Recently I read an article in The New York Times that highlighted some of the important information that drug ads often fail to disclose. As illustrated in the tweet below, the part that really got my attention was that it drew attention to the fact that drug ads rarely discuss the number of patients who need to take a drug before a benefit can be seen.
— The New York Times (@nytimes) April 25, 2016
In health care, we call this the number needed to treat (NNT), a concept that helps us describe the effectiveness of a treatment in practical terms. Even the most effective therapies will only benefit a fraction of the patients that receive it. For others, a therapy may have no effect at all or it may even be harmful. In the latter case, the number needed to harm (NNH) describes the number of patients who need to receive treatment before one of them is harmed by it.
When I teach students about analyzing a clinical trial, I emphasize NNT and NNH as being two of the most important concepts to understand, as they support clinical decision-making and aid in communicating to patients the relative benefits and harms of a treatment. The Times article was the first time I had seen this concept outside of the biomedical literature, and it occurred to me how valuable it would be if non-health professionals knew how to determine NNT/NNH and interpret what it means to patients.
Determining the Number Needed to Treat
To determine NNT, there must be a difference between the treatment and placebo (or the alternative treatment if two treatments are being compared), and it has to be statistically significant – a term used to describe when the difference is not likely a result of random chance. Pharmaceutical manufacturers must demonstrate this before they can claim a drug is effective. The NNT can then be calculated in three simple steps:
- Determine the event upon which the trial is based (for example, a reduction in the number of heart attacks). Subtract the percentage of patients who had an event despite receiving treatment from the percentage of patients who had an event while receiving placebo (or the alternative treatment). This is known as the absolute risk reduction (ARR).
- Divide 1 by the ARR.
- Multiply the resulting number by 100%. This is the NNT.
Here is a practical example:
The NNH can also be calculated using the above steps, except that the events used are the harms observed in the two groups.
Interpreting the Number Needed to Treat
The ideal NNT is 1, meaning that every patient who receives treatment obtains benefit. This is extremely unlikely, so the closer the NNT is to 1, the better. That being said, there are several important considerations for interpreting the NNT, especially when comparing it to other therapies.
Not all outcomes upon which the NNT is based are equal. The benefit associated with the NNT depends on which outcome was studied in the clinical trial. For example, let’s say the administrator of a health plan is comparing two drugs for blood pressure to determine which to add to the prescription drug benefit. The first drug has an NNT of 50 for preventing one stroke whereas the second drug has an NNT of 25 for reducing blood pressure by 5 mmHg. Although the second drug has a lower NNT, the outcome for the first drug is more compelling.
The same thing goes when comparing NNT and NNH. If a drug has an NNT of 25 for preventing one heart attack and an NNH of 25 for causing one nosebleed, the benefits of the drug likely outweigh its risks in most patients.
Duration of Time to Observe Benefit or Harm
The duration of clinical trials vary. Some trials are completed in under a day whereas others take years. When a benefit is observed in a clinical trial, it must be interpreted in context with how long the trial was conducted. For example, if the trial for a drug was conducted over the course of 5 years and an NNT of 20 to prevent one death was observed, then it is generally assumed that 5 years are required for the benefit to emerge. However, if a new drug is discovered that demonstrates the same NNT after only 1 year, it likely exerts greater overall benefit because it required less time to match the NNT and additional benefits are likely to accrue over time.
Prevalence of Disease in the Population
From a public health standpoint, the potential impact of a treatment may also depend on the population prevalence of the disease it treats. For example, let’s say one drug treats a rare disease and demonstrates an NNT of 25 in a clinical trial. However, because only 1000 people will get the disease annually, only 40 patients will likely benefit from the drug each year. On the other hand, a drug used for heart attacks may have an NNT of 100. Although its NNT is larger, the number of people who have heart attacks approaches 800,000 annually, meaning that 8000 patients may obtain benefit each year.
Understanding how to determine and interpret NNT/NNH could benefit a wide range of people. Health reporters could use this information to help explain the potential benefits and risks of a new therapy to their target audience. Government officials could use it to determine funding priorities for public health initiatives. Finally, it could help the public discern from drug ads what benefit is being purported as well as the likelihood of obtaining it.
It is hard for me to see a down side, so what if we required pharmaceutical manufacturers to report it?
Agree? Disagree? Have ideas for other important considerations related to NNT/NNH? Please feel free to leave comments below!