In a previous post, I described the difference between relative risk (efficacy), absolute risk and the related concept of NNT (number needed to treat). In that post I focused on the NNT to get the benefit of the treatment. In this post I will focus on the essential other half of NNT: the NNT to be harmed.
I will mostly focus on the direct harms of the treatment itself. However, in some cases, there are harms that come from other actions taken to treat or avoid a medical problem. Sometimes the harms can be large. The study of these indirect harms is not as advanced in the scientific literature as the direct harms, but given how large the scale of the indirect harms can be, they should be made standard. practice.
NNT for Harms
NNT is a simple way to understand how probable a given outcome is likely to be in absolute terms.
Sometimes there aren't any harms, as in this meta-analysis of over 240,000 patients in 18 studies.
What's important to note is that the researchers looked not only for the benefit of fever reduction, but also for the harms that had been suspected for one of the treatments.
Here is one where the NNT for harms is crucially important -- because the treatment that is supposed to prevent heart attacks caused more of them than it prevented!
The case above illustrates an important aspect of NNT: it should cover (if appropriate) multiple possible benefits and multiple types of harms.
Just because NNT harms outweigh benefits for a treatment doesn't mean that medical practice responds appropriately. For a long time, high blood cholesterol was thought to cause heart attacks. Statins became widely prescribed to lower the number. But now it is scientifically proven that blood cholesterol should not be lowered and therefore statins should not be taken. In spite of the fact that NNT harms are strong with no benefits, it remains standard practice for doctors to prescribe statins to lower the cholesterol level to meet now-disproven standards..
Sadly, this raises the issue of conflicts of interest and transparency in scientific research, and the readiness of the medical profession to update practices when the science demonstrates that it should. It's even trickier when a pharmaceutical company conducts studies to prove that a drug it developed has important benefits and minimal harms.
NNT Harms for covid vaccination
The FDA's EUA (Emergency Use Authorization) Issued in December 2020 for Pfizer's covid drug claimed 95% effectiveness, and listed minor side effects which lasted just a couple days. The FDA gave full approval for the drug in August 2021.
The full approval document stated that "the vaccine was 91% effective in preventing COVID-19 disease." No explanation was given for the reduced effectiveness. Unlilke the EUA document, the absolute numbers of infections were not disclosed, therefore giving a highly misleading impression of how likely any person who got the vaccine would be helped by it, implying that 90% of the vaxed would be protected vs. the actual number of under 1%.
For harms, most of the minor harms of the EUA were repeated. However, they disclosed that myocarditis and pericarditis were suffered by young males: "Available data from short-term follow-up suggest that most individuals have had resolution of symptoms. However, some individuals required intensive care support. Information is not yet available about potential long-term health outcomes." Sadly, they provided no data, no NNT for Harm.
I have yet to find good numbers for NNT Harms for covid. This should be easy, but as it turns out, the vast majority of the relevant data is secret. Yes, secret by approval of the FDA.
However, I've dug into a couple of issues based 100% on published scientific data. For example, I found a paper published in April 2021 in the New England Journal of Medicine on Vaccine Safety in Pregnant Persons. The paper showed that the mRNA vaccines were safe for pregnant people to receive. Here is Table 4 from the original paper, showing that there were 104 spontaneous abortions out of 827 vaccine recipients, about 12%, which is within a normal range.
Here is the footnote to the last column, about the numbers of people involved.:
A correction was published in October 2021 in the same journal, after the FDA's full approval had been issued. A casual reading of the correction, including the summary and abstract, makes it seem as though nothing significant was changed.
Here is Table 4 in the corrected paper:
The number of spontaneous abortions remained at 104, but the totals and percentages were dropped. The explanation is found in the footnote:
The footnote leaves the impression that nothing can be concluded. However, returning to the footnote in the original paper, we read "...based on 827 participants ... who received a Covid-19 vaccine ... A total of 700 participants (84.6%) received their first eligible dose in the third trimester..." So 700 participants could not have had spontaneous abortions, since all those took place in the first 20 weeks of pregnancy.
The arithmetic leads us to 827-700=127 participants were vaccinated earlier, and 104 of those participants had spontaneous abortions. The vast majority. This is clearly something that the authors should have pointed out and explained. Maybe my logic here is wrong.
This leads us to wondering what should happen:
What should happen
First of all, the authors should have made clear the implications of their correction. If indeed the data shows that spontaneous abortions were excessive, they should have said so, and promised further study to confirm.
Second, data about medical treatments of all kinds, including drugs, should be fully open source, the way some software is. That way, others could do the job that the authors of the study failed to do. The developer of the drug should open its data to the public, just like the source code to software like Linux is 100% open for copying, testing and use. This by itself will solve many problems. It will also enable problems to be surfaced quickly, so that a minimum of people are hurt by the problems. If drug makers were truly interested in safety and effectiveness, they would welcome the additional scrutiny.
Conclusion
NNT is an essential measure for treatment effectiveness. Every time a treatment is proposed to a patient, NNT should be part of the discussion. Certainly NNT for benefits is important -- that's the whole point of the treatment. But NNT for harms is regularly left out of the discussion. Instead, it should be brought to the forefront.
Comments