Sample Size Is Also a Feasibility Decision

A sample-size calculation can be mathematically correct and operationally impossible. A protocol may require 480 analyzable participants, while the institution sees only 300 potentially eligible people each year, expects half to consent, loses one fifth to follow-up, and has funding for twelve months. Approving that study without reconciling the numbers does not protect rigor. It postpones a predictable failure.

Feasibility is not permission to shrink the sample until the project fits the budget. It is a governance process for aligning the scientific objective, participant population, recruitment system, measurement burden, resources, and timeline before the study begins.

01Start with the scientific requirement

The statistical team should state why the target is needed: the effect the study is powered to detect or the precision required for an estimate. Every input should be visible, including event rate, variability, allocation, design effect, expected loss, and analysis method. The DELTA² guidance recommends choosing a target difference that is important, realistic, and informed by evidence and stakeholders—not simply the difference that yields a convenient sample.

Only after that requirement is clear should leaders test whether the organization can deliver it. Otherwise, feasibility discussions can quietly redefine the scientific question without acknowledging the consequence.

02Build a recruitment denominator

“We see many patients” is not a recruitment estimate. Construct a funnel using auditable data:

1. people seen in the relevant setting;
2. people meeting broad clinical criteria;
3. people meeting all eligibility criteria;
4. people reachable during the recruitment window;
5. people likely to be approached;
6. people likely to consent;
7. people expected to provide the primary outcome; and
8. people retained in the final analysis.

Use local historical records where lawful and appropriate, supplemented by prospective screening or a pilot when necessary. Avoid double-counting repeat visits, assuming referral populations are unique, or applying consent rates from a different burden and context.

03Test capacity across the whole workflow

Recruitment is only one constraint. Can the team perform the intervention or measurement at the required frequency? Are laboratory, imaging, pharmacy, data-management, and statistical resources available? Can multiple sites use consistent definitions and quality controls? Is the follow-up period compatible with the funding and staff contracts? Will participant burden increase dropout or missing data?

ICH E8(R1) frames quality as a design responsibility and encourages prospective attention to factors critical to reliable conclusions. An overcomplicated study that cannot be executed consistently is not made rigorous by an ambitious protocol.

04Use scenarios, not one optimistic projection

Build expected, conservative, and stress scenarios. Vary eligibility yield, consent, enrollment per site, retention, event rate, site-opening delay, and data completeness. The purpose is not to predict the future perfectly. It is to find which assumptions make the study fail and whether those assumptions can be monitored.

Predefine feasibility checkpoints: participants screened and enrolled by month, retention at the primary time point, completeness of critical variables, and site performance. Also define the response to missing milestones—support, redesign, additional sites, extension, or an independent decision about stopping. Avoid unplanned sample changes driven by emerging treatment results.

05Choose among defensible responses

If the target is infeasible, leaders have several options: narrow the population only if the scientific question permits; extend recruitment; add qualified sites; simplify nonessential measurements; improve follow-up; select a different justified primary outcome; redesign the study; or conduct a smaller feasibility study with appropriately limited aims.

The unacceptable option is to retain the language of a definitive study while recruiting a convenience sample that cannot answer the stated question. A smaller study may be valuable, but its objective and claims must match what it can estimate.

06Record the joint decision

The approval package should include the statistical calculation, recruitment funnel, capacity assessment, scenario analysis, assumptions and data sources, feasibility milestones, and named decision owners. Scientific, clinical, operational, data, financial, and ethical perspectives should be represented.

Before approving the next protocol, ask two separate questions: “Is this sample sufficient for the scientific objective?” and “Can this institution responsibly deliver that analyzable sample?” A defensible study requires both answers to be yes—or a transparent redesign before participants carry the cost of the mismatch.

Revisit the forecast at predefined milestones using aggregate operational data that do not compromise blinding. Feasibility monitoring should detect execution failure early without becoming a back door for result-driven changes to the scientific target.

This article is educational and intended for research purposes. It does not provide individual medical advice, diagnosis, or treatment. No patient data were used.