A Story about Surrogate Outcomes

I was part of a big meta-analysis of heart failure trials looking at how we measure clinical outcomes. We have more to come on this topic.

JAMA Network Open published our paper—which is open access. You can read the whole thing.

What follows is a preliminary report. Senior author Andrew Foy and I are writing more on our findings. We believe the question of how to measure outcomes in heart failure is a huge issue in this era of cardiology.

Also, kudos to first author Ahmed Sayed, MBBS, who has strong skills in data analytics and maybe even a stronger work ethic.

Clinical trial expert Dr. Sanjay Kaul wrote the accompanying editorial

---

The question is how to measure success of new therapies for heart failure.

This is a modern question. In days of old, heart failure trials measured one bias-free endpoint—alive or dead.

The foundational therapies of HF—ACE-I, ARBs, beta-blockers and mineralocorticoid receptor antagonists—extended survival. Period.

Times have changed. With these agents as background therapies, newer therapies cannot extend life any further. Mainly because humans have an expiration date. If you stabilize heart failure, patients live long enough to die of something else. (This is something I tell patients in my ICD clinic.)

Yet lots of things in medicine improve life without extending life. AF ablation improves life without lowering mortality.

So we need another outcome for our heart failure therapies.

The current choice is something called HHF or hospitalization for heart failure. New drugs for heart failure may not reduce death, but if they reduce the need for a hospital admission to treat excess fluid build-up, that is a positive.

And. In the days of old, the drugs that extended life in heart failure also reduced HHF. Things were congruent.

But there are now new trends in medicine that made us question the value of HHF as an important surrogate for benefit.

Trend 1 is that patients with HF increasingly present in older age and have more co-morbid conditions. That complicates matters because an older person may be hospitalized for many different things and reducing only one kind of hospitalization may make little overall difference.

Here is a slide I made regarding the ratio of HHF to total hospitalizations in the recent empagliflozin vs placebo trial in patients with heart failure and preserved ejection fraction. You can see that HHF represents only a small fraction of total hospitalizations.

Trend 2: Older patients and those with other illness may be more apt to have adverse effects from new therapies.

This is another reason we think all-cause hospitalizations is a better outcome. Let’s say a new drug causes low blood pressure, urinary tract infections, or electrolyte issues. Older patients with more co-morbidity may be more susceptible to harms. (Proponents of sticking with HHF will argue that trials measure adverse events, but we believe all-cause hospitalization is a more inclusive way to measure harm.)

Trend 3: Patients these days often deal with multiple conditions—not just a weak heart.

A final reason to care about all-cause hospitalization rather than a cause-specific hospitalization is patient-centered.

You are a 75-year-old person. You have had heart failure and high blood pressure and diabetes and obesity. You take numerous medicines. Your doctor wants to add another new drug that reduces HHF. You ask about total hospitalization and it either does not change this outcome or it has only a modest benefit. Is this enough to bother with another potentially expensive medication? Many patients think not.

---

This is the background for our meta-analysis of more than 100 trials in heart failure.

We discovered some interesting patterns and in future posts will discuss how our findings inform the way we measure net benefits in modern heart failure.

Two things to think about as you read our open-access paper: one is David Brown’s analysis of MI as a surrogate measure. We used to think MI was a valuable outcome to reduce. But his analysis showed that, in modern times, MI may not be such valuable surrogate of overall cardiac benefits.

Second similar issue: Vinay Prasad has often opined on the value of CT scanning for lung cancer. Advocates argue that screening reduces the chance of dying from lung cancer. But that reduction may not translate to living longer—which is, of course, the ultimate goal.

One of the reasons that CT scanning could reduce cancer deaths but not change overall death rate is that screening can cause harm. The endpoint of all-cause death better captures both benefits and harms.

---

Stay tuned. Foy and I have more to say about heart failure outcomes. It is super-important.