Most people will naturally assume that when a doctor prescribes them a drug, it’s because the doctor thinks they will receive a meaningful benefit from it. Most people have never heard the term NNT, which stands for Number Needed to Treat, or to put it another way, the number of people who need to take a drug for one person to see a noticeable benefit. It’s a bit of a counterintuitive concept for people outside medicine, since most people probably assume the NNT for all drugs is 1, right? If I’m getting this drug, it must be because it is going to help me. Well, wrong.
Before we move on, I want you to perform two small thought experiments:
Say you were suffering from depression, and there was a drug that could potentially improve your mood. But it’s not certain that the drug will work for you. And there’s a catch – the drug has side effects, which you are likely to experience regardless of whether you get the benefits of the drug or not. This particular drug causes a reduction in sexual desire and increased difficulty achieving orgasm during sex.
It also causes subtle changes to your personality, making you more prone to take risks, less emotional, and less empathic. It increases your tendency to engage in addictive behaviours, and it’s been known to cause addictions to alcohol and gambling. Additionally, withdrawal is common, so many people have trouble getting off the drug once they’re on it.
How good would the NNT for this drug need to be for you to be willing to take it? Would you want absolute certainty that it would end your depression, considering the harms? or would 50:50 odds be enough? Or even less?
Keep whatever odds you decide on in mind. If you, for example, think one in two odds are good enough, then that gives an NNT of 2 (you need to treat two people to get a noticeable benefit in one of them).
Ok, next scenario. Say you’d had a heart attack, and there was a drug that could decrease your risk of another heart attack. But just as with the previous drug, there are no certainties that you will actually get any benefit from taking this drug. And this drug also has side effects. Many people who take the drug develop chronic aches and pains. The drug also causes noticeable cognitive impairment in a proportion of those taking it, and some even end up being diagnosed with dementia – how big the risk is unfortunately isn’t known, because proper studies haven’t been carried out that could answer that question. Additionally, the drug causes blood sugar levels to rise, resulting in type 2 diabetes in around 2% of those taking the drug – it is in fact one of the most common causes of type 2 diabetes.
How good would the NNT for heart attack prevention need to be for you to take this drug?
Again, keep the number in mind.
As many of you have probably guessed, the first drug I described is an SSRI (examples of this type of drug are sertraline, citalopram, and fluoxetine). Currently, around 15% of adults in western countries take an SSRI every day.
So, what is the actual NNT for SSRI’s when used as a treatment for depression?
In other words, you need to treat seven people for one to experience a noticeable effect on their depression. The other six just get the side effects but no benefit. And when I say “effect”, I don’t mean that the depression resolved in the one person lucky enough to see a benefit. Far from it. I mean that on a certain numerical rating scale (MADRS, if you must know), they experienced an improvement in mood that was just big enough to be detectable using statistical methods.
What NNT number did you choose? Are 7:1 odds good enough for you to take an SSRI if you get depressed, knowing the harms?
When a doctor prescribes an SSRI to a depressed patient, they (hopefully) know that the odds of the patient benefitting even slightly are only 1/7 (or 14%). Which doesn’t seem like a very good deal to me. Yet SSRI’s are widely considered to be an “effective” drug.
The second drug, as many of you have probably also guessed, is a statin (examples include atorvastatin, simvastatin, and pravastatin). More than a quarter of adults over the age of 40 take a statin every day in western countries.
So, what is the NNT for statins?
Well, if you’ve already had a heart attack, i.e. you’ve already been established to be at high risk for heart attacks, then the NNT over five years of treatment is 40. In other words, 39 of 40 people taking a high dose statin for five years after a heart attack won’t experience any noticeable benefit. But even if they’re not the lucky one in 40 who gets to avoid a heart attack, they’ll still have to contend with the side effects.
What NNT did you decide on personally? Are 40:1 odds good enough for you to decide the benefits of a statin outweigh the harms?
Of course, patients rarely get presented with this type of information, and are thus rarely able to make an informed choice of their own. I once sat in on a conversation between a cardiologist and a patient who’d recently had a heart attack. The patient was skeptical about statins. He said that he’d read on the internet that they had side effects, and he wasn’t sure he wanted to take one.
The cardiologist gave the patient a long, withering stare, and then responded that there’s a lot of misinformation on the internet, and that the statin was the number one most important thing he could do if he wanted to not die prematurely.
Which I thought was a bit arrogant. Why?
Because the probability that the statins would prevent a future heart attack, let alone premature death, was in the low single digits, and the patient might quite reasonably have felt that that marginal benefit was outweighed by the various harms (which the cardiologist incidentally hadn’t mentioned at all – and which the patient thus wouldn’t have even known about if he hadn’t read “misinformation” on the internet).
Doctors have been conditioned by the pharmaceutical industry to think that drugs that provide very low probability of benefit are effective. An NNT of 10 is often considered good, and an NNT of 5 is considered excellent. Even an NNT of over 100 is often considered acceptable! Patients are rarely informed that the odds of them getting any benefit from the new drug they’re being prescribed are far less than 50:50. And they’re rarely informed about what the harms are, and how likely they are to experience them.
Just in case you think I’m picking on a few particularly ineffective drugs with my two examples, I’m not. NNT’s of five or worse are typical for many of the most commonly prescribed drugs.
What that means is that the average 70 year old who is on five drugs continuously will probably at best only benefit in any measurable way from one of those drugs. The other four are not providing any benefit, they’re just contributing to side effects (which become increasingly likely, and increasingly deadly, the older you get). Things get even worse when you consider that drugs interact in unpredictable ways to increase the risk of side effects, so the risk of harms increases exponentially with each additional drug added. Which is why it used to be considered bad form to have a patient on more than five drugs simultaneously.
The number of drugs the average person is on has increased massively over the last few decades. Polypharmacy (people taking multiple different drugs continuously) is now one of the top five leading causes of death in the western world – which is a little ironic when you consider that people are taking all those drugs in order to live longer. The best way to avoid becoming another polypharmacy death statistic is to be careful about which drugs you take, and only take those for which it’s clear that the benefits outweigh the harms.
From my perspective, a good drug is a drug for which the benefits clearly outweigh the harms. I’m not saying that all drugs with high NNT’s are inherently useless. A drug with an NNT of 40 might be worth taking, if the risks of harm are sufficiently low and the outcome is sufficiently important. Only the patient can make that decision.
Whether a drug is good for you as an individual is clearly context specific. The decision whether or not to take a certain drug requires a deep understanding of the drug (provided by the physician) and a deep understanding of personal values and wishes (provided by the patient). It requires a holistic perspective and a meeting of two minds that is literally the opposite of what doctors are asked to practice today, where we’re continuously pestered with various treatment guidelines and targets that turn physicians in to unthinking automatons and patients in to featureless blobs.
PS. I’m genuinely curious to know what you, dear reader, think would have been reasonable NNT’s for the two examples discussed above. Please post in the comment section, along with your reasoning.