Cohort study is one type of study design that is normally used to address risk factors/incidence of a certain disease. This type of study usually starts at certain time point with baseline measurement of participants and divides them into different groups, and then follows participants for a long time period to see whether those participants develop a certain disease. In this way, it is easy to address the predictor/factors that would cause/influence this disease/outcome.
As I said in my previous post, my colleagues and I designed a cohort study regarding to test whether people’s clothing preferences would influence people’s weight status. Also, my current research project is to discover predictors for breast cancer-related lymphedema. So, what is lymphedema? Lymphedema is a chronic progressive disease often caused by cancer treatment, especially in patients who require surgical removal of or radiation to lymph nodes. Breast cancer survivors are at life-time risk of developing lymphedema their cancer treatment likely included surgery or radiation treatment, which may adversely affect the lymphatic system. Therefore, it is a big population based chronic disease.
In this study, we followed breast cancer patients for five years starting from pre-operation as baseline measurement, which means the time period before the patients having breast cancer surgery. We then follow them based on the timeline as shown in Figure 1  and see whether they develop lymphedema afterwards. At every lab visit, we interviewed patients for their symptoms and measured their BMI, etc. The ultimate goal for this study is to discover risk factors of developing lymphedema among all different potential exposures (such as a certain symptom or combination of some symptoms, BMI, etc.) for breast cancer survivors.
Figure 1. Timeline for data collection, where Ti represents the ith visit of a patient. 
This study started in 2007, we’ve been following our patients (n=316) for 4 years. The main challenge that I found is follow-up bias. It is really hard to follow people for such a long time period. Even though we did call them several times before their next visit, patients might still not come. Also, even if they did come, they might not obey the timeline that we expect. In addition, patients might exit study or be lost contact. This study is only five year study which is considered as a very short time period for a cohort study, regarding to what Gordis suggested, which is twenty years.  Also, there is a funding issue. For this five year study, we got to apply for funding every two/three years. Normally, funding won’t have such a long period. It is hard to continue if it is not funded. And we need to report exciting findings in our annual report in order to keep funding…
Even though I think it is a wonderful and promising research project which could increase awareness of lymphedema, it is hard to get funding and follow participants over time. And there must be unmeasured confounders in our study design, because it is not practical to cover all possible confounders. Additionally, we didn’t have a hypothesis at the beginning which makes the study even harder to measure, because we want to discover as many as possible risk factors of developing lymphedema. Therefore, we could publish a guideline for practitioners of how to manage and control lymphedema. Although, this study has some drawbacks, we could still learn from the findings of this cohort study.
1. Xu S, Shyu CR. Efficient selection of association rules from lymphedema symptoms data using a graph structure. AMIA Annu Symp Proc. 2010 Nov 13; 2010:912-6.
2. Gordis, L. (2009). Epidemiology. Philadelphia, PA: Saunders.