Introduction. Step 2: Student Role - Your Plan of Action. Confounding. Good luck and have fun!

Transcription

1 Confounding Introduction You have learned that random error and bias must be considered as possible explanations for an observed association between an exposure and disease. This week we will examine the role of confounding. Unlike random error or bias, confounding is a property of the study population, and occurs when the effect of an exposure on an outcome is mixed together with the effect of a third variable. The following exercise will examine the properties of confounders and describe methods to adjust for confounding through both study design and analysis. (Please see Aschengrau Chapter 11 for more information). Good luck and have fun! Step 1: Learning Objectives A. Basic elements of confounding 1. Define confounding and distinguish it from bias and chance error 2. Identify three criteria a variable must fulfill to be a confounder in an epidemiological study 3. Diagram the relationship of a confounder with exposure and outcome B. Explain methods to adjust for confounding 1. Describe ways of handling confounding at the design phase of a study a. Randomization b. Restriction c. Matching 2. Describe ways of handling confounding at the analysis phase of a study a. Stratification b. Multivariate adjustment techniques C. Describe how to evaluate potential confounding in epidemiological data 1. Explain the difference between a crude and adjusted effect estimate 2. Discuss what is meant by "residual" confounding Step 2: Student Role - Your Plan of Action Reading and understanding scientific articles takes a critical mind. You must carefully evaluate the methodologies employed and conclusions that are formed. At the end of the semester you will be asked to critique a published article. Your work must include some of the core elements highlighted in the summary of Dr. Stellman's hypotheses presented in the paragraphs that follow. Pay particular attention to the areas that are highlighted. If you have any questions, do not hesitate to discuss them with your seminar leader. Please refer to the Student Role section of the Bias module to read the synopis of Dr. Spellman's study. Work through this [interactive exercise ] which demonstrates the "mixing of effects" when confounding is present in the data. 1

2 Step 3: Design Questions 1. Hospital room status, a marker of socioeconomic status, was one of the potential confounders considered in the study of artificial sweetener use and bladder cancer. Please explain why hospital room status was considered to be a potential confounder. a. Hospital room status, a marker for socioeconomic status, was considered a confounder because it is a risk factor for bladder cancer, and is not in the causal pathway of interest between artificial sweetener use and bladder cancer. Correct. For a variable to be a confounder, it must fulfill three basic properties: 1) associated with the exposure, 2) a risk factor for the disease, and 3) not in the causal pathway of interest. Hospital room status, a marker of socioeconomic status, is a potential confounder because it is associated with artificial sweetener use, is a risk factor for bladder cancer, and is not in the causal pathway between sweetener use and cancer. b. Hospital room status, a marker of socioeconomic status, is in the causal path between artificial sweetener use and bladder cancer. Incorrect. A confounder cannot be an intermediate step in the causal pathway of interest. If a third variable is in the causal pathway of interest, it is not a confounder but a mediator. c. Hospital room status, a marker of socioeconomic status, is a risk factor for bladder cancer but is not associated with artificial sweetener. Incorrect. A variable must fulfill three basic properties to be considered a confounder- one, that it is associated with the exposure of interest. 2. How was confounding handled in the design stage of the study? a. Randomization of subjects into cases and controls Incorrect. Subjects were not randomized in this study. Subjects can only be randomized in experimental designs (i.e., RCT s). The behavior of subjects cannot be manipulated by investigators conducting observational studies (i.e., cohort or case-control studies) b. Restriction of cases and controls within a narrow age category Incorrect. In this study cases and controls were not restricted to any specific age category. c. Matching controls to cases on selected characteristics Correct. Controls were matched to cases by age (in decades), sex, hospital, and hospital-room status (private, semiprivate, or ward). Matching in a case-control study is intended to created comparability in the underlying source population of exposed and unexposed by creating mini-studies in which all individual in the study are, for example, in a private ward such that ward can have no effect on the exposure disease relationship. Step 4: Data Analysis Questions 3. Explain how you would assess whether a potential confounder alters an effect estimate after adjusting for it in a multivariable model. a. Look at the crude OR 2

3 Incorrect. You must compare the crude and adjusted OR s to evaluate confounding. Remember, the crude estimate simply reflects the association between the exposure and outcome; it does not take into account the effect of potential confounders. b. Look at the adjusted ORs Incorrect. You must compare the crude and adjusted OR s to evaluate confounding. Remember, the adjusted estimate simply reflects the association between the exposure and outcome after controlling for a potential confounder. Without the crude to compare back to, we would not know what happened to the OR after taking other risk factors into account. c. Compare the crude OR to the adjusted OR Correct. It is important to compare the adjusted OR with the crude OR to see the change in the effect estimate. In this study, it was important to compare the crude OR which assessed the association between artificial sweetener use and bladder cancer only, with the adjusted ORs, which were calculated by taking into account age, hospital, hospital room status, year of interview, and education (see Table 5 of Step 2: Student Role - Your Plan of Action). [ Follow this link to learn more about the evaluation of confounding.] (Note: In the popup window, be sure to scroll down after each correct answer.) 4. Would confounding due to socioeconomic status still be a problem if the investigators chose to conduct a cohort study instead of a case-control study? a. Confounding would not be a problem in a cohort study Incorrect. Confounding is a problem in all study designs. Remember, confounding is a mixing of effects between an exposure, outcome, and third variable. It is not the result of study design or investigators. Instead, confounding results from the fact that risk factors are not evenly distributed between comparison populations (i.e., exposed and unexposed groups). b. Confounding would still pose a problem in a cohort study Correct. Confounding is a problem for all observational study designs. Because epidemiology research concerns human populations, we must always consider that certain characteristics (e.g., age, sex, income, etc.) may be unevenly distributed in our study populations. c. Confounding would be minimal in a cohort study compared to case-control study Incorrect. Confounding can be just as large in a cohort study as it is in a case-control study. Regardless of design, it is important to consider potential confounders in your work and adjust for them appropriately. 5. Stellman et al. matched controls to cases on four factors: age, sex, hospital, and hospital room. Other case-control studies match on fewer factors. Which technique is better and why? a. The other studies are better because it is best to match controls to cases on as few factors as possible. Incorrect. One should match for the minimum amount of factors necessary, but that amount differs from study to study. 3

4 b. Stellman et al. s study is better because it is best to match controls to cases on as many factors as possible. Incorrect. While it may be helpful to match controls to cases on several key confounders, matching on too many factors may be harmful to your study. Once controls are matched to cases by a particular variable, you can no longer assess that variable s effect on the outcome of interest. Furthermore, it is both difficult and expensive to find appropriate controls when matching on several factors. c. It is not possible to determine which study would be better at controlling for confounding by looking at the number of matched factors. Correct. It is not possible to determine whether one study better controls for confounding based on the number of matched factors. 6. What if during data analysis investigators found that the use of vitamin supplementation was associated with artificial sweetener use and was an independent risk factor for bladder cancer. Should they attempt to control for this potential confounder? a. Yes, investigators should control for vitamin supplementation as they did for other potential confounders and add this variable to the list of hypothesized confounders in the Methods section. Incorrect. Many variables may act as confounders in one study. While it is important to hypothesize which factors may confound an association, it is also important to evaluate other potential confounders during the analyses as well. In doing so, you must report the process of how you selected potential confounders (i.e., a priori confounders in the Methods section and a posteriori confounders in the Results section). b. Yes, investigators should control for vitamin supplementation and describe the process of confounder selection in their Results section. Correct. It is important to report the selection process of confounding variables in your work. A priori confounders should be reported in the Methods section, a posteriori confounders in the Results section. c. No, it is inappropriate to control for variables if they were not hypothesized as confounders a priori. Incorrect. It is not always possible to know all potential confounders at the beginning of a study. This may happen when investigating an exposure-disease association which has not been studied well, or if cost and feasibility make it impossible to address all potential confounders at the design phase of a study. Therefore, it is necessary to consider confounding at the analysis phase of a study as well. 7. Suppose investigators wanted to determine if confounding was present during the analysis stage of the study. What can be done at this stage to assess confounding? a. Perform stratified analyses Correct. Stratified analysis is one way to control for confounding at the analysis phase of a study.stratification means the effect of an exposure is evaluated within strata (levels) of a confounder (e.g., gender, looking at the exposure-disease association between males only and then females only). Once you calculate the OR s for each stratum (and if they are similar to one another) you then compare them with the crude OR. If there is a large difference (i.e., >10%) between the stratified and crude OR s, you can conclude that confounding is present. b. Use restriction 4

5 Incorrect. Restriction is used to control for confounding at the design phase of a study. For example, researchers can restrict the study to women aged Thus, both cases and controls would be women aged As a result of such restriction, it would no longer be possible to investigate the effects of sex and age on the outcome of interest. c. Conduct matched analyses Incorrect. As Stellman et al s work has shown, matched analysis is another way to control for confounding at the design phase of a study. 8. Shapiro, et al. matched controls to cases on two factors: age and geographic area, whereas Stellman and Wynder matched controls to cases on four factors: age, sex, hospital, and hospital room. Which study do you think is best at controlling confounding at the design phase of the study? a. Shapiro et al's study is better because it is best to match controls to cases on fewer factors. Incorrect. Although it is not good to match controls to cases on too many factors, matching on fewer factors in itself does not guarantee that confounding is accurately controlled. b. Stellman and Wynder's study is better because it is best to match controls to cases on more factors. Incorrect. A confounder cannot be an intermediate step in the causal pathway of interest. If a third variable is in the causal pathway of interest, it is not a confounder but a mediator. c. It is not possible to determine which study is better at controlling confounding by looking at the number of matched factors. Correct. It is not possible to determine whether one study is better than the other at controlling confounding by the number of factors matched. What should be of foremost importance when controlling confounding is whether confounding variables were measured properly and that their effects were removed at the analysis stage. 9. Age was a potential confounder in this study. Choose an appropriate diagram representing the relationship of this potential confounder with exposure and outcome. a. Correct. In this diagram age meets the requirements to be a confounder because, as depicted in the diagram, age is a risk factor for endometrial cancer and is associated with estrogen use, but it is not a result of estrogen use. b. Incorrect. This diagram illustrates that age is an intermediate in the pathway between estrogen use and endometrial cancer. If a factor is in the pathway between exposure and outcome, it is called a mediator. 5

6 c. Correct. Controls were matched to cases by age (in decades), sex, hospital, and hospitalroom status (private, semiprivate, or ward). Matching in a case-control study is intended to created comparability in the underlying source population of exposed and unexposed by creating mini-studies in which all individual in the study are, for example, in a private ward such that ward can have no effect on the exposure disease relationship. 10. Suppose that during data analysis, investigators found that the number of abortions performed was associated with estrogen use and was an independent risk factor for endometrial cancer. Should they attempt to control for this confounder? a. Yes, investigators should still control for this confounder as they did for other confounding variables. Incorrect. Since attempts to minimize confounding can only be made for known confounders, it is necessary to look for confounders during the analyses as well. However, it is also important to report the process of selection of confounding variables in the description of results (i.e., a priori confounders vs confounders which were identified in the data analysis step). b. Yes, investigators should still control for this confounder and describe the process of confounder selection in the description of the results. Correct. It is important to report the process of selection of confounding variables in the description of results (i.e., a priori confounders vs confouders which were identified in the data analysis step). The information on the process of adjustment for confounders is usually placed in the Methods section for published articles. c. No, investigators should not control for any confounders which they did not specify a priori. Incorrect. It is not always possible to know all the potential confounders at the beginning of the study. This may happen when investigating an exposure-- disease association which has not been studied well or if cost and feasibility may make it impossible to address all potential confounders at the design phase of a study. Therefore, it is necessary to consider confounding at the analysis phase of a study as well. Step 5: Discussion Questions 1. Which study design offers the best opportunity to control for confounding -- randomized clinical trial, cohort study, or case-control study? Explain your reasoning and make examples to prove your point. 2. Suppose that in the study of artificial sweetener and bladder cancer you wanted to evaluate the hypothesis that bladder cancer varies by geographic area. Would you match on geographic area? Please, explain your answer. 3. A principle of case-control studies is that controls should be selected independently of exposure status. Under what circumstances would matching violate this principle? What can be done about this type of violation? Questions for the Intellectually Curious: Why is it important to distinguish between confounding and confounders? 6