What is the Number to Treat?
When it comes to drug studies the number to treat is a most useful number to know, yet the hardest to find, because rarely do drug companies report it. The number to treat provides a good indication of how likely you are to benefit from a drug.
By the end of this article it will be obvious why drug companies do not shout out the numbers to treat in their studies. In most cases you have to calculate the number to treat yourself, and that is something many physicians do not take the time to do, let alone the average consumer. Fortunately, calculating the number to treat is not as nearly complicated as the formula in the above photo. Here is an online tool to help you calculate the number to treat.
How do you know if a drug is truly effective? Pharmaceutical companies are masters are reporting the results of drug studies in the most positive light. And, most physicians fall for the deceptive reporting practices. What does it mean when a study shows a 36.6% relative risk reduction in heart attacks or that a vaccine is 95% effective? The number to treat puts these types of reporting that rely on percentages into perspective.
Very simply, the number to treat is the number of patients who must take a drug over a defined period of time (usually the duration of a drug study) for one person to receive the intended benefit(s) of a drug. A number to treat of 1 means that every person who takes the drug receives the intended benefit. No drug has a number to treat of one. A number to treat of 1.2 means that 12 people must take a drug for 10 people to benefit. A number to treat of 2 means that two people must take the drug for one person to benefit from the drug. A number to treat of 3 means three people must take the drug for one person to obtain the intended benefit of the drug.
So you can see once you get above a number to treat of 2, the chances of benefiting from a drug is no better than a flip of the coin.
Did you know that there are FDA approved drugs that have numbers to treat of 100? That means that 100 people must take a drug for ONE person to benefit from the drug. Would you consider such a drug effective that has a number to treat of 100? Probably not, but there is a pretty good chance you or someone you know is taking such a drug.
Let’s use a real life example to illustrate how the number to treat brings to light the deceptive reporting of drug companies. We will take a look at the drug Lipitor (atorvastatin) which is the most popular drug ever prescribed. But, before we go there we should discuss the difference between statistical significance and clinical relevance.
Statistical Significance versus Clinical Relevance
New drugs are compared to placebos to determine whether or not they are effective. Now a placebo is not expected to show any benefit in a drug study but often they show some benefit. In many depression studies up to 50% of the participants will respond to a placebo, as an example. If a drug performs better than a placebo it is determined to be effective – if that difference is statistically significant. Just because a drug receives FDA approval does not mean that most people who take the drug will benefit from it. It just means they are more likely to benefit from it compared to a placebo (which is not expected to benefit at all). So besting a placebo in drug study in one sense is a low threshold to beat.
Here is a fictious example to illustrate this. For instance, a placebo may grow hair in 1% of the cases and a tested hair tonic grows hair in 3 % of the patients where that difference is statistically signifiant. So that hair tonic could be FDA approved for hair loss even though 97% of the patients are not going to benefit from the hair tonic.
Statistical significance means that the results of a study (drug) are not likely to have occurred by chance alone. It means that the results are real or true. But results of any study could reach statistical significance, and yet not be clinically relevant. Do you consider a number to treat of 100 or as high as even 22,000 to be clinically relevant even if the results are statistically significant? Probably not. Again, statistical significance just means the results of a study are not likely to have occurred by random chance.
Clinical relevance means just that – that the results are relevant or meaningful or have purpose in a clinical setting – that the results can change a clinical outcome. In the end, something can be statistically significant without being clinically relevant, but has to be statistically significant in order to be clinically relevant. Back to Lipitor
Number to Treat: Lipitor
In the ASCOT-LLA study patients who had high blood pressure and three other risk factor for heart disease, but without known heart disease or prior heart attacks, were studied to see if the addition of 10 mg of Lipitor would lower the rate of heart attacks over a 40 month period. This is called a primary prevention trial. If this same study were done on patients who already had heart attacks or coronary bypass surgery or coronary artery stenting, it would be called secondary prevention trial. Here is an article that summarizes the ASCOT-LLA trial and the number to treat for Lipitor.
In the study, for every 100 patients who received the placebo, 3 had heart attacks over the 40 month period or 0.9 heart attacks per 100 people per year. For every 100 patients who received Lipitor over the 40 months, 1.9 had heart attacks. That is 0.57 heart attacks per 100 people per year.
The first thing that should stand out is that for being the leading killer in this country causing more than 600,000 deaths a year, heart disease/heart attacks are still relatively uncommon. Of all the people you know, how many had a heart attack this past year? Many of you probably cannot think of a single person.
Now in its ads, Pfizer promoted a 36% relative risk reduction in heart attacks in those taking Lipitor. This is a true statement. But, is it relevant? Maybe. Maybe not. How do we get to a 36% relative risk reduction. Here’s how. Subtract 1.9 heart attacks in treatment group from 3.0 in the placebo group and we get 1.1. Then divide 1.1 by 3.0 and we get 36.67%.
That’s impressive number — if everyone got heart attacks. But, not everyone gets heart attacks. Only a relatively small number gets heart attacks as we alluded to above. What we really want to know is not the relative risk reduction but the absolute risk reduction. Calculating that is easy. Simply subtract 1.9 heart attacks/100 in the treatment group from 3.0 heart attacks/100 in the placebo group and we get an absolute risk reduction of 1.1%. One point one percent does not sound nearly as impressive as 36% which why pharmaceutical companies rarely mention the number to treat.
For the sake of illustration lets round up the 1.9 heart attack per 100 patients in the Lipitor group to an even 2.0. This means that for every 100 patients that took Lipitor for 40 months only one patient received the intended benefit (3.0 minus 2.0). Thus, in the ASCOT-LLA study the number to treat is 100. Ninety nine of 100 did not get the intended benefit of Lipitor – at least for that 40 months of the study. And, that’s an important point.
The number to treat is a function of time on a medication. The longer you take a medication, the more likely you are to obtain some benefit, but that is offset by the fact that you are also more likely to get a side effect. It has been estimated that for statin drugs like Lipitor that after 30 years of taking a statin drug the number to treat decreases to 7 which is a lot better than 100. This means that for every 7 patients who take a statin for 30 years one will be spared a heart attack. Is that clinically relevant? That’s up to you to decide. Do you want to take a drug for 30 years to have a one in seven chance of benefiting from it? The reality is most patients on statins will get a side effect than be spared a heart attack.
Again, this study only applies to those who are at risk for heart disease but do not yet have it. These numbers do not apply to those who have had heart attacks or coronary artery bypass surgery. It is also important to recognize that the outcome of the study could have looked more impressive if a stronger dose of Lipitor was used in the study.
Now if you hear about 36% risk reduction in heart attacks you would probably would want to be on that drug, right? But, if you hear a number to treat of 100 or even 7 you might think twice about starting a medication.
The sad reality is that many drugs have high numbers to treat with many being in the 20s.
Number to Treat: Pfizer COVID Vaccine
Now that you know something about the number to treat, let’s apply the concept to a situation that affects us all – COVID. I am going to give you some homework. How many patients had to be vaccinated in the Pfizer 6-month clinical trial in order to prevent one COVID death? Keep in mind the vaccine is being touted as being more than 95% effective. What does that mean?
I will direct you to the article (Pfizer’s 6-month trial) and to the supplementary appendix where you can find all the information you need to calculate the number to treat to prevent one COVID death. Post your answer in the comment section.
Let’s see how well you do.
Here is the link to the article published in the New England Journal of Medicine. Once you go to the article, scroll down to the methods section and you will see a hyperlink to the supplementary index. You can link here, if you prefer, to access the supplementary appendix.
Scroll down to Table S4 found on page 11. You should see on the left side Reported Cause of Death. The number of patients who received the BNT-Pfizer vaccine is 21,926 (vaccinated) and the number of patients who received the placebo is 21,921 (unvaccinated). If you see all of that then you are on the right table. For sake of simplification, let’s round up the number of patients in both groups to 22,000.
Here is the million dollar question. What is the difference in the reported cases of death due to COVID 19/COVID pneumonia (note that there are 2 categories – “COVID” and “COVID pneumonia” which you can treat as a single category) between the vaccinated and unvaccinated groups. Divide that number (difference) into the number of patients who received the Pfizer vaccinations and voilà, you have the number to treat.
What is the number to treat to prevent one COVID death via the Pfizer vaccine?
I sadly tell you that most doctors do not know what you are about to uncover.
The clock starts now.
For extra credit. Take the number of COVID deaths in the placebo or unvaccinated group and extrapolate that for 12 months rather than six (the trial study period). Then take that number and adjust/extrapolate for 333,000,000 Americans. How many COVID deaths in the United States in one year does this project? Is that number anywhere close to what is being reported by the media?