Here is an article from nytimes.com about a study done about the risks about radioactive iodine after the Chernobyl accident. The scariest part is that the exposure to the radioactive material wasn't direct , but through dairy products. The study reveals the exposure leads to increase chance of thyroid cancer. Interesting stuff, especially with what is going on with Japan right now. Hopefully Japan does a better job of protecting people from the potentially harmful effects of radiation.
http://www.nytimes.com/2011/03/18/health/research/18cancer.html?_r=1&ref=health
Friday, March 18, 2011
Wednesday, March 2, 2011
Elephants on LSD, Wilford Brimley, and gas masks. Let's talk about studies.
The title might be a little misleading, but all those are a part of this article in some capacity. Studies are an important part of epidemiology, as you probably have noticed with all the articles based on studies that I have posted so far. Epidemiological studies test whether an exposure has an association with the development of disease (Gordis, 2009). After this, epidemiologists attempt to find whether or not the association is also causal. I'll define/describe some of these words so that makes more sense.
Exposure -think food, sunlight, cigarettes (anything a person can be exposed to)
Disease - this includes actual diseases (lupus, cancer, etc.) and also many other outcomes (buying a car, lifting weights, etc.)*
Association - some kind of relationship exists between exposure and disease, but it's unclear what that relationship is
Causal - the exposure causes the disease
* I know that this is a little confusing but think of a study about whether smoking cigarettes makes a person lift more weights.
It is important to note the difference between association and causation. Just because a study finds an association between eating jelly beans and liver cancer does not mean that eating jelly beans will cause cancer (I just made that up by the way, so no need to be terrified of jelly beans). It is also important when reading an article to be critical of the findings, like whether they are saying an exposure is associated with a disease or if an exposure causes a disease. Big difference. News sites sometimes have a habit of leaving important details out or of making assumptions that the actual study may or may not support. Making bold headlines is a good way to attract viewers, but it is also a good way to send the wrong impression to readers. Take, for example, this article about a breastfeeding study and various headlines that were produced from it:
http://ksjtracker.mit.edu/2011/01/20/breastfeeding-study-a-case-study-in-bad-headlines/
How about this one:
http://theweek.com/article/index/203843/can-secondhand-smoke-make-you-crazy
Can secondhand smoke make you CRAZY?!?!?!? The picture to the left was attached to the article. I love that caption too, which I kept exactly as it appeared in the article. A little sensationalized in my opinion. That's not to say there isn't some validity to the study (notice the epidemiologist who led the study making sure to say that the results do not prove causality), which you can investigate yourself if you wish. What should you look for if you want to know more about the validity of a study? Well I will investigate an article and the associated study and show some things to keep an eye out for.
Keeping on the smoking trend (there is never a shortage of smoking studies and related articles) I chose this article to look at. First, lets take a look at that headline: Study: Quitting smoking raises diabetes risk. If you just read that and are a smoker you might think to yourself, "Well to heck with quitting if it will give me diabetes." If you were a little more critical you might read on and find that the study authors emphasize that quitting benefits far outweigh those of continuing to smoke and the authors associate the increased risk to weight gain often attributed to quitting smoking. Now you might have a different mindset about the topic. Just reading the article is just where you should start to investigate the study. Before you make a life-changing decision to cut out jelly beans (or whatever the exposure is in the study), you should also look at the original study.
Here is a link to the study that the article we are looking at discussing. So lets get some of the basics out of the way.
What is the exposure? = Quiting Smoking
What is the disease? = Type 2 diabetes mellitus
What is the hypothesis that is being tested? = Smoking cessation increases diabetes risk in the short term, possibly owing to cessation-related weight gain.
What kind of study is it? = Prospective cohort study
Who was studied? = 10,892 middle-aged adults who initially did not have diabetes
How do you find the answers to these questions? Well luckily in this study they pretty much just lay it out. Some studies may not and you might have to do some reading to figure it out from the article. The exposure, disease, and hypothesis should be fairly easy to pick out. The kind of study might be more difficult to find if it isn't listed. In a cohort study "the investigator selects a group of exposed individuals and a group of nonexposed individuals and follows up both groups [note: can be more than two groups] to compare the incidence of disease" (Gordis, 2009, p. 167). In this case the exposed individuals are smokers who quit and the nonexposed are continuing smokers and individuals who have never smoked. It is prospective because the people studied were identified at the beginning and then assessed later (17 years in this case) to see if they had developed the disease. This differs from a retrospective cohort study which chooses individuals based on whether they have a disease or not and then the researchers look back in time (using past records like surveys, census, etc.) to see whether or not each group being compared has the exposure being tested (Gordis, 2009). So prospective = future, and retrospective = past. Another common type of study is a cross-sectional study which determines exposure rates in the groups being compared at a point in time. So cross-sectional study = present. (Note: There are other types of studies. I just used cross-sectional studies as a comparison to cohort studies.) Prospective cohort studies are good because they allow researchers to follow groups of individuals that they have selected over a period and time and see if an outcome occurs. This is important because it allows researchers to (possibly) determine whether a temporal relationship exists between an exposure and a disease. A temporal relationship means that the exposure preceded the disease. If a cross-sectional study (remember, one point in time) is done the researchers cannot determine if a temporal relationship exists and decide which came first, the exposure or the disease. Temporal relationships are an important part of supporting causation between an exposure and a disease. Prospective studies do have some downsides though. Participants in the study could drop out (or die) at any time during the experiment. There is also a higher chance that more confounding factors are involved.
Okay, lets take a second and discuss confounding factors. Confounding factors are other exposures that could have an association with the disease and thus makes it difficult to ascertain whether the exposure studied or confounding factors led to the outcome. For example in this study, age, race, sex, education, adiposity (huh?), physical activity, lipid levels, blood pressure, and location were all found to be possible confounding factors. These researchers accounted for these particular confounding factors by adjusting their numerical results. So in theory, these factors are not affecting the final results. The problem is that not all confounding factors can be accounted for, but a study that adjusted for as much as possible is preferable to one that doesn't account for confounding factors. Just keep in mind when reviewing studies that these confounding factors are out there and could be affecting the end result.
Back to the study. What should we really be looking for to assess whether what they say is true or not? The most important thing we look for is a valid statistical association. In other words, do the statistics support the hypothesis and can we say that the results can be applied to whole populations (white females, teenagers, Chinese men, etc.). I know statistics are boring to a lot of people, but they are important. This post is already long, so I won't go into details about how to interpret statistical results, because that could get very detailed. But it is important to realize that the results could just be due to chance, confounding factors, or bias on the part of the researchers in selecting study individuals or what data to collect. Some questions you might ask yourself when analyzing a study:
Should you be alarmed if you are a smoker and want to quit? There are a million studies out there trumpeting the benefits of quitting smoking (or even better never starting), so I wouldn't take this as a sign to not stop. To really help yourself avoid an increased risk of diabetes after quitting, the best advice would be to eat healthy and watch your diet.
Well, that's my spiel. I'll finish this post with a little entertainment.
First, an article about some silly experiments you might find amusing. Bad science is out there, so be careful to be critical of what you hear.
And second, Wilford Brimley! (Note: Not laughing at the disease diabetes, I just find this amusing and it somewhat relates to the study I have been talking about.)
Gordis, L. (2009). Epidemiology (Fourth Edition ed.). Philadelphia: Saunders Elsevier.
Corbis (2010). She's just trying to stay sane [Photograph] Retrieved from http://theweek.com/article/index/203843/can-secondhand-smoke-make-you-crazy
Exposure -think food, sunlight, cigarettes (anything a person can be exposed to)
Disease - this includes actual diseases (lupus, cancer, etc.) and also many other outcomes (buying a car, lifting weights, etc.)*
Association - some kind of relationship exists between exposure and disease, but it's unclear what that relationship is
Causal - the exposure causes the disease
* I know that this is a little confusing but think of a study about whether smoking cigarettes makes a person lift more weights.
It is important to note the difference between association and causation. Just because a study finds an association between eating jelly beans and liver cancer does not mean that eating jelly beans will cause cancer (I just made that up by the way, so no need to be terrified of jelly beans). It is also important when reading an article to be critical of the findings, like whether they are saying an exposure is associated with a disease or if an exposure causes a disease. Big difference. News sites sometimes have a habit of leaving important details out or of making assumptions that the actual study may or may not support. Making bold headlines is a good way to attract viewers, but it is also a good way to send the wrong impression to readers. Take, for example, this article about a breastfeeding study and various headlines that were produced from it:
http://ksjtracker.mit.edu/2011/01/20/breastfeeding-study-a-case-study-in-bad-headlines/
 |
| She's just trying to stay sane. (Corbis, 2010) |
http://theweek.com/article/index/203843/can-secondhand-smoke-make-you-crazy
Can secondhand smoke make you CRAZY?!?!?!? The picture to the left was attached to the article. I love that caption too, which I kept exactly as it appeared in the article. A little sensationalized in my opinion. That's not to say there isn't some validity to the study (notice the epidemiologist who led the study making sure to say that the results do not prove causality), which you can investigate yourself if you wish. What should you look for if you want to know more about the validity of a study? Well I will investigate an article and the associated study and show some things to keep an eye out for.
Keeping on the smoking trend (there is never a shortage of smoking studies and related articles) I chose this article to look at. First, lets take a look at that headline: Study: Quitting smoking raises diabetes risk. If you just read that and are a smoker you might think to yourself, "Well to heck with quitting if it will give me diabetes." If you were a little more critical you might read on and find that the study authors emphasize that quitting benefits far outweigh those of continuing to smoke and the authors associate the increased risk to weight gain often attributed to quitting smoking. Now you might have a different mindset about the topic. Just reading the article is just where you should start to investigate the study. Before you make a life-changing decision to cut out jelly beans (or whatever the exposure is in the study), you should also look at the original study.
Here is a link to the study that the article we are looking at discussing. So lets get some of the basics out of the way.
What is the exposure? = Quiting Smoking
What is the disease? = Type 2 diabetes mellitus
What is the hypothesis that is being tested? = Smoking cessation increases diabetes risk in the short term, possibly owing to cessation-related weight gain.
What kind of study is it? = Prospective cohort study
Who was studied? = 10,892 middle-aged adults who initially did not have diabetes
How do you find the answers to these questions? Well luckily in this study they pretty much just lay it out. Some studies may not and you might have to do some reading to figure it out from the article. The exposure, disease, and hypothesis should be fairly easy to pick out. The kind of study might be more difficult to find if it isn't listed. In a cohort study "the investigator selects a group of exposed individuals and a group of nonexposed individuals and follows up both groups [note: can be more than two groups] to compare the incidence of disease" (Gordis, 2009, p. 167). In this case the exposed individuals are smokers who quit and the nonexposed are continuing smokers and individuals who have never smoked. It is prospective because the people studied were identified at the beginning and then assessed later (17 years in this case) to see if they had developed the disease. This differs from a retrospective cohort study which chooses individuals based on whether they have a disease or not and then the researchers look back in time (using past records like surveys, census, etc.) to see whether or not each group being compared has the exposure being tested (Gordis, 2009). So prospective = future, and retrospective = past. Another common type of study is a cross-sectional study which determines exposure rates in the groups being compared at a point in time. So cross-sectional study = present. (Note: There are other types of studies. I just used cross-sectional studies as a comparison to cohort studies.) Prospective cohort studies are good because they allow researchers to follow groups of individuals that they have selected over a period and time and see if an outcome occurs. This is important because it allows researchers to (possibly) determine whether a temporal relationship exists between an exposure and a disease. A temporal relationship means that the exposure preceded the disease. If a cross-sectional study (remember, one point in time) is done the researchers cannot determine if a temporal relationship exists and decide which came first, the exposure or the disease. Temporal relationships are an important part of supporting causation between an exposure and a disease. Prospective studies do have some downsides though. Participants in the study could drop out (or die) at any time during the experiment. There is also a higher chance that more confounding factors are involved.
Okay, lets take a second and discuss confounding factors. Confounding factors are other exposures that could have an association with the disease and thus makes it difficult to ascertain whether the exposure studied or confounding factors led to the outcome. For example in this study, age, race, sex, education, adiposity (huh?), physical activity, lipid levels, blood pressure, and location were all found to be possible confounding factors. These researchers accounted for these particular confounding factors by adjusting their numerical results. So in theory, these factors are not affecting the final results. The problem is that not all confounding factors can be accounted for, but a study that adjusted for as much as possible is preferable to one that doesn't account for confounding factors. Just keep in mind when reviewing studies that these confounding factors are out there and could be affecting the end result.
Back to the study. What should we really be looking for to assess whether what they say is true or not? The most important thing we look for is a valid statistical association. In other words, do the statistics support the hypothesis and can we say that the results can be applied to whole populations (white females, teenagers, Chinese men, etc.). I know statistics are boring to a lot of people, but they are important. This post is already long, so I won't go into details about how to interpret statistical results, because that could get very detailed. But it is important to realize that the results could just be due to chance, confounding factors, or bias on the part of the researchers in selecting study individuals or what data to collect. Some questions you might ask yourself when analyzing a study:
- How did the researchers select the study participants?
- How did the researchers measure the exposure and the disease?
- Do the researchers discuss confounding factors and limitations of the study?
- Can you see an apparent bias on the part of the researchers?
- Who sponsored the study? (i.e. Are there conflicting interests?)
- What assumptions do the researchers make?
- Do the statistics support the researchers' conclusion? (Might need a background in statistics to really interpret this.)
- Remember, be critical!
Should you be alarmed if you are a smoker and want to quit? There are a million studies out there trumpeting the benefits of quitting smoking (or even better never starting), so I wouldn't take this as a sign to not stop. To really help yourself avoid an increased risk of diabetes after quitting, the best advice would be to eat healthy and watch your diet.
Well, that's my spiel. I'll finish this post with a little entertainment.
First, an article about some silly experiments you might find amusing. Bad science is out there, so be careful to be critical of what you hear.
And second, Wilford Brimley! (Note: Not laughing at the disease diabetes, I just find this amusing and it somewhat relates to the study I have been talking about.)
I hope this has been another informative and and interesting post. Feel free to leave comments or ask questions. I will do my best with my admittedly limited knowledge to provide an adequate answer.
Oh and lastly, here is a list of sources that I used in this post:
Subscribe to:
Posts (Atom)