Knowledge centre

Research summaries

Plain-language summaries of significant trials, with study design, population, endpoints and limitations stated openly.

Evidence · 6 min read

Randomised controlled trials: why the design matters

Randomisation distributes known and unknown confounders evenly between groups. Blinding reduces measurement bias. Together they are why RCTs sit near the top of the evidence hierarchy.

Intention-to-treat analysis includes every randomised participant in their original group, preserving the benefit of randomisation even when people withdraw or switch treatment.

A statistically significant result is not automatically clinically meaningful. A hazard ratio of 0.95 may reach significance in a very large trial while representing a difference of days.

Check the population. Trials frequently exclude older patients, those with renal impairment and those with poor performance status, so results may not generalise to the person in front of you.

Statistics · 4 min read

Absolute versus relative risk reduction

A treatment that cuts risk from 2% to 1% delivers a 50% relative reduction but only a 1% absolute reduction. Both are true; only one tells you how likely you are to benefit.

Number needed to treat is the reciprocal of the absolute risk reduction. A 1% absolute reduction means 100 people must be treated for one to benefit.

Ask the same question about harm: number needed to harm places adverse effects on the same footing as benefit and supports genuinely informed consent.

These articles are general education reviewed by our clinical editorial team. They are not tailored to your circumstances and do not constitute medical advice, diagnosis or a recommendation to start, stop or change any treatment.