11. Efficacy evaluation

Section 11 of a clinical study report under ICH E3 shows whether the treatment worked and how far the study supports that answer. It sets out whose results enter each analysis, what the participants were like at baseline and whether they took the treatment as intended, then reports how large the benefit is and how precisely it is estimated, which analytical choices could change the answer, the individual data behind the group results, and what the efficacy conclusions may claim. CORE Reference moves its first three parts into Section 10, and a phase I oncology report may place it after the safety evaluation.

Which questions Section 11 answers

Section 11 examines whether the treatment provides the intended benefit, how large that benefit is, and how convincingly the data support it. It also examines whether missing measurements, treatment adherence, participant differences, or analytical choices change the answer. Section 9 establishes the question and analysis plan; Section 10 accounts for the participants and their study history. Section 11 brings those foundations together to interpret efficacy. ¹

For a study comparing a new blood-pressure medicine with placebo, each part of Section 11 answers one everyday question.

Everyday question	Where Section 11 answers it
Whose results enter the analysis?	11.1 — Data sets analysed
What are participants like at baseline?	11.2 — Baseline characteristics
Do participants take treatment as intended?	11.3 — Treatment compliance
How much benefit does treatment produce?	11.4.1 — Analysis of efficacy
Can the analysis method change the answer?	11.4.2 — Statistical issues
What happens to each participant?	11.4.3 — Individual response data
Does benefit vary with dose or concentration?	11.4.4 — Dose–response relationships
Do other medicines or illnesses influence response?	11.4.5 — Interactions
Do individual graphs help explain the results?	11.4.6 — By-patient displays
What can the study conclude?	11.4.7 — Efficacy conclusions

²

The first distinction is between getting better during treatment and getting better because of treatment. ² If average blood pressure falls by 12 mmHg with the new medicine and by 7 mmHg with placebo, the observed benefit over placebo is 5 mmHg, not 12 mmHg. ² Changes within a group can also reflect background care, natural variation, and other influences. ²

The endpoint definitions in 9.5 establish what counts as benefit and which findings carry the main claim, and Section 11 compares the observed results against those definitions. ²

11.1 — Data sets analysed

A study might randomise 400 people, but its different analyses may contain different numbers:

• Some participants never start treatment.
• Some miss the final assessment.
• Some stop treatment early.
• Some have major protocol deviations.
• Some have measurements outside the intended time window.

³

E3 asks the report to define exactly which participants and observations enter each efficacy analysis. ^{4, 3} Its examples of such definitions are all patients receiving any test drug, all patients with any efficacy observation or with a certain minimum number of observations, only patients completing the trial, all patients with an observation during a particular time window, and only patients with a specified degree of compliance. ⁴

Common analysis populations include:

Analysis population	Basic idea
Intention-to-treat, ITT	Analyse participants according to their randomised assignment, regardless of subsequent adherence
Full analysis set, FAS	A defined set intended to approximate the ITT principle as closely as possible
Per-protocol, PP	A defined subset meeting specified requirements for following the protocol
Completers	Participants who reach the completion point defined in 10.1

³

A population's name alone does not define it. A report must state the actual rules, because two studies using “full analysis set” may have different inclusion criteria. ³

E3 specifically asks:

• How inclusion and exclusion rules are developed.
• Whether they are established before or after unblinding, using the procedures described in 9.4.6.
• Which participants, visits, and observations are excluded, with reasons, in 16.2.3.
• How exclusions differ between treatment groups and over time.

^{3, 4}

Annexes VI and VII illustrate listings of excluded people or observations and summaries of exclusion reasons over time. ^{4, 5, 6} The Annex VI listing gives, centre by centre and separately for the test drug, the active comparator and placebo, each excluded patient's number, sex and age, the observation excluded and the reasons. ⁵ The Annex VII table counts, for the test drug with its N, the patients excluded for each reason at weeks 1, 2, 4 and 8, with a total, and similar tables are prepared for the other treatment groups. ⁶

For studies intended to establish efficacy, E3 also generally calls for a broad analysis using all randomised or otherwise entered participants with on-treatment data, even when the proposed primary analysis uses a smaller subset. ^{4, 3}

Excluding people can change the comparison: removing participants who stop because treatment is ineffective may make the remaining results look better. ³

The estimand in 9.7.1 specifies the treatment effect of interest. The handling of missing data in 11.4.2.2 examines whether the available observations support its estimation. ³

11.2 — Demographic and other baseline characteristics

Before outcomes are compared, the report establishes who enters the study and how similar the groups are at baseline. ⁷

Randomisation reduces systematic allocation differences, but it does not guarantee identical groups in a particular trial. ⁷

E3 asks this section to present group data for the critical demographic and baseline characteristics, and for other factors arising during the study that could affect response. Which variables are critical depends on the specific nature of the disease and on the protocol. ⁸

The prognostic factors in 9.3.1 guide which baseline characteristics deserve particular attention. ⁷

E3 asks for relevant characteristics such as:

Category	Examples
Demographics	Age, sex, race
Disease characteristics	Duration, stage, severity, relevant clinical classifications
Starting measurements	Baseline blood pressure, symptom score, or other important measurements
Other illnesses	Kidney disease, diabetes, heart failure; relevant previous illnesses
Treatment history	Previous treatment for the condition
Other medicines	Treatments continued, stopped, or changed around study entry
Biological factors	Weight, antibody levels, metabolic status, or other pertinent markers
Lifestyle and reproductive factors	Smoking, alcohol, diet, menstrual status where relevant

^{7, 8}

E3 asks for more than one baseline table when necessary:

• Describe the broad “all patients with data” population first.
• Describe other principal analysis populations.
• Show the characteristics of the corresponding excluded participants.
• Where appropriate, compare centres as well as treatment groups.
• Include a diagram showing how the analysis populations relate.

^{7, 8}

Group tables and graphs belong in 14.1, with relevant participant-level information and concomitant medication in 16.2.4. ^{7, 8} E3 asks for those by-patient listings to be broken down by treatment, and by centre in a multicentre study, and limits the appendix to the most relevant data, generally the characteristics listed above, although some regulatory authorities require all baseline data in tabular listings elsewhere. ⁸

The baseline data help answer two different questions: are the treatment groups reasonably comparable, and how well do these participants represent the people who might eventually receive the medicine? ⁷

11.3 — Measurements of treatment compliance

The compliance methods in 9.4.8 establish how treatment use is measured. Section 11.3 reports those measurements and their pattern over time. ⁹

E3 asks for summaries by:

• Treatment group.
• Time interval.

^{10, 9}

Individual measurements and relevant concentrations are tabulated in 16.2.5. ^{10, 9}

For example, a participant prescribed one tablet daily for 100 days who takes 80 tablets has 80% adherence under a simple count-based definition. That number does not show whether they miss scattered doses or stop entirely for the final 20 days. ⁹

Adherence affects interpretation, but poor adherence is not an automatic justification for removing someone from the main analysis. ⁹

11.4 — Efficacy results and individual participant data

With the analysis populations, baseline characteristics, and adherence results established, the report can examine the treatment comparison. It moves from the main result to its analytical assumptions and then to the individual observations supporting it. ¹¹

11.4.1 — Analysis of efficacy

E3 asks for treatment comparisons covering all critical efficacy measures:

• Primary and secondary endpoints.
• Relevant pharmacodynamic endpoints, using the biological-effect concept introduced in 9.2.
• Individual benefit–risk assessments, where used.
• Analyses planned in the protocol.
• The broad analysis of participants with on-study data.

^{12, 13}

The central result usually contains three elements:

Element	What it answers
Point estimate	What is the numerical estimate of the treatment difference?
Confidence interval, CI	How precise is that estimate?
Hypothesis test / p-value, where used	How compatible are the data with the specified null hypothesis?

¹²

For the blood-pressure example:

The estimated additional reduction with treatment is 5 mmHg, with a 95% confidence interval of 2 to 8 mmHg.

¹²

The point estimate is 5 mmHg, and the interval expresses the uncertainty around it under the analysis assumptions. Strictly, a 95% confidence-interval procedure would contain the true parameter in 95% of repeated comparable studies; it is not a 95% probability statement about this particular fixed interval. ¹²

A p-value is calculated assuming a specified null hypothesis—often no treatment difference—and the statistical model. It describes how unusual results at least as extreme as those observed would be under those assumptions. It is not the probability that the medicine works. ¹²

Statistical significance and clinical importance are separate. A tiny difference can be statistically significant in a large study; whether it matters to patients depends on its size, the endpoint, the disease context and the uncertainty. ¹² Reporting should emphasise the effect size and confidence intervals rather than rely on p-values. ¹² E3 asks the analysis to show the size of the difference between the treatments and the associated confidence interval, with the results of hypothesis testing where it is used. ¹³

E3 also distinguishes different ways to express response:

• Continuous: how much blood pressure or a symptom score changed.
• Categorical: whether a participant met a defined response threshold, such as cure/no cure.
• Time-to-event: an outcome such as relapse, analysed with the method introduced in 9.7.1.

^{12, 13}

E3 holds that analyses of continuous variables and of categorical responses can be equally valid. ¹³ A responder analysis applies the response criterion specified in 9.7.1. ¹² If both continuous and categorical analyses are planned and are available, both ordinarily should be reported. Newly created categories need explanation. ^{12, 13}

E3 also asks Section 11.4.1 to:

• Describe how response develops over time.
• Briefly assess other reasonable measurements, not only the favoured one.
• Present relevant centre-level results, especially for larger centres.
• If investigators and an independent committee assess the same outcomes, show disagreements and identify affected participants.
• Make clear which assessment is used in each analysis.

^{12, 13}

An endpoint can inform both efficacy and safety. Death, for example, may be the main efficacy outcome in a life-threatening disease trial. ¹² E3 says that efficacy and safety endpoints are in many cases difficult to distinguish, and that many of the principles it sets out for the efficacy analyses should be adopted for critical safety measures as well. ¹³

11.4.2 — Statistical and analytical issues

Section 11.4.2 is the most technical part of Section 11. Its purpose is to make the analytical choices understandable and to assess their implications. ¹⁴

E3 asks for an explanation accessible to clinical and statistical reviewers, with detailed methods in 16.1.9. ^{14, 15} It asks for any changes in the analysis made after blind-breaking to be identified. ¹⁵ The discussion covers the important features of the analysis: the particular methods used, adjustments for demographic or baseline measurements or concomitant therapy, the handling of dropouts and missing data, adjustments for multiple comparisons, special analyses of multicentre studies, and adjustments for interim analyses. ¹⁵

E3 then names eight specific issues, 11.4.2.1 to 11.4.2.8, that the report should address unless they do not apply. ^{15, 14}

11.4.2.1 — Adjustments for covariates

The covariates specified in 9.7.1 enter the analysis. The report explains how they affect the treatment estimate. ¹⁶

Where participants with higher starting blood pressure tend to experience larger reductions, an adjusted analysis can account for starting blood pressure when estimating the treatment difference. ¹⁶

Common methods include:

• ANCOVA — analysis of covariance: often used for a continuous outcome while adjusting for baseline measurements.
• Cox regression: used to examine time-to-event outcomes while accounting for relevant factors.

¹⁶

E3 asks the report to explain:

• Which covariates are selected and why.
• How adjustment is performed.
• The results and supporting statistical output.
• Any departures from the planned covariates or methods.
• Planned analyses as well, where relevant and possible.

^{16, 17}

E3 places the methods of adjustment, the results and the supporting output in the detailed documentation of statistical methods, and notes that comparing covariate adjustments and prognostic factors across studies, though not part of an individual study report, may be informative in a summary of clinical efficacy data. ¹⁷

Adjustment is not a licence to try many models and report the most favourable one. Variables affected by treatment also require particular care because adjusting for them can change the question being answered. ¹⁶

A report should state the direction and size of an association, not merely that variables are “associated”. ¹⁶

11.4.2.2 — Handling dropouts or missing data

Suppose Week 12 blood-pressure measurements are missing for 5% of the placebo group and 25% of the treatment group. ¹⁸

The meaning depends partly on why measurements are missing. Participants who leave because treatment is ineffective may have different outcomes from those who miss a visit because of travel. ¹⁸ E3 names the duration of the study, the nature of the disease, the efficacy and toxicity of the drug under study, and factors unrelated to therapy among the factors that may affect dropout rates. ¹⁹

E3 asks for examination of:

• Reasons for dropout.
• Timing.
• Proportions affected in each group over time.
• Potential effects on the conclusions.
• Methods used to estimate or derive missing values.
• The assumptions underlying those methods.

^{18, 19}

Analysing only completers can be misleading. ^{18, 19} E3 asks for the results to be assessed not only for the participants who completed the study but also for the entire population as randomised, or at least for all those with any on-study measurements. ¹⁹ Including people through a statistical model also does not automatically eliminate bias. ¹⁸ E3 warns that a large number of dropouts may introduce bias even when they are included in an analysis, particularly if there are more early dropouts in one treatment group or the reasons for dropping out are related to treatment or outcome. ¹⁹

E3 asks that the possible effects of early dropouts be explored as fully as possible, although they, and sometimes even the direction of the bias, can be difficult to determine. It suggests examining the observed cases at various time points or, where dropouts were very frequent, concentrating on time points when most participants were still under observation and the full effect of the drug had been realised, and it suggests modelling approaches to the incomplete data. ¹⁹

Imputation means supplying estimated values for missing observations. Other approaches model the observed data without filling every missing cell explicitly. Both depend on assumptions. ¹⁸

Stopping treatment is not the same as withdrawing from the study: a participant who stops medication may still provide the outcome measurements the estimand requires. Rescue medication likewise raises a question about which effect is being estimated, even when every measurement is available. ¹⁸

ICH E9(R1) connects the chosen estimand to the analysis assumptions. Under it, a sensitivity analysis tests whether conclusions remain stable under alternative assumptions while targeting the same estimand; an analysis answering a different question is supplementary rather than a sensitivity analysis in that strict sense. ¹⁸

11.4.2.3 — Interim analyses and data monitoring

The report needs to account for how the interim analyses planned in 9.7.1 were actually carried out. ²⁰

Examining accumulating results can help determine whether a trial should stop, but repeatedly looking for a positive result also creates more opportunities for a misleading finding. ²⁰

The type I error control described in 9.7.2 needs to account for relevant repeated looks at the data. ²⁰

The issue is both statistical and operational: seeing emerging results can influence subsequent decisions or behaviour. ²⁰ E3 states that examining and analysing accumulating data, formally or informally, can introduce bias and/or increase type I error. ²¹ It asks for every interim analysis to be described in full even if the treatment groups were not identified, and for data monitoring without code-breaking to be described even if it is considered to cause no increase in type I error. ²¹

E3 asks for disclosure of:

• Formal and informal analyses.
• Planned and unplanned analyses.
• Analyses by investigators, sponsor personnel, or monitoring groups.
• What information the reviewers examine and whether they know treatment identities.
• Procedures and any necessary statistical adjustments.
• Monitoring that remained blinded.

^{20, 21}

E3 holds that the minutes of meetings of any data monitoring group, and the data reports reviewed at them, may be helpful, particularly for a meeting that led to a change in the protocol or early termination of the study, and asks for them to be provided in 16.1.9. ²¹

11.4.2.4 — Multicentre studies

For the multicentre design introduced in 9.1, the report examines whether results differ meaningfully across sites. ²² E3 means by a multicentre study a single study under a common protocol whose centres’ data are intended to be analysed as a whole, as opposed to a post-hoc decision to combine data or results from separate studies. ²³

The main analysis usually combines the centres’ data. ^{22, 23} E3 nevertheless asks for individual centre results to be presented where appropriate, for example when the centres have enough patients to make such analysis potentially valuable. ^{23, 22}

A treatment-by-centre interaction means the treatment effect differs across centres:

• Quantitative interaction: benefit points in the same direction but differs in size.
• Qualitative interaction: the direction reverses—for example, benefit at one centre and worse outcomes at another.

²²

E3 asks for the possibility of either kind of interaction to be explored. ²³

Extreme or opposite results need discussion. ^{22, 23} Possible explanations include patient characteristics, clinical setting, study conduct, or random variation. ^{22, 23}

Relevant analyses should account for centre differences, with baseline and other data shown by centre where appropriate. ^{22, 23} Very small centres may provide little reliable evidence on their own. ²²

11.4.2.5 — Multiple comparisons / multiplicity

Multiplicity means having several opportunities to produce a statistically significant result. ²⁴ False positive findings increase in number as the number of significance tests performed increases. ²⁵

It can arise from:

• Multiple primary endpoints.
• Several doses or treatment groups.
• Repeated analyses of an endpoint.
• Multiple time points.
• Many subgroup comparisons.

²⁴

For illustration, if 20 independent tests each use a 5% false-positive threshold and all null hypotheses are true, the chance of at least one false positive is about 64%. ²⁴

Trials therefore may use procedures such as a prespecified testing order or adjusted significance thresholds. ²⁴ E3 asks the report to describe the adjustment or explain why one is unnecessary. ^{24, 25}

A nominal p-value is one not adjusted for the relevant multiplicity. A value below 0.05 does not necessarily satisfy the study’s formal success criterion. ²⁴

11.4.2.6 — Use of an “efficacy subset”

Section 11.4.2.6 returns to the participant exclusions of 11.1, now for their effect on the result. ²⁶

If the broad analysis shows a 2-unit benefit but the per-protocol subset shows an 8-unit benefit, the report needs to explain the discrepancy rather than simply favouring the larger estimate. ²⁶

E3 highlights exclusions for:

• Poor compliance.
• Missed visits.
• Ineligibility.
• Other reasons despite available data.

^{26, 27}

It asks for a broad analysis and explicit discussion of substantial differences between analysis populations. ^{26, 27}

Robustness means that a conclusion remains reasonably stable under relevant alternative assumptions or choices. Agreement between analyses can be reassuring; disagreement is itself a finding requiring interpretation. ²⁶ E3 considers it advantageous to demonstrate that the principal conclusions are robust to alternative choices of the population analysed. ²⁷

11.4.2.7 — Active-control studies intended to show equivalence

The design discussion in 9.2 defines active control, superiority, non-inferiority, equivalence, and the justified margin, and Section 11.4.2.7 compares the observed confidence interval with that margin. ²⁸

Where higher response rates are better and the non-inferiority margin is −10 percentage points:

• Difference of −2 points, CI −6 to +2: the interval excludes a loss of 10 points.
• Difference of −2 points, CI −15 to +11: it does not.

²⁸

The same point estimate leads to different conclusions because precision differs. ²⁸

Failure to demonstrate superiority does not establish equivalence or non-inferiority. ²⁸ The trial needs the appropriate design, margin, analysis, and evidence supporting the active-control comparison. ²⁸ When an active-control study is intended to show equivalence, meaning the lack of a difference greater than a specified size between the test drug and the active comparator, E3 asks the analysis to show the confidence interval for the comparison on the critical endpoints and its relation to the prespecified degree of inferiority that would be considered unacceptable, and it refers back to its Section 9.2 for important considerations in using the active-control equivalence design. ²⁹

11.4.2.8 — Examination of subgroups

The subgroup analyses specified in 9.7.1 examine whether response differs across relevant participant categories. ³⁰

E3 asks for examination of important demographic or baseline-defined subgroups when the study is large enough, including:

• Age, sex, and race.
• Baseline disease severity.
• Prognostic categories.
• Previous treatment with the same drug class.

^{30, 31}

E3 asks for these subgroups to be examined for unusually large or small responses and the results presented. ³¹ It asks the report to state when the study is too small for useful analyses. ^{30, 31}

E3 also asks the report to describe the demographic and other potentially predictive characteristics of the study population and, where the study is large enough to permit this, to present data for demographic subgroups, such as age, sex, race and weight, and for other subgroups, such as renal or hepatic function, so that possible differences in efficacy or safety can be identified. Usually, however, subgroup responses should be examined in the larger database used in the overall analysis. ³²

Two cautions apply:

1. A favourable subgroup should not be used to rescue an otherwise unsupported overall result. ^{30, 31}
2. A difference being statistically significant in one subgroup and non-significant in another does not itself demonstrate that the treatment effects differ. ³⁰

Whether the treatment effects differ is a question of treatment-by-subgroup interaction. ³⁰

Where there is a prior hypothesis of a differential effect in a particular subgroup, E3 asks for that hypothesis and its assessment to be part of the planned statistical analysis. ^{31, 30} Subgroup analyses, and unexpected findings in particular, may suggest hypotheses worth examining in other studies, or help refine labelling information, patient selection or dose selection. ^{30, 31}

Documentation of the statistical analysis

Annex VIII of E3 guides Section 11.4.2 and Appendix 16.1.9. It asks for the details of the statistical analysis of each primary efficacy variable to be presented in an appendix, including at least:

• a) The statistical model underlying the analysis, presented precisely and completely, using references if necessary.
• b) A statement of the clinical claim tested in precise statistical terms, such as null and alternative hypotheses.
• c) The statistical methods applied to estimate effects and construct confidence intervals, with literature references where appropriate.
• d) The assumptions underlying the statistical methods. Insofar as is statistically reasonable, the documentation should show that the data satisfy crucial assumptions, especially when that is necessary to confirm the validity of an inference. When the applicant has performed extensive statistical analyses, it is essential to consider how far they were planned before the data were available and, where they were not, how bias was avoided in choosing the analysis the conclusions rest on. This matters particularly for subgroup analyses, which, if not preplanned, ordinarily provide no adequate basis for definitive conclusions.
  • (i) A data transformation needs a rationale for its choice and an interpretation of the treatment effects estimated on the transformed data.
  • (ii) A discussion of the appropriateness of the chosen statistical procedure and of the validity of the statistical conclusions guides the regulatory authority's statistical reviewer in deciding whether the data need reanalysis.
• e) The test statistic, its sampling distribution under the null hypothesis, its value, the significance level (p-value), and intermediate summary data, in a format that lets the regulatory authority's statistical reviewer verify the results quickly and easily. The p-values are designated as one- or two-tailed, and a one-tailed test needs its rationale.

³³

For example, the documentation of a two-sample t-test gives the t-statistic, its degrees of freedom, the p-value, the two sample sizes, the mean and variance of each sample and the pooled variance. Multicentre studies analysed by analysis of variance need at least a table with terms for centres, treatments, their interaction, error and total, and crossover designs need sequences, patients within sequences, period baselines, washouts and their length, dropouts in each period, treatments, periods, treatment-by-period interaction, error and total. For each source of variation other than the total, the table gives the degrees of freedom, sum of squares, mean square, F-test, p-value and expected mean square. ³³ The intermediate summary data display the demographic and response data, averaged or otherwise summarised, for each centre-by-treatment combination, or other design characteristic such as sequence, at each observation time. ³³

Annex VIII also calls for listings of the participant data used in the statistical analyses, which the sponsor may be asked to supply in a computer-readable form. ³⁴

11.4.3 — Tabulation of individual response data

Group averages must be traceable to individual observations. ³⁵

E3 asks for participant-level tables containing relevant information such as:

• Participant identifier.
• All critical measurements, including baseline.
• Timing, including time of day where relevant.
• Treatment and dose at measurement.
• Compliance information.
• Concomitant medicines around the assessment.
• Overall responder/non-responder classifications.
• Which efficacy analyses included the participant.
• Relevant baseline characteristics and disease severity.
• CRF location, if the case report form is supplied.

^{35, 36}

The baseline values of critical measurements are ordinarily included as zero-time values for each efficacy measurement. ³⁶

These basic efficacy listings ordinarily belong in 16.2.6. ^{35, 36} Some regulatory authorities may require all individual data in archival case report tabulations. ³⁶

The report should distinguish material included in its appendices from more extensive archival tabulations or information available on request. ^{35, 36} What needs to be included varies from study to study and from one drug class to another, and E3 leaves the applicant to decide, if possible after consultation with the regulatory authority, what goes into the appendix. ³⁶

Because complete listings can be unwieldy, E3 also encourages focused displays—for example, one row per participant showing the most important visits. ^{35, 36}

11.4.4 — Drug dose, drug concentration, and relationships to response

Section 11.4.4 asks whether benefit changes with the amount of medicine given or the amount present in the body. ³⁷

The PK measurements in 9.5.4 link administered dose to measured concentration. The body-size units in Section 4 support appropriately scaled dose comparisons. ³⁷

E3 asks for:

• Actual doses, particularly when they vary within participants.
• Individual dose tabulations.
• Examination of available dose–response information.
• Body-size adjustment, such as mg/kg or mg/m², where helpful.
• Concentration tables in 16.2.5, PK analysis, and relationships to response where possible.

^{37, 38}

It acknowledges that a study not designed to investigate dose response may provide limited evidence. ^{37, 38}

For example, if clinicians increase the dose in poor responders, the highest-dose group may have the worst outcomes because poor response prompts escalation. The observed association alone does not show that higher doses reduce efficacy. ³⁷

For the design and analysis of studies that explore dose response or concentration response, E3 points to the ICH guideline Dose-Response Information to Support Drug Registration. ³⁸

11.4.5 — Drug–drug and drug–disease interactions

Section 11.4.5 asks whether response appears related to:

• Other treatment taken alongside the study drug.
• Previous or concurrent illnesses.

^{39, 40}

The drug–drug interactions introduced in 9.4.7 concern other medicines. A drug–disease interaction concerns an illness modifying the response to treatment. ³⁹

For example, impaired kidney function may alter drug exposure; another medicine may independently improve the same endpoint. ³⁹

E3 asks for apparent relationships to be described. ^{39, 40} An association within the trial may justify further investigation without establishing a causal interaction. ³⁹

11.4.6 — By-patient displays

Sometimes an individual timeline explains a pattern more clearly than a large listing. ^{41, 42}

Such a timeline might combine:

• A symptom score over time.
• Dose changes.
• Addition of another medicine.
• An adverse event.
• Treatment interruption or discontinuation.

⁴¹

The reader can then see, for example, whether improvement preceded or followed a dose increase. ⁴¹

E3 holds that such displays may offer little advantage where group mean data represent the principal analyses, but may help where overall evaluation of individual responses is a critical part of the analysis. ^{42, 41} They are not universally required. ⁴¹

11.4.7 — Efficacy conclusions

The efficacy conclusions should concisely integrate:

• Primary endpoint results.
• Secondary endpoint results.
• Prespecified analyses.
• Alternative analytical approaches.
• Exploratory findings.

^{43, 44}

They should preserve their different evidential roles. ⁴³

A fictional conclusion might read:

Treatment produces a greater reduction in blood pressure than placebo at Week 12. The estimated difference is 5 mmHg, with a 95% confidence interval of 2 to 8 mmHg. The primary analysis meets the prespecified success criterion, and sensitivity analyses support the finding. The exploratory age-subgroup analysis is inconclusive.

⁴³

Such a conclusion tells the reader the size of the benefit, uncertainty, formal result, robustness, and remaining limitation. ⁴³

Where the supporting material lives

Location	Material supporting Section 11
14.1	Demographic and baseline tables and figures
14.2	Efficacy tables, figures, and graphs outside the main text
16.1.9	Detailed statistical methods and relevant analytical documentation
16.2.3	Excluded participants, visits, or observations, with reasons
16.2.4	Relevant individual demographic, baseline, and concomitant-treatment data
16.2.5	Compliance and drug-concentration data
16.2.6	Individual efficacy response data

⁴⁵

Section 11 under CORE Reference

CORE Reference recommends moving data sets analysed, demographic and other baseline characteristics, and measurements of treatment compliance out of the start of Section 11 and into Section 10, as Sections 10.3, 10.4 and 10.5, so its Section 11.1 starts with the efficacy results. ^{46, 47} Its reasons are that the data sets analysed describe the subject population and so apply to more than the efficacy evaluation, demographic and baseline data relate to the safety as well as the efficacy population, and treatment compliance relates to all study subjects, not only those with efficacy data. ⁴⁷

Within Section 11, CORE Reference supplies subheadings for each potential type of analysis: primary, secondary and other, which may include exploratory and post-hoc analyses. Where pharmacokinetics was the primary endpoint, it suggests presenting the pharmacokinetic data in Section 11.1 and changing the titles accordingly. ⁴⁸

Section order in a phase I oncology report

The main aim of a phase I trial of an anticancer agent is to find the maximum tolerated dose and the recommended dose for phase II through a dose-escalation design that reports dose-limiting toxicities, so its main endpoints are all safety variables. Under the standard E3 order these results come in Section 12, after the efficacy results, and describing efficacy first needs many cross-references to the safety evaluation to explain how the cohorts were distributed. A more logical, reader-friendly way to show the data is to reverse the order and numbering of the two sections, presenting the safety evaluation as Section 11 and the efficacy evaluation as Section 12, and it is useful to add a note at the beginning of the report stating that the two sections were reversed and renumbered in accordance with the trial's primary and secondary objectives. ⁴⁹ Electronic submission with a hyperlinked table of contents may reduce the concern about efficacy appearing before safety, and by 2012 the reversed structure had been used in several reports submitted to the European Medicines Agency in marketing authorisation applications and accepted by the agency in validation. ⁴⁹