L5 - Case-Control Study Design

Question 1

Why Learn Case-Control Studies?
(EBP Framing)

Answer 1

1. Clinical decisions rely on research evidence
2. Many rehab questions involve risk factors
3. Case-control studies are commonly misused

Question 2

What type of study design are Case-Control Studies?

Answer 2

Retrospective: the outcome (disease) has already occurred.

*observational research design that focuses on events, data, or exposures that have already happened

Question 3

How misused are case-control studies?
What is the impact?

Answer 3

97% misclassified in rehab literature (Mayo & Goldberg, 2009)
–> Incorrect design leads to wrong statistics and flawed conclusions =risking patient care

Question 4

When are case-control studies useful? (3)

Answer 4

1. Outcome is rare
2. Exposure occurred in the past
3. Prospective follow-up is impractical or unethical

Question 5

Can we calculate incidence or risk of developing the outcome with case-control studies?

Answer 5

Case-control studies oversample the desired outcome (disease) on purpose

We can’t calculate incidence or risk of developing the outcome –> cohort studies! (follow ppl)

Question 6

Name examples of cases where case-control studies are useful. (4)

Answer 6

1. Smoking → Lung cancer
2. Obesity → Diabetes
3. Rheumatoid arthritis (RA) → Fracture
4. Knee injury → Knee osteoarthritis (OA)

Question 7

Are case-control studies used for common outcomes?

Answer 7

Case-control studies not used for common outcome.
–> instead: cross-sectional studies (sample that has), cohort design (at risk)

Question 8

Flow Chart
What type of study design are case-control studies?

Answer 8

Analytical –> Observational –> Case-Control Studies

Question 9

Flow Chart
What is the difference between cross-sectional and case-control studies?

Answer 9

Different questions:
- Cross-sectional: snapshot of what is happening now
- Case-control: what might have caused what we see now (outcome)

Question 10

Flow Chart
Does time direction matter in case-control studies?

Answer 10

Time direction matters: outcome comes first and exposure is second
–> retrospective

Question 11

What are the 4 majors characteristics of case-control studies?
- type of design
- participation
- focus

Answer 11

1. Observational study design
2. Retrospective: looks back in time to assess exposure
   –> Exposure must have occurred before the outcome
3. Participants selected based on outcome:
4. Focus: Etiology (what caused the outcome)

See NDC p.7 for illustration of retrospective design

Question 12

How are participants selected based on outcome?

Answer 12

Cases = with outcome
Controls = without outcome

Question 13

What are the 4 key features of case-control studies?
- time
- type of sampling
- selection (smt important)
- measure of association

Answer 13

1. Retrospective exposure assessment
2. Outcome–based sampling
3. Careful control selection required – matching to minimize confounding –> same age, sex…
4. Measure of association: Odds Ratio (OR)

Question 14

Bias in Case-Control Studies
What are the types of bias?

Answer 14

1. Selection Bias (sampling)
2. Recall Bias
3. Confounding Bias

Question 15

Bias in Case-Control Studies
In what cases can there be selection bias? (2)

Answer 15

1. Cases and controls not from the same population
2. Controls do not represent the population that produced the cases

Question 16

Bias in Case-Control Studies
In what cases can there be recall bias? (3)

Answer 16

1. Ascertainment of past exposures can be difficult
2. Cases remember exposure differently & may overreport prior exposure
3. Controls underreport exposure or forget exposure

Question 17

Bias in Case-Control Studies
What can cause confounding bias?

Answer 17

Third variable distorting association b/w exposure and outcome.
–> ex: sex division is uneven between groups

Question 18

Bias in Case-Control Studies
Name a design strategy to reduce selection bias.

Answer 18

1. Selection: Select controls from the same source of population as cases

Question 19

Bias in Case-Control Studies
Name some design strategies to reduce recall bias. (2)

Answer 19

1. Avoid emphasizing the exposure of interest
2. Use same standardized assessment/questionnaires for cases and controls
   –> We don’t want to draw attention to the outcome of interest, to not influence their answers
   BREF: Mask exposure of interest with other questions

Question 20

Bias in Case-Control Studies
Name a design strategy to reduce confounding bias.

Answer 20

Match or statistically adjust for key confounders (e.g., age , sex).

Question 21

Selecting Controls
Where do the case and control participants come from?

Answer 21

Case an control are determined by the outcome
- case = people with the outcome
- control = people without the outcome

Cases and controls must come from the same source of population.
–> Same place where the cases arose: hospital, clinic

Question 22

Selecting Controls
What are the characteristics of good controls? (3)

Answer 22

1. Represent the population that produced the cases
2. Have had the same opportunity for exposure
3. Be proven not to have the outcome

–> Controls are NOT “healthy volunteers by default”, good controls could develop the outcome

Question 23

Bias in Case-Control Studies
Describe matching.
- what
- purpose
- caution

Answer 23

1. Making cases and controls similar on: age, sex, SES
   –> ex: level of education, PA level, place of residence, time of hospitalization, comparable access to healthcare
2. Purpose: reduce confounding
3. Avoid overmatching (hides real effect)
   –> don’t match for variables that are caused by the exposure

Question 24

Odds Ratio
What is OR?
Does it give use info about the risk of developing?

Answer 24

OR compares odds of exposure in cases vs. controls
–> OR (odds) ≠ Relative Risk (probabilities)

Don’t use “risk” –> used in cohort studies!

Question 25

Odds Ratio Name the significance of OR value. (3)

Answer 25

- OR = 1 → no association b/w exposure and outcome - OR > 1 → Exposure increases the odds - OR < 1 → Exposure decreases the odds BREF: 1 is the “no effect” value for an odds ratio

Question 26

Odds Ratio What is the formula for OR? **NEED TO KNOW

Answer 26

OR = odds of exposure in cases / odds of exposure in controls --> odds of exposure in cases = cases that were exposed/cases that weren't exposed See NDC p.14-15 for table + OR calculation

Question 27

Confidence Intervals (CI) What does it mean if the CI includes 1?

Answer 27

CI includes 1 = not statistically significant

Question 28

Confidence Intervals (CI) What does it mean if the CI is wide? (3)

Answer 28

1. Low precision 2. More uncertainty 3. Estimate of OR is less precise See NDC p.17 for illustration

Question 29

Confidence Intervals (CI) What does it mean if the CI is narrow? (3)

Answer 29

1. High precision 2. More confidence 3. Estimate of OR is more precise See NDC p.17 for illustration

Question 30

Odds Ratio (OR) and Confidence Intervals (CI) Analyze this example. - OR = 1.71 - 95% CI = 0.89-3.29

Answer 30

OR > 1 : risk of outcome is exposed > unexposed BUT CI includes 1 = not statistically significant

Question 31

Logistic Regression When is logistic regression used?

Answer 31

Used when the outcome is binary (case vs control). Lets us look at one main exposure while taking into account other important factors!

Question 32

Logistic Regression What does logistic regression look at? (4)

Answer 32

1. Estimates the odds of being a case 2. Adjust OR to account for confounders 3. Model multiple predictors = Looks at several risk factors at the same time 4. Refine associations

Question 33

Strengths and Limitations What are the strengths of case-control study design? (4)

Answer 33

1. Often used when the outcome is rare (rare disease), and exposure Hx is needed 2. Efficient for early-stage research questions (etiology-based) 3. Less time & cost than cohort studies 4. Ethical for harmful exposures --> ex: we are not exposing subjects to environmental hazard, fall hazard

Question 34

Strengths and Limitations What are the limitations of case-control study design? (3)

Answer 34

1. Cannot calculate incidence 2. Prone to bias (recall bias!!!) 3. Cannot directly establish causality, but raise hypotheses

Question 35

Strengths and Limitations How are case-control study designs ranked according to the Centre for Evidence-Based Medicine?

Answer 35

Level 3B: 3rd worst out of 10 --> moderate level of evidence Lower than: RCTs, cohort studies --> control bias better --> establish cause better See NDC p.20 for illustration

Question 36

Mislabeling What are common mislabelings in the literature? --> ie what study designs are sometimes classified as case-control?

Answer 36

1. Cross-sectional studies: snapshot on time 2. Prognostic cohort studies: current exposure 3. Intervention studies

Question 37

Mislabeling How to identify a study that ISN'T a case-control? (3) What is the right answer in case-control?

Answer 37

Not a case-control if: 1. Group selected by exposure --> case-control select sample based on outcome 2. Exposure and outcome measured at the same time: this is cross-sectional --> in case-control exposure precedes the outcome 3. Intervention assigned: this is experimental --> case-control are observational: no intervention, just exposure

Question 38

Mislabeling What are the negative consequences of mislabeling?

Answer 38

1. Invalid statistical analysis 2. Wrong conclusions 3. Biased interpretations 4. Difficulty replicating or applying findings

Question 39

What makes a good case-control study? (5)

Answer 39

1. Clear etiologic question (PECOT) 2. Clear case and control definition 3. Appropriate control selection --> Appropriate sampling frame and sampling strategy 4. Same exposure measurement 5. Bias acknowledged

Question 40

Name the key points of case-control study design. (6)

Answer 40

1. Outcome first, exposure second --> Sample based on outcome from an underlying population 2. Retrospective (looking back in time) 3. Exposure occurred before the outcome 4. OR + 95% CI as measure of association 5. Powerful but bias-prone 6. Evidence level: moderate

Question 41

See NDC p.26-28 for exercise!

Answer 41

🏞