CONFOUNDING AND EFFECT MEASURE MODIFICATION

CONFOUNDING AND EFFECT MEASURE MODIFICATION

Class Activity

1. A case-control study was conducted to investigate the hypothesis that lifetime coffee consumption is associated with the risk of liver cancer. For the study, 300 patients under 55 years of age with incident liver cancer, diagnosed between 1994 and 1997, were identified from the Detroit SEER registry records. A random sample of 300 individuals with a distribution of age, sex, and race that was similar to the cases was also identified. All of the study participants were interviewed to assess average lifetime coffee consumption and other covariates, including smoking. Average lifetime coffee consumption was determined as drinking at least 5 cups of coffee per week. The data from the case-control study are as follows:

 

Average Coffee Consumption	Cases	Controls	Total
≥ 5 cups	203	138	341
< 5 cups	97	162	259
Total	300	300	600

 

 

1. Compute the Odds Ratio:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Coffee drinkers are often smokers, so the data from the case-control study were stratified by smoking, as shown below:

Average Coffee Consumption	Never Smokers		Ever Smokers
	Cases	Controls	Cases	Controls	Total
≥ 5 cups	27	114	175	25	341
< 5 cups	31	76	67	85	259
Total	58	190	242	110	600

 

1. Compute the OR for the association between average coffee consumption and liver cancer in never smokers and ever smokers:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1. Is smoking a confounder, an effect modifier, or neither? Explain why.

 

 

 

 

 

 

1. If appropriate, calculate the Mantel-Haenszel odds ratio. If it is not appropriate to calculate the Mantel-Haenszel odds ratio in this situation, what would you report?

 

 

 

1. What are the potential sources of bias in this investigation?

 

 

 

 

 

 

 

 

 

1. If cases and controls were misclassified according to their exposure status (coffee drinking), but misclassification was similar among cases and controls, what is the effect on the odds ratio?

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1. Suppose misclassification was not similar among cases and controls (for example, if cases were more likely to report coffee consumption as compared to controls), what is the effect on the odds ratio?

 

 

Question 2

2. Between 1993 and 1996, investigators from the University of Coffeeville enrolled 1,000 participants aged 50-70 years from the Light Brew community based on their physical activity levels. The aim of the study was to examine the association between coronary heart disease (CHD) and exercise. Data on physical activity, race, gender and other covariates were collected at baseline through face-to-face interviews, with no subsequent reassessment. On the basis of the information provided in the interview, study participants were classified as either “low physical activity” or “high physical activity”. Study participants were followed for 5 years. The following data were obtained:

 

Physical activity	CHD	No CHD	Total
Low	300	350	650
High	60	290	350
Total	360	640	1000

 

 

1. Calculate the relative risk:

 

 

 

 

 

 

 

 

 

 

 

The investigators were concerned that gender might be influencing theobserved association, so the data were stratified as follows:

Physical Activity	Women		Men
	CHD	No CHD	CHD	No CHD	Total
Low	291	297	9	53	650
High	50	162	10	128	350
Total	341	459	19	181	1000

 

1. Calculate the relative risks for the physical activity and CHD association in males and females. Then use the 10% rule to determine whether there is confounding present.

 

 

 

 

 

 

 

 

1. Is gender associated with activity levels? Is gender associated with CHD?

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1. Why is it important to study the association described in part c?

 

 

 

 

 

1. Is gender a confounder, an effect modifier, or neither in this study? Explain why.

 

 

 

 

 

 

 

 

1. If appropriate, calculate the Mantel-Haenszel relative risk. If it is not appropriate in this situation, what would you report?

 

 

 

 

 

 

 

 

1. What are the potential sources of bias in this investigation?