turns a lab value from a verdict into a trajectory. The information that mattered here was the direction the number moved over three years, which is a different and more useful question than whether a single draw crossed a threshold.
The remission finding is where the reflexive read goes wrong. Nobody was assigned to bring their number down. A value that drifts back toward normal can reflect regression to the mean, better blood pressure control, less fluid on board, weight change, or a different day and a different lab. It does not establish that a heart repaired itself, and it does not establish that pushing the number down would move the risk. The authors say this plainly: heart stress here is a marker of subclinical cardiac dysfunction, not a modifiable therapeutic target.
Two limits are worth holding onto. The design is observational, so adjusted absolute risks describe association, not cause. And the participants were predominantly White, so the numbers may not transfer cleanly to other populations.
Where this fits: most of what drives these events is not exotic. Nearly 9 in 10 US heart disease deaths trace back to modifiable risk factors, and diet is one of the levers with long-term cardiovascular data behind it.
What it means for you
If you are over 70 and have an NT-proBNP on a past lab report, the single value tells you less than a repeat value would. A rise from normal into the elevated range carried the larger added risk of cardiovascular disease, 7.4 percentage points against 6.7 for a level that stayed elevated, while a level that stayed elevated carried the larger added risk of death, 7.5 points against 6.1. Those ranges overlap, so neither pattern is clearly worse than the other, and both are worth flagging.
None of this makes NT-proBNP a treatment target. Nothing in this study shows that acting to lower the number lowers risk. What it supports is using change over time as a sharper signal of who deserves closer attention to blood pressure, weight, activity and heart failure symptoms.
A number with something to compare it against is worth far more than a number standing alone.