Chapter 5: Cohort and Case-Control Studies

CHAPTER OBJECTIVES

• Describe the design characteristics of case-control and cohort designs

• Describe the strengths and limitations of case-control and cohort designs

• Identify common biases in case-control and cohort designs

• Calculate and interpret common measures of association used in case-control and cohort designs

KEY TERMINOLOGY

• Attributable fraction in the exposed

• Attributable risk

• Attributable risk percent

• Base population

• Bias

• Case-control

• Cases

• Closed cohorts

• Cohort study

• Concordant pair

• Confounding factor (or confounder)

• Controls

• Detection bias

• Diagnostic bias

• Differential misclassification of the exposure

• Discordant pair

• Effectiveness

• Efficacy

• Exposure

• Exposed group

• Fixed cohort

• Hospital or clinic-based case-control

• Incident cases

• Induction period

• Information bias

• Latency period

• Matching

• Measures of association

• Measures of effect

• Nested case-control

• Nondifferential misclassification

• Odds

• Odds ratio

• Open cohort

• Population-based case-control

• Prevalent cases

• Prevented fraction in the exposed

• Primary data

• Prospective cohort studies

• Rate ratio

• Recall bias

• Reporting bias

• Retrospective cohort studies

• Risk difference

• Risk factor

• Risk ratio

• Secondary data

• Selection bias

• Source population

• Temporal ambiguity

• Unexposed group

• Validity

INTRODUCTION

Observational epidemiologic studies such as the cohort and case-control designs are used to inform patients, clinicians, and policy makers on a wide variety of topics, including the effects of drugs and the influence of different pharmacy services on a range of outcomes in humans. Though there are more challenges in establishing causal inferences with observational studies compared to randomized-controlled trials (RCTs), these studies are the only ethical way to gain insights on the effects of exposures known to have harmful effects that would be of interest to clinicians. For example, the effect of exposure to tobacco smoke, illicit drug use, or a nutrient poor diet on disease risk cannot be examined using RCTs because it would be unethical to purposefully expose persons to these types of conditions that pose harmful effects while having virtually no health benefits. The only ethical way to estimate the effect of these exposures on disease risk in humans is by carefully conducting observational studies.

Although RCTs are considered the “gold standard” to estimate the effect of an exposure (or treatment) on a disease outcome in humans, they are still subject to bias and other limitations. RCTs often have limited generalizability; they are primarily designed to assess efficacy, the effect of a treatment in ideal settings on disease outcomes, rather than effectiveness, the effect of the treatment in typical clinical settings on disease outcomes. Most often clinical trials are not adequately powered to estimate the effects of treatment on rare clinical endpoints and side effects (adverse events). For example, a cohort study detected an elevated risk of cardiac death associated with azithromycin that was undetected in clinical trials.¹ RCTs also often do not follow study subjects long enough to estimate the effect of treatments on clinical endpoints which take a long time to develop.² Further, if patients in clinical trials do not adhere to (or comply with) their randomly assigned treatment, ...