Primary Prevention: What Medicine Can Learn From Monte Carlo

John Mandrola, MD


March 29, 2016

Medicine is easier when people are sick. In treating heart attack or stroke, certainty rules over uncertainty. The best outcome of a heart attack or stroke, however, is not to have had one.

Prevention is where medicine gets hard, very hard. To prevent something that may or may not happen in the future, we operate with probabilities. Preventive medicine also requires us to treat people who are not sick. Given our oath to first, do no harm, that gets tricky.

Current thinking on primary prevention involves applying results from clinical trials to individuals in your exam room. We say taking this drug or having this cancer screening test will reduce your future risk of dying from that disease by x% or y%. But people don't care about population averages; they want to know their odds of benefit and how much benefit to expect.

A new study,[1] published in Open Heart , provides a novel way to think and discuss these odds. I will try to convince you that this paper may be one of the most important medical studies of recent times.

The Study

Prof Darrel Frances (National Heart and Lung Institute, UK) and colleagues performed a novel three-part study to explore the uncertainty inherent in primary-prevention decisions. They looked at the effect of statins, but it's important to set out that this kind of trial could be applied to any intervention known to reduce disease-specific mortality.

In the first part of the study, they used published mortality data (cardiovascular [CV] and non-CV) to calculate the mean lifespan gains from statin therapy. The model assumed statins reduced CV mortality by 30%, a figure taken from this Cochrane Database Systematic Review [2] of statins in primary prevention.

The second part of the study looked at the lifespan gains distributed among individuals with the same risk factors. That is a key sentence. We know intuitively that not all individuals who take a statin enjoy the calculated average lifespan gain. Some die early of another cause, some live a long life and die of something else, and some have a heart attack even though they took the drug. In Gaussian speak, not everyone lives in the middle of the curve; some live or die at the tails. To sort out the probability distribution of these gains or nongains, the research team devised a mathematical model based on a Monte Carlo simulation—a method named after the casino town in Monaco.

The authors explain this simulation with a dice analogy:

"I magine mortality being determined purely by throwing a pair of dice every day. If an individual throws a six on either of their dice, their life ends. Running this for many days permits a lifespan to be calculated. The same dice throws can then be reevaluated to deliver reduced mortality risk but identical play of chance. For example, if a double six was now required for a fatal event, then many of the throws that had been considered fatal would now not be fatal, so lifespan would likely be longer."

This model is unique because it can be carried out for multiple simulated individuals at identical baseline risk, whereas in real life, each patient lives only once.

In the third part of the study, the research team surveyed the general public to assess preferences on how to convey lifespan gains. Researchers actually went out to town centers and asked more than 350 people their preferences between a certainty of a small gain in healthy lifespan (1 year) vs increasing percentage chance of a larger gain in healthy lifespan (10 years).


  • Younger individuals gain the most lifespan from statins. Although CV risk and risk reduction with statins increases with age, so do competing causes of death. At any level of baseline risk, a 50-year-old enjoys more than twofold greater gains in lifespan than an 80-year-old.

  • Lifespan gains concentrate within an unpredictable minority. For example, men aged 50 years with average CV risk have a mean lifespan gain of 7 months. However, 93% of these identical individuals add no time to their lifespan, while the remaining 7% gain a mean of 99 months.

  • In higher-risk individuals; what increases is not the duration of lifespan gained but the proportion of people who benefit.

  • People vary in their preference for gains. Some individuals prefer the certainty of a small gain while others prefer the lottery approach—a smaller chance for a bigger gain.


Dr Richard Lehman (University of Oxford) tweeted that this study is a "game changer." Before I tell you why I agree with him, there are three caveats of this study that deserve mention.

First, this model did not consider the downsides of long-term statin use. Not only could these drugs confer harm, which would mitigate benefits gained, but as Dr Francis's group has previously published[3], people have different views on the burden (disutility) of taking preventive pills. Second, this model did not account for nonfatal events that could decrease quality of life, eg, nonfatal stroke or heart failure. That weakness could increase the benefit of statins. Third, this is a mathematical modeling study that had to make assumptions using point estimates of statin benefit and mortality data taken from death certificates—both are imperfect estimates.

Caveats aside, this is a momentous paper. Four aspects of it changes common thinking.

Asymptomatic people don't accept preventive treatment because they want to live 5 or 10 years, they accept these interventions because they want to live a longer life. This study forces doctors to look at the right time course for prevention—a lifetime.

Another reason this study is important is that it disrupts the notion that older patients with higher risk benefit more from prevention with statins. In fact, it's the opposite. This is new thinking for preventive medicine. The authors note that the lifetime risk of CVD for a man with average characteristics is 37%. Many of you may shudder at the thought of using statins in lower-risk younger people for decades, but the other side of this finding is earlier deprescribing in the elderly. Either way, the decision to take, not take, or stop a prevention treatment is not up to us. It's up to the person who swallows the pill each day. Our job is to help people make the decision.

This new way of understanding distribution of benefits changes how doctors advise people. Before this analysis, most doctors, and decision aids, translated the average absolute risk reductions of statins into a number needed to treat. We can debate the actual number, but let's agree that many patients would need to take the pill for one person to benefit. What the findings of this study suggest is that, yes, many people who take the statin get no benefit, but those who do benefit, may benefit a lot. That makes sense: If a statin prevents a left main occlusion in a 60-year-old, avoiding that event delivers many months of extra life. The gamble is that it's hard to know whether that 60-year-old avoided the occlusion because of the drug or just chance. (See, I told you probability is harder than procedures.)

The findings from the survey part of this study also change the conversation about prevention. Not surprisingly, people differ in their preferences. Some like the certainty of a small gain while others like the idea of big gains. Doctors can't know these preferences unless we ask. If we want patient-centered (not population-centered) decisions, we have to discuss these issues and then not be attached to what patients choose. You may not buy lottery tickets, but lots of people do.

Finally, please don't mistake this as a statin study. The issue of large gains concentrated in an unpredictable minority can be applied to lots of things we do in the name of preventing future events. Think blood-pressure treatment, for example. I would also urge you not to quibble too much with the assumptions and modeling. The exact numbers are not the main point. The point of this work is that it brings statistics, probability, and cognitive psychology to the doctor-patient relationship.

When it comes to treating people with risk factors, not diseases, embracing uncertainty has always been important. But, now, as technology increasingly measures the human condition and creates more risk factors, comfort with gambling in medical decisions has never been more vital.



Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.