Important Note

Tufts ended funding for its Open Courseware initiative in 2014. We are now planning to retire this site on June 30, 2018. Content will be available for Tufts contributors after that date. If you have any questions about this please write to

Tufts OpenCourseware
  • To understand the concepts of confounding and effect
  • Use these concepts in developing a hypothetical research

Outside Preparation: Due Small Group # 2


The baseline characteristics of subjects in a trial of Medicine A vs. Medicine B with myocardial infarction as an endpoint follow:

Medicine A Medicine B
% Type A Personality 30 05
% Diabetics 20 20
% Physically Inactive 80 80

The authors reported that 10% of subjects on Medicine A had a myocardial infarction compared to 6% of subjects on Medicine B, with p=0.20. Alpha was set at 0.05. The authors said that the trial had a 70% power to detect a true difference of 10% or more between the two treatment arms. The authors reported that after adjusting for type A personality, there was no change.

  1. Was type A personality a confounder in this study? Explain your answer.
  2. It is known that physical inactivity is a risk factor for myocardial infarction. The authors did not report a physical inactivity adjusted relative risk. Should the authors have reported this adjustment? Explain your answer.
  3. What is the likelihood that a Type I error occurred in this study?
  4. The authors wonder if diabetics on Medicine A might have fewer myocardial infarctions than diabetics on Medicine B. Is this an internal or external validity issue? Explain your answer.
  5. The trial was double-blinded. A committee of cardiologists, also blinded to the assignment groups, was responsible for assessing whether or not a subject had a myocardial infarction. They noted that it was difficult to determine whether or not several subjects actually had a myocardial infarction or not. However, they took a vote on these cases and the majority determined whether or not a subject had a myocardial infarction. Given the information provided in this paragraph, what type of bias could have occurred?
  6. This was a large randomized controlled trial having more than 5,000 subjects with little loss to follow-up. Soon after the results were published in a medial journal, another group of researchers reported that C reactive protein, a non-specific marker for inflammation, is an important risk factor for myocardial infarction. Some physicians then criticized the findings of the Medicine A vs. Medicine B study because the authors did not have baseline data on C reactive protein levels for subjects in each group. Do you think it is more likely than not that the Medicine A vs. Medicine B trial had confounding by C reactive protein levels? Explain your answer.
  7. The trial was also criticized because there were relatively few women in the trial. Is this an internal or external validity issue? Explain your answer.


  1. Review lecture notes on Lecture 4 - Threats to Validity and Hypothesis Testing.

Approximate Class Schedule:

30 minutes Instructor review of key concepts from Lectures 4 and 5
30 minutes Review of homework assignment
30 minutes The class will be divided into small groups to answer the questions
30 minutes Review of lab assignment

To Be Completed in Small Group Session

After reviewing some molecular science data, you wonder if there is an association between cigarette smoking and colon cancer. It is already known that family history in a first-degree relative, a type of inflammatory bowel disease (IBD), age and certain rare hereditary bowel diseases are risk factors for colon cancer. Smokers can be current or former smokers with varying amounts of smoking history. When former smokers quit smoking is also variable. You also wonder if passive exposure to cigarette smoke, so-called secondary smoking, might be an important factor to consider. However, to keep the analyses simple, you decide to dichotomize smoking into current or not current.

    1. Design a study that would assess your hypothesis given the information provided above.
    2. Do you anticipate any difficulties getting the study approved by the Institutional Review Board, which is responsible for approving “experiments” on humans?
    3. Specify why you selected your type of study design.
    4. What study population will you use to recruit subjects? Why did you select that study population?
    5. State the null and alternate hypotheses.
  1. What power did you select for your study and what considerations were parts of this decision. Remember that power also requires you to specify a minimum difference you’d like to detect if the difference really exists.
  2. Your study is underway. Following are the baseline characteristics of the subjects in the study:
    Current Smokers Non Current Smokers
    Family History in First Degree Relatives 20% 10%
    IBD History 0 0.5%
    Hereditary Bowel Disease 0 0
    College Graduates 5% 25%
    Vegetarians 20% 20%
    Mean Age 45 65
    Active Exercisers 2 15%
    Females 60% 60%
    1. Based on the data in this table, list your concerns about potential confounding.
  3. How could you address your concerns about confounding?
  4. Assume your study results indicate that current smokers have a 10% higher risk of developing colon cancer vs. individuals who are not current smokers and that result is statistically significant.
    1. When you adjust for active exercising, the relative risk does not change. Was active exercising a confounder? Explain your answer.
    2. When you adjust for family history of colon cancer in first-degree relatives, the relative risk does not change. Does this mean that family in first-degree relatives is not a risk for colon cancer? Explain your answer.
  5. Could gender be a confounder is this study? Explain your answer.
  6. The study is over and you are analyzing the data. You want to determine if gender is an effect modifier. How would you set up an analysis to make this determination? Draw the 2X2 diagrams below.
  7. What would you expect to see if gender were an effect modifier in the study?
  8. What is the essential difference between confounding and effect modification?
  9. Why is it important for physicians to understand effect modification as it relates to giving advice to patients?