Statistical & Psychometrics
Fragility Quotient: A Sample-Size Ratio, Not a Trial Verdict
Calculate a percentage from an independently supplied fragility index and a stated sample-size denominator, and understand the limits of interpreting it.
What the quotient does and does not measure
For a reported fragility index (FI) and a chosen sample-size denominator N, the fragility quotient is FI/N. Multiplying by 100 expresses that ratio per 100 participants. For example, FI = 3 and N = 240 give 1.25%; FI = 12 and the same denominator give 5.00%. The numbers are hypothetical and refer to outcome-status changes in the particular statistical procedure, not actual errors or proven changes in patient outcomes.
In the classic binary-outcome procedure, FI is obtained by changing observed outcomes in a specified group and repeatedly calculating a two-sided Fisher exact test until a chosen significance threshold is crossed. Procedures, comparisons and denominators vary. State how FI and N were obtained before comparing results. This worksheet accepts an already computed FI; it does not calculate or validate FI or the original P value.
Walsh and colleagues reviewed 399 eligible trials and reported a median FI of 8. Among the 306 trials that clearly reported loss to follow-up, 162 had more participants lost to follow-up than the FI. Those counts describe their sample and do not establish a universal threshold for individual trials.
| Hypothetical supplied FI | Stated sample size N | FI / N | As a percentage |
|---|---|---|---|
| 3 | 240 | 0.0125 | 1.25% |
| 12 | 240 | 0.05 | 5.00% |
Calculate the quotient
Enter a nonnegative whole-number FI and a positive whole-number N, with FI no greater than N. N must be the denominator chosen for the quotient; say whether it is randomized or analyzed participants and use definitions consistently. A valid-looking ratio does not establish that the supplied FI or sample size is correct.
=IF(COUNT(B2:B3)<>2,"Enter numeric counts",IF(OR(B2<0,B3<=0,B2>B3,B2<>INT(B2),B3<>INT(B3)),"Check whole-number counts",ROUND(100*B2/B3,3)))Interpret missing outcomes separately
Do not label a result robust or fragile solely by whether the number lost to follow-up is above or below FI. FI alters known outcome statuses, whereas loss to follow-up concerns missing outcomes; the counts do not answer the same question. A 2026 simulation analysis challenges direct comparisons of the two, and other authors have proposed missing-data-specific sensitivity methods.
The quotient is not an effect size, a probability that a conclusion is correct, a measure of clinical benefit, or a replacement for confidence intervals, risk-of-bias assessment and an appropriate missing-data analysis. The classic binary-outcome method does not reproduce adjusted or time-to-event analyses simply by counting events.
Try the example
Use the original values or change them to test your understanding.
Further reading
- Walsh et al. (2014): The statistical significance of randomized controlled trial results is frequently fragile
- Ahmed et al. (2016): Does Sample Size Matter When Interpreting the Fragility Index?
- Vivekanantha et al. (2026): Simulation analysis of fragility and loss to follow-up
- On clinical trial fragility due to patients lost to follow-up