Chelation Did Not Work But Science Did

When the TACT 2 trial failed to confirm TACT 1, the winner was science.

I remember when Dr. Gervasio Lamas stunned the world of cardiology when he presented the results of the TACT 1 trial. I chose the verb stun because chelation was way outside the box that contains medical thinking.

This was not a drug or a procedure. It is a way to remove heavy metals from the body. And it was practiced not by doctors, but alternative medicine practitioners.

Published in JAMA, the placebo-controlled randomized Trial to Assess Chelation Therapy found that chelation therapy induced an 18% reduction in major cardiovascular events. The confidence intervals and p-value met the threshold for significance.

Even more remarkable, a subgroup of patients with diabetes garnered an even greater benefit from chelation. In these patients, chelation reduced cardiac events by 41%.

---

Chapter 1 What to think about a surprise finding?

Let’s pause here and pretend we are neutral Martians unencumbered by priors regarding alternative vs traditional medicine.

NIH sponsored the trial. Dr. Lamas is a seasoned and established trialist. Co-investigator Daniel Mark also boasts traditional academic credentials. TACT had a proper-placebo control and endpoint. It had long follow-up and enrolled patients who had established heart disease.

If the treatment were a drug or lucrative procedure, Lamas’ presentation of a 41% reduction in major cardiac events in patients with diabetes would have been celebrated and likely sailed through FDA review.

But chelation was not a traditional therapy. Heavy metals were not an established cause of atherosclerosis. So, this was, as Dr. Harlan Krumholz wrote, “inconvenient” evidence.

The editors of JAMA published a letter explaining why they published the study. Steve Nissen wrote a highly skeptical editorial.

---

Chapter 2 Chelation Therapy — My Favorite Part

Drs Lamas and Mark did not promote chelation therapy. They did not recommend that people—even those with diabetes—seek chelation.

They were cautious about their surprise findings. The clinical equivalent of their approach was similar to stress testing for finding coronary stenoses. It was as if they had found a positive stress test in a patient with low-risk for heart disease. In the clinic and in trials, you have to consider the priors.

Instead, they went back to the NIH and sought funding for a second trial—TACT 2. NIH funded TACT 2, a trial that would test chelation in patients with diabetes.

This, by the way, is what subgroup analyses are for. We like to say that subgroups are not for clinical action, they are “hypothesis generating.”

Chapter 2 is my favorite because this is how science is supposed to work. Instead of promoting a surprise finding in one trial, medical science should seek to confirm the finding in another trial. (See Chapter 4).

---

Chapter 3 - Science is Hard

Earlier this month, at the American College of Cardiology meeting, Lamas presented the results of the 1000 patients enrolled in TACT 2.

There were no significant effects of chelation on major adverse cardiac events. The hazard ratio was 0.93 and confidence intervals went from 0.76-1.16. The Kaplan-Meier curves detailing the rate of events over time were nearly super-imposable. There were no signals of benefit in any subgroups.

---

Chapter 4 Possible Explanations for Divergent Trials

Lamas explained at the meeting that lead levels had dropped over time. One study had found that lead levels were 17 mcg/L back in the early 2000s. Now levels are about 11 mcg/dl.

This is a compelling explanation. Conditions change. We have seen previously effective therapies (aspirin, defibrillators, and post-MI beta-blockers) all lose their efficacy over time because the conditions they treat have become less severe.

I offer another explanation. One that involves chance.

This stems from a paper from Dr. Erik van Zwet—a statistician from the Netherlands. I recently did a podcast with him.

When van Zwet and his colleagues analyzed tens of thousands of trials in the Cochrane database, they found that most trials have low power to sort signal from noise. The consequences of this were that even when statistical thresholds are met (p < 0.05), as in TACT 1, trials are unlikely to replicate again with a significant finding.

They write:

An initial P value between 0.001 and 0.005 in a trial implies only a 58% chance of getting P< 0.05 upon attempted replication. It’s even more unreliable in the stratum of -p-values of 0.01-0.05 is only 37%.

My thesis, therefore, is that TACT 1 was positive and TACT 2 negative for the same reasons that a lot of trials are positive—low power and chance.

(To my friends in statistics, I also add that TACT 2 was underpowered. Carried on for only 2.5 years, the confidence intervals were wide—allowing for a 24% benefit and 16% harm. I would not call it unhelpful though. These were post-MI patients with diabetes, a group that garnered a huge signal in TACT 1. The lack of signal in TACT 2 is telling.)

Chapter 5 The Lessons of the TACT trials

The lack of a positive finding in TACT 2 likely closes the chapter on chelation to prevent atherosclerosis.

It also somewhat dampens our enthusiasm for heavy metals as a modifiable source of atherosclerosis. I added “somewhat” because lead levels were less over time.

I would not use the non-significance of TACT 2 to conclude that heavy metal exposure is fine. It is clearly not.

The main lessons here are humility and skepticism. Medical science is hard. We have made great progress. If TACT 1 had studied a new drug or appendage closure device, I am afraid we would have been too accepting.

Lamas showed us the way. He sought more data—which did not confirm the original surprise findings. In the face of a possible discovery back in 2013, he was humble and skeptical. And for this he has earned my respect.