A paper arrives titled "Efficacy of technique X in the management of condition Y". The data are eleven patients the surgeon operated on over four years, with no comparison group. The design cannot support the word "efficacy", and the paper is declined — not for the quality of the work, but for a claim it was never in a position to make.
This is the most common avoidable rejection in clinical publishing. The fix is to identify what you actually have before you write, and then claim precisely that.
Three questions that settle it
- How many patients, and how did they get into your study? Selected by you, or consecutive by criteria?
- Is there a comparison group? Anything at all to compare against?
- Did you follow them forward in time, or look backwards at what already happened?
| Design | Patients | Selection | Comparison | Can claim |
|---|---|---|---|---|
| Case report | One | Notable | None | This occurred |
| Case series | Several | Ideally consecutive | None | This is what happened in this group |
| Cross-sectional | Defined sample | Systematic | Within-sample | Prevalence, associations at one point |
| Case-control | Cases + controls | By outcome | Yes | Association with prior exposure |
| Cohort | Defined group | By exposure | Yes | Incidence, risk over time |
| RCT | Randomised | By chance | Yes | Causation |
The case report
One patient, described in enough clinical detail that a reader could recognise the same situation. Its value lies entirely in being notable — the first reported presentation, an unexpected adverse effect, a diagnostic trap, an unusual complication of a common procedure.
"A patient with condition X was treated with the standard approach and recovered" is not a case report. It is a clinical note. The question a case report must answer is: why should another clinician change what they do, or what they look for, having read this?
What it can claim: this occurred; this is possible; watch for this.
What it cannot: anything about frequency, risk or effectiveness. One patient generates hypotheses; it tests nothing.
Reporting guideline: CARE, or SCARE for surgical cases. Patient consent for publication is required — see our guide to consent.
The case series
Here is where most clinical papers sit, and where most overclaiming happens.
A case series describes a group of patients sharing a condition, exposure or procedure, with no comparison group. The absence of a comparison is not a minor limitation — it is the defining structural feature, and it determines everything you may conclude.
Selection is the whole ballgame
Because there is no control group, the only thing standing between your series and pure selection bias is how you chose the patients.
Consecutive enrolment — every patient meeting stated criteria in a stated period, at stated centres — is the strongest available design feature. It removes your judgement from who entered the study.
Selected cases — patients chosen because they did well, because the notes were complete, or because they were interesting — describe your selection as much as your intervention, and readers cannot adjust for a filter you have not described.
So state it explicitly: how many patients met criteria, how many were included, how many were excluded and why. "Twenty-nine consecutive patients undergoing X between January 2022 and June 2024 at two centres; four excluded for incomplete follow-up" tells a reader what they need. "We reviewed twenty-five cases" does not.
What it can claim: what happened in this group; feasibility; a hypothesis worth testing; a complication rate within this group.
What it cannot: that the treatment caused the outcome; that it is better than an alternative; that findings generalise beyond the setting.
Reporting guideline: PROCESS for surgical series. Ethics approval or a documented waiver is usually required.
The words to avoid
Without a comparison group, these are unsupportable: efficacy, effectiveness, superior, improved outcomes, reduced complications, safe and effective. Every one implies a comparison you did not make.
Available instead: we observed, in this series, feasibility, our experience suggests, this warrants comparative evaluation. Less exciting, and defensible.
The cohort study
A defined group followed over time, with comparison built in — typically exposed versus unexposed, or one management approach versus another.
The comparison is what changes the epistemics. You can now estimate incidence, compare rates and adjust for confounders. Prospective cohorts define variables before outcomes occur, which is stronger. Retrospective cohorts assemble the group from existing records — faster and cheaper, but constrained by data collected for clinical rather than research purposes. See our guide to writing Methods for a retrospective chart review.
What it can claim: incidence, relative risk, association over time, adjusted comparisons.
What it cannot: causation. Unmeasured confounding always remains possible, and the reason one group received one treatment may itself predict the outcome.
Reporting guideline: STROBE, or STROCSS in surgery; RECORD for routinely collected data.
Case-control and cross-sectional, briefly
Case-control starts from the outcome and looks backwards at exposure. Efficient for rare outcomes and long latencies; vulnerable to recall bias and to how controls were chosen — which is the design's central difficulty.
Cross-sectional measures exposure and outcome at one moment. Good for prevalence, weak for causality, because you usually cannot establish which came first.
Choosing well
Two practical points.
Design determines what you may conclude — so decide before you collect. If your question is whether a technique is better than the standard, a series will never answer it, however many patients you accumulate. Building in a comparison at the outset, even a historical or contemporaneous cohort, changes what the paper can say.
A well-executed simple design beats an overclaimed complex one. Reviewers respect a clearly reported consecutive series with honest limitations. They are markedly less generous to a series presented as though it were a trial.
Label it correctly
Name the design in the title or abstract, and again in the first line of Methods: "We conducted a retrospective review of consecutive patients..." This is not modesty. It tells readers immediately how to weigh what follows, and it signals that you understand your own study — which is exactly what a reviewer is assessing in the first thirty seconds.
See also types of research articles and our Types of Articles We Publish page.