Will Biomarkers Eventually Replace Physicians?

For my entire career, I have been an outspoken critic of the thoughtless use of biomarkers in the practice of medicine. Biomarkers are measures of a substance (typically in blood) or a bodily structure or function (often through an imaging study). Originally developed for making a diagnosis, they have subsequently been widely used to estimate the severity of a disease or assess the adequacy of the response to its treatment.

There are an extraordinarily large number of biomarkers, and more are developed with each passing day. Physicians routinely measure or evaluate countless biomarkers in clinical practice. Some rely on them heavily.

If a patient has diabetes, a physician is likely to monitor HbA1c to guide the use of medications. If a patient has heart failure, a physician may perform repeated measurement of ejection fraction or natriuretic peptides.

Biomarkers have long been an important part of medicine. For many decades, patients routinely had blood tests, a urinalysis, and an electrocardiogram. These are all biomarkers.

So why have I been so critical of our reliance on biomarkers?

In the past, physicians would spend time interacting with patients, taking a careful history and assessing each patient’s progress while answering questions and providing comfort. When biomarkers were assessed, they typically supplemented the personal insights gained from directly talking to patients. The human interaction was primary; the biomarkers were secondary.

However, over the past several decades, the time spent with a patient has dwindled, and the number of biomarkers has skyrocketed. Now many physicians are judged and paid according to the biomarkers that they check.

If the biomarkers look good, the patient must be doing well. And conversely, if the biomarkers look worrisome, the patient must be doing poorly. In either case, all too often, there is no need to ask the patient how he or she is doing. The biomarkers now have a higher priority than the patient-physician interaction.

So why is that a problem?

Most biomarkers are really awful. Even though physicians rely on them, most do not reflect how the patient is feeling or whether their condition is improving or worsening. We have few truly valid biomarkers.

HbA1c levels reflect control of blood glucose. And over a period of 10-15 years, that may be important in determining the effect of the disease on some organs. But for the older patient with diabetes, over the time course of 1-5 years, adjusting drugs based on their effect on HbA1c does not change their high risk of cardiovascular or renal disease, which can inflict life-threatening injury, as I wrote last year. Physicians can reduce these risks, but doing so does not involve adjusting drugs based on the measurement of HbA1c.

The same applies to heart failure. Many physicians compulsively measure blood levels of natriuretic peptides, and adjust medications to bring the levels towards a normal range. We have many drugs that have been shown to reduce the risk of death and disease progression in heart failure. But in the trials that showed their benefits, none of the drugs were adjusted based on the level of natriuretic peptides. In a large randomized trial, when physician judgment was supplemented by knowledge of natriuretic peptides, patients did not do better, but their care was significantly more expensive.

Given these important observations, how are physicians managing patients with diabetes or heart failure?

Most are still measuring biomarkers to determine which drugs to use and what doses to prescribe.

The biomarkers are easy to measure; they consume little physician time; and they do not require any real thought process.

Do you think this is bad? It is going to get worse.

If current trends continue, the number of biomarkers will soar to a point where it may be possible — by the analysis of a single tube of blood — to measure hundreds (perhaps thousands) of biomarker proteins. Eventually, assay results might be available almost instantaneously.

Think about it. At the start of each patient visit, all biomarkers are measured. And within minutes, the computer displays a full detailed profile of their biochemical health — each value displayed to 3 decimal points!

If the biomarkers actually became good enough to accurately measure every disease process, there might never be a reason for the patient to say anything. Just go to the lab, give a sample of blood, and wait a few minutes. And voila! The patient is given a list of diagnoses along with a precise quantification of their severity. And just before leaving, a pharmacist delivers a host of prescriptions that are designed to bring the biomarkers into the normal range.

Here is the most interesting part.

The entire process does not even need a physician.

A computer algorithm can manage the whole visit. There is no need for the patient to talk to a healthcare provider. Arguably, there is no purpose for a history or physical exam. No time is needed for patient questions or the delivery of comforting words. It is all efficient and very cost-effective.

In essence, the widespread use of biomarkers eliminates the need for and the expense of a middleman in the practice of medicine.

Who is the middleman? The physician.

Physicians who do procedures need not worry. It seems unlikely that robots will be able to do unassisted surgery or perform invasive diagnostic tests or treatments on their own. (I may be wrong about this!)

But computers may soon be able to read imaging studies faster, better, and less expensively than radiologists.

And if you are a non-procedural physician — someone who is paid to interact with patients, make a diagnosis, order tests, and prescribe non-procedural treatments — your days may be numbered. Artificial intelligence can easily do all of these things, with no need for an expensive intermediary.

Artificial intelligence can even be programmed to say hello, ask about symptoms, and provide comforting words. The computer would be very thorough and can be programmed to be very empathetic. Arguably, some patients might not even notice the absence of a human presence, or miss it.

Fortunately, that day has not yet arrived.

However, if you are a physician who robotically moves through their daily routine with minimal patient interaction and with a heavy reliance on ordering and treating biomarkers, you are only one step away from making yourself obsolete. In ten years from now, who will need you?

Remember that your humanity is the only part of the interaction that the patient really values.

So here is my advice to all healthcare providers: The mission of delivering healthcare is being a healer — a uniquely human experience. If you do not want to be replaced by a computer, then you should stop acting as if you are following a programmed algorithm.

Think about that the next time you order your next routine biomarker.