Melatonin and heart failure: Something doesn't add up

Image by ACSH using AI

I'm no fan of supplements, as evidenced by the roughly one bazillion articles and op-eds in which I've warned against them. [1] Yet, I've tried melatonin because there is a biological rationale for its use in treating insomnia, a miserable condition I've been dealing with forever.

Aside from the usual concerns with supplements (e.g., ethical lapses in the industry, lack of quality control, and "let's guess a dose," among others), these "magic pills" remain popular because many people genuinely believe that a) doctors get paid to push pills, b) the stuff in the bottle is natural, so therefore it can't harm you, and c) there are a whole bunch of things we don't get in our diet, hence the need to supplement them.

This is mostly a gigantic pantload, but even among the mountains of dubious products, some are bound to be both safe and effective, perhaps only by chance. [2] Anyhow, I'm still awake AND alive, so a recent study on melatonin and heart disease piqued my interest, not least because I'd rather stay that way. Before we get going, keep in mind that the findings have been published only as a conference abstract, not as a full peer-reviewed paper.

What the study shows

The abstract, which appears in Circulation, describes a large observational study examining the relationship between long-term melatonin use and the risks of heart failure, hospitalization, and death. Well, that's not so wonderful, is it? Let's see if the carpet matches the drapes.

Researchers used a large international database of electronic health records to compare 65,414 patients with recorded long-term melatonin exposure with an equal number who had no recorded melatonin use. All had been diagnosed with insomnia, and patients with a history of heart failure were excluded.

To make the comparison as fair as possible, the researchers used a statistical technique called propensity-score matching, which attempts to create two groups with similar characteristics. For example, a 65-year-old melatonin user with diabetes and high blood pressure might be matched with a nonuser of similar age and medical history. They matched patients on more than 40 variables, including age, medical conditions, medications, and laboratory measurements. This is a legitimate and useful technique, although, as you'll see, it has limitations.

The researchers then examined three outcomes over five years: new cases of heart failure, hospitalizations related to heart failure, and deaths from any cause. The results were not exactly reassuring.

Among melatonin users, 4.6% developed heart failure, compared with 2.7% of nonusers. Deaths occurred in 7.8% and 4.3%, respectively. And the reported rates of heart failure-related hospitalization were 19% and 6.6%. [3] Those are substantial differences, especially for a supplement that millions of people consider harmless.

But before we start flushing our melatonin down les toilettes, a few things are worth examining (my esteemed colleague Dr. Chuck Dinerstein recently wrote a different piece about melatonin and magnesium. You can read it here).

What the study doesn't show

First, this was an observational study, not a randomized clinical trial. The researchers examined existing medical records rather than randomly assigning patients to receive melatonin or a placebo. This distinction matters because an association between two things doesn't necessarily mean that one causes the other, a common limitation of observational studies.

For example, people who take melatonin regularly may have more severe insomnia than those who don't. They may also have other medical problems, such as sleep apnea, that increase their risk of cardiovascular disease. In other words, melatonin might really be doing something nasty, or it might simply be along for the ride. The researchers attempted to address this problem by matching patients on more than 40 characteristics. That's a strength of the study, but it doesn't eliminate the possibility of confounding, in which an unmeasured factor influences the results. You can match patients on 40 variables and still miss the one that matters.

There's another problem, possibly the worst one. Melatonin is available over the counter in the United States, yet the researchers relied on prescription records to identify users. Do you know anyone who has even gotten a script for the stuff? I don't. To be fair, melatonin requires a prescription in some countries, including the UK. But that doesn't solve the problem of unrecorded use elsewhere. How many supposedly melatonin-free patients were buying it at CVS or Walmart? Nobody knows. Nor does having a prescription prove that someone actually took the supplement.

Consequently, the researchers couldn't reliably establish who was taking melatonin, how much they took, or how consistently they used it. That's a rather important limitation when you're trying to determine whether a substance is harmful.

Something doesn't add up

There's a bizarro finding in the study. Among melatonin users, 4.6% developed heart failure, yet 19% reportedly experienced heart failure-related hospitalization. How can four times as many people be hospitalized for heart failure as developed the disease? Did 14.4% of the people forget that they were hospitalized?

No, that's stupid. There has to be a better explanation.

And there is, sort of. According to the American Heart Association, the hospitalization figures were higher because hospitals may use a broader range of diagnostic codes that don't necessarily include the code used to identify a new case of heart failure. Fair enough. But that raises another question: What exactly were they counting?

Without knowing precisely which codes were used, it's difficult to interpret the hospitalization numbers. This doesn't necessarily invalidate the findings, but it certainly makes the reported threefold increase in hospitalization hazard harder to evaluate.

Don't chuck out the study

Despite these limitations, the study has some legitimate strengths. It included more than 130,000 patients, followed outcomes over five years, and used statistical methods to account for numerous differences between melatonin users and nonusers. The researchers also performed a sensitivity analysis requiring multiple melatonin prescriptions, and the association with heart failure remained.

Furthermore, the differences weren't small. The reported hazard of developing heart failure was 89% higher among melatonin users, while the hazard of death was approximately doubled. These are substantial associations, even if they don't establish causation.

Dismissing the entire study simply because it's observational would be a mistake. Observational research has identified genuine drug safety problems, sometimes long before randomized trials could address them.

Does this study prove that melatonin causes heart failure? No. But are the findings sufficiently concerning to warrant further investigation? Possibly.

I'm still skeptical that relying on melatonin prescriptions as a measure of actual consumption will tell us what's really going on. Mais qu'est-ce que j'en sais? [4]

NOTES

[1] As has my good friend and colleague Dr. Henry Miller. We've even done so together.

[2] Yes, some people have genuine nutritional deficiencies that require supplementation. I'm not talking about them. They get a pass.

[3] The percentages tell us how many patients experienced each outcome during the study. Hazard ratios also account for when those events occurred, not just how many occurred. Therefore, dividing one percentage by another won't necessarily give you the reported hazard ratio.

[4] No, I don't know why I'm writing some of this in French. Je n'en parle pas un traître mot.