How Many Controls Per Case Do You Need?

Is One Control Per Case Enough, or Should You Recruit More?

You have identified 80 eligible cases of a relatively rare disease. Recruiting another 80 participants as controls seems manageable, but you wonder whether the study would be stronger with 160, 240, or even 400 controls.

Some researchers recommend one control per case. Others suggest two, three, or four, particularly when cases are difficult to find. You may also encounter the claim that recruiting more than four controls per case is unnecessary.

These recommendations are useful starting points, but they are not universal statistical rules.

The number of controls affects precision, statistical power, recruitment costs, and sometimes the complexity of the analysis. The appropriate ratio also depends on whether cases and controls are individually matched, sampled from risk sets, or selected through an unmatched design.

The real question is how much additional information each control contributes, and whether that contribution justifies the resources required to obtain it.

Choose the Ratio Based on Statistical Efficiency and Feasibility

How the Control-to-Case Ratio Affects Study Efficiency

What Does a Control-to-Case Ratio Mean?

The control-to-case ratio describes how many controls are included relative to the number of cases.

A ratio of 1:1 means one control for each case. A ratio of 1:2 means two controls per case, and a ratio of 1:4 means four controls per case.

For example, if a study has 100 cases, a 1:1 ratio requires 100 controls, while a 1:4 ratio requires 400 controls.

The notation should be stated clearly because some authors describe ratios as cases to controls, while others use controls to cases. Reporting the actual numbers avoids ambiguity.

In an unmatched case-control study, the ratio usually refers to the total number of controls divided by the total number of cases. In an individually matched study, it may refer to the number of controls selected for each case or matched set.

These distinctions matter because matching can change the amount of information available for estimating the association.

Why Does Adding Controls Improve Statistical Precision?

Cases and controls contribute information about the relationship between exposure and outcome. Increasing the number of controls generally improves estimation of the exposure distribution in the source population.

However, once the control group is substantially larger than the case group, uncertainty associated with the cases becomes an increasingly important limitation.

For a simple unmatched case-control study with a fixed number of cases, a useful approximation to the sampling-variance component is proportional to:

The reduction illustrates why additional controls can improve precision. Yet the benefit becomes progressively smaller as more controls are added while the number of cases remains fixed.

Actual odds-ratio precision also depends on the exposure distribution and the numbers in each cell of the case-control table. Consequently, the simplified expression should not replace a formal study-size calculation.

Why Is a 1:1 Ratio Often Considered Efficient?

When cases and controls have similar recruitment costs, measurement quality, and contribution to statistical information, a balanced allocation is often efficient for a fixed total sample size.

For example, if researchers can recruit 200 participants in total, allocating approximately 100 cases and 100 controls will often provide better precision than allocating 50 cases and 150 controls, assuming comparable exposure distributions and other conventional conditions.

This does not mean that 1:1 is always optimal.

In many case-control investigations, the number of available cases is limited by disease occurrence. Researchers cannot simply recruit more cases to achieve balance. Additional controls may then offer a practical way to improve precision without waiting for further cases.

Why Are Ratios of 1:2, 1:3, and 1:4 Common?

When cases are scarce but controls are comparatively accessible, recruiting more controls can increase statistical information.

Under a simplified equal-information approximation, the relative efficiency of a design with r controls per case, compared with an infinitely large control group and the same number of cases, is:

These percentages describe efficiency relative to an idealized design with unlimited controls and the same fixed number of cases. They are not percentages of statistical power, nor do they indicate the probability of detecting a true association.

The pattern explains the familiar recommendation to consider up to four controls per case. The largest gains occur when moving from one control to two or three, while additional controls beyond four generally produce smaller marginal improvements.

Nevertheless, ratios above four can be justified when controls are inexpensive, cases are exceptionally scarce, or the study's precision requirements warrant additional sampling.

Does Adding Controls Increase Statistical Power?

Usually, yes, when additional controls contribute relevant independent information and other design conditions remain unchanged.

Statistical power is the probability that a specified statistical test will reject the null hypothesis under a particular alternative hypothesis, given the assumed effect size, significance level, and study design.

Increasing the number of controls can reduce standard errors and improve the ability to detect an association of a specified magnitude.

However, the improvement is not linear. Doubling the number of controls does not double statistical power.

Power also depends on the expected exposure prevalence among controls, the anticipated odds ratio, the number of cases, the significance level, and the statistical method.

A study with very few exposed cases may remain underpowered even after recruiting many additional controls.

How Does Exposure Prevalence Affect the Required Number of Controls?

The prevalence of the exposure among controls influences the number of exposed and unexposed observations available for comparison.

If an exposure is uncommon, a small control group may contain very few exposed individuals. This can produce unstable estimates and wide confidence intervals.

For example, if the expected exposure prevalence among controls is 5%, a hypothetical sample of 100 controls would contain approximately five exposed controls on average. Increasing the control sample to 400 would yield approximately 20 exposed controls on average under the same prevalence assumption.

These are expected counts, not guaranteed observations.

The improvement may be valuable, but researchers must also consider how many cases are expected to have the exposure. If exposed cases remain extremely scarce, additional controls cannot fully compensate for that limitation.

Study-size calculations should therefore incorporate plausible exposure frequencies and the minimum association magnitude that would be scientifically meaningful.

How Does Matching Change the Optimal Ratio?

In a matched case-control study, controls are selected according to predefined matching characteristics. The information contributed by additional controls depends partly on how much exposure variation exists within the matched sets.

For example, suppose each case is matched with three controls on age and recruitment site. If all members of most matched sets have the same exposure status, those sets may contribute little information to estimating the exposure association in a conditional logistic regression.

Adding more controls can increase the chance of observing informative exposure differences within matched sets, but the gain depends on the matching variables, exposure distribution, and correlation within sets.

Consequently, the simple unmatched efficiency formula should not be applied mechanically to matched studies.

Researchers should use sample-size methods appropriate to the intended matching strategy and analysis.

Can You Use Different Numbers of Controls for Different Cases?

Yes. Variable matching ratios are possible, particularly when eligible controls are difficult to identify for certain cases or risk sets.

For example, one case might have four eligible matched controls, while another has only two.

Such differences do not automatically invalidate the study. However, they can affect precision and should be accommodated by the analytical method.

In an unmatched study, researchers may also recruit controls without maintaining a fixed one-to-one relationship with individual cases.

The important distinction is whether the control-selection process remains consistent with the source population and prespecified sampling scheme.

Researchers should not selectively add controls to particular cases merely because doing so produces a more favorable association estimate.

When Can More Than Four Controls Per Case Be Worthwhile?

Ratios greater than 1:4 may be reasonable when the marginal cost of additional controls is low and the number of cases cannot be increased.

For example, a large electronic health record database may contain thousands of eligible controls whose relevant variables are already recorded. Sampling additional controls may involve little incremental data-collection cost.

In that situation, using five, ten, or more controls per case could be justified after considering the expected precision gain and computational or analytical requirements.

Conversely, if each additional control requires an expensive interview, biological assay, or lengthy chart review, the marginal gain may not justify the cost.

The appropriate decision is an efficiency calculation, not adherence to a fixed numerical ceiling.

How Do Recruitment Costs Affect the Optimal Ratio?

When cases and controls have different recruitment or measurement costs, the economically efficient allocation may be unequal.

Under a simplified variance model in which the variance is proportional to the sum of reciprocal group sizes, and the per-participant costs are constant, the optimal control-to-case ratio for a fixed budget is:

This expression is most useful for understanding why unequal costs can favor unequal allocation. In real studies, case availability, exposure frequencies, matching, missing data, and other design features can change the optimum.

If the number of cases is already fixed, the problem is different: researchers must determine whether the expected improvement from each additional control justifies its marginal cost.

Does the Ratio Matter More Than the Total Sample Size?

No. A well-chosen ratio cannot compensate for an inadequate number of cases or insufficient information about the exposure.

A study with 20 cases and 80 controls has a 1:4 ratio, but it may be substantially less precise than a study with 200 cases and 200 controls.

The ratio describes allocation, not overall adequacy.

Researchers should first define the intended association measure and meaningful effect size, then determine the sample size required for acceptable precision or statistical power.

The control-to-case ratio is one component of that calculation.

Why Must Control Quality Come Before Control Quantity?

Increasing the number of controls improves precision only when the controls provide valid information about the intended source population.

If controls are selected from a population with systematically different exposure opportunities, adding more of them may produce a more precise but biased estimate.

Before considering additional controls, establish that the control-selection strategy represents the appropriate source population.

Controls must also satisfy the relevant eligibility conditions. In incidence-density sampling, for example, they should represent individuals at risk at the case event times rather than a convenience sample of people who remain outcome-free at the end of follow-up.

The principle that controls must be eligible to contribute cases is more fundamental than the number recruited.

Choosing Between One, Two, Four, and Eight Controls Per Case

The simplified efficiency model shows that moving from one to two controls per case provides a substantial gain. Moving from four to eight controls adds considerably more recruitment work but increases relative efficiency by only about 8.9 percentage points.

Assume that recruiting and measuring each control costs 1,000 monetary units. Increasing the ratio from 1:4 to 1:8 requires 300 additional controls, costing another 300,000 units.

The team must decide whether that expenditure is justified by the expected reduction in uncertainty.

The hypothetical decision illustrates a defensible approach, not a universal recommendation to use four controls per case.

If controls were available from an inexpensive, reliable database, the balance could favor a larger ratio. If the exposure were extremely uncommon or the analysis required matching, the calculations might also produce a different result.

The ratio should follow the study's information requirements and practical constraints rather than precede them.

Common Misconceptions About Control-to-Case Ratios

How Should You Choose a Defensible Control-to-Case Ratio?

Start with the number of cases you can realistically identify and the association you want to estimate. Then evaluate the additional information gained from alternative numbers of controls.

For many conventional unmatched studies, comparing ratios of 1:1, 1:2, 1:3, and 1:4 provides a useful starting point. Larger ratios should remain available when justified by precision requirements and recruitment economics.

The final protocol should report the anticipated number of cases, the proposed number of controls, the ratio, the assumptions underlying the sample-size calculation, and the rationale for the allocation.

When case availability is uncertain, researchers may also plan sensitivity calculations for different attainable case counts.

A defensible ratio is one that meets the study's analytical needs with an appropriate use of resources, not one selected merely because it is customary.

Before Finalizing the Number of Controls, Check These Factors

Frequently Asked Questions About Controls Per Case

Use Enough Controls to Achieve the Required Precision

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.