Study Design Notes

NOTES NOTES STUDY DESIGN SAMPLING osms.it/sampling ▪ Selection of individuals for study from specific population ▪ Aims to represent, estimate characteristics of that population PLACEBO EFFECT & MASKING osms.it/placebo-effect-and-masking WHAT IS THE PLACEBO EFFECT? ▪ Refers to situation where study participant’s belief in treatment brings about positive effect ▫ E.g. individuals given placebo drug tend to report improvements even when treatment has no real effect ▪ Placebos can be affected by study participant’s psychological responses to context in which treatment is taking place ▪ Placebo can be drug/pharmacologically inactive substance indistinguishable from an active treatment/can be based on any expectation the person may have about intervention under study ▪ Useful in studying rate of side effects, reactions to drug WHAT IS MASKING? ▪ Subjects and/or investigators are unaware of treatment assignment ▫ Single blind: subjects are unaware of treatment assignment ▫ Double blind: subjects, investigators are unaware of treatment assignment ▫ Triple blind: treatment administrator unaware of treatment assignment CASE-CONTROL STUDY osms.it/case-control_study ▪ Study that determines potential risk factors in individuals with condition ▪ May rely on individual recall, past medical history, autopsy ▪ Example: Percentage of people who gave 74 OSMOSIS.ORG birth to child with condition A who had previously taken drug B during pregnancy ▫ All children either do or do not have condition A ▫ We assess whether they did/did not

Chapter 12 Epidemiology: Study Design take drug B during pregnancy Pros ▪ More easily examines rare diseases than prospective studies; less expensive and time-consuming ▪ Individuals not exposed to possible risk factors ▪ Past medical history used to determine potential multiple risk factors Cons ▪ Potential problems matching cases and controls ▫ E.g. study may be influenced by characteristics not being studied (confounding variables) ▪ Potentially biased (relies on individual recall) ▫ E.g. study candidates may emphasize potential risk factors rather than controls Figure 12.1 Case-control study design. COHORT STUDY osms.it/cohort-study ▪ Measures disease within group of individuals (cohort) over period of time ▪ Focuses on disease development ▪ Two types: prospective cohort, retrospective cohort PROSPECTIVE COHORT STUDY ▪ AKA longitudinal, concurrent cohort study ▪ Results not known until after intervention ▪ Used to follow up on people who received treatment/were exposed to risk factors ▪ Laboratory tests often used as surrogate markers – for example, increase in hemoglobin immediately after blood transfusion assumed to mean that transfusion was effective ▪ Example: RSV rates of premature birth cohorts Pros ▪ Easier to conduct than randomized controlled studies ▪ Useful information on risk ▪ Matching decreases influence of confounding variables Cons ▪ Expensive, time-consuming ▪ Follow-up with people over time can be difficult; subjects may be lost RETROSPECTIVE COHORT (HISTORICAL COHORT, NONCONCURRENT PROSPECTIVE) STUDY ▪ Same prospective cohort study design but uses past data to determine future time frame; study and obtention of results faster ▪ Use pre-existing population to decrease study duration ▪ Can be conducted relatively quickly, inexpensively ▫ E.g.mortality rates according to duration of smoking OSMOSIS.ORG 75

Figure 12.2 Design of prospective and retrospective cohort studies with hypothetical time frames. Exposed = smokers, not exposed = non-smokers, disease = lung cancer. CROSS-SECTIONAL STUDY osms.it/cross-sectional_study ▪ Study that observes a group of people at one point in time ▪ Examines relationship between an exposure (variable), disease being investigated ▪ Example: the relationship between endometrial cancer, hormone replacement therapy (HRT) Pros ▪ Less time-consuming, expensive than longitudinal studies, as individual follow-up not necessary ▪ Good for establishing overall association between exposure and disease ▪ Can establish disease prevalence (number of individuals with particular disease in their lifetime) 76 OSMOSIS.ORG Cons ▪ Establishes disease prevalence but not incidence (percentage of individuals who may develop a particular disease within a year) ▪ Does not establish temporal relationship between exposure and disease ▪ Potentially biased if surveys used ▪ Retrospective studies: data quality may be compromised due to poor recall/“recall bias,” where people are more likely to recall certain events

Chapter 12 Epidemiology: Study Design Figure 12.3 Design of a cross-sectional (prevalence) study. Example: obesity is the exposure, and high cholesterol is the outcome. ECOLOGIC STUDY osms.it/ecologic-study ▪ Observes at least one variable ▫ Exposure/outcome ▪ Measured at group level ▪ At least one comparison group, disease occurrence compared between groups ▪ Often used to make large-scale comparisons ▪ Examples ▫ Rate of cancer occurrence in one population ▫ Average sunlight exposure at different geographical locations ▫ Comparing per capita dietary fat consumption, cardiovascular disease mortality ▫ Disease occurrence compared between groups OSMOSIS.ORG 77

RANDOMIZED CONTROL TRIAL (RCT) osms.it/randomized-control-trial ▪ Examines effectiveness of intervention (e.g. medications, treatment protocols) ▪ Three features: randomization, control, manipulation ▪ Considered gold standard of experimental research, identifying cause-and-effect relationships ▪ Study participants randomly assigned either experimental group or control group ▪ Example: Effects of drug A versus drug B on hypercholesterolemia in individuals with type 2 diabetes mellitus Figure 12.4 A summary flowchart of the different types of study designs. 78 OSMOSIS.ORG