Iron supplements, ferritin levels, and VO2max gains in athletes

There’s a new study out this month on iron supplements for athletes with low ferritin. The study, published by Anja Neža Šmid and colleagues at the University of Ljubljana in Slovenia, is a meta-analysis, meaning it pooled data from several different randomized trials that took athletes, assigned them to either a placebo group or an iron supplement group, then measured the difference in ferritin levels between groups after the study’s conclusion.

The results of this meta-analysis are some of the strongest evidence that I’ve seen that support the notion that athletes with ferritin levels circa 20 ng/mL will benefit from an iron supplement. However, I wanted to write up this post because that’s actually not how the authors of the study interpret their own results!

Here’s what Smid et al. say in the abstract:

Increase in serum ferritin concentration after [oral iron supplementation] was evident in subjects with initial pre-supplementation serum ferritin concentration ≤12 µg/l [ ng/mL], while only minimal, if any effect, was observed in subjects with higher pre-supplementation serum ferritin concentration.

Šmid et al 2024

How can I justify the benefits of iron supplementation at 20 ng/mL, while the authors cite a cut-off value of barely half that level? Well, it all comes down to the way you analyze the data.

Here’s the bottom line up front: when treating ferritin level as a continuous number—which it is—the study’s data clearly show that runners with ferritin levels of around 20 ng/mL or less will experience significant increases in ferritin and gains in VO2max when taking an iron supplement over the course of 6–8 weeks.

But don’t take my word for it. Observe:

A plot showing the effect of iron supplements on ferritin levels, as a function of initial serum ferritin levels. Athletes with lower ferritin levels see dramatically bigger gains in ferritin.

What the research says on iron supplementation and ferritin levels

First, let’s back up for a moment and look at the basic outline of this study. Using a standardized search strategy, the authors pulled every study on iron supplementation and ferritin levels that could possibly be relevant.

Then, they applied some filters: studies had to be randomized controlled trials (the gold standard for evidence-based medicine), they had to include people who were at least “recreationally active,” they had to compare an oral iron supplement against a placebo, and they had to record ferritin levels at the beginning and end of the study. Some of the studies recorded other outcomes as well, like gains in VO2max or increases in hemoglobin.

This search yielded 13 studies totaling 449 athletes. In practice, this study’s results are based on female athletes, since aside from a single three-day study on 14 men, 432 of the remaining 435 athletes were women.

The main part of the meta-analysis (which I have no major qualms with) involved pooling the results of each study, weighting them according to the variability in the results. In effect, this process down-weights smaller, less-reliable studies, and up-weights larger and more reliable ones.

The result is the “forest plot” below, which might be a familiar sight if you’ve done any reading on evidence-based medicine.[1]

Forest plot showing meta-analysis of iron supplement benefits on ferritin levels. The results show that iron supplements clearly increase ferritin levels in athletes.

To homogenize differences across studies, meta-analyses often measure their outcome using effect sizes, so the numbers we’ll be looking at here are, unfortunately, not ferritin increases in ng/mL but ferritin increases in “standardized mean differences”, or SMDs. An SMD of 1.0 represents a one standard deviation increase in ferritin within each study’s population.[2]

In the forest plot above, the large black diamond at the bottom represents the pooled results of this “meta-study,” and it shows a powerful effect for iron supplements on ferritin levels. The pooled effect size of 1.27 is pretty big as far as meta-analyses go.

Moreover, the 95% confidence interval—the width of the diamond—very clearly does not include zero, so we can conclude with high confidence that taking an iron supplement does indeed increase your ferritin levels. So far so good.

Ferritin levels and the benefits of an iron supplement

One benefit of pooling data from many studies is that you can look at moderators—factors that alter the relationship between an intervention (like taking an iron supplement) and an outcome (like increases in ferritin).

One very obvious moderating factor when it comes to iron supplement is your initial ferritin level, i.e. your serum ferritin concentration when you started taking the iron supplement.

Because iron uptake in the body is increased when your ferritin levels are low, we’d expect athletes with lower initial levels of ferritin to benefit more from an iron supplement, compared to athletes with higher initial levels of ferritin.

The best way to go about this analysis is where I depart from the approach taken by Smid et al.

Ferritin levels in athletes are not “low” or “high,” they are continuous values

In graduate school I learned two very important lessons about data analysis.

The first was to avoid dichotomizing your data—that means don’t pretend something is a binary yes/no when it’s a continuous number. This applies to things like age (young/old), weight (normal weight/overweight), or footstrike (heelstrike/forefoot strike). These are all continuous numbers, but we're often so used to categorizing them that we forget it!

“Binning” numbers into arbitrary categories is wasteful, statistically speaking, and also creates the illusion of a discontinuity where there often is none.

Ferritin levels are a clear example of a continuous value: it is a number, and we shouldn’t expect any abrupt discontinuities: low ferritin is bad, of course, but very low ferritin is likely to be very bad, and modestly low ferritin should only be modestly bad.

Smid et al., by contrast, dichotomized the data from the iron supplementation research into studies on people with “low initial ferritin” (<12 ng/mL) or “high initial ferritin” (>12 ng/mL). Then they looked at this binary yes/no as a moderator, and concluded that iron supplementation is only beneficial for raising ferritin levels if you are in the “low” category.

Why 12 ng/mL? There’s a bit of a throwaway line in the paper about this cut-off being based on a literature review, but this is one of the big problems with dichotomizing data: where you draw your cut-off is totally arbitrary and can affect your results.

Let’s take a look at how to do this analysis in a way that respects the continuous nature of ferritin levels and that doesn’t require choosing arbitrary cut-points.

The benefits of iron supplementation are a smooth function of your ferritin levels

Smid et al. do include a plot of the initial ferritin values in all 13 studies, which gives us everything we need to re-do the analysis. Here’s the raw study-level data plotted:

A scatter plot of ferritin increase as a function of initial ferritin level in 13 studies on 449 people.

Just eyeballing it, it sure looks like there’s a progressively larger benefit (in SMDs) for taking an iron supplement when your ferritin is low.


Your first inclination might be to fit a line to these data—i.e. to do linear regression—but we’ve now arrived at the second important thing I learned about data analysis in graduate school: assuming something is linear is often a very bad assumption.

Indeed, just thinking from first principles, we’d expect a more drastic response to iron supplementation in athletes with very low levels of ferritin, and a more muted response among athletes with higher levels of ferritin.

Fortunately, modern statistical software makes it very easy to relax the assumption of linearity and let the raw data tell us how initial ferritin levels moderate the benefits of an iron supplement.

Here’s a spline function—a smooth curve—fit to the same data from above:

Meta-regression showing that initial ferritin level is a strong moderator of the benefits of iron supplementation on ferritin increase in athletes.

Viewed this way, it is quite obvious that there are real benefits to taking an iron supplement for athletes with ferritin levels in the 12-20 ng/mL range. They’re not as big as the benefits of taking iron at 12 ng/mL, but the benefits at 12 are also not as big as the benefits at 5 or 10 ng/mL!

Another important thing to notice is that our initial “overall” estimate was only 1.27. That estimate gets dragged downwards by the studies on athletes with ferritin levels >20 ng/mL.

Iron supplementation’s effects on VO2max

Higher ferritin is all well and good, but what about the effects on performance?

Ten of the 13 studies in Smid et al. also measured VO2max. We can do the same analysis, looking at whether initial ferritin level is a moderator of the VO2max increase that occurs as a consequence of taking an iron supplement.

First, here’s the overall analysis, which again matches up with what’s in the original paper.

Forest plot showing meta-analysis of iron supplement benefits on VO2max in athletes. The results show that iron supplements likely lead to an increased VO2max in athletes.

In aggregate, the effects are only marginally significant. But, as with the effects of iron supplementation on ferritin levels, we should expect some strong moderating effects from initial ferritin level: athletes with very low ferritin are probably more likely to see large increases in VO2max when they take an iron supplement.

Here’s the same spline analysis fit to the VO2max data.

Meta-regression showing that initial ferritin level is a potential moderator of the benefits of iron supplementation on ferritin increase in athletes.

Again, we see fairly solid evidence that athletes with ferritin levels below about 15 ng/mL are going to see meaningful increases in their VO2max. The results aren’t quite as strong, in part because we’re going off ten instead of 13 studies.[3]

Some limitations to this analysis

I should note that I don’t take these data to be the end-all be-all of iron supplement research. There’s quite a lot of heterogeneity between the studies included in this meta-analysis, which makes it hard to tell to what extent the results are being driven merely by differences in study design and who participates.

Meta-analyses like this are great, but when looking at moderators, I would much prefer seeing data from the individual athletes. Other fields of research are moving to “Individual Participant Data” to be able to evaluate moderating factors at a more fine-grained level. I’d love to see sports medicine do this as well. You can’t blame Smid et al.—they provided their raw data; it’s the original studies that are lacking in this regard.

I’d be interested in seeing individual-level results from a large study on both men and women, with repeated measurements over time, different dosages of iron, and other improvements. I’ve got an article in the works about what the ideal iron supplementation study would look like—I’ll link to it here when I’m done.

Recap

My analysis of the data from Smid et al. make a strong case that athletes with ferritin levels of ~20 ng/mL or less will see significant gains in their ferritin level if they take an iron supplement. Below ~15 ng/mL, there’s a strong chance that athletes will see VO2max increases as well.

The benefits of iron supplementation are a smooth function of your initial ferritin level—the lower your ferritin, the greater the benefit you’ll see.

Don’t get me wrong: Smid did a great study and I agree with many of the points they make. I especially think their call for more research on intravenous iron infusions is timely, as even the impressive-looking SMD gains from above often did not restore athletes to the “normal” range of ferritin levels.

But I do believe that analyzing ferritin levels for what they are—a continuous number, with no sudden breakpoint defining “low” versus “normal”—leads to results that are more helpful for runners, coaches, and clinicians.

You can find data and code for this analysis on my GitHub.

Footnotes


[1] I made this forest plot in R using the metafor package using the study’s raw data; the 95% confidence intervals differ a small amount from Smid et al, likely due to computational differences between metafor and SPSS, which is what the original authors used.

[2] This study is a great example of why I frequently find effect sizes like SMDs unhelpful—for ferritin, the “clinically relevant” number is the actual measurement! Doctors and coaches know what a 5 ng/mL increase represents, much moreso than a standardized mean difference of 2.0. Admittedly, the studies themselves varied quite a bit in the standard deviation of ferritin within their subject pools, which I expect is for two reasons: (1) some of the studies had different inclusion/exclusion criteria, and (2) biomarkers like ferritin tend to be log-normally distributed, as opposed to normally distributed, which can inflate the standard deviation quite a bit in larger subject pools. This second reason is another reason why I stridently support randomized controlled trials publishing their raw, individual-level data, versus just summary statistics.

[3] As it happens, our p-value in this analysis is now p=0.036, though (again!) we should not obsess over arbitrary thresholds

Related articles

Low ferritin and iron deficiency anemia in distance runners: A scientific guide for athletes and coaches

When I see a runner getting fatigued early on in workouts or struggling mightily in races for no good reason, there's one potential cause I always consider first: low iron levels. Iron deficiency is a significantly underdiagnosed problem in distance runners. Low levels of hemoglobin in the blood, or low levels of the iron storage ... Read more

Notes on Randall Wilber's "Altitude Training and Athletic Performance" for runners

If you are interested in altitude training, Randall Wilber’s book “Altitude Training and Athletic Performance” is a fantastic read. I apologize for the lack of posts recently, but I’m recovering from arthroscopic hip surgery last week and am not up for doing a whole lot of sitting (in fact, I’m typing this out standing at ... Read more

About the Author

John J Davis, PhD

I have been coaching runners and writing about training and injuries for over ten years. I've helped total novices, NXN-qualifying high schoolers, elite-field competitors at major marathons, and runners everywhere in between. I have a Ph.D. in Human Performance, and I do scientific research focused on the biomechanics of overuse injuries in runners. I published my first book, Modern Training and Physiology for Middle and Long-Distance Runners, in 2013.

1 thought on “Iron supplements, ferritin levels, and VO2max gains in athletes”

  1. As a runner, athletics coach and a medical research scientist, who has studied iron homeostasis for several decades, I clicked on the link to this article expecting the normal misinformation that is found on so many websites. So, it was an unexpected pleasure to read such a well-informed analysis of this research article. Keep up the good work!

    Reply

Leave a Comment

Did you know I have a book? Check it out here!