HMB342 Lecture 5 Feb 8/11 Risk: Looking Backward (Ch. 6)
Revisiting Cohort Studies
• Good, direct method for estimating risk, but drawbacks:
1) Often few people develop the outcome but all must be followed.
2) Cohort studies are resource-intensive: time, effort, and money.
3) More resource-intensive (inefficient) for uncommon outcomes.
• More efficient cohorts: Retrospective cohort & case-cohort, or the case-control method.
Recall: Anatomy of a Cohort Study
• Cohort is assembled (none have outcome of interest, but all are at risk for the outcome).
• Risk factor is identified & cohort segregated: 1) With the risk factor; 2) without the risk factor.
• All members of the cohort are then observed over time to see which experience the outcome.
• Rates of outcome are then estimated for the exposed vs unexposed members of the cohort.
Anatomy of a Case-Control Study (retrospective studies)
• In the present, cases (all have outcome) and controls are identified (none have outcome).
• Control = individual as similar as possible to each case but have not developed the outcome.
• Look back in time to measure frequency of exposure to risk factor in each of case and controls.
• Rates of exposure to the outcome are then estimated for the case and control & compared to estimate the
risk (but not the RR) of developing the outcome from the exposure.
Advantages of Case-Control Studies
• Efficient: instead of many people who don’t get the disease (as with cohort studies), data is collected on
those with the outcome & compared to matched controls.
• Speed in generating results: no wait for the outcome to develop--it has already appeared.
Disadvantages of Case-Control Studies
• Bias: managing study bias is more challenging than with cohort studies.
• Estimate RR, but not a true RR: case-control studies produce only an estimate of RR (odds ratio) and no
information on other measures of effect (absolute risk, attributable risk, population risk).
Designing Case-Control Studies
• The validity of case-control studies hinges upon 3 aspects of the study design:
1) Selecting cases 2) Selecting controls 3) Measuring the exposure of interest.
Case-Control Study Design--Selecting Cases
• Explicit criteria to define case/control, else lead to miscategorization, erroneous/biased estimate
• Cases should represent all possible cases. This ensures that cases selected do not represent
extreme/atypical cases of outcome or exposure & is simpler to find matched controls.
• Cases should be new (incident), not prevalent ones. Prevalent cases are more likely to contribute bias due
to factors related to disease duration.
• Sample: Li et al, 2009 “BMI and risk, age of onset, and survival in pancreatic cancer.”
Designing case-control studies: the validity of case-control studies hinges upon 3 aspects of the study design, selecting cases 2) selecting controls 3) measuring the exposure of interest. Case-control study design--selecting cases: explicit criteria to define case/control, else lead to miscategorization, erroneous/biased estimate, cases should represent all possible cases. This ensures that cases selected do not represent extreme/atypical cases of outcome or exposure & is simpler to find matched controls: cases should be new (incident), not prevalent ones. Prevalent cases are more likely to contribute bias due to factors related to disease duration: sample: li et al, 2009 bmi and risk, age of onset, and survival in pancreatic cancer. 2 www. notesolution. com: explicit criteria should be used to define cases & controls. Case: confirmed pancreatic ductal adenocarcinoma (controls are the opposite but at risk): cases selected for inclusion should be representative of all possible cases.