Bisphosphonates are one of the most commonly prescribed drug classes. Their intended purpose is to strengthen bones, and thereby prevent fractures. They are primarily prescribed to postmenopausal women with evidence of bone thinning, either because a bone scan has shown that they have a weak skeleton, or because they’ve experienced a fracture from relatively low energy trauma. Examples of drugs in this class are alendronate, risedronate, and zoledronic acid. Basically, any drug that has a name ending in “-dronate” or “-dronic acid” is a bisphosphonate.
The mechanism of action is pretty ingenious. Our skeleton is continuously recycling itself throughout our lifetimes, and the work of recycling is primarily carried out by two types of cell: osteoblasts and osteoclasts. Osteoblasts build new bone and osteoclasts destroy old bone. What bisphosphonates do is to selectively poison and kill osteoclasts. This changes the balance of bone recycling to one that favours build-up of bone, which is why bisphosphonates have been unequivocally shown to make bone thicker.
But do they actually prevent fractures? And what are the downsides to a drug that inhibits the normal recycling of bone?
There are about a bazillion different studies of bisphosphonates, so for the sake of brevity and simplicity, I’m going to focus in this article on just a few, which are representative of the whole.
The first study we’re going to talk about was published in JAMA (the Journal of the American Medical Association) in 1998. It was a four year long randomized controlled trial in which roughly 4,300 post-menopausal women with evidence of bone thinning on scans were randomized to receive either alendronate or placebo. Alendronate is a bisphosphonate that is taken as a pill, either as a lower dose once a day or a higher dose once a week. In this particular study the participants were taking the lower dose once a day.
The study was funded by Merck, the company that held the patent for alendronate. Unfortunately, big independent studies of bisphosphonates don’t exist, so we’re forced to rely on what the pharmaceutical industry feeds us. It’s been shown that industry funded studies systematically overestimate benefits and underestimate harms, so the results presented here should be considered best case scenarios. As I’ve written about previously, there is evidence that in the real world, side effects are three to four times more common than they appear to be in industry funded randomized trials.
Another thing worth noting is that 1998 is pre-2000, which means that the study was carried out before the birth of clinicaltrials.gov, and before there was any serious expectation that companies register trials before beginning them. In theory that means that Merck could have run five different studies, and only published the results from the one the provided the best results, while “disappearing” the other four. In this particular case, that is unlikely, for the simple reason that this was a 4,000 person study, which is probably just too damn big and expensive to “disappear” if the outcome proves unsatisfactory.
The fact that it’s a pre-2000 study however also means that Merck could easily have engaged in various post-hoc manipulations of the study design, in order to maximize apparent benefits and hide harms, without anyone being any wiser. That’s unfortunately something we have to live with. So, again, the results should be considered best case scenarios.
Anyway, let’s take a look at what those results were.
14.1% of women in the placebo group experienced an osteoporotic fracture during the four year study period, as compared with 12.3% in the alendronate group. This gives a 1.8% absolute reduction in risk of a fracture after four years of treatment, which means that you would need to treat 55 women for four years in order to prevent one fracture.
On closer inspection, however, the alendronate only appeared to provide a benefit to those with the very weakest skeletons (defined as having a bone density more than 2.5 standard deviations below the mean bone density of a healthy young person). Among these women, 19.6% of those in the placebo group experienced a fracture over the four years, compared with 13.1% in the alendronate group. This gives a 6.5% absolute reduction in risk, which would mean that one in 15 women treated for four years would benefit from the treatment.
That really isn’t bad for a drug that is taken for the purpose of prevention. For comparison, statins generally require that 200 people be treated for five years in order to prevent one heart attack. And blood pressure lowering drugs require that 70 people be treated for five years to prevent one stroke. So by those standards, one in fifteen odds of benefit are really good.
Alendronate somewhat prevented the loss of height that gradually happens due to compression of the spine when you have a weak skeleton, by about 1.5 mm over four years. The placebo group lost 8.5 mm over four years, while the alendronate group lost 7 mm. The result is statistically significant. Clearly, however, 1.5 mm over four years is a difference that is too small to be noticeable to the patient, so it isn’t clinically significant, and thus isn’t a reason to take the drug.
Ok, so now we know the size of the benefit. What about harms?
Unfortunately, the information on harms provided in the trial report is extremely limited. We are told that 1.8% of participants in the placebo group died, compared with 1.7% in the alendronate group, so there is no signal that the alendronate is causing excess deaths, which is good. There were slightly more hospitalizations in the alendronate group, with 29.1% in the alendronate group being hospitalised, compared with 26.9% in the placebo group. This is concerning, because the whole point of taking alendronate is to avoid ending up in hospital, and this suggests that the alendronate actually increases the risk slightly. Of course, the increase is small enough that it could just be due to chance, but it certainly doesn’t suggest that the alendronate decreases risk.
The only more specific information that is provided on harms in the trial report concerns stomach issues. The reason stomach issues are given special attention is that earlier studies have found that bisphosphonates when given daily can cause inflammation in the esophagus. There was a signal of that here also, with 10 cases of esophagitis in the placebo group, compared with 19 in the alendronate group, which would, if taken at face value, mean that roughly one in 230 patients develop esophagitis as a result of the treatment (although the small absolute numbers make it impossible to draw any firm conclusions).
Nowadays, bisphosphonates are often given weekly in order to avoid the risk of esophagitis. Unfortunately, weekly dosing has instead been found to cause its own set of side effects: aches in muscles, bones and joints during the days immediately following intake of the drug. Researchers at the Charité hospital in Berlin looked at the frequency of these side effects in their patients treated weekly with alendronate or risedronate and found that the risk of muscle pain, joint pain, or bone pain developing during the first 48 hours after a weekly oral dose, and which is severe enough to cause the patient to decide to stop taking the bisphosphonate, is 20-25%. If one in four to one in five patients have aches and pains severe enough that they decide to stop the drug, then it’s reasonable to think that there are many more who experience at least some aches and pains as a result of the bisphosphonate, but who don’t decide to stop the drug.
There is an upside to this though – daily dosing causes one type of side effect, and weekly dosing causes another type of side effect. Since it’s unlikely that it’s the exact same group that experiences the stomach issues and the musculoskeletal pain issues, that means that someone who experiences stomach issues with daily dosing can try switching to weekly dosing instead, while someone who experiences aches and pains with weekly dosing can try switching to daily dosing.
Let’s move on to the next study. Like I said earlier, bisphosphonates can either be taken as a pill or be given as an annual injection. With this in mind, let’s take a look at a trial that was published in the New England Journal of Medicine in 2007. Like the previous trial, it was double-blind, randomised, and placebo-controlled. And like the previous trial, it was funded by the company that held the patent on the drug, in this case Novartis.
8,000 postmenopausal women with osteoporosis were randomised to receive either an injection of zoledronic acid or placebo once a year, and then followed for three years to see whether they had any fractures. So, what were the results?
At three years, 12.8% of participants in the placebo group had a clinically noticeable (i.e. symptomatic) osteoporotic fracture, compared with 8.4% in the zoledronic acid group. That gives an absolute reduction in risk of 4.4%. In other words, for every 23 postmenopausal women with osteoporosis that are treated with zoledronic acid for three years, one fracture is prevented. That is very similar to the benefit seen in the previous trial.
In terms of height loss, the patients in the zoledronic acid group lost an average of 4 mm of height, while the patients in the placebo group lost 7 mm. So that’s a 3 mm difference after three years, which is more impressive than the previous trial, but still so little that it’s unlikely to be enough to cause a meaningfully noticeable difference in height, even with extended use.
What about side effects?
Serious adverse events (i.e. death, hospitalisation, or lasting disability) were very similar between the two groups. 30.1% experienced a serious adverse event in the placebo group, and 29.2% did in the zoledronic acid group. So, unlike in the previous trial, there’s no signal that the drug increases the risk of a serious adverse event, which is encouraging. However, there were slightly more deaths in the zoledronic acid group than in the placebo group (3.4% vs 2.9%), which isn’t. Both sets of differences are so small though that they can easily be due to chance.
People who received the injection with zoledronic acid were quite likely to experience influenza-like symtoms (fever, muscle pain, joint pain, headache) during the days immediately post-injection. After the first of the three annual injections, fully 32% of patients in the zoledronic acid group experienced these symptoms, compared with only 6% in the placebo group. With each additional injection, this side effect became less and less common, however.
People might be willing to put up with a few days of influenza-like symptoms for improved odds of avoiding a broken bone. What’s more concerning, however, is that the study participants on zoledronic acid were more likely to develop atrial fibrillation than the participants in the placebo group, with 50 serious events in the zoledronic acid group, compared with only 20 in the placebo group. This works out to roughly one in 260 people who are treated with zoledronic acid developing a serious case of atrial fibrillation (i.e. requiring hospitalisation) over three years of treatment.
Of course, in a study you look at lots of different adverse events, and so a difference between the groups in any one might appear just by chance. However, in this case the risk does appear to be a real – a meta-analysis of observational studies and randomised trials of bisphosphonates published in 2013 found that they do significantly increase the risk of atrial fibrillation.
There are two very serious side effects associated with bisphosphonates that I haven’t yet mentioned – osteonecrosis of the jaw (a debilitating condition in which the jawbone breaks down) and atypical hip fractures (fractures that happen in weird parts of the hip, that are never seen in people not treated with bisphosphonates). Does this massive trial with 8,000 people shed any light on how frequent these side effects are?
Well, in the trial there was one case of osteonecrosis in the placebo group, and one case in the zoledronic acid group, so no signal of harm. With 8,000 people in the trial, that is reassuring, suggesting that osteonecrosis caused by zoledronic acid is very rare (at least in otherwise healthy elderly women with osteoporosis). So we have to look elsewhere for information on how frequently osteonecrosis occurs in patients treated with bisphosphonates.
An observational study carried out in Australia found that osteonecrosis occurs in roughly one in 4,000 people treated with bisphosphonates, although this rises massively to one in 500 people who have a tooth extraction while undergoing treatment. That is why it’s generally recommended that patients have a dental appointment and get any bad teeth taken out before they start taking bisphosphonates. The risk of osteonecrosis rises even higher to one in 100 people who have cancer in their bones. So for this group bisphosphonates are quite risky.
What about atypical hip fractures? Both this trial and the one I discussed earlier found a reduction in overall hip fractures among those treated with bisphosphonates, with a decrease from 1.1% to 0.9% over four years in the first trial, and a decrease from 2.5% to 1.4% over three years in the second trial. No specific information is provided about atypical hip fractures, but whatever increase in atypical fractures is seen with the drug appears to be more than made up for by a decrease in regular hip fractures.
Ok, so let’s sum up. With the caveat that we are as usual forced to rely on industry funded studies, bisphosphonates do appear to result in a meaningful reduction in risk of fractures in patients with radiographically confirmed osteoporosis (i.e. a T-score of -2,5 or worse), with roughly one in 20 treated patients benefiting after a couple of years of treatment.
What about harms? Well, that’s where it gets complicated, and where benefits and risks need to be weighed carefully. The alendronate trial showed a two percent increased risk of ending up in hospital overall, while the zoledronic acid trial showed a one percent decreased risk. This suggests that the net effect of these drugs on risk of hospitalisation is somewhere close to zero. That would mean that the overall benefit in terms of a decreased risk of experiencing a fracture is made up for by an increased risk of other things (atrial fibrillation, esophagitis, osteonecrosis, etc). So a conversation really needs to be had with the patient about which set of risks they’d rather be taking.
Even if the patient does decide to take the bisphosphonate, a high protein diet prevents just as many fractures as bisphosphonates do without the risks, and resistance training also offers massive benefits, decreasing the risk of a fracture while improving overall quality of life. Bisphosphonates should therefore primarily be used as a complement to lifestyle changes, not as a stand-alone treatment.
Please support my work by becoming a patron. Your support is what allows me to dedicate time to producing new content. Regardless of whether you can contribute a lot or a little, it all makes a difference – the more patrons I have, the more of my time I am able to dedicate to this work. As a bonus, patrons get access to my private forum, and also gain the ability to send me private messages through Patreon – I always respond to patrons.
21 thoughts on “Bisphosphonates: pros and cons”
I am a 70-year-old male who has just been diagnosed with severe osteoporosis, despite having normal levels of calcium throughout my history and until recently playing a lot of sport. I suspect my previous high fructose high carb diet may have led to osteoporosis – I’m carnivore now.
I have purchased a vibration platform and propose to use that as the primary method of compensating for my lack of exercise – I have Parkinsons and severe scoliosis so I’m very sedentary.
I have an appointment with an endocrinologist and will be interested to see what they recommend. You have made me much more able to decide on a plan of action!
Thank you Sebastian
I love your articles, I conclude that not taking meds is the best medicine, so far so good, I’m a 74 year old female, fat, and dieting…eating organic. Blessings to you!
Excellent article. Lays out the facts very well .
Did the authors report what the fractures were ie hip, spine ,wrist as hip fractures are more severe and need in patient care.
Excellent review. I would have expected more atypical fractures. Remember these were industry studies. Thanks, Roger MD
While on this subject. Maybe worth reviewing vitamin K2, its role to strengthen bones. Studies carried out appeared positive?
Japanese studies on K2 Mk-4 (the synthetic version) found a positive effect. No side effects either. Dosing was 15 mg three times a day. Active in body for about 2 hours.
The Mk-7 natural version (activity lasts a day or so) dosing is 15 mcg/three times a day.
I appreciate your thoughtful approach to this and other issues. My background as a nephrologist and bone expert has given me insight into the bisphosphonate scam.
The problem with these drugs, and all the antiresorptives, is their MOA. Osteoclasts are crucial for bone remodeling, and their inhibition is often terminal, with apoptosis. Blocking bone remodeling leads to fragility, and is clearly the explanation for the atypical fractures, only seen in patient taking bisphosphonates. These fractures involve cortical, as opposed to trabecular bone.
Osteopenia is WHO-invented disease based on arbitrary bone density scores. Bone density numbers are not adequate to determine bone quality and strength. There have been shockingly few bone histology studies of these agents.
I did publish a case series of bone biopsies in patients with varying degrees of chronic kidney disease who were given bisphosphonates. (Amerling R, Harbord NB, Pullman J, Feinfeld DA: Bisphosphonate Use in Chronic Kidney Disease: Association with Adynamic Bone Disease in a Bone Histology Series. Blood Purif 2010;29:293–299 DOI: 10.1159/000276666). 100% of our patients had adynamic bone.
I share your suspicion of industry-funded studies. The ARR for fractures is probably less than the studies report, but even at 2% does not justify the risk of a cortical femoral fracture.
It’s akin to giving drugs for depression that increase the risk of suicide!
Treatment of osteoporosis with antiresorptive medications is an example of treating numbers via the junk science called EBM. No one should be prescribing or taking them.
Now their advocating treatment for osteopenia where the benefits, if there are any, are even smaller. The impression is that they restore bone, they don’t. The type of fracture is important a fractured scaphoid is nowhere as dangerous as a fractured hip. When I last looked most of the fractures were spine and peripheral.
I was always sceptical about the impact of osteoporosis in the community however our betters pointed out we’d been overlooking it for years!!
Sebastian – as a maxillofacial surgeon I can personally attest to a high incidence of avascular necrosis of the the jaw associated with medication with bisphosphonates , especially when taken due to osseous metastatic malignant disease ie. in prostatic carcinoma .
I was diagnosed with osteoporosis in my 40’s and was on Fosomax for 11 years. I took myself off it when I realized it was doing no good and has horrible potential side effects i.e. jaw necrosis. Early onset osteoporosis is a red flag for celiac disease which I was diagnosed with in my 60’s, but the possibility of celiac disease was never mentioned. I should have been on a gluten free diet instead of Fosomax! Over the last few years I have taken control of my health, done my own research, and now feel better than I have in years. You must be your own advocate!
Good analysis to refute the bone doctors’ one sided push for drugs. I’ve noticed that doctors down-play adverse effects and pump-up benefits just like pharma, like they’d do if shilling for pharma. Never do they mention the number needed to treat. Instead they let you assume that all get benefits.
They must be thinking, “Well, who knows, maybe there’ll be some benefit, and if not I can adjust the dose or take them off it. Only one way to find out and I’ve got nothing to lose for trying.”
I’m convinced every doctor thinks that with every drug. And if the person dies, well you can’t win them all and that’s what malpractice insurance is for.
In the article you make the following comment.
“blood pressure lowering drugs require that 70 people be treated for five years to prevent one stroke”
Have you written an article on Blood pressure meds? If so would you please provide a link.
I’ve written a couple of articles about blood pressure lowering drugs. Here: https://sebastianrushworth.com/2021/03/07/blood-pressure-lowering-drugs-pros-and-cons/ and here: https://sebastianrushworth.com/2020/07/30/how-aggressively-should-high-blood-pressure-be-treated/
Have you considered interviewing a specialist in integrative medicine on the topic of combining conventional allopathic and alternative treatments of high blood pressure?
Interesting article. Got my mum off this poison a few years ago & now her bone density has improved back to “normal”. As you say, a good diet is better than dodgy drugs.
I’ve approached this topic from a position of skepticism toward our current practice of screening for osteoporosis and its management. I believe that this is another area of medicine in which we are overdiagnosing and overtreating based on low quality evidence.
Regarding your analysis of the 1998 NEJM paper by Cummings et. al.,
I think it is important to lead with the fact that this was a negative study. They failed to meet their primary endpoint of all clinical fractures. The difference of 1.8% absolute risk reduction in clinical fractures was not statistically significant.
There were positive findings within subgroups, such as a statistically significant reduction in fractures in women who had low BMD (T score < -2.5). Though this finding raises optimism for tailoring therapy to those with the lowest BMD, this study was not designed to answer this question with confidence.
My understanding of biostatistics is limited, but I’ve learned that as a general rule, subgroup analyses and secondary endpoints are hypotheses generating at best, and not to be practice changing data.
You have more experience critically appraising literature, so may I ask: What are appropriate conclusions one can draw from positive findings of secondary endpoints or subgroup analyses?
In contrast, the study you introduced regarding zoledronic acid produced a positive outcome for the primary endpoint of hip fractures, which also happens to be clinically relevant.
One question I had was why they had separated the patients into different stratums. Stratum one was for patients with no concurrent use of osteoporosis medications. Stratum two was for patients taking concurrent osteoporosis medications throughout the study.
Interestingly, the two primary endpoints, vertebral fractures and hip fractures, were assessed with different patient populations (i.e stratum one vs stratum 1+2, respectively).
Would you know why this was done, and if is this typical for trials of this type? I can’t help but imagine that there was some post-hoc analysis and manipulation unless this type of protocol is routine.
Regardless, it’s not standard of care for women in this population to be treated with two concurrent medications for osteoporosis. I’m not sure how to interpret this positive finding for most patients I encounter who are treatment-naive, or have failed alendronate due to intolerance.
This study design can’t help but raise suspicions. It makes me wonder if this protocol was pre-registered to a database prior to the conduction of the study. Yes, the primary outcomes were pre-defined in clinicaltrials.gov but I don’t know if the investigators pre-specified stratums one and two for measuring the two primary endpoints.
Finally, in regards to the treatment of osteoporosis diagnosed on screening, do you know of any other studies that showed that an intervention improved clinical outcomes? E.g. decreased risk of hip fractures, hospitalizations, or death, as opposed to merely a reduction in asymptomatic vertebral fractures found on xrays.
Most importantly, I’d appreciate any feedback/ criticism you have about the way I’m approaching these questions. All in all, I appreciate your conclusion that ultimately what needs to be stressed in clinic should be protein intake and resistance training, regardless of a patient’s decision to proceed with pharmacotherapy.
You make a lot of very good points. If you do a study, you should base your conclusions on the primary end points, not the subgroup analyses. It prevents the authors of a study making dodgy claims. However, when I’m doing my own analysis, I’m not constrained by what the authors considered their primary end point. I look at what I consider the most important end point. In bisphosphonate studies the authors generally choose to look at vertebral fractures as diagnosed by screening, since this shows the best effect, even though it isn’t a clinically relevant end point. I instead look at clinical fractures (i.e. symptomatic fractures), hospitalizations, and deaths, since these are clinically relevant. And I look at the patients with osteoporosis (T-score less than -2.5), because that’s the only group where it’s really reasonable to expect a benefit. If I was the study author it would be bad practice to focus in on a subgroup in this way, but I’m not, which is why I can. I pretty much always ignore what the authors choose as their primary endpoint and instead look at what I consider to be the most important endpoint. If I ran the study, I would have done it differently, and only looked at people with osteoporosis and made clinical fractures my primary end point.
Since this study is so massive, even the subgroups provide pretty reliable data (which often isn’t the case, often the subgroups become too small to draw any meaningful conclusions).
As to doing different analyses on different groups, as they’ve done in the zoledronic acid study, I agree, that’s fishy. My guess would be that they thought it would be easier to show a difference on screening x-rays in patients not on other osteoporosis drugs. They appear to have limited to stratum 1 for the morphometric fractures, while using the entire patient population for the clinical fractures. That means good conclusions can still be drawn concerning the clinical fractures, which are the only thing I care about personally.
Nothing to do with this article. Excuse me! I like how soberly and factually you argue. I would be very interested in an article from you on the subject of vaping. One could argue that if all smokers were to switch to vaping, significantly more lives (or lifetimes) could be saved than ever could be saved by all Corona measures. Maybe I’m wrong. It is very difficult for a layman to distinguish the good studies from the bad.
Thank you Sebastian. Great review. For any who may read this comment, Lani Simpson, DC has written a great book on the subject which basically comes to a similar overall conclusion.
Ok, let’s get some more comments on this post. Maybe time to Jump the Shark?
Recall that the FDA issued a recall of its EUA for hydroxychloroquine and recommended only for hospital use and this was its reasoning…
“Hospitalized patients were likely to have greater prospect of benefit (compared to ambulatory
patients with mild illness) and could be more closely monitored for potential toxicity, although it
was recognized that enrollment in a clinical trial would be the best option when using these drugs
so that data on safety and effectiveness could be obtained.”
(note the date…this is a revision of the previous revocation back in May which relied mostly on the Surgisphere study)
Does anyone have any comments about this statement from the FDA? I know I do, but I’ll wait for others to weigh in first.
(Yes, this was the same revocation that relied on the fraudulent Surgisphere study.)