Epidemiologic Methods
Case Control Study and Confounding.
The following set of questions is based on a short report by Karami et al. "A case-control study of occupational sunlight exposure and renal cancer risk." https://www.ncbi.nlm.nih.gov/pubmed/26505275
1. Briefly discuss two reasons why a case-control study is (or is not) well suited to examine risk factors for renal (kidney) cancer. (2 points/reason)
Please support each answers/reasons with the information/statistics provided in the article.
2. The authors describe the study design of their previous study assessing renal cell carcinoma (RCC) as "a hospital-based case-control."
Is the same study design used in this study?
a) Please state yes/no and
b) Provide brief rationale for your answer.
As you provide the rationale, include issues regarding the selection of cases, selection of controls, and implications for study validity. Include supporting information from the article to receive full credit.
3. Cases were selected from the Chicago hospital pathology reports and the Metropolitan Detroit Cancer Surveillance System. Which of the following categories of study design best describes this method of case finding? Choose one best answer.
A. Community-based screening
B. Cross-sectional survey
C. Hospital-based surveillance
D. Passive surveillance
E. Prospective follow-up
4. The authors indicate the controls were "frequency-matched to cases on sex, age, race and residential area."
A. Briefly describe what is meant by "frequency matching", how it is implemented, and how it differs from individual matching. State why frequency matching was used (instead of individual matching).
B. Please write "true" or "false" for the following statements:
Matching is a common strategy to control for confounding in cohort and case- control studies.
If matching is done according to a strong confounder, it tends to increase the statistical power (efficiency) of the study.
When multiple variables are being matched for, it may be difficult or impossible to find a matched control for a given case.
When matching was done, the association between sex, age, race, and residential areas and the outcome (i.e., RCC) can be assessed.
It is possible to assess additive interaction in this study between matched case- control studies between sex, age, race, and residential areas and the exposure of interest (i.e., vitamin D).
Matching may result in a reduced "external validity" and an increased "internal validity."
Matching can be undone during the analysis stage.
When matching is conducted according to categorical definitions of continuous or ordinal variables, there will be no residual confounding (i.e., residual differences between cases and controls).
5. Name and describe a type of bias that may have been introduced by the authors' methods of collecting controls' data?
6. Even if the authors were careful in the selection of cases and controls, selection bias can make the interpretation of results difficult. Which of the following was NOT a situation that can produce selection bias? Choose one best answer.
A. The exposure has some influence on the process by which controls are selected.
B. The exposure has some influence on the process of case ascertainment.
C. The disease status has some influence on the recall of exposures.
D. The exposed cases are reported to registries more than unexposed.
E. All of the above will produce selection bias.
7. The authors state, "...black and white participants, 20-79 years, with histologically confirmed incident RCC from 2002 to 2007 were identified from Chicago hospital pathology reports and the Metropolitan Detroit Cancer Surveillance system." (page 2) What are the advantages of selecting incident (as opposed to prevalent cases)? (Hint: think of definition of prevalence in terms of incidence and duration).
A. Selecting from a pool of incident cases creates a more homogenous case group with regard to unknown confounding factors.
B. Selecting from a pool of incident cases reduces misclassification bias.
C. Selecting from a pool of prevalent cases would make exposure assessment more difficult because of pre-existing disease status.
D. Selecting from a pool of prevalent cases would make separation of factors associated with risk and those with survival more difficult.
8. Group/distribute the following biases in one of two groups (0.5 point each): Group A: Selection Bias Group B: Information Bias
Neyman bias
Berkson's bias
Publication bias
Nonresponse bias
Loss to follow-up bias
Recall bias
Observer/interviewer bias
Ecological fallacy
9. Misclassification bias:
(a) define misclassification bias and
(b) could there be a misclassification bias in the study? Why or why not?
10. Is there a difference in weighted estimates reported in Table 1 vs. Table 2? Briefly explain to receive full credit (HINT: think in terms of the underlying analyses conducted to produce the reported estimates and the concept of confounding).
11. Refer to Table 1 to answer the following set of questions:
(a) What is the overall odds ratio for the risk of RCC from having a family history of cancer (including kidney and any other cancer)? (If you provide details of your calculations you may receive partial credit even if the result is incorrect).
(b) Please briefly interpret your results.
(c) Can a relative risk be calculated instead? Why or why not?
12. On page 3, the authors state "Compared with participants in the lowest exposure quartile, significant risk reductions were observed for those in the highest two quartiles for cumulative, frequency-adjusted duration and frequency-adjusted duration of exposure among those who held only low-intensity jobs." In this context, the term "risk" refers to which of the following measure(s)?
A. Attributable risk
B. Cumulative incidence
C. Estimated odds ratio
D. Incidence density
13. Examining the data in Table 2 and answer the following set of questions (1 point each except question (f)):
(a) Define a confounder.
(b) How do you decide before a study whether a variable is likely to act as a confounder?
(c) How do you decide after a study whether a variable is likely to act as a confounder?
(d) What confounding variable(s) is (are) being controlled for in Table 2?
(e) What methods was/were used to control for potential confounding?
14. Please focus on estimates of RCC odds given the frequency-adjusted duration of exposure among any high-intensity jobs (years) in Table 2.
(a) What is the crude odds ratio from 8.68+ years of exposure among those with high-intensity jobs? (i.e., you will need to estimate it using the data provided in the Table 2. Please pay attention to your reference group).
(b) How is your estimate differ from the one reported in the table? Why does it differ or does not differ?
(c) Are sex, race, smoking status, age, BMI, hypertension, family history of cancer, education, center, and dietary vitamin D intake positive or negative confounders? Please elaborate.
(d) What is the common (unadjusted) odds ratio for exposure among those with high intensity jobs across both white and blacks, estimated via the Mantel- Haenszel formula?
(e) Is race a positive or negative confounder? Please elaborate.
(f) Calculate the Mantel Haenszel chi-square to test for an association, and derive a test-based confidence interval for the odds ratio in "d" above.
(g) What other possible confounder(s) should we suspect? Can data about these confounder be obtained from study participants or other sources?