This is what happens when legitimate research and medical science, big-pharma, high end medical equipment manufacturers, the government and private insurance industries, compliant doctors, and statistics all converge on a single topic. As you might expect the outcome is less than ideal. That is not exactly correct. It is quite ideal for all involved but one, the older woman incorrectly diagnosed with osteoporosis, told she must take medication for the remainder of her life, and instilled with a sense of fear that at any moment she might fall and fracture a hip, or a leg, or an arm.
My 74 year old mother approached me the other day holding two pieces of paper in her hands. I had known that a few years earlier she had been diagnosed with osteoporosis and honestly had not given it a second thought. Like most Americans I just assumed she was like any other woman her age and that it was really no big deal. Mom is just getting older. It happens. For her it was a big deal. In fact I could see the fear in her eyes as she handed me the pages. Danny, she said, can you look at these for me and tell me what they mean?
My eyes scanned the first page very quickly. A few things immediately caught my attention. First I saw the words: diagnosis: osteoporosis and just below that I saw the words, fracture risk: high. It was this part that really hit me. My mom told me she was really worried about physical activity and was not sure what to do. She was scared of fracturing a bone and thus had been purposely limiting her physical activity. Of course this is the exact opposite of what a person in this position should do. Building muscle strength is one of the best ways to compensate for reduced bone density.
Now I was a little bit angry. So I started to really focus. I was looking at a DEXA scan report produced by a company called Hologics that makes and markets these machines. Essentially it is a bone density measuring device. No need to get into the gory details of how it works. Suffice to say this technology is considered the definitive tool for measuring bone density and thus for use in diagnosis of osteoporosis. There are other large medical machine manufacturers such as GE that market similar devices. They all work on similar principles and are generally considered extremely precise. In fact some have suggested they are the most precise mass marketed medical instrument ever produced. I looked at some of the published reproducibility data for a couple of them and I have to say it was mostly rock solid with some notable exceptions. I can only dream of having such reproducible diagnostics technology in my own field.
Having said all that, precision is only one of the metrics by which we evaluate a particular diagnostic test or technology. After all an instrument that generated the exact same value again and again would be very precise even if the value it produced was nowhere near the actual value. Accuracy is the other key component. For DEXA scans this is a bit more complicated. DEXA measures bone density in real time. In order to determine if the DEXA value matches the actual true value (as determined by various well accepted and robust physico-chemical methods) we would need to excise and process the bone of the measured subject immediately following DEXA analysis. Obviously, unless we are in Nazi Germany, such experiments cannot be conducted on living humans. They are also of limited value with human cadavers for various complicated reasons I will not discuss here. Luckily for us (not so lucky for them) our good friend the lab mouse has let us do these types of experiments. I will leave aside all the arguments about the usefulness of mouse models for drawing conclusions about humans and simply accept at face value an absolute correlation between the two.
When these experiments have been done the accuracy values have been only ok. Acceptable but nothing to write home about. So we have a very precise fairly accurate diagnostic technology for measuring bone density. What do bone density measurements actually tell us? This is also a very long and complicated story but ultimately it boils down to this. Reduced bone density seems to correlate strongly with increased risk of fracture. One of the major sequella of osteoporosis is reduced bone density. Therefore reduced bone density is a strong indicator of osteoporosis. Thus DEXA, which can precisely and accurately measure bone density, should be an excellent tool for diagnosing osteoporosis.
Logically this chain of reasoning is suspect but for purposes of this piece I will accept it. The next question you might be asking yourself is reduced bone density compared to what, compared to who? Now we approach the heart of the issue. The way that a diagnosis of osteoporosis is determined is by the use of a statistical test. Two different but very similar statistical tests. The T-score and the z-score.
The T-score compares the patients bone density vs the average for the young normal population. The Z-score is the age adjusted factor. It compares a patients bone density versus the population of their peers of similar age. The scores are also weighted by ethnicity (at least for the Hologic diagnostic system they used for my moms scan), which may inflate the theoretical risk even more than is warranted. Other instrument platforms such as GE also factor in patient weight.
Irrespective of how they are calculated, the importance of these scores, in particular the T-score cannot be emphasized enough. It is this one number, and this one number only, that totally and completely determines the diagnosis of osteoporosis. If you meet or exceed the threshold value you have the disease and need treatment. Moreover ones “fracture risk” is based entirely on it.
Ultimately, it seems these scores are more important in terms of determining if insurance will pay for treatment vs. an honest assessment of risk. To understand why I say this you need at least a little deeper understanding of the statistics behind these scores. In essence t and z scores reflect a deviation from a mean value. In the case of t scores, deviation from the mean bone density of the youthful healthy population. The z score represents the same but the mean is determined by the average bone density of the patients peer group (i.e. other persons of the patients same age).
Ok, so what? In traditional diagnostic testing there is a generally agreed upon requirement that for a particular effect to be deemed statistically significant it must show a consistent and reproducible deviation from the mean of at least 3 standard deviations. The 3SD rule has been around forever and is a long standing core principle among scientists and statisticians. For reasons of which I cannot seem to find any basis, in the case of bone density testing, this rule has been modified. Changed so that any deviation of 1SD or greater is now deemed significant. Significant enough in fact to trigger a diagnosis of disease, indicate treatment, and require that insurance pay for it. There is little motivation to challenge this assumption among any of the impacted parties. After all they all benefit from more cases of disease, more treatment. One would think the insurance companies might protest. A few did for a while. However once Medicare agreed that this was appropriate and acceptable the private insurance industry had no choice but to acquiesce.
When pressed, which has almost never happened, the pushers of this technology and statistical approach have responded by arguing that the precision of their test method is such as to warrant a tightening of the 3SD model. This is clearly bullshit. The precision of any test has almost nothing to do with its performance in terms of deviation from the mean true value. For tests that push the boundaries of detection sensitivity this argument may come in to play. However when the test method is only as sensitive as the reference method, deviation from the mean accepted value may occur with even the most precise test. Going back to my earlier example. A precise wrong value is still a wrong value.
As an example consider the case of my mom. Her spine T-score average of -2.1 puts her into the clinical definition of osteroporosis. As a reminder a -2.1 translates to a bone density that falls 2.1 SD below the mean of the young healthy population. Sounds bad right? As I discussed above, historically, and for any other diagnostic test, at any other time in history, this would have been dismissed as insignificant. No concern, no disease, no treatment. Her z-score was only -0.1. In other words her spinal bone density is essentially average for her age. A similar story can be told about her hip scores. If you do the math, and believe me I did, it works out that almost half of all women my moms age should have osteoporosis.
After I did this calculation I had to pause. That cannot be fucking right? So I did a quick Google search on the topic. It turns out that the Cleveland clinic estimates 3o% of post menopausal women have osteoporosis and close to half have osteopenia a milder form of the disease. Guess how they determined these estimates? Based on the exact same calculations I did from published t-scores, and more importantly, using the clinical definition of disease that says only a 1SD deviation from the mean is significant enough to diagnose a disease state. Frankly I don’t give a shit what they want to call it. Osteoporosis is ok by me. What I am mostly concerned about is the supposed association of this ‘disease” and fracture risk. After all what most bothered my mom and me about her diagnosis was not the disease itself but rather the implied increased risk of fracture it carries with it.
I talked through many of these arguments with a very intelligent person for whom I have great respect. He listened patiently, mostly agreed with my points, but then said the following: But what about all the studies showing correlation between osteoporosis and bone fractures? Yes these exist. However, one could just as easily substitute the word “post menopausal women” for osteoporosis in the title of these studies and reached the exact same conclusions. Older women are at increased risk for fractures. That is a truism. Older people in general have an increased risk of many bad things happening.
If disease is from now on to be diagnosed by comparison of an older persons “fill in the blank measurement” with that of a young “healthy” persons similar measurement than the menu of diseases will expand infinitely. Ultimately old age itself will be defined as a disease. Many researchers already have come to this conclusion. once that is accepted the focus will become how to diagnose “old age.” What set of markers or age ranges will be considered old and thus diseased? Moreover if the statistical requirements for deviation from the mean continue to shrink the disease model of aging will grow even faster.