Evaluating evidence
Surrogate Endpoints Versus Outcomes in Diabetes Trials
A surrogate endpoint is a measurement that stands in for the thing you actually care about. In diabetes that thing is usually a complication you want to avoid: a heart attack, kidney failure, vision loss, an amputation, an early death.
A surrogate endpoint is a measurement that stands in for the thing you actually care about. In diabetes that thing is usually a complication you want to avoid: a heart attack, kidney failure, vision loss, an amputation, an early death. HbA1c, blood pressure, and LDL cholesterol are the markers we measure instead, because they move faster and cost less to track. Here is the part that gets lost in headlines: moving a surrogate is evidence that something happened in the body, not proof that a patient will live longer or avoid harm. The two coincide often enough to be useful and diverge often enough to be dangerous. This article is educational and not medical advice; for your own care, talk with a clinician who knows your history.
I have watched this distinction matter from a few different chairs. In the EASY-1 randomized controlled trial (NCT03258268) we measured outcomes and workflow changes for an AI decision-support system against standard of care, and in global development work the gap between a marker moving and a patient benefiting was a daily conversation. The lesson keeps repeating: a marker is a messenger, and the messenger is not the message.
What does HbA1c actually promise?
HbA1c reflects the average amount of glucose attached to hemoglobin over roughly the prior three months. It is a genuinely good marker. It is reproducible, it integrates glucose over time so a single good or bad day does not dominate it, and decades of observation tie higher values to more microvascular damage in the eyes, kidneys, and nerves. When a clinician tracks your HbA1c, they are using one of the most validated surrogates in medicine.
What it promises is narrow, and worth stating precisely. A lower HbA1c tells you average glucose exposure fell. It is most trustworthy as a stand-in for the small-vessel complications, where the chain from glucose to tissue damage is fairly direct. It is a weaker promise for the large-vessel events, the heart attacks and strokes, where glucose is only one of several drivers and the others (blood pressure, lipids, clotting, smoking) carry much of the weight. The same number on the same lab report can be strong evidence for one outcome and merely suggestive for another. So the honest reading of "this drug lowered HbA1c by half a point" is "average glucose exposure dropped, which tends to reduce small-vessel complications," not "this drug prevents heart attacks."
Why is moving a surrogate not the same as preventing a complication?
The cleanest way to see the gap is to picture the causal chain. A treatment acts on the body, the body changes the surrogate, and the surrogate sits somewhere along the path toward the outcome. Using a surrogate assumes the treatment helps the outcome mainly by moving the marker, and that the marker captures most of what the treatment does. Either assumption can fail.
The marker can move while the outcome does not, because the drug nudged the number through a pathway that does not protect tissue. The outcome can improve while the marker barely moves, because the drug helps through a channel the surrogate never measured. The hardest case is a drug that improves the surrogate and harms the patient at once, lowering the marker by one mechanism while doing damage through another the marker cannot see. This is not hypothetical. In more than one cardiometabolic area, agents that reliably pushed a marker in the "right" direction proved, once tested against events, to be neutral or harmful. The marker behaved as predicted. The patient did not.
What makes a surrogate trustworthy?
Not all surrogates are equal, and the field has learned to grade them. A surrogate earns trust when two things hold up. First, it has to be on the real causal path to the outcome, not merely correlated with it; many markers travel alongside a disease without driving it, the way a fever tracks an infection without being what hurts you. Second, and this is the demanding part, the treatment effect on the surrogate has to predict the treatment effect on the outcome across multiple drugs and trials. That is why validation happens at the level of trials rather than individuals. When the relationship is tight, the surrogate is "validated" for that outcome. When it is loose, the marker is a hypothesis generator, useful for deciding what to test next, not a substitute for the outcome trial.
HbA1c sits at an interesting spot on this scale. For microvascular outcomes it is about as well-supported as surrogates get. For cardiovascular outcomes the link between lowering it with a particular drug and preventing events has been inconsistent enough that glucose-lowering agents are now increasingly expected to show cardiovascular benefit directly, on hard outcomes, rather than resting on the HbA1c they produce. That shift is one of the healthier corrections the field has made.
How should you read a study that reports a surrogate?
Start by naming the endpoint out loud. If the primary endpoint is a marker, the study is a surrogate study, whatever the framing. A pre-specified surrogate in a short trial is an early signal, not a verdict.
Next, ask what the surrogate is standing in for, and whether it is qualified for that job. HbA1c standing in for retinopathy risk is on solid historical ground; the same HbA1c standing in for cardiovascular mortality is on softer ground, and a thoughtful paper will say so. Watch for the quiet slide from "improved a marker" in the results to "reduces complications" in the discussion, with nothing in between to earn the upgrade.
Read the size and the spread, not the significance alone. A statistically real change can be clinically trivial, and an average hides its tails: half a point of HbA1c in a group near target is a different thing from half a point in someone running very high. Then look at duration. Complications accrue over years, while many surrogate trials run for months.
None of this argues against surrogates. They make trials feasible and let us screen many ideas before committing to long, expensive outcome studies. The discipline is to hold the claim at the altitude the evidence supports. A surrogate that moves tells you the machinery responded. Whether the patient is better off is a separate question, and the best studies never let the first answer stand in for the second.
References and sources
How this was researched. This explainer is built from the primary sources listed above and reflects Dr. Tojjar's own critical appraisal of that evidence. It explains and evaluates research and does not provide medical care.
This article is for general education and is not medical or professional advice. For guidance about your own health, talk with a qualified clinician.
Cite this article
Tojjar, D. (2024). Surrogate Endpoints Versus Outcomes in Diabetes Trials. Dr. Damon Tojjar. https://readingtheevidence.org/articles/surrogate-endpoints-vs-outcomes/
This article is part of Dr. Tojjar's guide to Evaluating evidence.
Part of the reading path How to read a risk or benefit number (step 7 of 7).
Part of the reading path How to Read a Drug Trial (step 5 of 10).
Part of the reading path How to Read an Oncology Trial (step 5 of 9).