First identify who took part, what they practised, what it was compared with and what was measured. Then examine the size of the difference, observation time and limitations. Start with the question, not the largest percentage.
Use four questions to complete the headline
A headline says tai chi reduced falls. Before asking how many days to practise, use the 2018 balance-programme trial to answer four questions.
| Question | Answer in this trial | Why it matters |
|---|---|---|
| Who took part? | Community-dwelling adults aged 70 or older with a fall history or impaired mobility, able to travel and exercise safely. | Not every age or level of mobility was tested. |
| What did they practise? | Eight adapted movements plus functional training, in supervised 60-minute classes twice weekly for 24 weeks. | This is not interchangeable with any 24-form video or home routine. |
| What was the comparison? | Multimodal exercise and stretching, both organised classes. | The size of the difference depends on which group is compared. |
| What was measured? | The rate of fall events over six months. | Different from a standing-balance score or a guarantee that a person will not fall. |
The trial answers a question about these older adults under these conditions. It does not directly test young adults learning at home or compare Baduanjin with tai chi. This table describes a study; it is not a home exercise prescription.
Is it a trial, a review or institutional information?
- Randomised trial: assigns participants to different programmes for comparison. Randomisation helps, but dropout, measurement and delivery still matter.
- Systematic review: uses explicit methods to find and combine studies addressing a question. Different outcomes may include different studies. It does not simply add every paper and participant together.
- Institutional information: introduces definitions, evidence and safety questions. Follow citations to individual papers for exact results. An institution citing a trial does not create another trial.
A main trial, its follow-up paper and a review that includes it may all describe the same people. Count independent studies, not webpages.
Which numbers and dates are easy to confuse?
The Baduanjin sleep abstract reports 202 people screened, 87 eligible and 55 completing the study. These describe different stages. Methods and a participant flow diagram are needed to explain the reduction and identify who entered the analysis. The abstract alone cannot supply those missing reasons.
In a review, check the sample for the particular outcome. The 2017 Baduanjin review reported 1,535 participants overall, but its pooled sleep analysis used 597. The overall number must not inflate each individual finding.
Separate the length of the course from later observation. During follow-up, did people keep practising, and how many completed assessment? Publication year, search cutoff and website check date are also different. A recently updated summary may still describe an old study.
How can I record a useful conclusion?
Write: “In this group of people, this programme compared with this alternative differed by this amount on this outcome after this time. These questions remain unanswered.” If a detail is missing, mark it as missing; do not borrow a parameter from another paper.
Then ask whether the difference could be practically important and how adverse events and withdrawals were recorded. A small P value cannot replace either answer. An unclear between-group difference does not automatically establish equivalence.
Use the small sleep trial to practise recognising an abstract’s limits, or the tai chi balance trial for the four questions. For percentages and intervals, continue to outcomes and certainty.
References
- Li et al.: 2018 tai chi balance-programme trial
Relevant full-text methods and results checked through the official Europe PMC XML copy; the four-question example uses this trial.
- Chen et al.: 2012 Baduanjin sleep trial
Complete public abstract checked. It distinguishes screened, eligible and completing participants; we did not reconstruct unreported randomisation details.
- Cochrane: evidence and systematic reviews
Further reading retained from the Chinese edition. This page returned HTTP 403 during the English source check on 20 October 2026; its current text was not verified.
- Cochrane Handbook, Chapter 14: certainty of evidence
Further reading. The current page and the training-site alternative returned HTTP 403 during this check; we do not claim to have read the current chapter.
- Cochrane Handbook, Chapter 15: interpreting results
Further reading. The current page and the training-site alternative returned HTTP 403 during this check; we do not claim to have read the current chapter.
