Randomised Controlled Trials: Randomisation, Allocation Concealment, Blinding and Trial Registration

Hammad Claims His Territory

Dr. Hammad Ali did not walk into the Research Room. He arrived.

He had a printed draft in a plastic sleeve, the first time anyone in the group had brought a synopsis in a plastic sleeve, and he placed it on the table in front of me as if it were a lab report from the Mayo Clinic.

“Sir, last week you said that the moment a researcher assigns the exposure, it becomes an experiment. And that it was my territory.” He tapped the sleeve. “So I have claimed my territory.”

The title read: “A Randomized Controlled Double Blind Trial Comparing Low Pressure and Standard Pressure Pneumoperitoneum for Shoulder Tip Pain After Laparoscopic Cholecystectomy.”

“Read me your method,” I said.

He cleared his throat. “Patients admitted on odd dates will get low pressure, ten millimetres of mercury. Patients admitted on even dates will get standard pressure, fourteen. That way it is random, because nobody chooses the date. I will see every patient at 24 hours and ask about shoulder-tip pain on a zero to ten scale. If the consultant has to raise the pressure during surgery because he cannot see properly, I will exclude that patient, because he did not really get low pressure. And it is double blind.”

“Who is blind?”

He paused. “The… patient? And…” The pause grew longer. “Double means two, Sir. I thought two was the standard number.”

Dr. Sumaira Talib looked at the ceiling. Dr. Junaid Rashid put his pen down to listen properly. Even Dr. Hassan Raza put his phone away.

“Hammad,” I said, “you have written four sentences, and every one of them contains a mistake that has sunk a published trial somewhere in the world. That is actually very useful. It means tonight we can learn the whole design from your page.”

He looked at the plastic sleeve as if it had betrayed him.

What Makes a Trial a Trial

I wrote one line on the whiteboard.

In an observational study, life decides who is exposed. In a trial, the researcher decides.

“In the cohort session,” I said, “Bushra could only watch which mothers received steroids. Her biggest problem was confounding: the exposed and unexposed groups might differ in a hundred ways she never measured. A randomised controlled trial solves that problem in one move. If chance alone decides who gets which treatment, then age, weight, gallbladder inflammation, pain threshold and every factor you never thought of are shared out evenly between the groups, on average. Any difference at the end can then be put down to the treatment.”

“That is why the trial sits so high in the evidence pyramid,” Junaid said. “We discussed it in the EBM session.”

“Exactly. And notice what the definition requires. The International Committee of Medical Journal Editors, the ICMJE, calls any study that prospectively assigns people to an intervention to study its effect on a health outcome a clinical trial [1]. Drugs, yes. But also a surgical technique, a dressing, a counselling session or a change in ward routine. Hammad’s pressure setting is a clinical trial, with every obligation that comes with the name.”

Hammad sat up a little. He liked the word “obligation” less than the word “territory”, but he wrote it down.

Odd and Even Dates Are Not Random

“Now your first sentence,” I said. “Odd dates and even dates.”

“Nobody chooses the date, Sir.”

“No. But everybody knows it. On the 14th, the ward knows the next patient will get standard pressure. Suppose the consultant has a difficult, inflamed gallbladder on the list and feels he needs a good view. Will he quietly move that patient to an even date? Suppose a registrar thinks low pressure is better and sends anxious patients on odd dates. Nobody has to be dishonest. People simply nudge, because they can see what is coming.”

Sumaira finished the thought. “So the groups are no longer chosen by chance. They are chosen by the people who know the schedule.”

“That is selection bias. Alternate days, odd and even registration numbers, days of the week and bed numbers are all called quasi-random allocation. They look like randomisation, but they are predictable. Real randomisation has two parts, and reviewers check both.”

Part one: generate an unpredictable sequence.

• Simple randomisation: a computer-generated list, or a coin toss for each patient. Fair, but in a small trial the groups can end up unequal, for example 27 against 37.
• Block randomisation: the sequence is made in small blocks, for example blocks of four, each containing two low-pressure and two standard-pressure allocations in random order. After every block the groups are equal. This is the usual choice for a trainee’s trial of 60 to 100 patients.
• Stratified randomisation: separate sequences for important subgroups, for example men and women, so that a factor that strongly affects the outcome is balanced.

“Free online randomisation tools and SPSS can produce the list in minutes,” I said. “Keep a printout of the list and the method, because CONSORT asks you to report exactly how it was made.”

Hassan raised a finger. “Sir, in my synopsis I wrote ‘random’ for how I select my patients. Is that the same thing?”

“No, and I am glad you asked. Random sampling is how you choose who enters the study. Random allocation is how you divide the people already in the study into groups. Hammad will enrol patients consecutively from his OT lists, which is not random sampling at all, and then allocate them randomly. We will deal with your sampling properly when we reach the sampling session.”

The Envelope Held Up to the Light

“Part two,” I said, “is the one most trainees have never heard of. Allocation concealment.”

I told them about trials where the allocation was kept in envelopes, and someone held the envelopes up to a bright lamp to read the group before deciding whether to enrol a patient. Kenneth Schulz and David Grimes described exactly this kind of deciphering in a well-known Lancet paper, and showed why inadequate concealment tends to exaggerate treatment effects [2].

“Concealment means that the person enrolling the patient cannot know the next allocation until the patient has been consented and formally entered into the trial. Once it is revealed, it cannot be undone.”

“How do we do that in a government hospital without a computer system?” asked Dr. Bushra Fatima, who had come straight from the labour room.

“The classic low-cost method is sequentially numbered, opaque, sealed envelopes. Someone not involved in recruiting, for example the OT in-charge nurse, prepares them from the computer list. Each envelope is numbered, the paper is opaque or wrapped in foil, and it is sealed. The envelope is opened only after the patient is anaesthetised and enrolled, and in strict number order. The nurse writes the patient’s name and the time on the envelope before opening it.”

Hammad was writing quickly. “So randomisation makes the list, and concealment protects the list.”

“That is the best summary anyone has given in this room.”

Who Is Actually Blind?

“Now,” I said, “double blind.”

“Two people, Sir.”

“Which two? The term ‘double blind’ means different things in different papers, which is why the CONSORT 2025 statement asks authors to say exactly who was blinded after allocation and how, rather than relying on labels [3]. In a trial, there are several people who could know the allocation:”

• the patient
• the surgeon
• the anaesthetist
• the person who measures the outcome
• the person who analyses the data

“Can the surgeon be blinded to the pressure he is operating at?”

“No,” said Sumaira. “It is on the insufflator screen.”

“Correct. In surgical trials, the surgeon usually cannot be blinded. That is acceptable, if you say so honestly. But the patient can be, because he is asleep when the pressure is set. And the outcome assessor must be, because shoulder-tip pain is a subjective score.”

I looked at Hammad. “Your draft says you will assess every patient yourself. You designed the trial. You want low pressure to win. When a patient says ‘maybe three, maybe four’, which number will you write?”

He was quiet. Then, honestly: “Three.”

“Not because you are dishonest. Because you are human. So a ward nurse or a fellow trainee who does not know the allocation records the pain score, and the allocation stays hidden in the theatre records until the data are complete. You will even give the groups codes, A and B, for the person running the analysis.”

“Concealment and blinding are different, then,” Junaid said slowly. “Concealment protects the allocation before the patient enters. Blinding protects the trial after.”

“Write that down too,” I said.

The Patient You Wanted to Exclude

“Your third sentence,” I said. “If the consultant raises the pressure, you will exclude that patient.”

“He did not get low pressure, Sir. It would be unfair to count him.”

“Think about who those patients are. The obese patient, the inflamed gallbladder, the difficult anatomy. If you remove them from the low-pressure group only, you remove the hardest cases from one side. The two groups are no longer the groups chance created. You have quietly undone your own randomisation.”

I wrote two terms on the board:

• Intention-to-treat analysis: every patient is analysed in the group to which he was randomised, whatever happened in theatre. This keeps the protection of randomisation and answers the real clinical question: what happens when a surgeon sets out to use low pressure?
• Per-protocol analysis: only patients who received the treatment exactly as planned are analysed. Useful as a secondary check, but more open to bias.

“Report intention-to-treat as your main analysis, and record how many patients needed the pressure raised, and why,” I said. “That number is a result in itself. The Cochrane review on this question found that laparoscopic cholecystectomy could be completed with low pressure in about 90% of people [4]. If your unit converts far more than that, readers will want to know.”

Is Another Trial on This Question Ethical?

Dr. Zunaira Malik had said nothing all evening. Now she asked her question in her usual quiet way.

“Sir, if many trials have already compared these two pressures, is it right to randomise more patients? Someone must already know the answer.”

The room went still. It was the right question, and Hammad had not thought of it.

“A trial is only ethical when there is genuine uncertainty about which treatment is better,” I said. “We call that equipoise. So Hammad must read the evidence honestly before his ethics committee asks him to. The 2014 Cochrane review included 21 trials and found that only one was at low risk of bias; the overall quality of evidence was low or very low [4]. A larger 2022 meta-analysis of 44 randomised trials found less post-operative pain, including shoulder pain, with low pressure [5].”

Hammad looked worried. “So my question is already answered?”

“Perhaps partly. So your rationale cannot say ‘no study has been done’. It must say what remains uncertain in your setting: your unit’s pressure, your analgesia protocol, your patients, or an outcome earlier trials measured poorly. And here is the lesson I want you to take home. Look again at why those older trials were rated at high risk of bias. Poor randomisation. No allocation concealment. Unblinded outcome assessment. Patients excluded after allocation.”

He looked down at his plastic sleeve. “My four sentences.”

“Your four sentences. A small trial that fixes them is worth more than ten that repeat them.”

We discussed whether the topic is feasible and novel in the FINER session, so I will only add this: the ethics committee will read your rationale with exactly Zunaira’s question in mind.

Register Before the First Patient

“One more obligation,” I said. “Registration.”

“Sir, I thought the ERC approval was the registration.”

“The ERC approval comes first, and it is essential. Trial registration is a separate, public step. You enter your trial on a public registry, ClinicalTrials.gov or one of the World Health Organization’s primary registries listed on the ICTRP platform, with your primary outcome, sample size and design, before the first patient is enrolled. The ICMJE requires registration at or before the time of first patient consent as a condition for publication [1]. Registration is free on ClinicalTrials.gov.”

“And if I register after starting?”

“Many journals will not consider the paper. The Journal of the College of Physicians and Surgeons Pakistan, for example, states in its author instructions that randomised trials must provide a trial registration number. A trial registered after recruitment has started looks as if the outcome could have been chosen after seeing the data. The whole point of registration is to fix your primary outcome before anyone knows the result.”

“So the order is,” Sumaira said, counting on her fingers, “synopsis approved, ERC approval, registration, then the first patient.”

“Yes. And if a trial tests a new drug or device rather than a surgical technique, the Drug Regulatory Authority of Pakistan has its own rules for clinical trials, so check with your ERC early. We will cover ethics review and consent in detail in a later session.”

The Flow Diagram

I drew the last thing of the evening: a set of boxes running down the board.

• Assessed for eligibility (n = ?)
• Excluded, with reasons (n = ?)
• Randomised (n = ?)
• Allocated to low pressure / allocated to standard pressure
• Lost to follow-up or discontinued, with reasons
• Analysed

“This is the CONSORT flow diagram,” I said. “The CONSORT 2025 statement is a 30-item checklist plus this diagram, and journals expect it with every trial report [3]. Start filling it from your first patient, not on the night before submission. Every box needs a number, and every number has to add up. When we reach manuscript writing, you will be grateful.”

Hammad’s Second Draft

By the end of the session Hammad had crossed out most of his page. Before he left, he read out the new version of his method, and I put it on the board as a table.

Element	First draft	Revised draft
Allocation	Odd and even admission dates	Computer-generated block randomisation (blocks of four)
Concealment	None	Sequentially numbered, opaque, sealed envelopes held by the OT in-charge nurse, opened after induction
Blinding	“Double blind”	Patient and outcome assessor blinded; surgeon not blinded, stated openly
Outcome measurement	Hammad himself at 24 hours	A blinded ward nurse records shoulder-tip pain on a 0 to 10 numerical rating scale at 24 hours
Patients whose pressure is raised	Excluded	Kept in their randomised group (intention-to-treat); number and reasons reported
Registration	Not mentioned	Registered on ClinicalTrials.gov after ERC approval, before the first patient
Reporting	Not mentioned	CONSORT 2025 checklist and flow diagram

“And the hypothesis?” Bushra asked, with a small smile. She had been on the wrong end of that question herself.

“Non-directional, as we did in the objectives session,” Hammad said at once. “There is a difference in mean shoulder-tip pain score between the two groups.”

“And the title?”

He thought for a moment, then struck out “Double Blind” from the heading of his plastic sleeve.

“Randomised controlled trial,” he said. “Nothing more until I can say who is blind.”

At the door he turned back. “Sir, the trials that were rated at high risk of bias. Their authors probably thought they were doing a proper randomised trial too.”

“Almost certainly,” I said. “That is the reason we do this slowly.”

He nodded, and for once Hammad Ali left the Research Room without running.

Key Takeaways

• A randomised controlled trial is the design in which the researcher assigns the intervention, and chance decides who gets which. Randomisation balances known and unknown confounders between groups.
• Alternate days, odd and even numbers and similar schemes are predictable, not random. Use a computer-generated sequence; block randomisation keeps a small trial’s groups equal.
• Allocation concealment (for example sequentially numbered, opaque, sealed envelopes) protects the sequence before enrolment; blinding protects the trial after allocation. They are not the same thing.
• Do not write “double blind”. State who was blinded: patient, surgeon, outcome assessor, analyst. A subjective outcome such as pain must be measured by a blinded assessor.
• Analyse patients in the group they were randomised to (intention-to-treat), and report deviations rather than excluding them.
• Get ERC approval, then register the trial publicly before the first patient is enrolled, and report it using the CONSORT 2025 checklist and flow diagram.

References:

1. International Committee of Medical Journal Editors. Recommendations for the conduct, reporting, editing, and publication of scholarly work in medical journals: clinical trial registration. Available from: https://www.icmje.org/recommendations/browse/publishing-and-editorial-issues/clinical-trial-registration.html
2. Schulz KF, Grimes DA. Allocation concealment in randomised trials: defending against deciphering. Lancet. 2002;359(9306):614-8.
3. Hopewell S, Chan AW, Collins GS, Hróbjartsson A, Moher D, Schulz KF, et al. CONSORT 2025 statement: updated guideline for reporting randomised trials. BMJ. 2025;389:e081123.
4. Gurusamy KS, Vaughan J, Davidson BR. Low pressure versus standard pressure pneumoperitoneum in laparoscopic cholecystectomy. Cochrane Database Syst Rev. 2014;(3):CD006930.
5. Ortenzi M, Montori G, Sartori A, Balla A, Botteri E, Piatto G, et al. Low-pressure versus standard-pressure pneumoperitoneum in laparoscopic cholecystectomy: a systematic review and meta-analysis of randomized controlled trials. Surg Endosc. 2022;36(10):7092-113.