Case control studies Flashcards
Where in the hierachy of evidence does a case-control study lie?
3rd
What is a case control study?
Recruit as cases or control (based on whether they have the disease or not) and then look back on their lives
= know outcome when they are recruited
How are people recruited for a case control study?
- Identify cases (people with outcome)
- Obtain information about exposure
- Identify controls (people without outcome)
- Obtain information about exposure
When is frequency of exposure measured?
Retrospectively
What does the way a case is defined depend on?
the outcome of interest e.g. pathological, radiological, microbiological, self report
How are controls defined?
Must be individuals who would have been cases if had developed outcome (represent population from which the cases came from)
What are the possible sources of controls?
- Other disease controls (have other disease not related to outcome) e.g. Hospital controls & cancer registry controls
- Community based controls = PREFERABLE (healthy people) e.g. Electoral register, random digit dialling, primary care controls, spouse or family controls
What are the advantages of using other disease controls?
Easy to locate subjects
Subjects more likely to participate in studies = cheaper
What are the disadvantages of using other disease controls?
Hospital populations may be more likely than the general population to have the exposure of interest
What are the advantages of using community-based controls?
Prevalence of exposure not likely to be due to other disease
What are the disadvantages of using community-based controls?
More difficult to locate and less likely to participate (more expensive)
Which ratio of controls to cases should be used?
- At least 1 control per case
- Up to 4 controls per case if few cases (to improve power of study)
- Negligible benefit of more than 4 controls per case (not worth the expense)
How is exposure measured in case control studies?
Retrospectively
Reported by subject or from records
(similar to historical cohort study)
What are the differences between case control studies and historical cohort studies?
Case control study:
Recruit cases and controls over short defined time period
Can not calculate incidence rates/risk ratios (as total number in population usually unknown)
Historical cohort:
All those with outcome are included even if died years before study if exposure data still availiable
Can calculate incidene rates/risl ratios
What is odds of outcome?
= odds ratio
The equivilent of risk ratio (which cannot be calculated from the date we have)
n. b. for rare outcome odds ~ risk
no. without outcome is approx. number in study
How do we calculate odds?
Odds = no. with outcome/ no. without outcome
What is an odds ratio?
Odds ratio = odds in exposed group/odds in unexposed group
n.b. the odds in each group are calculated seperately
Which two test statistics can be calculated for an odds ratio?
P value (strength of evidence against null hypothesis - i.e. no association between exposure and outcome… OR = 1)
CI intervals (precision of estimated OR)
Why do we use an odds ratio instead of a risk ration in case control studies?
In case controls studies:
- Cannot calculate risk of outcome or risk ratio
- Can calculate odds ratio (ORexposure= ORoutcome)
- Rare outcomes (<10%) odds ratio is approximately equal to risk ratio
How do you interpret an odds ratio <1?
Exposure reduces risk
How do you interpret an odds ratio >1 ?
Exposure increases risk
How do you interpret a RR > 1 when the outcome is not rare?
OR > RR
How do you interpret a RR < 1 when the outcome is not rare?
OR < RR
Would a case-control study require more or less subject than a cohort study?
Less = more powerful
e.g. case control based on 350 people nearly as powerful (precise) as cohort study based on > 20,000 people
What are the strengths of case-control studies (4)?
- Relatively quick (no need to wait for incident cases)
- Efficient for rare outcomes (large smaples sizes and long follow ups NOT required)
- Allows examination of >1 exposure for 1 outcome
- No loss to follow up bias
What are the weaknesses of case control studies (5)?
- Can not be sure exposure came before outcome = reverse causality (less likely when cases are a new diagnosis rather than prevalence of disease)
- Inefficient for rare exposures
- Can not calculate incidence rates
- Selection bias (cases and controls must be selected irrespective of their exposure)
- Measurement bias (recall = those with outcome more likely to think about exposure/ think they have when they haven’t & interviewer = may investigate differently for cases and controls)
What is the null value for an odds ratio?
1 = no association
In general what does an odds ratio assess?
The association between exposure and outcome (binary)
When is the odds ratio likely to be a good approximation of the risk ratio?
When the outcomes are rare
If you have an OR of 5.44 what does this show?
The odds (approximate risk) of those exposed getting the outcome are more than 5 times higher than in those who were not exposed
What are the possible explanations for the association (other than causality)?
- Chance (v. small p value = unlikely)
- Bias (selection -> must be selected irrespective of their exposure; measurement = recall & interiew; loss to follow up unlikely)
- Confounding (e.g. age, gender, social class)
- Reverse causality (may change after diagnosis … BUT… lifetime exposure i.e. lifetime drinking not more recent increase in drinking in the more recent years due to diagnosis)