Horizon is a randomized controlled trial. Half of the 8,000 women were randomly chosen to receive the infusions of the drug.
The assumption is that these women will be split so that all important causes of hip fractures would be equally represented in each half: their ages, the thinness of their bones, the degree to which they already had spinal fragility fractures, and even their body mass index.
This was all checked, and indeed, these features split 50-50 between the treated and untreated halves. But what if something important wasn't measured and did not split 50-50?
The most important cause of hip fractures in the elderly is not the thinness of bones. The most important cause is falling down.
Furthermore, the most important cause of falls in the elderly is instability. Attention should be focused on the living environment to remove obstacles that could trip the person up, and to provide railings and other assistive devices that promote stability of gait.
There are promising trials that explore the role of strength and balance training and of hip protectors to decrease the likelihood of hip fractures, but the results are inconsistent.
The reason so many elderly die within the year after a hip fracture is not a result of the fracture itself, which orthopedists can stabilize expeditiously. It's because the instability is but one feature of their decrepitude — it's a reflection of their ripe old age. Their time is near.
The Horizon trial design did not include a measurement of stability. What if instability did not randomize 50-50? What if 51 percent of the most unstable women were in the untreated group and 51 percent of the most stable women in the treated group? That's not too unlikely. Such skew could easily account for a 1 percent absolute difference in the result.
For that reason, I simply discount 1 percent absolute differences in outcomes.
If I have to treat 100 patients in order to get an important benefit in just one of those patients, I consider the treatment to be ineffective. Other clinical scholars start to gain confidence with a 2 percent absolute difference. I require 5 percent; I am comfortable advising my patient that a treatment is beneficial if I can expect benefit in one out of every 20 patients I treat.
In the Horizon trial, as many as a third of the subjects were ill for a few days following the infusion of the drug. There was about a 1 percent absolute difference in decrease in kidney function or a serious problem with heart rhythm; women who did not receive the drug were spared. These small effects are as impressive, or unimpressive, as the hip fracture difference.
However, I worry about the rare effect that would not happen at all without the drug.
The reason this drug can be infused only once a year is that this class of drug enters into the bone and stays there for a very long time, all the while influencing the fashion in which the cells in the bone try to keep the bone healthy.
There is also a rare, unique and horrifying complication that is well described. Drugs of this class can cause the bone to die, particularly the jaw bone. I am not reassured by a three-year experience in 4,000 women. I am particularly wary since I am not convinced of any benefit.
For Whose Benefit?
The debate on efficacy commences at a 2 percent absolute difference in outcome.
The misleading nature of relative results, such as relative risk reduction, has been widely decried.