Group 1 minus group 2. Use change scores if the paper reports them.
Risk difference, group 1 minus group 2, in percentage points.
Paste numbers (one per line, or separated by commas or spaces). Nothing leaves your browser.
Sentence for your results or discussion
Is a new option (cheaper, faster, less invasive) no worse than the standard by more than a margin? Enter the difference as new minus standard.
Example loaded: UKSTAR, functional brace vs cast after Achilles tendon rupture, ATRS at 9 months, pre-specified 8-point MCID.
Sentence for your results or discussion
Required sample size
Plan around the smallest effect that matters, not around an effect reported in an earlier small study. Calculations use t distributions and assume normally distributed outcomes.
SWC = factor × between-athlete SD. SEM = SD × √(1 − ICC). MDC = z × √2 × SEM. Small-telescopes bound: the standardized effect the original study had 33% power to detect (Simonsohn 2015); multiply by the SD of the paired differences (or the pooled SD) for raw units.
Published thresholds for common sports medicine outcomes. Thresholds depend on population, anchor and method, and most are individual-level values: used as a between-group bound they make equivalence easier to declare. Read the source before using one.
How to read the five patterns
| Pattern | What it means |
|---|---|
| Clinically important | The whole confidence interval is beyond the threshold. The effect is real and large enough to matter. |
| Real, importance unclear | The interval excludes zero but reaches past the threshold. There is probably an effect, but we cannot tell whether it matters. |
| Real but trivial | The interval excludes zero but stays inside the threshold. There is probably an effect, but it is too small to matter. |
| Practically equivalent | The interval includes zero and stays inside the threshold. The options can be treated as the same for patients. |
| Inconclusive | The interval includes zero and reaches past the threshold. The study cannot tell. This is not evidence of no difference. |
Zero is judged with the reported CI; the threshold is judged with the 90% CI (alpha = 0.05), which matches an equivalence test (two one-sided tests, TOST). Standard errors are recovered from the reported limits, so results from published summaries are approximations. With raw data, analyze the data directly (for example with the TOSTER or clinthresh R packages). The threshold must be chosen before looking at the result.
Cite
Papadakis Z. Statistically significant, clinically meaningful, or neither? Clinical thresholds, equivalence testing, and replicability in sports medicine rehabilitation. Curr Phys Med Rehabil Rep. 2027.
Software: Papadakis Z. clinthresh: clinical thresholds and equivalence tests (R package and web app). Zenodo. https://doi.org/10.5281/zenodo.24022248
R package, source code and the threshold table: github.com/zpapadakis/clinthresh. Worked examples from the review: doi.org/10.17605/OSF.IO/ZQM24.
Free and open source (MIT license). The calculations are a JavaScript port of the clinthresh R package and were checked against it and against the TOSTER R package (same P values to four significant figures for the published examples in the review and for the means, paired and proportion inputs). The author developed this app and the R package with the help of an AI assistant (Claude, Anthropic), tested them and is responsible for them. Report errors or suggest thresholds through the GitHub repository.