Thank you for taking the time to share your expertise. You may have missed one of my questions in a previous post. You introduced the concept of NNT in the discussion of statins and gave examples of 60 and 45 for two groups of people. I was curious as to what would be the threshold NNT to consider a drug to be ineffective or of marginal value?
In the American Heart Association item I showed, in a particular group of 100 people not taking statins 5 people will have a heart related incident in a six year period. If the 100 people took statins, that number is reduced to...not 0, not 1, not 2, not 3 but 4. The vast majority will receive no benefit and 2 will have side effects to varying degrees. Would that translate into a NNT of 100 for this specific group? Since the AHA was touting these results, I gather an NNT of even 100 is considered as showing a drug's effectiveness. How high is too high?
My pleasure. I think these are important things to discuss.
NNT is a fairly involved topic to dig into, especially if we’re taking about medications used in primary prevention and health outcomes ideally measured on the scale of decades. Such is the case with stations and CV events.
The ideal NNT would be 1, meaning every patient treated has a benefit. The value of NNT in the real world is dependent on a lot of factors, like how serious outcome is of the potential condition you’re trying to prevent, the costs of that undesirable outcome (incl financial, effect on quality of living, productivity, etc), the potential side effects and risks of the intervention, the cost of the intervention, the benefits beyond preventing the condition, how prevalent the condition is, how well the undesired outcome can be treated if not prevented, what alternatives exist, etc. An epidemiology textbook might say 5-10 is generally ideal, but in reality considering an intervention’s value really depends on a multitude of factors beyond its clinical utility, including economic, social, etc.
NNT is also highly dependent on how studies are conducted, especially the what baseline risk of the participants is and how long the study was conducted. We know that statins benefit people at the highest risk who have been taking the medication the longest. So NNT is and will be skewed if relatively healthy young people with high cholesterol and high risk old people who just started taking statins late in life vs. people who are high risk who have been taking statins for many years. And a study over the course of 2 vs 5 vs 20 vs 40 years will likely demonstrate different results. So as it stands, many of the NNT’s are based off of a broad spectrum of patients over a short period. And it’s worth noting it can be difficult to predict what a 35 year old’s health situation will be in 30-40 years.
It should also be noted that statin guidelines have changed over time- at least several times in my career (depending on if you follow USPSTF or ACC/AHA guidelines). The USPSTF just came out with the newest guidance last year and in some respects are actually more conservative.
The treatment algorithms are based on an individually calculated ASCVD (atherosclerotic cardiovascular disease) risk score. This informs the chance of having a cardiac event in the next 10 years. The newest guidelines generally recommend initiating statins for those with an ASCVD score of 10%.
I won’t get into the technical definitions of side effect vs. adverse effect vs. adverse event. Side effects are predictable but unintended effects, usually
undesirable (but in some cases can they can be beneficial). Adverse reactions are harmful effects caused when the drug is otherwise properly taken.
Regardless, to avoid being technical, not all side effects (in the vernacular sense) are harmful, but they may be tolerable or inconvenient. Or they may so intolerable in which case another of action needs to be considered.
Harm generally means causing substantial distress, physiological damage, or risk of death. Slight fatigue caused by a drug is not considered harm. Kidney damage would be.
It’s not like patients are forced to take medication and that prescribers ignore the experience of the patient. If a drug’s side effects are that bad, then they’re less likely to take it, and it defeats the entire purpose.
That said, the benefit of the drug is always weighed against the risks, including side effects. In the case of cancer and cytotoxic chemotherapy, there’s obviously terrible and debilitating side effects, but it’s a life or death situation. Thus most people, at least for some time, endure. If the condition is less serious, there’s going less of a tolerance for significant side effects.
It should also be mentioned there is often a false equivalency of potential benefit vs side effects. If they were in reality equal, no one would ever take any medication. The reduction in risk of a stroke isn’t comparable to a mild headache that subsides in 2 days.
Side effect statistics usually don’t take into consideration the severity of general adverse effects or how long they last or the cause ie adverse effects seen in the placebo groups are not subtracted and in many cases are the same or higher. Obviously however there are some side effects than are predictable and easily attributable to the mechanism of the drug.
———
Re: Pharmaceutical companies, there is always reason to be skeptical of their reporting given their conflict of interest. The government has cracked down significantly on companies hiding clinical data with much greater oversight. And given numerous controversies, it’s not surprising the public is skeptical. And I’m not telling you to inherently trust them.
The thing often forgotten is the government and research institutions (universities, hospitals) does their own independent studies. Statins in particular are generic and have a flooded market, so the original inventors are not really making any money anymore. After 30 years and billions of prescriptions, I think we have a pretty good understanding of what risks they pose. Measuring benefits as described above is much more complex to measure given a long list of significant variable discussed earlier.
I should also mention I think the impact of money on individual providers is vastly overestimated. It’s not like a couple+ decades ago where a pharma kickbacks were legal and everywhere.
