Welcome back, readers. As I demonstrated in my last post, epidemiology is everywhere. It smacks you in the face whenever you open up a newspaper or log onto your favorite news outlet. While the publicity of epidemiological studies is a good thing, often times important details can get lost in translation. When you read a newspaper or magazine article, you have to consider that Mr. or Mrs. Reporter did not come up with the study or it’s findings all on his or her own. Rather, the reporter has likely been tasked with writing about ANOTHER article, originally published in a scholarly journal unfamiliar to the average American. The reporter takes the information, polishes it up, and presents it to the laymen in a sparkly article with a catchy headline. Call it sensationalism, call it recycled journalism, but whatever you call it, make sure the condensed version of the story does not get the final word.
If you are thinking that you don’t have any interest in reading an article from The New England Journal of Medicine or The Journal of the American Medical Association, I get it. They use big complicated words and scary charts and graphs. Sure, it would be useful to have a background in statistics or the biological sciences, but it not necessary to understanding a published research study. Today, we will examine an article found on MSNBC, as well as the companion journal article published by National Cancer Institute. We are going to discuss specific attributes of the articles to give to an idea of what you should be looking for when you evaluate these news findings yourself.
.jpg) |
| Source: DrSoram.com |
Now that you have the synopsis, we’ll break it down a bit by answering a few basic questions.
What exposures are being assessed?
When you begin reading any article, be it a scholarly journal or a magazine, you want to ask yourself two basic questions. What is the exposure, or the variable that increases the risk of disease? Secondly, and usually most obvious, what is the disease in question? Every study is different and sometimes both the disease and the exposures are known up front. Other times, the end result, disease, is known and researchers are trying to measure exposure to determine if there is an association. What every study wants to answer is whether or not the disease and exposure are related, and sometimes evidence can go so far as to prove one causes the other. In the example case study, we know that breast cancer is the disease, and we can tell from the article that researchers are studying several possible exposures for this disease. Although the newsworthy association gleamed from this study is reproductive history and triple-negative breast cancer, the study also examined associations between this type of cancer and menstrual history, breastfeeding, and use of oral contraceptives. Additionally, these same factors were measured for women with other types of breast cancer. The researchers describe these exposures as “hormonally mediated risk factors[1].”
When you begin reading any article, be it a scholarly journal or a magazine, you want to ask yourself two basic questions. What is the exposure, or the variable that increases the risk of disease? Secondly, and usually most obvious, what is the disease in question? Every study is different and sometimes both the disease and the exposures are known up front. Other times, the end result, disease, is known and researchers are trying to measure exposure to determine if there is an association. What every study wants to answer is whether or not the disease and exposure are related, and sometimes evidence can go so far as to prove one causes the other. In the example case study, we know that breast cancer is the disease, and we can tell from the article that researchers are studying several possible exposures for this disease. Although the newsworthy association gleamed from this study is reproductive history and triple-negative breast cancer, the study also examined associations between this type of cancer and menstrual history, breastfeeding, and use of oral contraceptives. Additionally, these same factors were measured for women with other types of breast cancer. The researchers describe these exposures as “hormonally mediated risk factors[1].”
Are there possible confounders?
For those of you new to the epidemiology game, a confounder is an additional variable that may account for the effect of exposure on disease. Researchers must account for the possibility that any observed association may be due to differences that exist between study groups, not including the exposures being studied. Leon Gordis notes that a confounder is risk factor for both the disease and the exposure in question[2]. Accounting for confounders is important because they may inflate or deflate the findings of the study. In this particular study, hormones are believed to influence breast cancer development. I am therefore concerned about outside variables that may influence hormone levels. Diet, physical activity, and age may have an impact on hormone levels. The researchers clearly accounted for age in this study, but were less clear in their measurements of diet and exercise. While they mentioned these measurements were collected at baseline, it is unclear how this data figured into the final analysis, if at all. Given the known information about hormone levels in our food supply, especially in dairy and meat products, I am interested to know if this would influence overall hormone levels as they relate to breast cancer.
What is the Study Design?
There are several study design types and the design used can influence the success or failure of the study. Rather than trying to teach you the details of each design, I'll give you a tip that I find most useful. When reading a journal article, ask yourself if the researchers were truely able to answer their research question. Sometimes you may find that they were not able to give conclusive answers. In this situation, ask yourself what could have been done differently in order to gleam the most information out of the data collected. Researchers often do what they can with the resources available, and unfortunately this can lead to a study that tells us a whole lot of nothing.
This data for this study came from the Women’s Health Initiative (WHI), which detailed the reproductive history of 155,723 postmenopausal women. The WHI was a longitudinal study where participants were followed over a long period of time, in this case a median of 7.9 years. The study can also be called a cohort study, as a group of participants, known as a cohort, is followed to see if a particular disease develops.
There are several study design types and the design used can influence the success or failure of the study. Rather than trying to teach you the details of each design, I'll give you a tip that I find most useful. When reading a journal article, ask yourself if the researchers were truely able to answer their research question. Sometimes you may find that they were not able to give conclusive answers. In this situation, ask yourself what could have been done differently in order to gleam the most information out of the data collected. Researchers often do what they can with the resources available, and unfortunately this can lead to a study that tells us a whole lot of nothing.
This data for this study came from the Women’s Health Initiative (WHI), which detailed the reproductive history of 155,723 postmenopausal women. The WHI was a longitudinal study where participants were followed over a long period of time, in this case a median of 7.9 years. The study can also be called a cohort study, as a group of participants, known as a cohort, is followed to see if a particular disease develops.
Here we are dealing only with a cohort study. The study was further broken down into two components; an observational study of 93,676 women and randomized clinical trials with 68,132 women. 6085 women were excluded from the study for reasons that included history of breast cancer or mastectomy or lack of follow-up information. 307 of these women developed triple-negative breast cancer and 2610 developed other types of breast cancer.
What are the strengths and weaknesses of the Cohort study design?
The strength of this study is increased by large a sample size and the considerable length of time in which the cohort was observed. I personally favor this type of study because I think it tends to show real results as opposed possible associations. If I'm going to spend time reading about a study, I like to know a firm conclusion can be drawn from the results. Certainly, some studies are more of a starting point for further research, which is great too, but I like something more concrete. In the example, the study period of around 8 years gave me the confidence that the researchers really did find what they say they did. Another advantage to this design is the elimination of recall bias. Recall bias is a bias that occurs because people may remember past events or exposures differently depending upon their current disease status. Because participants in a cohort study are observed in real time, they do not have to recount exposures that may or may not have existed. Accuracy is much improved with this design. This study design also allows researchers to study the effects of more than one exposure and to compare/contrast different levels of exposures. As we see in the example study, researchers did not just examine whether a women had children, but the number of children she had as well as several other exposures. The researchers were concerned with a dose-response relationship, meaning as the exposure increases, the risk of disease increases as well[1]. They were also able to study the different stages of the disease. The researchers themselves note that that the analysis is strengthened by completeness of follow-up and exposure information.
The cohort study design has disadvantages as well, the greatest being the difficulty in executing such a study. It is very costly in terms of resources, manpower and time. Additionally, because of the length of time required in such a study, it may be difficult to track participants throughout the entire length of the study. It can also be difficult to control for and measure extraneous factors outside the realm of the study variables.
My Take
I believe the implications of this study are fairly significant. The results are consistent with previous findings and also improve upon existing evidence. Additionally, the findings warrant further research into the exposures associated with ‘triple-negative’ breast cancer. One limitation of the study, which was addressed by the research team, is that the study was restricted to postmenopausal women, which may mean the results are not generalizeable to younger women. The findings related to oral contraceptive use and breastfeeding may not be accurate due to the population studied and results may have been markedly different had a younger demographic been included.
The journal article may have more implications for medical and research communities than for the general public. While technology is being developed to help women understand their genetic risk for certain types of cancer, this is not available, or even desired, for all women. Even if a woman did know her genetic risk, this information alone may not be enough to influence her reproductive choices. A women who is prone to ‘triple-negative’ breast cancer is not, in my opinion, likely to use this risk factor as a basis for deciding how many children she should have. Therefore, this information in this study will be more pertinent to professionals that will work to further understand these risk factors and to eliminate them.
If you would like to read the on-line article, the link is:
If you would like to read the entire journal article, the citation is as follows:
Phipps, A. I., Chlebowski, R.T., Prentice, R., McTiernan, A., Wactawski-Wende, J., Kuller, L.H., Adams-Campbell, L.L., Lane, D., Stefanick, M.L., Vitolins, M., Kabat, G.C., Rohan, T.E., and Li, C.I. (2011). Reproductive history and oral contraceptive use in relation to risk of triple-negative breast cancer. Journal of the National Cancer Institute, 103, 1-8. doi: 10.1093/jnci/djr030.
[1] Phipps, A. I., Chlebowski, R.T., Prentice, R., McTiernan, A., Wactawski-Wende, J., Kuller, L.H., Adams-Campbell, L.L., Lane, D., Stefanick, M.L., Vitolins, M., Kabat, G.C., Rohan, T.E., and Li, C.I. (2011). Reproductive history and oral contraceptive use in relation to risk of triple-negative breast cancer. Journal of the National Cancer Institute, 103, 1-8. doi: 10.1093/jnci/djr030.
