Study Design and Protocol Development
This domain asks whether you can read a protocol critically: is the question answerable, does the endpoint actually measure the objective, and is the design protected against bias? Questions frequently hinge on the difference between an objective and an endpoint, on why a particular control or blinding level was chosen, and on whether the sample size justification is coherent.
Objectives and endpoints
- An objective is the question; an endpoint is the variable that answers it. A primary endpoint must be pre-specified, validated or well justified, and measurable within the visit schedule.
- Secondary and exploratory endpoints do not drive the sample size but still require a multiplicity strategy; otherwise positive findings are hypothesis-generating only.
Design elements
- Randomisation: simple, permuted blocks, stratified, or adaptive; allocation must be concealed until assignment (sequentially numbered kits, central randomisation).
- Blinding: open-label, single, double, triple. Blinding protects against performance and detection bias; open-label designs need objective endpoints or blinded adjudication.
- Controls: placebo, active comparator, historical (weakest, most bias-prone).
- Analysis populations: intention-to-treat (all randomised), per-protocol, and safety set must be defined before unblinding.
Required protocol content (ICH E6 §6)
Background and rationale, objectives and endpoints, design and flow diagram, inclusion/exclusion criteria, treatments and accountability, efficacy and safety assessments, statistics (sample size justification, analysis populations, handling of missing data), ethics, data handling, quality control, and stopping rules or independent data monitoring committee arrangements.
Worked example
A proposal offers 40 participants total to detect a 10 mmHg difference with a standard deviation of 15, two-sided alpha 0.05, 80% power. Reasoning: for a two-arm parallel design, n per arm ≈ 2 × (1.96 + 0.84)² × 15² / 10² = 2 × 7.85 × 225/100 ≈ 35, so about 36 per arm and 72 total. Forty is underpowered; the protocol must either justify the smaller effect size or raise enrolment, and the time-to-event or dropout inflation must be shown explicitly.
Traps that cost marks
- Quoting a primary objective when the question asks for the endpoint (or vice versa).
- Forgetting dropout inflation, so the "completer" number looks adequate but randomised enrolment is not.
- Treating ITT as "everyone who finished" rather than everyone randomised.
- Assuming double-blinding is always feasible, or that a placebo arm is always ethical.
- Writing a protocol with no pre-specified stopping rule or multiplicity control.