Researching collagen to help his achy knees, a statistician explores the painfully weak evidence

One Thursday in September, I woke up with achy knees. That was nothing new. I’ve had knee pain on and off for a few years. Before the pandemic, I had visited an orthopedist, who told me that I’d lost some cartilage in my knee joint — an early sign of osteoarthritis.

“Come on, I’m just over 50,” I told her. “That’s much too young for arthritis.”

“Yeah,” she laughed, “I hear that a lot.”

My knees aren’t nearly bad enough for surgery, but I’ve tried most everything else. A steroid injection provided relief, but only for a week. Injections of rooster comb (it’s for real) didn’t help me, and certainly didn’t help the roosters. Physical therapy worked when I did the exercises religiously, but I’d lapsed in recent months.

So I thought I might try collagen supplements.

Collagen is a protein that makes up two-thirds of the dry weight of cartilage in knees and other joints. It’s also a major component of tendons, ligaments, and skin. With age, the body makes less collagen, and that helps explain a lot of what happens to skin and joints — including knees. In 2019, consumers worldwide bought $1.8 billion in collagen supplements, hoping to replace some of what they’d lost.

There’s a certain common sense to the idea, but it’s not clear how the collagen in oral supplements is supposed to get from the mouth to the knees. Whether collagen is ingested from supplements or from foods such as eggs, bone broth, or chicken wings, digestion breaks collagen down into its component amino acids and peptides. To make new collagen, the body has to put those components back together again — and that’s exactly what the body gets worse at with age. It’s a bit like bringing extra mortar to a construction site after the bricklayers have gone home for the day.

A 2012 Science magazine article on cartilage described it as “intrinsically unable to heal.” Despite decades of research, involving technologies much more sophisticated than taking collagen supplements, the goal of “cartilage regeneration remains elusive.”

Yet if you search popular health sites for collagen and arthritis, you can find just about anything you want to hear. Penn Medicine, quoting a staff physician, claims that collagen “may help fight the symptoms of osteoarthritis in the knee with few side effects.” VeryWellHealth claims the research is “mixed but shows promise.” A WebMD page on collagen says that “thanks to a small but growing body of evidence suggesting it can improve skin, ease arthritis symptoms, promote wound healing, and fend off muscle wasting, former skeptics in the medical field are…beginning to come around.” Yet on another WebMD page, a distinguished professor of rheumatology dismisses the idea that collagen can “help you grow cartilage. … It does not,” she says.

Perusing these sites, I got a little confused. But my confusion was only beginning.

Since even the most skeptical WebMD page said that collagen supplements “won’t hurt if you want to try them,” I figured I might as well order some. I started shopping on the internet and within minutes I was overwhelmed. Collagen supplements come in many brands and forms — powders, liquids, capsules, tablets, gummies, “super” collagen, vegan collagen, hydrolyzed collagen, “beauty collagen,” collagen for hair loss, and more. I had no idea what kind I should take. And whatever I chose, I had no idea how much I should take, how often I should take it, or how soon I should expect relief.

I turned to the scholarly literature, and quickly found a meta-analysis, published in the journal International Orthopaedics, claiming to summarize the most rigorous research on oral collagen supplements and osteoarthritis. The authors found 114 published articles and winnowed them down to just five that had assigned patients at random to take collagen or a placebo. After calculating a weighted average of the trial results, the authors concluded that “collagen is effective in improving [osteoarthritis] symptoms.”

Halfway through the article, though, I was starting to have doubts. I’m a statistician who’s published research on meta-analysis. This one, as I pointed out in a letter to the journal that had published it, was full of red flags.

The first red flag was that the results were all over the place. One article reported large benefits of collagen supplements, two reported much smaller benefits, and two found no benefit at all. Trials can yield different results for various reasons, but none of the obvious explanations fit the results on collagen. For example, trials with larger doses or longer treatment periods didn’t report larger effects. To the contrary, trials that lasted just 10 to 13 weeks claimed greater benefits than trials that lasted 24 to 48 weeks. And trials in which patients took just 2 to 10 grams a day claimed larger effects than trials in which patients took 10 to 40 grams a day.

The results didn’t make sense. I had the same questions as when I started: If I ordered a collagen supplement, how large a dose should I take? How soon I should expect relief? I still had no idea, and I was starting to think that no one else knew, either.

If dose and duration didn’t explain the disparity in results, what did? The only thing that seemed to matter was the number of patients in the study. An article reporting two very small trials, each of which had just 18 to 19 patients taking collagen, appeared to show dramatic benefits. But the largest trials, with 54 to 111 patients taking collagen, found that the supplements improved symptoms by less than five points on a scale that ran from zero to 100 — a difference so small that patients typically say they feel “much the same.”

That’s important, because studies with more patients tend to give more precise estimates. The larger studies were probably closer to correct in suggesting that collagen didn’t make patients feel noticeably better. So why did the smaller studies suggest larger benefits?

To check, I examined the article that claimed the largest effects and immediately saw a huge red flag, right on the title page. Four of the five authors were employed by Nitta Gelatin, “a trusted producer of gelatin and collagen ingredients for the food, dietary supplement, and pharmaceutical markets.”

That turned out to be not the exception, but the rule. As I learned with a little digging, all five articles in the meta-analysis were either funded by a collagen company or included some authors who were being paid by a collagen company, either as employees or consultants.

Industry-sponsored research isn’t necessarily biased but, for a variety of reasons, it is more likely to suggest that the sponsor’s products are beneficial. Among the reasons is what’s called “reporting bias.” Instead of conducting one large study, a company can sponsor a bunch of small studies and only publish the ones that appear to show benefits. Because the published studies aren’t typical of what most studies found, they tend to exaggerate the benefit that consumers can typically expect.

In recent years, the U.S. and some other countries have taken steps to reduce reporting bias in clinical trials. One of the most important steps is called preregistration. Authors planning a clinical trial must state publicly and in advance how the trial is designed and what results they plan to report. Preregistration is required by a variety of organizations, including the Food and Drug Administration, the National Institutes of Health, the World Health Organization, the European Union, and the International Committee of Medical Journal Editors.

But collagen, as a minimally regulated supplement, slips through the cracks of many preregistration requirements that apply to drugs. And two of the collagen trials were conducted in Mexico and Ecuador, where no registry exists in which researchers can preregister trials.

Just one of the five collagen articles was preregistered. It concluded that collagen might have benefits, but the results showed patterns that I found hard to understand. True, the group taking collagen showed dramatic improvements over the course of the six-month trial — but so did the control group, which took placebo pills that weren’t supposed to help. In fact, the improvements in the placebo and collagen groups differed by a practically unnoticeable amount — again just over five points on a scale that ran from 0 to 100. Fewer patients in the treatment group reported taking acetaminophen (Tylenol), but that was true even in the first week of the study, when the treatment and control groups were reporting practically identical levels of pain. Although another collagen trial also found dramatic improvements in both the treatment and control group, I found the idea of large improvements in the control group hard to swallow. Osteoarthritis is not a condition that usually gets better on its own.

Regardless of the results, a trial in which only 54 participants took collagen wouldn’t be nearly enough to get collagen approved if it were classified as a drug rather than a supplement. A trial that small would be a Phase 2a trial, also known as a pilot or proof-of-concept trial. Still to come would be the Phase 2b trial to estimate the optimal dose, and then Phase 3 trial to test the drug in a larger group of patients. Phase 3 trials typically enroll hundreds or thousands of participants — more if the effect is small, as the effect of collagen on arthritis likely is. If there is an effect at all.

All this brings me to a question that has bothered me since I started trying to sort out the evidence. Why is the research on collagen seemingly stuck at a preliminary stage? If executives of collagen companies truly believe that they are selling a product that relieves symptoms of knee osteoarthritis — a condition that afflicts more than 600 million adults worldwide — why haven’t they sponsored an independent, preregistered trial with hundreds or thousands of participants, the kind of trial whose results could be published in a top medical journal?

If these executives thought there was a decent chance of positive results, sponsoring a trial would be a much better use of money than anything they might spend on advertising or marketing. With a large and rigorous clinical trial to support the use of collagen for osteoarthritis, the market for collagen would expand dramatically. Doctors would routinely recommend it. Patients would experience relief. Companies making these supplements would profit handsomely, and the executives who sponsored the trial could write their own tickets.

The fact that there hasn’t been such a trial suggests that collagen executives aren’t at all confident about how it would come out. They’re not actually sure that their product works. And if the people who sell collagen supplements don’t really believe in them, why should anyone with achy knees buy them?

Paul T. von Hippel is a statistician and professor with the Center for Health and Social Policy at the LBJ School of Public Policy, University of Texas, Austin.