Case-Control Recruitment for Rare-Variant Association

Why Recruitment Design Matters

Gene-based sequencing compares rare-variant patterns between people with a phenotype of interest and people who do not meet the case definition. The comparison is informative only when case and control status is defined consistently and the groups arise from a defensible shared source population. Poor recruitment can make a gene-level difference reflect how participants were selected rather than how genetic variation relates to the phenotype.

The recruitment plan should therefore settle three decisions before enrollment begins: what qualifies a participant as a case or control, where each group will be recruited, and how many controls will be sought per case. The target population, as defined in the earlier material on target populations for gene-based sequencing studies, anchors these decisions. Recruitment criteria should operationalize that population rather than silently replace it with whoever is easiest to enroll.

Make Phenotype Eligibility Reproducible

A diagnosis label alone may not be enough to define cases. Before recruitment, specify the clinical features, diagnostic criteria, data sources, and time window that establish case status. For example, a plan might require a confirmed diagnosis under stated clinical criteria, documented in a medical record by a defined cutoff date. If a phenotype can arise through different presentations, specify whether all presentations are included and how uncertain or borderline diagnoses are handled.

Control eligibility also requires an explicit rule. Controls should not meet the case definition at the assessment time, and their phenotype status should be assessed with a method appropriate to the condition. A record search may be adequate for a well-documented diagnosis; a condition that is commonly unrecognized may require a clinical assessment or additional records. “No diagnosis listed” is not automatically equivalent to “does not have the phenotype.”

Eligibility criteria should also define the participant characteristics that are shared across groups: for example, age range, residence, relevant clinical setting, and study period. Apply these rules to both groups unless a difference is necessary to define the phenotype. If a condition usually develops later in life, a young control may not yet have reached an age at which absence of the condition is meaningful. The protocol should address that issue with a suitable age criterion or assessment plan rather than treating every currently unaffected person as an equivalent control.

Write the criteria so that a recruiter can apply them without knowing a participant's sequencing results. Eligibility based on genotype can distort the comparison: selecting controls because they lack variants in a candidate gene would remove precisely the genetic variation the study is intended to compare. Phenotype eligibility may use genetic information when it is part of a prespecified clinical diagnosis, but the rule should be clear and applied consistently.

Choose Recruitment Sources That Support Comparability

The central source-population question is: could a person recruited as a control plausibly have been recruited as a case if that person met the case definition? Controls should come from the population, geography, calendar period, and eligibility framework that gave rise to the cases. This does not require identical recruitment channels, but differences between channels need to be understood and should not make one group a fundamentally different population.

Potential case sources include disease registries, specialty clinics, hospital systems, or population-based screening. Potential control sources include the same healthcare catchment, an existing cohort, or community recruitment within the same region. Each source has limitations. Clinic cases may have more severe or unusual disease than cases in the wider population. Controls recruited from a volunteer cohort may differ from cases in health-care access or willingness to participate. These differences are reasons to define the intended population carefully and assess whether both sources represent it.

A control group recruited from patients with another illness is not automatically suitable. Such controls may be useful when their source population is appropriate, but the other illness or its treatment could be related to the target phenotype or to factors that influence participation. The study should justify that choice instead of assuming that “not the target disease” is sufficient.

For each recruitment route, document where participants are found, how they are invited, and how eligibility is checked. Keep case and control procedures as similar as practical for consent, assessment, and data collection. This makes the contrast easier to interpret and helps reveal whether one group was selected through a distinct process. Recruitment plans should also provide a clear rule for repeat invitations, withdrawals, and replacements, so that enrollment does not drift toward ad hoc selection.

State the Sampling Ratio and Enrollment Targets

A recruitment ratio is an operational plan, not a description of how common the phenotype is in the population. For example, a 1:2 case-to-control ratio means that the plan seeks two controls for each enrolled case. It does not mean that two-thirds of the source population are controls, nor does the proportion of cases in the recruited study estimate population prevalence.

Convert the ratio into integer enrollment targets and state whether it applies overall or within recruitment sites or groups. If the target is \(1:r\), where \(r\) is the number of controls per case, then for \(C\) cases the planned number of controls is \(rC\). The total planned enrollment is \(C+rC=(1+r)C\). This arithmetic makes the target transparent; it does not establish that the ratio is statistically optimal.

In practice, specify what happens if the planned ratio cannot be reached. For example, the protocol might recruit eligible cases consecutively and recruit up to a fixed number of eligible controls for each case. If case and control recruitment occurs at multiple sites, give site-specific targets or rules for pooling participants. A ratio stated only as an overall aspiration can conceal substantial imbalances between sites or recruitment periods.

A Recruitment Plan in Practice

A useful plan connects phenotype rules, recruitment frames, and enrollment targets in a sequence that can be implemented and audited. The steps below make the key decisions explicit.

Common Mistakes and a Short Justification

A common mistake is to treat “healthy volunteers” as a complete control definition. That phrase says little about whether controls came from the source population that produced the cases, whether they received an adequate phenotype assessment, or whether they met the same age and residence requirements. The correction is to define eligibility and recruitment source explicitly, then explain how a control could have become a case under the study's rules.

The logic can be stated as a simple eligibility check. Let \(E\) denote meeting the shared non-phenotype eligibility rules, and let \(D\) denote meeting the case phenotype definition. A case is eligible when \(E\) and \(D\) both hold; a control is eligible when \(E\) holds and \(D\) does not hold at the stated assessment time. The shared \(E\) rules make the intended contrast about phenotype status within a common eligible population, rather than about unrelated eligibility differences.

This notation is a design aid, not a guarantee that recruitment is unbiased. The same written eligibility criteria can still produce incomparable groups if one group is recruited through a narrow referral pathway and the other through a broad volunteer source. The source and invitation processes must also be examined.

• Using an easy-to-recruit control group without a source-population rationale. Convenience can improve enrollment speed but does not establish comparability.
• Defining controls only as “not known to have the disease.” This can misclassify people whose phenotype has not been adequately assessed.
• Changing criteria during recruitment without documentation. Different rules across sites or time periods can make group status inconsistent.
• Interpreting the sample case fraction as population prevalence. A case-control recruitment ratio is chosen by design and is not a population frequency estimate.