An aspirin tablet is like a key that doesn't fit every lock. In a large study of older adults, this key usually refused to turn: life without disability was not extended. But for some locks, especially those belonging to people with diabetes, excess weight and high blood pressure, the key did work. Why that happens remained a mystery for a long time.
Scientists decided not to look at everyone as a single group, but to learn to predict who exactly aspirin would help. To do this, they built a model that learns from each participant's data: how their health changed with and without aspirin. The model split people into three groups by expected benefit. In the first group, the most favorable, aspirin reduced the risk by 64%, that is, by 11.1 percentage points over five years. In the second group the effect was weaker, and in the third there was no benefit at all, with some even showing signs of harm.
What set apart those whom aspirin helped? They were younger, more often lived alone, smoked, had diabetes or obesity, higher blood pressure and greater grip strength, but lower cholesterol. So aspirin worked like a key for locks of a particular shape: cardiometabolic risk, meaning a tendency toward heart and metabolic diseases, made a person more receptive to the benefit.
Previously it was unclear why aspirin did not help on average, even though some minority groups showed signs of benefit. Now it has become clear: the benefit is concentrated in a subgroup with high cardiometabolic risk, rather than distributed evenly. This is the first test of the idea that aspirin may be useful not for all older minorities, but only for those at higher risk. Such an approach could change the way clinical trials are analyzed: instead of searching for an average effect, search for those whom the treatment helps.
For now this is pure knowledge, with no clinical applications: validation on other data is needed. But the very idea that a key can be matched to a lock in advance is inspiring. If the model is confirmed, a doctor could tell a specific patient: aspirin will most likely help you, and it won't help you. And then the tablet will stop being a lottery.
In the group with the highest predicted benefit from aspirin, participants were more likely to live alone but had stronger grip strength. Grip strength is considered a marker of overall health and survival in older adults.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.