---
title: "NNT/NNH Calculator & Log-rank to Hazard Ratio"
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{NNT/NNH Calculator & Log-rank to Hazard Ratio}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(collapse = TRUE, comment = "#>")
```

## Overview

Systematic reviews frequently encounter trials that report incomplete survival data  --  a log-rank p-value but no Hazard Ratio, or probabilities without a directly computed NNT. ParCC bridges these gaps with two tools in the **HR Converter** module.

## Tutorial A: Extracting a Hazard Ratio from a Log-rank Test

### The Scenario  --  Adjuvant Chemotherapy in Colon Cancer

An older trial (published 2005) reports:

- Log-rank chi-squared = **6.8**
- Total events (deaths) across both arms = **142**
- The treatment arm had better outcomes

The paper does not report a Hazard Ratio, which you need for your meta-analysis.

### The Peto Approximation

When only summary log-rank statistics are available, the Peto method estimates:

$$\ln(HR) = \pm \frac{\sqrt{\chi^2}}{\sqrt{E/4}}$$

with a 95% confidence interval:

$$\ln(HR) \pm \frac{1.96}{\sqrt{E/4}}$$

where $E$ is the total number of events.

### In ParCC

1. Navigate to **Convert > HR -> Probability & NNT > Log-rank -> HR** tab.
2. Select input type: **Chi-squared statistic**.
3. Enter chi-squared = **6.8**, Total events = **142**.
4. Select direction: **Treatment is better** (HR < 1).
5. Result: **HR = 0.68** (95% CI: 0.51 - 0.91).

### Alternative: From a p-value

If the paper reports only "log-rank p = 0.009":

1. Select input type: **p-value**.
2. Enter p = **0.009**, Total events = **142**.
3. ParCC converts the p-value to a z-statistic via $z = \Phi^{-1}(1 - p/2)$, then applies the same Peto formula.

## Tutorial B: Computing NNT for a Formulary Decision

### The Scenario  --  Hospital P&T Committee

A Pharmacy & Therapeutics committee asks: "How many patients must we treat with Drug X to prevent one additional death?" The trial reports:

- 12-month mortality: Control = **18%**, Intervention = **12%**

### The Formula

$$NNT = \left\lceil \frac{1}{ARR} \right\rceil = \left\lceil \frac{1}{p_{control} - p_{intervention}} \right\rceil$$

### In ParCC

1. Navigate to **Convert > HR -> Probability & NNT > NNT/NNH** tab.
2. Select input mode: **Two Probabilities**.
3. Enter Control = **0.18**, Intervention = **0.12**.
4. Result: ARR = **6.0%**, NNT = **17**.

**Interpretation:** For every 17 patients treated with Drug X for 12 months, one additional death is prevented.

### Other Input Modes

ParCC supports four ways to compute NNT:

| Input Mode | You provide | ParCC calculates |
|-----------|------------|-----------------|
| Direct ARR | Absolute risk reduction | NNT = ceil(1/ARR) |
| Two Probabilities | Control & intervention probabilities | ARR, then NNT |
| RR + Baseline | Relative Risk + control probability | ARR = p0 x (1 - RR), then NNT |
| OR + Baseline | Odds Ratio + control probability | Converts to probabilities via Zhang & Yu, then NNT |

### NNT vs NNH

When the intervention *increases* risk (ARR < 0), the result is reported as **NNH** (Number Needed to Harm) with an orange warning. This happens when testing safety endpoints rather than efficacy.

## When to Use These Tools

- **Log-rank -> HR:** Systematic reviews where older trials lack HR estimates; indirect treatment comparisons needing HR inputs from published statistics.
- **NNT Calculator:** Communicating treatment effects to clinicians and formulary committees; sensitivity analyses varying NNT across plausible baseline risk ranges.

## References

1. Tierney JF, Stewart LA, Ghersi D, Burdett S, Sydes MR. Practical methods for incorporating summary time-to-event data into meta-analysis. *Trials*. 2007;8:16.
2. Parmar MKB, Torri V, Stewart L. Extracting summary statistics to perform meta-analyses of the published literature for survival endpoints. *Statistics in Medicine*. 1998;17(24):2815-2834.
3. Altman DG, Andersen PK. Calculating the number needed to treat for trials where the outcome is time to an event. *BMJ*. 1999;319(7223):1492-1495.