A primer on reading scientific studies
Learning to read scientific research can feel like you are trying to read a whole other language. And you’re right, it does have its own unique terminology. However, with a little patience and effort, I believe you can master it. Let’s start with a few resources:
Medical dictionary https://medlineplus.gov/mplusdictionary.html
Scientific dictionary (for non-medical terms) http://www.worldofscience.in/dictonary.aspx
You can also look for apps for your smartphone or tablet.
PubMed (government website that only catalogs medical research)
https://www.ncbi.nlm.nih.gov/pubmed/
Google scholar searches primarily research journals and books (when it can). https://scholar.google.ca/
Sci hub grants full access to scientific research to bypass pay barriers (you may need to try google for the latest version of the URL). http://sci-hub.tw/
Sci hub is particularly awesome because research scientists are not losing anything since they don’t get paid by the journals from those access fees.
Essential components of a study:
Title -Tells you EXACTLY what they were studying, does not automatically tell you if it is relevant to your research.
Abstract -This gives a general summary of what the paper contains. Do not use this to form the basis of an argument because it can be misleading (thanks commercially motivated conflicts of interest!).
Introduction -The authors will explain their reasons for conducting the research and often mention other research related to their work.
Methodology -This tells you how big the study sample was, what type of study it was, who/what was in the study sample, ect. This section should be detailed enough that another scientist should be capable of redoing the study without contacting the original authors. This section also tells you if the work and evidence for their argument/hypothesis is strong or weak.
Results -This should contain the answers they got from the work mentioned in the methodology section.
Conclusions -The scientist’s interpretation of the answers they got. Comparing this section with the previous 2 is one of the fastest ways to spot fraudulent work because it will not match the actual data. Many professionals are pressed for time and will read the abstract and conclusions to decide if they want to read the entire study.
In the Methodology and Results sections you should find lots of numbers (yay, can you feel the sarcasm?). When it comes to the statistics remember that there is an entire field of study just on that. Don’t sweat it, just stay focused on the bigger, overall picture. P values are Probability Values, for example, a P value of 0.000001 is a 1 in a million chance, however, a P value of 0.0482 is a 4.8% chance. Both of these results would be considered statistically significant because they move the proverbial needle away from zero by a noticeable amount. However, that does not automatically make it medically significant because a 4.8% chance is bad odds for the result you may be looking for. Additionally, the statistics can only account for the conditions and data that was recorded, anything that occurs outside of those parameters cannot be accounted for. Make sure to read the methodology before the results or the results could be misleading.
There are different types of study methods that serve different purposes.
Review — summarizes existing research and compares results to come to a conclusion.
Original research — summarizes researchers’ own work which may be in the form of data collection and/or experiments. Original research has quite a few sub-types;
Experimental studies can establish cause and effect, randomized control trials (considered the best type) is one kind.
Observational studies can establish correlation. These cannot establish causation, however, if enough studies come to the same conclusion then it is a safe bet that the association being studied is important.
Cohort studies examine groups of people with similar characteristics over a period of time. This is to examine what changes occur and then check what is similar between the ones who experienced the change. There are 2 types of cohort studies, retrospective (measures data already collected to generate a cohort and examine outcomes), and prospective (already described).
Case-control studies examine 2 separate groups with different outcomes and then examine existing data to figure out why they had different results for each group.
Cross-sectional studies take a snapshot of a population to see correlations between risk factors and disease outcomes. For example, looking at glyphosate exposure and cancer outcomes.
Some additional terminology you will run into;
In-vitro — experiment done in a petri dish or test tube.
In-vivo — experiment done in animals or humans.
Confidence interval — This relates back to the statistics, Stattrek.com defines it as “…a confidence interval to describe the amount of uncertainty associated with a sample estimate of a population parameter.” That should be as clear as mud I’m guessing. I linked their explanation of how it works to the site name. Many times when reading any medical study it always says “confidence interval of 95%” because it is almost always the same it makes this important part of the data otherwise pointless to the average reader until it deviates from the norm. I’ve included it for thoroughness.
Every study that gets published will have a DOI number which stands for Digital Object Identifier, this number is unique for each study and permanently assigned to it. This means if you have the DOI number you can find it by simply putting “DOI insert number here” into google and it will bring it up. If the study is cataloged in PubMed then it will also have a unique PMID (PubMed ID) number assigned to it. This number is only applicable to the PubMed database.
If you feel overwhelmed don’t fret. Professional scientists of all levels feel that way too. Check out this piece on reading studies that were published in Science Magazine: How to (seriously) read a scientific paper.
Examine.com also did an excellent write-up on reading scientific papers, you can find it here.
