Blood-pressure drug trial (fictional)

A fictional, simulated randomized double-blind trial of an antihypertensive tablet against placebo in 200 adults with raised systolic blood pressure (≥ 140 mmHg at screening), run at three invented clinics. Each row is one participant with arm, age, sex, clinic, baseline and week-8 systolic BP, adverse-event flag and dropout. Built for paired and two-sample t-tests, confidence intervals for a treatment effect, regression to the mean and handling of missing outcomes. Not real medical evidence.

Open in the lab Download CSV

Rows: 200
Columns: 9
One row is: one trial participant
Version: v1

Data dictionary

Column	Type	Unit	Summary	Description
`participant_id` Participant	id			Participant code.
`clinic` Clinic	category		3 levels: Clinic 1, Clinic 2, Clinic 3	Trial site.
`arm` Arm	category		2 levels: Treatment, Placebo	Randomized arm.
`age` Age	integer	years	mean 57.09, median 58, SD 9.447, range 30–80	Age at enrolment.
`sex` Sex	category		2 levels: Female, Male	Sex.
`baseline_sbp` Baseline SBP	integer	mmHg	mean 153.7, median 153, SD 10.11, range 140–191	Systolic blood pressure at baseline.
`week8_sbp` Week-8 SBP	integer	mmHg	mean 147.5, median 148, SD 14, range 114–208; 19 missing	Systolic blood pressure after 8 weeks.Missing: Null for participants who dropped out before week 8.
`adverse_event` Adverse event	boolean		2 levels	True if the participant reported any adverse event.
`dropped_out` Dropped out	boolean		2 levels	True if the participant left the trial before the week-8 visit.

What it is good for

Compute within-person changes and run a paired t-test in each arm.
Compare mean change between arms with a two-sample (Welch) t-test and a confidence interval.
Explain why the placebo arm also improves (regression to the mean plus placebo effect).
Compare adverse-event rates between arms and discuss complete-case analysis vs dropouts.

Methods: paired t-test, Welch two-sample t-test, confidence interval for a difference in means, ANCOVA, two-proportion test, Fisher's exact test, missing data.

Data quality notes

week8_sbp is null for the 19 dropouts (9.5% of participants). The simulation made dropout likelier after an adverse event, but in this sample it is not (1 of 18 vs 18 of 182), so check missingness against the data instead of assuming a pattern.
Both arms fall from baseline; only the difference between arms estimates the drug effect (true effect −8 mmHg).
Entirely fictional: no real drug, trial, clinic or patient.

Provenance and licence

Each candidate has a true usual SBP ~ Normal(148, 12) and a single baseline measurement = true + Normal(0, 7); only candidates with baseline ≥ 140 are enrolled (this creates regression to the mean). Block randomization gives exactly 100 per arm. Week-8 SBP = true − 10 (treatment) or − 2 (placebo) + Normal(0, 7). Adverse events: 15% treatment, 8% placebo. Dropout probability 6% (28% after an adverse event); dropouts have no week-8 value.

Synthetic data · CC0-1.0 (synthetic data generated by Statistics Quest) · generated deterministically (seed 20241014). It does not describe real people or organisations; any resemblance is coincidental.

Preview

Dataset

Blood-pressure drug trial (fictional)

200 rows · one row = one trial participant · synthetic data

About this dataset Open in the lab

Participant	Clinic	Arm	Age (years)	Sex	Baseline SBP (mmHg)	Week-8 SBP (mmHg)	Adverse event	Dropped out
P001	Clinic 1	Treatment	56	Female	158	146	false	false
P002	Clinic 2	Placebo	80	Male	152	139	false	false
P003	Clinic 3	Placebo	63	Female	177	missing	false	true
P004	Clinic 1	Treatment	51	Male	153	142	false	false
P005	Clinic 2	Placebo	50	Female	167	157	false	false
P006	Clinic 3	Treatment	55	Male	154	144	false	false
P007	Clinic 1	Treatment	54	Male	140	missing	false	true
P008	Clinic 2	Placebo	46	Male	143	139	false	false
P009	Clinic 3	Placebo	50	Male	144	148	false	false
P010	Clinic 1	Treatment	42	Female	165	148	false	false
P011	Clinic 2	Placebo	42	Male	143	160	false	false
P012	Clinic 3	Placebo	64	Female	150	165	false	false

Used in

Observational studies and experiments · Statistical Thinking
Interval for a mean with unknown σ · Estimation and Confidence Intervals
Interpreting confidence correctly · Estimation and Confidence Intervals
Intervals for differences in means · Estimation and Confidence Intervals
Paired t-test · Hypothesis Testing
Comparing two proportions · Hypothesis Testing
Effect sizes · Hypothesis Testing
Reporting a test responsibly · Hypothesis Testing
Evaluating a (fictional) blood-pressure trial · Case