Two Synopses, One Wrong Word
Dr. Hassan Raza and Dr. Hammad Ali put their synopses on the table at the same moment, and I noticed it before I had turned a page. Under the heading Sampling Technique, both had written the same three words: “Simple random sampling.”
“Hassan, you already told us in the trial session that your patients are not random. Why is it still here?”
“Sir, I meant to change it. But every synopsis I copied the format from says random. It sounds scientific. If I write ‘consecutive’, the examiner will think my study is weak.”
“And Hammad?”
Hammad looked genuinely surprised. “Sir, mine is a randomised trial. Block randomisation, sealed envelopes, everything we did. So my sampling is random. Isn’t it?”
“No,” I said. “And today both of you will see why the two words in your synopses describe something neither of you is doing. But we start one step earlier, with a question every synopsis must answer before it says anything about sampling: who belongs in your study?”
Three Populations, Not One
I wrote three words down the board: TARGET, STUDY, SAMPLE.
“Most of us use the word ‘population’ as if it means one thing. In a study it means three, and confusing them is one of the commonest reasons a methods section comes back with red ink.”
| What it means | Hassan’s study | |
|---|---|---|
| Target population | Everyone you want your results to apply to | Adults with displaced distal radius fractures treated by open reduction and internal fixation (ORIF) in Pakistan |
| Study population | The part of the target population you can actually reach, defined by place and time | Patients undergoing ORIF for distal radius fracture in the Department of Orthopaedics, UPMED Hospitals, Lahore, for one year after the approval of the synopsis |
| Sample | The patients you actually enrol, chosen by a stated method | The eligible patients enrolled by the sampling technique he writes in his synopsis |
“Your results travel from the sample back to the study population, and from there, with care, to the target population. Every step is a bridge. Your inclusion criteria, exclusion criteria and sampling method decide how strong each bridge is.”
Random Sampling and Random Allocation Are Different Jobs
I turned to Hammad. “In the trial session I told you in one line that random sampling and random allocation are not the same. Now the full answer.”
“Sampling decides who enters your study. Allocation decides which arm a patient goes to after entering. They protect different things.”
| Random sampling | Random allocation (randomisation) | |
|---|---|---|
| Question it answers | Who gets into the study? | Which group does each enrolled patient join? |
| When it happens | Before enrolment | After enrolment and consent |
| What it needs | A complete list of eligible people (a sampling frame) | A random sequence, kept concealed |
| What it protects | External validity: do the results apply to the wider population? | Internal validity: are the two groups comparable, so the difference is due to the intervention? |
| Hammad’s trial | Not used. Patients enter consecutively | Used. Block randomisation with sealed opaque envelopes |
“Almost no hospital trial uses random sampling. Nobody has a list of next month’s gallbladder patients to draw from. Patients are enrolled as they arrive, and then randomised. So your synopsis says ‘non-probability consecutive sampling’ under Sampling Technique, and describes your block randomisation separately, under Randomisation. Both are correct, and both are needed.”
“Then how do my results apply to anyone outside UPMED?” asked Hammad.
“Through honest description. Rothwell’s review of trial external validity asks one plain question: to whom do these results apply? He lists the setting, the selection of patients, the exclusions and the run-in as the things that decide the answer, and he notes that most trial reports describe them poorly [1]. Report who was eligible, who was screened, who was excluded and why. A reader in Multan or Peshawar can then judge whether your patients look like theirs.”
Probability and Non-Probability Methods
“In probability sampling, every eligible person has a known chance of selection, and chance does the choosing. In non-probability sampling, the researcher’s situation does the choosing [2]. Neither is shameful. Only mislabelling is.”
| Technique | How it works | Where it fits |
|---|---|---|
| Simple random (probability) | Every eligible person on a complete list has an equal chance; selected by random numbers | When a full list exists: a registry, a school roll, a household census |
| Systematic (probability) | Every k-th person from an ordered list or queue, after a random start | Busy OPDs with a register, where every patient cannot be enrolled |
| Stratified (probability) | The list is divided into groups (for example, male and female) and each group is sampled separately | When small subgroups must be represented |
| Cluster (probability) | Groups such as villages, schools or wards are selected at random, then people within them are studied | Community surveys and district-level studies |
| Consecutive (non-probability) | Every eligible patient, in order of arrival, during a defined period, until the sample size is reached | The usual method in CPSP and UHS hospital-based synopses |
| Convenience (non-probability) | Whoever is easy to reach: your duty days, your ward, the patients who are willing and nearby | Common in practice, highest risk of selection bias; avoid where possible |
| Purposive (non-probability) | Participants chosen deliberately because they hold the information needed | Qualitative interviews and focus groups |
Hassan’s Sampling Frame
“Hassan, tell me exactly how you will pick patients.”
“Sir, the patients come through emergency and the fracture clinic. Whoever is listed for distal radius plating, I take.”
“Every one of them, or the ones operated on when you are on duty?”
He hesitated. “Mostly my unit’s lists.”
“Then today you have convenience sampling, and you have been calling it random. To make it consecutive, decide where every eligible patient passes, and enrol from there without skipping. The OT list for distal radius ORIF is your natural sampling frame for the study period, across all the units that operate on these fractures. If the other units will not cooperate, say in your synopsis that the study is limited to your unit’s patients. That is honest, and it is still consecutive, because you take every eligible patient on your unit’s lists, not the ones you happen to operate on.”
“And random sampling, sir? Could I do it?”
“Only if you had the whole year’s list in advance and drew names from it by chance. You cannot, because those patients have not broken their wrists yet. Write ‘non-probability consecutive sampling’. Examiners do not reject a study for honest consecutive sampling. They return synopses that claim random sampling with no sampling frame.”
When the OPD Is Too Big to Take Everyone
Dr. Junaid Rashid had been doing arithmetic on a notepad. “Sir, my medicine OPD sees more than two hundred hypertensive patients on a busy day. One data collector cannot interview them all. If I take the first fifteen who walk in, is that consecutive?”
“It is convenience wearing a consecutive label. Who arrives first in your OPD?”
He thought. “Old patients from the city who come for refills. The ones from villages arrive late, after the bus.”
“And whose adherence do you suspect is worse?”
“The late ones.” He put the pen down. “So the first fifteen would make my non-adherence look lower than it is.”
“Exactly. Your OPD has something most wards do not: a register, filled in order through the whole day. Use systematic sampling. Pick a random start between 1 and k, then interview every k-th hypertensive patient on the register, with k set from your daily numbers and how many interviews one person can manage. The early and the late patients both get their chance. That is a genuine probability method, and you may call it one.”
Inclusion Criteria: Who Is In
“Before any sampling, you decide who is eligible. Inclusion criteria are the characteristics a patient must have to enter. They define the study population so precisely that another doctor could apply them without calling you [3].”
I took Hammad’s synopsis. His inclusion criteria read: “Adult patients. Gallstones. Laparoscopic cholecystectomy.”
“Hammad, is a 17-year-old adult? Is a patient with a gallbladder polyp included? Emergency or elective? And can your patients give the pain score your outcome depends on?”
Good inclusion criteria usually state four things:
- The condition or procedure, with how it is confirmed (for example, symptomatic gallstones on ultrasound).
- Age as a range in years, never the word “adult” alone. If you set an upper limit, give a clinical reason for it; an unexplained cap only shrinks the population your results apply to.
- The setting: elective or emergency, OPD or ward, which department.
- The ability to take part: consent, and whatever your outcome measurement needs from the patient.
Exclusion Criteria: Who Is Out, and Why
“Exclusion criteria remove patients who meet the inclusion criteria but have something that would distort the result, put them at risk, or make the outcome impossible to measure [3]. Two rules. First, an exclusion must never repeat an inclusion in reverse. If you include ages 18 to 65, writing ‘age below 18 or above 65’ under exclusions adds nothing. Second, every exclusion needs a reason you can defend.”
Dr. Bushra Fatima looked up. “Sir, in my cohort I was going to exclude twin pregnancies and mothers with hypertension. They make the data messy.”
“Bushra, open your rationale. What did we agree was still uncertain about antenatal corticosteroids?”
She read it, and her face changed. “Multiple pregnancies. Mothers with hypertension.” A pause. “I was about to exclude the exact patients my study is supposed to tell us about.”
“Simplification without justification is bias with a polite name. Exclude a group only when there is a clinical or safety reason. If you keep them, plan to report them as subgroups. If you must exclude them, say so under limitations, because then your results do not apply to them.”
“One more rule, for the trial in the room,” I said to Hammad. “Exclusions are applied before randomisation. A patient who is converted to open surgery after being randomised is not ‘excluded’. He stays in the group he was allocated to, as intention-to-treat requires.”
The Patients Who Say No
Dr. Zunaira Malik had been quiet. “Sir, if sampling is consecutive and a patient refuses consent, is the sampling broken? And how will anyone know how many refused?”
“That is the question that separates a study from a collection of patients,” I said. “Refusal does not break consecutive sampling. Hiding it does. Keep a screening log from the first day: every patient screened, whether eligible, the reason if not, whether consent was given, and the reason if refused, as far as the patient is willing to say. At the end, your flow of patients comes straight from that log, and a reader can see whether the refusers looked different from the participants.”
“For my OSMF cases I can enrol consecutively from the Oral Pathology clinic,” she said. “And the controls?”
“By the same rules, from the same source population, as we discussed in the case-control session. Your synopsis must state where the controls come from and how they are picked, in a way another person could repeat. And both cases and controls go into the screening log.”
Hammad’s Revised Section
By the end of the session, Hammad’s methods page looked like this.
Setting: Department of General Surgery, UPMED Hospitals, Lahore.
Duration: six months after the approval of the synopsis.
Inclusion criteria:
- Patients aged 18 years and above, of either gender.
- Symptomatic gallstones confirmed on abdominal ultrasound.
- Listed for elective laparoscopic cholecystectomy, ASA class I or II.
- Able to understand and report pain on the 0 to 10 numerical rating scale, and willing to give written informed consent.
Exclusion criteria:
- Acute cholecystitis or common bile duct stones on ultrasound or liver function tests (different operation and pain profile).
- Pre-existing shoulder or neck pain, or regular analgesic or opioid use (would confuse the shoulder-tip pain score).
- Previous upper abdominal surgery.
- Pregnancy.
Sampling technique: non-probability consecutive sampling. All eligible patients during the study period will be screened, and every patient screened will be recorded in a screening log.
Randomisation: block randomisation in blocks of four, using sealed opaque envelopes, as already planned.
Hassan was already crossing out “simple random” in his own copy. “Sir, it does not look weaker.”
“It looks true,” I said. “That is stronger.”
Key Takeaways
- Define three populations: the target population your results should apply to, the study population you can reach (place and time), and the sample you actually enrol.
- Random sampling decides who enters the study and needs a complete list; random allocation decides which arm an enrolled patient joins. A randomised trial almost always uses consecutive sampling plus randomisation, and the synopsis should state both separately.
- Never call a non-probability method random. Consecutive sampling (every eligible patient, in order, without skipping) is the usual and defensible method for hospital-based CPSP and UHS synopses; convenience sampling carries the highest risk of bias.
- When there are too many patients to enrol all of them, systematic sampling from a register with a random start keeps every patient’s chance equal.
- Inclusion criteria state the condition and how it is confirmed, an age range in years, the setting, and the ability to consent and be measured. Exclusion criteria must not mirror the inclusions and must each have a clinical or safety reason.
- Keep a screening log from the first day: screened, eligible, consented, refused, with reasons. It becomes your patient flow and lets readers judge to whom your results apply.
References:
- Rothwell PM. External validity of randomised controlled trials: “to whom do the results of this trial apply?”. Lancet. 2005;365(9453):82-93.
- Elfil M, Negida A. Sampling methods in clinical research; an educational review. Emerg (Tehran). 2017;5(1):e52.
- Patino CM, Ferreira JC. Inclusion and exclusion criteria in research studies: definitions and why they matter. J Bras Pneumol. 2018;44(2):84.
Rate this article
How useful was it for you?
UPMED reads every review; some are shown on this page.
Log in to write a few lines too.