How to read a study (plain language)
Types of evidence, weakest to strongest (roughly)
- Lab or computer studies: ideas, not proof in people.
- Case report: what happened to one person or a few. Can be dramatic; cannot show a treatment works in general.
- Registry or cohort study: follows groups over time. Doctors chose the treatments, so groups can differ in hidden ways.
- Randomized controlled trial: chance decides who gets which treatment. The fairest test.
- Systematic review or meta-analysis: combines many studies.
Trial phases
Phase 1 tests safety in a small group; phase 2 looks for signs of benefit and the right dose; phase 3 compares with standard care in a larger group. Registration tells you the question, not the answer.1,2
Numbers in one sentence each
- Hazard ratio: how fast an event (like relapse) happened in one group compared with another; 0.41 means it happened at well under half the rate during the study.
- AUC or C-index: how well a model separates people who have an event from those who do not; 0.5 is a coin flip, 1.0 is perfect.
- Absolute vs relative risk: "halves the risk" can mean 2 in 100 down to 1 in 100.
- Small numbers: 8 of 72 people is 11%, but a few cases either way would change it a lot.
Press releases, abstracts and retractions
A press release is not a result; companies choose what to highlight. Conference abstracts are short and not fully peer-reviewed. Even published trials can be retracted: the main avacopan trial paper was retracted in 2026.3 See avacopan.
Thinking about joining a trial?
Ask about eligibility (kidney function limits, transplant exclusions), costs and travel, placebo, how to withdraw, and how your data are protected. Talk with your own specialist first.1,2
Sources
- ClinicalTrials.gov: Learn about clinical studies
- NIH: NIH Clinical Research Trials and You: the basics
- PubMed record of NEJM notice (29 Jun 2026): Retraction: Jayne DRW et al. Avacopan for the treatment of ANCA-associated vasculitis (ADVOCATE) [doc ref 19]
Links checked 15 November 2026. Full list: all sources.