Before any treatment is given, a participant travels a defined path — recruitment → screening → consent → eligibility → enrolment → randomisation → allocation. The CONSORT flow diagram captures this journey in four stages, and three concepts on the way (randomisation, allocation concealment, blinding) are the ones examiners most often confuse you with.
Screening
Screening evaluates potential participants against the inclusion and exclusion criteria to decide eligibility.
- Informed consent must be obtained BEFORE any screening procedure (tests, investigations) — a GCP requirement.
- Each candidate gets a screening number; a screening log is kept (an essential document).
- Screen failure = a candidate who does not meet eligibility; the reason is documented. Re-screening may be allowed per protocol.
- A washout period may precede enrolment to clear prohibited prior medication.
Eligibility criteria
- Inclusion criteria — characteristics a participant MUST have (age, confirmed diagnosis, disease severity).
- Exclusion criteria — characteristics that DISQUALIFY (pregnancy, comorbidity, contraindication, prohibited concomitant drugs).
- Purpose — ensure scientific validity, participant safety and homogeneity, while balancing generalisability.
Enrolment
Enrolment registers an eligible, consented participant into the trial — at which point they become a study subject. It requires signed consent AND confirmed eligibility; a subject identification code is assigned and a subject enrolment log kept (essential documents). Baseline assessments are recorded before randomisation.
Randomisation — what and why
Randomisation allocates participants to treatment groups purely by chance, so each has a known (often equal) probability of assignment. Why randomise?
- Eliminates selection (allocation) bias.
- Balances known and unknown confounders between groups (comparability).
- Provides the basis for valid statistical inference.
The allocation ratio is usually 1:1 but may be unequal (e.g. 2:1).
Methods of randomisation
| Method | Description |
|---|---|
| Simple | Coin toss / random-number table / computer; can cause group-size imbalance in small trials. |
| Block (permuted) | Randomise within blocks so group sizes stay balanced; block size kept unknown to prevent prediction. |
| Stratified | Randomise within strata of prognostic factors (centre, age, severity); usually combined with blocks. |
| Cluster | Randomise groups/clusters (hospitals, villages) rather than individuals. |
| Minimisation | Covariate-adaptive — allocate each new subject to minimise imbalance in chosen prognostic factors. |
| Response-adaptive | Allocation probabilities shift toward the better-performing arm (e.g. play-the-winner). |
Allocation concealment (≠ blinding)
Allocation concealment prevents foreknowledge of the next treatment assignment, so whoever enrols a participant cannot predict or influence the group they receive. Methods: central randomisation (web/voice — IWRS/IVRS), SNOSE (sequentially numbered opaque sealed envelopes), and pharmacy-controlled allocation. Poor concealment exaggerates treatment effects by introducing selection bias.
| Concept | What it does | Stage | Bias prevented |
|---|---|---|---|
| Randomisation | Assigns to groups by chance | Sequence generation | Selection & confounding |
| Allocation concealment | Hides the next assignment until enrolment | Up to allocation | Selection |
| Blinding | Hides treatment identity after allocation | Treatment & assessment | Performance & detection |
Exam tip: consent before screening; screen failure is documented. Randomisation removes selection bias and balances confounders; block keeps groups equal, stratified balances prognostic factors. Allocation concealment ≠ blinding — concealment protects the assignment up to allocation; blinding hides identity after. Report an RCT with CONSORT; analyse by ITT.
Picture one subject moving through the CONSORT funnel: consented, screened, found eligible, enrolled, then allocated by chance behind a concealed sequence and followed under blinding to an intention-to-treat analysis. Hold that picture and you can place every term — randomisation, concealment, blinding — exactly where it belongs.