Data Collectionhard
0:00.0
A randomized controlled trial compares a new antihypertensive medication to placebo with 600 participants (300 per arm). At 12 weeks:
- Medication group: 240 patients took ≥80% of doses; mean BP reduction = 14 mmHg
- Placebo group: 270 patients took ≥80% of doses; mean BP reduction = 2 mmHg
A per-protocol (PP) analysis comparing compliant patients shows a 12 mmHg difference. An intent-to-treat (ITT) analysis using all 600 participants finds an 8 mmHg difference.
Why is the ITT estimate preferred for causal inference about the medication's effect?
A randomized controlled trial compares a new antihypertensive medication to placebo with 600 participants (300 per arm). At 12 weeks:
- Medication group: 240 patients took ≥80% of doses; mean BP reduction = 14 mmHg
- Placebo group: 270 patients took ≥80% of doses; mean BP reduction = 2 mmHg
A per-protocol (PP) analysis comparing compliant patients shows a 12 mmHg difference. An intent-to-treat (ITT) analysis using all 600 participants finds an 8 mmHg difference.
Why is the ITT estimate preferred for causal inference about the medication's effect?