This blog has MOVED!

Please visit for the most updated content. All these posts and more can be found over at the new URL.

Friday, February 8, 2013

Mel's Literature cheat sheet

We are continuing on the topic of how to evaluate a study --> the scenario is this: you have an issue, you have sought the best advice you can from friends, mentors, and vets, and you want to see what the most current literature has to say on a particular subject. How do evaluate studies? Chances are you will find lots of conflicting information and it can be confusing with all those statistics and "45 dollar" words........which brings me to my first point.

1. Read the Abstract - but don't make a decision yay or nay for a particular topic until you've read the actual paper. An abstract can be so clearly written and make so much sense. Sometimes it's the only thing that is available to read online for free. HOWEVER, I can't tell you the number of times that I've been burned by just reading the abstract - think of the abstract as a series of claims by the author. Without seeing their evidence, even if their hypothesis makes logical sense, it's no better than asking yahoo answers what to do and then taking the advice of a random stranger. The abstract is a STARTING point that helps you decide whether the article is related to your topic of interest enough to put in the time to actually evaluate the rest of the paper. It's perfectly fine to tell someone you saw a study on a particular subject, and it looked interesting and the authors made such and such a claim......BUT you cannot claim the study as fact until you have examined whether the authors conclusions were supported by their data (and more papers than you would think have serious enough flaws that I would be hesitant to say that anything was proven or even strongly correlated!).

When I'm reading a paper I like to start with the abstract and write down in sentence form what I think the authors are claiming before I read the rest of the paper, so I can keep straight in my mind what the point was suppose to me.

2. Background information - here I try to decide what kind of study it is. Retrospective? Case control? Cohort? Each study has it's own strengths and drawbacks. Each study type is suited to something different and the data collected can be used to calculate things like risk, relative risk etc. It's important to know what KIND of study it is so you can determine whether the conclusion of the paper is VALID. Some types of studies can tell you risk, and some CAN'T - but it's amazing how many papers I've read in school journal clubs that we get to the results section and are all like "WTF? you can't use that stat with this kind of data?!!!?????"

Disclaimer. I'm not a statistician. This is a brief summary of what I've learned in class and how I've tried to remember and applied it! I welcome corrections :).

Retrospective - Often looking back at records, or using surveys that rely on people's memory - these can be useful for gathering information about a certain condition/treatment etc over a long period of time.


Often start with a "healthy" cohort, meaning that it's a population of critters that are all basically the same and no one has been "exposed". You then expose some of your cohort with a risk factor and you see who in what group develops some specific condition. For example, you take a group of endurance riders. You put salmonella in the ride meeting potluck potato salad of half of them, and you give unadulterated potato salad to the other half. You then measure the number of Salmonella cases from each group - the exposed group and the non-exposed group.

You can use Cohort studies to calculate "relative risk". Relative risk is a the risk of developing the condition/disease RELATIVE TO BEING EXPOSED. This is different than absolute risk. Absolute risk would be your risk of getting salmonella if you eat the salad. You can see that this is a stupid and ridiculous conclusion to try and make from our cohort study. The ride meeting cohort study isn't measuring your risk of getting salmonella from eating ride meeting potato salad - it's measuring your risk of getting salmonella if you eat infected potato salad, and your risk of getting sick if you eat the "normal" salad. Thus its measuring your relative risk of developing a condition having been exposed (or not) to the risk factor.

It would be WRONG to publish a paper on the risk of developing salmonella at a ride meeting from our cohort study and conclude that the risk of getting sick from the potato salad is X%. It would be CORRECT to say that your relative risk of getting salmonella from a ride meeting salmonella-contaminated salad is X%.

Releative risk can be given as a ratio that expresses the relationship of condition developing in the non-exposed and exposed groups. A ratio of 1.0 means that the relative risk of developing the condition is the same in both exposure groups.

Limitations of Cohort studies: less useful for rare diseases. it might not develop at all if it's rare enough, even in your "exposed group".

Case Control

You have cases (endurance riders with salmonella) and you have controls(endurance riders at the same ride without salmonella). You then try to decide what the exposure factors are. You are basically working in reverse of a cohort study! Case control studies are much more useful for rare events/diseases/conditions etc.

For our salmonella stricken riders, you must first try to find the exposure factor. Perhaps you chose the chicken as your "exposure".

Perhaps you calculate chicken exposure for your salmonella verus non-salmonella riders and find that your salmonella riders ate more chicken than your non-salmonella riders. Great! Riders who ate the chicken were more at risk for the Salmonella infection!!!! Yeah!

But wait.......the salmonella was in the potato salad. Oops.

See - that's what can happen with a case control study - you can pick the wrong exposure factor to study. However, with rare diseases it allows you to study a group of individuals with a disease instead of hoping that the disease develops in your healthy cohort as you fling exposures onto them.

While looking at what kind of study it is, try to decide whether you agree with the authors - are the controls really controls? Are they measuring the correct thing? Are they using the appropriate type of study?

This subject will continue in a future post!!!! This infectious disease block is sucking the time out of my life to do anything but study, so it's off to antibiotic therapies for now.


1 comment:

  1. (remind me to stick to powerbars if you show up at the potluck...)



Note: Only a member of this blog may post a comment.