Randomized trial data suggests that the medication, which strengthens bones, increases the risk that they will break.

I will show you a graph without any labels.

What could this line represent? Maybe it’s the stock price of some company that made a big splash but failed to live up to expectations. Maybe it’s an outbreak curve – charting the introduction of a new infectious agent to a population. Maybe it’s the performance of a particularly viral tweet?

I’ll tell you what it is in a moment, but I wanted you to recognize that there is something inherently wistful in this shape – something that speaks of past glory and inevitable declines. It’s a graph that induces a feeling of resistance – no – do not go gently into that goodnight.

What the graph actually represents, roughly, is the normal level of serum testosterone in otherwise healthy men as they age.

A caveat here, these numbers are not as well-defined as I made it seem on the graph – particularly for those above age 65. But it is clear that testosterone levels decline with time, and the thought of supplementing testosterone is hardly new. But like all treatments, testosterone supplementation therapy has risks and benefits. Some risks are predictable – like supplementation exacerbating symptoms of BPH. Some risks seem to come completely out of left field. And that’s what we have today in a study that suggests that testosterone supplementation increases the risk of bone fractures.

Let me set the stage here by saying that nearly all prior research into the effects of testosterone supplementation has suggested that it is pretty good for bone health. It increases bone mineral density, bone strength, and improves bone architecture.

So, if you were to do a randomized trial of testosterone supplementation and look at fracture risk in the testosterone group compared to the placebo group, you would expect the fracture risk would be much lower in those getting supplemented. Of course, this is why we actually do studies instead of assuming we know the answer already – because in this case, you’d be wrong.

I’m talking about this study, appearing in the New England Journal of Medicine.

It’s a pre-specified secondary analysis of a randomized trial known as the TRAVERSE trial, which randomized 5246 men with low testosterone levels to transdermal testosterone gel vs. placebo. The primary goal of that trial was to assess the cardiovascular risk of testosterone, and the major take-home was that there was no difference in cardiovascular event rates between the testosterone and placebo groups.

This secondary analysis looked at fracture incidence. Researchers contacted participants multiple times in the first year of study and yearly thereafter. Each time, they asked if the participant had sustained a fracture. If they answered in the affirmative, a request for medical records was made and the researchers, still blinded to randomization status, adjudicated whether there was indeed a fracture or not, along with some details as to location, situation, and so on.

The breaking news is that there were 154 confirmed fractures in the testosterone arm, 97 in the placebo arm. This was a big study though – that translates to just a 3.5% fracture rate in testosterone versus 2.5% in control, but the difference was statistically significant.

This difference persisted across various fracture types – non-high impact fractures, for example- and after excluding the small percentage of men taking osteoporosis medication.

So the question is… um… what? How does a drug that increases bone mineral density and bone strength increase the risk of fracture?

Well, one clue – and this was pointed out in a nice editorial by Matthis Grossman and Bradley Anawalt – is that the increased risk of fracture occurs quite soon after starting treatment – which is not consistent with direct bone effects. Rather, this might represent behavioral differences. Testosterone supplementation does seem to increase energy levels – might it lead men to engage in activities that put them at higher risk of fracture?

Regardless of the cause, this adds to our knowledge about the rather complex mix of risks and benefits of testosterone supplementation, and probably puts a bit more weight on the “risks” side. The truth is that testosterone levels do decline with age – as do many things – and it may not be appropriate to try to fight against that in all people. It’s worth noting that all of these studies use low levels of total serum testosterone as entry criteria. But total testosterone is not what your body “sees” – it sees free testosterone – that which is not bound to sex hormone binding globulin. And that binding protein is affected by lots of stuff – diabetes and obesity lower it for example – making total testosterone levels seem low when free testosterone might be just fine.

In other words, testosterone supplementation is probably not terrible, but it is definitely not the cure to aging. In situations like this, we need better data to guide exactly who will benefit from the therapy, and who will only be exposed to the risks.

A version of this commentary first appeared on Medscape.com