The relative risk of population-based studies

I’m with my wife in Boston for a conference this Memorial Day weekend, and I caught a presentation by two very knowledgeable women on what “risk” means to everyday medical practice. The presenters did a heck of a job in explaining risk and what it means to a provider as they speak to their patients about possible therapies. For example, if a patient comes in with pericarditis, how would you explain to them that a therapy was found to have a relative risk reduction of 10% in patients who took it compared to patients who didn’t?

I’m not a healthcare provider, so I have no idea how you would explain this to a patient. Also, I’m an epidemiologist, so I understand what a relative risk reduction is. So the presenter was kind of preaching to the choir when it came to myself, but others in the room had different approaches. Some said that they would tell their patients that they would be 10% more likely to be cured with the therapy. (This was wrong, and I’ll tell you why in a little bit.) Someone else said that 1 in 10 people benefit from the therapy, so the patient had a 1 in 5 chance of getting better. (This is a better answer, and I’ll tell you why in a little bit.)

The thing about population-based studies is that they’re based on, well, populations. For example, a randomized clinical trial has inclusion criteria to try and control for confounders before the trial starts. For example, if your testing a drug for heart disease, you may keep diabetics out of your trial because your results might be confounded by diabetes. So how do you apply the findings of that study on a diabetic patient sitting in front of you?

Likewise, population studies have told us that some screenings are not necessary in some people, while other screenings are necessary in others. But these findings are usually presented as guidelines for providers to follow. First and foremost, they need to look at the patient in front of them before making the decision to treat or not, or to screen or not.

When telling the patient about relative risk reduction, you need to explain what relative risk is. (We also call it a “risk ratio”.) Relative risk is the risk in the treated divided by the risk in the untreated. Risk is the number who developed the condition divided by the number in the treatment group. For example, you randomize 2,000 people into two groups of 1,000 people each. Each group then gets either a medication or a placebo. You then follow them to see who gets the disease and who doesn’t.

Now, let’s say that the medication group has 250 heart attacks (or whatever). And the placebo group has 350 heart attacks. The risk in the medication group is 250/1000, or 0.25. The risk in the placebo group is 350/1000, or 0.35. The relative risk is 0.25/0.35, or 0.71. The relative risk reduction is 0.35 minus 0.25, or 0.10. The number needed to treat is 1/0.10, or 10. You would need to treat ten people with the medication to prevent one heart attack.

How would you translate all this to your 50 year-old, overweight, hypertensive smoker patient sitting in front of you for their annual physical? The presenters at the session this morning recommended — quite rightly — that providers look at the details of the study. Who was admitted into the study? Were there people in their 50s? Overweight? Hypertensive? Because, if there were not, your patient sitting in front of you may have a greater risk of a heart attack than the people in the study, skewing the way you look at the results in the study.

Someone asked how the patient would see that they had only a 1 in 10 chance of being that person who is benefited from the medication. This is a valid point. If that patient is already taking medication for their hypertension, etc., would they want to take an additional one in this situation to prevent a heart attack? Well, that’s a discussion to have with the patient.

In a couple of days, I’ll be taking my comprehensive exams at Hopkins. One of the days is dedicated to looking critically at two published study and answering questions based on them. While I’m not a healthcare provider, I am training in how to be a public health practitioner. It’s going to take a lot of skill to be able to separate the wheat from the chaff when it comes to epidemiological studies. You just can’t take a study and run with it when none of the participants were comparable to the population you’re working with on many levels. We’re all human, yes, but the determinants of health are based a lot on where you live and how you live, and then your biology.

Things to think about… I’m glad I came to the conference with my wife.

I'm a doctoral candidate in the Doctor of Public Health program at the Johns Hopkins University Bloomberg School of Public Health. All opinions posted here are my own, of course, and they do not necessarily reflect the opinions of my school, employers, friends, family, etc. Feel free to follow me on Twitter: @EpiRen