THIS ARTICLE is a SUPERB assessment of physicians overtreating patients. (A corollary to this would be over-ordering fancy expensive tests to rule out exotic diagnoses, but that’s for another day.) Recent personal experience includes one 86-year-old Caucasian male concomitantly taking 26 prescription medicines; another includes a geriatric patient taking (daily) 13 prescription medicines plus 15 over-the-counter (OTC) medications.
The topic [below] is also clearly within the realm of GEITP — because diagnosis and treatment of patients for various complex diseases (multifactorial traits) involves gene-environment (GxE) interactions. And we all know that complex diseases (which can also be called “genetic architecture”, i.e., the underlying genetic basis of a phenotypic trait and its variational properties) are manifested by: genetic differencess, epigenetic factors, environmental effects (includes drugs as well as other chemicals), endogenous influences (i.e., concurren lung, heart, kidney etc. disease), and each individual’s microbiome.
This topic also falls under the realm of “Personalized Medicine,” which has recently been renamed by some as “Precision Medicine.” What we like most about this article is the quantitation [in number of days’ increased longevity (i.e., benefit) — versus number of days of potentially shortening longevity (i.e., detriment) due to one or more adverse drug reaction(s) (ADRs)]. Other caveats not covered in this beautiful article include cost/benefit ratios and drug-drug interactions (DDIs). ☹
Should the patient really get the drug?
Sebastian Rushworth, M.D.
Jun 14, 2022
I recently gave a lecture to 70 primary care physicians here in Stockholm, titled “should the patient really get the drug?”. The lecture generated quite a bit of cognitive dissonance among some in the audience, based on the somewhat aggressive discussion that followed the lecture, which suggests to me that much of what I was saying was stuff they had literally never been exposed to before – not at any point in medical school, and not at any point during their careers after medical school either. Cognitive dissonance is good. It’s the first step towards change.
I thought it would be interesting to re-write the lecture as an article, so that more people can hopefully achieve similar levels of cognitive dissonance. Please feel free to share it with any doctors you know that you think might benefit from an expanded perspective. Anyway, here we go.
Let’s imagine a common patient. Every primary care physician meets this patient, or someone much like her, on an almost daily basis. She’s 75 years old, and overweight. She experienced a wrist fracture two years ago, and was subsequently diagnosed with osteoporosis. She has high cholesterol levels, but she’s never had a heart attack or other “cardiovascular event”. On top of that, she has type-2 diabetes, chronic knee pain due to osteoarthritis, and high blood pressure. She was diagnosed with depression a few years ago, after her husband died.
Our patient takes seven drugs every day:
1. Alendronate, because of her weak bones.
2. Atorvastatin, because of her high cholesterol levels.
3. Sertraline, because of her depression.
4. Metformin, because of her type 2 diabetes.
5. Insulin, also because of her type 2 diabetes.
6. Paracetamol (a.k.a. acetaminophen), because of her knee pain.
7. Enalapril, because of her high blood pressure.
So, the question is, are these drugs doing her any good?
Well, to answer that question, we need to consider NNT (Number Needed to Treat). NNT is the number of patients who need to take a drug for one patient to achieve a noticeable benefit.
For alendronate, the NNT is 20, i.e. if you treat 20 people for a couple of years, you prevent one fracture. For atorvastatin the NNT is 200, i.e. you need to treat 200 people for five years or so in order to prevent one heart attack. For sertraline, the NNT is 7, which means that you need to treat seven people in order to have a noticeable effect on depression in one patient. Note that this doesn’t mean that one out of seven gets cured of their depression, it just means that there is a noticeable difference on a rating scale for depression.
For metformin, the NNT is 14 – If you treat 14 type 2 diabetics with metformin for ten years, you prevent one death. For enalapril, the NNT is 70 – If you treat 70 people with high blood pressure with enalapril for five years or so, you prevent one stroke.
For insulin, however, there is no NNT, because insulin has not been shown to result in any benefit on any clinically relevant outcome, even though big studies have been carried out that have included thousands of patients and followed them for five or ten years. Note here that we’re talking about insulin for type 2 diabetics. When it comes to type 1 diabetes, insulin is pretty much magical – you don’t even need to do a randomised trial in order to show benefit. People with type 1 diabetes virtually return from the dead when treated with insulin. But when it comes to type 2 diabetes, there is no benefit, at least not to any hard outcomes. All insulin has been shown to do is reduce blood sugar, but it’s never been shown to result in any meaningful patient oriented benefit for type 2 diabetics.
The same is true for paracetamol/acetaminophen. When it comes to patients with knee pain due to osteoarthritis, the drug doesn’t provide any benefit whatsoever.
Ok, so we have seven drugs, and we know what their NNT’s are. If we plus the probabilities of benefit together, then we get the probability that our 75-year old woman will benefit in some way from at least one of the drugs she’s taking. So, what probability of benefit do we get?
We get 30%. Only 30%.
What that means is that there is a 70% probability that this woman doesn’t benefit at all from any of the seven drugs that she takes every day for years on end!
If you told her, I’d say there are pretty good odds she’d decide to stop taking her pills. Seven drugs a day, every day, and two to one odds of zero benefit.
And we haven’t even talked about harms yet. Because none of these pills are inert. All have widespread biological effects. And all can cause harms. So any rational treatment decision must include not just the potential benefits, but also the potential harms.
For figuring out harms, we have NNH (Number Needed to Harm), which is the counterpoint to NNT. NNH is the number of patients who need to get a drug for one to be harmed. Like I said, the drugs all have widespread biological effects, so there isn’t just one NNH – there is an NNH for each possible harm. That means that there are multiple NNH’s for each drug.
With our 75-year-old woman and her seven drugs, we don’t have time to go through the NNH for every possible side effect, so we’re just going to look at a few, and put them side by side with the NNT, to get a somewhat more complete picture of benefits vs harms. I’ve tried to make sure that the NNH numbers apply to the same time period as the NNT numbers, since otherwise it’s an apples to oranges comparison.
If we do that, we get something like this:
NNT: 20 (fractures)
NNH: 200 (esophagitis), 260 (atrial fibrillation), 4,000 (osteonecrosis)
NNT: 200 (cardiac infarction)
NNH: 20 (myalgia), 20 (type 2 diabetes)
NNT: 7 (depression)
NNH: 2 (sexual disturbance), 10 (hyponatremia)
NNT: 14 (death)
NNH: 2 (stomach upset), 5 (B12 deficiency), 1,000 (lactic acidosis)
NNT: 70 (stroke), 125 (death)
NNH: hyperkalemia (10), acute kidney failure (100)
NNT: 0 (no benefit to clinically relevant outcomes)
NNH: severe hypoglycemia (5), weight gain (1)
NNT: 0 (no benefit to clinically relevant outcomes)
NNH: Hypertension (30), liver damage (?)
It’s possible to quibble here about specific NNT and NNH numbers. Different studies show different things. And many of the numbers come from studies carried out by pharmaceutical companies, which generally means that the risk of a certain side effect is massively underestimated (as we will discuss shortly). The point here isn’t to get hung up on any of the specific numbers. It’s to illustrate that we quickly end up with a very complex equation, where it in many cases isn’t clear at all whether the benefits outweigh the harms.
Take alendronate, as an example. We know that it decreases fractures in elderly osteoporotic women. But it doesn’t decrease hospitalisations. The only reasonable conclusion is that the reduction in hospitalisations that is seen due to the reduction in fractures is made up for by an increase in hospitalisations due to the many and varied side effects. So at the end of the day the only way to decide whether or not to take the drug is to have a detailed discussion with the patient and get them to decide which set of risks they’d rather be taking.
Hippocrates is supposed to have said “primum non nocere”, which is latin for “first, do no harm”. Actually he didn’t say that, and couldn’t have even if he wanted to. Hippocrates was greek, and didn’t speak latin. The quote comes from a 19th century American physician, Worthington Hooker.
Of course, as doctors, we all know that “first, do no harm” is completely unrealistic. Every intervention we do carries som measure of risk. If our primary guiding principle was to never do harm, we literally would never be able to do anything. A more reasonable principle is “only do something if the benefits clearly outweigh the risks”. If it isn’t clear to you that the benefits of a drug outweigh the harms, then don’t give it to the patient.
That’s a good general rule to stick by. However, it probably isn’t enough, for a few reasons we’re now going to discuss.
A study was published in JAMA Internal Medicine in 2021 that sought to establish how good physicians are at estimating the likelihood that a patient has a certain disease. 500 primary care physicians in the US were asked to consider various hypothetical scenarios, and then answer what they thought the probability of disease was. Here’s an example of a scenario that they were asked to consider:
Ms. Smith, a previously healthy 35-year-old woman who smokes tobacco presents with five days of fatigue, productive cough, worsening shortness of breath, fevers to 102 degrees Fahrenheit (38.9 degrees centigrade) and decreased breath sounds in the lower right field. She has a heart rate of 105 but otherwise vital signs are normal. She has no particular preference for testing and wants your advice.
How likely is it that Ms. Smith has pneumonia based on this information? ___%
Ms. Smith’s chest X-ray is consistent with pneumonia. How likely is she to have pneumonia? ___%
Ms. Smith’s chest X-ray is negative. How likely is she to have pneumonia? ___%
Go ahead and make your own guesses in relation to each of the three questions.
Once you’ve done that, you can take a look at the table below, and the answers will be revealed.
So, for our pneumonia example above, we see that the actual initial risk of disease based on the presented information was around 35%. If we then move along and look at what the doctors answered, they thought the risk was 80-85%. In other words, they thought pneumonia was more than twice as likely as it actually was!
The same phenomenon was seen in all clinical scenarios tested. The doctors consistently overestimated the initial risk, and they continued to overestimate the risk after both a positive and a negative test result. In some cases the difference between reality and what the doctors thought was huge, with the doctors overestimating risk by a factor of ten or more.
What can we conclude from this?
Doctors consistently overestimate disease risk.
Hold that thought, as we move on to take a quick look at another study, which was published in BMJ Open in 2015. This study sought to do something about a problem inherent in statin trials (and for that matter, all trials in medicine), which is that the results they produce, in the form of percent absolute risk, percent relative risk, and NNT, are so abstract that they’re completely meaningless to patients (and for that matter, to doctors as well). We know that statins have an NNT of 200 when used for primary prevention (to prevent a heart attack in someone who has risk factors but hasn’t already has a heart attack), and 40 when used for secondary prevention (to prevent additional heart attacks in someone who has already experienced a heart attack). But what do those numbers actually mean? Are they good or bad?
What the patient really wants to know is “how much longer will I live if I take this drug?”
So, what the researchers did was gather together data from all the big randomised trials of statins, and use the survival curves provided to estimate how much longer the patients actually lived. Here’s what they came up with:
All the big statin trials are included here. What’s interesting to do is look at the NNT provided, and then compare that with the number to the right of it, which is how much longer the patients actually lived, on average. So, for the ALLHAT trial, to take the topmost example, we have an NNT (for primary prevention) of 250, which comes down to a postponement of death of 4.96… well, 4.96 what?
Is it years? No.
Is it months? No
The patients in the statin group lived 4.96 days longer than the patients in the placebo group. That is what the NNT of 250 means in real terms.
Let’s look instead at 4S, which was published in 1994 and is the statin trial that has produced the best results of any statin trial ever. It’s the trial that initiated the massive boom in statin prescribing that we still see today. In 4S, the NNT (for secondary prevention) is 27.8. So, in other words, one in 27.8 patients benefited from the treatment.
But what does that actually mean in terms of life extension?
It means 27 days.
Not as impressive as you would have thought, right?
When the researchers put all the data together, from all the trials, in order to get an overall average, what they found was that when statins are used for primary prevention they prolong life by 3 days. When they are used for secondary prevention, they prolong life by 4 days.
I can imagine quite a few patients turning down the offer of a statin if they knew that it will on average only prolong their life by days.
The purpose of bringing up this study was to illustrate the following general point:
Doctors consistently overestimate the benefit of the drugs they prescribe.
Hold that thought in your mind as we move on and look at a third study.
This one was published in The Lancet Healthy Longevity in 2021. It compared the rate of serious side effects seen in randomised trials with that seen in the real world. If randomised trials give us good information about what to expect in reality, then the rate of serious side effects in the trials should be the same as that seen in reality.
But that isn’t what the researchers found. What they found was that serious side effects were three to four times more common in reality than they are in the randomised trials! Three to four times!
How is this possible?
Well it’s important to remember that the randomised trials are funded and run by the drug companies, and the drug companies want to sell their drugs, so they will do what they can to make side effects appear as rare as possible.
Why is this a problem? Because it’s the randomised trials that doctors mostly use as a basis for determining whether a drug is safe to give to a patient or not.
So, what can we conclude from the study?
Doctors consistently underestimate side effects of drugs.
Ok, so we have three conclusions, that are all pointing us in the same direction:
1. Doctors consistently overestimate disease risk.
2. Doctors consistently overestimate drug benefit.
3. Doctors consistently underestimate drug harm.
What does this lead to?
Massive overprescribing of drugs.
Peter Gotzche, a founding member of the Cochrane Collaboration and former director of the Nordic Cochrane Center, has estimated that prescription drugs are now the third biggest cause of death in the western world, after heart disease and cancer.
That on its own should lead to massive humility among all doctors about our drug prescribing. It should make us much more careful every time we think about prescribing a drug to a patient.
Ok, so we’ve identified the problem. The causes of this problem are many and complex, so I’m just going to bring up one that each of us as doctors can actually do something about – industry sponsored meals.
A study was published in JAMA Internal Medicine in August 2016 that sought to estimate the extent to which physicians are influenced by partaking in industry sponsored meals, which often take the form of a lecture about a specific drug given by an drug company salesperson, which the physician is supposed to sit and listen to in return for getting a free meal. Industry sponsored meals are very common. Most physicians probably take part in at least a couple of these per year, and many take part in far more than that.
As the saying goes, “there’s no such thing as a free lunch”. The drug companies are not charities whose goal it is to keep starving doctors alive. If they spend vast sums of money of sponsored meals, it’s because they’re pretty damn sure that it increases sales of their drugs, and thereby their profits.
So, anyway, the study sought to estimate the extent to which industry sponsored meals influence physician prescribing patterns, by comparing participation in such meals with later prescribing behaviour. Here’s what they found:
They looked at four different drugs. As I think is clear from the tables, participation in industry sponsored meals increased prescribing of the drug the meal was about, and the more such meals a doctor participated in, the more often he or she prescribed that drug.
The purpose of these meals is not to educate us, or make us better doctors. It’s the opposite – the purpose is to make us do a specific profit-driven company’s bidding. And it works.
If you’re a doctor, and you think you don’t get influenced by participating in industry sponsored meals, then you are very naive. The more industry sponsored meals we participate in, the worse doctors we become.
Doctors in general massively underestimate the extent to which their thoughts, beliefs, and opinions are influenced by the pharmaceutical industry. We like to think that we are evidence based. But the truth is that much of what we think we know is not based on sound scientific knowledge, but on pharmaceutical industry propaganda, which quickly becomes clear to anyone who starts going through the studies in detail themselves.
On that note, I strongly recommend reading these three books, all written by physicians, to help get some perspective on the scale of the problem we face in relation to the pharmaceutical industry.
1. Bad Pharma by Dr. Ben Goldacre
2. Doctoring data by Dr. Malcolm Kendrick
3. Deadly medicines and organised crime by Dr. Peter Gotzsche
There is one very simple thing every doctor can do, to at least partially free themselves from the onslaught of drug company propaganda, and that is to refuse to take part in industry sponsored lunches, and all other forms of industry sponsored “education”. Just say No.
Ok, so, that’s number one: refuse to take part in industry sponsored lunches.
What else can you do as a doctor?
Well, something that was once considered standard, but has fallen by the wayside in recent decades, is to never have a patient on more than five drugs at the same time. With drugs, as with everything else, there is a state of diminishing returns – the more you add, the less benefit (and more harm) each additional drug confers. So try to keep a patient on at most five simultaneous drugs. If you want to add a sixth, then rank them all, and get rid of the one that you think is least important. Most likely, the sixth least important drug in a list of six is not going to do anything useful for the patient anyway, just increase their risk of harm.
Ok, so that’s number two: try to avoid having your patients on more than five drugs simultaneously.
Number three: go through the patient’s drug list with them once a year, and get rid of anything that isn’t clearly conferring a benefit. As any doctor will know, it’s common for patients to stay on drugs for years, even though the original reason they were put on the drug resolved itself a long time ago. The patient often doesn’t remember why they were put on the drug in the first place, but they keep taking it dutifully. Drug lists require regular pruning or they will become increasingly bloated as the years go by, which is one reason why so many elderly people are on 15 simultaneous drugs or more.
Number four: only prescribe a drug if the benefits clearly outweigh the harms. This should be obvious, but it requires a deep knowledge of the size of both potential benefit and potential harm, which unfortunately most doctors lack. And what they think they know is often incorrect because it’s based more on pharma propaganda than real science.
As a doctor, the only way to get around this is to start doing your due diligence and getting in to the weeds of the scientific studies. Do that for the ten drugs you prescribe most commonly, so that you’re an expert on those ten drugs, and you’ve already done a lot. If a patient asks you about the probability of benefit and the probability of harm, you should be able to answer that question correctly, at least for the ten drugs you use most frequently. It requires an up-front investment of time, but it will pay massive dividends to your patients over the remainder of your career.
Ok, so that was number four: only prescribe a drug if the benefit clearly outweighs the harm.
Here’s number five: prioritise lifestyle changes. Most of the diseases that doctors spend most of their time dealing with are caused by poor lifestyle choices. And most can be rectified by switching to good lifestyle choices, which invariably produce greater benefits than any drug can, with less risk of harm.
Doctors can accomplish a lot with their patients with simple lifestyle coaching. To take one example, a primary care clinic in the UK decided to try putting their type 2 diabetic patients on a ketogenic diet, since the drugs they were using clearly weren’t making the patients better. They published their six year follow up results in BMJ Nutrition, Prevention, and Health in 2020.
Over six years, the patients following the ketogenic diet decreased their median HbA1c (a measure of average blood sugar over the preceding few months) from 66 to 48. Normally, that would be unheard of. HbA1c doesn’t decrease over time in a type 2 diabetic. It increases. Yet here it was far better at the end of the six years than at the beginning. The same goes for body weight. Normally it goes up over time. But here the median decreased from 99 kg to 91 kg. And on top of that, median systolic blood pressure dropped from 152 to 141.
All this just with a simple diet intervention. Thanks to the improvements in all health markers, the patients were able to get off a lot of their drugs. This meant that after six years, the clinic was spending less than half as much money on anti-diabetic drugs as the other primary care clinics in the region.
To take another example of a simple lifestyle intervention, a randomised trial published in BMJ in 2021 that was carried out in nursing homes in Australia found that a diet high in protein has an effect on fracture risk that is equivalent to that seen with bisphosphonates.
There is a massive amount that can be accomplished with simple lifestyle interventions, and since they are much less risky than drugs, and actually treat the underlying problem rather than just putting a patch on top of it, they should be the primary intervention we use whenever possible. Drugs should be viewed as a complement to lifestyle interventions. It shouldn’t be the other way around.
Ok, so that was my fifth and final point. I’ll repeat the five points here again. These are five things that you as a doctor can do about the situation we currently find ourselves in, where prescription drugs are the third biggest killer in the western world:
1. Refuse to participate in industry sponsored lunches and other industry sponsored “education”.
2. Try to avoid having your patients on more than five drugs simultaneously.
3. Go through the patient’s drug list with them once a year, and get rid of anything that isn’t clearly conferring a benefit.
4. Only prescribe a drug if the benefits clearly outweigh the harms.
5. Prioritize lifestyle changes.COMMENT
COMMENT: Dear Dan:
I read the entire piece and indeed it was excellent.
A couple of comments:
1) It would have been nice if the author had posted a link as to where non-MDs might find NNT for pharmaceuticals. ——John, this URL should help you: https://www.thennt.com/thennt-explained/
2) It would have been nice if the author had explained the important difference between relative benefit and absolute benefit. ——The best URL on this topic that I could find was:
https://dearpandemic.org/difference-between-ar-and-rr/ —DwN 😊