Treatment Efficacy
The Treatment Efficacy calculator compares event rates between a treatment and a control group — for example, the proportion of patients who experienced a stroke in a drug arm vs. a placebo arm.
Step 1 — Choose the outcome direction
| Outcome direction | When to use | Example |
|---|---|---|
| Higher rates are better for patients | The event is a good outcome | Treatment response, recovery rate |
| Lower rates are better for patients | The event is a bad outcome | Mortality, stroke, adverse events |
Step 2 — Choose the hypothesis type
| Hypothesis type | What it tests |
|---|---|
| Superiority | Show that treatment is better than control |
| Equivalence | Show that treatment is equivalent to control |
| Non-inferiority | Show that treatment is not significantly worse than control |
For Superiority, set a Superiority threshold (default 5%). For Non-inferiority, set a Non-inferiority margin.
Step 3 — Enter your data
Four numbers:
- Treatment Events / Total
- Control Events / Total
Also supports Sample Data Generator and Raw Data Import from Excel.
Step 4 — Calculate
Results panel shows:
- Treatment/Control proportions (%)
- Relative Risk (RR) with 95% CI
- Absolute Risk Reduction (ARR) with 95% CI
- Relative Risk Reduction (RRR) with 95% CI
- Odds Ratio (OR) with 95% CI
- NNT or NNH with 95% CI
- Chi-square test with p-value
- Fisher's exact test with p-value
- Statistical decision aligned with chosen hypothesis type
Step 5 — Sample Size Assessment
Evaluates whether your sample size is adequate. A Decision Explanation block walks through the hypothesis test logic.
A UniversalChatBot is available for follow-up questions.
Statistical methods used
Relative Risk (RR)
| Component | Formula |
|---|---|
| RR | p_T / p_C |
| SE of log(RR) | sqrt((1-p_T)/events_T + (1-p_C)/events_C) |
| 95% CI | exp(ln(RR) +/- 1.96 * SE_logRR) |
Absolute Risk Reduction (ARR)
ARR = |p_C - p_T| with Wald SE and 95% CI.
Odds Ratio (OR)
Woolf method: OR = (a*d)/(b*c) with log-scale CI.
NNT / NNH
NNT = 1 / ARR. Labelled NNT (beneficial) or NNH (harmful) based on outcome direction.
Chi-square test
Standard 2x2 chi-square without Yates' correction. df = 1.
Fisher's exact test
Hypergeometric point probability on the 2x2 table.
Hypothesis designs
| Design | Decision rule |
|---|---|
| Superiority | Treatment effect > threshold |
| Equivalence | Effect lies within equivalence margins |
| Non-inferiority | Effect not worse than margin |