Sample Size Calculator
Works out how many specimens a sensitivity or specificity target needs, on the MFDS or the CLSI basis.
Results update as you type. There is no calculate button.
A tool for drafting a study plan. Its output is meant to be reviewed by the person responsible, who remains accountable for the final sample size.
Driven by power. Secures the chance of clearing a target lower bound.
Result
Positive specimens · sensitivity
Positive specimens · sensitivity
164specimens
minimum 149 plus a 10% margin
Coefficient (1.96 + 0.84)² = 7.8400
Tolerated errors, by the exact (Clopper-Pearson) method
False negatives up to 8
At 164 including the margin, against a 90% target, by the exact method at two-sided 95%
How it was calculated
- Formula
- n= (Z₁₋α/₂ + Z₁₋β)² × P̂(1−P̂) ÷ (P̂ − P_L)²
- Values
- = (1.96 + 0.84)² × 0.95 × 0.05 ÷ (0.95 − 0.90)²
- Working
- = 7.8400 × 0.047500 ÷ 0.002500
- Rounded up
- = 148.96 → round up → 149
- Margin applied
- n′= 149 × (1 + 0.10) = 163.90 → round up → 164
Negative specimens · specificity
Negative specimens · specificity
189specimens
minimum 171 plus a 10% margin
Coefficient (1.96 + 0.84)² = 7.8400
Tolerated errors, by the exact (Clopper-Pearson) method
False positives up to 3
At 189 including the margin, against a 95% target, by the exact method at two-sided 95%
How it was calculated
- Formula
- n= (Z₁₋α/₂ + Z₁₋β)² × P̂(1−P̂) ÷ (P̂ − P_L)²
- Values
- = (1.96 + 0.84)² × 0.98 × 0.02 ÷ (0.98 − 0.95)²
- Working
- = 7.8400 × 0.019600 ÷ 0.000900
- Rounded up
- = 170.74 → round up → 171
- Margin applied
- n′= 171 × (1 + 0.10) = 188.10 → round up → 189
Total specimens needed
353specimens
164 positive + 189 negative
Z values are taken to two decimal places, per calculation rule 1.
Conditions
Extra specimens for dropouts and unusable results. Multiplied onto the minimum, not divided out of it.
Sample size against the target lower bound
Raising the target lower bound a little raises the sample size a lot. Use this to find the trade-off.
estimate held at 95%, coefficient 7.8400
| Target lower bound | Minimum | With a 10% margin |
|---|---|---|
| 94% | 3,724 | 4,097 |
| 93% | 931 | 1,025 |
| 92% | 414 | 456 |
| 91% | 233 | 257 |
| 90%current setting | 149 | 164 |
| 89% | 104 | 115 |
| 88% | 76 | 84 |
| 87% | 59 | 65 |
| 86% | 46 | 51 |
| 85% | 38 | 42 |
Wording for the study plan
A paragraph to paste straight into a plan. It carries the track, the formula, the inputs and the result.
Basis: MFDS (Korea) Formula: n = (Z₁₋α/₂ + Z₁₋β)² × P̂(1−P̂) ÷ (P̂ − P_L)² Sensitivity — estimate 95%, target lower bound 90%, confidence 95% Two-sided, power 80%, coefficient 7.8400 → minimum 149 With a 10% margin → 164 - Sensitivity: at a final sample size of 164, up to 8 error(s) can be tolerated while still holding the 90% target lower bound (exact method, two-sided 95%). Specificity — estimate 98%, target lower bound 95%, confidence 95% Two-sided, power 80%, coefficient 7.8400 → minimum 171 With a 10% margin → 189 - Specificity: at a final sample size of 189, up to 3 error(s) can be tolerated while still holding the 95% target lower bound (exact method, two-sided 95%). Total specimens needed: 353 (164 positive + 189 negative) Source: Ministry of Food and Drug Safety, Guideline on Clinical Performance Studies of In Vitro Diagnostic Medical Devices (안내서-0640-03 (2020-11-26)), checked 2026-07-29 This figure supports drafting the plan. The final sample size is settled by the responsible reviewer.
Conditions shared by every row
7.84
Sensitivity
| Estimate % | Target lower bound % | Minimum n | With margin | Errors tolerated | Row actions |
|---|---|---|---|---|---|
Copy two columns out of a spreadsheet and paste into a cell to fill several rows at once.
Specificity
| Estimate % | Target lower bound % | Minimum n | With margin | Errors tolerated | Row actions |
|---|---|---|---|---|---|
Copy two columns out of a spreadsheet and paste into a cell to fill several rows at once.
Source documents
Ministry of Food and Drug Safety, Guideline on Clinical Performance Studies of In Vitro Diagnostic Medical Devices
안내서-0640-03 (2020-11-26) · Checked 2026-07-29 · Program v1.0
The final judgement rests with the responsible reviewer.