RESEARCH GUIDE
Case-Control Study Design Guide
Control Selection, Matching and Odds Ratios
A case-control study design starts from people who have an outcome and compares their past exposures with those of a sample of people who do not. Its validity turns on whether controls represent the population that produced the cases, so selection and matching come first.
A case-control study samples by outcome: cases are identified, controls are drawn to represent the population that produced the cases, and exposure is then compared between the two groups. It is efficient for rare outcomes and for several exposures at once, but it cannot estimate incidence directly, and its validity depends on controls reflecting the exposure distribution in the source population. The odds ratio is the measure the sampling supports, and under suitable sampling it estimates a rate ratio. OnlineSPSS, a specialist statistical consultancy for academic, clinical and professional research, supports case-control studies from control sampling and matching to the matched analysis and STROBE reporting, while the research question and its conclusions remain the researcher’s.
TYPES OF CASE-CONTROL STUDY
Types of Case-Control Study and How Cases Are Paired
The variants differ in where controls come from and how cases and controls are paired, and each choice changes which biases are most likely and which analysis fits.
Population-Based Study Cases come from a registry or defined area and controls from the same population, which gives a clear source population but usually costs more and has lower participation.
Hospital-Based Study Cases and controls are drawn from a hospital or clinic, which is convenient, though patients admitted for other reasons may differ from the source population in their exposure.
Nested Within a Cohort Cases and controls are sampled from within an established cohort, so exposure data are often recorded before diagnosis and the source population is fully defined.
Case-Cohort Study A random subcohort serves as the comparison for all cases, which lets one comparison group be reused for several outcomes, and the data are usually analyzed with weighted survival models.
Matched Case-Control Study In case-control matching, each case is paired with one or more controls who share characteristics such as age and sex, which improves efficiency and calls for a matched analysis.
Case-Crossover Study Each case acts as its own control by comparing exposure just before the event with exposure at other times, which suits brief exposures and acute outcomes.
DESIGN COMPARISON
Three Ways to Obtain Controls
The source of controls shapes validity more than any other design choice, and the table compares population controls, hospital controls and controls sampled within a cohort.
| Feature | Population controls | Hospital controls | Nested, density-sampled controls |
|---|---|---|---|
| Where controls come from | The community that gave rise to the cases | Patients admitted for other conditions | Cohort members still at risk when a case occurs |
| Source population | Defined by the study base | Often unclear, as admission may relate to exposure | Fully defined by the cohort |
| Exposure data | Interview or records, recalled after the event | Interview or records, recalled after the event | Often recorded before diagnosis |
| Main threat | Low participation and recall differences | Selection, when the control conditions relate to exposure | Confounding and attrition of the parent cohort |
| Estimate | Odds ratio | Odds ratio | Odds ratio estimating a rate ratio |
KEY PLANNING DECISIONS
Setting Up a Valid Comparison of Cases and Controls
Control selection in case-control studies cannot be fixed after the data are collected, so the case definition, the source population, the controls and the matching are decided and written down at the start.
Defining the Case Specify the diagnostic criteria and whether only new cases are included, because including long-standing cases favors exposures linked to survival rather than to onset.
Defining the Source Population Name the population that produced the cases, since controls are valid only if they represent the exposure distribution among people who could have become cases.
Sampling the Controls Select controls with the same opportunity for exposure as cases, and where possible sample them at the time each case occurs so the odds ratio estimates a rate ratio.
Matching Choices Match only on factors that are related to the outcome and not caused by the exposure, and remember that a matched variable cannot then be studied as a risk factor.
Sample Size for the Odds Ratio Base the numbers of cases and controls on the smallest odds ratio worth detecting and the expected exposure prevalence among controls, and for matched sets on the correlation of exposure within them.
Measuring Exposure Alike Use the same sources and procedures for cases and controls, with interviewers unaware of case status where possible, so that recall and recording do not differ by group.
COMMON ISSUES
Selection, Recall and Matching Errors in Case-Control Work
Errors in sampling and in matching tend to be silent in the output, so each problem below is looked for in how participants were identified before any odds ratio is read.
Control Selection Bias Controls who do not represent the source population, for example because of who agrees to take part, give an exposure distribution that makes the odds ratio too large or too small.
Recall Bias People with a disease often search their memory for causes more thoroughly than healthy controls, which inflates reported exposure among cases unless records or blinded measures are used.
Overmatching Matching on a variable linked to the exposure but not the outcome, or on an intermediate step, reduces efficiency and can bias the association toward the null.
Matching Ignored in the Analysis Individually matched sets are analyzed with methods that use the pairing, such as conditional logistic regression, since an unmatched analysis can give biased estimates.
Odds Ratio Misread Whether the odds ratio approximates a risk ratio depends on how controls were sampled and, for controls drawn from non-cases, on the outcome being rare, and absolute risks need outside data on incidence.
Reverse Causation The disease or its early symptoms may change the exposure, as when illness alters diet before diagnosis, so the timing of exposure relative to onset is examined.
REPORTING
Reporting Standards for Case-Control Studies
A case-control report is read first for how cases and controls were chosen, so the checklist items on selection and matching deserve the most care.
Case-control studies are reported with the STROBE checklist for case-control designs, which asks for the eligibility criteria, the sources and methods of case and control selection, the rationale for the choice of controls, the matching criteria and the number of controls per case. The RECORD extension applies when the data come from routinely collected records. The flow of cases and controls, including how many were approached and how many took part in each group, is shown so that selection can be judged. Reviewers tend to read the selection items first, so the rationale for the controls is argued in the methods and not left to the checklist alone.
STATISTICAL METHODS
Methods Used to Analyze Case-Control Data
Case-control sampling fixes the effect measure as an odds ratio, and the model then depends on matching, sparse cells and how many exposures are adjusted for.
Relative Risk & Odds Ratios
Odds ratios with confidence intervals
Logistic Regression
Adjusted odds ratios across exposures
Firth's Logistic Regression
Sparse data and rare exposures
Chi-Square Test
Crude association in a two-by-two table
Missing Data Analysis
Incomplete exposure and covariate records
Clinical, Survival & Epidemiological
Odds, risk and other epidemiological measures
ANALYSIS SERVICES
Statistical Support for a Case-Control Project
Control sampling and matching are settled at the start and carried into the model at the end, and these four services follow the project through planning, analysis, reporting and review.
Research Planning & Study Design
The case definition, the source of controls, the matching scheme and the sample size, written down before participants are identified or records are extracted.
Quantitative Data Analysis
The conditional and unconditional models for your case-control data, with the odds ratios, confidence intervals and diagnostics that the sampling supports.
Results Reporting & Interpretation
Tables of case and control selection, matched-set summaries and methods wording prepared to STROBE, with each odds ratio read against the controls actually used.
Independent Statistical Review
An analysis already run on case-control data is checked for control selection, whether the matching was kept in the model and how each odds ratio is worded.
FREQUENTLY ASKED QUESTIONS
Questions About Case-Control Study Design
Controls, matching and recall are where case-control questions concentrate, and the final two answers explain data protection and what an assessment involves.
How can recall bias be reduced in a case-control study?
Recall bias is reduced by measuring exposure in the same way for cases and controls, preferably from records made before the outcome occurred, and by keeping interviewers and, where possible, participants unaware of the study hypothesis. Structured questions with fixed reference periods support this, and controls with another illness can equalize the motivation to remember, at the risk of selection bias if that illness is related to the exposure. Remaining concern is examined in a sensitivity analysis.
Who should be selected as controls?
Controls should come from the same source population as the cases, so that they show how common exposure is among people who could have become cases. Practical choices include population registers, neighbors, hospital patients with unrelated conditions or members of a cohort. Each has its own risk of bias, and the rationale for the choice is stated in the methods.
Should I match cases and controls?
Matching is a choice and not a requirement. It can improve efficiency for strong confounders such as age, but it cannot control confounding by itself, it prevents study of the matched variable and it commits you to a matched analysis. Adjusting for the same variables in a regression model is often an effective alternative for an unmatched design.
How many controls should be selected for each case?
One control per case is efficient when controls are as easy to find as cases. When cases are few, adding controls improves precision, with gains that become small beyond about four controls per case, so the cost of recruitment is weighed against power. In Epidemiology, registry or cohort data can make extra controls inexpensive, and the ratio chosen is stated and justified in the methods.
How do you protect my research data and confidentiality?
Case-control data usually contain identifiable clinical and personal information, and your data and project details are handled confidentially. Project information is not published, and researcher names and institutions are not used publicly without permission. An NDA is available where required and can be arranged before sensitive materials are shared. Read more about confidentiality and data security.
What happens when I request an assessment, and how long does it take?
You share your research question, design and relevant materials, such as a protocol, an analysis plan or a dataset, and a statistically trained Research Coordinator reviews the request and responds the same day. The analysis, deliverables, timeline and one fixed project fee are then confirmed before any work begins, so the timeline follows from the agreed scope.
Still have a question? Find more answers in our full FAQ.
NEXT STEP
Review Your
Controls and Matching Plan
Share how cases and controls were or will be identified, the matching used and the data or protocol. We confirm an analysis that respects the sampling and agree the scope first.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014