Longevity medicine has a measurement obsession and a causal inference problem.

Patients bring me studies all the time. A paper linking sauna use to lower mortality. A trial showing that an intervention moved a biological age clock backward. They want to know whether the finding should change what they do.

Should I be doing this?

That question is harder than it looks. Answering it requires understanding how the study was done, what it actually established, and how much of it applies to the person asking. It also requires being willing to say that an interesting result has not yet earned a place in someone's medical care.

Much of what is marketed as longevity medicine moves past those questions too quickly. An association becomes a recommendation. A change in a biomarker becomes evidence of rejuvenation. Sophisticated measurements give the whole process an appearance of precision.

The patient is left trying to decide which parts to believe.

The discipline that helps us work through this is causal inference. Its central question is straightforward: what would have happened if the person had done something different?

Suppose you start taking rapamycin at forty-five and remain healthy for the next twenty years. Did the drug help? Would you have stayed healthy anyway? Could it have caused harm that was outweighed by other things you were doing?

We can observe what happened to you. We cannot observe the same twenty years of your life without the drug. Epidemiologists call that missing comparison the counterfactual.

This problem applies to almost everything we do in medicine. Randomized trials help by assigning treatment independently of the characteristics that might otherwise influence who receives it. They give us a more reliable comparison than watching people choose their own treatments.

Preventive medicine has benefited enormously from these trials. We have outcome data for blood pressure treatment and lipid lowering, among many other interventions. We do not have to wait for an entire lifespan to learn whether a treatment prevents meaningful disease.

But many newer longevity interventions have little comparable evidence. We have animal experiments, observational studies, and human trials that are often small or short. These can justify further research. Deciding what they justify in clinical practice takes more care.

Consider sauna bathing.

The association between frequent sauna use and lower mortality is interesting. There are plausible physiological reasons heat exposure might help. There are also reasons frequent sauna users might differ from less frequent users in ways that affect their health.

The researchers know this. The widely cited Finnish study adjusted for factors including smoking, physical activity, fitness, and socioeconomic status. Those adjustments strengthen the analysis. They cannot guarantee that all relevant differences were measured or adequately accounted for.

We should be careful in both directions. The study does not establish that prescribing more sauna sessions will produce the mortality reduction observed. We also cannot dismiss the result simply by imagining a list of possible confounders. The association may reflect a real benefit, differences between the people being compared, or some combination.

That leaves room for a reasonable conversation about sauna. Someone may enjoy it and find it relaxing. What I cannot responsibly do is promise that it will reduce their chance of dying by the percentage in the headline.

Biological age testing presents a different version of the same problem.

A clock may predict who is more likely to develop disease or die. That makes it potentially useful. It does not automatically mean that an intervention that lowers the clock's estimate will improve those outcomes. Prediction and response to treatment are different questions.

A recent analysis of multiple human intervention studies found that epigenetic clocks varied in their responses across interventions and populations. Establishing their usefulness as clinical surrogate endpoints requires more than showing that their readings can move.

If a test says your biological age fell by two years, I want to know what that means for you. Are you functioning better? Has your risk of disease changed? Do we have evidence connecting an intervention-induced change in that score to either outcome?

The result may be scientifically interesting while those questions remain unanswered. Charging a patient to optimize it requires a separate justification.

This is where discussions about evidence can become frustrating. People hear the qualifications and conclude that medicine is telling them to wait.

We still have to make decisions. Patients are getting older while the studies are being done. The amount of uncertainty we can reasonably accept depends on what we are considering, the potential benefit, the possible harm, and what else the intervention asks of someone's life.

We have strong reasons to treat elevated blood pressure and reduce exposure to atherogenic lipoproteins in appropriate patients. We have good reasons to help people exercise and maintain their physical capacity. Exercise trials have demonstrated benefits that matter in their own right, including reducing mobility disability in vulnerable older adults. We can value those benefits without claiming that every improvement in fitness or muscle mass has been proven to extend life.

We should be equally specific about metabolic health. The Diabetes Prevention Program showed that an intensive lifestyle intervention reduced progression to diabetes in people at high risk. That is a substantial clinical result. It gives us something firmer to discuss than a claim about improving every pathway of aging.

Being precise about these distinctions helps us act. It tells us where to put our effort and what we can honestly promise.

Even then, interpreting the evidence gets us only part of the way to a decision.

A trial estimates what happened in a group under particular conditions. In the exam room, I need to consider the patient's baseline risk, medical history, current treatment, tolerance for side effects, and priorities. The likely absolute benefit matters. So does the burden of getting it.

Two people can reasonably make different decisions after reading the same study. One may have more to gain because their underlying risk is higher. Another may place greater weight on avoiding a side effect or preserving something important in their daily life.

That distance between a research finding and a decision is the judgment gap. Clinical judgment has to remain accountable to the evidence while making sense of the circumstances the study could not fully represent.

This takes conversation and follow-up. I need to understand what a patient hopes to achieve, explain the tradeoffs, and agree on what we will watch. We need to know whether treatment is tolerable, whether the intended effects occur, and whether the plan remains practical.

Following someone over time does not solve the counterfactual problem. I generally cannot tell a particular patient that our treatment prevented their heart attack. I can explain the evidence behind our decision, assess their response, and revise the plan as their circumstances and the evidence change.

That ongoing responsibility is a large part of why we built PrimaryMD.

Patients should be able to bring us a study and expect us to take it seriously. Sometimes it will change our advice. Sometimes it will raise a question worth following. Sometimes we will explain why the conclusion goes beyond the evidence.

They deserve to understand how we reached that judgment, and to have a physician who stays involved in what happens next.

Sukhjit "Sarge" Takhar, MD, SM is a physician and Harvard-trained epidemiologist and the Co-Founder and Chief Medical Officer of PrimaryMD.