9. Investigational plan

Section 9 of a clinical study report sets out how the study was planned and why that plan can answer the study’s questions: the overall design and the reasons for it, who could take part and when they were removed from treatment, which treatments were given and how they were assigned and blinded, which measurements count as evidence and why they can be trusted, how the data were kept reliable, how the analyses and the sample size were planned, and what changed after the study began. ICH E3 sets out what each subsection carries and which appendices hold the supporting documents, and the CORE Reference manual proposes changes to the study schematic and to how the methods that prevent bias are grouped.

Which questions Section 9 answers

Section 9 explains the study plan and how its methods support a reliable answer, covering eligibility, treatments, measurements, data quality, planned analyses and changes during the trial. ¹ The introduction explains the study’s place in development and Section 8 establishes its objectives; Section 9 then shows how the plan makes those questions answerable, before Section 10 follows participants through the study’s execution. ¹

For a study of a new blood-pressure medicine, each subsection answers an everyday question: ²

Everyday question	Section
What happens to participants, from beginning to end?	9.1 — Overall design
Why is this a sensible and fair comparison?	9.2 — Reasons for the design
Who can join, and when should someone stop treatment?	9.3 — Study population
What exactly do people receive, and how?	9.4 — Treatments
What does the study measure, when, and why?	9.5 — Efficacy and safety variables
How does the study keep the data trustworthy?	9.6 — Data quality assurance
How does the study analyse the data, and why this many participants?	9.7 — Statistical plan and sample size
What changes during the study?	9.8 — Changes

²

The protocol is the central source for Section 9, which gives the reviewer an intelligible account of the study’s methods, supported by the protocol and other documents. ²

E3 asks the report to give a clear explanation of how the critical design features of the study were chosen, and enough information on the plan, methods and conduct of the study that there is no ambiguity in how it was carried out. ³ The detailed description of how the study was carried out may simply restate the description in the initial protocol, but often the methodology can be presented more concisely in a separate document. ³ In each section describing the design and conduct of the study, it is particularly important to clarify features of the study that are not well described in the protocol, to identify ways in which the study as conducted differed from the protocol, and to discuss the statistical methods and analyses used to account for these deviations from the planned protocol. ³

9.1 — Overall study design and plan

Section 9.1 is the overview before the detailed explanations. ⁴ E3 asks for the overall plan and design of the study, such as parallel or cross-over, to be described briefly but clearly, using charts and diagrams as needed. ⁵

An arm is a treatment group. ⁴ A parallel-group study might have: ⁴

• Arm A: new medicine for 12 weeks.
• Arm B: comparison medicine for 12 weeks.

Participants remain in their assigned groups during the comparison. ⁴

In a cross-over design, the treatment sequences might be A then B or B then A, and comparing treatments within the same person can reduce the influence of differences between people but creates the challenges discussed in Section 9.2. ⁴

E3 asks the overview to identify:

• Treatments, doses, and procedures.
• Intended population and planned participant numbers.
• Who is blinded and how.
• Control treatments and study configuration.
• How participants are assigned to groups.
• Sequence and duration of all study periods.
• When randomisation occurs.
• Safety, monitoring, steering, or evaluation committees.
• Any interim analyses.

^{4, 5}

E3’s examples of levels of blinding are open, single-blind and double-blind studies, and blinded evaluators with unblinded patients or investigators. ⁵ The study periods it names include pre-randomisation and post-treatment periods, therapy withdrawal periods, and single- and double-blind treatment periods. ⁵

The periods may include screening, a pre-treatment period, treatment, treatment withdrawal, and follow-up. ⁴ A timeline helps the reader see how they relate. ⁴

Screening precedes treatment assignment. ⁴ Follow-up observes participants after a defined point, potentially after treatment ends. ⁴ The baseline is the starting reference for each measurement. ⁴

E3 considers it usually helpful to display the design graphically, with a flow chart that includes the timing of assessments, and points to its Annexes IIIa and IIIb for an example. ⁵ The two annex examples chart the study periods, the dose steps or dose arms, and the assessments scheduled at each visit or study week. ^{6, 7} Annex IIIb also marks the point of randomisation. ⁷

Case report forms (CRFs) capture the observations specified by the protocol. ⁴

E3 also specifies:

• Protocol and amendments in 16.1.1.
• A sample CRF, with unique pages, in 16.1.2.
• Identification of information obtained from sources other than the protocol.
• Where useful, reference to similar studies and important differences.

^{4, 5}

The sample CRF need not repeat identical pages from forms for different evaluations or visits. ⁵

9.2 — Discussion of study design

Describing a design and justifying it are different tasks. ⁸ Section 9.1 gives the arrangement, and Section 9.2 explains why it can produce an interpretable answer. ⁸ E3 asks for the specific control and the study design to be discussed as necessary, including known or potential problems with them in light of the specific disease and therapies studied. ⁹

Choosing the control

A control helps distinguish the treatment’s effect from changes that would occur for other reasons. ¹⁰ E3 names five generally recognised kinds of control group: ⁹

Control type	What the comparison is
Concurrent placebo	Participants studied at the same time receive an inactive comparison product
Concurrent no-treatment	A group receives no study treatment
Concurrent active treatment	A group receives an established treatment
Concurrent dose comparison	Groups receive different doses
Historical control	Current participants are compared with earlier patients or data

^{10, 9}

“Concurrent” means studied during the same general period. ¹⁰

Improvement alone is difficult to interpret: disease may fluctuate, other care may help, or participants may improve naturally. ¹⁰ A suitable control helps separate these influences. ¹⁰

Historical controls require particular explanation because patients, diagnostic methods, supportive care, and disease management may differ across periods. ¹⁰ Bias is a systematic distortion of a result, and the absence of randomisation raises the risk of selection bias, in which the way people enter groups creates systematic differences between those groups. ¹⁰ Among the known limitations of historical controls, E3 names the difficulty of assuring comparable groups, the inability to blind investigators to treatment, changes in therapy or disease, and differences due to placebo effect, and holds that they deserve particular attention. ⁹ Where randomisation was not used, E3 asks how other techniques, if any, guarded against systematic selection bias. ⁹

Cross-over complications

Two terms apply to cross-over designs:

• Carry-over effect: the first treatment continues influencing outcomes during the next treatment period.
• Washout period: an interval intended to allow previous treatment effects to dissipate.

¹¹

Cross-over designs also become difficult if the disease changes substantially over time: a treatment received second could appear different because the condition changes. ¹¹

For a cross-over design, E3 asks the design discussion to consider the likelihood of spontaneous change in the disease and carry-over effects of treatment. ⁹ E3 also names the presence or absence of washout periods among the other design features that may deserve discussion. ⁹

Superiority, non-inferiority, and equivalence

Aim	Meaning
Superiority	One treatment is better on the specified outcome
Non-inferiority	The new treatment is not worse by more than a justified margin
Equivalence	Differences lie within specified acceptable limits in both directions

¹²

E3’s 1995 wording uses “equivalence” for the absence of a specified degree of inferiority, and later terminology distinguishes these designs more precisely. ^{12, 9}

The margin, sometimes written Δ or delta, defines the difference the study aims to exclude. ^{12, 9} It requires justification and cannot be selected merely to make the result favourable. ¹² E3 asks the design discussion to cover the degree of inferiority the study was intended to show was not exceeded. ⁹

A failure to find a statistically significant difference does not establish equivalence or non-inferiority. ¹² A study might simply be too imprecise. ¹²

E3 also asks whether the trial could distinguish an effective treatment from an ineffective one. ¹² Later guidance calls this ability assay sensitivity, where “assay” refers to the trial’s ability to detect treatment effects, not to a laboratory test. ¹²

If both groups receive treatment poorly, or the measurements cannot detect improvement, similar results may be uninformative. ¹² When efficacy is to be shown through equivalence, E3 asks for a basis for considering the study capable of distinguishing active from inactive therapy. ⁹ Support may come from an analysis of earlier studies that resemble the present one in important design characteristics, such as patient selection, endpoints, duration, dose of active control and concomitant therapy, and that consistently showed the active control superior to placebo. ^{9, 12} The report should also discuss how to assess this ability in the present study, for example by identifying from past studies a treatment response, such as a change from baseline or a healing or survival rate, that would clearly distinguish a treated population from an untreated group. ⁹ Attaining such a response would support the expectation that the study could have distinguished the active drug from an inactive one. ⁹

Treatment duration and dose intervals

E3 asks whether the study lasts long enough and whether the dosing schedule makes sense. ^{13, 9} It holds that the duration of the treatment period may deserve discussion, especially for a chronic illness, and that the rationale for dose and dose-interval selection should be explained if it is not obvious. ⁹

A drug’s half-life describes how quickly its concentration decreases, and its pharmacodynamic effect is what it does to the body. ¹³ The effect may last longer than the measurable concentration suggests. ¹³

If a short-lived drug is given once daily, the report should explain why this schedule is reasonable. ^{13, 9} Measurements just before the next dose can investigate whether the effect wears off, which E3 calls “escape” from drug effect, and E3 asks for the procedures used to seek evidence of it at the end of the dose interval to be described. ^{13, 9} E3 notes that once-daily dosing of a short-half-life drug whose effect is closely related in time to blood level is not usually effective, and a design that uses such dosing should be explained, for example by pharmacodynamic evidence that the effect is prolonged compared with blood levels. ⁹

The design discussion also covers selection of patients based on previous response or non-response to treatment and the rationale for doses in dose-response studies. ^{13, 9}

9.3 — Selection of the study population

Section 9.3 defines the people about whom the study can say something meaningful. ¹⁴

9.3.1 — Inclusion criteria

The report should describe the patient population and the selection criteria used to enter patients into the study, and discuss the population’s suitability for the purposes of the study. ¹⁵

Inclusion criteria are the conditions participants must satisfy, such as a confirmed diagnosis, a particular disease severity, a specified age range, or inadequate response to existing treatment. ¹⁶

E3 asks for:

• Diagnostic criteria.
• Disease severity and duration.
• Required test, examination, or rating-scale findings.
• Relevant treatment history.
• Prognostic factors.
• Applicable age, sex, and ethnic criteria.
• Separate requirements for screening, randomisation, and treatment entry.

^{16, 15}

A prognostic factor is a characteristic associated with future outcome, for example baseline disease severity. ¹⁶

Passing screening does not necessarily mean immediate randomisation. ¹⁶ A participant may need to satisfy additional requirements after a preliminary period. ¹⁶

If there is reason to believe that entry criteria not defined in the protocol were applied, the report should discuss their implications. ^{15, 16} For example, if investigators systematically avoid enrolling the sickest eligible patients, the actual population may be narrower than the eligibility rules suggest. ¹⁶ E3’s example is investigators who excluded patients who were particularly ill or had particular baseline characteristics, or entered them into other studies. ¹⁵

9.3.2 — Exclusion criteria

Exclusion criteria are the conditions that prevent entry, and the report should give them with their rationale. ^{17, 18} Reasons can include safety, inability to complete necessary assessments, or factors making the person unsuitable for the study’s purpose. ¹⁸ E3’s examples are safety concerns, administrative reasons and lack of suitability for the trial. ¹⁷

Exclusions affect generalisability: how well findings apply beyond the enrolled population. ^{18, 17} A study excluding people with kidney impairment provides limited direct evidence about that group. ¹⁸ E3 places the fuller discussion of this consequence in Section 13 or an appropriate overview. ^{18, 17}

9.3.3 — Removal from treatment or assessment

The report should describe the predetermined reasons, if any, for removing a participant from treatment or from assessment, and the nature and duration of any planned follow-up of those participants. ^{19, 20}

For example, a specified laboratory abnormality might require stopping the study medicine while safety follow-up continues. ²⁰

The plan specifies separately when treatment stops and whether assessments continue. ²⁰

9.4 — Treatments

Section 9.4 has eight subsections because treatment identity, assignment, dosing, blinding, and actual use each affect interpretation. ²¹

9.4.1 — Treatments administered

For every arm and every study period, the report should describe:

• Product or diagnostic agent.
• Dose.
• Route and mode of administration.
• Schedule.

^{22, 23}

Route means how the treatment enters the body: oral, intravenous, inhaled, and so on. ²² Mode adds relevant detail, such as an intravenous infusion over a specified duration. ²²

“Participants receive Drug A” is insufficient, whereas “Participants receive Drug A 10 mg orally once daily for 12 weeks” is much more informative. ²²

9.4.2 — Identity of investigational products

Section 9.4.2 identifies the physical material used:

• Formulation: tablet, capsule, solution, etc.
• Strength: amount per unit, such as 5 mg per tablet.
• Batch number: the manufacturing identifier.
• Sources of placebo and active comparator.
• Changes made to a commercial comparator.
• Storage conditions and resupply arrangements where relevant.
• Any use after expiry, including affected participants.

^{24, 25}

E3 asks for the formulation, strength and batch numbers to be given briefly in the text of the report. ²⁵ For long-duration trials of products with limited shelf-lives or incomplete stability data, the report should describe the logistics of resupply, and it should describe any specific storage requirements. ²⁵

Strength and dose differ: two 5-mg tablets deliver a 10-mg dose. ²⁴

If multiple investigational-product batches are used, 16.1.6 identifies which participants receive each. ^{24, 25}

Changing a comparator, for example by placing it inside a capsule to conceal its appearance, may affect its bioavailability: the extent and rate at which its active ingredient becomes available in the body. ²⁴ E3 asks how unchanged bioavailability is assured. ^{24, 25}

9.4.3 — Assignment to treatment groups

The report should describe the actual assignment method and its execution, and explain any unusual features. ^{26, 27} Assignment methods are described in these common terms: ²⁶

Term	Meaning
Centralised allocation	Assignment managed through a central process
Site-level allocation	Assignment managed within individual study centres
Stratification	Separate allocation within categories of important baseline factors
Blocking	Allocation arranged to maintain the intended group ratio within blocks
Adaptive allocation	Assignment depends on information such as earlier assignments or outcomes, according to the specified method

^{26, 27}

Randomisation reduces systematic selection differences, but it does not guarantee identical groups in every realised trial. ²⁶

Allocation concealment prevents foreknowledge of the next assignment when enrolling a participant. ²⁶ It protects the assignment process, whereas the blinding introduced in Section 9.1 protects subsequent treatment and assessment. ²⁶

E3 asks for detailed randomisation methods, random-number generation, codes, participant identifiers, and assigned treatments in 16.1.7, organised by centre where applicable. ^{26, 27}

For a historically controlled trial, the report should explain how the particular control is selected, what other historical experience is examined, if any, and how its results compare with the control used. ^{27, 26}

9.4.4 — Selection of doses in the study

Section 9.4.4 explains why the study uses its dose levels. ²⁸ It should give the doses or ranges for every treatment and their basis, for example previous human experience or animal data. ^{28, 29}

This is the study-level rationale: Why investigate 5, 10, and 20 mg? ²⁸

9.4.5 — Selection and timing of dose for each participant

Section 9.4.5 explains how each participant arrives at their dose. ³⁰ Possibilities include:

• A fixed assigned dose.
• Titration: stepwise dose adjustment.
• Adjustment according to response or tolerability.
• Back-titration: reducing a previously increased dose.

^{30, 31}

The report should describe the adjustment rules, dosing intervals, time of day, relationship to meals, and instructions to participants, and should say so if timing or food conditions are not specified. ^{30, 31}

This is the participant-level procedure: Start at 5 mg; increase according to specified criteria. ³⁰

9.4.6 — Blinding

The report should explain exactly who can know the treatment assignment and how that knowledge is controlled. ³² Unblinding, or breaking the blind, reveals an assignment. ⁴

Labels such as “double-blind” are useful shorthand, but the report should identify the roles: participants, treating investigators, outcome assessors, sponsor staff, and committee members as applicable. ³²

E3 asks Section 9.4.6 to cover:

• Packaging, labels, and matching appearance, smell, and taste.
• Code storage and access.
• Conditions and procedures for breaking the blind.
• Protection of blinded staff when others must remain unblinded.
• Laboratory results or characteristic effects that might reveal treatment.
• Unblinded monitoring committees.
• Protection of blinding during interim analyses.
• Why blinding is unnecessary or infeasible, where applicable.
• Known or potential failures and mitigation.

^{32, 33}

A double-dummy arrangement helps when treatments look different: to compare a tablet with an injection, one group might receive active tablet plus placebo injection and the other placebo tablet plus active injection, so both groups experience the same administration forms. ³² E3 names double-dummy techniques among the blinding procedures the report should describe. ³³

E3’s other examples of blinding procedures are how bottles were labelled, labels that reveal blind-breakage, and sealed code lists or envelopes. ³³ It asks for the circumstances in which the blind would be broken for an individual or for all patients, for example for serious adverse events, and for who had access to patient codes. ³³ The report should describe the measures taken to make the test product and placebo indistinguishable and the evidence that they were. ³³

Where blinding was considered unnecessary to reduce bias for some or all observations, the report should explain why: for example, a random-zero sphygmomanometer eliminates possible observer bias in reading blood pressure, and Holter tapes are often read by automated systems presumed immune to observer bias. ³³ If blinding was considered desirable but not feasible, the reasons and implications should be discussed. ³³

Blinding is sometimes known to be imperfect because of obvious drug effects in at least some patients, such as dry mouth, bradycardia, fever, injection-site reactions or changes in laboratory data. ³³ The report should identify such problems and describe any attempts to assess or manage them, for example by having some endpoint measurements made by people shielded from information that might reveal treatment assignment. ³³

9.4.7 — Prior and concomitant therapy

Concomitant therapy means other treatment taken alongside the study treatment. ³⁴ The report should describe the treatments allowed or prohibited before and during the study, how their use is recorded, and the relevant rules. ^{34, 35} Rescue medication is additional treatment available when disease control is inadequate, and its permitted use and recording need clear specifications. ³⁴

Other treatment can affect interpretation through:

• Drug–drug interactions: one drug alters another’s effects or handling.
• Direct effects on endpoints: another treatment improves the same measured outcome.

^{34, 35}

For example, additional blood-pressure medication could affect the apparent effect of the investigational medicine. ³⁴ The report should explain how these influences are addressed. ^{34, 35}

9.4.8 — Treatment compliance

Treatment compliance, also commonly called adherence, is whether participants take treatment as instructed, and Section 9.4.8 describes the measures taken to ensure and document it. ^{36, 37} E3 gives examples:

• Drug accountability, such as returned-tablet counts.
• Diaries.
• Drug concentrations in biological samples.
• Medication-event monitoring.

^{36, 37}

These are evidence about treatment use, each with its own limitations. ³⁶ Section 9 describes the assessment method, and later sections report the findings. ³⁶

9.5 — Efficacy and safety variables

The study’s benefit and risk questions require measurements. ³⁸ A variable is something measured or classified: efficacy variables assess benefit and safety variables help assess harm. ³⁸ A treatment effect compares outcomes under the treatments of interest. ³⁸

Section 8, by connecting objectives to endpoints, establishes why each measurement matters, and Section 9.5 specifies exactly how the study obtains it. ³⁸

9.5.1 — Measurements and their schedule

For every relevant assessment, the report should explain:

• What is measured.
• How it is measured.
• Who measures it.
• When it is measured.
• Relevant timing relative to doses, meals, or time of day.
• Instructions given to participants.
• Changes in personnel performing critical measurements.

^{39, 40}

Timing matters: a measurement two hours after a dose may answer a different question from one immediately before the next dose. ^{39, 40}

E3 recommends a schedule showing meaningful times, not visit numbers alone. ^{39, 40} It also asks for full definitions of clinical outcomes: if “stroke” is an endpoint, the report should explain what qualifies as a stroke and how it is classified. ^{39, 40} E3’s other examples of such definitions are the criteria for an acute myocardial infarction and the designation of its location, the distinction between TIA and stroke, and the assignment of cause of death. ⁴⁰

Standardisation makes measurements comparable across people and centres. ³⁹ Central reading uses a central assessor or group to interpret measurements such as images or ECGs. ³⁹ An adjudication committee may evaluate whether reported events meet predefined endpoint criteria. ³⁹ The report should identify who performs these evaluations and describe their procedures and how blinding is maintained. ^{39, 40} The report should also describe the techniques used to standardise or compare laboratory and other clinical measurements, which is particularly important in multicentre studies, and for these studies it should indicate how methods were standardised. ⁴⁰

The definition of an adverse event (AE) covers symptoms, diagnoses, and abnormal test results, and the collection method must capture relevant events without assuming causation. ^{39, 41} An adverse reaction adds a suspected causal relationship. ^{39, 41} Severity describes intensity, whereas causality, or relatedness, concerns whether treatment is suspected of causing the event. ³⁹

For safety data, the report should explain:

• Whether AEs are volunteered, elicited through questions, or collected using checklists.
• Rating scales.
• Severity and causality criteria.
• Who assigned the ratings.
• Planned follow-up and any procedure for restarting treatment after an event.
• Definitions behind categorical ratings or numerical scores.

^{39, 40}

This account of collection methods is the counterpart to Section 12: different ways of asking and measuring can affect what is recorded. ³⁹

9.5.2 — Appropriateness of measurements

Section 9.5.2 explains why the measurements can be trusted to answer the question. ⁴² For a nonstandard assessment, the report should document:

• Reliability: consistency.
• Accuracy: closeness to what is intended to be measured.
• Relevance: usefulness for the clinical question, including ability to distinguish effective from ineffective treatment.

^{42, 43}

E3 counts an assessment as standard when it is widely used and generally recognised as reliable, accurate, and relevant, and adds that it may be helpful to describe the alternatives considered but rejected. ⁴³

A surrogate endpoint substitutes a laboratory or physical measurement for a direct measure of clinical benefit, and its use needs justification. ^{42, 43}

Being easy to measure does not establish that a surrogate predicts how people feel, function, or survive, and the supporting evidence depends on the particular surrogate and context. ⁴² E3’s examples of such justification are references to clinical data, publications, guidelines, or previous actions by regulatory authorities. ⁴³

9.5.3 — Primary efficacy variables

The objective hierarchy in Section 8 determines the roles of the primary and secondary endpoints. ⁴⁴ Prespecified means defined before the relevant results are known, and post hoc describes a choice developed after examining the data or results. ⁴⁴ An exploratory analysis investigates patterns or generates hypotheses, whether planned or post hoc. ⁴⁴

The definition of a primary variable must specify the relevant measurement, its timing, and any rule combining information. ⁴⁴ With many measurements or visits, the reader should not have to guess which determines the principal result. ⁴⁴ For example: ⁴⁴

Difference between treatment groups in mean change from baseline in systolic blood pressure at Week 12.

The report should explain why the primary variables are chosen and describe any predefined efficacy threshold. ^{44, 45}

Where there are multiple variables or variables measured repeatedly, E3 expects the protocol to identify the primary ones, or to designate the pattern of significant findings or other method of combining information that would be interpreted as supporting efficacy. ⁴⁵

If the protocol does not identify the primary variables, E3 asks how and when selection takes place, including its timing relative to completion and unblinding. ^{44, 45} Choosing after viewing results can favour whichever outcome looks best. ⁴⁴ E3’s examples of how the selection is explained are references to publications, guidelines or previous actions by regulatory authorities. ⁴⁵

9.5.4 — Drug concentration measurements

The pharmacokinetic (PK) questions in Section 8 require concentration measurements. ⁴⁶ People can handle the same administered dose differently. ⁴⁶ Renal means kidney-related, and hepatic means liver-related. ⁴⁶

Section 9.5.4 describes how the study collects that information. ⁴⁶ The report should specify:

• Concentrations measured and biological samples used.
• Sampling times relative to dosing.
• Food, posture, and relevant concomitant medicines or substances.
• Sample handling.
• Measurement methods and references to assay validation.
• Timing of other relevant factors, such as renal or hepatic function.

^{46, 47}

An assay here is the laboratory measurement method, and validation provides evidence that it performs adequately for its intended purpose. ⁴⁶

9.6 — Data quality assurance

Even a good design needs dependable execution and records. ⁴⁸

E3 asks for a brief account of systems used to obtain accurate, consistent, complete, and reliable data. ^{48, 49} If the study uses none, the report should state this. ^{48, 49}

Examples of such systems include:

• Investigator training and meetings.
• Standard instructions and terminology.
• Site monitoring.
• Verification and cross-checks.
• Central laboratories and central ECG interpretation.
• Audits.

^{48, 49}

Broadly, quality assurance concerns the systems supporting quality, and quality control concerns operational checks that requirements are met. ⁴⁸

An audit is an independent examination. ⁴⁸ E3 asks that independent internal or external audit arrangements be mentioned, with supporting information and available certificates in 16.1.8. ^{48, 49} Inter-laboratory standardisation and relevant quality procedures belong in 16.1.10. ^{48, 49}

ICH E6(R2) asks the sponsor to describe the quality management approach implemented in the trial and to summarise, in the clinical study report, the important deviations from the predefined quality tolerance limits and the remedial actions taken. ^{50, 51} Reporting audits in Section 9.6 is not enough to comply with this requirement. ⁵¹

9.7 — Statistical plans and sample size

Section 9.7 sets out how the plan turns observations into a defensible comparison. ⁵²

9.7.1 — Statistical and analytical plans

The emphasis is on planned analyses, including changes made before outcome results become available, and not on the analyses actually used. ^{53, 54} The report should describe:

• Comparisons and statistical tests.
• Which measurements and time points contribute.
• Whether the analysis uses changes, averages, slopes, or time-to-event methods.
• Planned adjustment for covariates.
• Reasons for excluding available participant data.
• Prespecified subgroup analyses.
• Definitions of categorical responses.
• Monitoring committee composition and procedures.
• Interim analyses, stopping rules, and associated statistical adjustments.

^{53, 54}

The terms in such a plan mean the following: ⁵³

Term	Meaning
Change from baseline	Difference from the starting measurement
Slope analysis	Analysis of how a measurement changes over time
Time-to-event analysis	Analysis of time until an outcome occurs
Covariate	A variable included in the analysis, such as baseline severity
Subgroup analysis	Separate examination of a defined subset
Responder	Participant meeting a specified response threshold
Interim analysis	Examination of accumulating results before final completion

^{53, 54}

Repeatedly checking for a favourable result creates additional opportunities for chance findings. ⁵³ Planned interim analyses therefore need appropriate statistical handling and protection against operational bias. ⁵³

A statistical analysis plan (SAP) supplies detailed analysis specifications. ⁵³

9.7.2 — Determination of sample size

Section 9.7.2 explains why the study plans, for example, 100 participants rather than 20 or 1,000. ⁵⁵

E3 asks for the planned number, calculation method or reference, assumptions, and the basis for those assumptions. ^{55, 56} Practical constraints should also be explained where they determine the number. ^{55, 56}

A null hypothesis is the statistical claim tested, often that treatments do not differ. ⁵⁵ Typical statistical concepts include:

• Effect size: the magnitude of an effect, such as the treatment difference targeted in the sample-size calculation.
• Variability: how much outcomes differ between people.
• Type I error: falsely rejecting a true null hypothesis, such as concluding that treatments differ when they do not.
• Significance level: the threshold controlling that error under the planned test.
• Power: the probability of detecting the specified effect if it exists, under the calculation’s assumptions.
• Expected missing data or dropout: factors that can reduce available information.

⁵⁵

An 80% power calculation does not mean an 80% chance that the medicine works. ⁵⁵

For a superiority study the report should specify the difference targeted, and for a non-inferiority study the unacceptable difference the study needs to exclude. ^{55, 56}

Sample size is tied to the particular question and assumptions, so a large study is not automatically capable of answering every safety or subgroup question. ⁵⁵

9.8 — Changes after the study began

When a study changes after it begins, the reviewer needs to understand both the change and the information available at the time of the decision. ⁵⁷

Examples include changing eligibility, dose, sample size, or analysis methods, and dropping a treatment group. ^{57, 58}

A protocol amendment formally updates the plan. ⁵⁷ Departures from the applicable plan during execution are assessed in Section 10.2 on protocol deviations, and following an applicable amendment is not itself such a departure. ⁵⁷

E3 asks for:

• What changed.
• When and why.
• Decision-making procedure.
• Responsible people or groups.
• Data available at the time and who could access them.
• Whether a formal protocol amendment documents the change.
• Consequences for interpretation.

^{57, 58}

Relevant changes should be reported whether or not a formal amendment documents them. ^{57, 58} Routine personnel changes need not be included in Section 9.8, although Section 9.5.1 separately asks about changes in personnel carrying out critical measurements. ^{57, 58} The possible implications of a change for interpreting the study should be discussed briefly in Section 9.8 and more fully in other appropriate sections of the report, and every section of the report should distinguish clearly between conditions planned in the protocol and amendments or additions. ⁵⁸

The timing of a change relative to unblinding and the availability of outcomes matters: a change made without knowledge of comparative outcomes generally raises different concerns from one chosen after seeing which analysis favours treatment. ⁵⁷ E3 holds that changes in planned analyses made before the blind is broken generally have limited implications for interpretation, so it is particularly critical that the timing of changes relative to blind breaking and to the availability of outcome results be well characterised. ⁵⁸

A change explicitly specified as part of an adaptive design also differs from an unplanned change introduced in response to emerging findings, and the report must make clear which kind each change is. ⁵⁷

Where the supporting documents live

Appendix	Supporting material
16.1.1	Protocol and amendments
16.1.2	Sample CRF
16.1.6	Participants receiving specific product batches
16.1.7	Randomisation scheme and codes
16.1.8	Audit documentation and available certificates
16.1.9	Documentation of statistical methods
16.1.10	Inter-laboratory standardisation and quality procedures

⁵⁹

The investigational plan in CORE Reference

The CORE Reference manual proposes changes to the study schematic in Section 9 and to the grouping of two of its subsections. ^{60, 61}

CORE Reference redesigns E3’s two design annexes, IIIa and IIIb, and combines them into a single example figure, Figure 9.1. ⁶⁰ It recommends placing the figure early in Section 9.1, because the schematic of the study design highlights the link between the report and the protocol. ⁶⁰ The figure’s title carries the protocol number. ⁶⁰

CORE Reference holds that E3’s 9.4.3, on the method of assigning participants to treatment groups, and 9.4.6, on blinding, often overlap and are both methods designed to prevent bias, and it therefore suggests merging their content under a suitable new name, such as avoidance of bias. ⁶¹ It also suggests that the procedures for unblinding accompany the blinding procedures, so within the merged section E3’s blinding subsection becomes 9.4.3.2, on blinding and unblinding. ⁶²