HEC508 Applied Epidemiology Unit 5 Chapters 7, 9, 10, 11: Randomized Trials, Cohort Studies, Case-Control Studies, and Estimating Risk

Sequence of studies in human populations Initial step: clinical observations at bedside Identify routinely available data Carry out new studies (such as cohort and case-control studies to determine whether there is an association btwn exposure and disease, and whether relationship is causal)

Guidelines for Judging Whether an Association is Causal 1. Temporal relationship 2. Strength of the association 3. Dose-response relationship 4. Replication of the findings 5. Biologic plausibility 6. Consideration of alternate explanations 7. Cessation of exposure 8. Consistency with other knowledge 9. Specificity of the association

Randomized trials Study design used for evaluating new approaches to treatment and prevention “Gold standard” of study designs Begin with a defined population and randomly assign subjects into two groups 1 to receive new treatment 1 to receive current treatment (or no treatment) Follow the subjects in each group to see how many improved in one group compared to the other

Randomized trials—selection of subjects Criteria for who will be included/excluded in study must be explicitly stated/written All participants are volunteers, and informed consent is required regardless of which group they are randomized to Participants should be informed that they may be in the intervention group (new treatment) or control group (current treatment)

Randomized trials—random allocation of study subjects Random assignment gives greatest confidence that groups are alike at the beginning of the trial If you have two groups, and you apply a different treatment to each group, you can only ascribe a difference in outcome to the differing treatment if that is the only factor that differs between the groups Studies without comparison are case studies Studies with comparison (historical controls) and simultaneous nonrandomized controls (may be predictable) Best approach is to use randomization (unpredictability of the next assignment, like in flipping a coin)

Randomized trials Elimination of bias using blinding techniques; in double-blind trials, neither the subject nor the investigator know to which group the subject is assigned Subjects can be blinded (masked) by using a placebo; control group receives some type of intervention, but they won’t know if it is the “new” treatment or the “current” treatment

Randomized trials—sample size Four possibilities in testing whether the treatments really differ 1. The treatments do not differ, and we correctly conclude that they do not differ 2. The treatments do no differ, but we conclude that they do differ 3. The treatments differ, but we conclude that they do not differ 4. The treatments do differ, and we correctly conclude that they do differ. REALITY DECISION Treatment not different Treatment different Conclude treatments are not different Correct decision (1) Type II error (3) Conclude treatments are different Type I error (2) Correct decision (4)

Possible outcomes of randomized trials In the best of worlds, all of the possibilities would be “correct decision” ( cells 1 and 4) Sometimes there is no difference between the therapies, but on the basis of the samples of subjects included in our study, we erroneously conclude that they differ (type I error) Also possible that there really is a difference between therapies, but on the basis of the samples included in our study we erroneously conclude that there is no difference (type II error) Internal validity (whether the study was well done and whether the findings are valid) and external validity (generalizability) are basic concerns in conducting any randomized trial

Cohort Study Design—investigator seeks a group of exposed individuals and a group of non-exposed individuals and follows up both groups to compare the incidence of disease (or rate of death) in the two groups If a positive association exists between exposure and disease, we would expect incidence in the exposed group to be greater than incidence in non-exposed group “Exposure” in most randomized trials is exposure to a treatment or preventive measure (educational intervention)

Cohort Study--types Prospective refers to the fact that the study group is followed forward in time to the future and is contrasted to retrospective, which goes back in time Prospective cohort study—exposure and non-exposure are ascertained as they occur during the study; the groups are then followed up for several years into the future and incidence of disease is measured; it identifies the original population at the beginning of the study and accompanies the subjects until the disease develops (or not) Retrospective cohort study (or historical)—exposure is ascertained from past records and development of disease (or not) is ascertained at the time the study is begun

Case-control study Identify a group of individuals with that disease (called cases) and, for the purposes of comparison, a group of people without that disease (called controls) Determine what proportion of the cases (and of the controls) were exposed and what proportion were not The case-control study begins with people with the disease (cases) and compares them to people without the disease (controls) Case-control study begins with diseased and nondiseased people; cohort study begins with exposed and nonexposed people

Case-control study Limitations in recall and recall bias Matching is the process of selecting the controls so that they are similar to the cases in certain characteristics, such as age, race, sex, socioeconomic status, and occupation Group matching (frequency matching) selects controls so that the proportion of controls with a certain characteristic is identical to proportion of cases with same characteristic Individual matching (matched pairs)—for each case selected for the study, a control is selected who is similar to the case for a certain variable of concern

Cross-sectional studies Both exposure and disease outcome are determined simultaneously for each subject, like viewing a snapshot of the population at a certain point in time We identify prevalent cases because we know that they existed at the time of the study but do not know their duration Cross-sectional study tells us about the distribution of a disease in a population rather than its etiology

Estimating risk Risk is the probability of an event Absolute risk is incidence of a disease in a population To determine whether there is an association between a certain disease and a certain exposure, determine if there is an excess risk of disease Excess risk can be calculated by ratio (disease risk in exposed/disease risk in nonexposed) or by difference (disease risk in exposed – disease risk in nonexposed)

Estimating risk Relative risk asks “What is the ratio of the risk of disease in exposed individuals to the risk of disease in nonexposed individuals?” in a cohort study Odds ratio is a useful measure of association in both case-control and prospective studies In a case-control study, the odds ratio is used as an estimate of the relative risk when the risk of disease is low

Estimating risk Relative risk and odds ratio are important measures of the strength of association, which is important for deriving causal inference Attributable risk is the amount or proportion of disease incidence (or disease risk) that can be attributed to a specific exposure; useful in learning how much of the disease can be prevented if we have an effective means of eliminating the exposure Calculate attributable risk for exposed persons or for the total population (includes both exposed and nonexposed persons)

HEC508 Applied Epidemiology Unit 5 Readings: Chapters 7, 9, 10, 11