Question

A hospital-based case-control study of alcohol consumption in relation to a particular type of cancer was conducted. The investigators selected patients admitted for conditions that were not related to alcohol consumption as the control group. 200 controls were individually matched to 200 cases by age and sex. Alcohol consumption was assessed by an in-person interview conducted by a trained interviewer in the hospital (suggested length, a brief paragraph each for 1) to 3) below).

1. Using epidemiological terminology, describe three potential biases which you may be concerned about in this example, and why you think such biases might have arisen.

2. For each of the three sources of bias, briefly describe steps that could be taken to control or remove the potential biases in the design of your study and analysis of study data.

3. Why was it important that the control group included patients with conditions that were not related to alcohol consumption?

4. The matched odds ratio for the association between alcohol consumption and cancer was 2.0.  Given that, create a contingency table that would result in this odds ratio based upon the matched data (note:  there are many possible correct responses to this question).

Answer 1

1. three potential biases

• 1. Selection bias - it occures when subjects are selected to the control group, that they are not stem from the same source population that produced in the case.
• 2. Information bias or recall bias- The informant's answer may be influenced by knowledge about topic under investigation or disease experience itself.
• 3. Confounders- confounders are variables other than study variables. These factors can affect disease directly or indirectly . It should be identified earlier prior to the start of the study. In this study, smoking, life style practices, genetic factors are also causes cancer. These factors affect association.

2. steps that could be taken to control or remove the potential biases in the design

1. selection Bias- select subjects to case and control group with same risk exposed and potential to develop outcome/ disease.

• matching - it ensures comparability between cases and controls and reduces variability and systematic differences,due to background variables. Can be done by individual or group matching such as age, sex.
• 2. Information bias - can be reduced by collecting accurate data along with biomedical data.
• 3.confounding- it can be controlled by study design . Conduct or analysis level.
• Study level:
• restricted to one group. in this study, samples included in the case have alcohol consumption and cancer and they are compared with non alcoholic groups( control).
• matching- cases and controls are matched by age and sex
• Analysis level-collect relevant information about confounders and adjust during analysis-Stratification and mutivariate adjustment.

3. Why was it important that the control group included patients with conditions that were not related to alcohol consumption?

• To restrict confounding variables to one group.
• If controls selected from the same exposure(Alcohol consumption) may treated for other diseases like myocardial infarction, liver disease. This may result in an underestimation of strength of association between exposure and outcome.

4. create a contingency table that would result in this odds ratio based upon the matched data

2x2 contingency table

Predictor varable / cause alcohol consumption	case cancer -present	control cancer- abscent
yes	alcohol consumption +, disease+	Alcohol consumption +, no disease
no	No alcohol consumption, disease present +	no alcohol consumption, no disease