Case-Control Studies
Case-control studies compare the exposure history of individuals with a disease (cases) to those without (controls), providing an efficient method for investigating rare diseases and measuring the odds ratio as a proxy for relative risk.
Key Facts
Case-control studies compare exposure in cases (disease present) with controls (disease absent); retrospective design They measure the odds ratio (OR): odds of exposure in cases / odds of exposure in controls OR approximates RR when disease prevalence is low (rare disease assumption) Efficient for rare diseases: can study diseases with very low incidence Strengths: quick, cheap, good for rare diseases, can study multiple exposures Weaknesses: recall bias (cases remember exposures differently), selection bias, cannot measure incidence or RR directly Matching: controls matched to cases on potential confounders (age, sex) to reduce confounding Nested case-control: cases and controls selected from within a defined cohort; reduces selection bias
Overview
Key Facts
Case-control studies start with the outcome and look backwards for exposure. They are retrospective, efficient, and particularly valuable for rare diseases. The key measure is the odds ratio.
Design
- Define cases: individuals with the disease/outcome of interest
- Select controls: individuals without the disease from the same source population
- Measure past exposure in both groups (questionnaires, records, biological samples)
- Compare exposure prevalence using odds ratio
Control Selection
- Controls should come from the same source population as cases
- Common sources: hospital controls, community controls, friend/neighbour controls
- Matching: individual matching (each case matched to control on key confounders) or frequency matching (groups matched on proportions)
Measure of Association
- Odds ratio (OR) = (a × d) / (b × c) from 2×2 table
- Where a = exposed cases, b = exposed controls, c = unexposed cases, d = unexposed controls
- OR >1 indicates positive association; OR <1 indicates protective association; OR = 1 indicates no association
Variants
- Nested case-control: cases arise from a defined cohort; controls sampled from same cohort at time of case occurrence
- Case-cohort: controls sampled from entire cohort at baseline regardless of disease status
- Case-crossover: each case serves as their own control (exposure at time of event vs earlier period)
Clinical Presentation
Classic Case-Control Studies
- Doll & Hill (1950): smoking and lung cancer — established link using hospital-based case-control design
- Thalidomide and phocomelia: case-control study by McBride (1961) identified association
- Oral contraceptives and VTE: multiple case-control studies established risk
- SIDS and sleeping position: case-control studies led to 'Back to Sleep' campaign
When to Use Case-Control Design
- Disease is rare
- Long latency period between exposure and disease
- Limited resources/time
- Multiple exposures to investigate for single outcome
- Outbreak investigation (rapid assessment of risk factors)
Differential Diagnosis
| Feature | Case-Control | Cohort | Cross-Sectional |
|---|---|---|---|
| Direction | Retrospective | Prospective/retrospective | Snapshot |
| Starting point | Disease status | Exposure status | Neither |
| Measure | Odds ratio | Relative risk | Prevalence ratio |
| Good for | Rare diseases | Common exposures | Prevalence estimation |
| Can measure incidence | No | Yes | No |
| Temporal sequence | Inferred | Established | Not established |
| Common bias | Recall bias | Attrition bias | Prevalence-incidence bias |
Diagnosis / Investigation
Critical Appraisal (CASP Case-Control Checklist)
- Did the study address a clearly focused question?
- Was an appropriate case-control design used?
- Were cases defined and selected appropriately?
- Were controls selected appropriately from same source population?
- Was exposure measured accurately (validated tools)?
- Were confounders identified and controlled (matching, adjustment)?
- What are the results (OR, CI)?
- How precise are the results?
- Do you believe the results?
- Can results be applied locally?
Key Biases in Case-Control Studies
- Recall bias: cases may recall exposures differently from controls (particularly for traumatic exposures)
- Selection bias: controls not representative of source population
- Interviewer bias: knowledge of case/control status influences questioning
- Survival bias (Neyman bias): prevalent cases may differ from incident cases
Management
Addressing Bias
- Recall bias: use objective exposure measures (records, biomarkers), blinding of exposure assessors
- Selection bias: use population-based controls, nested case-control design
- Confounding: match controls, stratify analysis, multivariable logistic regression
Odds Ratio Interpretation
- OR = 1: no association
- OR > 1: exposure associated with increased odds of disease
- OR < 1: exposure associated with decreased odds (protective)
- 95% CI not crossing 1.0: statistically significant
- OR approximates RR when disease is rare (<10% prevalence)
Prognosis
- Case-control studies have been instrumental in identifying risk factors for many diseases
- They are particularly valuable in outbreak investigation (rapid identification of exposure source)
- Well-conducted case-control studies can provide strong evidence of association
- Limitations in establishing causality (retrospective, recall bias, cannot measure incidence)
- Nested case-control designs within large cohorts combine efficiency with reduced bias
Other Relevant Information
2×2 Table for Case-Control Studies
| Cases | Controls | |
|---|---|---|
| Exposed | a | b |
| Unexposed | c | d |
OR = (a × d) / (b × c)
OR Calculation Example
| Lung Cancer (Cases) | No Cancer (Controls) | |
|---|---|---|
| Smokers | 90 | 60 |
| Non-smokers | 10 | 40 |
OR = (90 × 40) / (60 × 10) = 3600/600 = 6.0 Smokers have 6× the odds of lung cancer compared to non-smokers