Lecture 5 (Case-Control Studies) Flashcards
Case control study
-Group of ppl w/disease + comparison grp w/o disease compared w/respect to history of past exposures
Diff between case control and cohort study
- Case control: starts w/ppl with and w/o disease, compares history of past exposures
- Cohort: starts w/ppl with and w/o exposure, follows to compare future disease
Incidence or prevalence for case control study?
- Incidence preferred
- Prevalence can’t distinguish between risk factors for development vs survival (smtms used so we don’t have to wait for incidence though)
Source pop
-Pop that gives rise to cases
Case def
-Definite medical criteria to ID cases
Case ID
-System for finding all cases who meet criteria and are part of source pop
Source pop types
- Pop-based
- Hospital-based
- Nested
Pop-based: source pop
-All residents of defined geo area w/o disease X
Pop-based: cases
-All new cases of disease X in source pop, over specified period of time
Pop-cased: controls and their selection
- Sample (ideally random) of source pop over same period of time
- Selection: random sample f/pop registry, neighbourhood controls, random digit dialling
Hosp-based: source pop
- All ppl w/o Disease X who would attend Hosp A if they had it
- Impossible to ID in practice, random sample in general =/= source pop b/c it doesn’t account for referral patterns
Hosp-based: cases
-All new cases ID’d in Hosp A over specified period
Hop-based: controls
-Sample of patients in Hosp A w/diagnoses other than Disease X over same period (shared risk factors means exposures won’t be representative of general pop)
OR
-Non random sample of pop, most of whom are healthy (but distribution of exposures may not be = between grps)
Effective strategies for hosp-based controls
- Limit controls to those hospitalized for disease w/no shared risk factors
- Include variety of diseases in control grp to dilute biasing effects of unknown shared risk factors
- When excluding diseases, only consider diagnosis at current hospitalization
Nested: source pop
-Subjects in ongoing concurrent cohort study who did not have Disease X at baseline
Nested: cases
-All new cases in cohort over defined follow-up period
Nested: controls
-Random sample of cohort who didn’t develop Disease X over same time period
Pros of a nested case control study (vs cohort)
- Requires analysis of way less samples/etc. (cohort: all, this: only cases/controls)
- Preservations of specimens for other research
Case control: matching
-Selection of controls similar to cases w/respect to factors other than exposure
Frequency matching
-Selection such that distribution of matching factors is similar in case/control grps
Individual matching
-Each control individually matched to case w/respect to specific factors
Implications of matching
- Association between matching factors + disease can’t be studied
- Overmatching: if matching factor is associated w/the exposure of interest –> similar factors between grp
- Must be taken into account throughout analysis
Case-control: RR
- Can’t accurately measure incidence b/c we picked #s in both grps, therefore no RR
- Can do OR instead
Odds Ratio
- Ratio of odds that cases were exposed vs odds that controls were exposed
- Good approximation of RR in most cases
- Same interpretation as RR
Analysis of matched pairs w/dichotomous exposure
++/–: concordant (uninformative)
+-/-+: disocrdant (can produce OR)
Pros of case control studies
- Way less time than cohort
- Requires much smaller sample size (if exposure is common, if it’s rare: cohort)
- Much less costly
- Can test hyps
- Useful for studying rare diseases
Cons of case control studies
- Past exposures ascertained after the onset of disease (uncertainty for causality, recall bias)
- Prone to selection bias
- Can estimate incidence using OR
- Observational study
Reliability
-Reproducibility/replicability of measurement
Validity
-Degree to which a measurement measures what it purports to
