Opinion

Vinay Prasad: How to read a science paper

Photo from Vinay Prasad/Twitter

Dec 29, 2023, 11:29

Vinay Prasad: How to read a science paper

Vinay Prasad, Professor in the Department of Epidemiology and Biostatistics at the University of California San Francisco, shared a post on X/Twitter:

”How to read a science paper. There are lots of tips and tricks to read certain types of studies, but we seldom discuss how to approach a paper in first place. Some thoughts.

Papers are individual pieces in a broader puzzle. Before I read a paper, I ask myself: what question am I trying to answer? Often it is: Should I incorporate this drug, device, or surgery into my practice? Sometimes it is: Should I alter my behavior? Other-times it is…

Is this finding concerning enough that we should look into it more or communicate it broadly? Sometimes it is… Does the world really work this way? Is this really the mechanism?

Before you approach a paper, you should have some idea which category you are in. Often in medicine, we are reading with the idea of whether this paper helps or supports recommending some specific medical practice, the… Should I do this in my practice?

Then you need to know a fair bit before diving in. What is the specific population this question is about. Does the trial include that population? What do we normally do outside of the study? Is the control arm doing that, or deviating? If it is deviating, in which way?

Does the study apply to your practice— no matter the results– is the first key Q? Often the answer is no, and should be acknowledged.

The next Q is: who has the burden of proof? Is it on the authors to convince me, or am I going to do this till proven otherwise. And why? How did that come to be? Is that framework consistent? Burden of proof is a key q.

Then of course, the tricky bit is teasing out whether the data presented support the inferences made. Is the study confounded? Issues with time zero? Other types of bias? Are there errors, but that don’t matter?

Then finally, it has to be put in context. What other studies are out there? Are they better or worse? Why is this one the one I am prioritizing, or instead, does it fall short?

A big error I see is argument that goes nowhere. People have forgotten the central question that is at stake, and are distracted by tangents. Ultimately, a simple test of how you’ve done is presenting the article to others who have read it, and see what they think.”

Source: Vinay Prasad/X