Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one
Contents

Academy · Research Data, Analytics & AI · Trial Design & Sample Size Planning

Designing Controlled Clinical Trials: From Research Question to Sample Size Calculation

The ICH E9 guidance on statistical principles and a growing set of reporting checklists treat the sample size calculation as a specific, verifiable commitment made before a trial begins rather than an administrative afterthought, because an underpowered trial may fail to detect a real effect however well it is otherwise conducted. This article examines the design decisions that precede enrolment.

Pubrica AcademyWritten by a clinical trial design writer Reviewed bya biostatistics reviewer Published10 October 2026 Last checked10 October 2026 Reading time8 minutes Sources10 peer-reviewed

Designing a controlled clinical trial comprises the planning undertaken before any participant is enrolled: framing a testable research question, choosing a design capable of answering it, defining eligibility, and calculating the number of participants needed to detect a meaningful effect if one exists. International Council for Harmonisation (ICH) E9 guidance on statistical principles for clinical trials treats this planning stage as the foundation of the trial’s credibility, because design decisions made before enrolment, and not only conduct decisions made during it, determine whether a trial is capable of answering its own question.

Learning objectives

  • How to frame a testable research question using the PICO format
  • The principal trial designs (parallel-group, crossover and factorial) and the questions to which each is suited
  • Why the sample size calculation is a specific, reportable commitment rather than an approximate estimate

Common misconceptions

  • “A larger sample size is always preferable, so trials should enrol as many participants as is feasible.” ICH E9 guidance frames sample size as a calculation tied to a pre-stated effect size, significance level and statistical power; a trial far larger than its calculated requirement raises its own ethical and resource-use questions, whereas a trial far smaller than its calculated requirement risks being unable to detect a real effect even if one exists. For this reason, reviewers and reporting guidelines expect the calculation itself, and not merely the final number, to be justified.

Framing a testable research question with the PICO format

The design of a trial follows directly from the precision with which its question is framed. The PICO format (Population, Intervention, Comparison, Outcome) structures a research question into four decidable components: who the trial will enrol, what intervention is being tested, what it will be compared against (placebo, standard care or an active alternative), and what outcome will be measured to judge success1. This framing matters for feasibility as much as for clarity: a vaguely stated question, such as “does this drug help patients”, cannot be powered, analysed or reported against, whereas a PICO-framed version, “in adults with stage 2 hypertension (P), does drug X compared with placebo (C) reduce systolic blood pressure at 12 weeks (O) when added to standard care (I)”, can be. The Cochrane Handbook for Systematic Reviews of Interventions recommends the same PICO framing for this reason when defining the scope of a systematic review, since both trial design and evidence synthesis depend on a question specific enough to be answered with a defined set of methods2.

Selection of a trial design appropriate to the question

Once the question is framed, the choice of design follows from practical and scientific constraints. A parallel-group design, in which each participant is randomised to one arm and remains in it for the duration of the trial, is the most common choice and the easiest to interpret, but it requires enough participants to detect a between-group difference. A crossover design, in which each participant receives multiple interventions in sequence with washout periods between them, can detect smaller effects with fewer participants because each participant serves as their own control; it is, however, appropriate only for stable conditions and for interventions whose effects do not persist after treatment stops, since a carryover effect from an earlier period would otherwise contaminate a later one3. A factorial design tests two or more interventions and their combination simultaneously, which is efficient when a research programme needs to test interaction effects between treatments, but it adds analytical complexity that is not justified when the research question concerns only a single intervention. ICH E9 frames this as matching the trial design to the specific question and population, rather than defaulting to the design that is most familiar or most commonly published in a given field4.

Blinding and comparator choice as design decisions

Whether a trial can be blinded at all, and to whom, is a design-stage decision constrained by the intervention itself, not a conduct detail to be resolved later. A double-blind design, in which neither participants nor the clinicians and assessors evaluating them know the assignment, is feasible when an indistinguishable placebo or sham comparator can be constructed, but it becomes infeasible for interventions such as surgery, physical therapy or many behavioural programmes; in these cases, the design team must instead plan for single-blinding of outcome assessors only, or accept an open-label design with other bias-mitigation strategies built in4. The choice of comparator is an equally consequential design decision. A placebo-controlled design isolates the absolute effect of the intervention but can raise ethical concerns when an established effective treatment already exists, whereas an active-comparator design, testing against standard care, answers a more clinically relevant question (whether the new intervention is better than current practice) but typically requires a larger sample size to detect the smaller difference expected between two active treatments than between treatment and placebo.

Sample size as a specific and reportable commitment

Sample size calculation translates four specific, pre-stated choices into a required number of participants per arm: the minimum effect size considered clinically meaningful, the acceptable risk of a false positive (the significance level, conventionally 0.05), the acceptable risk of a false negative (1 minus statistical power, conventionally targeting 80% or 90% power), and the expected variability in the outcome measure4. The CONSORT reporting checklist requires authors to state how the sample size was determined, including these underlying assumptions, because a trial that enrolled too few participants to detect its own pre-specified effect size is at risk of a false-negative conclusion, reporting “no significant difference” when a real effect existed but the trial lacked the statistical power to detect it5. Reviewers and readers assessing the sample-size justification of a published trial examine whether the stated effect size was realistic and clinically meaningful, rather than chosen to produce a convenient sample size.

The statistical analysis plan as a design decision

Whether a trial will be analysed by intention-to-treat, which includes every randomised participant in the arm to which they were assigned regardless of adherence to or completion of the assigned intervention, or by a per-protocol approach that excludes participants with major protocol deviations, must be decided and documented at the design stage, not chosen after the results are known. Intention-to-treat analysis preserves the comparability that randomisation was designed to create, since excluding non-adherent participants after the fact can systematically remove the participants for whom the intervention was least tolerable or effective, thereby biasing the remaining comparison4. The ICH E9(R1) addendum on estimands, published in response to growing recognition that “the treatment effect” can mean different things depending on how intercurrent events such as treatment discontinuation are handled, requires trial teams to pre-specify exactly which clinical question the primary analysis is designed to answer, and to align the chosen statistical approach with that pre-specified estimand, rather than adapting the analysis strategy to whichever approach produces the most favourable result once the data are available6.

Eligibility criteria and their effect on generalisability

Inclusion and exclusion criteria determine the population to which the results of a trial can be generalised, which involves a direct trade-off with internal validity. Narrow, highly specific eligibility criteria, such as the exclusion of participants with comorbidities, of specific age ranges or with prior treatments, produce a more homogeneous study population in which the effect of the intervention is easier to isolate statistically, but the resulting findings may not generalise well to the broader, more heterogeneous population that clinicians treat. ICH E9 and CONSORT both require eligibility criteria to be stated explicitly and completely in the published report so that readers can judge whether the trial population matches the patients for whom they are considering the intervention, rather than assuming that the findings apply uniformly to everyone with the condition under study45.

The role of pilot and feasibility data at the design stage

Several design inputs, including the expected variability used in a sample-size calculation, the likely dropout rate used to inflate the enrolment target, and whether a planned recruitment strategy can realistically reach the target population within a reasonable timeframe, are frequently unknown with any precision before a formal pilot or feasibility study has been conducted. A dedicated pilot or feasibility trial, distinct from the full-scale trial that follows it, is designed to generate these inputs empirically rather than relying on assumption or literature estimates alone, and its results are intended to inform, and sometimes substantially revise, the sample-size calculation and operational plan of the full trial before the larger, more resource-intensive trial begins enrolment. Treating pilot data as a design-stage input, rather than proceeding directly to a full trial on assumed parameters, reduces the risk that a trial is found to be underpowered or operationally infeasible only after substantial resources have been committed to it.

Key points

Key points

Design decisions made before enrolment, namely how the question is framed, which design is chosen, how the sample size is calculated and who is eligible, determine whether a trial is capable of answering its own question, which is why reporting guidelines require each of these to be stated and justified explicitly rather than assumed.

Practical application

Practical application

  1. The research question should be framed in full PICO format before a design is chosen. Population, Intervention, Comparison and Outcome should each be stated explicitly; vagueness in any component will be carried into the design and the sample size calculation.
  2. The design should be matched to the actual constraints of the question rather than to convention. Investigators should consider whether a crossover design is appropriate, given the stability of the condition and the absence of carryover effects, before defaulting to a parallel-group design.
  3. The sample size should be justified with stated, realistic assumptions. The minimum clinically meaningful effect size, significance level, power and expected variability should be documented, and each should be defensible as realistic rather than convenient.
  4. Eligibility criteria should be stated in full in any published report. Stating every inclusion and exclusion criterion, rather than a representative subset, allows readers to judge generalisability for themselves.

Frequently asked questions

An underpowered trial has a reduced chance of detecting a real effect even if one exists, so a non-significant result from such a trial should be interpreted as inconclusive rather than as evidence that the intervention is ineffective; this is why CONSORT requires sample-size justification to be reported and checked.

A crossover design is inappropriate for conditions that change substantially over time (since each participant's own baseline shifts between periods) or for interventions whose effects persist after treatment stops (since a carryover effect from an earlier period can contaminate results in a later period), unless an adequately long washout period can rule this out.

Because PICO requires each design-relevant decision (who is eligible, what is being tested, what it is compared with, and what will be measured) to be made explicit before the trial design and sample-size calculation can proceed; a vague question cannot be powered or analysed.

Not weaker, but narrower in scope: a more homogeneous population improves internal validity by isolating the effect of the intervention more clearly, but readers require the full eligibility criteria to judge whether the trial population matches the patients to whom they intend to apply the findings.

A pilot or feasibility study generates empirical estimates of outcome variability, likely dropout rate and recruitment feasibility, which are inputs on which the sample-size calculation and operational plan of the full trial depend, so that the full-scale trial is not conducted on untested assumptions about parameters that materially affect its prospects of success.

References

  1. Cochrane Handbook for Systematic Reviews of Interventions. Chapter 3: Defining the criteria for including studies and how they will be grouped for the synthesis. read 10 October 2026. View the Cochrane Handbook
  2. Cochrane Handbook for Systematic Reviews of Interventions. Chapter 8: Assessing risk of bias in a randomized trial. read 10 October 2026. View the Cochrane Handbook
  3. Jones B, Kenward MG. Design and Analysis of Cross-Over Trials. 3rd ed. Chapman and Hall/CRC; 2014. read 10 October 2026. View publisher page
  4. International Council for Harmonisation. ICH E9: Statistical Principles for Clinical Trials. read 10 October 2026. View ICH E9 Guideline
  5. Schulz KF, Altman DG, Moher D, for the CONSORT Group. CONSORT 2010 Statement: updated guidelines for reporting parallel group randomised trials. BMJ. 2010;340:c332. read 10 October 2026. View on BMJ
  6. International Council for Harmonisation. ICH E9(R1): Addendum on Estimands and Sensitivity Analysis in Clinical Trials. read 10 October 2026. View ICH E9(R1) Addendum
  7. Piaggio G, Elbourne DR, Pocock SJ, Evans SJ, Altman DG, for the CONSORT Group. Reporting of noninferiority and equivalence randomized trials: extension of the CONSORT 2010 statement. JAMA. 2012;308(24):2594-2604. read 10 October 2026. View on JAMA
  8. Montori VM, Guyatt GH. Intention-to-treat principle. CMAJ. 2001;165(10):1339-1341. read 10 October 2026. View on CMAJ
  9. Friedman LM, Furberg CD, DeMets DL, Reboussin DM, Granger CB. Fundamentals of Clinical Trials. 5th ed. Springer; 2015. read 10 October 2026. View publisher page
  10. Eldridge SM, Lancaster GA, Campbell MJ, et al. Defining Feasibility and Pilot Studies in Preparation for Randomised Controlled Trials: Development of a Conceptual Framework. PLOS ONE. 2016;11(3):e0150205. read 10 October 2026. View on PLOS ONE

Every source above is a peer-reviewed article, a publisher’s record of one, or a standard issued by the body that owns it. Each carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.

How to cite this article

Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.

Pubrica Academy. (2026, October 10). Designing Controlled Clinical Trials: From Research Question to Sample Size Calculation. Pubrica. https://pubrica.com/academy/study-guide/designing-controlled-clinical-trials/

This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.

Trial Design Planning Checklist

This checklist should be completed before the protocol is finalised and before enrolment begins.

[ ] Research question framed explicitly in PICO format (Population, Intervention, Comparison, Outcome)
[ ] Trial design (parallel, crossover, factorial, etc.) chosen to fit the question's specific constraints
[ ] Minimum clinically meaningful effect size stated and justified
[ ] Significance level and statistical power targets stated explicitly
[ ] Expected outcome variability estimated from prior data or pilot work
[ ] Sample size calculated and documented using the above assumptions
[ ] Full inclusion and exclusion criteria listed and justified

Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.

How Pubrica can help

Pubrica’s biostatistics and trial design team works with investigators on PICO framing, design selection and sample size calculation before the protocol is finalised, producing a documented and defensible justification for each design choice.

Find something else

Search the Pubrica Academy

Access is free and requires no account. Enter a topic, or open the full library to filter by category, sub-topic or resource type.

Trial designs capable of answering the research question

A scoping assessment is free and takes two business days. It provides a written assessment of what the work requires and which specialist would undertake it.

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.