Duration of Trials vs Duration of Therapy

The Study of the Week exposes one of biggest blind spots in all of evidence-based medicine.

Co-editor of Sensible Medicine Vinay Prasad and two colleagues published a remarkable study last week regarding the length of antidepressant drug trials vs duration of therapy in the real world.

The median trial duration was 8 weeks while real-world use of these drugs is 5 years. Few trials lasted long enough to monitor for withdrawal symptoms or included taper protocols.

I am not an expert in psychiatry, nor pharmacology, but the topic of trial duration vs long-term use is a key area of evidence-based medicine.

I asked Claude to calculate the duration of the most cited trials in cardiology—using trial manuscripts because many were terminated early. Below is the screenshot. The median duration was 5 years. Yet many of these therapies are carried on for decades in patients.

Three Examples

Intensive BP Control

The BP-target SPRINT trial stands out. SPRINT compared an “intensive” 120 mmHg vs standard 140 mmHg target for treating patients with high blood pressure. After 3.3 years the trial was stopped prematurely because the intensive arm led to lower rates of major adverse cardiac outcomes.

SPRINT has been contested; there are many caveats in its interpretation. Let’s ignore those and assume that intensive therapy is better over the 3+ years of follow-up.

The average age of patients in SPRINT was 67 years. What happens when this patient is 75 or 80 years old? We know that aging brings extra conditions, and drug metabolism slows. Does the benefits of intensive treatment persist longer than the trial period? I am asking.

Statins

I don’t want to wade into the statin controversy. Similar to the BP example, let’s assume that the approximate 25% relative risk reduction of non-fatal cardiac events is true.

Yet the average duration of the ten-most cited trials is 5-6 years.

Let’s pick a well-known trial called the JUPITER trial of rosuvastatin vs placebo in patients with elevated CRP levels suggestive of inflammation.

The trial planned to go five years but was stopped after 2 years because of benefit. The average age was 66 years. Same question as before: what happens with statin use after 10 years, at age 76, or 20 years, at age 86?

The 4S trial of simvastatin vs placebo raises similar questions. The average follow-up time of these 58 year-old patients was 5 years. Same question again: what happens after decades of therapy?

---

Let me now point you to one of the best pieces I’ve ever read on statin use. It’s from Professor Rod Hayward at the University of Michigan.

The title is When to Start a Statin is a Preference-Sensitive Decision. He has a paragraph subtitled “Unknowable Unknowns.” Do read the whole piece as it is open access. Here is a quote:

Could being on a statin for 25 years result in a 25% increase in decline in cognitive function or a 35% acceleration in the decline in muscle health associated with aging? Such effects are unknowable given any data source currently available.

Because starting statins at 40 years of age would result in 10s of millions of people being put on a daily, highly biologically active medication 10 to 20 years earlier than under current guidelines, such questions are relevant. For example, if being on a daily statin for 25 years had even a modest negative impact on common aspects of aging, then it would not just reverse the marginal benefits of routinely starting a stain at 40 years of age, it also would be a public health disaster.

The implantable cardioverter defibrillator or ICD.

Twenty years ago, two seminal trials found that the ICD vs medical therapy alone reduced the risk of dying in patients with heart failure.

There is debate about whether the ICD retains its efficacy given the improvement of medical heart failure therapy, but for the sake of argument, let’s assume the ICD still has benefit.

The MADIT 2 trial duration was 20 months, and the average age of patients was 64 years. Pause there. 20 months!

Nearly every week, I see a patient whose ICD battery has reached depletion (usually 10 years out from initial implant) and is considered for replacement. Sometimes it is for a third or fourth replacement.

It is one of the great tragedies in cardiology that we have no idea whether an ICD replacement will confer the same benefit as the original implant. It almost certainly does not. For two reasons: one is that the patient is older and has acquired more illnesses that act as competing risks. And the risk of a generator change is higher.

Conclusion

I offer you no solution how to handle translation of evidence well beyond trial duration. The antidepressant example is extreme, but the concept pervades all of medicine. If I was a czar of evidence based practice, I would work with experts to gain more evidence on long-term therapy.

The ICD problem could be solved with a simple trial. That it has not been done is likely because the major funder of medical evidence—industry—has no interest in a trial that could lessen units sold. (That is not nefarious; it just is.)

A pragmatic approach I have learned over the years is to accept uncertainty and shun hubris. We simply don’t know in most cases of long-term therapy. Guidelines are especially unhelpful because the colored boxes recommending therapies don’t include such nuance.

When there is uncertainty, the best answer is transparent discussions with patients so as to align treatment with their goals. Some people are maximizers, some are minimizers. Many defer to us as trusted advisers.

I often tell 90-year-old patients that the purpose of primary prevention meds (BP and cholesterol drugs) is to make it to 90. Now there, there is unknown benefit.

I am sorry that the study of the week decreases our confidence in evidence. But this is the reality of medical practice.