Why Science? | ICH E4 Dose-Response Information to Support Drug Registration

A dose-response relationship describes how an administered amount is associated with a beneficial or harmful outcome. E4 treats this relationship as central evidence because a medicine’s effect is rarely captured by a single yes-or-no comparison. The curve can identify onset, steep change, a plateau and a region where toxicity rises.

The useful question in dose-response connects amount with outcome is not simply whether a medicine works. A dose-response relationship describes how an administered amount is associated with a beneficial or harmful outcome. E4 treats this relationship as central evidence because a medicine’s effect is rarely captured by a single yes-or-no comparison. The curve can identify onset, steep change, a plateau and a region where toxicity rises. Keep the curve inside this boundary: Association across assigned doses still needs sound randomisation, measurement and adherence to support a causal interpretation.

A strong science note separates observation from interpretation. Draw separate benefit and harm curves and label the dose range where the balance appears most favourable. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Next chapter

FDA identifies E4 as final guidance dated July 1996 under docket FDA-1993-D-0183 and issued by CDER. Its purpose is to help obtain dose-response information for regulatory decisions, including the starting dose, dose steps, highest recommended dose and approaches to individualisation.

For the july 1996 e4 guidance is final, dose is an instruction while exposure is the biology that follows. FDA identifies E4 as final guidance dated July 1996 under docket FDA-1993-D-0183 and issued by CDER. Its purpose is to help obtain dose-response information for regulatory decisions, including the starting dose, dose steps, highest recommended dose and approaches to individualisation. Preserve this guardrail: General ICH guidance does not replace product-specific evidence or current legal requirements.

A strong science note separates observation from interpretation. Create a source receipt with the official title, status, date, docket, centre, purpose and scope boundary. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

The dose that moves an efficacy endpoint may not be the dose at which an adverse effect appears. E4 encourages examining both desirable and undesirable responses across dose. Looking at only the efficacy curve can conceal a narrowing safety margin; looking only at adverse events can miss a meaningful gain at modest exposure.

A clear explanation of benefit and harm can follow different curves puts benefit and harm on the same decision map. The dose that moves an efficacy endpoint may not be the dose at which an adverse effect appears. E4 encourages examining both desirable and undesirable responses across dose. Looking at only the efficacy curve can conceal a narrowing safety margin; looking only at adverse events can miss a meaningful gain at modest exposure. The honest limit is: An average benefit-risk picture can hide people with unusually high exposure or vulnerability.

A strong science note separates observation from interpretation. Place efficacy and safety estimates with uncertainty on the same dose grid, without forcing them onto one invented score. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

A starting dose should be tolerable for people beginning treatment, including those who may be sensitive. A target or usual dose reflects the range expected to deliver the intended benefit for many patients. Titration can connect them, but the label should make the rationale and steps clear rather than implying one amount suits everyone.

The useful question in starting dose and target dose answer different questions is not simply whether a medicine works. A starting dose should be tolerable for people beginning treatment, including those who may be sensitive. A target or usual dose reflects the range expected to deliver the intended benefit for many patients. Titration can connect them, but the label should make the rationale and steps clear rather than implying one amount suits everyone. Keep the curve inside this boundary: A cautious starting dose does not prove that later escalation is safe for every patient.

A strong science note separates observation from interpretation. Write separate rationales for starting, target and maximum doses and link each to the evidence it uses. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

In a fixed-dose parallel study, participants are randomised to remain on one of several doses, often with placebo or an active control. Separation between assigned doses can reveal the population dose-response curve with less confounding by response-driven titration. Enough spacing and adequate duration are needed to see meaningful differences.

For parallel fixed-dose studies reveal the shape directly, dose is an instruction while exposure is the biology that follows. In a fixed-dose parallel study, participants are randomised to remain on one of several doses, often with placebo or an active control. Separation between assigned doses can reveal the population dose-response curve with less confounding by response-driven titration. Enough spacing and adequate duration are needed to see meaningful differences. Preserve this guardrail: A flat result can reflect poor dose separation, nonadherence, insensitive endpoints or a real plateau.

A strong science note separates observation from interpretation. Sketch the planned doses and explain what biological feature each comparison is intended to reveal. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

A titration design adjusts dose according to response, tolerability or a preset schedule. It can show how a practical regimen works and may protect participants from unnecessary exposure. Because dose after titration depends on prior outcome, comparisons among final-dose groups can mix drug effect with patient characteristics and clinical decisions.

A clear explanation of titration studies resemble care but complicate inference puts benefit and harm on the same decision map. A titration design adjusts dose according to response, tolerability or a preset schedule. It can show how a practical regimen works and may protect participants from unnecessary exposure. Because dose after titration depends on prior outcome, comparisons among final-dose groups can mix drug effect with patient characteristics and clinical decisions. The honest limit is: Patients who finish at the same dose may have reached it for very different reasons.

A strong science note separates observation from interpretation. Map every titration rule and identify the response that causes movement up, down or out of treatment. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

When a condition is stable, effects are reversible and carryover can be controlled, participants may receive multiple doses in different periods. Within-person comparison can reduce noise from stable individual differences. Washout, period effects and dropout need careful handling, especially when high doses cause more discontinuation.

The useful question in crossover designs can compare doses within a person is not simply whether a medicine works. When a condition is stable, effects are reversible and carryover can be controlled, participants may receive multiple doses in different periods. Within-person comparison can reduce noise from stable individual differences. Washout, period effects and dropout need careful handling, especially when high doses cause more discontinuation. Keep the curve inside this boundary: A crossover design is unsuitable when effects persist, disease changes quickly or prior treatment alters later response.

A strong science note separates observation from interpretation. Create a timeline showing treatment order, washout, measurement windows and the carryover assumption. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

A factorial study varies two treatments or components so main effects and interactions can be examined. For combination therapy, it can help show how each component contributes and whether one changes the dose-response of the other. Efficiency depends on plausible assumptions, adequate sample size and interpretable interactions.

For factorial designs can study combinations efficiently, dose is an instruction while exposure is the biology that follows. A factorial study varies two treatments or components so main effects and interactions can be examined. For combination therapy, it can help show how each component contributes and whether one changes the dose-response of the other. Efficiency depends on plausible assumptions, adequate sample size and interpretable interactions. Preserve this guardrail: A strong interaction can make a simple average main effect misleading.

A strong science note separates observation from interpretation. Draw every treatment cell and state the contribution and interaction question attached to each comparison. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

The same administered dose can produce different blood concentrations because absorption, distribution, metabolism, elimination and adherence vary. Exposure-response analysis links concentration or a summary such as area under the curve with efficacy or safety, helping explain why dose alone may be an incomplete predictor.

A clear explanation of concentration-response explains variable exposure puts benefit and harm on the same decision map. The same administered dose can produce different blood concentrations because absorption, distribution, metabolism, elimination and adherence vary. Exposure-response analysis links concentration or a summary such as area under the curve with efficacy or safety, helping explain why dose alone may be an incomplete predictor. The honest limit is: Concentration measured after treatment begins can be influenced by timing, adherence and illness, creating bias if the analysis ignores them.

A strong science note separates observation from interpretation. Trace dose to concentration to outcome and mark every measurement time and source of variation. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

A pharmacodynamic measure records a biological response to exposure, such as receptor activity, blood pressure or a laboratory marker. It can help detect an active range or compare populations before a slower clinical endpoint is available. Its meaning depends on how closely it relates to the outcome that matters to patients.

The useful question in pharmacodynamic markers can bridge mechanism and benefit is not simply whether a medicine works. A pharmacodynamic measure records a biological response to exposure, such as receptor activity, blood pressure or a laboratory marker. It can help detect an active range or compare populations before a slower clinical endpoint is available. Its meaning depends on how closely it relates to the outcome that matters to patients. Keep the curve inside this boundary: A responsive biomarker is not automatically a validated substitute for clinical benefit.

A strong science note separates observation from interpretation. Write the causal chain from dose through marker to patient outcome and identify the least certain link. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Population pharmacokinetic and exposure-response models can examine how body size, age, organ function, genotype, interacting drugs or other factors relate to exposure and response. E4’s individualisation goal becomes stronger when covariates have a plausible mechanism and are supported by adequate data across their ranges.

For population analyses can reveal influential covariates, dose is an instruction while exposure is the biology that follows. Population pharmacokinetic and exposure-response models can examine how body size, age, organ function, genotype, interacting drugs or other factors relate to exposure and response. E4’s individualisation goal becomes stronger when covariates have a plausible mechanism and are supported by adequate data across their ranges. Preserve this guardrail: A model-selected covariate may be unstable, confounded or too small to justify a dosing rule.

A strong science note separates observation from interpretation. For each proposed covariate, record mechanism, data coverage, effect size, uncertainty and possible action. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Children, older adults and people with renal or hepatic impairment may reach different exposure at the same dose or tolerate effects differently. Dedicated studies, modelling or subgroup evidence can support an adjustment, depending on risk and uncertainty. A subgroup label should state the exact population and evidence rather than extending by analogy.

A clear explanation of special populations may need different dose information puts benefit and harm on the same decision map. Children, older adults and people with renal or hepatic impairment may reach different exposure at the same dose or tolerate effects differently. Dedicated studies, modelling or subgroup evidence can support an adjustment, depending on risk and uncertainty. A subgroup label should state the exact population and evidence rather than extending by analogy. The honest limit is: Absence of a statistically significant interaction does not prove equal response in a small subgroup.

A strong science note separates observation from interpretation. Compare exposure, efficacy and safety evidence for the main and special populations in parallel columns. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Dose individualisation can use measured response, adverse effects, concentration, organ function or another patient feature. The rule should specify what is measured, when, the action threshold and the next dose. A vague instruction to adjust “as needed” transfers uncertainty to clinicians without explaining how evidence supports the change.

The useful question in individualisation needs a usable decision rule is not simply whether a medicine works. Dose individualisation can use measured response, adverse effects, concentration, organ function or another patient feature. The rule should specify what is measured, when, the action threshold and the next dose. A vague instruction to adjust “as needed” transfers uncertainty to clinicians without explaining how evidence supports the change. Keep the curve inside this boundary: A precise rule is only as reliable as the measurement and evidence behind its threshold.

A strong science note separates observation from interpretation. Turn one proposed adjustment into an if-then rule and test it on three contrasting patient scenarios. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Some responses rise nearly linearly over the studied range; others are steep near a threshold, then plateau. A U-shaped or delayed pattern is also possible. Study doses should cover the range needed to distinguish plausible shapes, and analysis should not assume a convenient straight line without checking alternatives.

For curve shape changes the meaning of a dose difference, dose is an instruction while exposure is the biology that follows. Some responses rise nearly linearly over the studied range; others are steep near a threshold, then plateau. A U-shaped or delayed pattern is also possible. Study doses should cover the range needed to distinguish plausible shapes, and analysis should not assume a convenient straight line without checking alternatives. Preserve this guardrail: Sparse doses can make very different biological curves appear equally compatible with the data.

A strong science note separates observation from interpretation. Fit or sketch several plausible shapes and name the additional dose or time point that would separate them. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Adherence, formulation, food, timing, assay performance, rescue treatment and dropout can blur exposure differences. A well-spaced protocol may produce overlapping actual concentrations. E4 therefore connects dose selection with careful conduct and with concentration information when it improves interpretation.

A clear explanation of dose separation depends on conduct and measurement puts benefit and harm on the same decision map. Adherence, formulation, food, timing, assay performance, rescue treatment and dropout can blur exposure differences. A well-spaced protocol may produce overlapping actual concentrations. E4 therefore connects dose selection with careful conduct and with concentration information when it improves interpretation. The honest limit is: Randomisation to distinct nominal doses does not guarantee distinct biological exposure.

A strong science note separates observation from interpretation. Audit assigned dose, taken dose and measured exposure as three separate variables. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Earlier dose-ranging evidence should support the dose or doses taken into confirmatory trials. Selecting only the largest tolerated dose wastes information if lower doses approach the efficacy plateau. Carrying more than one dose can clarify benefit-risk, but endpoints and multiplicity should be planned before results are known.

The useful question in confirmatory trials should test a justified regimen is not simply whether a medicine works. Earlier dose-ranging evidence should support the dose or doses taken into confirmatory trials. Selecting only the largest tolerated dose wastes information if lower doses approach the efficacy plateau. Carrying more than one dose can clarify benefit-risk, but endpoints and multiplicity should be planned before results are known. Keep the curve inside this boundary: A successful confirmatory comparison cannot recover the dose-response information that was never collected.

A strong science note separates observation from interpretation. Write the evidence chain from early curve to selected regimen and the unanswered question retained for confirmation. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

The receipt records intended population, formulation, doses, titration, exposure measures, efficacy and safety endpoints, curve models, covariates, adherence, missingness, subgroup evidence, selected starting and target doses, maximum dose, uncertainty and sources. It shows why the chosen regimen is reasonable and where future learning may change it.

For a complete e4 receipt keeps the dose decision auditable, dose is an instruction while exposure is the biology that follows. The receipt records intended population, formulation, doses, titration, exposure measures, efficacy and safety endpoints, curve models, covariates, adherence, missingness, subgroup evidence, selected starting and target doses, maximum dose, uncertainty and sources. It shows why the chosen regimen is reasonable and where future learning may change it. Preserve this guardrail: A complete record makes uncertainty visible; it does not turn a weak dose-ranging programme into strong evidence.

A strong science note separates observation from interpretation. Ask an independent reader to reproduce the dose recommendation and identify the assumption that matters most. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter