Question
The following questions use data and situations adapted from: Nickson C, Mason KE, English DR, Kavanagh AM. Mammographic screening and breast cancer mortality: a case-control study and meta-analysis. Cancer Epidemiol Biomarkers Prev. 2012 Sep;21(9):1479-88.
This study was designed to investigate the association between mammograms and breast cancer mortality among women over the age of 50. “Cases” were breast cancer patients who had died as a result of their cancer, and “controls” were women who were still living at the time of the study, regardless of their breast cancer diagnosis. Participants were categorized as “exposed” if they had received at least one mammogram any time after their 50th birthday. The investigators collected the following data:
Cases
|
Controls
| |
Exposed
|
167
|
2051
|
Unexposed
|
264
|
1599
|
The investigators were also concerned that socioeconomic disadvantage might be a potential confounder or effect modifier. They measured SES and found the following:
Low SES:
Cases
|
Controls
|
Total
| |
Exposed
|
56
|
1095
| |
Unexposed
|
131
|
1107
| |
Total
|
2389
|
High SES:
Cases
|
Controls
|
Total
| |
Exposed
|
111
|
956
| |
Unexposed
|
133
|
492
| |
Total
|
1680
|
a. Calculate the crude OR and the stratum-specific ORs for those with low and high SES.
b. Calculate and interpret the adjusted OR.
c. The p-value on the chi-squared test for homogeneity is 0.98. State the null hypothesis being tested. Given these results, what do you conclude about the role of SES? Is it a confounder, effect modifier, or neither? Justify your answer in 1-2 sentence
d. The investigators excluded from their study any breast cancer patient who had had a mammogram before the age of 50. However, they note that this excluded women who were screened earlier in their lives because they have a particularly high risk of developing a rapidly fatal form of breast cancer. What type of bias might this introduce, and how would you expect it to affect your OR?
e. The authors note that family history of breast cancer might be a confounder, but they did not have the data necessary to control for this. If family history makes women more likely to receive a mammogram, and more likely to die from breast cancer, in which direction would you expect this unmeasured confounding to bias the results?
f. Name two options the investigators would have had to account for family history of breast cancer in the design phase of the study.
Source: 250B Problem Set
