How to Read a Clinical Paper
A few weeks ago, I was talking with a good friend who’s a lawyer about the strange language in contracts. I’ve worked on a lot of legal agreements and am always flummoxed by clauses like
“Any notice required or permitted by this Agreement shall be in writing and shall be deemed sufficient upon receipt, when delivered personally or by courier, overnight delivery service or confirmed facsimile, 48 hours after being deposited in the regular mail as certified or registered mail (airmail if sent internationally) with postage prepaid, if such notice is addressed to the party to be notified at such party’s address or facsimile number as set forth below, or as subsequently modified by written notice.”
When writing a contract you trust your lawyers (and pay them a lot) to understand it. But why? If the point of a contract is for two people to agree on something, shouldn’t those two people understand it?
Which got me thinking about the weird language that clinical studies are written in, for basically no reason.
There are a bunch of hand-wavy words that could be stated in plain English, and a structure that makes no sense to the average person. We are admonished not to share click-bait without checking the source, but what good is that if we can’t understand the source?
This is an explanation of the different parts of clinical papers, what they are for, and what you should pay attention to.
Abstract:
Reading an abstract is a good way to help you quickly decide if the type of information you are looking for can be found in this particular paper. The abstract is the TL;DR of a paper. It should tell you three things
The level of clinical study this is: a report of a single patient, a study of a small number of patients, or a randomized controlled study, or a meta-analysis. The higher the level of study, the stronger the conclusion
The question they were asking. A big mistake people make is to take results out of context. Always know that a study can only really speak to its starting question.
The most impressive / impactful result of their work. This is what the authors most want you to take away.
Background:
This section tells you the history of the question this paper is asking. A good paper will explain why this question is being asked -- why it matters. The background should go through all of the previous research that has asked and answered similar questions. It should also explain why the authors believe that the question they are asking is a good one and if their assumptions make sense. When you are reading a few papers on a topic, the background section can get very repetitive.
Pro-tip: Skim the background when you are first reading it. After reading the conclusions, come back and reread this section carefully. Very often the conclusions that are drawn depend on the assumptions the authors are making. But it’s hard to know which assumptions to pay attention to.
Pro-tip 2: Check the references. If the authors make a statement in the background and cite it with a footnote, do NOT assume that the footnote actually proves what they say it does. I cannot tell you how often I’ve checked a reference only to find that it says nothing. My experience is that in the peer review process, no one checks references, allowing authors to get away with a lot.
Methods:
I believe that the Methods section is the most important part of a paper. Even more so than the Results. The Methods should tell you:
What was studied? (Humans in a lab/ humans in the real world / animals / models)
How many samples were in the study?
What was measured? You need to ask if that measurement is able to answer the question that was posed.
How were errors, biases, or confounding variables controlled for? Usually a discussion of why the authors believe they did or did not control for things is in the Conclusions.
A good Methods section should have the right detail to allow another scientist to replicate the study that was done to confirm the findings.
Results:
The results section is usually very brief and provides a summary outcome (i.e. “the drug reduced pain in 80% of patients”) of the study as well as some statistics (the dreaded p-value). The point of the statistics is to understand how likely it is that the result could have happened by chance.
Smaller studies and some case reports will provide no statistics. This is okay. These studies are only intended to create the information others need to do further research.
A good results section should also explain if anything changed about the Methods during the study or if the Methods were not able to be followed, and how they handled that. Studies are designed with the best intentions, but real humans often cannot perfectly follow instructions.
Discussion:
The final section of the paper is space for the authors to explain how they believe the results of the study should be interpreted. You shouldn’t read the Results without also reading the Discussion. The authors will discuss why they think they didn’t get the results they expected, what further research is needed to explain a strange outcome, and what they are most excited about.
Discussion sections are particularly weird because the formula for writing a paper demands that authors try to sound unbiased and that they talk about where they might have gone wrong. It’s forced humility. When the results are really impressive, a Discussion can make it sound like an accident not to be trusted.
In Summary
Start with the abstract. Don’t throw out small studies, but take them for what they are. Read in this order
Methods to make sure they are pretty sound
Results for the highlight
Discussion is where you really think about what’s being said
Background to verify the Discussion
Never read only one paper. Always look to see if the Background papers have built a solid foundation for what this paper is arguing. There’s nothing wrong with a new idea, just important to recognize it for what it is.
