Protocol Design, Estimands & Study Design
The protocol turns strategy into an executable study: objectives and estimands, endpoints, sample size, eligibility, the assessment schedule, and the identification of what is genuinely critical to quality. It is the single document governing everything sites do, which is why complexity added here multiplies across every participant at every visit — and the burden falls on people who had no part in writing it.
What an explainer is not
A topic explainer is SPEQ’s synthesis of what a practice involves, cited to the standards that govern it. It does not reproduce their text, and it does not determine which of them apply to your product or process.
[ POSITION IN THE FRAMEWORK ]
7 DIMENSIONS · 20 LINKSThe estimand fixes what the trial is actually measuring before anyone sees data — including what happens when patients stop treatment — which is precisely the question a protocol written without one leaves to be argued afterwards.
06 · QUALITY MATURITY — PROTOCOL DESIGN, ESTIMANDS & STUDY DESIGN, REACTIVE TO ADAPTIVE
The protocol states an endpoint and an analysis. What happens to patients who discontinue is decided when it happens.
An estimand is described, but it was added to satisfy the template and the analysis does not follow from it.
The estimand is defined before the design — population, variable, intercurrent event strategy, summary measure — and the analysis and data collection follow from it.
Intercurrent events are anticipated in the operational design, so the data needed to handle them is actually collected rather than reconstructed.
Protocol, estimand, analysis and data collection are one argument, so a messy trial still answers the question it was designed to answer.
SPEQ’s shared five-stage progression, labelled synthesis — not the FDA QMM rating scale. Where does your organization sit? Score your quality system →
07 · REGULATORY & EVIDENCE
GOVERNING STANDARDS · 4
Derived from the 4 standards SPEQ maps to this subject, across 2 regulatory bodies: ICH, FDA.
RECORDS & OBJECTIVE EVIDENCE
- The protocol with the estimand stated in full, including intercurrent event strategies
- The statistical analysis plan, traceable to the estimand
- Data collection specifications supporting the intercurrent event strategies chosen
- Amendment history, with the reason and timing of each change
- Records of analysis decisions taken before database lock
COMMON INSPECTION FINDINGS
- An estimand stated in the protocol that the analysis plan does not implement
- Intercurrent event handling decided after the data was seen
- Data required by the chosen strategy not collected, so the estimand cannot be estimated
- Repeated amendments to the primary endpoint or analysis during conduct
- A sample size derived from a different question than the one the estimand asks
The estimand framework asks what is being estimated
ICH E9(R1) introduced the estimand: a precise description of the treatment effect the trial is estimating, specified through the population, the variable, the handling of intercurrent events, and the population-level summary. Its purpose is to close a gap that was previously filled by convention — two analyses of the same data could answer different questions without anyone stating which question was intended.
Intercurrent events are the load-bearing part: treatment discontinuation, rescue medication, death. How they are handled changes what the estimate means, and that decision belongs in the protocol as a scientific choice rather than in the statistical analysis plan as a technical one. Trials that specify the estimand well tend to have fewer arguments about the analysis afterwards, because the argument was had in advance.
Critical to quality factors, and the things that are not
ICH E6(R3) and E8(R1) both direct sponsors to identify the factors critical to the quality of the trial — the ones whose failure would undermine participant safety or the reliability of results — and to focus oversight there. That is a licence to stop treating every data point as equally important, and most organisations have not used it.
The tell is a protocol collecting data nobody has a plan to analyse. Every extra assessment adds site burden, participant burden and a new opportunity for a deviation, and unanalysed data carries all of that cost for no benefit. The discipline is to ask of each collected variable which analysis uses it, and to remove the ones with no answer.
Complexity is paid for by people who did not choose it
A protocol amendment adding one assessment adds it for every participant at every visit at every site. Eligibility criteria that are individually reasonable can combine into a population that barely exists. An assessment schedule designed around scientific ideal rather than clinical reality produces visit windows sites cannot meet, and the resulting deviations are recorded against the site.
Site and patient input during protocol design is the standard countermeasure and is still treated as optional. It is the cheapest available way to discover that a schedule is unworkable, and it is dramatically cheaper than discovering it through an amendment after activation.
SPEQ interpretation — the deviation profile is designed, not discovered
When a study reports high protocol-deviation rates, the response is usually operational: more monitoring, site retraining, corrective actions. But a large share of deviations in most studies trace to a small number of protocol features — an impractical visit window, an assessment requiring equipment sites do not have, an eligibility criterion sites read differently.
Those are design defects, and they were fixable before the study opened. Reviewing the deviation profile of previous studies against their protocol features, before writing the next one, is the highest-return protocol-quality activity available and one almost nobody performs — because deviation data lives in operations and protocol authorship lives in development.
FREQUENTLY ASKED
What is an estimand?
A precise description of the treatment effect a trial is estimating, specified through the population, the variable, the handling of intercurrent events, and the population-level summary. ICH E9(R1) introduced it to close the gap where two analyses of the same data could answer different questions without either being stated.
Why do intercurrent events matter so much?
Because how treatment discontinuation, rescue medication or death are handled changes what the estimate means. That is a scientific choice belonging in the protocol, not a technical one to be settled in the statistical analysis plan — and specifying it in advance prevents the argument from happening after unblinding.
How do you decide what data to collect?
Ask which analysis uses each variable and remove the ones with no answer. ICH E6(R3) and E8(R1) direct sponsors to identify the factors critical to trial quality and focus there, which is explicit permission to stop treating every data point as equally important. Unanalysed data costs site burden, participant burden and deviation opportunities for no benefit.
Where do protocol deviations actually come from?
Disproportionately from a small number of protocol features — an impractical visit window, an assessment needing equipment sites lack, an eligibility criterion read inconsistently. Those are design defects fixable before activation, and reviewing prior studies’ deviation profiles against their protocol features is the highest-return protocol-quality activity available.